MicroRNAs (miRNAs) are small, non-coding, endogenous RNA molecules approximately 14–22 nucleotides in length, which function in gene silencing by post-transcriptional regulation of protein expression. miRNAs hybridize to the 3′-untranslated regions (UTRs) of target mRNAs, and either inhibit translation or promote degradation of transcripts. Since miRNAs are not a perfect complement of their target mRNA, they have the ability to target tens to thousands of genes whose protein products function in various signaling pathways. This combined effect on multiple target genes may dysregulate multiple pathways and networks in carcinogenesis, which brings promise to miRNAs as effective therapeutic targets.
Based on their expression levels and major target oncogenes (or tumor suppressor genes), miRNAs may act as oncogenic miRNAs (onco-miRs) (or tumor suppressors miRNAs) in cancer development and progression [1
]. Onco-miRs are generally upregulated in cancer, and contribute to tumorigenesis by silencing tumor suppressor genes, while tumor suppressor miRNAs are generally down-regulated in cancers, and function by silencing target oncogenes involved in tumorigenesis [1
]. Given their critical roles in carcinogenesis, miRNAs themselves are also subject to regulatory controls at both post-transcriptional and epigenetic levels. The mutations in epigenetic modifiers, such as isocitrate dehydrogenase (IDH1/2), telomerase reverse transcriptase (TERT), and alpha-thalassemia/mental retardation syndrome X-linked (ATRX), which lead to global changes in the epigenome, are common drivers of gliomagenesis [2
]. The roles that these mutations play in miRNA dysregulation and glioma development are poorly understood.
In this review, we discuss dysregulated miRNAs and their involvement in glioma development and progression. More specifically, we have performed a comprehensive review of miRNAs targeting a number of hallmarks of malignant gliomas, which include angiogenesis, invasion, and altered metabolism. To this end, we have identified and discussed the most frequently cited miRNAs involved in angiogenesis, invasion, and glioma cell metabolism. Therapeutic implications for these miRNAs and the challenges for their clinical translation are also discussed.
2. miRNA Biogenesis
The generation of mature miRNAs is a multi-step process. Biogenesis starts with the initial transcription of miRNA genes by RNA polymerase II or, alternatively, by RNA polymerase III as co-transcripts with neighboring repetitive elements [4
]. The initial miRNA transcript is several hundred nucleotides long, capped, polyadenylated, and contains an imperfectly double-stranded region within a hairpin loop. The 5′ and 3′ ends of miRNAs’ primary transcripts are processed by the ribonuclease (RNase) III Drosha-DGCR8 nuclear complex, in an endoribonucleolytic cleavage, into the hairpin structure precursor miRNA of less than a hundred nucleotides. The resulting pre-miRNA is then transported through the nuclear pore into the cytoplasm by exportin-5, in complex with Ran and GTP. Once in the cytoplasm, the pre-miRNA is further cleaved by the RNase enzyme DICER1, an Argonaute protein, and either TARBP2 or PRKRA into double-stranded miRNAs, with protruding single-stranded 3′ ends of 2–3 nucleotides. The two strands are separated by helicases, and the mature strand incorporates into the RNA-induced silencing complex (RISC) [5
]. Here, the passenger strand is degraded and the guide strand is targeted to specific mRNAs (Figure 1
). The selection of the strand that can enter the RISC and become the predominant product depends on several factors, including thermodynamic stability, base pairing, and the position of the stem-loop [8
Typically, mature miRNAs regulate gene expression through sequence-specific binding to the 3′-UTR of an mRNA, though several lines of evidence indicate that miRNAs can also bind to 5′-UTR or other regions of target mRNAs. The seed sequence, located 2–7 nucleotides from the 5′ end of the miRNA, most commonly determines the specific mRNA targets that it binds to. The miRNA-mRNA interaction usually causes translational repression and/or mRNA cleavage, and may reduce the final protein level. Alternatively, miRNAs are also known to increase the translation of a target mRNA by recruiting protein complexes to the AU-rich elements of the mRNA, or indirectly by de-repressing the mRNA translation via interaction with proteins that block the target gene translation [5
While the majority of miRNAs remain intracellular, there is a small population of miRNAs termed cell-free miRNAs or circulating miRNAs that are ejected into the circulation. Alternatively, miRNAs can also be packaged into multivesicular bodies (MVBs) and released into the extracellular environment as exosomes. Following their release, the circulating miRNAs are taken up by the recipient cells in which they execute specific functions, or they may also act as hormones, triggering a receptor-mediated response in a different cell or tissue [10
