The recent discovery of non-coding RNAs, such as long non-coding RNAs (lncRNAs) and microRNAs (miRNAs), has begun to direct more and more attention to these potentially very powerful regulatory molecules. The first miRNA was discovered in 1993 during studies of the timing of embryonic development of different larval stages of the worm Caenorhabditis elegans (C. elegans). In this experiment, Lee and colleagues observed that the RNA transcribed from the lin-4 locus did not encode a protein but instead silenced the gene encoding Lin-14, an important protein in larval development (1). Since then, the number of studies on miRNAs has been rising rapidly. Indeed, a PubMed search (August 4, 2016; keyword “microRNA”) revealed an increase from 5 entries in 2001 (the first year this term appeared) to 10,189 entries in 2015. MiRNA-encoding genes comprise only 1–5% of the animal genome but have been estimated to affect approximately 30% of all protein-coding genes (2, 3). It is being recognized that their regulatory roles are much more sophisticated than initially thought, owing to the cooperativity (i.e., more than one miRNA species can target the same mRNA) and the multiplicity of their targets (i.e., one miRNA can target hundreds of mRNA species) (4). miRNAs have been shown to play important roles in essentially all biological processes (5), and the differential expression of host miRNAs during infection (1, 6) supports the idea that they may constitute key players in the host response to invading pathogens. We recently summarized trials of therapeutic interventions based on small non-coding RNAs for treatment or prevention of infectious diseases of veterinary importance (7). The current review presents an update on miRNA biogenesis and profiling, discusses some of the challenges encountered when studying them in animals, and summarizes current knowledge of the roles of miRNAs in viral infectious diseases in their respective natural animal hosts. In addition to this, information obtained from cellular and laboratory animal models is presented in cases where data from natural infection are not available or are difficult to obtain. Furthermore, we include some examples of important animal viral diseases where in vitro studies have revealed roles for miRNAs. We also discuss miRNA involvement in prion diseases as examples of fatal, untreatable diseases caused by infectious proteins. Viruses, particularly DNA viruses [Marek’s disease virus (MDV), bovine herpesvirus] and even retroviruses (e.g., bovine leukemia virus), can also encode their own miRNAs, but due to space limitations, this topic is not emphasized in this review, and we refer the reader to excellent existing reviews [e.g., Ref. (8, 9)].

MicroRNAs are non-coding single-stranded oligoribonucleotides of about 22 nucleotides (nts) in length. Their biogenesis and mode of action is illustrated in Figure 1, but the reader is also referred to excellent recent reviews of miRNA biogenesis, e.g., Ref. (10). miRNAs can be transcribed from within protein-coding genes (intragenic miRNAs), from dedicated miRNA coding genes (intergenic miRNAs), or from genes encoding other ncRNA classes, such as small nucleolar RNA (snoRNA) and lncRNAs. More than half of vertebrate miRNA genes lie in introns, implying that most miRNAs are co-expressed with specific host mRNAs (11, 12), although specific transcription start sites for miRNA coding sequences also exist (13). While the majority of miRNA genes are physically separated on the genome (14, 15), many functionally related miRNA genes often, but not always, reside in clusters within the genome (16), probably because they are processed from the same polycistronic transcript (6). When they have their own promoter, miRNA genes are transcribed mainly by RNA polymerase II and, rarely, by RNA polymerase III (17) to form primary miRNAs (pri-miRNAs) (18), which are then folded to produce hairpin structures. Each hairpin structure consists of a 32-nt-long imperfect stem and a large terminal loop. The enzyme Drosha and its cofactor DiGeorge syndrome critical region 8 (DGCR8) cleave the 22 nts downstream of the stem to yield 60-nt-long precursor miRNAs (pre-miRNA) (18). The pre-miRNAs are then exported to the cytoplasm via exportin 5 where the terminal loops are excised by Dicer and tar RNA binding protein (TRBP) to produce short imperfect miRNA duplex intermediates (19). This duplex is then unwound by a helicase into two miRNA strands. One strand (the 5′-strand or guide strand for most miRNAs) is incorporated into the RNA-induced silencing complex (RISC) to target mRNAs, and the other strand (3′-strand, also referred to as miRNA*) is usually degraded but can also persist and take on regulatory roles of its own.

Apart from the classic biogenesis pathway, Dicer or AGO-2-independent (non-canonical) pathways have also been described (20). For instance, mirtrons are miRNAs that are produced from introns by the splicing machinery instead of the Drosha processing complex (21). The maturation of some miRNAs (for example, miR-451) is Dicer independent, but AGO dependent. This is possibly because the stem part of its pre-miRNA is too short (17 bp) to be processed by Dicer (22).

