RISCy Business: MicroRNAs, Pathogenesis, and Viruses*

Proteins serve important regulatory and effector functions inside cells. However, the recent discovery that plants and animals have thousands of genes that encode non-protein-coding (nc)RNAs has opened a new vista on RNA-mediated biology. NcRNAs include rRNA, tRNA, small nuclear (sn)RNA, small nucleolar (sno)RNAs, micro-(mi)RNAs, and some of the lesser known RNAs such as vault RNAs, Y RNAs, repeat-associated small interfering (rasi)-RNAs, and PIWI-interacting (pi)RNAs (reviewed in Ref. 1). Emerging data now suggest that whereas 2% of the human genome encodes for protein-coding RNAs, 60–70% of our DNA is transcribed into ncRNAs (2, 3). Thus, the earlier view that ncRNAs are largely, if not exclusively, constituted by the relatively abundant rRNA, tRNA, snRNA, and snoRNAmoieties is likely an oversimplification.More sensitive analytical methods such as reverse transcription-PCR and DNA tiling arrays have revealed that genomes of complex organisms are replete with numerous less abundant ncRNAs, which can contribute a hitherto unrecognized regulatory dimension.

Proteins serve important regulatory and effector functions inside cells. However, the recent discovery that plants and animals have thousands of genes that encode non-protein-coding (nc)RNAs has opened a new vista on RNA-mediated biology. NcRNAs include rRNA, tRNA, small nuclear (sn)RNA, small nucleolar (sno)RNAs, micro-(mi)RNAs, and some of the lesser known RNAs such as vault RNAs, Y RNAs, repeat-associated small interfering (rasi)-RNAs, and PIWI-interacting (pi)RNAs (reviewed in Ref. 1). Emerging data now suggest that whereas 2% of the human genome encodes for protein-coding RNAs, 60 -70% of our DNA is transcribed into ncRNAs (2,3). Thus, the earlier view that ncRNAs are largely, if not exclusively, constituted by the relatively abundant rRNA, tRNA, snRNA, and snoRNA moieties is likely an oversimplification. More sensitive analytical methods such as reverse transcription-PCR and DNA tiling arrays have revealed that genomes of complex organisms are replete with numerous less abundant ncRNAs, which can contribute a hitherto unrecognized regulatory dimension.

miRNA Biogenesis and Mechanisms of Action
The first miRNA, 2 lin-4, and its target mRNA, lin-14, were described in Caenorhabiditis elegans by Ambros, Ruvkun, and colleagues in 1993 (4,5). Subsequently, computational analysis of aligned regions between human, mouse, and puffer fish genomes initially led to the prediction of ϳ255 discrete miR-NAs in the Homo sapiens genome (6). That early number has been quickly exceeded by the latest enumeration of 474 characterized human miRNAs in the Sanger miRBase sequence data base (release 9.1). Later in silico estimates have posited ϳ1000 or more human miRNAs (7,8); some of these newer suggestions have been verified by direct cloning and RAKE (RNAprimed, array-based Klenow enzyme) assay (9). Currently, difficulties with the reliability of computer-based miRNA prediction can be attributed to the absence of any single property sufficient for accurate determination and the realization that novel miRNAs, by definition, may contain characteristics not fully recognized by extant pattern algorithms. Additionally, experimental confirmation (e.g. direct cloning) of miRNAs remains challenging because expression profiles of ncRNAs can be constrained temporally, spatially, and in a tissue-specific fashion. Moreover, the expression levels of individual miRNAs can vary by 3 orders of magnitude (10). Indeed, in a recent extensive miRNA cloning attempt to identify novel miRNAs, the investigators reported that 76% of their new miRNAs were cloned only once and that they failed to clone ϳ100 discrete human miRNAs that had been documented previously by others (9). This experience illustrates the rigor and challenges encountered in attempts to validate rare miRNAs that are expressed at low levels.
