RNA Modalities in Huntington’s Disease Therapy

Reported by: Jill Crittenden

Abstract:

Key deliberation at the workshop RNA Modalities in Huntington’s Disease Therapy concerned how to tailor therapeutic RNAs for the treatment of Huntington’s disease (HD). This meeting was spurred by the recent discovery that small interfering RNAs (siRNAs) can be used in mammals to degrade complementary messenger RNA. Discussions focused primarily on the mechanism and design of siRNAs, but also touched on recent progress in the therapeutic application of ribozymes and aptamers (RNAs that cleave or that inhibit/activate target molecules, respectively). The choice target for destruction by therapeutic RNAs was the huntingtin transcript, although notable mention was given to other targets that might contribute to the disease phenotype such as the aspartic protease and CREB binding protein (CBP). For both target-specificity and overall safety it was deemed essential to develop vectors in which the expression of RNA can be tuned to therapeutic levels. Proposals for how to specifically disrupt the mutant huntingtin transcript without destroying the function of the normal allele included 1) targeting an siRNA to a single nucleotide polymorphism (snp) specific to the disease transcript, 2) introducing a wildtype HD transgene that is resistant to the therapeutic RNA and 3) designing a ribozyme that cleaves only transcripts with an extended polyglutamine repeat. Several participants described their success in using the first strategy to specifically disrupt disease transcripts, thus underscoring the need to identify ample snps within the mutant HD gene to test for use in siRNA-mediated degradation (not all siRNA sequences are effective). It is not yet known whether introducing siRNAs directly into the brain would be sufficient to confer long-term suppression, therefore, several participants arranged collaborations to test this using reporter genes in rodent models. The latter part of the workshop focused on a familiar topic: how best to deliver therapeutic agents to the brain. Attending virologists gave promising reports of how new methods combined with the latest generations of adenovirus, adeno-associated virus and lentivirus are improving delivery in animal models. The use of synthetic delivery systems may be further from clinical application.

Workshop Introduction

            The December 2002 workshop on designing RNA technologies to cure HD included scientists with expertise ranging from RNA biochemistry to human drug delivery systems. This diverse group was brought together by the Cure HD Initiative, the main funding branch of the Huntington Disease Foundation (HDF), to explore the application of recently discovered siRNA functions toward the treatment of HD. The meeting opened with a warm greeting from Nancy Wexler who described how her father came to establish the HDF in 1968. Today, the HDF has grown to provide millions of dollars in annual funding for HD research and sponsors frequent meetings to promote scientific collaboration and speed the discovery of a cure.

            Richard Mulligan, HHMI professor in Genetics at Harvard Medical School, served as chair of the workshop which covered two major topics: how therapeutic RNAs, particularly siRNAs, should be tailored to the treatment of HD, and how these RNAs could be delivered to brain tissue. Mulligan succeeded in steering the two days of discussion and debate such that all of the major topics were introduced early, and the final hours focused on defining key questions and the experiments to answer them.

Patient Interview

          Opening statements were followed by the introduction of Gary Fujiwara and his wife Sarah, who discussed living with HD during a touching two hour interview led by the neurologist Steven Hersch. Gary was diagnosed with HD six years ago at the age of 60 and falls into the 1-2% of HD patients who have no family history of the disease. Despite having no previous familiarity with HD, Gary and Sarah have made concentrated efforts to follow therapies that can extend a reasonable quality of life. Although Gary displayed uncontrollable choreic movements during the entire interview, he is devoted to physical therapy and has even taught himself to ski within the past three years. Gary also works hard to maintain an active mind by reading articles in the daily paper two or three times so that he can remember them for dinner conversations with his family. Besides adherence to physical and mental exercise, maintaining a healthy weight can also improve life with HD. For reasons still unclear, but thought to be metabolic rather than neurological, HD patients must consume as much as 5,000-6,000 calories per day to avoid cachexia (accelerating weight loss). Gary is barely able to maintain his modest weight by supplementing three full meals per day with one or two chicken pot pies, several bowls of ice cream and 3,000 calories of Ensure. Gary stated that what bothers him most is the impatience and humiliation associated with being unable to complete trivial tasks. He strongly recommended participation in support groups and praised the Cambridge Early HD group led by Kathy Knoblauch. In addition to psychological and physical therapies Gary takes the antipsychotic Seroquel to suppress insomnia, a common symptom of HD. Gary also takes creatinine, which provides relief from chorea in the early stages of HD but loses effectiveness as the disease progresses; more powerful movement-suppression medications are unsuitable because they also inhibit voluntary motions.

