Investigating molecular alterations to profile short- and long-term recurrence-free survival in patients with primary glioblastoma
Affiliations: Department of Translational Research and of New Surgical and Medical Technologies, University Hospital of Pisa, Pisa I‑56126, Italy, Genomic Section, Pisa Science Foundation, Pisa I‑56121, Italy, Department of Neurosurgery, University Hospital of Pisa, Pisa I‑56124, Italy, Radiotherapy Unit, University Hospital of Pisa, Pisa I‑56126, Italy
- Published online on: September 23, 2015 https://doi.org/10.3892/ol.2015.3738
- Pages: 3599-3606
This article is mentioned in:
Glioblastoma (GB) is the most common and aggressive type of malignant primary brain tumor. The current standard of care for GB, as established by Stupp et al (1), is maximal safe surgical resection, followed by temozolomide chemotherapy and radiation. This current standard therapeutic regimen has increased the median survival rate of patients with GB to 12.1–14.6 months, with a median 2-years survival rate of 26% (2). However, mortality from GB inevitably occurs, due to the recurrence or progression of the disease (3). Previous studies have indicated that recurrence of GB is unavoidable, following a median survival time of 8–9 months (3). Once a patient with GB relapses following initial surgery, radiotherapy and chemotherapy, the salvage strategies available are often limited, and the survival time is short (4,5). This is most likely due to molecular and genetic alterations within the tumor itself, or as a consequence of preceding therapies (3).
Therefore, novel therapeutic approaches to treat GB, and scientific and clinical advances in the treatment of this disease are required. The identification of molecular biomarkers for the management and monitoring of patients with cancer may improve their clinical outcome. Numerous biomarker candidates have been generated by high-throughput technologies, which are powerful and promising methods for evaluating the expression of a large number of genes and detecting alterations in genome-wide expression analyses (6). An important aspect of GB invasion is the elaboration by the tumor cells of a migration-enhancing extracellular matrix (ECM), and the secretion of proteolytic enzymes that permit cell invasion through this matrix (7). This hypothesis is consistent with previous gene expression profiling studies, which identified a subset of tumors with increased expression of ECM components and intracellular proteins associated with cell motility (8,9). The interplay between the various matrix and growth factor receptors, and the activation of signaling pathways that facilitate the invasion of tumor cells, has been recently recognized as a composite, dynamic consequence of altered cell-cell adhesion, proteolytic remodeling and synthesis of ECM, and selective expression and activation of integrins (10). Furthermore, in the past few years, several signaling pathways have been associated with reduced sensitivity of GB cells to radiation and chemotherapy (11). Numerous genes involved in these pathways have been demonstrated to be regulated by microRNAs (miRNAs), which are important regulators in cancer cell biology, and promising biomarkers or therapeutic targets in GB (11,12).
However, despite all the molecular information on GB gathered thus far, the optimal management of patients with GB remains elusive, due to the absence of validated data from clinical studies, and the great heterogeneity of this fragile subpopulation, in terms of their physical condition, co-morbidity status, tolerance to treatment and clinical prognosis (13). The invasive nature of GB growth and its fast proliferation rate are the major reasons for therapeutic failure in patients with GB (14). Therefore, a detailed characterization of the molecular mechanisms involved in the development of GB is required, in order to improve the accuracy of the prediction models, regarding the prognosis and response to therapies in patients with GB.
In order to improve the understanding of the dynamics of the genomic alterations associated with tumor relapse, and to provide novel information on the aggressive behavior of GB tumors, primary GB tumors of patients with GB displaying short or long recurrence-free survival (RFS) outcome, were molecularly characterized and compared in the present study.
In particular, 19 GB samples were characterized according to the presence of mutations in the isocitrate dehydrogenase 1 (IDH1) gene, amplification of the epidermal growth factor receptor (EGFR) gene, presence of the EGFR variant III (EGFRvIII), and methylation of the promoter region of the O6-methylguanine-DNA methyltransferase (MGMT) gene. Furthermore, the expression of 84 genes known to be important for cell-cell and cell-matrix interactions, and that of 84 miRNAs known to switch their expression pattern during nervous system-associated carcinogenesis, were profiled. In addition, a copy number variation analysis of 23 genes, whose expression has been previously reported to be frequently altered in human glioma tumors (15), was also performed.