Human miRNAs are present in introns of coding genes and introns and exons of noncoding transcripts (11). MiRNAs are small ncRNAs of 18 -25 nucleotides. A mature miRNA begins with the transcription by RNA polymerase II of a long primary transcript (pri-miRNAs) that contains a "hairpin" structure. pri-miRNA is first cropped in the nucleus into an ϳ70-nucleotide stem-loop RNA intermediate (pre-miRNA) by a protein complex that contains the Drosha ribonuclease with an RNAbinding protein, known as DGCR8 in humans (12). The pre-miRNA is then ferried by Exportin 5 into the cytoplasm (13) and further processed by a cytoplasmic ribonuclease III enzyme, Dicer, to mature miRNA (Fig. 1A). One strand of this mature double-stranded miRNA is destined as the guide strand, with the other as the passenger strand. The guide strand is channeled by Dicerinteracting proteins PACT and TAR RNA-binding protein (TRBP) (14 -16) into an RNA-induced silencing complex (RISC). A detailed review of miRNA biogenesis and RISC complex formation has been presented elsewhere (17).
An miRNA-armed RISC (mi-RISC) mediates miRNA function(s) inside cells. In plants, mi-RISC use miRNA guides that are perfectly complementary to either the coding region or the 3Ј-untranslated region of cognate mRNAs. In the setting of an miRNA-mRNA interaction driven by perfect complementarity, plant mi-RISC can mediate mRNA cleavage/degradation similar to that described for siRNA (or si-RISC)-mediated silencing (18,19). By contrast, animal and human mi-RISC recognize target mRNAs using base-pairing in a manner tolerant of mismatches. Thus, in animal cells, the imperfect miRNA-mRNA complementarity is commonly composed of matched nucleotides at positions 2-7 (termed the seed sequence) in the 5Ј-portion of the miRNA (20) (Fig. 1B) with mismatched nucleotides at positions 10 and 11. These mismatches preclude the endonucleolytic cleavage of mRNA, a phenomenon normally observed with si-RISC-mediated RNA interference (RNAi), by mi-RISC.
Once a mi-RISC-mRNA interaction forms in a human cell, how does the resulting complex trigger mRNA silencing? As yet, the answer to this question remains incompletely elucidated and somewhat controversial. Nonetheless, current data are compatible with multiple miRNA mechanisms that either repress mRNA translation or enforce premature mRNA decay. There is evidence that mi-RISC-mRNA interaction can promote inhibition of translational initiation, increase co-translational degra-dation of nascent proteins, reduce the elongation rate of translation, and/or increase the rate of mRNA deadenylation (21)(22)(23)(24)(25)(26). More recent data suggest that the eIF6 component of miRNA-RISC can prevent productive assembly of the 80 S ribosome complex (27). Which mechanism operates under what conditions remains contested. However, the retention of mi-RISC-mRNA in ribosome-free translationally silent cytoplasmic organelles termed processing bodies (P-bodies) does appear to account for some aspects of silencing (28,29).

Roles Played by miRNAs in Development and Disease
What roles are served by the several hundred characterized human miRNAs? Currently, only a handful of human mRNAs have been validated as specific targets for miRNAs. Extant bioinformatic predictions suggest that a single miRNA through imperfect complementarity can potentially target ϳ100 different mRNAs (30). A reasonable extrapolation from these predictions argues that up to 30% of all mammalian genes are under some degree of miRNA regulation (31). Experimental observa-tions do support important physiological roles contributed by miRNAs. For instance, in C. elegans, zebrafish, Drosophila, mice, and humans, miRNA expression occurs with tissue-restricted profiles (32)(33)(34)(35)(36) and differential timing during development (37)(38)(39). Both patterns suggest that miRNAs contribute to morphological development and organogenesis.
Studies that have depleted Dicer, the RNase III enzyme critical to miRNA maturation, from zebrafish and mice have been functionally informative. Knock-out of Dicer in zebrafish arrests embryo development 8 days after fertilization (40), and dicer-deficient mice lose viability prior to axis formation during gastrulation (41). Conditional depletion of Dicer in mice has shown that general loss of miRNA function(s) affects T-cell development (42,43), limb formation (44), and organ maturation (45,46). Some of these results should be interpreted cautiously because Dicer is known to play roles in heterochromatin formation and chromosome segregation in yeast Schizosaccharomyces pombe, which apparently does not encode any known miRNAs. Hence, Dicer has functions beyond miRNA processing, and some of the multicellular loss-of-Dicer phenotypes could occur independently of miRNA effects.
