Potential conflict of interest: Nothing to report.
Posttranscriptional regulation of gene expression is increasingly recognized as a model for inherited and acquired disease. Recent work has expanded understanding of the range of mechanisms that regulate several of these distinct steps, including messenger RNA (mRNA) splicing, trafficking, and/or stability. Each of these pathways is implicated in disease pathogenesis, and each represents important avenues for therapeutic intervention. This review summarizes important mechanisms controlling mRNA processing and the regulation of mRNA degradation, including the role of microRNAs and RNA binding proteins. These pathways provide important opportunities for therapeutic targeting directed at splicing and degradation in order to attenuate genetic defects in RNA metabolism. We will highlight developments in vector development and validation for therapeutic manipulation of mRNA expression with a focus on potential applications in metabolic and immunomediated liver disease. (HEPATOLOGY 2010.)
RNA Synthesis, Processing, and Posttranscriptional Regulation
Messenger RNA (mRNA) synthesis is accompanied by cotranscriptional addition of two major stability elements. These include 7-methyl guanosine (m7G) capping at the 5′ end of the mRNA transcript, which prevents 5′ exonuclease digestion and addition of a variable length poly-A tail to the 3′ end of the mRNA. Splicing of intron-containing transcripts occurs in the nucleus and is mediated by spliceosomal apparati composed of distinctive small nuclear ribonucleoprotein complexes1 and amplifies the number of translation products generated from a single genomically templated transcript. Enhancers and repressors of splicing exert additional combinatorial control over other posttranscriptional modifications including RNA editing.1, 2 Following or simultaneous with 5′ and 3′ end modifications and splicing of the nuclear transcript, mRNAs are bound to a complex of nuclear export factors to form a ribonucleoprotein complex. The ribonucleoprotein complex interacts with the nuclear pore complex for translocation to the cytoplasm.3 The m7G cap is bound by elongation initiation factor 4E and other cytoplasmic eIF4 proteins, which then engage by way of eIF3 directly to ribosomes to facilitate mRNA translational initiation.4 Similarly, poly-A binding protein also accelerates mRNA translation through activation of the helicase activity of the eIF4 complex.5
Eukaryotic mRNA half-lives range from minutes to months, the predominant mechanism limiting mRNA longevity being transcript deadenylation. mRNA transcripts earmarked for destruction undergo initial removal of the 5′ m7G cap by decapping enzymes6 or alternatively the 3′ poly-A tail may be targeted by deadenylases.7 Poly-A tails are deadenylated at differing rates, mRNAs containing long poly-A tails (>55 adenosines) being generally more efficiently translated.8 During cellular stress (such as glucose deprivation),9, 10 or when the mRNA decay machinery is overloaded, mRNA transcripts may be shuttled to cytoplasmic processing bodies where they are sequestered for degradation or later translation, providing an additional layer of regulation of gene expression and a defense against translating aberrant proteins.11
In addition to the 5′ cap and 3′ poly-A tail, key mRNA stability elements are found within the 3′ untranslated region (UTR). The 3′ UTR of transcripts manifesting rapid turnover or tight regulation commonly contain AU-rich elements (AREs). Most AREs contain the pentanucleotide AUUUA consensus sequence and are classified as class I or II, reflecting the number and context of pentanucleotides, although non-AUUUA AREs may also direct mRNA decay.12 ARE domains are targeted by ARE-binding proteins, including Hu antigen R (ELAVL1), tristetraprolin (ZFP36), apolipoprotein B mRNA editing enzyme, and catalytic polypolypeptide 1 (APOBEC1) that modulate mRNA stability.13, 14 CUG triplet repeat, RNA binding protein 1 (CUGBP1) is an ARE-binding protein that modulates mRNA splicing as well as stability and directly recruits the deadenylase poly(A)-specific ribonuclease (PARN), linking AU-rich elements and deadenylation.15 