Abstract
Ribonucleic acid (RNA) interference (RNAi) is a recently discovered phenomenon whereby the introduction of double stranded (ds) RNA into the cytoplasm of the cell results in the specific and efficient degradation of complementary messenger (m) RNA and, therefore, reduced protein production. It was discovered by chance during attempts to develop flowers with increased colour intensity. The specific nature of the inhibition of protein production of cells has resulted in an explosion of research to understand and exploit RNAi. The technique is now established in in vitro systems, and much work is focussed in adapting RNAi for in vivo application. The potential of the technology in understanding physiological and pathological processes is significant, while its development as a therapeutic agent holds much promise as targeted agents. This review will describe the basic biological processes that drive RNAi, indicate current areas of areas research, and forecast future areas of development.
Keywords
Introduction
Ribonucleic acid (RNA) interference (RNAi) is a naturally occurring intracellular mechanism, which causes sequence specific posttranscriptional gene silencing. The reaction is triggered by the introduction of double-stranded (ds) RNA into the cytoplasm of the cell, and results in the specific targeted destruction of mRNA and a subsequent reduction in protein production (Elbashir et al., 2001). Its initial discovery (Napoli et al., 1990) was followed by an explosion of research in this field, and by 2002 it had earned the title of “Scientific Breakthrough of the Year” (Couzin, 2002).
When successfully manipulated, RNAi can result in the knockdown of single or multiple genes so providing a quick and convenient method of analysing gene function (Dykxhoorn et al., 2003). RNAi libraries have been developed as a useful screening method for assessing functional consequences of inhibiting multiple proteins within given pathways (Devi, 2006). Examples of experimental use include identification of key steps in the p53 pathway (Berns et al., 2004) and verification of the role of Fas in apoptosis regulation in murine liver and kidney (Song et al., 2003; Hamar et al., 2004). In the pharmaceutical industry, investigation into its potential for use in pharmaceutical target validation and as a therapeutic tool is ongoing (Lavery and King, 2003; Lu et al., 2003).
As a model therapeutic agent, RNAi has achieved regression of clinical signs in neurodegenerative disease models (Xia et al., 2004) and is currently in phase 1 clinical trials to treat age related macular degeneration (Alnylam Pharmaceutical, Acuity Pharma). It also has potential for use as an antiviral therapeutic with examples of targeted viruses in animal disease models including Herpes simplex virus-2, respiratory syncytial virus, parainfluenza virus, SARS, and influenza A (Cristofaro and Ramratnam, 2006).
In addition, RNAi can be used to generate animal disease models, e.g., Parkinson’s disease has been modelled through targeted knockdown of Th within neurons in the mid brain of adult mice. (Hommel et al., 2003). Finally, germline transmission of RNAi in mice has also been achieved (Carmell et al., 2003). This review aims to summarize the current opinion on the mechanisms underlying RNA interference, discuss key in vitro and in vivo existing applications and outline its prospects for future use within the pharmaceutical industry.
The Discovery of RNA Interference
First recognized by Rich Jorgenson and his team in plants (Napoli et al., 1990), the discovery of RNAi was the unexpected result of attempts to make the colour of petunia blooms more purple. An experiment was designed to overexpress the gene responsible for the production of anthocyanin pigments (chalcone synthetase (CHS)) through the introduction of a CHS transgene attached to a powerful promoter sequence. However, instead of increasing anthocyanin levels and therefore the deepness of bloom hue, the transgenic model showed apparent silencing of both the endogenous and exogenous CHS genes, and the flowers appeared variegated or white.
The mechanism was recognized as a form of posttranscriptional gene silencing (PTGS) and termed “co-suppression.” Further work found that transcripts produced from both loci were immediately degraded in the cytoplasm. In this case, activation of PTGS was thought to be due to the production of aberrant dsRNA by the transgene, which resulted in silencing of the mRNA (Zamore, 2002).
A few years later, Fire and his colleagues were investigating the efficiency of injecting single-stranded antisense RNA as a method of gene silencing in the nematode Caenorhabditis elegans by using its ability to hybridise with endogenous mRNA and inhibit translation (Fire et al., 1998). The discovery that introduction of the sense strand also resulted in silencing, and that dsRNA was even more effective at inhibiting the target gene led to the conclusion that both original single-stranded sense and antisense samples may have been initially contaminated with dsRNA (Fire et al., 1998). Similarities between the underlying mechanism in nematodes and co-suppression in plants was soon recognized, and the term RNAi was adopted for both (Ketting and Plasterk, 2000).
