Non–Small Cell Lung Cancer Exhibits Transcript Overexpression of Genes Associated with Homologous Recombination and DNA Replication Pathways

Published OnlineFirst April 7, 2009; DOI: 10.1158/0008-5472.CAN-08-2981

Research Article

Silvia Saviozzi,1 Paolo Ceppi,2 Silvia Novello,2 Paolo Ghio,2 Marco Lo Iacono,1 Piero Borasio,2 Alberto Cambieri,2 Marco Volante,3 Mauro Papotti,3 Raffaele A. Calogero,1 and Giorgio V. Scagliotti2

1Genomics and Informatics Unit, 2Thoracic Oncology Unit, and 3Pathology Department, University of Torino, Department of Clinical and Biological Sciences, San Luigi Hospital, Orbassano, Turin, Italy

Abstract conditions of continuous genotoxic stress, a coordinated DNA Genes involved in DNA repair and replication have been damage surveillance system and several DNA repair pathways have recently investigated as predictive markers of response to been evolutionally developed. Subjects with reduced DNA repair chemotherapy in non–small cell lung cancer (NSCLC). capacity have an increased risk of developing lung cancer (4), and However, few data on the expression of these genes in tumor conversely, high DNA repair capacity has correlation with resistance to chemotherapy (5, 6). Many of the cytotoxic agents compared with corresponding normal lung are available.The aim of this study was to evaluate differential mRNA levels of 22 used in the systemic treatment of non–small cell lung cancer DNA repair genes of five different DNA repair pathways: (NSCLC) are interfering with DNA activity and the possibility of direct, base excision, nucleotide excision (NER), double-strand individualizing DNA repair profiles is becoming a central issue in break (DSBR), and postreplicative repair.In addition, six the search for improved chemotherapy results. Indeed, most of the genes involved in DNA replication (REP) and three telomere molecular markers currently used to predict the responsiveness to chemotherapeutic treatment are genes involved in DNA damage maintenance genes were investigated.Total RNAs extracted p53 BRCA1 ERCC1 MLH1 MSH2 from fresh-frozen tumors and corresponding normal tissues response such as (7), (8), (9), and (10), Rad51 (11), ERCC2, XRCC1, XPA (12), XPC, or involved in DNA of 50 consecutive chemo-naı¨ve resected NSCLC patients were RRM1 TYMS analyzed.Transcript levels were quantified by real-time PCR. replication such as (13) and (14). However, no exhaustive data on the differential expression of these DNA repair/ A significant overexpression was detected in 20 of 30 (67%) genes, mostly belonging to DSBR pathways, whereas others DNA replication genes in lung tumor versus corresponding normal (XPA, XPC, and UBE2N; 10%) were significantly underex- tissue are currently available. The aim of the present study was to pressed.For 7 of 30 (23%) genes, mostly belonging to NER evaluate, in 50 tumor and paired normal lung tissues, the pathway, no significant difference between paired tumor and differential transcript expression of the following 22 human DNA repair genes belonging to five DNA repair pathways: MGMT in normal samples was observed.Transcript overexpression of OGG1, UNG, XRCC1 DSBR and REP genes was significantly higher in poorly direct repair; and in base excision repair (BER); XPA, XPC, ERCC1, ERCC2, ERCC4, ERCC5, ERCC6, and XAB2 in differentiated carcinomas and DSBR levels were higher in XRCC2, XRCC3, XRCC4, XRCC5, men compared with women.The transcriptional overexpres- nucleotide excision repair (NER); XRCC5, TOP3B, TYMS UNG BRCA1, BRCA2, and UBE2V2 in double-strand break repair (DSBR); sion of four genes ( , and ) showed UBE2A, UBE2B, UBE2N significant correlation with a shorter patients’ outcome at the and and in postreplicative repair (PRR). In addition, nine genes involved in DNA replication (REP; TYMS, univariate, whereas only stage of disease appeared as an RRM2B, RRM2, RRM1, TOP3A, TOP3B independent factor affecting prognosis, as assessed by and ) and in telomere maintenance (TERT, TERF1, and TERF2) were also investigated multivariate analysis.In conclusion, genes belonging to DNA repair/replication pathways are overexpressed in NSCLC and (see Supplementary Data Table S1). are associated with a more aggressive phenotype. [Cancer Res 2009;69(8):3390–6] Materials and Methods Patients and samples. Primary tumor samples and paired normal lung Introduction tissues of 50 consecutive NSCLC patients who received radical surgery for early NSCLC at the San Luigi Hospital, Division of Thoracic Surgery, The high incidence of genomic instability in lung cancers is well- Orbassano (Italy) between December 2003 and March 2004, were established and in some cases it has been associated with poor immediately snap-frozen in liquid nitrogen and stored at 80jC until prognosis (1–3). To guarantee genomic stability under the RNA extraction. The age of the patients (37 males and 13 females) ranged from 41 to 82 years (median, 69 years) and none of the patients received either preoperative/postoperative radiation or chemotherapy. Based on the negative outcomes of a large phase III study performed in Europe (15), and Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/). as policy by the investigators, patients with stage II and III completely Requests for reprints: Silvia Saviozzi, Department of Clinical and Biological resected NSCLC did not receive any adjuvant treatments at that time. Sciences, University of Turin, San Luigi Hospital, Regione Gonzole 10, 10043 Median survival time was 56 mo and 21 out of 50 (42%) had died at the end Orbassano, Turin, Italy. Phone: 39-11670-5410; Fax: 39-11903-8639; E-mail: of the study. Histologic examination was performed on adjacent formalin- [email protected]. I2009 American Association for Cancer Research. fixed fragments in all cases to evaluate tumor histotype according to WHO doi:10.1158/0008-5472.CAN-08-2981 criteria (adenocarcinomas, 29; squamous cell carcinomas, 18; and large cell