MicroRNAs mainly act in the cytoplasm, namely, by repressing mRNA translation and, to a lesser extent, by inducing mRNA degradation. Actually, the two processes can be connected in that decreased translation may lead to loss of stability. Usually, the seed region (seven or eight consecutive nts at the miRNA 5′-end) binds to complementary sequences in the 3′-untranslated region (UTR) of the target mRNA (23), but miRNAs can also target other sequences, such as the 5′-UTR or coding sequences (24). Remarkably, some miRNAs have been implicated in actually stimulating translation (25, 26). miRNAs can also affect gene expression at the level of pre-mRNA splicing. Usually, this occurs indirectly in that a miRNA regulates translation of a protein involved in differential splicing. miR-124 represents a good example of this: during neuronal differentiation of the mouse, it binds to the mRNA encoding a repressor of alternative splicing, PTBP1, leading to increased synthesis of the PTBP2 protein. The latter, in turn, induces alternative splicing programs leading to neuronal differentiation (27).

Interestingly, mature miRNAs can also be found in the nucleus (28–30). Profiling studies of fractionated cells revealed the enrichment of miRNAs, including miR-320, miR-373, and miR-29b, in the nucleus of cell lines derived from various origins (28, 29, 31–34). The driving force of this cytoplasmic-nuclear shuttle, at least for miR-29b, is thought to be a miRNA-associated hexanucleotide terminal motif (AGUGUU) which increases the stability of the miRNA (29). Their localization in the nucleus suggests that these nuclear miRNAs affect gene expression at critical steps that take place in the nucleus, notably transcription. One mechanism by which they can affect transcription is to promote gene silencing by inducing epigenetic changes, such as histone modifications or methylation of promoter elements (35, 36). Remarkably, it has been shown that a nuclear miRNA can even stimulate transcription by binding to a complementary promoter sequence (34). Taken together, these studies suggest that the scope of miRNA functions has become broader than previously thought. miRNAs themselves can be regulated by several mechanisms. In general, miRNA half-life ranges from hours to days, but this obviously varies depending on the organ, body fluid, and cell type (37). Compared to other RNA species, their stability in stored biosamples is generally much higher than that of other RNA species, which increases their value for biomarker studies using biobanked animal or human biosamples (7). For instance, in contrast to mRNA, they are highly stable in formalin-fixed paraffin-embedded tissue blocks, allowing for localization and expression studies even after years of storage (38). Active degradation can be mediated by small RNA degrading nucleases (SDNs) [reviewed in Ref. (39)]. miRNAs can also be extruded from the cell into body fluids (e.g., blood, CSF, urine) in exosomes or microvesicles. On one hand, this may be an additional mechanism by which intracellular miRNA populations can be regulated. On the other hand, these vesicles can also function as vehicles by which miRNA can spread systemically and potentially be taken up by accessible cells (40, 41). More recently, it was found that miRNA activity can be inhibited by sequestration by circular RNAs (“miRNA sponges”) (42).

