Mario Stevenson   about the author

Program in Molecular Medicine, University of Massachusetts Medical School, 373 Plantation Street, Suite 319, Worcester, Massachusetts 01605, USA.
Mario.Stevenson@umassmed.edu

In cells of organisms, ranging from nematodes to primates, there is a process known as RNA interference (RNAi) that effects the degradation of RNA in a highly sequence-specific manner. Scientists have figured out a way to co-opt elements of the RNAi machinery such that almost any RNA can be targeted for degradation. It is now clear that HIV-1 is fair game for RNAi; viral RNA intermediates have been targeted as well as messenger RNAs for cellular co-factors that are required for replication of HIV-1. The hope is that RNAi can be used not only as a research tool, but also as a therapeutic strategy for infection with HIV-1.

Introduction of double-stranded RNA (dsRNA) into cells activates a process known as RNA interference (RNAi) that ultimately promotes the degradation of messenger RNAs that are homologous to the dsRNA trigger. The phenomenon whereby genes are silenced by homologous sequences was first recognized in transgenic plants and termed co-suppression. In an effort to increase flower pigmentation, a transgene encoding chalcone synthase was introduced into Petunias. However, the plants developed variegated petals and endogenous chalcone synthase mRNA levels were reduced 50 fold¹. Therefore, the exogenous chalcone synthase gene was, in some way, affecting the expression of the endogenous gene in trans. Transgenic tobacco plants that express untranslatable sense or antisense forms of the gene encoding tobacco etch virus (TEV) coat protein were resistant to infection with TEV by a process that was mediated by RNA and was sequence specific². This was subsequently shown to involve degradation of the transgene RNA³. Parallel studies further enforced the existence of an RNA-mediated defence against viruses in plants when it was shown that tobacco plants that were transgenic for a potato virus X (PVX) replicon had marked strain-specific resistance to PVX⁴. Some investigators who were using antisense RNA to regulate gene expression noted that sense RNA also worked⁵. The breakthrough came when Fire and colleagues⁶ found that dsRNA was more efficient at silencing gene expression in Caenorhabditis elegans than either strand individually. Furthermore, injection of dsRNAs into adult worms led to gene silencing in the progeny (heritable silencing; Box 1) and as only a few molecules of injected dsRNA were required in each infected cell, Fire et al.⁶ proposed that the mechanism of RNAi involves a catalytic or amplification component (Box 1). It is now clear that RNAi, co-suppression and virus-induced gene silencing are triggered by dsRNA and operate by similar mechanisms⁷.

The mechanism of RNA interference

As the demonstration that RNAi was activated by dsRNA and the indication that RNAi might involve a derivative of the dsRNA⁶, studies in plants and in Drosophila have provided fundamental insight into its mechanism. Arabidopsis plants undergoing either transgene- or virus-induced gene silencing were found to contain 25 nucleotide RNAs that were complementary to the gene being silenced⁸. Sequence-specific gene silencing could be recapitulated after injection of dsRNA into Drosophila embryos⁹. Subsequent studies indicated that dsRNA promoted gene silencing in extracts of Drosophila embryos, that silencing correlated with the processing of the dsRNA to small 21–23 nucleotide-duplex-interfering RNAs (siRNAs)¹⁰ and that a dsRNA-specific endoribonuclease known as DICER was responsible for processing dsRNA to siRNAs^{7, 11}. In Drosophila embryo extracts, artificial 21- and 22-nucleotide duplex RNAs with overhanging 3' ends promoted the cleavage of target RNA^{12, 13}, and 21-nucleotide duplex RNAs triggered RNAi in mammalian cell lines such as embryonic kidney and HeLa cells¹². Furthermore, introduction of long dsRNAs (500 nucleotides) into some cell lines and undifferentiated embryonic stem cells resulted in the production of small RNAs and specific gene silencing^{14, 15}.

These studies have prompted a model for RNAi (Fig. 1) by which a long dsRNA trigger is first recognized and cleaved by DICER to siRNAs, which are subsequently incorporated into an RNA-induced silencing complex (RISC). This protein–RNA effector complex unwinds the duplex siRNAs and the antisense strand guides the RISC to homologous RNAs, which undergo endonucleolytic cleavage. Artificial siRNAs can also be incorporated into the RISC to induce degradation of target RNA¹² and this obviates the requirement for either the long dsRNA trigger or DICER. This has been important for the activation of RNAi in mammalian systems because long dsRNAs induce an interferon response, which promotes sequence-non-specific degradation of mRNA through the activation of RNAse L, as well as a global inhibition of mRNA translation following activation of protein kinase R¹⁶. Surprisingly, a recent study indicates that siRNAs, can induce global upregulation of expression of interferon-stimulated genes¹⁷. Although the interferon response did not affect the specificity of RNAi, this effect warrants consideration in in vivo use of siRNAs.