            The end of the interview touched on the controversial issue of genetic testing, which less than 3% of people at risk for HD in the United States and Europe choose to do. Sarah does not advise her children to be tested because she has seen the regret and trauma caused by the news of facing an incurable disease. However, Sarah did agree that patient-blind testing of individuals who wish to participate in clinical trials might be appropriate. Further, both Sarah and Gary were enthusiastic about parent-blind testing of pre-implantation embryos with the purpose of providing non-HD offspring to at-risk individuals; this procedure is not widely available and the cost is rarely covered by insurance companies. In terms of middle-aged people like himself, Gary said that he finds hope in studies like those of Yamamoto et al., which demonstrated that turning off transcription of a mutant HD transgene in a mouse model can reverse the disease symptoms. The in-depth discussions and forthrightness of the Fujiwaras led several seasoned HDF members to state that this was one of the most insightful interviews they had seen. One thing is certain, meeting Gary and Sarah set the stage for intense discussion to find a cure that is applicable at any stage of HD.

Huntington’s Disease Overview

Genetics and neuropathology

            The scientific discussions began with an overview of HD given by Ethan Signer, Director of the High Q Foundation. HD is an autosomal dominant disease caused by the expansion of a CAG repeat within the first exon of the HD gene. People carrying an HD gene with 15-36 CAG repeats are normal whereas an expansion beyond 39 repeats invariably causes disease. CAG encodes the amino acid glutamine which, when expanded in the huntingtin protein (htt), is thought to mediate the neuropathology of HD; however, a role for the mutant huntingtin transcript has not been ruled out. Although HD expression is nearly ubiquitous, post-mortem analysis of early-stage HD brains reveals that specific regions are particularly vulnerable. The first cells to amass the stereotypical aggregates of htt and die are the medium spiny neurons of the striatum and a subset of cortical cells. Steven Hersch said that one study found cortical cell death mainly within layers 3, 5 and 6; however, it has been difficult to determine which types of cortical cells are affected because the aggregates are localized to the dendrites, rather than the nucleus, of these cells. Because neurons in layers 3 and 5 of the cortex heavily innervate the striatum, it has been suggested that malfunction of these cells might be the primary cause of striatal degeneration. Hersch mentioned supporting experiments in mice where the resection of cortex reduced the severity of striatal malfunction. Carl Johnson said that this issue of whether striatal neurons are “murdered” by cortical afferants or die by “suicide” is being further addressed by experiments in Dan Goldowitz’s lab: chimeric mice that express mutant huntingtin in the cortex but not in the striatum are being generated to compare to mice that have striatal-specific expression. Results from Elena Cattaneo’s lab, described by Signer, suggest that the striatum and cortex are mutually dependent in supporting striatal viability. Previous studies demonstrated that the viability of medium spiny neurons is dependent on brain-derived neurotrophic factor (BDNF), a protein released from cortical afferants. Further investigation by the Cattaneo lab showed that wildtype htt potentiates the release of BDNF from cortical afferants and that this effect is inhibited by the presence of mutant htt; thus, function of both the striatum and the cortex may be key to striatal viability.

Monitoring disease progression

            A question from Phil Sharp sparked a short discussion of what biomarkers can be used to monitor progression of early HD, with the goal of measuring therapeutic efficacy. Signer said that structural magnetic resonance imaging (MRI) and neurological studies are being carried out with at-risk populations to improve the methods of early diagnosis. Emphasis on MRI of cortical function is being tested as a means of measuring early disease progression in a study cited by Steven Hersch. These are critical studies since striatal atrophy and cortical aggregate formation occur presymptomatically in humans and, therefore, patients often do not present before significant striatal loss has already occurred.