Cancer Res 2009; 69: (8). April 15, 2009 3390 www.aacrjournals.org

DNA Repair/Replication Transcription Profiling in NSCLC

Figure 1. Differential expression of pathway genes in tumor vs. normal lung tissue. Box-whisker plot represents relative expression level, reported as DDCt values, of each target gene. HKGs, housekeeping gene variability; dashed line, cutoff threshold indicating 2-fold change variation. Target genes have been grouped into highly (dark gray), moderately (light gray), poorly (dotted), and not (white) regulated based on their prevalence in our cohort of patients with NSCLC. carcinomas, 3), grade of differentiation (grade 1, 10; grade 2, 18; grade 3, 22), retrotranscribed totRNA), 1 TaqMan Universal PCR Master Mix pT status (pT1, 11; pT2, 34; pT3, 5), and pN status (pN0, 35; pN1,8;pN2, 7). (2 ; Applied Biosystems) and 1 Assay-on-Demand Gene Expression Assay Forty-four patients were smokers, two (both women with adenocarcinomas) Mix (20). Cycle conditions were as follows: after an initial 2-min hold at had never smoked, and for the remaining four patients, smoking history 50jC to allow AmpErase-UNG activity, and 10 min at 95jC, the samples was unknown. Follow-up until death or to the last follow-up visit were were cycled 40 times at 95jC for 15 s and 60jC for 1 min. Baseline and available for all cases. The study was approved by the ethical review board threshold for Ct calculation were set manually with the ABI Prism SDS 2.1 of our institution. software. Automation allowed negligible intra-assay variation (V5% RNA extraction and cDNA synthesis. Total RNA (totRNA) was isolated coefficient of variance) and low inter-assay variation (V10% coefficient of from lung specimens with the RNeasy 96 Kit and Biorobot 8000 (Qiagen) variance) when evaluated on raw linear expression quantities. according to the manufacturer’s instructions. RNA was extracted from 15 to Quantitative PCR data analysis. For each target gene, fold change in 25 mg and 60 to 80 mg of tumor and normal lung tissue specimens, expression levels between normal and tumor specimens were evaluated DD respectively. Genomic DNA contaminations were removed by on-column using the 2 Ct method (16). Raw data were normalized using the DNaseI treatment (Qiagen). TotRNA was then quantified with an Agilent geometric average value (17) of two suitable reference genes (POLR2A and 2100 Bioanalyzer (Agilent Technologies) and stored at 80jC. Two 18SrRNA; ref. 18). For each patient, the normalized values of each biological micrograms of totRNA were finally retrotranscribed with random hexamer triplicate were averaged before the calculation of the DDCt using normal primers and Multiscribe Reverse transcriptase contained in the High- tissue as a calibrator. The calibration of tumor samples to matched normal Capacity cDNA Archive Kit (Applied Biosystems), in accordance with the tissues allowed the assessment of transcript deregulations strictly manufacturer’s suggestions. associated with malignant transformation, independently from patients’ Real-time PCR. Expression levels of all target genes and two reference genetic alterations such as single nucleotide polymorphisms (19, 20). genes were evaluated with TaqMan Probes commercially available as ‘‘Assay Moreover, to take into account tissue heterogeneity of fresh-frozen samples, on Demand’’ (Applied Biosystems) with optimized primer and probe transcript expression levels were evaluated on biological triplicates of both concentrations (Supplementary Data Table S1). Quantitative PCR was normal and tumor specimens. performed on an ABI PRISM 7900HT Sequence Detection System (Applied Ki67 immunohistochemistry. Formalin-fixed paraffin-embedded tis- Biosystems) in 384-well plates assembled by Biorobot 8000 and the reaction sues were cut into 4-Am sections and collected on charged slides in order to was performed in a final volume of 20 AL. All quantitative PCR mixtures perform immunohistochemical staining. After deparaffinization and contained 1 AL of cDNA template (corresponding to approximately 20 ng of rehydration through graded alcohols and PBS(pH 7.5), the endogenous www.aacrjournals.org 3391 Cancer Res 2009; 69: (8). April 15, 2009

Cancer Research

Table 1. Median ‘‘DDCt’’, SD and P values of Wilcoxon-matched pair test of 30 pathway genes