		Figure 1 |  Gene silencing by RNA interference.  Long double-stranded (dsRNA) is cleaved by DICER — a conserved RNase III protein — into RNA duplexes of 21–23 nucleotides that contain two unpaired nucleotides at the 3' end of each strand. This reaction has been documented mainly in lower eukaryotic systems. In higher eukaryotes, dsRNA promotes non-specific inhibition of gene expression through the activation of the interferon (IFN) response. Artificial small interfering RNAs (siRNAs) are converted into functional siRNAs that have 5' phosphate groups by a cellular kinase. These artificial siRNAs obviate the need for processing of siRNAs from long dsRNA, thereby, allowing activation of RNAi without concomitant activation of the IFN response. siRNAs assemble with cellular proteins to form an RNA-induced silencing complex (RISC), which contains a helicase that unwinds the duplex siRNA and a ribonuclease that cleaves target sequences. The RISC is directed by the antisense strand of the unwound siRNA, through sequence complementarity, to the target messenger RNA, which is subsequently cleaved. Micro-RNAs (miRNAs) and small temporal RNAs (stRNAs) are genomically encoded as 70 nucleotide stem-loop RNA precursors, which are then processed by DICER to single-stranded 21 nucleotide forms. The mature stRNA or miRNA binds to mismatched target sequences, and in this case, translational repression, rather than degradation of the target mRNA occurs. This process, which underscores a mechanism by which miRNAs regulate the expression of endogenous genes, has been comprehensively reviewed elsewhere63 (not shown).

Modulating HIV-1 replication by RNAi

HIV-1 is an obvious target for RNAi because it uses RNA intermediates at various points in its replication cycle (Fig. 2). The GENOMIC VIRAL RNA contained in the virion is the template for the viral complementary DNA, which integrates into the chromosomal DNA of the host cell. The integrated cDNA or PROVIRUS directs the production of nascent viral RNAs, which are templates for the synthesis of virus proteins that are required to perpetuate the infection (Fig. 2). Several studies have now shown that HIV-1 gene expression and replication in vitro can be inhibited by virus-specific artificial siRNAs^18-26 or expressed siRNAs^{18, 20} that are targeted to early or late phases of virus replication. After establishment of the provirus, the HIV-1 encoded protein Tat must be expressed before efficient virus transcription can occur. Therefore, targeting Tat by RNAi might also make it easier to target other viral transcripts by virtue of their lower abundance when Tat is compromised. CD4⁺ T cells²⁰ and macrophages²⁷, which are the natural targets for HIV-1 infection, are functional for RNAi. As RNAi operates in the cytoplasm²⁸, genomic viral RNAs that enter the cytoplasm after infection, and de novo synthesized genomic and subgenomic RNAs that are targeted to the membrane and to sites of protein translation might be degraded.

		Figure 2 |  RNA interference susceptible targets in the HIV-1 replication cycle.  After infection of a cell, a nucleoprotein reverse transcription complex containing genomic viral RNA associated with virion proteins is deposited in the cytoplasm. Among the virion proteins are enzymes that catalyse the synthesis of complementary DNA (reverse transcriptase; RT) and that catalyse the integration of the viral cDNA into genomic cellular DNA (integrase). Following reverse transcription, the viral cDNA in the complex is transported to the nucleus to form the integrated provirus. The virus-encoded protein Tat and its associated cellular partner, cyclin T, promote the expression of genomic and subgenomic viral transcripts, which are exported from the nucleus by the action of the virus protein Rev and its cellular co-factor, CRM1. Genomic viral RNAs are transported to sites of virus assembly for incorporation into progeny virions. Subgenomic RNAs act as templates for structural virion proteins and regulatory/accessory proteins that perpetuate the infection and have various roles in virus–host interplay. Intermediates in the replication of virus that can potentially be targeted by RNA interference (RNAi) are indicated by scissors. LTR, long terminal repeat; TSG101, tumour-suppressor gene 101.