Pathological agent

            A continuing debate in the HD field is what component actually causes neuronal malfunction and death. This is of relevance for RNA therapy to guide the choice of which transcripts to target for prevention or reversal of HD. One possibility is that the HD aggregates, which Signer defined as comprising only mutant htt, result in disease by blocking proteosomes and allowing the build-up of toxic proteins. Alternatively, HD might be caused by the sequestration of essential proteins within inclusion bodies (htt aggregates amassed with other components including heat-shock proteins, ubiquitin, and transcriptions factors). The latter hypothesis is supported by experiments that Signer described from Chris Ross’s lab where overexpression of CBP, a transcription factor found in HD inclusions, rescued mutant htt-associated toxicity in cultured cells. Carl Johnson mentioned that experiments from several groups support the idea that inclusions and aggregates are seeded by an amino-terminal fragment of htt that results from cleavage by an aspartic protease. It was concluded that, although a priority target for suppression is the transcript for mutant htt, work on other targets should not be slowed. Briefly, these targets include proteases that cleave near the amino terminus of htt, proteins that exacerbate the impaired mitochondrial function found in HD and proteins that shuttle mutant htt into the nucleus where it may interfere with transcription.

Overview of Small Interfering RNAs (siRNAs)

Mechanism of interference

            Phil Sharp and Phil Zamore provided a brief description of how siRNAs function to inhibit gene transcription and translation. In vivo, the process begins by a double stranded RNA molecule being cleaved into 21-23 nucleotide fragments by the RNaseIII enzyme Dicer. The resulting antisense fragment, termed siRNA, then associates with Argonaute/EIF2c proteins to form RISC (RNA-induced silencing complex). RISC can induce silencing by at least two mechanisms: 1) by cleaving mRNAs within a sequence that shares perfect complementarity to an siRNA and 2) inhibiting translation of mRNAs that are nearly complementary to an siRNA, but have a few mismatches. RNA interference (RNAi) refers to the first mechanism and is a well studied method of transcriptional control used by plants, worms and flies. This process has not yet been shown to occur naturally in mammals, but the introduction of exogenous siRNA to mammalian cells can result in RNAi. Moreover, it has recently been found that mammals do express short RNAs that interact with RISC to mediate translational interference; these microRNAs are produced by Dicer cleavage of longer precursor RNAs that are in the form of double-stranded hairpins as a result of intramolecular base-pairing. Numerous microRNAs have recently been cloned from mammals and, according to Phil Sharp, may reach a total of ~250 by David Bartel’s estimates. The mammalian microRNAs do not have exact complementarity to any known genes and are therefore proposed to be involved in translational control. Phil Sharp said that David Bartel’s group measured up to 40,000 copies of a single siRNA species per human cell, suggesting that there may be a total of ~250,000 RISC complexes per cell. Sharp’s lab has shown that knockdowns of 14-40 fold can be achieved with only 4,000 siRNA molecules, a population that may, by the estimates above, represent only ~2% of the total siRNA per cell. This calculation suggests that it may be possible to introduce therapeutic levels of siRNAs without titrating away RISC complexes from endogenous siRNAs.

            A third mechanism by which siRNAs can interfere with expression is by gene silencing. In plants, it has been shown that siRNAs can permanently silence gene expression by inducing methylation of homologous DNA within the promoter (the upstream region of a gene that regulates its transcription). John Rossi presented data demonstrating that siRNAs can cause spreading methylation of promoter DNA in a human cell line as well, although silencing was not conferred in this case. Further experiments along this line are important to determine whether transient expression of a therapeutic siRNA could induce permanent silencing of a disease-causing gene.

Generating an siRNA

            siRNAs may be produced by 1) chemical synthesis in an RNA synthesizer, 2) in vitro transcription of an expression construct or 3) in vivo transcription of an expression construct. The double-stranded siRNA may be assembled by annealing two complementary strands of RNA or by Dicer cleavage of a hairpin RNA either in vivo or in vitro. Phil Sharp recommended choosing a 21bp siRNA sequence that has ~50% G or C nucleotides, no homology in the sequence database to genes other than the intended target and no run of identical nucleotides. Because the efficacy of any given siRNA can vary greatly, there was a general consensus that three different siRNAs should be generated to ensure success in silencing the target. The major hurdle to designing an effective siRNA is determining which region of the target mRNA is accessible for binding, since mRNAs undergo intermolecular base pairing to form secondary structures. John Rossi said that as yet no effective algorithms exist to predict the secondary structure of large RNA molecules like mRNA. Zamore even pointed out that measuring siRNA effectivity might be the best available method to identify secondary structures in its target!

