|Home | About | Journals | Submit | Contact Us | Français|
Genetic therapy is undergoing a renaissance with expansion of viral and synthetic vectors, use of oligonucleotides (RNA and DNA) and sequence-targeted regulatory molecules, as well as genetically modified cells, including induced pluripotent stem cells from the patients themselves. Several clinical trials for neurologic syndromes appear quite promising. This review covers genetic strategies to ameliorate neurologic syndromes of different etiologies, including lysosomal storage diseases, Alzheimer's disease and other amyloidopathies, Parkinson's disease, spinal muscular atrophy, amyotrophic lateral sclerosis and brain tumors. This field has been propelled by genetic technologies, including identifying disease genes and disruptive mutations, design of genomic interacting elements to regulate transcription and splicing of specific precursor mRNAs and use of novel non-coding regulatory RNAs. These versatile new tools for manipulation of genetic elements provide the ability to tailor the mode of genetic intervention to specific aspects of a disease state.
Genetic therapy covers a range of methods for modifying the nervous system, including delivery of genes, sequence-targeted regulatory molecules, genetically modified cells and oligonucleotides. In this review, we focus on ways to change the genetic status of the nervous system, including routes of access and modes of delivery. Examples are provided of strategies used for a number of diseases, including recessive loss of function conditions [lysosomal storage diseases and spinal muscular atrophy (SMA)], dominant toxic mutations [amyotrophic lateral sclerosis (ALS)] and conditions of mixed etiology [Alzheimer's disease (AD), Parkinson's disease (PD) and brain tumors] (Fig. 1). Neurodegenerative diseases caused by triplet nucleotide repeats are discussed in an article on siRNAs in this issue. Many of these strategies have proven very promising in preclinical studies in mouse and larger animal models, and a substantial number are now being evaluated in clinical trials.
Entrance to the central nervous system (CNS) is limited by the skeletal structures that enclose it and the blood–brain barrier (BBB), which plays a central role in regulating the chemical microenvironment of the CNS by preventing simple diffusion of most small molecules and proteins from the bloodstream. The physical barrier properties of the BBB are a result of the tight junctions between brain microvascular endothelial cells and low vesicular transport (1; Fig. 2). As a result, most molecular traffic in and out of the CNS is tightly regulated by specialized transporter systems present on the luminal and abluminal membranes of the endothelial cells, with the exception of small solutes such O2 and CO2 gases and some lipophilic molecules (e.g. ethanol) that can diffuse freely across cellular membranes. Other components of the BBB are the basal lamina, astrocyte end-feet that surround all CNS blood vessels and pericytes. Interestingly, pericytes appear to play a central role in regulating BBB permeability via their influence over endothelial cells and astrocytes (2). Not surprisingly, the BBB has proved exceptionally efficient in excluding the vast majority of gene transfer vehicles from reaching the CNS via the vasculature. As a result, most CNS gene therapy approaches have utilized direct infusion of gene transfer vectors into the brain parenchyma to target disease-relevant structures. The distribution of viral vectors in the brain can be improved considerably by convection-enhanced delivery (CED), a slow pressurized infusion of a larger volume (3,4). Nonetheless, transduction of CNS cells after intraparenchymal injection of viral vectors remains mostly limited to the targeted structure.
An alternative route of entry into the CNS is by delivery of the gene transfer vectors directly into cerebrospinal fluid (CSF) via either the lateral ventricles (5) or intrathecal space (6). This approach is exceptionally effective in achieving widespread transduction in the neonatal mouse brain (5,7), but the distribution pattern is more restricted in adult animals (6,8,9). This suggests that there may be additional barriers to distribution of gene transfer vectors from CSF into adult brains, or that commonly used injection procedures disrupt normal CSF flow preventing widespread distribution. Delivery of recombinant lysosomal enzymes into CSF appears to be effective in providing therapeutic levels of these enzymes throughout the CNS (10,11). Similarly, CSF infusion of antisense oligonucleotides (ASOs) directed at altering the splicing pattern of the SMN2 gene has been dramatically effective in a mouse model of SMA (see below).
An ideal route of entry into the CNS to achieve widespread gene transfer would be through the vasculature. Until recently, the only exception to the BBB-imposed block to gene therapy vectors was PEGylated immunoliposomes (PILs) formulated with monoclonal antibodies specific for receptors, such as transferrin and insulin receptors which mediate transcytosis of their ligands across the BBB. These PILs appear to be quite effective in delivering expression plasmids (with expression under cell type-specific promoters) and RNAi to normal brain or brain tumors (12). In a parallel approach, therapeutic proteins are delivered to the CNS using chimeric recombinant molecules, e.g. growth factors, single-chain antibodies or lysosomal enzymes, fused to receptor-targeting monoclonal antibodies (12) or ligands, such as transferrin (13). Recently, adeno-associated virus 9 (AAV9) vectors have been found to enter the CNS of neonatal mice and young cats after intravascular (i.v.) infusion and to transduce large numbers of glia and motor neurons in the spinal cord (14,15). Transduction of other neuronal populations in the brain is found mostly in the hippocampus and Purkinje cells in the cerebellum (14). SV40 recombinant vectors also appear to mediate efficient gene transfer to certain regions of the CNS after i.v. infusion in adult mice combined with intraperitoneal mannitol infusion (16). Alternative approaches to deliver secretable therapeutic proteins to the CNS (e.g. lysosomal enzymes, growth factors, cytokines, tumor killing agents) are to target gene transfer to brain microcapillary endothelial cells (17), or use ex vivo genetically modified stem cells [hematopoietic stem cells (HSCs), neural stem cells, mesenchymal stem cells] which will themselves, or their progeny in the case of HSCs, migrate within the brain to regions of injury or tumors (18–20). In fact, the whole nervous system can be transduced with a gene for the lifespan by injecting a lentivirus vector into the amniotic sac of mouse embryos before the neural groove has closed (21).
