MiRNA expression patterns predict survival in glioblastoma

Glioblastoma multiforme (GBM) is the most common and aggressive primary brain tumor [1]. Malignant gliomas account for approximately 70% of new cases of malignant primary brain tumors diagnosed in adults. Median age at diagnosis of primary gliomas is 64 years and malignant gliomas are more common in men than in women [2].

Currently the treatment of GBM is based on a multidisciplinary approach including surgery and adjuvant radiochemotherapy followed by maintenance chemotherapy. Concomitant and adjuvant administration of temozolomide improved 2-year survival of patients with newly diagnosed malignant glioma (mainly GBM) from 11% to 27%, 3-year survival from 4% to 16% and 5-year survival from 2% to 10% [3, 4].

Despite all developments for primary and recurrent glioblastoma [3, 5], there is still extreme room for further improvement since glioblastoma has a dismal prognosis for most of the patients with a high rate of local recurrences [6]. At present, several strategies may lead to an optimization: Firstly, better imaging tools as well as improved image-guidance are available [7‐10] and dose escalation while sparing normal tissue has been achieved by new technical approaches such as intensity-modulated radiotherapy [11], volumetric single arc technique [12] or older techniques such as fractionated stereotactic boost/radiosurgery [13]. Secondly, in addition to an improved application of radiotherapy, the combination of radiation with targeted drugs may turn out to increase the therapeutic ratio. In this regard different targeted molecules are currently undergoing pre-clinical and clinical testing [14‐20].

Closely associated with the research fields mentioned above, it is of crucial importance to gain insight into the underlying biological reasons for different patient outcomes. It is well known that the prognosis of patients with glioblastoma differs considerably on an individual scale [21].

At present only very few prognostic factors such as a higher age or a bad ECOG score have been defined. Recently, certain molecular pathways and associated biomarkers have been established as prognostic and predictive markers. In this regard, Hegi et al. identified the O6-methylguanine DNA-methyltransferase (MGMT) promoter methylation status to be a potent prognostic factor for GBM patients with a potential predictive value for the efficacy of temozolomide-based radiochemotherapy [22, 23]. Additionally, IDH1 and IDH2 status have recently been introduced in GBM but their predictive role has not yet been defined though their prognostic power is well-known [24, 25]. However, neither MGMT methylation nor other markers are precise enough to enable individual assessments.

The use of microRNAs (miRNAs) as tumor biomarkers has gained growing interest in the last few years. Accumulating evidence indicates that miRNA expression can be used as a prognostic and/or diagnostic marker for human cancers [26]. The miRNAs consist of 18-25 nucleotides and are a class of endogenous ribo-regulators that modulate gene expression via the RNA interference (RNAi) pathway [27]. The discovery of miRNAs dates back to 1993 when Lee et al. described a small RNA, lineage-deficient-4 (lin-4), with antisense complementarity to lin-14 involved in the regulation of developmental timing in Caenorhabditis elegans[28]. MicroRNA deregulation is implicated in processes such as cell proliferation, cell cycle regulation, apoptosis, invasion, glioma stem cell behavior and angiogenesis [29, 30].

More than 1000 microRNAs (miRNAs) are present in the human genome. Expression is largely tissue and cell type specific, with some miRNAs considered to be housekeeping molecules. Each miRNA is predicted to target and possibly regulate multiple mRNA species. Diverse high-throughput screenings of various systems have as yet identified only a limited number of functional miRNAs [31].

Their role in regulating a great variety of targets and, as a consequence, multiple pathways, makes their use in diagnostics potentially a powerful tool to be exploited for risk assessment and prognosis and for the design of innovative therapeutic strategies [32].

With all the inherent limitations of a retrospective analysis in mind, we have determined the prognostic value of miRNA expression patterns for overall survival in 35 primary glioblastoma patients who were treated at our institution with radio(chemo)therapy following surgery. Follow-up data and MGMT methylation status were used to test whether miRNA profiles may be MGMT-independent prognostic factors for overall survival.

All glioblastoma patients treated with surgery and postoperative radio(chemo)therapy treated between 7/2005 and 8/2008 at our institution were identified using the departmental database. Radiochemotherapy with temozolomide was applied according to the EORTC/NCIC regimen [3, 4].

Our study was approved by the local ethics committee (No. 442-09), histopathologic diagnosis was confirmed by central pathology review in all cases.

Determination of MGMT promoter methylation was performed using both methylation-specific PCR and sequencing analysis as being published before [33, 34].