The notion that miRNAs are involved in human diseases arises from two sets of observations. A clue that dysfunction of miRNA-pathway(s) contributes to pathology came from the recognition that humans with mutations in DGCR8 (a Drosha cofactor) or fragile X (a RISC cofactor) suffer, respectively, from DiGeorge syndrome (47) and mental retardation (48,49). Second, Ͼ50% of human miRNA genes are present at genetic loci (such as fragile sites, common break point regions, etc.) implicated in cancers. Accordingly, miRNA expression patterns are invariably found to be very different in tumor tissues when compared with matched normals. Indeed, in studies of 334 leukemia and 540 primary tumors, Lu et al. (50) and Volinia et al. (51), respectively, observed miRNA cancer signatures that distinguished tumors based on their tissue origin. Other miRNA profiling studies have also substantiated malignancy-specific expression patterns in lung (52), colon (53), breast (54), and heptacellular (55) cancers. A recent study has added indirect support that miRNA changes are causal, rather than consequential, of cellular transformation (56).
How then do miRNA changes promote human carcinogenesis? The full answer is unknown, but there are several ways that one could consider mechanisms. One perspective posits that some miRNAs are tumor suppressors, whereas other miRNAs are oncogenes (see Table 1). Reduced expression of the former or gained expression of the latter would confer a growth advantage to cells. Experimental data support that miR-15a and miR-16-1 provide tumor suppressor function by targeting Bcl2 (57), and miR-155 is oncogenic through incompletely understood effects on perhaps PU.1 and C/EBP␤ (58). However, one must entertain the possibility of tissue-specific determinants of miRNA function when interpreting findings. Thus, although miR-15a and miR-16-1 are down-regulated in chronic lymphocytic leukemia (consistent with their postulated tumor suppressor function (59)), the same miRNAs are paradoxically overexpressed in endocrine pancreatic tumors (60).
MiRNA changes in cancers offer suggestive correlations, which are insufficient to prove conclusively their causality for FIGURE 1. miRNA biogenesis and function. A, processing of pri-miRNA transcript by Drosha and pre-miRNA by Dicer is shown. One strand of the miRNA becomes a guide strand and is incorporated into RISC, and the other (passenger) strand is discarded. B, perfect complementarity of miRNA with mRNA specifies cleavage, whereas imperfect hybrid formation leads to translational silencing. Further details regarding proposed mechanisms used by mi-RISC for silencing are discussed in the text.
carcinogenesis. Direct validation of causality can, however, emerge from studying transforming viruses that do not encode oncogenes but do integrate near genome loci whose expression drives tumorigenesis. Using such an approach, the first proof of an oncogenic miRNA causal of cancer, was miR-155/BIC (reviewed in Ref. 61), which was activated in chicken tumors by retroviral insertion (i.e. avian leukosis virus integration). Overexpression of human miR-155/BIC has subsequently been linked to the development of Hodgkin's (58,62) and Burkitt's (63,64) lymphoma. Additional evidence for oncogenic miRNAs used by viruses to transform cells comes from examples of integration induced activation of miR-17-92 and miR-106a-363 in SL3-3 murine leukemia virus tumors (65,66). Separately, findings of increased human papilloma virus insertion at miRNAcontaining fragile sites in cervical carcinomas (67) add credence to the general contribution of miRNA perturbations to cancers.

