Keywords

microRNA Biogenesis

The microRNAs (miRNAs) are a class of 18–25 nucleotides long RNAs involved in the repression of translation and in the adjustment of protein production in response to various stimuli [1–3]. Their expression must be accurately controlled to ensure plethora of cellular processes [4–6]. The miRNA biogenesis involves several steps, each step being subject to specific controls (for review [7]). Briefly, a long (thousand nucleotides long) RNA called the primary-miRNA (pri-miRNA) is transcribed from the genome mostly by the RNA polymerase II. This pri-miRNA contains one or several local stem-loop structures (called precursor(pre)-miRNA) in which the mature miRNA sequence is embedded. Next, a specific complex, called the Microprocessor and containing the RNAse III Drosha, crops the pre-miRNA from the pri-miRNA. The pre-miRNA is exported to the cytoplasm where another RNAse III, Dicer, processes the pre-miRNA into duplex of miRNAs. Only one strand of this duplex will guide a protein complex onto mRNAs harboring partial sequence homology and eventually trigger translation repression mostly by mRNA exonucleolytic cleavage [8]. The first two steps are believed to be the main control points for miRNA regulation [7, 9]. Similarly to protein coding genes (PCGs), control of pri-miRNA transcription involves DNA-binding proteins (i.e., transcription factors, TFs) that recognize specific cis-regulatory DNA motifs in the promoter region of the pri-miRNA. The definition of miRNA promoters remains elusive. The pri-miRNAs are unstable molecules making hard the precise identification of their 5′ end, i.e., miRNA Transcription Start Sites (TSSs) . Numerous studies have tackled that problem and proposed different approaches to characterize miRNA TSSs, mostly based on features of PCG promoters such as CpG content, epigenetic marks, nucleosome positioning [10–19] but the results are quite mixed. A precise and complete map of miRNA TSSs/promoters is thus still missing precluding a genome-wide view of miRNA transcriptional regulations and the identification of potential miRNA-specific regulations. This lack of knowledge does not impede the study of specific miRNA loci though. We and others have shown that miRNA genes and PCGs are regulated by the same TFs. For instance, we have demonstrated that the PML-RARA oncogenic protein, which is associated with the Acute Promyelocytic Leukemia, represses the transcription of retinoic acid-responsive miRNA genes similarly to its action on PCGs [20]. Likewise, we showed that the antagonism between retinoic acid and estrogen signaling initially reported for PCGs [21] is also observed on miRNA genes [22].

At the posttranscriptional level, control of the miRNA biogenesis can be subjected to RNA-binding proteins (RBPs), which recognize specific RNA motifs on or at the vicinity of the pre-miRNAs. For instance, the LIN28 protein, a developmentally regulated RBP, can recognize a specific motif in the loop of the pre-miRNAs belonging to the let-7 family and selectively blocks their processing [23]. Also the p72 DEAD Box RNA Helicase binds a motif located in the 3′ flanking region of the pre-miRNAs and this binding can be controlled—in a cell-density-dependent manner—by the sequestration of p72 by YAP, a downstream target of the tumor-suppressive Hippo-signaling pathway [24].

These transcriptional and posttranscriptional regulations make miRNA extremely sensitive to various intra- and extracellular stimuli (e.g., hormones, vitamins, nutrients, pharmacological molecules, or hypoxia). They notably ensure that the miRNA repertoire is controlled in a temporal and cell-specific manner. These features were first reported by Chen et al., who observed that the miR-181 was preferentially expressed in the B-lymphoid cells and that its ectopic expression in hematopoietic progenitor cells redirects lymphopoiesis towards the B-cell lineage [25]. On the other hand, these tight regulations can have severe consequences in human diseases in particular cancer [22, 26–30].

microRNA Deregulation in Cancers

The miRNAs are key players in cancer initiation and progression, including metastasis formation [31–33]. This field of research is probably one of the most productive in terms of publications (16,022 publications related to “miRNA and Cancer” listed in PubMed in March 2015 with an increase throughout the years). The miRNAs can act as oncogenes (“oncomirs”) or tumor suppressors [34]. He et al. first reported the potential of one miRNA cluster, the miR-17/92, to act as an oncogene [35]. In 2007, Chang et al. showed that the miR-34a, which is transcriptionally regulated by p53, has a tumor suppressor activity [36]. Several databases have now been created to list the miRNA activity in specific cancer type [37, 38]. As observed for PCGs [39–43], the oncogene/tumor suppressor activity of miRNAs depends on the cellular context and/or the type of cancer considered. For example, the miR-221 can act as an oncogene in liver cancer [44] while playing a tumor suppressor role in erythroblastic leukemia [45].