ARE-binding protein-mediated mRNA stabilization is incompletely understood, but likely involves competition for binding sites with destabilizing ARE-binding proteins or stabilizing interactions between the poly-A tail and poly-A binding protein.11 Modulation of ARE-binding protein activity occurs through combinatorial interactions with other signaling pathways, including p38 mitogen-activated protein kinase and Wnt/β-catenin.16 For example, activation of p38 mitogen-activated protein kinase results in phosphorylation of the ARE-binding protein KH-type splicing regulatory protein in developing myocytes, decreasing the affinity of KH-type splicing regulatory protein for ARE containing mRNAs and thus favoring transcript stabilization.17
Stability elements within mRNA transcripts may also include microRNA (miRNA)-binding sites (Fig. 1A). miRNAs are small single-stranded noncoding RNAs that are processed and incorporated into the RNA-induced silencing complex (RISC). miRNAs are transcribed as monocistronic or polycistronic transcripts to form primary microRNAs, which are cleaved in the nucleus by the RNAse III endonuclease Drosha to 60–90 nucleotide pre-microRNAs and exported from the nucleus by the karyopherin Exportin 5. Cytoplasmic endonucloeolytic processing by Dicer yields 20–22 nucleotide double-stranded miRNA fragments that bind complementary target mRNAs leading variably to translational suppression, trafficking to processing bodies or degradation.18
Imperfect complementarity between the miRNA and target sequence may result in translational repression. This may occur through alterations in m7G cap-dependent recruitment of the 80S ribosome to mRNA (cap-dependent repression) and interference with postinitiation translation (cap-independent repression) (reviewed by Wu and Belasco19) or alternatively, targeted transcripts may be trafficked to processing bodies for sequestration.20 In contrast, perfect or near-perfect complementarity generally results in transcript destabilization, which may occur either by direct endonucleolytic cleavage within the target sequence21 or by accelerated deadenylation, 5′ decapping, and 5′ exonuclease activity.19 There is still uncertainty, however, regarding the fate of the target transcript and the role of target-miR complementarity (reviewed by Brodersen and Voinnet22). For example, near-perfect miRNA-target mRNA interactions may lead to translational repression and not cleavage.23 miRNA-mediated translational repression is also reversible as evidenced by the cationic amino acid transporter-1, which is repressed in human hepatoma cells by miR-122, and reversed by stress (amino acid deprivation or induction of the unfolded protein response). This response is mediated by competition with Hu antigen R for binding sites within the 3′-UTR shared with miR-122. These results suggest that miRNA binding and ARE-binding proteins may interact to modulate overall mRNA stability.24
Mechanisms of Therapeutic Posttranscriptional RNA Regulation
Antisense oligonucleotides (ASOs) were discovered 30 years ago25, 26 as single-stranded DNA molecules (13-25 nucleotides) that complement target RNA through sequence-specific base pairing. ASOs with negatively charged backbones recruit RNase H to the temporary duplex causing endonucleolytic mRNA cleavage27 (Fig. 1B). Alternatively, ASOs targeting the 3′ UTR, 5′ region, or AUG initiation codon and those with uncharged backbones may suppress mRNA expression by steric mechanisms, leading to decreased or aberrant splicing, translation, or mRNA transport.28, 29 Although ASO-mediated mRNA targeting is susceptible to sequence-specific off-target effects as well as nonspecific effects resulting from enzymatic processing,30 the major limitiations are in ASO delivery and short half-life. Chemically modified derivatives, including phosphorothioates and peptide-linked morpholino derivatives, are currently in development with targets such as bcl-2 and protein kinase C-α (PKC-α, PRKCA).31, 32 To date, fomivirsen sodium (Vitravene, Isis, Carlsbad, CA), a phosphorothioate derivative targeting cytomegalovirus mRNA in CMV retinitis, is the only ASO therapy approved by the US Food and Drug Administration.
