Expression of DNA repair and replication genes in non-small cell lung cancer (NSCLC): a role for thymidylate synthetase (TYMS)

Current cytotoxic chemotherapy for NSCLC patients mostly includes doublets of platinum (alkylating agents causing DNA adducts) with taxanes or vinorelbine (plant alkaloids interfering with the stability of tubulins), with gemcitabine (a nucleoside analog interfering with DNA and RNA synthesis) or with pemetrexed (a folate antimetabolite). These are administered following surgical removal of the tumor (adjuvant setting), preceding surgical intervention (neoadjuvant) and in inoperable cases or metastatic disease (1^st line treatment). Cytotoxic chemo-therapeutics are applied based on in vitro and preclinical evidence of efficient cancer cell killing; after acceptable tolerance by the patients is ensured, drug combinations are mostly chosen empirically and cannot be distinguished for their efficacy in NSCLC patient outcome [1].

In an effort to rationalize treatment, predictive markers for cytotoxic drug efficacy in NSCLC patients have been extensively investigated over the last 15 years. Such markers mostly include molecules responsible for DNA synthesis and repair. With respect to DNA repair genes, low ERCC1 (excision repair cross-complementing rodent repair deficiency, complementation group −1) expression has been associated with a better outcome upon platinum compounds initially in patients with ovarian cancer [2] and later in patients with NSCLC [3‐11]. High BRCA1 expression was associated with a better response to taxanes and sensitivity to docetaxel or docetaxel/gemcitabine (predictive value) [12, 13] and with a worse overall survival in NSCLC patients both in early and advanced disease setting (prognostic value) [14, 15]. With respect to DNA synthesis genes, low expression of ribonucleotide reductase M1 (RRM1), one of the two subunits of an enzyme essential for the production of deoxyribonucleotides prior to DNA synthesis in S phase of dividing cells, was associated with clinical benefit from neoadjuvant cisplatin/gemcitabine in NSCLC patients [7, 11, 16]. Thymidylate synthetase (TYMS), an enzyme that is critical in maintaining the dTMP (thymidine-5-prime monophosphate) pool for DNA synthesis and repair, is considered among the targets of newer antifolate drugs, such as pemetrexed [17, 18], while allelic variants of TYMS have been shown to interfere with platinum activity in vitro [19]. In most studies, high TYMS expression is reported as an unfavourable prognostic factor [20‐23], although it was occasionally associated with favourable outcome [24].

As yet, however, no conclusive evidence has been provided on the prognostic and/or predictive value of ERCC1, RRM1, BRCA1 and less of TYMS, nor has any consensus been reached on how these markers could be used for the routine assessment of NSCLC patients (critically reviewed in [25‐28]). Reasons for this discrepancy include, among others, study design [29], method and methodology differences [25, 27, 30], type of tissue specimen [31], and primary or metastatic origin of the tissue examined [32, 33]. Beyond methodological issues, the inability of comprehending the value of BRCA1, ERCC1, RRM1, and TYMS (B, E, R, T) expression in the assessment of NSCLC patients may be related to the nature of these markers. B, E, R, T are in fact expressed in the normal lung tissue in the frame of normal regeneration [34‐36], and might thus contribute to the generic response of patients to cytotoxic therapy, which is given systemically and does not target tissues or tumors specifically.

In the present study, we investigated the prognostic/predictive impact of single and combined B, E, R, T mRNA expression profiles on the outcome of NSCLC patients who had been treated with platinum- and/or taxane-containing schemes in the adjuvant and 1^st line setting. Single and combined intratumoral B, E, R, T expression was compared with clinicopathologic parameters and interactions with treatment type in relevance to outcome. Other than has been practiced as yet though, we evaluated B, E, R, T expression in NSCLC in comparison to normal lung tissue located distally to the tumor site, in order to mitigate the effects of the field cancerization phenomenon [37].

The outline of tissue material and patient data involved in this study is presented in the REMARK diagram in Figure 1 . In total, 361 NSCLC tumors were examined, for which histologic material (routinely diagnosed formalin-fixed paraffin-embedded tissue [FFPE] material) was available. Tissue specimens were collected from the Tumor Tissue Repository of Hellenic Cooperative Oncology Group (HeCOG). Corresponding demographic, clinicopathological and follow-up data had been registered for these patients in the frame of clinical service in HeCOG-affiliated hospitals. All patients had signed an informed consent form permitting the use of their biologic material for research purposes. The study was approved by the Bioethics Committee, School of Medicine, Aristotle University of Thessaloniki.

The clinicopathologic characteristics of registered patients and tumors are listed in Additional File 1 Table 1. Patients were considered as never, current and former smokers, the latter if they had quit the habit more than one year before treatment initiation. Throughout this manuscript, stage corresponds to disease stage at initial diagnosis. The treatment regimens administered and the corresponding patients’ outcome are shown in Additional File 1 Table 2. In order to examine homogeneous groups of patients we analyzed the following three subgroups; Subgroup A, including 1^st line treated chemo-naïve patients who initially presented with advanced disease (n = 160); Subgroup B, including all 1^st line treated patients (1^st line chemo-naïve and patients who relapsed upon adjuvant or neoadjuvant treatment) (n = 180); and Subgroup C, including chemo-naïve patients who were treated either in the (neo)adjuvant or metastatic setting (n = 192).

