Abstract
This study determined whether expression levels of a panel of biologically relevant microRNAs can be used as prognostic or predictive biomarkers in patients who participated in the International Adjuvant Lung Cancer Trial (IALT), the largest randomized study conducted to date of adjuvant chemotherapy in patients with radically resected non–small cell lung carcinoma (NSCLC). Expression of miR-21, miR-29b, miR-34a/b/c, miR-155, and let-7a was determined by quantitative real-time PCR in formalin-fixed paraffin-embedded tumor specimens from 639 IALT patients. The prognostic and predictive values of microRNA expression for survival were studied using a Cox model, which included every factor used in the stratified randomization, clinicopathologic prognostic factors, and other factors statistically related to microRNA expression. Investigation of the expression pattern of microRNAs in situ was performed. We also analyzed the association of TP53 mutation status and miR-34a/b/c expression, epidermal growth factor receptor and KRAS mutation status, and miR-21 and Let-7a expression. Finally, the association of p16 and miR-29b expression was assessed. Overall, no significant association was found between any of the tested microRNAs and survival, with the exception of miR-21 for which a deleterious prognostic effect of lowered expression was suggested. Otherwise, no single or combinatorial microRNA expression profile predicted response to adjuvant cisplatin-based chemotherapy. Together, our results indicate that the microRNA expression patterns examined were neither predictive nor prognostic in a large patient cohort with radically resected NSCLC, randomized to receive adjuvant cisplatin-based chemotherapy versus follow-up only. Cancer Res; 70(21); 8288–98. ©2010 AACR.
Introduction
MicroRNAs are a class of small noncoding RNA species of 20 to 22 nucleotides that have been implicated in the control of many fundamental cellular and physiologic processes such as cellular differentiation, proliferation, apoptosis, and stem cell maintenance (1). MicroRNAs have been shown to play key roles in carcinogenesis, and some microRNAs have been categorized as “oncomiRs” as opposed to “tumor-suppressor miRs” (2). The expression patterns of microRNAs are often tissue specific, and certain cancer types can be classified based on microRNA expression profiles (3, 4). Importantly, the expression of certain microRNAs has been associated with chemoresistance (5, 6).
Early-stage non–small cell lung carcinoma (NSCLC) patients who undergo complete surgical tumor resection still develop distant metastases in 50% to 70% of cases, resulting in an overall 5-year survival rate of only 40% (7). The International Adjuvant Lung Cancer Trial (IALT) showed that adjuvant cisplatin-based chemotherapy improves the 5-year survival rate in this patient category by an absolute value of 4.1% (8). Two additional randomized studies have confirmed the prolonged 5-year survival rate in stage IB to IIIA NSCLC patients treated with adjuvant cisplatin-based chemotherapy. A third study, using carboplatin-paclitaxel, did not confirm this for stage IB disease (9–11). Currently, cisplatin-based adjuvant chemotherapy is considered part of the standard management of patients with completely resected stage II and III NSCLC. Nevertheless, the survival benefit of adjuvant chemotherapy remains limited to a subgroup of treated patients, and a recent update by the IALT investigators showed that the benefit of chemotherapy does not exist beyond 5 years of follow-up (12).
The IALT Biologic Program (IALT-Bio) was established, aiming to define biomarkers predictive for outcome of adjuvant chemotherapy as well as biomarkers prognostic for NSCLC overall survival. A predictive marker refers to a patient or tumor characteristic that is associated with therapy response. A prognostic marker refers to a characteristic of a patient or tumor at the time of diagnosis that can be used to estimate the outcome. It was shown that IALT cases with low excision repair cross-complementation group 1 (ERCC1), as well as low p27kip1 expression, benefit from adjuvant chemotherapy (7, 13). Importantly, ERCC1 expression remained predictive at 8 years of follow-up (12). Expression levels of multidrug resistance proteins MRP1 and MRP2 had no predictive value, but MRP2 was shown to be a strong prognostic factor (14). Based on our prior work as well as the literature, we analyzed in the IALT cohort the expression of seven biologically relevant microRNAs: miR-21, miR-29b, miR-34a/b/c, miR-155, and let-7a. Using a microRNA microarray approach, we have previously identified microRNA expression profiles unique for NSCLC subtypes. We showed that high miR-155 and low let-7a expression were associated with poor patient survival (15). Other reports also suggested a protective effect of let-7a expression in terms of survival and treatment outcome (16, 17) and showed that reduced let-7, which is known to interact with KRAS, is a prognostic factor in lung cancer and can contribute to carcinogenesis (18–20). Furthermore, we have previously reported on low miR-21 expression as a biomarker for favorable outcome of adjuvant chemotherapy in colon cancer, as well as increased miR-21 expression in epidermal growth factor receptor (EGFR) mutant tumors (21, 22). Numerous other studies support a prognostic or treatment-predictive role for miR-21 (23, 24).