The Mechanism of RNA Interference
RNA interference takes place predominantly within the cytoplasm of the cell and is triggered by the introduction of a double-stranded oligonucleotide into the cell cytoplasm (Figure 1). The mechanism is mediated by the activation of 2 major molecules; the initial activity of the endonuclease Dicer (an RNAse III family enzyme), followed by the activity of the RNA interfering silencing complex (RISC) (Chiu and Rana, 2002). An adenosine triphosphate (ATP) dependent reaction involving the endonuclease Dicer is responsible for cleaving the long ds nucleotide into short interfering (si) double-stranded RNAs, 21–23 nucleotides (nt) in length. RISC then unwinds the double-stranded siRNA, using a helicase, and subsequently binds to the free antisense strand.
This complex is able to identify the specific complementary strand of mRNA and degrades it with the help of one of its major components, Argonaute 2 protein (Liu et al., 2004; Song et al., 2004). The result is destruction of the mRNA that is complementary to the antisense strand of the original dsRNA introduced into the cytoplasm, and prevention of translation and protein production.
Initial experiments successfully manipulated RNAi in plants and invertebrates through the introduction of long-stranded dsRNA into the cytoplasm. In mammalian cells, however, similar techniques resulted in the initiation of the interferon response and cell death before cleaving by Dicer could occur. It was not until 2001 that Elbashir et al. reported an alternative method of RNAi induction in mammalian cells. Through the direct introduction of siRNAs under 30 base pairs in length, they successfully avoided the interferon response and activated the RISC complex and mRNA destruction in mammalian cells.
How the introduction of just a few strands of RNA results in the silencing of a large excess of target mRNA has also been investigated (Zamore, 2002). One explanation suggested the involvement of a family of RNA-dependent RNA polymerases (RdRp), which use the target mRNA as a template, and cleaved primary siRNAs as primers. These produce a population of secondary siRNAs, which lead to an increase in the number of activated RISC complexes formed and mRNA splicing. Members of the RdRp family have been identified in C. elegans, Arabidopsis, Neurospora, and Dictyostelium; but not Drosophila and humans. Other suggested mechanisms include the simple fact that Dicer cuts the dsRNA into shorter lengths, thereby increasing the number of siRNAs 10–20-fold, and that RISC itself acts as a multiple turnover enzyme and uses a catalytic mechanism to recycle already synthesised siRNAs (Plasterk, 2002). The latter two would be feasible explanations inDrosophila and mammals, which lack RdRp and thus replicative mechanisms for RNAi (Harborth et al., 2003).
The natural role of RNA interference is thought to be the protection of the plant/nematode from invasion by viral pathogens (Zamore, 2002). On infection, RNA viruses generate double-stranded (ds) RNA molecules either in activation or replication, and it is these molecules that are capable of activating the host RNAi defense mechanism. This results in the specific degradation of the viral RNA so preventing viral multiplication (Voinnet, 2005). This phenomena is highlighted by plants defective in RNAi demonstrating hypersensitivity to viral infection (Ding et al., 2004).
Some animal and plant viruses, e.g., Potato virus X, are also found to produce proteins capable of suppressing host-mediated RNA silencing. These proteins have also been shown to be responsible for successful viral spread within the host (Li et al., 2002; Ding et al., 2004).
Following this identification of the RNAi phenomenon in plants, invertebrates and mammalian cells, further investigations into its specific properties were carried out (Downward, 2004). Experiments to determine its potential for its systemic spread were performed using nematodes. These were fed, soaked, or injected with a dsRNA rich solution and in all cases this resulted in body-wide RNAi-induced gene silencing. Although the process behind this mechanism was not clear, a transmembrane protein Sid-1 that enables passive cellular uptake of dsRNA was identified that could play an active part in RNAi distribution (Feinberg et al., 2003).
RNAi silencing effects have also been observed in progeny in plants and invertebrates but again the underlying mechanism remains undetermined (Napoli et al., 1990; Fire et al., 1998). As a result of these numerous experiments, artificial manipulation of RNAi in cultured mammalian cells has been possible and its full potential for use in vivo hypothesized. It prompted the need for mass production of synthetic siRNAs, which is now the main business of several companies.