In vertebrates, miRNAs have been studied most extensively in humans and mice, in part because much fewer miRNAs have been annotated or made publicly available in other organisms, including animals of veterinary importance (43). Figure 2 shows the number of currently annotated mature and immature miRNAs in different animal species of veterinary importance, as well as in humans and mice, according to miRBase version 21 (43). Humans and mice had the highest number of annotated miRNAs, followed by chicken, cattle, and horse. miRNA expression profiling has assisted in identifying miRNAs that regulate a range of biological processes, and there are several established and emerging methods for measuring miRNA expression profiles in biological samples. The commonly used approaches include reverse transcription quantitative real-time PCR (RT-qPCR), microarrays, and RNA sequencing (RNA seq) (44, 45). In the process of discovering novel miRNAs, the analysis of profiling data follows the same principle in that the generated reads have to be mapped to the reference genome of the concerned species. In this context, the lack of a published reference genome always represents a limitation. The unique feature of cross-species conservation of miRNAs has aided in the identification of miRNAs or their targets especially in the species where the genome (or the miRNome) is not completely annotated (46). The miRNAs from such species have been identified by homology searches where the deep sequencing reads are aligned against the genome of the most closely related species (47). Indeed, most bovine miRNAs have been identified in this way, and a similar method was used to identify goat and sheep miRNAs (48). Researchers from China recently characterized miRNA species in skin and ovaries of ducks (49, 50). An alternative strategy was used in these studies. First, the reads were filtered and then mapped by blast alignment to all known mature chicken miRNA sequences present in miRBase. The sequences that did not correspond to chicken miRNAs were then mapped to miRNAs in other species. This strategy has also been applied to the Chinese hamster (51). There are several computational resources for the identification of miRNAs and their targets (52). Among these, miReader is a newly launched bioinformatics tool for the discovery of novel miRNAs that can be used to identify miRNAs that are not annotated in miRBase, yet without the need for reference genomic sequences or homologous genomes. It shows a high degree of accuracy in a wide range of animal species (47). The prediction and validation of miRNA targets are essential steps in determining their regulatory function. In this regard, the imperfect complementarity between the miRNAs and their mRNA targets represents a major challenge because of potentially false positive predictions. Nevertheless, various online target prediction tools have been developed. Prominent tools include TargetScan (53), miRanda (54), PITA (55), and RNA hybrid (56). Most of these tools rely on basic principles, such as miRNA/mRNA pairing, cross-species conservation of the mRNA 3′-UTR, and the free energy required to form the duplex (57, 58). Kiriakidou indicated that the agreement of mRNA targets predicted for a set of 79 miRNAs by several target prediction tools was only in the range of 10–50% (59). One reason for this may be the divergent use of the degree of conservation. For example, conservation may be used but not directly incorporated into the score (as in PITA), or not used at all (as in RNA22) (60). It is worth mentioning that these tools exhibit some differences, which might be independent of the algorithms, such as using different UTR databases for prediction, cross-species comparison, or alignment artifacts. Other among-tools variations that are related to the prediction algorithms themselves include the number of nts involved in pairing (canonical, marginal, and atypical seed-matched sites) (61), the method used to measure 3′-UTR conservation, the accessibility to the UTR, and the statistical approach used (62). The computational approaches for predicting miRNA–mRNA binding revealed wide variation and bias. Therefore, various experimental strategies have been developed in humans, for instance, “Cross-linking and sequencing of hybrids” (63), which allow the identification of miRNA–target pair chimeras in deep sequencing data (64). Some attempts have been successful to use the mRNA targets to predict biological pathways that are regulated by a set of miRNAs. For instance, the DNA intelligent analysis (DIANA) tool (65) offers the miRPath server, which can create hierarchical clusters of miRNAs and pathways based on the levels of the predicted miRNA/mRNA interactions. Generally, a major limitation of the available web tools is that they include a limited number of species (e.g., human, mouse, C. elegans), leaving many animal species underrepresented, in particular those of veterinary importance. However, some tools (e.g., TargetScan, miRDP, and Microcosm) include some species of veterinary importance, such as chicken, dog, and cow (3, 66, 67). Oasis is a comprehensive online tool for the analysis of RNAseq data that allows for target prediction and analysis of novel (putative) miRNAs, including prediction of miRNomes of species without annotated miRNomes (68). The growing list of annotated miRNAs in species of veterinary importance undoubtedly warrants better inclusion of such species in the online prediction algorithms. Considering the probability of obtaining false positive and negative results using the online prediction algorithms, it seems critical to confirm miRNA function using experimental work. The function of a miRNA can be validated by several experimental approaches. The luciferase reporter assay can be used to test for binding of a miRNA to a predicted target sequence (69). To detect the effect of miRNA–mRNA pairing on the protein level, immunoblotting can be employed. For a detailed analysis of procedures used for the experimental validation of miRNA targets, we recommend references (70–72). In situ hybridization using a labeled antisense probe can be used to investigate which cell type in a tissue expresses a certain miRNA (73), but this is technically not easy, for instance, because an endogenous counterpart miRNA that also contains the antisense sequence may compete with binding of the probe to the target miRNA.

Considering the implication of miRNAs in nearly all biological processes, links between miRNAs and disease status are expected. Earlier reports suggested that miRNAs are involved in the regulation of inflammatory pathways as well as adaptive and innate immunity, stress factors, and cytokine signaling (74). Diseased tissues may show unique miRNA expression patterns, which subsequently might affect virus replication and/or survival. For instance, miRNAs might promote virus replication by direct pairing with virus-derived transcripts, as has been shown for miR-122 and hepatitis C virus (75). A similar mechanism is utilized by bovine viral diarrhea and classic swine fever viruses, where miR-17 and let-7 were found to bind their 3′UTRs, thus enhancing stability and translation of viral mRNA (76). miRNAs may restrict viral replication, as exemplified by miR-32 in primate foamy virus type 1 infection (77).

When studying regulation of host- or virus-encoded RNA targets by miRNAs, the “abundance problem” needs to be considered. Based on studies in cell lines, it has been argued that approximately 100 copies of a miRNA are needed to affect host transcripts and, depending on the number and activity of viral genomes in the cell, likely a higher number to affect viral transcripts (78). Thus, expression changes of low copy number miRNAs during infection may be of little functional consequences. Studies on regulation of host miRNAs in infection often ignore this potentially important issue.