The miRNA deregulations observed in cancer (i.e., forced expression for oncomiRs and downregulation for tumor suppressor miRNAs) can occur at the gene (deletions, amplifications, or mutations of miRNA genes), the transcriptional (epigenetic silencing, deregulation of transcription factors), and/or the posttranscriptional (deregulation of the miRNA biogenesis pathway) levels (for review [29]). The action of miRNAs can also be impaired without affecting miRNA expression levels by, for example, genomic mutations that can modify either the sequence of the miRNAs and/or the sequence of their targets [46]. We provided earlier some examples of specific transcriptional regulations responsible for miRNA deregulations [20, 36]. Likewise, the miR-15a and miR-16 are downregulated in the majority of chronic lymphocytic leukemia cases because the corresponding gene is frequently deleted [47]. The transcription of miRNA genes can also be silenced by DNA methylation [48]. At the posttranscriptional level, the reactivation of LIN28 is many human tumors can lead to the exclusive downregulation of let-7 miRNAs [49]. The expression of several key proteins involved in the processing or the action of miRNAs (e.g., Dicer, Drosha, Argonaute 2) is perturbed in certain cancers [50, 51] with presumably broad impact on cell biology.

These deregulations ultimately generate miRNA profiles that can be associated with cancer types/subtypes and/or response to chemotherapies [52–57]. Most of these profiles have been made available in several databases. PhenomiR provides data from several studies that investigate deregulation of microRNA expression in various diseases (not only cancer) and biological processes as a systematic, manually curated resource [58]. OncomiRDB is specifically dedicated to cancers [37]. Wang et al. manually curated 2259 entries of cancer-related miRNA regulations with direct experimental evidence from approximately 9000 abstracts, covering more than 300 miRNAs and 829 target genes across 25 cancer tissues [37]. PROGmiR is aimed at providing potential prognostic properties of miRNAs in several cancer types derived from publicly available data from Gene Expression Omnibus (GEO) and The Cancer Genome Atlas (TCGA) [59]. The next question remains to determine whether these profiles contain clinically relevant biomarkers that could serve in diagnostic, prognostic, and/or theranostic tests.

Specific Advantages of microRNAs in Cancer Diagnosis

In addition to specific mutations associated to specific cancers [46], miRNA levels can also be indicative of cancer initiation, progression, and metastasis formation. Measuring miRNA levels is relatively straightforward. Several technologies are now available to profile either a specific set of miRNAs (RT-qPCR, Nanostring, microarrays) or the whole miRNA repertoire (small RNA sequencing). Advantageously, RT-qPCR does not necessitate large amount of RNA and is highly sensitive and specific. Moreover, several assays are commercially available rendering miRNA profiling easy even in clinical practice. It is important to note that each platform has, however, its advantages and drawbacks. For instance, the use of specific RT primers [60] could be a heavy procedure compared to the universal method, which uses linkers and one common RT primer. Problems with cross-priming can also lead to specificity issues and make it difficult to distinguish miRNAs belonging to the same family and differing by 1 or 2 nucleotides only. The Nanostring technology utilizes color-coded barcodes, which hybridize with the targeted miRNAs without the need of amplification thereby providing very sensitive digital data. However, similar to microarrays, RT-qPCR and Nanostring technologies are targeted approaches that do not allow the detection of novel miRNAs that can be species- and tissue-specific [61, 62]. In that context, RNA sequencing is definitely the best way to discover novel miRNAs. It can also detect sequence variation and posttranscriptional modifications thereby providing a more complete picture of the miRNA repertoire. However, its cost is still high to be envisaged in clinics. Besides, analysis of sequencing data is still a complex process, which requires rigorous bioinformatics approaches and refined sequence algorithms.

The miRNAs can be detected in a variety of human tissue specimens, fresh or Formalin-Fixed Paraffin Embedded (FFPE), and in almost all human biological fluids (e.g., serum, plasma, saliva, urine) [63–67]. In contrast to most RNAs, circulating miRNAs are remarkably stable [68]. In fact, circulating miRNAs represent a potent mode of intercellular communication [69, 70]. The secretion of miR-105 through exosome destroys tight junctions between endothelial cells thereby facilitating metastasis propagation [70]. The molecular mechanisms responsible for the secretion of miRNAs remain largely unknown. Circulating miRNAs can be free, packed into exosomes or other microvesicles present in body fluids [71] or can be associated with (lipo)proteins (HDL [72] and Argonaute 2 protein [73]). Plethora of studies showed association between the presence of one or several extracellular circulating miRNAs in a given biological fluid and cancer initiation/progression or response to chemotherapy. These profiles have been listed and classified in the miRandola database [74]. miRandola contains 2132 entries, with 581 unique mature miRNAs and 21 types of samples. miRNAs are classified into four categories, based on their extracellular form: miRNA-Ago2 (173 entries), miRNA-exosome (856 entries), miRNA-HDL (20 entries), and miRNA-circulating (1083 entries) [74]. miRandola is also connected to miRò, a compendium, which integrates various online resources (ontologies, diseases, and targets) to provide users with miRNA-phenotype associations in humans [75].