Materials and methods
Patients and tumors
The present study included 19 newly diagnosed cases of GB, who presented to the Department of Translational Research and of New Surgical and Medical Technologies of the University Hospital of Pisa (Pisa, Italy). The patients, who were initially diagnosed by resection with GB grade IV (according to the World Health Organization classification criteria) (16), did not present any previous record of brain neoplasm. The GB tumors isolated from the patients were stored as formalin-fixed, paraffin-embedded (FFPE) specimens.
The GB cases included in the present study were selected according to their differences in progression-free survival. Of the 19 included GB cases, 10 cases, were classified as ‘short RFS’ and first displayed recurrence earlier than 6 months, and 9 cases, were classified as ‘long RFS’ and first displayed recurrence later than 14 months.
The median age of the patients was 61 years, and the gender distribution was 9/19 (47%) males and 10/19 (53%) females. The short RFS GB patients presented a median age of 63 years; gender distribution of 4/10 (40%) males and 6/10 (60%) females; and a mean RFS of 4.5 months, ranging from 2 to 6 months. The long RFS GB patients presented a median age of 58 years; gender distribution of 4/9 (44%) males and 5/9 (56%) females; and a mean RFS of 27 months, ranging between 14 and 71 months (Table I). The information regarding the size of the tumors and their localization within a certain brain region in relation to the duration of RFS is presented in Table II.
Demographic characteristics of the patient population at the time of diagnosis, and molecular characterization of the glioblastoma tumors.
Brain region localization, size (height × width × length) and area of the glioblastoma tumors in the short (n=10) and long (n=9) RFS cohorts.
The present study was approved by the Internal Review Board of the University Hospital of Pisa. All patients provided their consent for participation in the study.
DNA and RNA extraction
DNA and RNA were extracted from 2 10-µm sections of each FFPE specimen, using NucleoSpin Tissue (Machery-Nagel, Düren, Germany) and RNeasy FFPE Kit (Qiagen GmbH, Hilden, Germany) to extract the genomic DNA and total RNA from the tissues, respectively. The extracted DNA and RNA were quantified using Qubit 2.0 Fluorometer (Life Technologies, Grand Island, USA), according to the manufacturer's instructions. The DNA and RNA yields ranged from 50 to 500 ng/µl.
The samples were molecularly characterized by investigating the most studied molecular markers alterations in GB tumors, including mutations in the IDH1 gene; amplification of the EGFR gene; presence of EGFRvIII, the active mutant variant form of the EGFR gene; and methylation of the MGMT promoter:
For the detection of IDH1 mutations, Primer3 software (http://primer3.ut.ee/) was used to design the following primers, in order to amplify the exon 4 of the IDH1 gene: Forward, 5′-AGC TCT ATA TGC CAT CAC TGC-3; and reverse, 5′-TTC ATA CCT TGC TTA ATG GGT GT-3). Polymerase chain reaction (PCR) amplification was performed with AmpliTaq Gold DNA Polymerase with Buffer II and MgCl2 (Life Technologies), according to the manufacturer's protocol. The PCR products were loaded onto a 2% agarose gel, subjected to DNA gel electrophoresis, stained with ethidium bromide and verified by sequencing. For this purpose, the PCR bands were excised from the gel and purified using QIAquick Gel Extraction Kit (Qiagen GmbH), prior to being subjected to direct sequencing in both directions with the aforementioned primers, using the BigDye Terminator v3.1 Cycle Sequencing Kit (Applied Biosystems Life Technologies, Foster City, USA). The sequencing reaction was conducted on an ABI PRISM 3130xl Genetic Analyzer (Applied Biosystems Life Technologies), according to the manufacturers protocol.
Copy number variation
Analyses of the number of copies of the EGFR gene and the presence of the EGFRvIII variant were performed with SALSA MLPA P105 Glioma-2 probemix (MRC-Holland, Amsterdam, The Netherlands), according to the manufacturers protocol. Fragment analysis was performed on an ABI PRISM 3130xl Genetic Analyzer (Applied Biosystems Life Technologies).
MGMT promoter methylation
Analysis of the methylation status of the MGMT promoter was performed with the MGMT plus kit (Diatech Pharmacogenetics, Jesi, Italy), according to the manufacturers instructions. Following bisulfite conversion, the DNA was amplified with bisulfite-specific primers, using TaKaRa Ex Taq Hot Start DNA Polymerase (Clontech Laboratories, Inc., Mountainview, USA), according to the manufacturer's protocol, and analyzed on a PyroMark Q96 ID (Qiagen GmbH), using Pyro Q-CpG software (Qiagen GmbH).