The ability to target incoming genomic viral RNA by RNAi has implications in terms of therapeutic use of RNAi in infection with HIV-1, as it offers a mechanism to sterilize the cell from infection — that is, by preventing the formation of the integrated provirus. In addition, at this early stage of virus replication, the RNAi machinery only has to deal with two genomic viral RNAs. However, after a provirus is established, many thousands of viral transcripts are generated de novo in the infected cell and the degradation of these is a far greater task for the RNAi machinery. Studies with the non-segmented, negative-strand RNA virus respiratory syncytial (RSV) indicate that viral mRNAs are highly susceptible to siRNA-mediated silencing, but genomic and antigenomic viral RNAs that are encapsidated in the virus nucleocapsid protein are not²⁹. Several groups have shown that when CD4⁺ T cells were transfected with artificial siRNAs and subsequently infected with HIV-1, there was a degradation of genomic viral RNA^{20, 21} and a reduction in the synthesis of viral cDNA^{20, 22}. However, other investigators have not seen an effect of siRNAs on incoming genomic viral RNA^{24, 26}. The position of the siRNA viral recognition sequence in target mRNA has a marked impact on the efficiency of silencing³⁰. As discussed later, some regions of genomic viral RNA might be less accessible to siRNAs when the RNA is contained in the virus reverse transcription complex. As a result, some siRNAs might be more effective than others at inhibiting this early step in virus replication.

Cellular HIV-1 co-factors as targets for RNAi

Investigators have begun to explore ways to inhibit HIV-1 replication by targeting the transcripts of cellular co-factors on which HIV replication depends, including receptors/co-receptor molecules, transcription factors and assembly factors^{24, 31-33} (Fig. 2). In this regard, RNAi offers a powerful tool with which to determine the role of cellular co-factors in HIV-1 replication. In fact, in the first study to use RNAi in HIV-1 research, Garrus and colleagues³³ used siRNAs to silence the expression of tumour-suppressor gene 101 (TSG101) — a component of the class E vacuolar protein-sorting pathway. This indicated a crucial role for TSG101 in the budding of HIV-1 virions³³. Antiviral strategies aimed at suppressing the expression of cellular co-factors by RNAi need to be balanced with the negative effects that inhibiting essential cellular genes are likely to have on the host cell function. In most cases, the cellular co-factors of HIV-1 have important roles in processes such as transcription (for example, cyclin T and nuclear factor-B; NF-B), NUCLEOCYTOPLASMIC SHUTTLING (for example, CRM1) and immunoregulation (for example, CD4). The CC-chemokine receptor 5 (CCR5) is an exception. Individuals defective for CCR5 expression, because they have a homozygous 32-base-pair deletion in the CCR5 gene, show increased resistance to infection with HIV-1, yet are otherwise healthy³⁴. Therefore, CCR5 is a potential antiviral target, and small molecule inhibitors of CCR5 are being developed at present³⁵. RNAi-mediated suppression of CCR5 could form the basis of a new therapeutic strategy for infection with HIV-1.

Strategies for delivery and expression of siRNAs

Most of the studies carried out with HIV-1 so far have involved the use of artificial siRNAs, which can be introduced into cell lines and primary cells by various transfection techniques. Surprisingly, these siRNAs seem to be relatively stable and the silencing effect persists for many days after their introduction^{12, 20}. The design of a biologically active siRNA seems to be somewhat of a hit-or-miss venture. The basics are still being established in terms of the characteristics of siRNAs that produce the most effective silencing effect. mRNA sequences that are sensitive to siRNA-mediated silencing can be relatively insensitive when placed into a highly structured region of RNA³⁰. This indicates that mRNA structure could dictate the degree to which target RNA is accessible to the siRNA³⁶.

It is now apparent that siRNAs can silence gene expression in vivo. Artificial siRNAs specific for FAS (CD95), delivered intravenously under hydrostatic pressure, reduced CD95 mRNA levels in mouse hepatocytes, and protected mice from liver injury when challenged with concanavalin A or CD95-specific antibodies³⁷. Artificial siRNAs have also been shown to silence expression of co-introduced transgenes containing viral or cellular sequences in vivo^{38, 39}. Although siRNAs can be taken up by cells in some tissues in vivo, there is no evidence that siRNAs enter mammalian cells in vitro without transfection. The mechanism by which siRNAs enter cells in vivo is not known nor is it clear whether the uptake mechanism is general or is restricted to cells in particular tissues, for example, the liver, spleen and lungs. Although these initial in vivo studies with exogenous artificial siRNAs are encouraging, it will be important to evaluate whether siRNAs can enter lymphoid tissue and whether siRNAs can be modified to promote this uptake.