]. miRNA biogenesis is therefore regulated at multiple levels by multiple mechanisms and, not surprisingly, each step of miRNA generation and function may be targeted for therapeutics [12
Gliomas are the most common (~80%) primary tumors of the human central nervous system [20
]. Despite recent advances in surgery, chemotherapy, and radiation, prognosis remains poor for many of these patients. Gliomas have traditionally been diagnosed and treated based on their histopathological grade (I through IV) defined by the World Health Organization (WHO) [21
]. This grading system was based on morphologic features of tumor tissues, including atypia, mitosis, endothelial proliferation, and necrosis. However, histopathological grading often produced inconsistent clinical outcomes owing to inter-observer variability [22
]. To overcome this limitation, the WHO recently reclassified malignant gliomas by integrating molecular biomarkers, including IDH1/2, ATRX, TERT, TP53, CI, and FUBP1 genes, and co-deletion of 1p and 19q chromosome arms [23
Lower-grade gliomas (LGGs, grades II and III) and glioblastomas (grade IV) have been redefined by the presence or absence of the IDH1/2 mutation [23
]. Notably, patients with gliomas harboring IDH1/2 mutations have an overall survival (OS) benefit over patients without an IDH1/2 mutation [28
]. The underlying biology accounting for these differences in clinical response remains elusive. Mutations in IDH1 are shown to induce the accumulation of methylated DNA via inhibition of DNA demethylation which, in turn, causes global promoter methylation and gene silencing [3
]. The specific changes in gene expression that result from these epigenetic changes are uncertain; however, they likely include changes in miRNA expression. Interestingly, Cheng et al. found that IDH1/2 mutation status in lower grade gliomas has more impact on miRNA expression profiles than other genomic changes. Moreover, miRNA profiles play a more significant prognostic role in IDH mutant tumors than IDH WT tumors, as evident from the identification of a four-miRNA risk classifier (miR-10b, miR-130b, miR-1304, and miR-302b) in IDH mutant patients [30
6. Prognostic and Predictive miRNA Biomarkers
Given the previously discussed roles of miRNAs in glioma development and progression, miRNAs have attracted a great deal of attention as potential biomarkers that may facilitate management decisions for glioma patients. The small size and stability of miRNA in clinical specimens, as well as the efficacy and affordability of miRNA assays, brings promise for the clinical implementation of miRNAs [97
]. After the recent addition of prognostic molecular alterations to the WHO classification of gliomas, significant emphasis has been put on identifying miRNA signatures that may help further refine the prognostic classification, as well as define treatment groups.
One widely reported molecular classification system involves glioma subtypes based on an unsupervised clustering of genome-wide mRNA expression, including proneural, neural, classical, and mesenchymal subtypes [98
]. However, only the proneural and mesenchymal subtypes have consistently been confirmed in multiple gene expression profiling studies [99
]. While gliomas of the proneural subtype tend to more closely resemble normal neurons, and also have better prognoses, gliomas of the mesenchymal subtype have increased invasive and angiogenic potential, in addition to worse prognoses [99
]. Interestingly, Ma et al. identified miR-128a, miR-504, miR-124a, and miR-184 to be significantly inversely correlated with mesenchymal marker expression, therefore suggesting that these miRNAs likely suppress mesenchymal signaling in gliomas [100
]. Moreover, functional studies inhibiting miR-128 and miR-504 resulted in increased levels of VIM and YKL-40 expression, both mesenchymal genes involved in invasion and angiogenesis. As previously mentioned, other miRNA including miR-21, miR-10b, and miR-221 have also been shown to play a role in mesenchymal glioma migration and invasion [35
]. Interestingly, Papagiannakopoulos et al. showed that, in proneural gliomas, miR-128 acts as a tumor suppressor by enhancing neural differentiation and repressing growth, secondary to targeting oncogenic receptor tyrosine kinases [101
]. Marziali et al. identified mir-23a, miR-27a, and miR-9-3p as a miRNA signature able to discriminate proneural vs. mesenchymal gliomas in both glioma stem cell cultures and The Cancer Genome Atlas (TCGA) glioblastoma cohort [102
]. Furthermore, Li et al. [103
] identified prognostic miRNA signatures corresponding to each molecular subtype, including proneural, neural, classical, and mesenchymal; however, only the mesenchymal signature has been validated in an independent cohort.