PCR commercial arraysECM and adhesion molecules
100 ng of isolated RNA was reverse transcribed to cDNA, and pre-amplified using the RT2 PreAMP cDNA Synthesis Kit (Qiagen GmbH). The cDNA was aliquoted with Mastermix (Qiagen GmbH) and added to the wells of an ECM microarray plate (catalogue no. PAHS-0013; SABiosciences, Frederick, USA), according to the manufacturers instructions.Brain cancer miRNA
250 ng of isolated RNA was reverse transcribed to cDNA using miScript II RT Kit (Qiagen GmbH). The cDNA generated was subsequently used as a template for qPCR, using the miScript SYBR Green PCR Kit (SABiosciences, Qiagen GmbH) and the Human Brain Cancer miRNA PCR Array (catalogue no. MIHS-108Z; SABiosciences), according to the manufacturers instructions.Glioma copy number
800 ng of extracted DNA was aliquoted with qBiomarker SYBR Mastermix (Qiagen GmbH, Hilden, Germany) and added to the wells of a Human Glioma Copy Number PCR Array Plate (catalogue no. VAHS-0053Z, SABiosciences), according to the manufacturers instructions.
All the above PCR array analyses were performed calculating the ΔCT value for each gene profiled in the plate, using the formula ΔCT = CT (gene of interest) - average CT (reference gene). The fold difference value, used for statistical analysis, was equated to 2–ΔΔCT.Statistical analyses
The χ2 test was used to establish whether there was any association between the 2 categorical variables, short/long RFS group, and any alterations in molecular markers, including mutations in the IDH1 gene, EGFR amplification, presence of the mutant form EGFRvIII and methylation of the MGMT promoter. Differences in the gene expression levels between the 2 groups were analyzed by one-way analysis of variance. Discriminant analysis was conducted to predict the probability of a particular sample to belong to the short or the long RFS group, based on the expression levels of the profiled genes exhibited by the sample.
Principal component analysis (PCA) was used to convert a set of observations of possibly correlated variables (gene expression levels) into a set of values of linearly uncorrelated variables termed PCs. PCs are guaranteed to be independent if the data set is jointly normally distributed (17). A tridimensional PCA scatter-plot was generated with R software (R Foundation for Statistical Computing, Vienna, Austria; https://www.r-project.org/).
Leave-one-out cross-validation (LOOCV) was used as resampling method to estimate the prediction error of the model. Its indicator, the error rate, defined as 1 - accuracy rate, states the probability of misclassification of a classifier. In LOOCV, 1 instance is eliminated from the dataset, and the classifier is created on the remaining instances, in order to assess the correctness of the prediction of the eliminated instance. This process is then repeated for all the instances.
Logistic regression analysis was performed using Stata SE12.0 software (StataCorp LP, College Station, USA), considering short/long RFS group as the dependent variable, and the area of the tumor and the patient's age and gender as the independent variables.
Demographic characteristics of the patients and molecular characterization of the tumors
Patients with GB were classified in 2 groups, based on their duration of RFS: Patients were defined as short RFS, if RFS = <6 months, and long RFS, if RFS = >14 months.
The comparison between the patient's characteristics, including age, gender and localization of the tumor, and the 2 lengths of RFS, did not identify any significant associations (data not shown). The calculated average tumor area was smaller in the long RFS group compared with the short RFS group (38 vs. 58 cm2, respectively; Table II), but this difference was not observed to be statistically significant (data not shown).
Furthermore, using logistic regression analysis, the role of potential interference factors, including the patient's age, size of the tumor and extent of surgery, were excluded from the short/long RFS classification (data not shown).
Codon 132 in the IDH1 gene was confirmed to be wild-type in all the 19 GB samples analyzed, while 9/19 (44%) GB samples presented with EGFR amplification, and 3/9 (33%) samples carried the EGFRvIII variant (Table I). Among the short RFS samples, 3/10 (30%) cases presented with EGFR amplification, and 1/3 (33%) carried the EGFRvIII variant; whereas in the long RSF samples, 6/9 (67%) cases presented with EGFR amplification, and 2/6 (33%) carried the EGFRvIII variant (Table I). Overall, 8/19 (42%) GB samples exhibited MGMT promoter methylation, 5 of which belonged to the 10 short RFS group (50%), whereas only 3 (33%) of those samples with MGMT promoter methylation belonged to the 9 long RSF group (Table I).