All these features make miRNAs appealing candidates for non-invasive diagnostic tests and several companies have indeed decided to meet the challenge (e.g., Santaris Pharma, Rosetta Genomics, Cepheid, Prestizia-Theradiag, and IntegraGen; [76]). However, at this stage, miRNA signatures per cancer type are still inconsistent [77, 78] impeding their usage in clinics and calling for further development and research.

Challenges in microRNA-Based Diagnosis

One important challenge in the field of microRNA-based diagnosis is to find the sources of inconsistencies in order to propose standardized protocols. Inconsistencies in miRNA signatures could come from sample procurement and could be the results of, for instance, platelet contamination of the plasma [79, 80] or hemolysis occurring during blood collection [81–84]. The protocols used to extract miRNAs also differ and can introduce significant variability. One important point to compare miRNA extraction protocols is to evaluate the quantity and the quality of the extracted miRNAs. Though the size and abundance of ribosomal RNAs is traditionally used as a quality marker for large RNAs, these RNAs cannot be informative on the quality of the miRNA extraction and specific methods are required (e.g., Agilent Small RNA Kit, synthetic miRNA standards). Moreover the quantification of miRNAs is only accurate in samples where larger RNAs are not degraded. The low concentration of RNAs in body fluids also makes the estimation of miRNAs abundance particularly difficult [85]. Besides, protein and lipid content of plasma and serum samples could affect efficiency of RNA extraction and introduce potential inhibitors of PCR [86]. This can be estimated using a spiked non-human synthetic miRNAs (typically from Arabidopsis thaliana or Caenorhabditis elegans) that will go through the entire RNA isolation procedure and will eventually be measured by RT-qPCR. Another aspect that should be considered is that the extraction methods could affect the nature (i.e., nucleotide composition) of the miRNAs extracted. Notably, depending on the protocol used, the quantity of the biological samples can impact the GC content of the miRNAs detected [87, 88]. Since these observations were also made in serum [88] (where large RNAs are barely detected), it is likely that the selective lost of miRNAs is linked to the presence of additional compounds (proteins and/or lipids), which are associated with miRNAs. Together these studies [87, 88] argue for standardization in quantities/volumes of starting materials to allow strict comparison of miRNA profiles. In fact, all these considerations point to the urgent need of consistency in all the steps of miRNA extraction procedures.

Analytical aspects also impact the definition of miRNA signatures. Among them, normalization of the data, which is required to remove unwanted technical variation present in the samples, is critical. On common approach is to use other abundant noncoding RNAs, such as U6 small nuclear RNAs, as normalizers of miRNA expression. However, the biology of such RNAs is quite distinct from miRNA biology in terms of transcription, processing, and tissue-specific expression [89]. An alternative is to use miRNAs whose expression is supposed to be stable in various conditions. However, this strategy can be limited by the fact that the chosen reference miRNAs are sensitive to other biological processes and/or other diseases commonly encountered in clinics. In that case, the expression of the normalizer miRNAs could fluctuate in patients and introduce serious bias. In fact miRNA levels are extremely sensitive to various stimuli and conditions, even nonpathogenic, from gender [90, 91] and age [92, 93] to nutrients such as amino acids, carbohydrates, fatty acids, vitamins, and phytochemicals (curcumin, resveratrol) [94, 95]. If clinically relevant, these aspects should be invariably taken into account in the cohort used to define a miRNA signature. The ideal strategy would be to restrict potential miRNA signature to miRNAs whose transcriptional/posttranscriptional regulations are relevant for the cancer or the chemotherapy considered. This is where translational research meets fundamental research as this strategy clearly depends on a better understanding of miRNA regulations.

Conclusion

The discovery of miRNAs [96] has opened up new avenues of research in biomedicine, in particular in cancer, and contributed to a large extent to the “Noncoding RNA revolution” [97]. It is remarkable to note not only the fast rate of fundamental discoveries made in two decades (illustrated by the exponentially growing number of publications) but also the velocity with which “miRNA gets to business” [76]. These molecules indeed harbor specific features (stability, easy manipulation, reasonably simple detection, tissue specificity) that make them appealing candidates as diagnostic, prognostic, or theranostic biomarkers and even therapeutic targets [64, 98]. However some uncertainties remain [77, 78] that may prevent their immediate large-scale exploitation. Collective efforts made by clinicians, academic and industrial researchers are needed to circumvent these limitations and promote the transfer of miRNAs from bench to bedside.