Therapeutic modality

            Methods for siRNA therapy include submitting the patient to multiple injections of siRNA or introducing a DNA expression construct for continuous expression of the siRNA. The latter has already met with success in a mouse model as shown by Beverly Davidson’s lab. John Rossi recommended the addition of an intron to the siRNA expression construct since this has been shown to improve the stability of small transcripts.

An open question is whether a single injection of siRNA could maintain long-term knockdown. Phil Sharp said that experiments in nondividing cells suggest that siRNAs may have a half-life as long as 40 days in cell culture. John Rossi, Natasha Caplen and others reported detecting the antisense strand of siRNA, but not its complement, following siRNA transfection or expression in cultured cells. Together, these results suggest that siRNAs might be specifically and enduringly protected by RISC. Further, if experiments such as those in the Rossi lab demonstrate a mechanism whereby siRNA can induce promoter methylation, transient siRNA application may suffice to stably silence genes.

Allele-specific siRNAs

            Although an essential role for wildtype htt in the adult has not been proven, mice with a constitutive knockout of the mouse HD gene die at embryonic day 8. Furthermore, knockout of mouse HD expression in the brain at postnatal day ~5 results in neurodegeneration within multiple brain regions including the striatum, hippocampus and cortex. In the likelihood that wildtype htt is required for striatal function, several workshop discussions focused on how to design an siRNA that disrupts the disease-causing transcript without effecting the normal allele. The most straight forward approach is to design siRNAs to a region in which there is a single nucleotide polymorphism (snp) between the two alleles. To reduce nonspecific binding to the wildtype huntingtin transcript, Sharp and Zamore suggested placing the snp within the center of the siRNA and, if possible, choosing an A/G snp to increase steric hindrance (purine-purine clash). Philip Zamore warned of an siRNA experiment where the uncleaved portions of a targeted mRNA were found to be translationally competent following processing by Dicer. Considering that aggregation is promoted by an amino-terminal fragment of htt that contains the glutamine repeat, it might be wise to design an siRNA that targets this region. Unfortunately, the region upstream of the CAG repeat comprises fewer than 400 base pairs and the average snp frequency throughout the genome is only ~1/500 base pairs. Although the chances may be low for finding a snp within an effective siRNA-target sequence in the amino-terminal portion of htt, the full-length transcript is 13.7 kilobases (kb), providing other potential target regions. The identification of appropriate snps across multiple HD-effected families is a justified funding priority, declared Carl Johnson.

            One unknown is whether an siRNA with a single nucleotide mismatch would interfere with translation of normal huntingtin as in the case of microRNAs. Zamore speculated that allele-specific knockdown may be a function of siRNA concentration since transcript degradation is a catalytic event and translational repression is stoichiometric; thus, it may be beneficial to put therapeutic siRNAs under the control of a promoter that is pharmacologically regulatable. In practice this may not be an issue however since Hank Paulsson’s and Philip Zamore’s labs said that they had both successfully used the nucleotide mismatch strategy for allele-specific knockdown in cultured cells. Paulsson made an allele-specific knockdown of a huntingtin transcript that contains a GAG deletion in exon 58. This is an important finding in light of the fact that this deletion is in significant linkage disequilibrium with HD chromosomes worldwide, i.e. an HD gene carrying the GAG deletion is more likely to be an HD patient’s disease-causing allele rather than his/her normal HD allele.

            Stefan Kochanek offered the view that an alternative to allele-specific knockdown would be to introduce a normal HD transgene that is resistant to siRNA. This transgene could be included in the siRNA construct such that it would only be delivered to cells undergoing RNAi. One impediment to this strategy is that not all vectors can deliver enough DNA to include the large HD gene. Another potential complication is the possibility that this would result in overexpression of wildtype htt, although in mice a five-fold overexpression has not been found to cause negative effects according to Carl Johnson.