SiRNAs and Their Design
Effective design of synthetic siRNAs relied on a detailed investigation of the characteristics of naturally occurring siRNA molecules, to ensure effective uptake by RISC and specific silencing of the targeted gene. Biochemical characterization of nautrally occurring siRNAs found them to be 21–23 nt dsRNA duplexes with symmetric 2–3 nt 3′ overhangs and 5′-phosphate and hydroxyl groups (Dykxhoorn et al., 2003). The 5′ phosphorylation was shown to be essential for RISC incorporation of the antisense strand illustrated by its inhibition resulting in a marked reduction in RNAi activity (Czauderna et al., 2003).
This work led to the development of a set of empirical guidelines to generate two 21-nt sense and antisense oligoribonucleotides for efficient siRNA (Mittal, 2004; Naito et al., 2004). Success is maximized by achieving looser binding of the 5′ end of the antisense strand to its complementary strand in order to promote its binding to RISC, and increasing the specificity of the siRNA sequence to reduce off-target effects (Schwarz et al., 2003). In addition, it is recommended that: each strand must have 2-nt 3′ overhangs; A/U base pairing at the 5′ end of the sense strand and G/C base pairing at the 5′ end of the sense strand; AU richness in the 5′ terminal third of the antisense strand; avoidance of introns, 5′ and 3′ untranslated regions, regions within 75 bases of the start codon and sequences with >50% Guanine and Cysteine content; maximizing sequence divergence from related mRNA.
Despite applying these criteria, differences in silencing efficiencies between siRNAs occur, and occurrence of off-target siRNA activity has been recognized as an important factor in siRNA experimental design (Reynolds et al., 2004).
Off-Target Effects of RNA Interference
Off-target activity is defined as the nonspecific silencing of genes other than those for which the sequence of siRNA has been manufactured (Reynolds et al., 2004). Despite the strict design criteria already mentioned, there is now evidence that up to 75% of siRNAs can tolerate mismatches and induce off-target activity while guiding mRNA cleavage (Snove et al., 2004). Proposed causes include cross- hybridization of the siRNA to transcripts of similar sequences, incorporation of the sense rather than the antisense strand by RISC, and the ability of the siRNA to inhibit mRNA translation by binding to the mRNA rather than destroying it (Dyxhoorn et al., 2005). The latter resembles the action of microRNAs—a type of siRNA more prone to mismatches.
Some studies have shown that as few as 11 contiguous matches between siRNA and the off-target mRNA can result in silencing of protein production (Jackson et al., 2003). In one experiment 16 siRNAs were designed to target the same specific coding region and the expression profiles of each siRNA, then compared. As well as silencing the target gene, it was found that each siRNA showed specific, repeatable, off-target gene silencing with only a small number of gene regulations in common, i.e., the off-target effects were specific to the siRNA, not the target (Jackson et al., 2003). Further work with siRNA targeting luciferase (an exogenous gene) in vivo showed repeatable regulation of off-target endogenous gene expression that also showed a dose-dependent response (Jackson et al., 2003).
Some domains of siRNAs appear to be more susceptible to off-target effects than others, and attempts have been made to identify specific patterns that might predict this (Jackson et al., 2003; Du et al., 2005). In the latter study, both the position of the mismatched base pair and the nucleotides involved were shown to influence silencing ability. Substitution of nucleotides at the 3′ and 5′ termini still resulted in knockdown of the off-target gene, whereas centrally located mismatches were poorly tolerated and gene silencing was abolished. Nucleotides involved in base pairing wobble were Guanine:Uracil (G:U) and Adenine:Cytosine (A:C) (instead of G:C and A:T respectively). These mismatches were well tolerated and resulted in efficient off- target silencing (Du et al., 2005).
To reduce the likelihood of off-target effects and identify more unique oligonucleotides, it is suggested to use more sensitive algorithms than the BLAST database, e.g., Smith and Waterman or siDirect (Naito et al., 2004; Reynolds et al., 2004; Snove et al., 2004). The design of such siRNAs has recently been described as one of the “hottest topics in molecular biology” (Yamada et al., 2004).
Other Forms of RNA Interference
In addition to naturally occurring and manufactured siR-NAs, there have been recent publications of alternative forms of both, which are listed here:
Micro (mi)–RNAs
These are an abundant class of short (19–25 nt) single-stranded RNAs that are expressed in all higher eukaryotes (Cullen, 2004). They are encoded in the host genome and are processed by Dicer from 70nt hairpin precursors (Novina and Sharp, 2004). Their function is to regulate endogenous gene expression during development by interfering with mRNA expression through imprecise base pairing and inhibition of protein translation (Carrington et al., 2003). Recent work has identified their specific roles in the regulation of early haematopoiesis and lineage commitment (Chen et al., 2004). They are sometimes referred to as small temporal RNAs as a reflection of their importance in the regulation of developmental timing (Medema, 2004).