Molecular markers and RFS association analysis
The comparison between the frequencies of alterations in the molecular marker that were identified in the patients with GB, and the differences in RFS displayed by these patients, did not reveal any significant differences. Furthermore, no significant correlations were observed among all the molecular alterations analyzed.
PCR array data and RFS association analysis
The differences between the short and long RFS groups, regarding the expression levels and copy number of 191 genes, were analyzed using 3 PCR arrays. The results revealed significant differences in the expression levels of miRNA-150-5p between the 2 groups (P=0.040; Table III).
Differences in the expression levels of the 5 selected genes among the patients with glioblastoma, belonging to the short (n=10) or long (n=9) RFS groups.
In order to perform a discriminant analysis and predict the probability of a GB case to belong to the short or long RFS group, based on their gene expression levels, 5 genes with the lowest P-value (P<0.095), including miRNA-150-5p and −328-3p, selectin E (SELE) and L (SELL), and a disintegrin and metalloproteinase with a thrombospondin type 1 motif, member 13 (ADAMTS13), were selected for the analysis (Table III).
The combination of the expression levels of the above 5 genes was able to correctly classify the GB samples into their corresponding short or long RFS group, with a precision of 90 and 100%, respectively (Table IV). Overall, 94.74% (18/19) GB cases were correctly classified with this method (P=0.008; Table IV). The probability of each sample to belong to one or the other group is also indicated in Table IV. Among the 18 samples correctly classified, 12 (66%) were assigned a probability >90%, including 7 cases (34%) whose probability was >99% (Table IV). Using the expression levels of the aforementioned 5 selected genes (miRNA150-5p, miRNA328-3p, SELE, SELL and ADAMTS13) as data variables, a PCA was performed, which clearly indicated a well-defined separation of the 2 RFS populations in a tridimensional space (Fig. 1).
Scatter-plot of principal component analysis. Visualization in a tridimensional space of the discriminative power of the expression levels of the 5 selected genes (microRNA-150-5p and −328-3p, selectin E and L, and a disintegrin and metalloproteinase with thrombospondin type 1 motif, member 13), considered as variables, in correctly classifying the 19 GB samples included in the present study into their corresponding short or long RFS group. The black and red dots represent the 10 short and 9 long RFS GB samples, respectively. GB, glioblastoma; RFS, recurrence-free survival; Comp., component.
Cross-validation of the 5-genes combination
The prediction error of the 5-genes combination approach was estimated via cross-validation, using a resampling method. In this type of methodology, the prediction error is calculated for each measured tree from a presumed distribution of errors, and a dataset is subsequently generated as the sum of a prediction and a random prediction error.
The percentage of cases correctly classified following this method was 90% (9/10) in the short RFS samples, and 78% (7/9) in the long RFS samples, with a total error rate of 0.1579 (Table V).
In regards to GB, it is of importance to identify key molecular markers, and to perform simple and prompt molecular diagnostic tests that are able to predict the survival time and therapeutic response of patients. In the present study, a wide molecular characterization, including analysis of miRNA, RNA and DNA, was performed on 19 GB cases with different RFS outcome, in an attempt to better understand the dynamics of the genomic alterations associated with relapse in patients with GB. A long-term recurrence GB group, with an RFS ranging from 14 to 71 months (mean RFS = 27 months), was included in the study. This group represents rare cases of GB, compared with the GB RFS reported in the literature (3).
Based on the clinical history of the patients and the molecular features of their tumors, all the cases were classified as primary GBs. The patients did not present any previous records of brain neoplasm, and their median age at the time of diagnosis was 60 years, which is in agreement with previous studies that reported the mean age of patients clinically diagnosed with primary GB to be 62 years, and 45 years in the case of patients clinically diagnosed with secondary GB (18). In addition, the patients of the present study exhibited a high frequency of overexpression of EGFR (44%) and absence of IDH1 mutations, which are typical features of primary GBs (18).
The expression levels of 84 miRNAs whose expression pattern is altered during nervous system-associated carcinogenesis, and 84 genes whose products are involved in cell-cell and cell-matrix interaction processes, were evaluated by qPCR array. Additionally, a copy number PCR array analysis was performed on 23 genes whose expression pattern has been reported to be frequently altered in human glioma tumors (15). The results revealed significant differences in the expression levels of miRNA150-5p between the short and long RFS groups.