MicroRNA-34a has been proposed as a prognostic marker of relapse in surgically resected NSCLC. Tumor-suppressor p53, frequently inactivated in NSCLC, is known to activate the transcription of miR-34a as well as miR-34b and miR-34c, containing a p53 binding site in their promoter (25–29). Finally, reduced expression of miR-29b, known to target de novo DNA methyltransferase 3A and 3B (DNMT3A and DNMT3B), can lead to global hypermethylation and silencing of various tumor-suppressor genes (30, 31). Methylation of tumor-suppressor p16 is a prognostic indicator in lung cancer. Thus far, a potential association between p16 expression and miR-29b expression has not been investigated (7).
Our main hypothesis was that expression of the seven selected microRNAs in tumor specimens from IALT patients could predict treatment response and survival benefit from adjuvant chemotherapy. Furthermore, we hypothesized that associations of microRNAs and prognosis would differ in lung adenocarcinoma patients compared with squamous cell carcinoma patients. As a secondary analysis, we evaluated whether microRNA expression status was associated with molecular markers relevant in lung cancer (EGFR, KRAS, and TP53 mutation status as well as p16 expression), for which data was available in the IALT cohort. Finally, we performed in situ investigation of microRNA expression in formalin-fixed paraffin-embedded (FFPE) tissue sections.
Materials and Methods
Patients and study design
Patients were enrolled in the IALT study, which randomized 1,867 patients with completely resected NSCLC, stages I through III, to receive adjuvant cisplatin-based chemotherapy or follow-up. See Supplementary Appendix S1 for a list of the IALT-Bio Participating Centers. FFPE tumor specimens were collected from patients at 28 centers in 14 countries that had recruited more than 10 patients (13). A total of 867 samples were reviewed centrally at the Centre Hospitalier Universitaire Albert Michallon (Grenoble, France), according to the histopathologic classification system adopted by the World Health Organization (WHO) in 2004. The amount and quality of 824 of 867 blocks were judged adequate for serial sectioning and experimental procedures. Ultimately, 783 were judged NSCLC after central review. Approval for the study was obtained from local institutional review boards according to the legal regulations in each participating country. For in situ hybridization experiments, additional tissue sections of NSCLC cases were obtained through the Department of Pathology at the University of Maryland, Baltimore, MD. The use of these sections was granted approval by the institutional review board of the National Cancer Institute and was approved by the institutional review board of the University of Maryland.
RNA isolation and quantitative reverse transcriptase-PCR
Sections were all processed at the National Cancer Institute by the Lung Cancer Laboratory (Medical Oncology Branch) and the Laboratory of Human Carcinogenesis. To ensure consistency in experimental conditions, uniform procedures, reagents, and equipment were used. Staff members involved in experimental procedures were jointly trained. All reagents, including all quantitative reverse transcriptase-PCR (qRT-PCR) reagents, were ordered in one batch, centrally stored, and distributed among staff members of the two participating laboratories. Furthermore, all qRT-PCR microRNA Taqman assays (see also below) were performed on the same equipment. Samples with insufficient or necrotic tumor material were omitted from further processing (n = 86). In total, 697 tumor samples, as well as 79 adjacent normal tissue specimens, were processed.