Gene symbol Pathway Median DDCt SD P Over (%) Equal (%) Under (%) Cluster

BRCA1 DSBR 1.82 1.38 0.000* 66 34 I BRCA2 DSBR 1.5 1.21 0.000* 62 36 2 I XRCC2 DSBR 3.51 1.43 0.000* 90 10 I XRCC3 DSBR 2.2 1.17 0.000* 76 24 I XRCC4 DSBR 0.55 0.55 0.000* 26 74 II XRCC5 DSBR 0.63 0.78 0.000* 38 62 II UBE2V2 DSBR 1.18 0.7 0.000* 64 36 I MGMT DR 0.21 0.93 0.063 16 72 12 II XRCC1 BER 0.45 0.98 0.001* 32 62 6 II OGG1 BER 0.01 0.64 0.233 14 84 2 III UNG BER 1.71 1.42 0.000* 70 30 I XPA NER 0.32 0.64 0.005* 4 84 12 III XPC NER 0.46 0.59 0.000* 74 26 II ERCC1 NER 0.04 0.6 0.633 14 82 4 III ERCC2 NER 1.12 0.81 0.000* 60 40 I ERCC4 NER 0.58 0.94 0.000* 36 64 II ERCC5 NER 0.12 0.66 0.111 6 88 6 IV ERCC6 NER 0.12 0.56 0.136 4 94 2 IV XAB2 NER 0.05 0.59 0.897 6 94 IV UBE2A PRR 0.69 0.94 0.000* 42 56 4 II UBE2B PRR 0.38 0.93 0.002* 32 64 4 II UBE2N PRR 0.84 0.83 0.000* 58 42 II RRM2B DNA-REP 0.09 0.98 0.318 20 68 12 II TYMS DNA-REP 3.28 1.6 0.000* 96 4 I RRM1 DNA-REP 1.5 1.03 0.000* 74 26 I RRM2 DNA-REP 3.62 1.37 0.000* 100 I TOP3A DNA-REP 0.62 0.69 0.000* 26 74 II TOP3B DNA-REP 0.41 0.52 0.000* 20 80 III TERF1 TM 1.18 1.24 0.000* 54 42 4 I TERF2 TM 0.21 0.41 0.000* 4 96 IV

NOTE: Target genes were further grouped into four clusters: highly (I), moderately (II), poorly (III), and not (IV) deregulated based on transcript deregulation observed in the cohort of 50 NSCLC patients. *Significant transcript deregulations.

peroxidase activity was blocked by absolute methanol and 0.3% hydrogen present study. Multivariate analysis was carried out with Cox’s propor- peroxide for 15 min. To perform the antigen-retrieval procedure, the slides tional hazard model and all the factors significantly correlated with were heated in a pressurized heating chamber (Pascal; DakoCytomation) in survival at the univariate analysis were included. Statistical significance EDTA at 10 buffer (pH 8.0) solution. Sections were incubated overnight was set at P = 0.05. with a mouse monoclonal antibody anti-Ki67 (MIB-1; DakoCytomation) diluted 1:300. Immunoreaction was revealed by a dextran-chain (biotin-free) detection system (EnVision; DakoCytomation), using 3,3¶-diaminobenzidine Results (DakoCytomation) as a chromogen. Finally, the sections were lightly counterstained with hematoxylin. Negative control reactions were obtained Variability of target gene transcript expression. All genes TERT by omitting the primary antibodies. Ki67 proliferation index was calculated except were detected in all normal and tumor specimens. as the percentage of positive nuclei among at least 200 nuclei counted at TERT transcript was detectable in at least one specimen of the high magnification in areas of highest labeling. biological triplicate in 96% of tumor samples but only in 76% of Statistical analysis. Differences in transcript expression levels between normal tissues. Each target gene showed a wide range of expression paired tumor and normal samples were evaluated using the Wilcoxon within the group of either normal or tumor tissues. Expressing such matched-pair test on DCt values. Association of patient’s clinicopathologic variability as coefficient of variance evaluated on normalized data features to transcript expression levels and Ki67 was evaluated with the (i.e., DCt), the range of expression was broader in tumors than in Mann-Whitney or Kruskal-Wallis tests. The Spearman rank correlation normal specimens for all genes. Coefficients of variance ranged method was used to evaluate transcript coexpression and genes showing TERF2 UNG R P from 3.68% for to 12.01% for in the tumor tissues and > 0.7 with < 0.01 were defined as significantly coexpressed. Univariate TERF2 XRCC1 analyses of survival were made by the Kaplan-Meier method and validated from 2.72% for to 7.28% for in the normal tissues. by log-rank test with the Benjamini correction for multiple comparisons. Moreover, for each target gene, the intra-sample variability The method by Benjamini was selected because it generated a lower (evaluated as the coefficient of variance of biological triplicates) number of false-negative results compared with other tests for multiple was always found to be lower than inter-sample variability and was comparison, possibly as a consequence of the limited sample size of the fairly comparable between tumor and paired normal specimens.

