Despite the above findings, there remains some contention about the physiological function of RNAi in regulating viruses in mammalian cells. This contention may be caused in part by expectations of mammalian cells/mammalian viruses based on results from invertebrate cells/invertebrate viruses. For instance, when invertebrate viruses infect mosquito cells, the incoming viral dsRNAs or replication intermediates are frequently processed into small exogenous siRNAs (exo-siRNAs) [34
]. By contrast, when mammalian viruses, such as HIV-1, infect human cells, the production of processed viral exo-siRNAs is rare [19
]. Some have interpreted these results to mean that invertebrate cells sense and process viral dsRNAs and then deploy exo-siRNAs in antiviral RISC complexes while vertebrate (mammalian) cells cannot perform these functions. This interpretation however is challenged by experiments in which an authentic short hairpin RNA (shRNA) was engineered into the HIV-1 genome and the genome introduced into human cells, with ensuing efficient siRNA production from the 'viral-shRNA' [36
]. Moreover, a version of the above experiment is performed hundreds, if not thousands, of times every day by investigators who use shRNA libraries cloned into lentivirus vectors to transduce human cells for the purpose of silencing specific target genes. In every instance, the lentivirus-shRNA is recognized and processed by human cells faithfully into the expected siRNA (Figure ).
How then does one reconcile the above observations? One possibility is that human (mammalian) cells, rather than being unable to efficiently recognize and process viral dsRNAs or shRNAs into siRNAs, may actually be more proficient than invertebrate cells at processing and using siRNAs in antiviral RISC complexes against invading viruses. In this case, RNA viruses with double-stranded shRNA-like sequences amenable for processing into siRNAs might be subject to more potent negative selection in human than in invertebrate cells. Over time, strong stringent selection in mammalian cells (and relaxed selection in invertebrate cells) would result in many human viruses being devoid of shRNA-like or dsRNA sequences, while the less robustly restricted invertebrate viruses would still keep shRNA and/or dsRNA sequences. Hence, today's human cells would face many viruses largely lacking shRNA/dsRNA sequences, accounting, in part, for the rarity with which viral exo-siRNAs are detected in mammalian infections (see, for example, [37
]). On the other hand, invertebrate cells would still encounter many viruses that harbor shRNA sequences, and consequently viral exo-siRNAs would more frequently be found in infected invertebrate cells [34
]. In short, the argument is that human cells do, in fact, efficiently process viral exo-siRNA and can use siRNA-RISC as an antiviral defense, but the rarity of this process in human cells may be because many mammalian viruses have already been tightly selected by human cells not to maintain siRNA-producing sequences.
They do, however, encode miRNA sequences [38
]. These viral miRNAs are processed efficiently in human cells, are engaged in miRNA-RISCs, and are frequently used by the cell (or the virus) to target other viral transcripts. Examples of this type of usage include the human cytomegalovirus (CMV) miR-112-1 targeting the CMV IE1 viral RNA [40
], the EBV BART miRNAs targeting the EBV LMP mRNAs [41
], and the HSV miR-H2, miR-H3, and miR-H4 targeting viral ICP0 and ICP34.5 mRNAs [43
]. Similar miRNA-mRNA targeting can also occur in the case of a model retroviral infection [44
]. Taken together, these findings support the view that mammalian cells employ viral exo-miRNA-RISCs to regulate viral infections, paralleling invertebrate and plant cells that use viral exo-siRNA-RISCs to regulate their cognate viruses (Figure ).