Specimens consisted of 10-μm FFPE sections. We disposed only of one section per patient case. Glass slides containing the tissue section were cut using a diamond pen in two parts. One part was stained with hematoxylin and eosin (H&E) and used for qRT-PCR. The other unstained part was available for in situ hybridization. H&E-stained tissues were marked by a lung pathologist for tumor area and, if present, normal tissue area under a BX40 light microscope (Olympus). Each area was macrodissected with sterile disposable scalpels (Cincinnati Surgical Company) for RNA isolation using the RecoverAll Total Nucleic Acid Isolation kit (Ambion). Forty nanograms of RNA were required for expression analysis of miR-21, miR-29b, miR-34a, miR-34b, miR-34c, let-7a, and miR-155.
qRT-PCR of microRNAs was performed using TaqMan microRNA assays (Applied Biosystems) and the 7900 HT-Fast real-time PCR system (Applied Biosystems). We used small nRNA U66 as endogenous normalization control, consistent with our prior report on miR-34a, miR-34b, and miR-34c expression (32). All assays were performed in triplicate by investigators who were blinded to clinical data of the sample cohort. MicroRNA expression was quantified as δCt values, where Ct = threshold cycle, δCt = (Ct target microRNA − Ct RNU66). δCt was calculated using RQ manager software, version 1.2 (Applied Biosystems).
Replicates with a Ct SD greater than 1, or, in case of U66 only, with an average Ct greater than 35, were omitted from further analysis (n = 58), resulting in a dataset of 639 cases. In case expression of a microRNA was available for both tumor and normal tissue, microRNA expression was additionally quantified as δδCt values, δδCt = (δCt target microRNA tumor tissue − δCt target microRNA matched normal tissue), for a separate analysis.
In situ hybridization
In situ investigation of microRNA expression in lung cancer was performed on a selection of the IALT-Bio cases as well as on lung cancer tissue sections procured from the University of Maryland (see also above). Probes for human miR-21, miR-34a, miR-155, and let-7a were used (Exiqon). U6 and Scramble probes were used as positive and negative controls, respectively. For in situ investigation of microRNA expression, we used the method as previously reported (21). In situ hybridization conditions for each individual probe were optimized using serial tissue sections from University of Maryland lung cancer cases because serial sections of IALT-Bio cases were not available. Methodology updates included the use of biotin-labeled microRNA probes (Exiqon) and a biotinyl tyramide–based system (GenPoint, Catalyzed Signal Amplification System; DAKO) and Vector NovaRed (Vector Laboratories) as a substrate (brown/red). Tissues were counterstained with Mayer's hematoxylin (blue). Images were taken on a BX40 light microscope using the Olympus DP70 digital camera and DP controller software (Olympus). Staining results were confirmed by an independent pathologist at the National Cancer Institute.
Statistical analysis
National Cancer Institute investigators involved in experimental procedures remained blinded to any of the clinical data. All statistical analyses were performed at the Institute Gustave Roussy (Villejuif, France). Initial analysis of microRNA expression values revealed heterogeneity in the data distribution of the tumor specimens between the two laboratories. The data were therefore standardized by subtracting the subgroup mean and division by the subgroup SD. Subgroups were well balanced with respect to treatment (adjuvant chemotherapy arm and control arm). To remove the potential bias in the results, the median standardized value was a priori chosen as cutoff to determine the microRNA expression status. MicroRNA expression was defined as negative when the expression value was lower than the median and positive when the expression value was equal or greater than the median. To test for differences between microRNA-negative and microRNA-positive samples, comparisons had to take into account the study centers. Therefore, logistic regression stratified by center was used both for univariate and multivariate analyses. The prognostic value of microRNA status and chemotherapy for survival were studied using a Cox model. As in the original IALT analysis, the Cox model included every factor used in the stratified randomization [center, tumor stage, and type of surgery (pneumonectomy, lobectomy, segmentectomy), plus clinical and histologic prognostic factors (age <55, 55–64, >64 years), sex, WHO performance status, nodal status, lymphoid infiltration (not intense, intense), and the revised histopathologic type (adenocarcinoma, squamous cell carcinoma, other NSCLC); ref. 8]. All other factors that were statistically related to microRNA expression in the multivariate logistic model (P < 0.05) were added to the survival Cox model. Trend tests were performed using the continuous standardized values instead of the dichotomized values (positive versus negative). Analyses were also performed using distribution quartiles of the standardized values.