Cancer Res 2009; 69: (8). April 15, 2009 3392 www.aacrjournals.org

DNA Repair/Replication Transcription Profiling in NSCLC

Differential expression of pathway genes in tumor versus 0.32, 0.46, and 0.84, respectively). For 7 out of 30 (23%) genes, normal lung. Relative expression levels (DDCt) were evaluated mainly associated with the NER pathway, no significant differences for all target genes, with the exclusion of TERT, for which only the between paired tumor and normal samples were observed. transcript overexpression in 78% of tumor samples but not the Quantitative PCR quantification allows us to qualify as ‘‘positive’’ extent of such deregulation was detected. TERT has therefore been only expression level variations greater than the reference gene(s) excluded from subsequent analysis. Relative expression levels were variability. Because this restriction eludes statistical assessment, we extremely wide and most of the target genes were significantly have decided to set the absolute value of ‘‘DDCt’’ to 1 as a differentially expressed between tumor and normal lung tissue threshold for evaluating the prevalence of overexpression or specimens. Significant overexpression has been observed in 20 out underexpression of our target genes in patients with NSCLC. This of 30 (67%) target genes (median DDCt ranging from 0.38 for threshold has been selected because it contains the entire variability UBE2B to 3.62 for RRM2). A positive deregulation of all genes of the reference genes (Fig. 1). Based on the pattern of deregulation belonging to DSB repair (seven of seven), two out of three in the BER in patients with NSCLC, we have arbitrarily segregated target genes pathway, two out of eight in the NER pathway, two out of three in into four clusters: highly deregulated (I, deregulated in at least 50% the PRR pathway, and seven out of eight of the genes involved in of patients), moderately deregulated (II, deregulated in at least 25% DNA replication and telomere maintenance was shown. Only three of patients), poorly deregulated (III, deregulated in 14–18% of genes (10%), XPA and XPC of the NER and UBE2N of the PRR patients), and not deregulated (IV, deregulated in <10% of patients; pathways, had significantly lower transcriptional expression levels see Table 1). Inconsistencies among our clustering and Wilcoxon’s in tumor specimens than normal matched samples (median DDCt analysis have been observed for TERF2 and MGMT.

Table 2. Association between relative gene expression and clinicopathologic variables

c b b b b Pathway Cluster Gene Gender* Age* Histotype*, Grade pT ,x pN Stage ,k Ki67 *,{

DSB I BRCA1 0.016** 0.348 0.012** 0.000** 0.065 0.332 0.072 0.000** DSB I BRCA2 0.039** 0.984 0.033** 0.000** 0.446 0.317 0.286 0.001** DSB I XRCC2 0.050** 0.107 0.020** 0.001** 0.407 0.778 0.834 0.012** DSB I XRCC3 0.018** 0.899 0.001** 0.000** 0.382 0.674 0.48 0.000** DSB II XRCC4 0.104 0.66 0.322 0.501 0.376 0.645 0.736 0.741 DSB II XRCC5 0.274 0.868 0.347 0.003** 0.314 0.453 0.244 0.033** DSB I UBE2V2 0.288 0.296 0.039** 0.049** 0.646 0.33 0.178 0.010** Direct II MGMT 0.309 0.762 0.088 0.27 0.715 0.833 0.813 0.101 BER II XRCC1 0.259 0.807 0.020** 0.022** 0.349 0.356 0.194 0.002** BER III OGG1 0.154 0.899 0.322 0.014** 0.147 0.475 0.281 0.057 BER I UNG 0.144 0.667 0.002** 0.002** 0.801 0.356 0.368 0.000** NER III XPA 0.877 0.822 0.895 0.938 0.893 0.37 0.224 0.565 NER II XPC 0.965 0.74 0.214 0.58 0.346 0.298 0.34 0.264 NER III ERCC1 0.55 0.401 0.014** 0.193 0.678 0.656 0.732 0.009** NER I ERCC2 0.842 0.86 0.137 0.030** 0.752 0.188 0.278 0.024** NER II ERCC4 0.588 0.5 0.176 0.08 0.588 0.14 0.266 0.005** NER IV ERCC5 0.376 0.309 0.599 0.169 0.193 0.29 0.545 0.58 NER IV ERCC6 0.493 0.732 0.233 0.851 0.255 0.83 0.701 0.482 NER IV XAB2 0.642 0.13 0.012** 0.075 0.404 0.447 0.879 0.006** PRR II UBE2A 0.868 0.938 0.179 0.027** 0.637 0.267 0.243 0.007** PRR II UBE2B 0.452 0.429 0.824 0.714 0.455 0.078 0.118 0.125 PRR II UBE2N 0.791 0.145 0.936 0.061 0.072 0.028** 0.004** 0.142 REP II RRM2B 0.699 0.087 0.531 0.982 0.457 0.687 0.907 0.064 REP I TYMS 0.063 0.591 0.06 0.001** 0.555 0.905 0.814 0.001** REP I RRM1 0.304 0.591 0.025** 0.004** 0.591 0.14 0.147 0.002** REP I RRM2 0.106 0.476 0.020** 0.012** 0.434 0.855 0.79 0.003** REP II TOP3A 0.627 0.19 0.022** 0.017** 0.387 0.602 0.487 0.005** REP III TOP3B 0.293 0.025** 0.16 0.049** 0.831 0.623 0.781 0.163 TM I TERF1 0.748 0.363 0.054 0.066 0.197 0.124 0.073 0.07 TM IV TERF2 0.32 0.458 0.571 0.724 0.99 0.131 0.11 0.194

*Mann-Whitney test. cSquamous vs. nonsquamous histotype. bKruskal-Wallis test. x pT1 vs. pT2 vs. pT3. k Stage I vs. II vs. III. { Ki67 < median vs. z median score. **Statistically significant values.