Safety considerations

            In the event that allele-specific knockdown proves infeasible for HD, it is important to understand the consequences of knocking down the normal allele. Natasha Caplen and Phil Sharp pointed out that the relevant functional studies of wildtype htt will be expedited by the advent of siRNA technology. A second important issue for siRNA therapy is whether generalized effects on transcription may result; Natasha Caplen and Ken Kosik are performing microarray analysis of cells following the introduction of a variety of siRNAs in an attempt to identify shared changes. One approach to reducing potentially harmful effects of therapeutic siRNA is to express it only in a subset of tissues. Many of the expression constructs for siRNAs that have been published-to-date utilize the constitutive U6 promoter. This promoter recruits Polymerase III, an enzyme that normally transcribes RNAs that are small and have short polyA tails, two features common to siRNAs. By placing the transcription start site immediately downstream of the promoter and using a minimal polyA signal, Beverly Davidson’s group succeeded in using a Polymerase II promoter, CMV, to transcribe siRNA. This is an important first step to driving tissue-specific siRNA expression since Polymerase II is recruited by tissue-specific promoters. Paulsson pointed out that the use of a promoter that can be regulated pharmacologically is especially important in clinical trials where adverse effects of siRNAs might be revealed. Clearly a multitude of HD siRNAs and constructs must be tested in models for 1) knockdown efficacy, 2) target specificity, 3) expression control, 4) toxicity and 5) effect on normal striatal function.

Other Therapeutic RNAs

Ribozymes

            John Rossi and Ron Breaker summarized recent advances in the development of ribozymes for therapy. Ribozymes, discovered in the early 1980s, are RNA molecules whose sequence and folding allow them to catalyze cleavage and/or ligation of RNA. Ribozymes with specificity for a desired transcript may be individually engineered by recombinant DNA technology. For example, the hammerhead ribozyme contains two variable regions (flanking the catalytic domain) that can be engineered to base-pair with a target RNA. Alternatively, in vitro evolution may be used to select for ribozymes with the specificity of interest: ribozymes can be converted to DNA and subjected to mutagenic PCR (polymerase chain reaction). Resulting mutant ribozymes can then be tested for binding to the target of interest and subjected to additional rounds of PCR for improvement. Ron Breaker and Beverly Davidson were both enthusiastic about data presented at a recent meeting showing that a snp-recognizing ribozyme in a viral vehicle could be delivered by subretinal injection to reduce levels of a disease-causing transcript by ~50%.  Breaker’s lab is designing ribozymes that can act as molecular measuring sticks (or “ribozyme logic gates”) by cleaving between two specific sequences that are separated by a defined interval. Obviously such an enzyme would be useful to specifically cleave expanded transcripts such as that for mutant htt. Experiments are ongoing in the Breaker lab to modify this ribozyme so that it will function only when the target sequences are part of the same transcript rather than situated in trans. John Rossi said that one of the barriers to ribozyme utility in vivo has been the low frequency of interaction between the ribozyme and its target. This problem was overcome for targeting the transcript that causes myotonic dystrophy­–in this case the mutant mRNA is localized to the nucleus so that addition of a nuclear localization signal to the ribozyme is sufficient to drive interaction with the mRNA substrate. However, since most mRNAs cytoplasmic, the ribozymes may be diluted to ineffectual levels in the cytoplasm or sequestered in a subcellular compartment away from their target. Rossi  and Breaker expressed hope that the dilution problem would be overcome by the rapid catalytic activity of a newly engineered hammerhead ribozyme that is capable of cleaving as many as 187 molecules per minute.