Piwi-Interacting (pi) RNAs
These are single-stranded 25–31 nt RNAs which have recently been detected in mouse, rat, and human testes. They have been shown to associate with Piwi protein (a subclass of Argonaute proteins) and the human RecQ1 protein to form a Piwi-interacting RNA complex (piRC). These complexes are thought to regulate the genome within developing sperm cells (Carthew, 2006).
Short-Hairpin (sh) RNAs
These are synthetically manufactured and their design is based on naturally occurring miRNAs. They can be expressed by plasmid vectors or endogenously inserted into the genome. The vectors contain an inverted repeat of 19–29 nucleotides with the desired sequence separated by a short loop of 6–9 base pairs. When the nascent RNA is synthesised, it will form a stem loop that can then be processed by DICER. These vectors commonly employ pol III promoters, normally used by the cell to drive expression of short RNAs, e.g., transferRNAs (Medema, 2004).
Small Modulatory (sm)RNAs
These are short, double-stranded RNAs which are found in the nucleus of neural stem cells of mice. They play a critical role in mediating neuronal differentiation through dsRNA/protein interaction (Kuwabara et al., 2004).
Making Rnai Work —Successful Transfection
Whether in vitro or in vivo, the RNA interference mechanism can only be initiated once the siRNA has been transported into the cytoplasm of the cell. A successful transfection requires that the siRNA molecule, which carries a net negative charge under normal physiological conditions, must come into contact with and cross a cell membrane that also carries a net negative charge. Several methods of siRNA transfection have been developed for use in vitro and in vivo.
In vitro Transfection Mechanisms and Protocols
Naked Delivery
The challenge of a successful transfection in vitro is to achieve intracytoplasmic delivery of siRNA to as many cells as possible while at the same time keeping the toxicity associated with transfection or the transfection reagent to a minimum. Optimization of both the method and conditions of transfection is essential for each experiment and is dependent on the cell line and siRNA being used.
The most basic approach involves the addition of naked siRNA directly to the cells without the use of a transfection reagent.
This was attempted in neurons by Lingor et al. (2004). The siRNA was shown to enter the endosomal compartment of the cells when left in suspension (identified with fluorescent tagging) but its release into the cytoplasm while still fully functional was not demonstrated. Therefore, to ensure functional delivery, the siRNA must either be forced into the cells under high pressure (microinjection) or enter via micropores formed by affecting the charge of the cell membrane with electric pulses (electroporation) (Ambion technotes). High levels of cytotoxicity associated with these methods means that they are generally reserved for the more difficult to transfect cell lines e.g., neurons (McManus et al., 2002; Ambion technotes, undated).
Chemical Transfection
Chemical transfection reagents are the most commonly used siRNA transfection vehicle, and of these cationic liposomes (CL) are the most popular. There are currently more than 30 different commercial varieties of CL formulations that have been developed specifically for siRNA delivery, e.g., Lipofectamine 2000 (Dass, 2004). They operate by surrounding the siRNA with a positively charged lipophylic shield, which is able to fuse with the cell membrane and transport the siRNA into the cytoplasm of the cell contained within endosomes. The precise mechanism governing the timing of their functional release from the endosomal compartment is undetermined and will affect the timing of silencing onset.
Peak mRNA destruction is said to occur within 24–48 hours transfection and is transient with an average duration of 2–3 days (Ambion technotes, undated). The exact length of silencing will depend on the growth rate of the cell line, the level of gene expression and the half-life of the protein being knocked down. The development of other transfection reagents is ongoing. These include: Membrane permeant peptides, which are short amphipathic peptides that translocate across lipid bilayers in an energy-independent manner; Atecollagen, a highly purified, pepsin-treated type 1 collagen from calf dermis which is positively charged; and organic-inorganic hybrid nanoparticles which are poly(ethylene glycol)—block—poly (aspartic acid) with calcium phosphate. All of these have been used to successfully deliver siRNA in vitro (Kakizawa et al., 2004; Minakuchi et al., 2004; Muratovska and Eccles, 2004).