www.aacrjournals.org 3393 Cancer Res 2009; 69: (8). April 15, 2009

Cancer Research

Correlation of relative expression of target genes with Table 3. Association between transcript deregulations clinical pathologic features. Generally, transcript overexpression and patient survival as assessed by univariate analysis was more frequent in squamous cell carcinoma than in BRCA1, BRCA2, XRCC2, XRCC3, adenocarcinoma, and in 12 genes ( Gene DDCt cutoff HR (95% CI) Unadjusted UBE2V2, XRCC1, UNG, ERCC1, XAB2, RRM1, RRM2, and TOP3A), the P value difference was statistically significant. For all genes belonging to BER, REP (except RRM2B), and DSB pathways (except XRCC4), and BRCA1 1.5 2.88 (1.01–8.18) 0.039* for ERCC2 associated with NER pathway and UBE2A belonging to BRCA2 1 2.1 (0.789–5.6) 0.13 PRR, a significant correlation with tumor grade was observed as XRCC2 3 1.56 (0.594–4.08) 0.36 being an overexpression rate higher in poorly differentiated XRCC3 2.5 2.7 (1.06–6.9) 0.04* XRCC4 carcinomas. Interestingly, most of the genes involved in DSBR 1 0.411 (0.12–1.41) 0.14 XRCC5 c pathway (BRCA1, BRCA2, XRCC2, and XRCC3) showed a significant 1 3.4 (1.38–8.38) 0.005 UBE2V2 1 3.72 (1.09–12.7) 0.024* correlation with gender, with overexpression being higher in MGMT 1 1.21 (0.352–4.16) 0.76 male patients compared with female patients. Finally, only XRCC1 1 2.01 (0.829–4.86) 0.12 UBE2N displayed a significant correlation with pN and stage of OGG1 1 4.02 (1.49–10.8) 0.003* c disease, whereas association with age was observed only for TOP3B UNG 2.5 2.06 (0.81–5.25) 0.008 (Table 2). For 57% (17 of 30) of the pathway genes, a significant XPA 1 1.52 (0.352–6.56) 0.57 association between transcript overexpression and proliferation XPC 1 0.741 (0.285–1.93) 0.54 index, determined as Ki67 protein expression, was observed. ERCC1 1 2.61 (0.947–7.22) 0.054 Coexpression of pathway genes. Coordinate expression was ERCC2 1 2.17 (0.786–5.98) 0.13 ERCC4 noted among the 29 target genes. Considering only Spearman 1 1.26 (0.516–3.09) 0.61 R ERCC5 —ND values >0.70, significant coexpression between genes belonging to ERCC6 BRCA1- —ND the same pathway have been observed for the DBSR ( XAB2 BRCA2, R BRCA1-XRCC3, R BRCA1-XRCC2, R —ND = 0.86; = 0.83; = 0.75; UBE2A 1 1.56 (0.649–3.75) 0.32 BRCA2-XRCC3, R BRCA2-XRCC2, R BRCA1-XRCC5, = 0.75; = 0.74; UBE2B 1 0.945 (0.363–2.46) 0.91 R = 0.70; all P < 0.01), PRR (UBE2A-UBE2B, R = 0.75; P < 0.01), and UBE2N 1 2.34 (0.901–6.06) 0.07 REP pathways (RRM1-TYMS, R = 0.76; RRM2-TYMS, R = 0.78; all RRM2B 1ND c P < 0.01). TYMS 4 3.21 (1.29–7.95) 0.008 Relative expression of target genes and association with RRM1 1.5 2.92 (1.12–7.61) 0.022* survival. To evaluate the correlation between expression regula- RRM2 3.5 1.34 (0.563–3.19) 0.51 TOP3A 1 2.11 (0.826–5.41) 0.11 tion of these genes and patients’ survival, relative expression was c TOP3B 1 3.57 (1.44–8.83) 0.003 categorized as 1 (overexpressed), 0 (not regulated), and 1 TERF1 DD 1 2.44 (0.933–6.37) 0.06 (underexpressed) using arbitrary Ct cutoff as thresholds. For TERF2 DD —ND most of the genes, a Ct cutoff = 1 was used, corresponding to a TERT 1 0.61 (0.24–1.54) 0.29 2-fold change, but for target genes characterized by higher transcript deregulation, the median DDCt values (from 1.5 to 4) were used. At univariate analysis, the deregulation of nine genes NOTE: P values are unadjusted log-rank tests. (BRCA1, XRCC3, XRCC5, UBE2V2, OGG1, UNG, TYMS, RRM1, and *Genes statistically significant at univariate analysis. c TOP3B) was found to be significantly associated with a poor Genes significant at the 0.05 level after correction by the Benjamini prognosis of patients [the hazard ratio (HR) and the unadjusted method. P values are shown in Table 3]. After the statistical correction following the Benjamini method, four genes were found to have significant correlation with survival at the P = 0.05 level: XRCC5, TYMS, UNG, and TOP3B. Figure 2 shows the Kaplan-Meier curves DNA replication genes analyzed, especially RRM1 and RRM2 and of the four genes with the adjusted P values. Among the TYMS, showed a significant overexpression in tumors. Secondly, clinicopathologic variables, tumor grade [3 versus 1–2; HR, 3.75 another feature of neoplastic cells is the unlimited replicative (1.48–9.52); P = 0.005] and stage [III versus I–II; HR, 5.63 (1.57– potential which is intimately related to the maintenance of 20.17); P = 0.008] resulted as significant prognostic factors at telomeres. Despite the inability to determine the fold change univariate analysis. The multivariate analysis by Cox was variation for TERT due to its low expression in normal lung tissues, performed including all the variables (genes and clinicopathologic a higher TERT transcript expression in f78% of tumor samples factors) significantly associated with survival at the univariate and has been detected. TERF1 overexpression was also significantly indicated that stage of disease was the only independent factor associated with squamous histotype and poorly differentiated predicting patients’ outcome [HR, 3.47 (1.05–11.4); P = 0.04]. tumors. Recently, the underexpression of the TERF1 transcript in lung tumor has been reported (21, 22), and this discrepancy may be related to the use of different housekeeping genes, as recently Discussion shown (18). High genomic instability resulting in frequent genomic The transcriptional differences of DNA repair/replication and rearrangements is another molecular feature of cancer cells. The telomere maintenance genes between tumor and normal lung results of the present study show a significant overexpression in 13 tissue reported in this study highlight some of the peculiar features out of 22 (59%) DNA repair genes analyzed in a cohort of patients of neoplastic transformation. One of these hallmarks is represented with NSCLC. Interestingly, among the DNA repair pathways, the by the increase of proliferation rate of neoplastic cells; indeed, all DSBR, mainly in its homologous recombination mechanism, was