The predictive value of each microRNA was studied by testing the interaction between microRNA expression and the attributed treatment (chemotherapy or no chemotherapy) in the same Cox model. To study the association between microRNAs and other markers, logistic regression of the marker on either positivity or standardized value of microRNA expression (denoted “trend”), stratified by center, was used. For analysis using normal and/or normal-tumor matched expression levels, the Cox survival model had to be simplified due to the low number of available cases. Stratification by two regions of the world (Western Europe versus other parts of the world) was used. Furthermore, a smaller number of adjustment variables were entered in the model: only stage (the only significant prognostic factor) plus histology and the variable(s) correlated with each microRNA. To study variation of the prognostic effect with histology, only squamous cell carcinoma and adenocarcinoma were considered.
All analyses were performed with long-term survival data (12). All reported P values were two-sided. P values <0.01 were considered statistically significant to limit the risk of false-positive results. All analyses were performed using SAS software, version 9.1 (SAS Institute, Inc.).
Results
After quality control, 639 IALT-Bio samples remained that had measurements of at least one microRNA. Patient characteristics are listed in Supplementary Table S1. Patient characteristics varied in relation to the sample size per evaluated microRNA.
Association of microRNA expression and clinicopathologic covariates
As microRNA expression is thought to vary according to histotypes and could be associated with tumor characteristics, we explored associations between individual microRNA expression patterns and the clinicopathologic variables.
The associations remaining significant in the multivariate analysis are summarized in Table 1. miR-21 status was associated with histology (P = 0.04) and lymphoid infiltration (P = 0.04); miR-29b status was associated with age (P = 0.03), histology (P ≤ 0.0001), and lymphoid infiltration (P = 0.005); miR-34a status was associated with histology (P = 0.0002), lymphoid infiltration (P = 0.03), and lymphatic invasion (P = 0.04); miR-34b status was associated with disease stage (P = 0.04) and histology (P = 0.01); miR-34c status was associated with histology (P = 0.0002); miR-155 status was associated with lymphoid infiltration (P = 0.001); and no covariates were associated with let-7a status. Supplementary Table S2 summarizes the association of microRNA expression with covariates (univariate analysis or trend test). See also Supplementary Tables S3 to S9 for a complete overview of all univariate analyses per microRNA.
Summary of significant associations between covariates and dichotomized microRNA expression values (multivariate logistic model)
Prognostic analysis
Next, we investigated the prognostic value of microRNA expression. The expression status of none of the microRNAs was significantly prognostic for outcome (see Fig. 1, as well as Table 2 for a summary of results). However, there is a borderline prognostic effect of miR-21 expression on overall survival with a worse survival for miR-21–negative cases (P = 0.06; trend: P = 0.01). Although there is no prognostic effect of miR-34b as well as miR-34c positivity on survival, there is nevertheless a heterogeneous effect between quartiles (P = 0.005 and P = 0.004, respectively), which is not associated with a specific trend.
Kaplan-Meier estimates of overall survival according to expression of miR-21 (A), miR-29b (B), miR-34a (C), miR-34b (D), miR-34c (E), miR-155 (F), and Let-7a (G).
Prognostic analyses
On the population evaluable for both let-7a and KRAS (n = 582), let-7a was a good prognostic indicator [hazard ratio (HR), 0.79; 95% confidence interval (95% CI), 0.64–0.99; P = 0.04]. However, in a total of 638 samples evaluable for let-7a, this prognostic association was no longer significant (HR, 0.84; 95% CI, 0.68–1.04; P = 0.11).
Predictive analysis
As microRNA expression status has been associated with treatment response, we next investigated the predictive value of the seven microRNAs (21, 33). Expression status of none of the microRNAs had a predictive effect on survival. See Table 3 for a summary of results as well as Supplementary Fig. S1 for Kaplan-Meier estimates of overall survival according to treatment in microRNA-positive patients compared with microRNA-negative patients.