Several strategies have been described for the expression of siRNAs from plasmid or transgene templates (Fig. 3). A common approach is to place siRNA cassettes under the control of RNA polymerase III (Pol III) promoters such as U6 or H-1 promoters⁴⁰. The siRNA can then be expressed as short hairpin RNAs (shRNAs) or formed from sense and antisense strands that are expressed from different promoters (Fig. 3). To facilitate the introduction of siRNA-expressing cassettes into cells, investigators have used viral vectors. HIV-1-based lentivirus vectors have received particular attention in this regard. These vectors make use of the ability of HIV-1 to infect non-dividing cells^{39, 41, 42}. HIV-1-based lentivirus vectors retain this central characteristic and as such, are particularly suited for the transduction of non-dividing cells such as neurons and haematopoietic progenitor cells⁴³. Several studies have begun to use lentivirus-delivered siRNAs to silence gene expression in vitro and in vivo. siRNAs expressed from a lentivirus vector have been shown to diminish the expression of exogenous and endogenous genes in vitro and in the brain and liver of mice⁴⁴. CD8-specific shRNAs expressed from an HIV-1-based lentivirus vector could silence the expression of CD8 in vitro and in vivo⁴⁵, and lentivirus vectors that express shRNAs were shown to promote specific gene silencing in primary dendritic cells⁴⁶. One study has used lentivirus-derived siRNAs to generate mice that are impaired in the expression of specific genes. In this case, eggs from green fluorescent protein (GFP)-positive transgenic mice were transduced with a lentivirus expressing GFP-specific siRNAs and mice established from these eggs showed decreased expression of GFP⁴⁷. Together, these studies illustrate the broad use of RNAi for the silencing of viral and cellular processes in vitro and in vivo.

		Figure 3 |  Strategies for the expression of interfering RNAs.  Duplex, short hairpin RNAs (19–32 base pairs)64, 65 can be generated by the expression of target gene sequences in sense and antisense orientations (purple boxes) separated by non-complementary intervening sequences of 4–23 nucleotides (dashed lines). When used at concentrations that induce silencing, perfect stems of less than 30 base pairs do not efficiently activate an interferon response in mammalian cells. Hairpin RNAs are substrates for DICER, which generates small interfering RNAs (siRNAs). Placement of target sequences in tandem under individual promoters, either on the same or on different expression constructs, generate sense and antisense strands that associate to form siRNAs18, 66. These designs can also be incorporated into lentivirus and adenovirus vectors. For example, a self-inactivating (SIN) lentivirus vector has a deletion in the U3 region of the 3' long terminal repeat (LTR). As this region normally contains the virus enhancer and promoter, expression from the SIN vector depends on internal promoters, which reduces the possibility of promoter insertion after integration of lentivirus. (Psi, RNA encapsidation signal). A minimal adenovirus vector that expresses short hairpin siRNAs from an internal promoter is shown. LITR, left inverted terminal repeat sequence; RITR, right inverted terminal repeat sequence.

Factors affecting HIV-1 susceptibility to RNA

Gene silencing by RNAi is highly sequence specific. Almost perfect complementarity is required between the siRNA and its target sequence. Unfortunately, HIV-1 is a moving target. Because of the infidelity of reverse transcription, new sequence variants are continually being generated in infected individuals. This allows the virus to evade inhibition by antiretroviral agents or surveillance by the host's immune system⁴⁸. As nucleotide mismatches between the siRNA and its target can abrogate silencing, it is expected that HIV-1 will rapidly evolve mutations that confer resistance to individual siRNAs. Escape from siRNA pressure has already been shown for poliovirus in vitro⁴⁹. Some regions of the HIV-1 genome, such as regions involved in binding transcription factors or transacting regulatory proteins, are highly conserved because mutations in these areas can markedly impair the fitness of the virus. Therefore, escape variants might be less probable when siRNAs are targeted to these regions. It might also be possible to express many siRNAs that target different regions of the virus genome. To escape, the virus would have to simultaneously acquire mutations in each of the target sequences. One caveat to this approach is that CD4-specific siRNAs were found to compete with CD8-specific siRNAs for silencing in T cells⁵⁰. Therefore, the RNAi machinery might be saturable and this could limit the use of many siRNAs in individual cells.