In tumors, B, E, R, T mRNA expression did not vary with age but it was higher in men than in women for BRCA1 and ERCC1 (Table 1). The expression of BRCA1 and TYMS was affected by smoking being higher in smokers than in never and former smokers (Figure 2A). Disease stage seemed to have the same impact on the expression of all genes, being higher in stage IIIA disease; however, significant differences were observed only for TYMS and, in particular, for BRCA1 (Figure 2B). Remarkably, B, E, R, T expression was significantly higher in SCC in comparison to adenocarcinomas and LCC, with adenocarcinomas expressing the lowest levels of these mRNAs (Figure 2C).

The difference observed for BRCA1 and TYMS with respect to smoking status could not reliably be assessed in the normal lung samples, since 35 out of 40 of these corresponded to smokers. No difference in B, E, R, T mRNA expression was observed in this group with regard to age, gender and stage, while BRCA1 and TYMS expression was lower in the normal lung parenchyma of patients with adenocarcinomas in comparison to SCC (Mann–Whitney p = 0.001 in both cases).

Within tumor groups (entire series and tumors paired to normal samples), the same pattern of B, E, R, T mRNA expression was observed (Table 2), involving strong correlations of all markers to each other. In comparison, however, TYMS expression in normal lung samples was not related to BRCA1 or ERCC1, while it correlated positively with RRM1. These data indicate that, other than in the normal lung, common regulatory stimuli affect B, E, R, T mRNA expression in NSCLC tissues.

So far, successful drug-specific predictive markers in NSCLC and in most malignancies include molecular phenotypes reflecting somatic genetic alterations; for example, small scale mutations or large scale structural gene changes. Regarding BRCA1, ERCC1, RRM1 and TYMS in NSCLC, however, differences in the “expression level” of the respective molecules are considered as markers, usually by referring to high vs. low expression. High vs. low is a qualitative description deriving from (semi)quantitative measurements in the case of q-PCR gene expression assessments. Undoubtedly, in order to understand high vs. low we need to refer to a standard. By using an external “normal” cell line RNA as a standard (as published before for ERCC1) [3], two of these genes, RRM1 and TYMS, appeared as very low expressed in tumors. However, even when coined as “normal”, cells in culture are, in fact, transformed; therefore, having acquired the ability of continuous division they inherently produce enzymes such as RRM1 and TYMS. In addition, cells in culture constitute a homogeneous system by definition. In comparison, cells in normal tissues comprise a heterogeneous environment and divide upon interaction with their surrounding cells. This explains the very low expression of RRM1 and TYMS in the normal lung tissue in comparison to the reference standard, as observed here.

Our approach was to use normal lung tissues located distally to tumors as a standard. Given the retrospective type of this study, it was difficult to find a large number of such normal samples with matched tumors. The limited number of samples in this reference group (12.5% of the total number of cases examined) may be considered as a drawback for the normal range of RQ values obtained in this study. However, with respect to the RQ values in the matched tumor samples at least, the group was considered representative for the remaining 87.5% of single tumor samples. The level of BRCA1, ERCC1, RRM1 and TYMS transcripts in NSCLC tumors of all histologic types, and especially SCC, was mostly within the normal range or increased. These observations are in line with previous reports for unchanged ERCC1 [36] or increased ERCC1 [35] and for increased TYMS mRNA expression [40, 41] in tumors vs. normal lung, whereby TYMS has also been associated with enhanced proliferation activity [22]. Genetic or epigenetic BRCA1 silencing in NSCLC could not be inferred from our study, contrary to a previous report [34], since only two tumors in our series exhibited very low expression of this gene. RRM1 mRNA expression was found lower in NSCLC compared to the normal lung and the condition was attributed, at least in part, to LOH at 11p15.5 [16], which may also have accounted for the low RRM1 observed in 22% of tumors in this study. For BRCA1, ERCC1 and TYMS, expression below the normal range was encountered in <10% of the tumors examined. Overall, only in 11 cases (<4%) tumor B, E, R, T RQ values were lower than any normal ones, which would indicate absence of gene expression. Thus, although genetic / epigenetic changes may have been present in these cases, low expression of B, E, R, T indicative of gene pathology in NSCLC tumors could not be considered as a frequent event. In addition, considering that all mRNA markers strongly correlated with each other in tumors, our findings on mRNA expression of the two DNA-repair and the two DNA-replication genes in NSCLC reflect the expected patterns of tissues responding to increased demands for both these cell functions. Of note, our results are limited to gene expression, which does not preclude that for example, these particular DNA repair genes would be dysfunctional in NSCLC. BRCA1 and ERCC1 mRNA expression has as yet been addressed in a rather simplistic way in tissue studies with q-PCR, abolishing the possibility to investigate which mRNA variants are being produced from these genes. At least for ERCC1, detection of a particular splice variant might be important for the evaluation of platinum resistance, as has been reported for ovarian cancer cells [42].