Predictive analyses
Additional prognostic and predictive analyses
As initial analysis of microRNA expression values revealed heterogeneity in the data distribution between the two laboratories, separate prognostic and predictive analyses were performed in both groups to exclude an effect of the standardization procedure on the study outcome and to confirm that negative results were not due to inconsistent assay results.
There was no difference in the prognostic effect of any of the seven microRNAs in both groups. There was no difference in the predictive effect of miR-21, miR-34a, miR-34b, miR-34c, or let-7a in both groups. There was a borderline difference in the predictive effect of miR-29b in both groups (P = 0.03 for comparing the interactions and P = 0.04 for comparing deviations from the group averages), but the heterogeneity and trend test did not show anything consistent. There was a difference in the predictive effect of miR-155 in both groups [P = 0.007 for comparing the treatment interaction; however, the difference was not significant for comparing deviations from the group averages (P = 0.12) or in the trend tests (P = 0.07)]. As the variations of the HRs in the quartiles within each group were not consistent, the likelihood of a predictive effect is low. Furthermore, for predictive analyses, the treatment effect of chemotherapy was borderline significantly different between both groups, and differences in predictive effects of microRNA expression profiles should therefore be interpreted with respect to the average treatment effect in the group. More details regarding these additional analyses are available upon request.
Association of microRNA expression with other IALT-Bio markers
As we did not find a significant predictive or prognostic effect of any of the candidate microRNAs, we subsequently assessed in a secondary analysis the association of microRNAs and other molecular markers relevant in lung cancer. We were able to perform these analyses because data were or became available in the course of the study. It was hypothesized that a combined protein and microRNA expression signature could also potentially be of predictive and/or prognostic value.
Because p53 activity is associated with miR-34a/b/c transcription, we looked at an association between TP53 mutations and miR-34a/b/c expression. IALT-Bio investigators previously reported on the prognostic and predictive values of TP53 mutations (exons 5–8) and KRAS mutations (codons 12 and 13) in the IALT-Bio cohort (34). TP53 mutations were shown in 46% of patients. After 8 years of follow-up, no prognostic value for TP53 mutation status was shown in all cases grouped together. It was recently suggested that patients with nonadenocarcinoma NSCLC and TP53 wild-type might benefit from adjuvant cisplatin-based chemotherapy, whereas it might be less beneficial in patients with mutated TP53 (34). Nevertheless, there was no global association between TP53 mutations and miR-34a status (positivity: P = 0.70; trend: P = 0.33). The association between TP53 mutations and miR-34a positivity did not vary with histology (P = 0.23). Equally, there was no global association between TP53 mutations and miR-34b or miR-34c (positivity: P = 0.91, trend: P = 0.49; positivity: P = 0.57, trend: P = 0.35, respectively), and this did not vary with stratification for histology (P = 0.47; P = 0.78, respectively).
Next, we looked at potential associations between miR-21 expression and EGFR mutation status, as increased miR-21 expression was previously detected in EGFR mutant tumors, EGFR being one of the key markers in NSCLC biology (22). Only 20 tumors had EGFR mutations and it was therefore not possible to perform a subanalysis by histology. We did not detect a global association between occurrence of EGFR mutations and miR-21 expression (positivity: P = 0.73, trend: P = 0.29).
In the aforementioned study, KRAS mutation was suggested to be a biomarker of poor prognosis in patients with nonadenocarcinoma (34). Because KRAS is one of the known targets of let-7a, we investigated the association of KRAS mutations and let-7a expression. We found that there is a global borderline association between KRAS mutations and let-7a (positivity: P = 0.10, trend: P = 0.03). The association between KRAS mutations and let-7a positivity did not vary with histology (P = 0.42).
Finally, we assessed a potential association between miR-29b and p16 expression. De novo DNMT3A and DNMT3B are targets of miR-29b, and p16 expression is silenced in lung cancer due to hypermethylation. However, we found that there is no global association between p16 expression as assessed by immunohistochemistry and miR-29b (positivity: P = 0.14, trend: P = 0.10). The association between p16 IHC positivity and miR-29b positivity did not vary with histology (P = 0.86).