Aptamers

            Richard Mulligan asked about the current status of controlling protein expression with aptamers­–DNA or RNA molecules that can be evolved in the test tube to activate or inhibit a molecule of interest. The most efficient way to design aptamers is to select for the appropriate function extracellularly and then work on methods of intracellular delivery according to Breaker. Breaker described how Michael Green’s group used in vitro techniques to generate aptamers that bind to the membrane-permeable Hoechst dye. These aptamers were cloned into the 5’UTR (untranslated region) of  reporter genes and demonstrated to inhibit translation in vivo upon application of Hoechst. Although the Hoechst experiment is an important proof-of-principle, for therapeutic purposes it might be necessary to design an aptamer that can hijack a naturally existing target. For example, stage III clinical trials that are underway to treat macular degeneration using an anti-VEGF DNA aptamer, described Breaker. Phil Sharp said that numerous instances of natural aptamer systems have been identified in the bacterium B.subtilis: this organism utilizes four receptors that are composed of RNA to regulate some 20 downstream genes. Sharp noted that aptamers could potentially be designed that would allow RNAi to be regulated by drugs. For example, erythromycin-binding aptamers could be added to the Dicer transcript that would make the translation of Dicer erythromycin-dependent. Or, proposed Zamore, production of a microRNA could be individually controlled. MicroRNAs are derived from hairpin precursors that undergo trimming of their single-stranded tail prior to export from the nucleus, processing by Dicer in the cytoplasm, and then assembly into RISC; therefore, an aptamer that protects the tail from cleavage could inhibit its interaction with RISC by preventing nuclear export. There was no lack of ideas for how aptamers and ribozymes might be further developed, and it was generally agreed that the promise of siRNA therapy should not slow such work. And, as John Rossi pointed out, it is important to gather and compare safety data from different types of therapies.

Overview of RNA Delivery Systems

Viral-mediated delivery

1) Adenovirus

            Delivery to the target tissue is a problem common to any type of therapeutic RNA. The employment of viruses to deliver therapeutic DNA is ongoing in hundreds of clinical trials and Beverly Davidson and Stefan Kochanek described three viruses that show the most promise. The discussion began with a review of the gutless adenovirus, a second-generation virus from which genes have been removed to reduce immunogenicity and allow for insertion of up to 34 kb of foreign DNA; Kochanek favors this virus because it has the capacity to accommodate both the wildtype HD gene and siRNA or ribozyme expression constructs. Adenovirus is capable of transducing nondividing cells such as neurons and does not integrate into the host genome, thereby reducing the risk of insertional mutagenesis. The absence of integration could potentially result in the loss of the expression construct over time; however, in some animal models, central nervous system expression has been stable for more than 6-10 months said Davidson. Furthermore, the transient expression of siRNAs might be sufficient for long-term benefit. The primary problem with this virus, according to Davidson, is that the two serotypes of adenovirus that have been genetically mapped, Ad2 and Ad5, primarily infect glial cells rather than neurons. Davidson’s lab has attempted to alter the target-specificity of Ad2 by replacing its coat fiber with that of Ad17, an adenovirus type that targets neurons. This experiment resulted in fiber instability and low viral titers however, two problems that are frequently encountered when pseudotyping (altering viral capsid or fiber proteins). Kochanek suggested adding polylysine to the Ad2 fiber to simply reduce its target-specificity, rather than attempting to switch it. Davidson said that she had not tried infecting neurons using the adenovirus with a polylysine fiber that was engineered by Wickham. Adenovirus is pervasive and can cause the common cold, which prompted Neil Aronin to inquire about complications of immunogenicity. Davidson and Mulligan responded that adenovirus is indeed highly immunogenic as demonstrated by priming experiments: animals that have had intracranial injections of adenovirus subsequently show skin inflammation when injected subcutaneously with the virus. Davidson said that the acute response to second-generation viruses is limited in the rodent brain however, with modest gliosis that wanes after a few weeks; therefore, if one injection is sufficient for therapy, immunogenicity may not be of major concern. Mulligan stated that if multiple injections were required, the immune response could be minimized by using a virus with a different serotype for each injection. Davidson said that adenovirus has 42 naturally-occurring serotypes, which increases the chances of finding several that are suited for a particular target tissue.