Targeting of mRNA expression through ASO mediated manipulation of mRNA splicing holds promise for treatment of diseases caused by a variety of hereditary mutations and may offer opportunities for therapeutic targeting in other settings. Morpholino-modified ASOs targeting mutant intronic splice sites in β-globin pre-mRNA from thalassemic patients33 corrected the splicing defects and restored hemoglobin synthesis. ASOs could thus theoretically be used to target disease-causing splicing mutations, such as in Wilsons disease,34 familial intrahepatic cholestasis, hepatocellular carcinoma,35 or alternative splice variants of hepatitis B36 and C. Therapeutic manipulation of Apo-B mRNA splicing offers an experimentally testable approach to reduce hepatic Apo B production.37, 38 In this situation, ASOs targeted to the physiological 5′ and 3′ splice sites within exon 27 of APOB induce exon skipping leading to a frame shift within exon 28 leading to production of a truncated protein that is less efficient at exporting lipid (discussed further below). Mutations in the dystropin gene cause Duchenne muscular dystrophy (DMD) as a result of a prematurely truncated and unstable transcript. Less severe forms of DMD result from mutations that maintain the reading frame by skipping the defective exon resulting in a partially functional protein. ASO-mediated rescue produced an mRNA transcript lacking exon 23 with validated cellular localization of the dystrophin gene product in mice with a mutant exon 23 that caused premature termination.39 Accordingly modification of mRNA splicing may provide a strategy to increase the abundance and stability of an otherwise rare or unstable transcript.
Catalytic RNAs (ribozymes) act in a sequence-specific fashion in cis or trans to cleave mRNA in the absence of proteins (reviewed by Scherer and Rossi28) and may direct trans-splicing reactions. Ribozyme-mediated trans-splicing therapeutic constructs contain a targeting domain, the ribozyme sequence and a replacement coding region (Fig. 1C). Base pairing of the guide sequence with the target transcript results in a trans-splice reaction and a chimeric transcript29 in which the disease causing mRNA segment is replaced by a functional coding exon. A shortened version of the Tetrahymena thermophila self-splicing intron coupled to a portion of the γ-globin transcript was used to target the mutant βs-globin mRNA transcript just upstream of the sickle cell mutation. The ribozyme-directed cleavage and ligation generated a chimeric transcript encoding a nonsickling but functional globin protein.40 Nonribozyme mediated trans-splicing reactions using the cellular spliceosomal apparatus are also under development.29
RNA interference (RNAi) encompasses two gene-silencing processes, small interfering RNA (siRNA) and miRNA (Fig. 1A). miRNAs are endogenously transcribed single-stranded RNAs that form hairpin structures, which are processed through a series of nucleolytic cleavages to form mature miRNAs. In contrast, siRNAs are endogenously or exogenously derived double-stranded noncoding RNAs that are also processed to their mature 21–23 nucleotide form. In both cases, the final “guide” strand is incorporated into the RISC directing endonucleolytic cleavage or translational repression of target RNAs.28, 41 Both mechanisms may be co-opted for therapeutic purposes and offer advantages over other sequence-specific targeting strategies. These include low effective concentration, limiting toxicity, and flexibility in targeting sites within a given mRNA. RNAi guide sequences can be delivered as preformed double-stranded RNA oligonucleotides (siRNA) or generated within cells using promoter-based expression systems (short hairpin RNA [shRNA]). Each approach has advantages and disadvantages.42 siRNA is attractive in that the delivered sequence activates RNAi directly with minimal cellular metabolism being required. The downside is that the effect is transient and administration must be repeated. In contrast, shRNA-based delivery involves delivering a DNA template to the target tissue from which an shRNA is transcribed (Fig. 1A). This shRNA undergoes Exportin-5–mediated nuclear export and processing by Dicer to yield more durable siRNA production. Saturation of the Exportin-5 pathway, however, by adenoviral-mediated shRNA is hepatotoxic,43 an effect not seen with in vivo delivery of siRNA.44
Specific targeting of endogenous miRNAs may also hold therapeutic benefit. Administration of a cholesterol-conjugated, 2'-OMe–modified RNA sequence (antagomir) targeting miR-122 produced potent suppression of endogenous hepatic miR-122 levels in mice and up-regulation of mRNAs enriched in miR-122 3′-UTR target sites (discussed below).45
Hepatotropism of RNA Targeting