Analyses using tumor adjacent normal tissue
In total, 79 samples have measurements of a least one matched tumor and normal microRNA. The median expression ratio for tumor and normal tissues (T/N) was, as expected, positive for the oncomiRs miR-21 [0.95 ± 1.33 (±SD)] and miR-155 (0.05 ± 1.13), indicating a higher expression in the tumor tissue compared with the normal tissue. Relative T/N expression was, as expected, negative for the tumor-suppressor miRs miR-29b (−0.68 ± 1.33), miR-34a (−0.40 ± 1.38), miR-34b (−0.37 ± 2.88), miR-34c (−0.51 ± 2.73), and let-7a (−0.61 ± 1.40).
In situ hybridization
We performed in situ hybridization of the seven tested microRNAs. For the IALT-Bio cohort, none of the tested microRNAs were prognostic or predictive. MicroRNA-21 was the only microRNA with a suggested deleterious prognostic effect of lower values. Figure 2 shows a representative pattern of in situ miR-21 staining in a NSCLC adenocarcinoma case with high miR-21 expression according to the qRT-PCR assay. In general, we observed that staining was exclusively present in the cytoplasm for all probes except, as expected, for the experimental control, endogenous small nucleolar RNA U6, which showed nuclear expression. As for miR-155 staining, we frequently observed that the staining was localized to the outer end/leading edge of each nodule of squamous cells within the tumor. Additionally, Fig. 2 shows miR-34a– and let-7a–positive cases (data for miR-29b, miR-34b/c not shown).
In situ hybridization for microRNAs in lung tumors. Lung tumors were hybridized with biotin-labeled microRNA probes that were detected using a biotinyl tyramide–based system and Vector NovaRed as a substrate (brown/red). Tissues were counterstained with Mayer's hematoxylin (blue). Tissue sections were from both the IALT cohort and the University of Maryland. Representative slides with positive staining cells are shown for various cancer types, but was not limited to these specific types. Positively staining cells are shown for miR-21 in adenocarcinoma (A), miR-34a in a large-cell carcinoma with neuroendocrine features (B), miR-155 in squamous cell carcinoma (C), and let-7a in adenocarcinoma (D). Staining was limited to the cytoplasm, as expected, and was diffusely localized within tumor types except for miR-155 in squamous cell carcinoma. Mir-155 was expressed in cells on the edge of squamous cell nodules within squamous cell carcinomas. U6 (E) and Scramble (F) probes were used as positive and negative controls, respectively.
Discussion
Based on preclinical studies as well as translational studies using clinical specimens, microRNA expression levels have been suggested to be biomarkers for prognosis as well as treatment outcome in numerous malignancies (35–37). Expression of seven candidate microRNAs was hypothesized to predict outcome of adjuvant chemotherapy in NSCLC, aiming to establish routine markers for selection of patients for adjuvant chemotherapy to improve treatment outcome.
We tested this hypothesis in a very large NSCLC cohort of 639 cases, randomized for treatment, using extensive quality and methodologic control measures. We disposed of only one tumor section per case and made a selection of seven biologically relevant microRNAs to be analyzed for associations with clinicopathologic covariates, prognosis, and treatment outcome.
Interestingly, microRNA expression correlated with histology for five of seven microRNAs assessed: miR-21, miR-29b, and miR-34a/b/c. Recently, a study on microRNA expression profiling was reported using 290 FFPE specimens from the EAGLE study, a population-based case-control study in lung cancer (38). A microRNA signature, including miR-29b and let-7a, was identified, which strongly differentiated histologic subtypes (i.e., adenocarcinoma and squamous cell carcinoma) in male patients who were smokers. Interestingly, cigarette smoking intensity showed an inverse correlation with let-7 expression in female adenocarcinoma patients and a positive correlation with miR-21 expression in male squamous cell carcinoma patients. Furthermore, only in the group of male smokers with squamous cell carcinoma, a prognostic miR signature could be established, which included miR-34c-5p and miR-34a (38). Unfortunately, we did not dispose of smoking status data for IALT cases to confirm these relationships.