ncRNAs and Mammalian Defense against Pathogenic Infections
Human infectious diseases in the 21st century represent the second leading cause of death and the leading global burden on disability-adjusted life years. Infections from HIV-AIDS and hepatitis B and C viruses alone cause Ͼ3 million deaths annually (68). Accordingly, an understanding of how mammals defend against viral infections and whether such defenses employ ncRNAs is important.
ncRNAs (e.g. RNAi activity) have been proposed to confer wide-ranging antiviral functions in bacteria (69,70), plants (reviewed in Ref. 71), and animals (reviewed in Refs. 72 and 73). However, some have argued that sequence-specific innate immunity in mammals has been replaced by a sequence-nonspecific, double-stranded RNA-triggered, interferon-based defense (74). Although this hypothesis frames one possible evolutionary scenario, several findings are incongruent with the extinction of mammalian sequence-specific RNAi in favor of an interferon defense. First, although RNAi is effective for silencing the replication of all classes of mammalian viruses (reviewed in Ref. 75), interferon in practical applications has been shown to be modestly efficacious only in the treatment of two viral infections (HBV and HCV (76,77)). Second, there is evidence that cellular miRNAs are employed in a sequencespecific fashion by primate cells either to restrict or augment ( Fig. 2) the replication of viruses such as primate foamy virus type 1 (PFV-1) (78), HCV (79), or the human immunodeficiency virus type 1 (HIV-1) (80,81). In response, viruses have evolved RNAi suppressor proteins or decoy RNAs to counter these cellular restrictions (82) (reviewed in Ref. 83). Third, several recent studies have found that humans and mice do process cellular siRNAs (84,85) and miwiRNAs (86), another class of small ncRNAs, for sequence-specific defenses against endogenous retroviruses/retrotransposons. Finally, emerging findings implicate the involvement of human miRNAs in inflammatory responses against pathogenic infections (87,88) (Table 1), providing further evidence for sequence-specific RNA function in mammalian hostpathogen interaction.
Mammalian viruses appear to encode viral miRNAs (vmiR-NAs) (89,90), which are processed in primate cells. This observation is consistent with the concept that RNAi pathways are generally important and are preserved from bacteria to plants to mammals. However, it has been argued that not all viruses necessarily encode miRNAs and that vmiRNAs exist selectively only in viruses (e.g. herpes) with large DNA genomes (74). This notion arose in part because some investigators (89) have failed to predict and clone small viral ncRNAs described by others (91,92). Nonetheless, this type of failing should be interpreted with circumspection. For instance, three separate studies on vmiRNAs encoded by herpes simplex virus type 1 (HSV-1) illustrate the potential for discrepant results. Two of the three studies (93,94) cloned discrete miRNAs that were not predicted by the third study (89) and failed to clone any of the eight predicted HSV-1 miRNAs (89). Drawing from the experience of identifying and cloning rare cellular miRNAs (9), it seems reasonable that temporal, spatial, and different tissue culture conditions could influence individual experimental successes at capturing vmiRNAs. Moreover, a recent bioinformatic analysis found that secondary structures frequent in ncRNAs are evolutionarily selected against in coding regions of genomes (95).  This suggests that smaller DNA and RNA viruses based simply on size constraints that require greater portions of their genomes for coding purposes are less likely to retain ncRNAs than their larger viral counterparts. We note that initial nucleotide sequence-specific selection by the cell against highly mutable viruses such as HIV-1 and HCV could quickly reshape viral genome sequences. Viruses may mutate to escape restriction and even evolve adaptations (e.g. the emergence of T20 (an anti-viral peptide)-dependent HIV-1 replication after initial negative selection against HIV-1 by T20 (96)) to turn negative effects into positive factors (Fig. 2). This type of virus-host, cat-and-mouse interplay may ultimately limit the antiviral effectiveness of RNAi and could explain how, unlike other viruses, HCV co-opts a cellular miRNA for enhancing, rather than inhibiting, viral replication (79) (Fig. 2).

Perspective
The past few years have been exciting and challenging for studying ncRNAs. It is a safe prediction that many new rules, principles, and functions of small RNAs await elucidation. On the horizon are nascent findings that small RNAs may also serve gene activating rather than gene silencing functions (97). New discoveries in the coming days will likely further advance our knowledge of miRNAs and their applications toward human diseases.