The utilization of miRNA signatures as independent diagnostic, prognostic, and predictive biomarkers has been extensively reviewed in glioblastoma patients [104
]; however, only two prognostic studies in lower grade gliomas have been published to date [30
]. While individual prognostic miRNAs have been confirmed in many studies, the most robust prognostic models likely consist of miRNA signatures. However, there have been inconsistencies among prognostic miRNA signatures reported in the literature, which may be attributed to small sample sizes, lack of long-term follow up data, utilization of diverse platforms, and normalization techniques. This is particularly true for studies including lower grade gliomas, as these patients are less common and are often lost to follow-up, due to longer survival times. Furthermore, many analyses include patient populations of heterogeneous grade, histology, and treatment modalities.
A recent study reported that glioblastomas (n
= 35) could be divided into two prognostic subgroups (early death <450 days vs. long-term survival >450 days), based on expression profiles of thirty miRNAs [106
]. However, due to the small sample size, this miRNA signature will need to be validated in a larger patient cohort. Additionally, Srinivasan et al. identified a ten miRNA profile that accurately predicts survival among glioblastomas (n
= 22) from the TCGA database [107
]. A more recent analysis using a larger glioblastoma patient population (n
= 563) from the TCGA cohort identified three miRNAs (miR-222, miR-302d, and miR-646) that independently predict survival among these patients [108
]. By also using the TCGA cohort of glioblastoma patients, Hayes et al. generated a risk score based on expression levels of nine miRNAs found to be significantly associated with survival [109
]. Intriguingly, miR-222 emerged as a common player in all three studies; however, despite the use of the same TCGA dataset, it was not found to be associated with disease free survival in another study [110
]. These inconsistencies across studies emphasize the need for not only utilizing large patient cohorts, but also for standardization and validation of data analyses among studies.
Similar limitations have prevented the identification of miRNA profiles to help predict responses to treatment among glioma patients. Many previously published clinical studies lack appropriate treatment and control arms to determine true predictive markers of treatment response, whereas other studies have patient cohorts too small to allow for the power necessary for interaction tests.
Single predictive miRNAs have been reported; however, to our knowledge there have only been a few reported miRNA signatures that predict treatment response among gliomas. Hayes et al. identified an eight miRNA signature (miR-124a, miR-202, miR-7, miR-222, miR-363, miR-630, miR-663, miR-204) that predicts overall survival only in those glioma patients treated with bevacizumab [104
]. Interestingly, increased expression of miR-7, an inhibitor of angiogenesis discussed earlier in this review, was shown to be associated with a poor response to bevacizumab, suggesting that tumors with less angiogenesis will have a worse response to this VEGF-targeted therapy. Additional studies have suggested that individual miRNAs and miRNA signatures are predictive of a treatment response, including those that predict a response to temozolomide (TMZ) in addition to radiation therapy [111
] and TMZ alone [112
]; however, it remains to be concluded whether or not these are truly predictive biomarkers. Further investigation is imperative to develop and validate clinically relevant miRNA profiles for predicting patients that may or may not respond to treatment.
8. miRNAs in Clinical Trials
A major advantage of miRNAs includes their ability to target multiple genes at once, and therefore they have the potential to effectively address cancer heterogeneity. However, the simultaneous targeting of multiple genes may also lead to unexpected side-effects and unwanted toxicities. The primary requirement for miRNA-associated therapies includes a meticulous selection of candidate miRNA. Ideally, the miRNA should target the desired oncogene(s) with minimal off-target mRNAs. In accordance with these measures, several miRNAs have successfully navigated across the preclinical stage, and are discussed below.