The 5 genes with the highest differences in gene expression levels that were identified in the study, including miRNA150-5p, miRNA328-3p, SELE, SELL and ADAMTS13, were selected for subsequent analysis, despite their P-values were not significant (P<0.095). The resulting 5-genes combination approach was able to correctly classify 94.74% of the 19 GB samples into their corresponding short or long RFS groups. To the best of our knowledge, the expression pattern of the above 5 genes has never been associated with RFS in GB thus far.
Next, an LOOCV was performed to evaluate the performance of this 5-genes combination approach, and a low error rate (0.1579) was obtained for this method, indicating that 84% of the 19 GB cases were correctly classified following this approach.
Of the aforementioned genes, SELE, SELL and ADAMTS13 were observed to be overexpressed in the short RFS GB samples, which is the group with the worst prognosis. ADAMTS13 has been previously associated with the degradation and remodeling of the ECM, while SELE and SELL are cell adhesion molecules (CAMs) that mediate cell attachment, migration and signaling to and from the ECM (15). Alterations of CAMs have been reported to be common in cancer, where the disruption of cell-cell or cell-ECM adhesion significantly contributes to uncontrolled cell proliferation and progressive distortion of the architecture of normal tissue (19). Furthermore, alterations in CAMs have been involved in tumor dissemination, since the loss of cell-cell adhesion enables the malignant cells to detach and subsequently migrate from the primary mass (19).
By contrast, miRNA-150-5p and 328-3p were observed to be overexpressed in the long RFS GB group, which is the group with the best prognosis. Previous studies have demonstrated that the increased expression of miR-150 inhibited breast cancer cell migration and invasion in vitro, and decreased cell proliferation and induced apoptosis in natural killer and T-cells (20,21). A previous study revealed that patients with GB whose expression levels of miR-328 were lower than the median value, exhibited poor survival, since miR-328 is associated with the cell cycle progression (22). Therefore, the ectopic expression of miR-328 in GB cells may suppress cell proliferation (22).
In the present study, the GB samples were divided and compared, according to their short or long RFS time, despite not being 2 distinct clinical or pathological entities. All the patients with GB displayed the same histological and pathological features, including high malignancy, invasive behavior and poor prognosis, regardless of their short or long RFS. Therefore, it is not easy to differentiate between short and long RFS in patients with GB. However, it is of importance to understand the mechanisms leading to aggressive GB tumors with high spreading capacities, which are not susceptible to standard treatments, including surgery, radiotherapy and chemotherapy, whereas other GB tumors are more manageable, and enable patients to live >1 year. Although tumor invasion and metastasis are known to be the main causes of mortality in patients with solid cancer, including GB, the characterization of the molecular and cellular mechanisms behind these processes is limited.
In conclusion, the present study has demonstrated that alterations in the expression pattern of certain miRNAs, mRNAs and genomic DNA constitute a particular molecular hallmark, which may increase or reduce the aggressive behavior of GB tumors, thus influencing the survival time, response to therapy and tendency to experience a tumor relapse of patients with GB. In particular, the 5-genes combination approach performed in the present study was able to correctly predict the short or long RFS outcome in the majority of patients with GB. The high sensitivity and precision of this approach, as confirmed by LOOCV, provide a strong foundation for further validation of the association between alterations in the expression of the aforementioned 5 genes and the RFS of patients with GB in a larger population.