There are a number of variables which are known to affect the transfection efficiency and these need to be optimized for each particular cell line. Cell density, concentrations of transfection reagent and siRNA, the time they are left together before adding to the well and the presence of antibiotics or serum in the cell medium will all influence the transfection rate. For example, when using Lipofectamine 2000 (a CL) a cell confluency of 30–50% is recommended and the Lipofectamine and diluted siRNA must be allowed to combine for 30 minutes to achieve optimal transfection efficiency (Dalby et al., 2004).
Vector-Mediated Delivery—Overcoming Transient Silencing
DNA-vector mediated, and virus-vector mediated mechanisms have been developed to overcome the limitation of transient silencing (Dyxhoorn et al., 2003). The siRNA-expressing vector works as a platform to produce a large amount of siRNA for a relatively long period. There are two main types of DNA-vector based RNAi systems both of which involve the use of Pol III promotors (U6 or H1) (Dykxhoorn et al., 2003). In the first, the sense and antisense strands of siRNA are expressed from different cassettes aligned in tandem in the same construct i.e., tandem type. These then bind and form a short double stranded (si) RNA, which is presented directly to RISC and activates the RNAi pathway.
In the second, the sense and antisense strands are expressed as a connected ribonucleic acid with several intermediate bases which form a stem loop structure (short hairpin (sh) RNA). This is then presented to Dicer for further cleavage to siRNAs and RISC presentation (Kobayashi et al., 2004). One of the first to describe such a plasmid was Brummelkamp et al. (2002). This particular vector was named pSUPER and it used an RNA polymerase III promotor to express short dsRNA in the form of an inverted repeat sequence containing a hairpin loop. The mammalian cells were transfected using electroporation methods and the targeted protein levels were suppressed for 5–7 days after 10 rounds of cell division. Comparison between these two methods has shown increased efficiency of silencing using shRNA production instead of the tandem type (Hutvagner et al., 2002).
Retrovirus vectors have also been developed, and these use either oncoretrovirus or lentivirus vectors (Dyxhoorn et al., 2003). Lentivirus vectors have the added advantage of being able to infect both actively dividing and non-dividing cells. They are also capable of generating transgenic animals as they are resistant to proviral silencing during development (Dyxhoorn et al., 2003).
In vivo Transfection
Once the specificity and efficiency of the siRNA molecule has been validated in vitro, the development of an in vivo application is possible. Unfortunately, due to the susceptibility of siRNAs to nuclease degradation in serum, and the difficulty in finding a delivery method that achieves effective systemic distribution to the cell cytoplasm in vivo, this is the area that has proved most elusive.
Naked siRNA
Initial studies were conducted using naked siRNA, which has the advantage of excluding toxicities associated with transfection reagents. Rotes of administration considered included intravascular, intratumoural, intracranial, intraperitoneal, intrasplenic, intramuscular, subretinal, subcutaneous, mucosal, topical application and oral ingestion (Templeton, 2002). Initial fears regarding the safety of introducing naked siRNAs in mice, with regard to their potential for inducing immune responses were investigated by Heidel et al. (2004). By measuring levels of plasma interleukins and interferons following intraperitoneal and intravenous injection of siRNAs, an immune response was not detected and systemic siRNAs were shown to be well tolerated by mice. To date, the most commonly reported technique of successful administration of naked siRNA in vivo has been the use of intravenous “hydrodynamic delivery.”
Hydrodynamic Delivery of siRNA
This introduces the nucleic acid via the intravascular route at high speed and volume, resulting in greater levels of transfection with a more diffuse distribution. Described sites of intravenous delivery include intraportal (Budker et al., 1998; Song et al., 2000) intra-saphenous (Hagstrom et al., 2004) and the tail vein (Zhang et al., 1999). Modification of this technique has been successful in the delivery of naked siRNAs silencing both exogenous (Lewis et al., 2002; McCaffrey et al., 2002) and endogenous (Song et al., 2003) genes in multiple organs in mice. The efficiency of intracellular delivery has been shown to be directly dependent on the volume of fluid given and the speed of injection (Liu et al., 1999). In mice this has been optimised at 1ml/10g bodyweight injected over a period of 3–5 seconds (Hodges et al., 2003).