Early work with modified encapsulated ASOs demonstrated effective hepatotropism after intravenous administration46 with >80% detectable in the liver at 1 hour and 26% by 4 hours. Further experiments demonstrated that subcutaneously injected ASO targeting Fas accumulated in the liver of mice for approximately 10 days.47 Naked siRNAs injected systemically concentrate predominantly in the kidney,48 but modifications to the siRNA backbone increase plasma stability and improve hepatic delivery. Synthetic duplex siRNAs containing one phosphodiester strand and one phosphothioate strand were stable up to 72 hours in 50% mouse serum compared with unmodified single-stranded RNA that was degraded in seconds.49 siRNA-lipid complexes including cholesterol, cationic lipid, or lipidoid nanoparticles increase transfection efficiency50 and have been used to optimize in vivo targeting of apolipoprotein B and proprotein convertase subtilisin/kexin type 9 (PCSK9) (discussed below).51–53
Potential for Off-Target Effects and Toxicity
siRNAs activate the immune system54 and double-stranded RNA can activate the intracellular receptors double-stranded RNA (dsRNA)-dependent protein kinase R (EIF2AK2) or retinoic acid inducible gene-I (DDX58). Activation of protein kinase R results in global suppression of protein synthesis, ultimately leading to apoptosis. Activation of either protein kinase R or retinoic acid inducible gene-I leads to interferon production and other antiviral responses, including Toll-like receptors TLR7 and TLR8 which are activated by siRNA in a sequence and delivery-specific fashion. U- and G-rich siRNAs, longer dsRNA sequences, single-stranded siRNAs, and cationic lipid-complexed siRNAs result in the most potent activation of immunostimulatory signaling pathways. Conversely, shorter siRNA sequences, low G/U content, double-stranded, naked, and cholesterol-complexed siRNAs appear to be least immunostimulatory.54In vivo trials of siRNA therapy have used chemical modifications such as 2 nucleotide 3′ overhangs and 2'-O-methyl modifications that diminish immunostimulatory effects of siRNAs as evidenced by attenuated production of interferon-α, interleukin-6, and tumor necrosis factor-α.51, 55
RNA therapy–mediated off-target effects can also occur by way of sequence homology to unintended transcripts. ASOs targeting protein kinase C-α RNA suppressed protein kinase C-ζ (PKC-ζ, PRKCZ) mRNA even though these isoforms share only 11/20 base pair homology.56 Additionally, whereas siRNA binding to target mRNA with perfect complementarity results in mRNA cleavage, imperfect complementarity suppresses translation, similar to that discussed above with miRNA.57
Experience with RNAi therapy using adeno-associated virus has revealed potential hepatotoxicity. Adeno-associated virus–mediated shRNA expression targeting transgenically expressed human α-1 antitrypsin mRNA43 suppressed human α-1 antitrypsin mRNA, but was complicated by lethal hepatotoxicity, likely due to oversaturation of Exportin-5 as mentioned above (Fig. 1A). Accordingly, successful delivery of an RNAi effector must be accompanied by careful modulation of intracellular concentrations in order to avoid toxicity.
Therapeutic Applications of RNA Modulation in Metabolic Liver Disease
Efforts are underway to use RNA-targeting therapy in metabolic conditions involving the liver, in particular the management of hypercholesterolemia, where despite the success of statin-based therapies, only two-thirds of patients achieve their target low-density lipoprotein (LDL) goal.58 Patients with hereditary mutations in either the LDL receptor (LDLR) (familial hypercholesterolemia) or in the LDLR binding region of apoB100 fail to clear apoB100-containing lipoproteins from the serum, leading to accumulation of lipid particles and hypercholesterolemia. ASO targeting of apoB-100 has been used in proof-of-principle experiments in ApoE-deficient and LDLR-deficient mice,59 where cholesterol reduction ranged from 25% to 55%, with LDL cholesterol reductions as high as 88% without hepatic steatosis. A double-blind, placebo-controlled, randomized trial of mipomersen (apoB ASOs) in 36 volunteers with mild sporadic dyslipidemia demonstrated decreased serum LDL and apoB levels of 35% and 50%, respectively. There were no serious adverse events, with the most common side effect being mild injection site erythema.60 More recently, phase 3 studies with mipomersen in familial hypercholesterolemia subjects revealed a 25% reduction of LDL cholesterol levels after 26 weeks of treatment.32, 61 Further studies are directed toward assessing whether this approach places patients at risk for hepatic steatosis as seen in patients with some naturally occurring apoB mutations or with familial hypobetalipoproteinemia.62
Targeting the 3′ end of apoB mRNA with cholesterol-conjugated siRNA in mice lowered serum apoB levels and decreased serum LDL52 and stable nucleic acid lipid particle–formulated siRNA targeting the 5′ end of apoB mRNA53 in cynomolgus monkeys suppressed apoB mRNA by 90% at 48 hours after a single dose, an effect that persisted for 11 days, accompanied by >60% reductions in serum cholesterol.