Studies of changes in miRNA populations in infections in the natural host are clearly limited by the fact that they usually do not allow to distinguish between cause and effect (and therefore do not allow to draw conclusions as to mechanisms) but are mostly suitable for biomarker studies and for formulating hypotheses based on the descriptive data obtained. These can then be tested in the appropriate experimental models, if available. Another point to be considered is that a good part of the host miRNome response to infection may be due to collateral cell and tissue damage and turnover and not due to the immune response per se. Clearly, our understanding of the mechanistic associations and implications of miRNAs in animal viral infectious diseases is still far from complete. Here, we review the literature that describes the expression of miRNAs in the context of viral infectious diseases that affect farm and pet animals, with special emphasis on infections in the natural host. Additionally, we discuss miRNA expression in infectious viral diseases in laboratory in vivo and in vitro models where there are no sufficient data involving the natural host.

Many miRNAs have been identified that may affect pathogenesis and outcome of viral infectious diseases of veterinary importance. Nevertheless, many open questions remain. Inter-host differences in susceptibility to a given viral infection are often due to differences in mRNA and, subsequently, protein expression, and the intriguing role of miRNAs in regulating these differences certainly requires further investigations. In this regard, we propose that the study of highly pathogenic avian influenza (HPAI) infections in different avian and mammalian hosts would constitute a promising model system of high relevance to veterinary practice. While infection is asymptomatic in waterfowl, humans and chicken are more susceptible and develop a concurrent strong inflammatory response and high tissue cytokine levels (178). Also, pathogenicity of HPAI H5N1 viruses varies among various breeds of ducks (179). In the same context and as a deviation from the normal ecology of the virus, some recent isolates of H5N1 proved to be lethal for waterfowl species, including ducks (180). Given these observations, comparing the expression patterns of host-encoded miRNAs in response to HPAI in different hosts, organs, and disease states might explain their difference in susceptibility to HPAI infection and help to identify additional elements of the host response that affect disease severity. One research area of great scientific interest and of clinical importance as well is to investigate the role of miRNAs in the development of viral and bacterial coinfections, such as IAV/S. pneumoniae in humans and IAV/Escherichia coli in chicken and ducks (181), as the role of host miRNAs in modulating potential synergies between two pathogens may be substantial. Human populations experienced the emergence of zoonotic diseases, in particular those caused by viruses that cause varying numbers of human fatalities. It is important to invest more efforts to delineate the role of miRNAs that are associated with these zoonotic viral diseases, especially when considering their cross-species conservation. This will improve our understanding of the complex nature of zoonotic pathogens as well as their potential to further establish interhuman transmission.

The discovery of small non-coding RNAs was a turning point in biology. The role of miRNAs has grown with an unprecedented speed from research on worms to a wide variety of physiological and pathological processes in humans and animals. With genome-wide profiling techniques and the tools of bioinformatics, considerable information about miRNAs and their role in animal viral diseases is now available. The current experimental data on the role of miRNAs in host–virus interaction upon infection of a natural host, in laboratory models and in cell-based systems, indicate that miRNAs can contribute to both pathogenesis and clinical outcome of many diseases affecting animal populations on the individual and farm level (Figure 4). However, a general conclusion on the role of miRNAs cannot be drawn yet. One possibility is that they favor the host as part of the antiviral response. This could occur in two ways. First, miRNAs could silence viral transcripts through sequence-specific binding, and thus protein expression. Second, they could indirectly modulate host transcripts in a way that creates a less favorable condition for virus propagation and survival. Vice versa, host miRNAs may be beneficial for the viral pathogens if they tend to increase their replication or survival. Finally, they can be beneficial for both sides as in the case of latent viruses. The role of miRNAs in veterinary medicine is receiving more and more attention. Considering the current efforts to include more hosts of veterinary importance in online miRNA databases, as well the increased availability of high-throughput profiling approaches and functional studies, our understanding of the roles of miRNAs in viral infections of veterinary importance will likely continue to improve and lead to tangible clinical applications in the foreseeable future.

This review paper is part of the PhD thesis of Mohamed Samir Ahmed, completed at the University of Veterinary Medicine (Stiftung Tierärztliche Hochschule), Hannover, Germany.

MA conceived the project, did the literature search, wrote the initial draft of the manuscript, and prepared the figures and tables. LV participated in the bioinformatics analyses and edited the manuscript. FP oversaw the project, edited the manuscript, and takes responsibility for the integrity of the data.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.