2) Adeno-associated virus

            Adeno-associated virus (AAV) can carry almost 5kb of foreign DNA and can integrate into the host genome, but rarely does so as a vector. AAV comprises at least 8 serotypes, of which AAV2, -4, and -5 have been studied by Davidson’s group. Davidson found that AAV2 specifically targets neurons, AAV5 transduces both neurons and astrocytes and AAV4 infects only empendymal cells. Both Davidson and Bankewicz found that AAV5 diffuses farther and more uniformly than AAV2; however, Bankewicz said that high titers of AAV5 can be difficult to obtain relative to AAV2. Ease of high-titer production is an issue for all of the viral vectors since they have similar threshold concentrations for transduction: injection of  ≥10¹⁰ infectious units/milliliter results in abundant infection while concentrations of <10⁸ infectious units/milliliter does not support any transduction. Of the three virus families discussed, AAV may be the easiest to produce at high titer, noted Davidson. One advantage of AAV over adenovirus is that it is less immunogenic, offered Mulligan and Davidson. Nevertheless it can still cause priming: Bankewicz found that a second injection of AAV2 induces an inflammatory response in the brain.

3) Lentivirus

            Lentiviruses are classified as retroviruses, a family with RNA genomes that integrate into the host genome subsequent to undergoing reverse transcription to DNA. Lentiviruses can infect nondividing neurons and carry up to 9kb of exogenous RNA. Davidson said that three different lentiviruses have been engineered for use in gene delivery: equine infectious anemia virus, feline immunodeficiency virus and human immunodeficiency virus. To target these viruses to neurons, their envelope protein has been replaced with VSV-G, a glycoprotein from the vesicular stomatitis virus. Bankewicz said that both lentiviruses and AAV2 are highly effective in targeting the medium spiny neurons of the striatum. Steven Hersch reminded that for the treatment of HD it may be necessary to transduce cortical neurons as well. Although  Bankewicz expressed skepticism at being able to target the entire cortex, he pointed out that viruses that are capable of retrograde infection may be useful to specifically transduce cortical areas with major input to the striatum. Like AAV, lentivirus is less immunogenic than adenovirus; nevertheless, Mulligan said that Inder Verma’s group has found that the VSV-G coat protein can prime an immune response. Bankewicz said that one advantage of lentivirus is that it does not inhibit a second transduction event, a competitive effect observed with some AAV serotypes. John Rossi added that viral production for human trials would need to be carried out in a GMP (Good Manufacturing Practice) facility such as that of Cell Genesys Inc., which currently manufactures all three of the viruses discussed at this workshop  (Ad, AAV, and lenti).

Synthetic delivery systems

            Workshop discussions also covered non-viral delivery systems for nucleic acids, which are usually less expensive but do not yet compete with the efficiency of viral-based delivery. Jon Wolff used high pressure tail vein injections to introduce siRNA into mice, resulting in effective gene knock-down in the liver and other tissues. Although in hepatocytes the knock-down was not persistent, more stable knockdown might be expected in postmitotic neurons. Further, the same mechanism can be used to introduce plasmid DNA and thereby an siRNA expression construct. Asked whether the mechanism is known by which nucleic acid enter cells in this procedure, Wolff replied that the process is inhibited by polyions but that no specific receptor has been identified. The same hydrodynamic technique may be used to transduce proteins he added, so if it is receptor mediated it is not via nucleic acid receptors. One possibility is that the high pressure causes transient disruption of the cell membrane that allows entry. The requirement for high pressure injection and tissue damage eliminates it from consideration for intracranial application. Arterial injection of siRNAs cannot be used to target the brain because of the blood brain barrier which prevents the non-specific entry of large molecules such as siRNA, estimated by Zamore to be ~12kiloDaltons (plus hydrodynamic radius). Furthermore, according to Wolff, small oligonucleotides are rapidly cleared to the urine, although the addition of polyethylene glycol enhances their persistence. Beverly Davidson asked about fusion to TAT (an HIV-derived peptide that can cross the cell membrane) for improving transduction. Wolff replied that there is no strong evidence that TAT will cross the blood brain barrier and that published results indicating otherwise may be artifactual. Davidson described using mannitol to facilitate penetration of the brain parenchyma. Mannitol increases systemic osmolality, thereby reducing brain pressure and drawing cerebrospinal fluid into the brain; in addition, it increases intercellular space to facilitate diffusion throughout the brain. Davidson’s experiments showed that although mannitol does not allow proteins to cross the blood brain barrier, it significantly enhances their diffusion from the brain ventricles. With this in mind, Davidson, Aronin and Zamore discussed plans to test whether intraventricular injections of siRNA with mannitol treatment could cause gene knockdown throughout the brain.