Although it is evident that almost any RNA can potentially be targeted by RNAi, the siRNAs must be able to locate the RISC on the target RNA to affect silencing. However, it is possible that in some cells, such as macrophages and dendritic cells, HIV-1 might persist in a form that is inaccessible to the RNAi machinery (Fig. 4). In infected macrophages, a marked amount of virus assembly occurs in the cytoplasmic membranes of MHC class II vesicles in which virions accumulate^{51, 52}. This is in contrast to the conventional situation in T cells where virus budding occurs at the plasma membrane. In macrophages, these intracellular virions seem to be stable and retain their infectivity for months (N. Sharova, unpublished observations). Furthermore, dendritic cells express C-type lectin receptors that can bind and internalize extracellular virions⁵³, which retain their infectivity after internalization⁵⁴ (Fig. 4). FOLLICULAR DENDRITIC CELLS might also retain virus particles in an infectious form for many months in vivo⁵⁵. The overall contribution of these reservoirs to persistence of the virus in infected individuals remains to be determined. Nonetheless, genomic viral RNA that is encapsidated in virus particles in these intracellular compartments is likely to be inaccessible to the RNAi machinery. Therefore, these compartments deserve consideration in any future RNAi strategies that involve systemic delivery of artificial siRNAs.

		Figure 4 |  Factors affecting susceptibility of HIV-1 to RNA interference.  As cleavage of the messenger RNA target requires a high degree of complementarity between the small interfering RNA (siRNA) and its target sequence, heterogeneity in the virus population might prevent efficient silencing of some virus variants by specific siRNAs. In addition, in the reverse transcription (RT) complex, some regions of genomic viral RNA that are tightly associated with virion proteins in a ribonucleoprotein (RNP) complex might be less accessible to siRNAs. In macrophages, viruses can also assemble into cytoplasmic MHC class II compartments52. In addition, in dendritic cells (DCs), extracellular virions can be captured and internalized into vesicles. Virions in MHC II compartments and endocytic vesicles might retain infectivity for long periods and might be inaccessible to the RNA interference (RNAi) machinery. Presumably, infection of macrophages might still be halted if genomic and subgenomic RNAs are silenced before establishment of intracellular virion compartments.

Conclusions

RNAi promises to have an important impact on AIDS research. Irrespective of its potential therapeutic applications, the ability to silence cellular genes offers the opportunity to validate putative cellular co-factors of HIV-1 replication. RNAi will therefore help to further our understanding of the interplay between HIV and its host, and indicate new cellular targets that could, in their own right, be new drug targets. Although the therapeutic application of RNAi in infection with HIV-1 faces many obstacles, these obstacles are no different to those faced by investigators developing genetic therapies for other diseases, such as metabolic disorders or cancer. Steady advances are being made in the design of vectors for the transduction of haematopoietic progenitor cells. The chemistry of artificial nucleic acids (including siRNAs) is being modified to improve their stability in body fluids and their uptake into cells. By virtue of its unique mechanism of action, an RNAi-based strategy would complement existing therapeutic strategies for AIDS. Because they target the RNA and not the protein, HIV-1-specific siRNAs would be equally effective against drug-resistant HIV-1 variants. It will be exciting to see how things develop over the next couple of years.

Boxes

		Box 1 | Aspects of RNAi not observed in higher eukaryotic systems
		There are several features of RNA interference (RNAi) that allow amplification and spreading of the silencing effect beyond the cell or tissue in which the initial triggering event occurred. These observations have been made in lower eukaryotic systems but have not, as yet, been recreated in more complex systems. Systemic RNAi: observed in plants56 and in worms6, the silencing effect spreads throughout the whole organism. This requires that some component of the RNAi machinery is passed from cell to cell and that there is a mechanism for amplifying the silencing signal. In Caenorhabditis elegans, the transmembrane protein systemic RNAi defective 1 (SID1), which is required for systemic RNAi57, might function as a channel for double-stranded RNA (dsRNA), small interfering RNAs (siRNAs) or an unidentified RNAi signal. RNAi amplification: the presence of an amplification mechanism is supported by the observation that only a few molecules of dsRNA per cell can silence a large excess of target messenger RNA molecules6, 9. The mechanism for amplification is not well understood, but studies in various organisms implicate a role for an RNA-directed RNA polymerase (RdRp), which is primed for RNA synthesis by the siRNAs58-61 and which amplifies the dsRNA trigger and the response. However, inhibiting RNA synthesis in mouse oocytes did not affect dsRNA-mediated silencing of an endogenous gene, arguing against a role for RdRp dsRNA-mediated mRNA degradation in mammalian cells62. It seems that mammalian cells lack RNAi amplification mechanisms and as such, do not sustain the silencing effect through many cell divisions. Heritable RNAi: the heritable nature of RNAi became evident when Fire and colleagues6 observed that injection of dsRNA into worms resulted in a silenced state in the progeny. The underlying mechanism has not yet been identified.