The exact mechanism behind the transfection of the gene/siRNA into the cells following hydrodynamic delivery is not fully understood. When delivered via the tail vein, it is suggested that as the injection rate exceeds cardiac output so the introduced fluid accumulates in the superior vena cava. This is then forced out into vessels within organs and subsequently through fenestrae in these vessels into extravascular spaces. The naked siRNA is then brought into contact with the cells of the organ before it is mixed with blood so reducing the chance of nuclease degradation (Hodges and Scheule, 2003). Zhang et al. (2004) further argued that actual transfection of the hepatocyte by this method is the result of “hydroporation.”
Membrane defects (pores) are generated following the high-pressure delivery and provide a route for introduction of the siRNA into the cell. Injection of a membrane permeability marker (Evans blue) showed that these pores closed up within 10 minutes, by which time the siRNA had entered the cytoplasm and the RNAi mechanism would be underway. Data regarding exact distribution of siRNA molecules following tail vein injection is scant, although initial experiments show a majority of the siRNA distribution is to the liver (Zhang et al., 2004). This can be explained by its proximity to the vena cava and ability to accommodate extra fluid.
Following hydrodynamic injection in mice, various side effects have been recorded. The heart rate decreases from 510 beats per minute to 280 and there are transient irregular rhythms lasting for up to 60 seconds (Liu et al., 1999). A concurrent increase in QRS complex size also suggests increased cardiac chamber size. The liver expands and turns whitish soon after injection (Zhang et al., 2004). Clinical side effects included apnoea directly after injection, although gentle massage of the abdomen is said to resolve this (Hodges et al., 2003). This is accompanied by a transient, right–sided, congestive cardiac failure. Histopathological examination showed single cell liver necrosis in half the mice after one day (Liu et al., 1999). The body weight remained elevated for 30 minutes but was back to normal after 2.5 hours due to urination (Liu et al., 1999).
Despite all these effects, the survival rate was reported to be 100%, and the method is being widely used for in vivo siRNA delivery in mice. Development of this method of delivery is in early stages. Ethical and animal welfare implications must also be considered as many would regard the intravenous injection of such high volumes of fluids over limited time periods unacceptable. Given the significant side effects, it is highly unlikely that this technique would ever be of use in humans.
The use of naked siRNA has been further advanced by the use of stabilization modifications. Unmodified siRNA molecules are susceptible to serum nuclease degradation, and without effective modification they will be immediately degraded on in vivo administration. Czauderna et al. (2003) identified a successful stabilizing 2′-O-methyl modification, which managed not to compromise the efficacy of the siRNA molecule. Layzer et al. (2004) used stability modified 2′–fluoro pyrimidines siRNAs molecules and found over 50% to be intact after 24 hours of suspension in plasma. This compared with complete degradation of normal molecules within 4 hours. The modifications did not affect their silencing capabilities in vitro or in vivo and acted to increase their resistance to nuclease degradation in plasma.
Chemical Modifications Enhancing Transfection in vivo
Transfection reagents are also a widely experimented technique for in vivo delivery of siRNA. Similar compounds developed for in vitro use have been trialled in vivo. The use of cationic liposomes has been problematic primarily due to their poor systemic distribution and toxicity. Preferential deposition of complexes in the lungs and liver, or areas containing proliferating vasculature (e.g., tumors) has been found to occur following intravenous administration.
Their potential toxicity is highlighted in a lengthy review by Dass (2004) and includes the induction of acute inflammation following intraocular, intra-articular, and intra-tracheal administration and emboli formation from liposome/siRNA complexes and agglutination of red blood cells in response to their net positive charge. Transient acute toxicities of leukopenia, thrombocytopenia and increases in alanine transaminase, have also been reported (Tousignant et al., 2000).
The development of second-generation molecules has helped, increasing stability in serum, improving target organ distribution and penetration, and helping the formation of a homogenous population of complexes with siRNA before injection (Sioud and Sorenson, 2003). siRNA molecules have also been stabilised for in vivo administration with cholesterol conjugation. This has been shown to increase resistance to degradation in vivo and to have good tissue distribution when injected via the tail vein after 24 hours. Using this technique, SiRNA has been detected in the liver, heart, kidney, lung, and adipose tissue (Soutschek et al., 2004). This method was shown to be successful in silencing an endogenous gene (apoB protein) in the liver and jejunum.