Cancer Res 2009; 69: (8). April 15, 2009 3394 www.aacrjournals.org

DNA Repair/Replication Transcription Profiling in NSCLC mostly affected, with five genes in cluster I and the remaining two in lated with tumor grade, which has been previously shown to be an cluster II. It is likely that overexpression of the homologous independent prognostic factor for survival (30). Because patients recombination pathway could be a reflection of the high prolifer- with poorly differentiated carcinomas have a higher risk of ation rate of the tumors, as suggested by a significant correlation recurrence after tumor resection (30), these correlations strongly with the Ki67 score. Several studies support the hypothesis that in support the hypothesis that overexpression of DNA repair genes preneoplastic lesions, replicative stress leads to DSBs, which in turn, could be associated with metastasis development (31, 32). results in activation of DNA damage response that delays or prevents Despite the fact that an intrinsic higher repair capacity of cancer progression (23, 24). At the same time, this protective barrier NSCLC cells has been reported (6), transcript overexpression is creates a selective pressure that eventually favors the outgrowth of not proof of an increased enzymatic activity. Further studies malignant clones with genetic or epigenetic defects in the genome should confirm this hypothesis and deepen the molecular maintenance machinery (i.e., p53 or ATM; refs. 23–26). It should also mechanism underlying the overexpression of so many genes: the be considered that telomere attrition results in the activation of DNA high correlation observed between transcript expression levels of damage response signaling (27) and that the homologous recombi- some genes could, for example, suggest the presence of (a) nation machinery can play a key role in maintaining telomere length common activator(s). by telomerase-independent mechanisms (28). Only three DNA repair genes were found to be significantly The reduced overexpression of BRCA1, BRCA2, XRCC2, XRCC3, underexpressed in lung tumor tissue: XPC, XPA, and UBE2N. XPC and XRCC4 in females compared with males is intriguing because underexpression by at least a 2-fold change has been observed in women seem to have an increased susceptibility to tobacco f26% of patients with NSCLC, in agreement with other reports carcinogens, but have a better survival in each stage of the disease which detected hypermethylation of XPC promoter in 34% of compared with men (29). NSCLC (22% in smokers; ref. 33). No data on UBE2N transcript Interestingly, for most genes associated with DSBR and DNA- deregulation are currently available. However, in addition to its role REP pathways, transcript overexpression was significantly corre- in post-replication repair, UBE2N has been recently involved in the

Figure 2. Kaplan-Meier survival curves of patients with NSCLC divided by the status of regulation of the XRCC5, UNG, TYMS, and TOP3B genes, where ‘‘regulated’’ indicates patients with a significant transcript overexpression for each gene. P values are log-rank test–adjusted with the Benjamini correction.