We found that positive expression of miR-21, miR-29b, miR-34a, and miR-155 was associated with intense lymphoid infiltration or the presence of lymphatic invasion (miR-34a only). In this regard, it was previously shown that miR-21 expression is associated with an increased rate of lymph node metastasis in breast cancer and colorectal cancer patients (39, 40).
Overall, we were not able to show a prognostic value of any of the seven microRNAs in this patient cohort. Upregulated miR-21 expression has been previously associated with worse outcome in NSCLC cases (23, 24, 41). However, the study by Markou and colleagues consisted of few cases, 48, which made use of matched tumor and normal tissues (24). Consequently, results reported were based on the T/N expression ratio. In contrast, for our main analysis, we made use of tumor tissue expression values. Additionally, the study by Raponi and colleagues made use of a microRNA array approach (23). Another characteristic of our study was that we disposed only of FFPE tissue sections instead of frozen tumor specimens.
We established an association between let-7a expression and KRAS mutation status, albeit only in the multivariate analysis on the means and not the dichotomized expression values. The significance of this finding is unclear. Thus far, it has been reported that the KRAS-LCS6 polymorphism results in upregulation of the KRAS gene and concomitant downregulation of let-7 (42). However, this polymorphism was not found to be associated with KRAS mutations (43).
We could not confirm an association between EGFR mutation status and miR-21 expression, as reported previously (22).
Additionally, we could not show a predictive effect of any of the seven microRNAs on treatment outcome, which consisted of cisplatin-based adjuvant chemotherapy. Drugs combined with cisplatin among the 639 patients were etoposide (54%), vinorelbine (33%), vinblastine (8%), or vindesine (6%; ref. 8). We previously showed that high microRNA-21 expression is related to fluoropyrimidine resistance in colorectal cancer as well as gemcitabine resistance in pancreatic cancer, and the predictive value of microRNA expression may be chemotherapy and/or tumor specific (21, 44).
As the prognostic and/or predictive value of microRNAs might be dependent on the expression of more than one microRNA, we performed cluster analysis using standardized values of the seven microRNAs. However, no combination of expression profiles of the seven microRNAs was found to be prognostic or predictive (data not shown). Finally, in situ hybridization was performed using specimens from this cohort as well as specimens obtained from the University of Maryland. Treatment conditions for microRNA in situ hybridization vary per probe and histology. Conditions had to be optimized using serial sections of an independent test cohort for which qRT-PCR data was available. For the IALT cases, a disadvantage was that we were completely blinded to origin of cases, specimens being collected from a worldwide multicenter study, with nonstandardized fixation and storage conditions. Fixation and procurement variations all affect the efficacy of in situ hybridization to a much bigger extent than microRNA isolation and qRT-PCR (45, 46). Nevertheless, in standardized test cases as well as IALT-Bio cases, in situ expression of microRNAs could be shown.
In conclusion, many reports exist on microRNA expression profiling and the prognostic and/or treatment predictive impact of microRNA expression in NSCLC. However, many studies are limited by small sample size and by not being randomized. This study constitutes the largest ever group of NSCLC patients analyzed for the prognostic and predictive value of microRNA expression. The seven target microRNAs chosen for evaluation, miR-21, miR-29b, miR-34a/b/c, miR-155, and let-7a, were neither prognostic nor predictive in this patient cohort. Further studies, for example, making use of a microRNA array approach, are warranted in NSCLC, to identify the prognostic or predictive value of expression of other microRNAs that were not included in this study.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Acknowledgments
Grant Support: Unrestricted research grant from Eli Lilly and grants from the Programme Hospitalier de Recherche Clinique 2005, as well as Cancéropôle Rhône-Alpes.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Footnotes
Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/).
- Received April 20, 2010.
- Revision received July 19, 2010.
- Accepted August 14, 2010.
- ©2010 American Association for Cancer Research.
References
Volume 70, Issue 21