Plasmid vectors have also been used for in vivo siRNA delivery (Brummelkamp et al., 2002). However, the relative ease of plasmid regeneration did not compensate for general poor transfection efficiency of the plasmid based vectors, and prompted the development of viral-/retroviral-based vectors for shRNA delivery. Vector systems that are based on the adenovirus are limited in that they do not integrate into the host genome (Tomar et al., 2003). In contrast integration of the shRNA expression cassette into the host genome can occur with lentiviral based vector systems, e.g., Moloney murine leukaemia virus, the Murine stem cell virus, and HIV-1 (Mittal, 2004). As transgenes expressed from these viruses are not silenced during development, these vector systems can also be used to generate transgenic animals through the delivery of shRNAs to embryonic stem cells of embryos (Rubinson et al., 2003).
Applications of RNA Interference in Pathology
The full potential of RNAi is yet to be recognized, but there are clear areas for which there are uses in both experimental and toxicological pathology.
RNAi in Pharmaceutical Target Validation and Toxicology
In the pharmaceutical industry, modern scientific techniques have greatly expanded the pool of potential drug targets for the treatment of disease. Unfortunately, the success rate of new drugs is declining. Two significant causes of failure are unacceptable toxicity, and failure of the mechanism to provide significant clinical benefit. Both of these sources of attrition occur at a late stage in development and are costly both financially and in terms of resources expended. SiRNA has the potential to provide an early, specific, and relatively inexpensive method for studying a pharmacological mechanism in both preclinical models of efficacy and toxicity. This could result in reduction in mechanisms with efficacy and safety issues being progressed into late stages, diverting resource into more productive targets. In particular, the use of siRNA in toxicity screening has the potential to reduce the number of animals used in toxicity testing.
RNAi in Mechanistic Pathology
The understanding of pathogenesis of lesions frequently relies on the testing of generated hypotheses in vivo. This usually includes the exaggerated action or inhibitions of cellular pathways via the over expression or inhibition of cellular proteins. RNAi offers the potential for rapid development of the specific inhibition of protein expression in vivo to either mimic pathological findings or to investigate the toxicity of administered compounds in animals devoid of pathways suspected of initiating pathological findings.
RNAi in Animal Models of Disease
Human diseases may be modelled in animals by the inhibition of key regulatory proteins. RNAi has the potential to produce rapid and cost effective models of human disease, especially for proteins whose inhibition in the embryo has significant effect on viability, precluding the use of gene knockout technology. This potential has been demonstrated by the recent use of adenovirus to introduce siRNA specifically targeting tyrosine hydroxylase mRNA within neurones of the mid brain. This is a key enzyme in the production of dopamine, a molecule that is involved in regulating food intake, addiction, and movement control (Hommel et al., 2003). These animals are a model of human Parkinson’s disease.
RNAi as a Therapeutic Agent
Probably of most commercial interest in the use of RNAi as a therapeutic agent. While there are many technical hurdles to be overcome before the technology would be of application to man, there has been much research in animals into the potential of RNAi as a therapeutic. Initially, the prevention of liver disease was attempted by targeting genes linked to apoptosis control in the liver in two models of autoimmune hepatitis. SiRNA was directed against caspase (Zender et al., 2003) and Fas (Song et al., 2003) and was delivered hydrodynamically via the tail and portal vein respectively.
In both cases this was successful in reducing hepatocyte necrosis and inflammation, and protected the mice from future chronic fibrosis. In another experiment siRNA targeting Fas in the kidney was used to reduce ischaemia- reperfusion injury. Hydrodynamic and normal volume intrarenal delivery both successfully reduced Fas protein expression 4-fold. In addition, the pathology of the renal tissue following the ishaemic insult was reported to be much improved compared to the control (Hamar et al., 2004).
Following this success one of the first examples of using RNAi to inhibit viruses in vivo was then attempted. Immuno-compromised and immunocompetant mice were hydrodynamically injected with plasmids expressing hepatitis B virus (HBV) and shRNAs targeting HBV. This resulted in a 99% reduction in HBV detection using antibodies to detect HBV core antigen by immunohistochemical methods, and it was suggested that RNAi could be used to treat viral disease in the future (McCaffrey et al., 2003).
Since then, several experiments using RNA interference to target respiratory viruses have been attempted. Initially Influenza virus was chosen due to its significant public health issues and lack of a wholly effective vaccine (Tompkins et al., 2004). Proteins were targeted that are highly conserved across several sub types of influenza and which are essential for viral replication. It was found that combined iv hydrodynamic and intranasal (in a lipid carrier) delivery was most effective at specifically inhibiting virus replication at the site of infection. It also reduced lung virus titres in infected animals and protected animals against lethal challenge. Another experiment using slow iv delivery of siRNA complexed with a polycation carrier, and its delivery using DNA vectors iv/intra nasally, also showed a dose dependant reduction in virus production (Ge et al., 2004).