www.aacrjournals.org 3395 Cancer Res 2009; 69: (8). April 15, 2009

Cancer Research repair of double-strand breaks by homologous recombination, and multigene predictive model, as recently reported (36). The UBE2N-deficient cells are sensitive to a wide range of DNA- prognostic indications emerging from the results of the present damaging agents (34). Data generated in this study indicate that study are therefore preliminary and require validations in larger NSCLC patients with underexpression of UBE2N have a better prospective studies of appropriate sample size. prognosis even if statistical significance was not reached (P = 0.07). In conclusion, this study found that lung tumors, compared with The large variability in interpatient transcript expression levels normal tissues, exhibit a significant overexpression of a wide could generate a substantial source of variable chemosensitivity in number of DNA-repair genes, mostly associated with DSBR and patients with NSCLC. However, because the homologous recom- PRR, DNA replication, and telomere maintenance pathways. bination machinery plays critical roles in regulation of sensitivity to Transcriptional overexpression of some of these genes showed the majority of chemotherapeutic drugs currently used in cancer strong correlation with a more aggressive clinical behavior and therapy (35), the broad overexpression of several components of could partly explain the low level of responsiveness of NSCLC to this repair pathway in lung tumor tissues could partially explain combination chemotherapy. the low chemosensitivity of patients with NSCLC. With regards to the prognostic effect of the genes investigated, 9 out of 26 had significant (unadjusted) P values at univariate Disclosure of Potential Conflicts of Interest analysis, and 4 genes maintained their prognostic relevance after No potential conflicts of interest were disclosed. the multiple testing correction: XRCC5 (belonging to DSBR), UNG (belonging to BER), and TOP3B and TYMS (both of the DNA-REP pathway). By contrast, no gene was retained in the multivariate Acknowledgments analysis, which indicated stage of disease as the only independent prognostic factor affecting patients’ outcome. This result was not Received 8/6/08; revised 1/15/09; accepted 1/16/09; published OnlineFirst 4/7/09. The costs of publication of this article were defrayed in part by the payment of page surprising and most likely due to the limited number of cases charges. This article must therefore be hereby marked advertisement in accordance collected, which similarly, did not allow the assessment of a with 18 U.S.C. Section 1734 solely to indicate this fact.