Currently, there are few ongoing Phase I clinical trials utilizing miRNAs as therapeutics in cancer. Unfortunately, there are no Phase I clinical trials for miRNA therapy in gliomas. One miRNA used in the treatment of cancer is miRagen-106, an LNA-modified antisense inhibitor of miRNA-155, for cutaneous T-cell lymphoma [63
]. Another is an miR-16 mimic therapy for non-small cell lung cancer patients [63
]. The delivery vehicle for miR-16 is EnGeneIC, a non-living mini-bacteria system that can be modified and used as targeted drug delivery vehicles [63
One of the first tested miRNA therapies for cancer is MRX34, a liposome-formulated synthetic miR-34a mimic (miRNA Therapeutics, Austin, TX, USA). As previously discussed, miR-34a levels are decreased in multiple cancers and function as a tumor suppressor. MRX34 directly inhibits at least 24 different oncogenes, including c-met, Notch, CDK4, and BCL2. Pre-clinical results in multiple mouse models were promising and revealed successful, safe systemic delivery of the miR-34a mimic with no change in cytokine profiles. These studies also demonstrated induction of apoptosis, with an associated tumor response [131
A multi-center phase I trial was then initiated in 2013, which involved treatment of primary liver cancer, lymphoma, small cell lung cancer, and melanoma patients with a miR-34a mimic, delivered systemically by intravenous infusion. Substantial evidence of antitumor activity and acceptable safety levels were highlighted in a subset of patients with refractory advanced solid tumors [62
]. However, the trial was suspended due to major immune-related adverse events, including severe (Grade 4) cytokine release syndrome. At present, the trigger for these immune reactions is unclear, and pre-clinical trials may have to be repeated [63
Neurotoxicity induced by miRNA-associated immunomodulation is an important area of investigation. MiRNAs excreted from cancer cells can directly bind to toll-like receptors (TLRs) at the surface of neighboring immune cells, which may lead to activation of the unwarranted signaling pathways in the recipient cells [123
]. This may result in neurodegeneration, as evident with the let-7b-mediated activation of TLR7 in neurons [135
Another immuno-toxicity with miRNA-based therapy is the aberrant activation of specific innate immune effector cells, including natural killer (NK) cells via the TLR1-NF-κB pathway. This may affect multiple NK cell functions, including cytokine production, proliferation, and cytotoxicity, all of which may alter the immune response and induce malignant transformation [134
]. Also, it may lead to secretion of inflammatory cytokines and type I interferons (IFNs) by TLRs, based on the structure, sequence, and the delivery system of specific miRNAs, thus affecting the innate and adaptive immune response. This may activate a cascade of events leading to the priming of surrounding immune cells, causing them to become more sensitive to RNA stimulation [123
]. These toxicity issues need to be addressed in order to better understand and prevent immune-related adverse events similar to those that occurred with MRX34.