Intranasal delivery was also found to be effective when targeting Parainfluenza virus and Respiratory Syncytial virus. Administration of siRNA with and without a vector (Transit TKO) was found to be both protective and therapeutic (Bitko et al., 2005).
The ability to induce RNAi across mucosal surfaces is also being explored as a means of treating sexually transmitted disease. Intravaginal delivery of RNAi targeting 2 viral genes have been shown to protect the mice from the otherwise lethal Herpes simplex virus-2 (Palliser et al., 2006).
RNAi has also been used to alleviate joint inflammation in experimental animals. siRNA targeting tumor necrosis factor alpha was injected into the knee joints of mice with collagen-induced arthritis (CIA). This was followed by electoporation. The development of arthritis was scored by assessing the inflammation of joints in the mouse paw, and in mice with CIA, joint inflammation was successfully inhibited (Schiffelers et al., 2005). Finally, diminished pain responses have been observed in rats following intrathecal delivery of an siRNA directed against the pain related cation channel P2X3 (Dorn et al., 2004).
Conclusion
The field of RNAi is still rapidly expanding and new discoveries are being made on a daily basis. SiRNAs are currently being used in gene function analysis, target identification and validation and as therapeutic agents. Their potential for use in evaluating target toxicity is significant and warrants further investigation. Although a viable technique for in vitro experimentation, success can still be hampered by problems with intracellular siRNA delivery and effective gene silencing.
Low transfection efficiency and excessive cytotoxicity are frequently encountered, and the development of an in vitro transfection method with an improved ratio of transfection to efficacy is a priority. For example, controlled forced intracellular delivery of siRNA by combining nucleotides with magnetic nanoparticles and applying a pulsed magnetic field could reduce the level of cytotoxicity and increase transfection efficiency above that achieved with electroporation of cells exposed to naked nucleic acids. This area is likely to see novel development in the near future.
Once delivered to the cell, effective knock-down of protein activity in the cell is influenced by the half-life of the protein under investigation. As RNAi inhibits novel protein production, long half-life proteins may maintain activity for long periods after effective gene silencing. Knowledge of protein half-life is helpful when planning and interpreting RNAi knock-down experiments. Deliveries of RNAi in viral vectors, particularly those that integrate into host genome, allow a prolonged knockdown of protein production. The short-lived nature of inhibition may be of advantage, however, when used as a therapeutic (to allow for rapid reversal following adverse events). Demonstration of effective gene silencing relies on the ability to detect both a reduction in mRNA and protein levels of the specified target in tissues – possibly in limited biopsy material. The availability of reagents to accurately measure reduced protein production may provide a further challenge in the interpretation of experimental results.
The transient nature of the inhibition of protein production is of particular concern in dividing cells, where cell division typically results in the loss of RNAi activity. This is of particular concern in vitro, where cell division rates are typically high. Certain in vivo organ systems, for example the gastrointestinal tract, also have high endogenous rates of division and gene silencing in these organ systems would be particularly challenging. Virally mediated integration of RNAi sequences into stem cell would overcome this issue, but would lead to a permanent knock down of protein expression, which may be of advantage for studying in vitro systems, but would carry liability as a therapeutic when adverse events were encountered.
The development of efficient in vivo delivery of RNAi is remains problematic. The paucity of publications describing successful in vivo delivery either suggests that either this is technologically demanding or else systems are being developed confidentially to ensure commercial exploitation. Efficient and effective in vivo transient and specific protein inhibition using RNAi clearly has significant potential to generate many therapeutics with associated financial reward. The search for a transfection method that achieves such efficient delivery of siRNA to a majority of body cell types and avoids significant toxicity is a demanding task. The current “gold standard” of hydrodynamic delivery raises issues of adverse events associated with the delivery system.
New developments in the area of in vivo delivery are anticipated. Finally, the cost or production of RNAi molecules has been a hurdle for the use, particularly the quantities required for in vivo studies. The cost of RNAi molecules has reduced significantly in recent years, and it appears that this is a trend which will continue. Hopefully, cost should not the barrier to use of this valuable tool.
In conclusion, the existing hype and excitement surrounding this mechanism of posttranscriptional gene silencing is justified, but experiments must now concentrate on in vivo validation of the technique to allow the path of discovery to continue.