References Khuder SA, Willey JC. ABCC5, ERCC2, XPA and XRCC1 instability in human precancerous lesions. Nature 2005; transcript abundance levels correlate with cisplatin 434:907–13. 1. Nakamura H, Saji H, Idiris A, et al. Chromosomal chemoresistance in non-small cell lung cancer cell lines. 25. Nuciforo PG, Luise C, Capra M, Pelosi G, d’Adda di in situ instability detected by fluorescence hybridization Mol Cancer 2005;4:18. Fagagna F. Complex engagement of DNA damage in surgical specimens of non-small cell lung cancer is 13. Bepler G, Kusmartseva I, Sharma S, et al. RRM1 response pathways in human cancer and in lung tumor associated with poor survival. Clin Cancer Res 2003;9: modulated in vitro and in vivo efficacy of gemcitabine progression. Carcinogenesis 2007;28:2082–8. 2294–9. and platinum in non-small-cell lung cancer. J Clin Oncol 26. Bartek J, Lukas J, Bartkova J. DNA damage response 2. Markovic J, Stojsic J, Zunic S, Ruzdijic S, Tanic N. 2006;24:4731–7. as an anti-cancer barrier: damage threshold and the Genomic instability in patients with non-small-cell- 14. Takizawa M, Kawakami K, Obata T, et al. In vitro concept of ‘conditional haploinsufficiency’. Cell Cycle lung-cancer assessed by the arbitrarily primed polymer- sensitivity to platinum-derived drugs is associated with 2007;6:2344–7. ase chain reaction. Cancer Invest 2008;26:262–8. expression of thymidylate synthase and dihydropyrimi- 27. D’Adda di Fagagna F, Reaper PM, Clay-Farrace L, et al. 3. Zhou X, Kemp BL, Khuri FR, et al. Prognostic dine dehydrogenase in human lung cancer. Oncol Rep A DNA damage checkpoint response in telomere- implication of microsatellite alteration profiles in 2006;15:1533–9. initiated senescence. Nature 2003;426:194–8. early-stage non-small cell lung cancer. Clin Cancer Res 15. Scagliotti GV, Fossati R, Torri V, et al. Randomized 28. Dunham MA, Neumann AA, Fasching CL, Reddel RR. 2000;6:559–65. study of adjuvant chemotherapy for completely resected Telomere maintenance by recombination in human 4. Wei Q, Cheng L, Hong WK, Spitz MR. Reduced DNA stage I, II, or IIIA non-small-cell lung cancer. J Natl cells. Nat Genet 2000;26:447–50. repair capacity in lung cancer patients. Cancer Res 1996; Cancer Inst 2003;95:1453–61. 29. International Early Lung Cancer Action Program I, 56:4103–7. 16. Livak KJ, Schmittgen TD. Analysis of relative gene Henschke CI, Yip R, Miettinen OS. Women’s suscepti- 5. Bosken CH, Wei Q, Amos CI, Spitz MR. An analysis of expression data using real-time quantitative PCR and bility to tobacco carcinogens and survival after diagno- DNA repair as a determinant of survival in patients with the 2(-DDC(T)) method. Methods 2001;25:402–8. sis of lung cancer. JAMA 2006;296:180–4. non-small-cell lung cancer. J Natl Cancer Inst 2002;94: 17. Vandesompele J, De Preter K, Pattyn F, et al. Accurate 30. Sun Z, Aubry MC, Deschamps C, et al. Histologic 1091–9. normalization of real-time quantitative RT-PCR data by grade is an independent prognostic factor for survival in 6. Zeng-Rong N, Paterson J, Alpert L, Tsao MS, Viallet J, geometric averaging of multiple internal control genes. non-small cell lung cancer: an analysis of 5018 hospital- Alaoui-Jamali MA. Elevated DNA repair capacity is Genome Biol 2002;3:RESEARCH0034. and 712 population-based cases. J Thorac Cardiovasc associated with intrinsic resistance of lung cancer to 18. Saviozzi S, Cordero F, Lo Iacono M, Novello S, Surg 2006;131:1014–20. chemotherapy. Cancer Res 1995;55:4760–4. Scagliotti GV, Calogero RA. Selection of suitable 31. Sarasin A, Kauffmann A. Overexpression of DNA 7. Tsao MS, Aviel-Ronen S, Ding K, et al. Prognostic and reference genes for accurate normalization of gene repair genes is associated with metastasis: a new predictive importance of p53 and RASfor adjuvant expression profile studies in non-small cell lung cancer. hypothesis. Mutat Res 2008;659:49–55. chemotherapy in non small-cell lung cancer. J Clin BMC Cancer 2006;6:200. 32. Kauffmann A, Rosselli F, Lazar V, et al. High Oncol 2007;25:5240–7. 19. Popanda O, Schattenberg T, Phong CT, et al. Specific expression of DNA repair pathways is associated with 8. Rosell R, Skrzypski M, Jassem E, et al. BRCA1: a novel combinations of DNA repair gene variants and metastasis in melanoma patients. Oncogene 2008;27: prognostic factor in resected non-small-cell lung cancer. increased risk for non-small cell lung cancer. Carcino- 565–73. PLoSONE 2007;2:e1129. genesis 2004;25:2433–41. 33. Wu YH, Tsai Chang JH, Cheng YW, Wu TC, Chen CY, 9. Ceppi P, Volante M, Novello S, et al. ERCC1 and RRM1 20. Gibson G, Weir B. The quantitative genetics of Lee H. Xeroderma pigmentosum group C gene expres- gene expressions but not EGFR are predictive of shorter transcription. Trends Genet 2005;21:616–23. sion is predominantly regulated by promoter hyper- survival in advanced non-small-cell lung cancer treated 21. Lin X, Gu J, Lu C, Spitz MR, Wu X. Expression of methylation and contributes to p53 mutation in lung with cisplatin and gemcitabine. Ann Oncol 2006;17: telomere-associated genes as prognostic markers for cancers. Oncogene 2007;26:4761–73. 1818–25. overall survival in patients with non-small cell lung 34. Zhao GY, Sonoda E, Barber LJ, et al. A critical 10. Brooks KR, To K, Joshi MB, et al. Measurement of cancer. Clin Cancer Res 2006;12:5720–5. role for the ubiquitin-conjugating enzyme Ubc13 in chemoresistance markers in patients with stage III non- 22. Hu J, Sun L, Zhang C, Zhou X. Expression of initiating homologous recombination. Mol Cell 2007; small cell lung cancer: a novel approach for patient telomeric repeat binding factor 1 in non-small cell lung 25:663–75. selection. Ann Thorac Surg 2003;76:187–93. cancer. J Surg Oncol 2006;93:62–7. 35. Miyagawa K. Clinical relevance of the homologous 11. Takenaka T, Yoshino I, Kouso H, et al. Combined 23. Bartkova J, Horejsi Z, Koed K, et al. DNA damage recombination machinery in cancer therapy. Cancer Sci evaluation of Rad51 and ERCC1 expressions for response as a candidate anti-cancer barrier in early 2008;99:187–94. sensitivity to platinum agents in non-small cell lung human tumorigenesis. Nature 2005;434:864–70. 36. Raz DJ, Ray MR, Kim JY, et al. A multigene assay is cancer. Int J Cancer 2007;121:895–900. 24. Gorgoulis VG, Vassiliou LV, Karakaidos P, et al. prognostic of survival in patients with early-stage lung 12. Weaver DA, Crawford EL, Warner KA, Elkhairi F, Activation of the DNA damage checkpoint and genomic adenocarcinoma. Clin Cancer Res 2008;14:5565–70.

Cancer Res 2009; 69: (8). April 15, 2009 3396 www.aacrjournals.org

Non−Small Cell Lung Cancer Exhibits Transcript Overexpression of Genes Associated with Homologous Recombination and DNA Replication Pathways

Silvia Saviozzi, Paolo Ceppi, Silvia Novello, et al.

Cancer Res 2009;69:3390-3396. Published OnlineFirst April 7, 2009.

Updated version Access the most recent version of this article at: doi:10.1158/0008-5472.CAN-08-2981

Supplementary Access the most recent supplemental material at: Material http://cancerres.aacrjournals.org/content/suppl/2009/04/06/0008-5472.CAN-08-2981.DC1

Cited articles This article cites 35 articles, 8 of which you can access for free at: http://cancerres.aacrjournals.org/content/69/8/3390.full#ref-list-1

Citing articles This article has been cited by 5 HighWire-hosted articles. Access the articles at: http://cancerres.aacrjournals.org/content/69/8/3390.full#related-urls

E-mail alerts Sign up to receive free email-alerts related to this article or journal.

Reprints and To order reprints of this article or to subscribe to the journal, contact the AACR Publications Subscriptions Department at [email protected].

Permissions To request permission to re-use all or part of this article, use this link http://cancerres.aacrjournals.org/content/69/8/3390. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.