ASO-mediated therapeutic apoB RNA splicing inferference was used to induce skipping of exon 2738 to generate a truncated apoB isoform predicted to decrease LDL cholesterol. The skip 27 mRNA translated product (apoB87) was exported from the cells at decreased efficiency compared with apoB100, an effect reminiscent of that associated with familial hypobetalipoproteinemia patients expressing an apoB truncating mutation.63 Differential processing of the truncated apoB mRNA and alterated posttranslational lipidation of the protein product might account for the therapeutic efficacy of this targeting strategy.
A second target of hypolipidemic therapy is proprotein convertase subtilisin/kexin type 9 (PCSK9), which regulates hepatic LDLR expression via altered protein processing.64 Gain-of-function mutations result in increased degradation of the LDLR and increased lipid levels. Conversely, loss-of-function mutations result in increased expression of LDLRs and decreased serum cholesterol.65 Epidemiologic studies link mutations in PCSK9 to low LDL cholesterol levels in approximately 2% of African Americans,66 with an approximately 40% reduction in LDL cholesterol and reduction in coronary artery disease,64 making PCSK9 an attractive hypolipidemic target. In vivo targeting of PCSK9 using ASOs in hyperlipidemic mice fed a high-fat diet increased hepatic LDLR expression and decreased serum cholesterol.67 siRNA targeting of PCSK9 in rodents and nonhuman primates using intravenous lipidoid nanoparticles also decreased serum cholesterol.51
ASOs have also been developed (antagomirs) to target miRNAs involved in hepatic lipid metabolism, including miR-122. In vivo silencing of miR-122 in mice using a miR-122 antagomir decreased 3-hydroxy-3-methylglutaryl-coenzyme A reductase and acyl-Co A synthetase 2 among other targets, accompanied by a >40% decrease in plasma cholesterol and without adverse effects on hepatic function or elevated aminotransferase levels.45 Analysis of miRNA expression patterns in patients with nonalcoholic steatohepatitis revealed reduced expression of miR-122 with corresponding alterations in several target genes that were replicated in HepG2 cells following either miR-122 knockdown or overexpression.68 These findings and the implications for nonalcoholic steatohepatitis pathogenesis need to be interpreted with caution, however, because miR-122 is known to modulate the expression of approximately 200 target genes69 and the precise mechanisms by which alterations in miR-122 expression influence hepatic lipid metabolism in human subjects remains to be defined.
Immunomediated Liver Disease
Hepatocytes are susceptible to Fas-mediated apoptosis, and disruption of this pathway protects hepatocytes from acute liver injury. Mice treated with ASOs targeting Fas were protected from Fas antibody-induced fulminant liver failure and from a sublethal dose of acetaminophen.47 Similar experiments using RNAi confirmed that interference with Fas signaling protects from fulminant hepatitis.70 Taken together, these data strongly suggest that RNA-directed therapy may have clinical use in treating disorders of immunomediated liver injury as well as fulminant liver failure with an immunomediated component.
Future Directions and Prospects
The last several years have witnessed great advances in understanding of the physiologic and pathologic roles of regulated mRNA stability. Among the challenges moving forward will be to define how aberrant posttranscriptional mRNA processing including mRNA trafficking and stability functions in disease pathogenesis and to define optimal mRNA targets for specific clinical conditions. From the standpoint of drug development, refining potency, specificity, minimization of side effects and tissue delivery will be crucial goals in coming years.
We apologize to colleagues whose work we were unable to cite due to space limitations.