Stupp R, Mason WP, van den Bent MJ, Weller M, Fisher B, Taphoorn MJ, Belanger K, Brandes AA, Marosi C, Bogdahn U, et al: European Organisation for Research and Treatment of Cancer Brain Tumor and Radiotherapy Groups; and National Cancer Institute of Canada Clinical Trials Group: Radiotherapy plus concomitant and adjuvant temozolomide for glioblastoma. N Engl J Med. 352:987–996. 2005. View Article : Google Scholar : PubMed/NCBI
Grossman SA, Ye X, Piantadosi S, Desideri S, Nabors LB, Rosenfeld M and Fisher J: NABTT CNS Consortium: Survival of patients with newly diagnosed glioblastoma treated with radiation and temozolomide on research studies in the United States. Clin Cancer Res. 16:2443–2449. 2010. View Article : Google Scholar : PubMed/NCBI
Hou LC, Veeravagu A, Hsu AR and Tse VC: Recurrent glioblastoma multiforme: A review of natural history and management options. Neurosurg Focus. 20:E52006.PubMed/NCBI
Wong ET, Hess KR, Gleason MJ, Jaeckle KA, Kyritsis AP, Prados MD, Levin VA and Yung WK: Outcomes and prognostic factors in recurrent glioma patients enrolled onto phase II clinical trials. J Clin Oncol. 17:2572–2578. 1999.PubMed/NCBI
Lamborn KR, Yung WK, Chang SM, Wen PY, Cloughesy TF, DeAngelis LM, Robins HI, Lieberman FS, Fine HA, Fink KL, et al: North American Brain Tumor Consortium: Progression-free survival: An important end point in evaluating therapy for recurrent high-grade gliomas. Neuro Oncol. 10:162–170. 2008. View Article : Google Scholar : PubMed/NCBI
Zhang W, Zhang J, Yan W, You G, Bao Z, Li S, Kang C, Jiang C, You Y, Zhang Y, et al: Whole-genome microRNA expression profiling identifies a 5-microRNA signature as a prognostic biomarker in Chinese patients with primary glioblastoma multiforme. Cancer. 119:814–824. 2013. View Article : Google Scholar : PubMed/NCBI
Knott JC, Mahesparan R, Garcia-Cabrera I, Bølge Tysnes B, Edvardsen K, Ness GO, Mørk S, Lund-Johansen M and Bjerkvig R: Stimulation of extracellular matrix components in the normal brain by invading glioma cells. Int J Cancer. 75:864–872. 1998. View Article : Google Scholar : PubMed/NCBI
Freije WA, Castro-Vargas FE, Fang Z, Horvath S, Cloughesy T, Liau LM, Mischel PS and Nelson SF: Gene expression profiling of gliomas strongly predicts survival. Cancer Res. 64:6503–6510. 2004. View Article : Google Scholar : PubMed/NCBI
Phillips HS, Kharbanda S, Chen R, Forrest WF, Soriano RH, Wu TD, Misra A, Nigro JM, Colman H, Soroceanu L, et al: Molecular subclasses of high-grade glioma predict prognosis, delineate a pattern of disease progression, and resemble stages in neurogenesis. Cancer Cell. 9:157–173. 2006. View Article : Google Scholar : PubMed/NCBI
Yin AA, Zhang LH, Cheng JX, Dong Y, Liu BL, Han N and Zhang X: The predictive but not prognostic value of MGMT promoter methylation status in elderly glioblastoma patients: A meta-analysis. PLoS One. 9:e851022014. View Article : Google Scholar : PubMed/NCBI
Colen RR, Vangel M, Wang J, Gutman DA, Hwang SN, Wintermark M, Jain R, Jilwan-Nicolas M, Chen JY, Raghavan P, et al: TCGA Glioma Phenotype Research Group: Imaging genomic mapping of an invasive MRI phenotype predicts patient outcome and metabolic dysfunction: A TCGA glioma phenotype research group project. BMC Med Genomics. 7:302014. View Article : Google Scholar : PubMed/NCBI
Qiagen: Glioma Copy Number PCR Array. simplehttp://www.sabiosciences.com/copynumber_product/HTML/VAHS-0053Z.html
Louis DN, Ohgaki H, Wiestler OD, Cavenee WB, Burger PC, Jouvet A, Scheithauer BW and Kleihues P: The 2007 WHO classification of tumours of the central nervous system. Acta Neuropatho. 114:97–109. 2007. View Article : Google Scholar
Huang S, Chen Y, Wu W, Ouyang N, Chen J, Li H, Liu X, Su F, Lin L and Yao Y: miR-150 promotes human breast cancer growth and malignant behavior by targeting the pro-apoptotic purinergic P2X7 receptor. PLoS One. 8:e807072013. View Article : Google Scholar : PubMed/NCBI
Caramuta S, Lee L, Ozata DM, Akçakaya P, Georgii-Hemming P, Xie H, Amini RM, Lawrie CH, Enblad G, Larsson C, et al: Role of microRNAs and microRNA machinery in the pathogenesis of diffuse large B-cell lymphoma. Blood Cancer J. 3:e1522013. View Article : Google Scholar : PubMed/NCBI
Wu Z, Sun L, Wang H, Yao J, Jiang C, Xu W and Yang Z: MiR-328 expression is decreased in high-grade gliomas and is associated with worse survival in primary glioblastoma. PLoS One. 7:e472702012. View Article : Google Scholar : PubMed/NCBI