In the noticeably heterogeneous patient cohort of the present study, none of the four markers was universally associated with OS and DFS in the adjuvant setting or PFS in the two different first-line settings. This finding was rather expected, since the prognostic value of the markers examined remains contradictory, as supported in recent meta-analyses for ERCC1, which also stress the need for standardization methods to assess gene expression by qPCR in tumors [27, 30]. The only independent predictors for OS, after adjusting the four mRNA markers for clinicopathologic parameters and treatment schemes, were disease stage (unfavorable if advanced) and histology type, meaning adenocarcinomas, which conferred a favorable outcome in comparison to squamous and large cell carcinomas. However, this latter finding must be considered as cohort specific, since the majority of patients in this study were chemo-naïve adenocarcinoma patients and received 1^st line chemotherapy, where adenocarcinomas are the best responders among NSCLC [43]. No interaction between histology type and BRCA1, ERCC1, RRM1, or TYMS transcript levels was observed in terms of outcome, despite the strong association of continuous RQ values with the same parameter. The relative expression of the four genes was similar in adeno- and large cell carcinomas but, in comparison to these, it was significantly increased in squamous cell carcinomas, in line with previously published data [4, 40]. Apparently, the expression of these genes did not predict for the outcome of adenocarcinoma patients in our series, in line with recent reports with immunohistochemistry for Ercc1 and Brca1 [44, 45].

The heterogeneous treatment regimens administered for advanced disease in this study constituted a very useful template to assess drug-specific interactions of BRCA1, ERCC1, RRM1, and TYMS, particularly with respect to patients’ PFS, an endpoint which is closely related to treatment efficacy although more prone to errors during assessment [46]. It was interesting to observe the same treatment-related patterns of patient outcome for tumors with low and high expression of the markers examined, while tumors with normal expression of the same marker had different prognosis. Plausibly, the first approach for the interpretation of this finding would be to consider it an artifact due to the small number of tumors with low BRCA1, ERCC1, RRM1, and TYMS mRNA expression. However, this finding may also suggest that drug interactions, if any, with the molecules examined at the protein level may follow bell-shaped curves, as known for various biologic response models involving enzymes (e.g., concerning the cellular fate upon different levels of superoxide dismutase [47]) or drug efficiency [48]. To prove this, functional studies in controlled systems with multicellular components are needed. It remains worthy considering, however, that the usually applied one-point cut-offs may mask the information needed for the evaluation of the herein investigated as well as for other mRNA markers.

In this context, normal ERCC1 transcript levels were associated with a favorable outcome in patients who received the platinum/taxanes doublet. Considering that normal ERCC1 mostly corresponds to low ERCC1 as addressed in the literature, this finding was in keeping with the reported platinum specificity of low ERCC1 [27, 30, 49]. However, a clear effect as would be expected for normal ERCC1 on patients treated with platinum-based regimens without taxanes was not observed in the present study.

Surprisingly, TYMS followed the same predictive pattern of significance as ERCC1 for platinum without taxanes, but an inverse pattern than ERCC1 for taxanes without platinum. For this molecule, significant results were obtained with respect to platinum (high TYMS – favorable) and taxanes (high TYMS – unfavorable), and less so for the combination of the two drugs (high TYMS – favorable). Since the sole independently significant interaction for patient outcome was TYMS with taxanes, patients with tumors expressing high TYMS mRNA, as compared to a normal lung tissue standard, may be spared receiving taxanes. This is new evidence, meriting clarification at the molecular level for the known microtubule stabilizers. In addition, there was an interaction of high TYMS with platinum as well, probably reflecting the increased efficacy of this drug in killing cells featuring active DNA replication. TYMS, an enzyme that is essential for DNA synthesis, has not yet been studied as a “platinum-target” molecule, as has been ERCC1 in the context of the nucleotide excision repair pathway. In NSCLC therapeutics, TYMS counts as an established target mostly being associated with the efficacy of the newer antifolate pemetrexed, which is a multi-targeted drug, nonetheless [50, 51]. Another observation from the present study concerns the widely used taxanes/platinum combination in advanced NSCLC patients. ERCC1 and TYMS were related in an opposing manner to patient outcome upon this treatment, with TYMS being more specific. It would be interesting to investigate the drug-specific predictive value of combined profiles of ERCC1 and TYMS in a larger series of patients with adequate statistical power.

This study shows that the expression of genes participating in DNA repair and replication are mostly up-regulated in NSCLC, as would be expected for proliferating cells, without indications for a severe dysfunction of these two cellular functions at the mRNA level. To further elucidate the cellular molecular involvement in cytotoxic drug targeting, an extended cohort of NSCLC patients is being treated with the Docetaxel / Carboplatin combination (plus bevacizumab when indicated) in the 1^st-line setting in HeCOG-affiliated hospitals. Detailed biomarker studies and correlations with outcome in this cohort are expected to shed more light on the interaction of DNA replication and repair molecules with chemotherapy in the quest of refining current treatment strategies for patients with NSCLC.