Another miRNA that has gained an edge in the clinical domain is Miravirsen, which is currently being evaluated in Phase II clinical trials for the treatment of Hepatitis C virus (HCV) infection. Miravirsen is a β-D-oxy-LNA-modified phosphorothioate antisense oligonucleotide that targets miR-122. miR-122 is endogenously expressed in the liver and important in hepatocyte development, differentiation, and metabolism. This miRNA is also involved in replication of Hepatitis C virus RNA when in complex with the Argonaute 2 protein. This miR-122/viral RNA/arogonaute 2 protein complex also helps prevent the nucleolytic degradation of hepatitis C. In the presence of Miravirsen, miR-122 is unable to associate with the complex and the virus cannot replicate [91
During a recent Phase 2a study for patients with chronic HCV infection, there was a significant dose-dependent decrease in HCV load with Miravirsen therapy. The majority of adverse side effects consisted of Grade I headaches [92
]. Miravirsen, in combination with other anti-viral therapies (telaprevir and ribavirin), for the treatment of chronic Hepatitis C virus infection, is currently being studied in a phase II trial [139
Clinical miRNA trials for other disease types such as type 2 diabetes, non-alcoholic fatty liver disease, and scleroderma also exist. An anti-miRNA therapy utilizing N-Acetylgalactosamine (GalNAc)-conjugated anti-miRNAs for miRNA-103/107 is currently a Phase I trial for patients with type 2 diabetes and non-alcoholic fatty liver diseases. Additionally, scleroderma patients have the opportunity to enroll in a miRNA-29 mimic Phase I trial, which uses a cholesterol-conjugated miRNA duplex-based delivery system [130
9. Conclusions and Perspectives
Although still in its infancy, the translation of miRNAs from the bench to the clinic has the potential to greatly impact personalized medicine for patients with gliomas or other malignancies, either as biomarkers, a monotherapy, or in combination with other treatment modalities. Given their pivotal roles in glioma development and progression, miRNAs have received a great deal of attention as potential therapeutic targets and biomarkers for glioma patients. These are attractive therapeutic candidates, due to their ability to simultaneously modulate multiple genes across signaling pathways, their small size, and their stability. In addition to the therapeutic potential of miRNA, research is currently underway to identify miRNA signatures that may serve as diagnostic, prognostic, and predictive biomarkers for glioma development, progression, and treatment in the clinic. In addition to tissue miRNA biomarkers, miRNAs are readily accessible in biofluids (blood and urine), which may offer an additional, non-invasive source of biomarkers that may not only facilitate diagnosis and prognosis, but may also help determine the best treatment options and monitor treatment response. Furthermore, the low cost, low RNA input requirements and rapid processing of miRNA assays make them efficient and suitable for use in the clinic.
One limitation to the clinical translation of miRNA is identifying of an ideal therapeutic miRNA candidate. An ideal miRNA candidate should have multiple oncogenic targets with limited non-specific targets. The most promising miRNA sequence would likely simultaneously target multiple features of tumor development and progression, including angiogenesis, invasion, and/or cell metabolism. Given that the targets of miRNAs may involve multiple pathways via imperfect matching with 3′-UTRs, off-target gene silencing of tumor suppressor genes may lead to toxicities and/or reduced therapeutic effects. A comprehensive knowledge of the mRNA targets of each miRNA is imperative for avoiding off-target effects and unwanted toxicities.
High-throughput miRNA profiling studies are being utilized to identify novel miRNA and their associated targets. While new strategies for the identification and characterization of targets of individual miRNAs have been developed, limitations still exist. Bioinformatics software use algorithms to predict potential mRNA targets based on the miRNA “seed” sequence, which is typically located on nucleotides 2–7 from the 5′ end of the miRNA [140
]. Hundreds of candidate mRNA targets are often generated based on these algorithms, all of which must be confirmed in vitro due to the high likelihood of false positives. While validating these targets in vitro can often be tedious and expensive, confirming targets by an experimental approach is more sensitive and accurate than a computational approach.
Previous studies have identified candidate miRNA by pre-clinical experiments, and while this is valuable, further validation in patient samples needs to be conducted. However, there are limitations involved with screening for miRNA biomarkers in patient samples, including small sample sizes, as well as a lack of long-term follow up and survival data. This is particularly true for lower grade gliomas, compared to glioblastomas, because of their rarity and much longer and variable survival times. Additionally, many of these analyses are performed using heterogeneous patient populations consisting of multiple grades, histologies, and treatment modalities, which create challenges during data analysis and interpretation. Furthermore, glioblastoma has a vastly different genetic make-up from lower grade gliomas, both pathologically and molecularly. While we discuss miRNAs and their involvement in gliomas in general in this review, more work needs to be done to distinguish the miRNAs important for gliomagenesis, and progression in low grade gliomas versus glioblastomas.
The recent failure of the MRX34 trial has provoked the development of a deeper understanding of the barriers and associated toxicities of miRNA-based therapy for cancer. Further pre-clinical and mechanistic studies are imperative for elucidating the signaling pathways and immune responses modulated by specific miRNA, as well as further improving our strategies for their safe and effective cellular delivery. While no clinical trials involving miRNA for gliomas have been instituted to date, recent research brings tremendous hope and potential for miRNA-based therapeutics as personalized therapeutic interventions for glioma patients.