            Bankiewicz said that he had experienced success using liposomes to deliver proteins to cells. Specifically, he used 30-100 nanometer liposomes to deliver the EGFR protein to cells in the presence of mannitol. Jon Wolff said that a fundamental problem with this technique is that the encapsulating efficiency is only 5-10%. The efficiency of
DNA encapsulation has been improved by using cationic liposomes, but this change impedes release from the capsule, reported Wolff.

Injection technique

            Krys Bankiewicz discussed his efforts to overcome the limited tissue access provided by standard intracranial injections that rely on diffusion. For example, AAV2 diffuses only 2-3 millimeters from the injection site, an acceptable range to target the striatum of rodents but not of monkeys and humans. Using the technique of convection infusion with a specialized cannula, Bankiewicz has succeeded in delivering compounds to the entire primate caudate with a single injection; a second injection site is required for delivery to the putamen since these two structures are separated by the internal capsule in primates. In this procedure, Bankiewicz’s group uses a microdialysis pump that gradually increases the rate of infusion over several hours to overcome intracranial pressure and achieve fluid convection. A telescoping cannula is used to block reflux along the cannula track, and the tip is fine enough to traverse blood vessels without causing hemorrhage. For maximum spread of non-viral therapeutic agents as chemotherapy, volumes of up to 20 milliliters can be infused into the brain parenchyma of humans and up to 0.5 milliliters can be delivered to rats. To target the human striatum with AAV vector, Bankiewicz estimated that volumes of 150-250 microliters should be sufficient. In response to a question from Paul Taylor, Bankiewicz said that he had not measured whether this method resulted in increased intracranial pressure during AAV administration, but this has been performed for small molecule administration by a National Institute of Health (NIH) group. Although Bankiewicz found that convection increased the spread of AAV2 and AAV5 up to 10 fold, he was unsure if this method would be effective with adenovirus and lentivirus which, at a diameter of almost 200 nanometers, are about ten times larger than AAV. Viral spread can be further enhanced by adding heparin to the delivery medium (both phosphate buffered saline and artificial cerebrospinal fluid are satisfactory media according to Bankiewicz and Davidson). Viruses bind to heparan sulfate proteoglycans in the extracellular matrix and competition for this binding by heparin results in a widespread and homogenous infection rather than a gradient from the site of injection. Steven Hersch asked whether the heparin promotes bleeding but Bankiewicz said that the concentration required for enhancement of AAV distribution has not been associated with bleeding.

Concluding Agreements

            The meeting concluded by prioritizing experiments toward the development and delivery of therapeutic RNAs for HD. First, multiple siRNAs need to be tested to identify which sequences most effectively disrupt htt expression. This can be carried out in HeLa cells, an immortalized human cell line that is easy to propagate and transfect and, according to Zamore, expresses htt. Second, the function of selected siRNAs must be verified in an animal model, preferably the R6/2 mouse line which has the most rapid disease onset. R6/2 only carries the 5’ portion of mutant HD, however, making it unsuitable to test siRNAs directed to other regions. All potentially therapeutic siRNAs should further be tested in mouse models with full-length htt that have slow disease progression and might more faithfully model human HD. Finally, siRNA-treatment of wildtype mice may be used to test for consequences of disrupting the normal HD allele.

            Anatomical details of therapeutic RNA delivery should be tested in primates where the brain architecture and volume are similar to that of the human.  Critical parameters are the volume, speed and pressure of vehicle delivery. The area of tissue penetration and percent of neurons transduced can be monitored by reporter genes such as green fluorescent protein and thymidine kinase. Safety tests for adverse effects of siRNA therapy may also be carried out in primates. Testing experimental treatments on the most desperate HD patients faces the problem that in late-stage HD the striatum has largely degenerated. Thus the neurons targeted for therapy are mostly gone and are replaced by tightly packed glial cells that impede drug delivery. Furthermore, late-stage patients are often incapable of informed consent. Although prior consent may be obtained, the procedures must be exactly as described in the original protocol- a potential problem in the rapidly changing field of gene therapy. Thus, early-stage HD patients might be the best suited for therapies to rescue striatal neurons.