For miRNA analysis the biochip "Geniom^® Biochip MPEA homo sapiens" from febit was used (Febit holding GmbH, Im Neuenheimer Feld, Heidelberg, Germany). The probes are designed as the reverse complements of all major mature miRNAs and are based on sequences as published in the current Sanger miRBase release (version 15.0 April 2010, see http://​microrna.​sanger.​ac.​uk/​sequences/​index.​shtml) for Homo sapiens. Additional nucleotides are bound on the 5'end of each capture oligonucleotide that serve as template for the enzymatic extension in the labelling procedure. The probes are synthesized with intra-array replicates to increase the statistical confidence and to compensate for potential positional effects. The intensities of blank probes consisting only of one single "T" nucleotide are used for background corrections.

In order to define new marker constellations for a more precise separation of different prognostic groups in patients with GBM we performed a miRNA array analysis including merely all currently known miRNAs.

In our analysis, two complementarily defined miRNA pattern predicted early death versus long-term survival (split at 450 days) in a significant way (p = 0.01) and this prediction was independent of the MGMT status. Thus, importantly the miRNA pattern was correlated with outcome independently of the MGMT status within the univariate analysis. This could not be proven within the multivariate analysis due to statistical limitations as the case number and the number of events was too small.

Our findings clearly indicate that complex alterations of the regulatory network involved in tumor gene expression are at least as important as a single disturbance of a single DNA repair enzyme.

At present, the role of individual miRNAs in GBM is poorly understood. MiRNAs are small noncoding regulatory RNAs that reduce stability and/or translation of fully or partially sequence-complementary target mRNAs [36]. In this sense, they are important post-transcriptional gene regulators and play an important role in response to cellular stress [37] as well as pathogenesis of cancer development and progression [38] with miR-17 and miR-184 being promising candidates for tumor progression [39].

It has been shown recently that GBMs display a distinct miRNA expression signature and a number of recent studies have linked these miRNA alterations to key hallmarks of GBM including proliferation, survival, invasion, angiogenesis and stem cell-like behavior [40, 41]. Moreover, resistance to temozolomide might be associated with miRNA deregulation [42]. In this regard, Ciafre et al. studied the global expression of 245 microRNAs in GBM using a microarray technique [43] in comparison to normal brain tissue. This approach enabled the identification of miRNAs whose expression is significantly altered in tumors compared with peripheral brain areas from the same patient, including miR-221, strongly up-regulated in GBM, and a set of brain-enriched miRNAs, miR-128, miR-181a, miR-181b, and miR-181c, which were down-regulated in glioblastoma [36].

In contrast to the available data on miRNA in GBM, our approach differed significantly. Based on the hypothesis that an equally complex network of different miRNAs might regulate complex metabolic interactions regulating processes such as the therapy response in tumor cells we speculated that a fingerprinting approach using biological extremes would reveal informative prognostic data. Indeed our approach revealed a miRNA pattern being significantly associated with the outcome of the pre-defined biological extremes.

However, at present one has to bear in mind that the approach chosen is purely retrospective and does not contain a control group of uninvolved brain parenchyma. It is only useful to generate a hypothesis but not to prove a predictive character of the observed miRNA pattern. Thus it is mandatory to validate the miRNA pattern in a prospective trial.

Besides the validation of the given pattern, it is of major interest to understand to role of individual miRNAs for the biological response of GBM. Among the panel of miRNAs, there are some with relatively unknown cell cycle function, such as miR-3163, miR-1305, miR-1260. Others have already been examined on a functional level such as miR-539 which is among the factors sensing biotin and regulating holocarboxylase synthetase. This enzyme plays an essential role in catalyzing the biotinylation of carboxylases and histones. Biotinylated carboxylases are needed for the metabolism of glucose, lipids and leucine; biotinylation of histones plays important roles in gene regulation and genome stability [44]. The most prominently deregulated miRNA is let-7a: Lethal-7a was recently found to be associated with several cancers, such as lung and colon cancers. It was also proposed that let-7a may be a tumor suppressor in laryngeal cancer by inhibiting cell growth, inducing cell apoptosis and down-regulating oncogene expression; in Hep-2 cells, let-7a induced apoptosis and downregulated RAS and c-MYC protein expression without affecting the mRNA levels [45]. But none of these miRNAs has been linked to glioblastoma up to now.

This study is - to our best knowledge - the first clinical miRNA screening trial in GBM patients resulting in the definition of a distinct miRNA pattern. This pattern may serve as potential new prognostic and/or predictive marker set allowing for patient stratification independent of the MGMT status. Although the trial was based on a small sample size and is limited by the retrospective character of the study, the data urgently mandate a prospective validation and additional research in order to define to biological role of miRNA alteration for the pathogenesis of glioblastoma.