The use of non-viral nucleic acid delivery to selectively modify cellular processes within the brain presents a number of challenges, including target cell specificity and transient gene expression duration. Pardridge and colleagues (22) have developed PILs target to the brain when administered intravenously (see above). Stachowiak et al. (23) have recently reported the use of organically modified silica-based nanoparticles to induce neurogenesis within the subventricular zone of adult mice via intraventricular delivery of DNA encoding a recombinant nuclear form of fibroblast growth factor receptor-1. Electroporation-based nucleic acid transfection has also been extensively documented in the prenatal and postnatal rodent brain [reviewed by De Vry (24)], providing the means to genetically modify significant numbers of neurons (25,26) within the living brain (27).
Recombinant viral vector systems remain the most efficient vehicles to achieve long-term stable gene expression in the CNS. Over the years, many different viral vector systems have been investigated for this purpose, including those derived from herpes simplex virus type 1 (HSV-1), adenovirus, AAV, lentiviruses—such as HIV-1, feline immunodeficiency virus or equine infectious anemia virus, and more recently SV40. AAV and lentivirus vectors have emerged as the vectors of choice for gene transfer to the CNS for non-oncological applications as they mediate efficient long-term gene expression with no apparent toxicity. A recent study has shown that AAV-mediated transgene expression in the primate brain continues for at least 8 years with no evidence of neuroinflammation or reactive gliosis (28). Moreover, several clinical trials have shown that direct infusion of AAV2 vectors into brain parenchyma in humans is well tolerated (29–33). The safety profile of direct infusion of lentivirus vectors into human brain remains to be evaluated. High-capacity adenovirus vectors are attractive because of their large transgene capacity (~30 kb) and ability to mediate long-term gene expression without immunological complications (34). These and the large capacity HSV amplicon vectors (150 kb; 35) may be ideal choices to transfer large genomic regions necessary to achieve physiological regulation of gene expression for particular genes/sets of genes in specific cell populations in the CNS. Full-length gene copies with intact regulatory elements (>100 kb) can by delivered in HSV-1 amplicon vectors (36), verging on the capacity to deliver virtual mini-chromosomes (37). The main difficulty for high-capacity adenovirus and HSV-1 amplicon vectors lies in difficulties in large-scale production. Recombinant SV40 vectors display some promising properties, namely their apparent ability to cross the adult mouse BBB after i.v. infusion and transduce neurons and astrocytes in specific brain regions and spinal cord (16). Moreover, these vectors can be administered repeatedly as they do not seem to elicit neutralizing antibodies in rodents. Whether this property is shared in other mammalian species remains to be determined. In the neuro-oncology field, recombinant vectors derived from HSV-1, adenovirus, measles virus and Newcastle virus appear to be the most promising in pre-clinical models of brain tumors (see below). After initial attempts to broadly use each viral vector system for any/all CNS gene transfer applications, now the particular strengths of each system are being exploited to meet the specific needs of different applications/diseases.
Lysosomal storage diseases are typically, but not exclusively, childhood diseases resulting from a genetic deficiency in a lysosomal enzyme involved in a particular metabolic pathway that results in lysosomal accumulation of its substrate(s). Many of these enzymes can be secreted from cells genetically engineered to overexpress them and then taken up by enzyme-deficient cells and correctly targeted to lysosomes via mannose-6-phosphate receptors where they degrade the stored substrate(s). This mechanism, known as cross-correction (38), is the basis for enzyme replacement therapy which is now available for a subset of these diseases without neurological involvement. Unfortunately, the BBB prevents recombinant lysosomal enzymes infused peripherally from entering the CNS. Alternative routes of delivery and molecules are being actively investigated (see above). These diseases are particularly well suited for gene therapy as they are monogenic diseases with very well established genotype–phenotype correlations, and the cross-correction mechanism makes it possible to devise gene delivery strategies to supply essentially the entire CNS with therapeutic levels of these enzymes. Two main approaches have shown dramatic therapeutic effects in animal models of lysosomal storage diseases, namely intraparenchymal infusion of recombinant viral vectors (39–46), and bone marrow transplantation with ex vivo lentivirus vector modified-autologous HSCs (see article by Biffi, Cartier and Aubourg in this issue).
The ability of focal viral vector-mediated gene delivery to supply the CNS with therapeutic levels of lysosomal enzymes is dependent on widespread distribution based on diffusion (47), axonal transport over long distances (48,49) and CSF flow in the perivascular space (8). AAV-mediated genetic modification of highly interconnected structures in the brain, such as deep cerebellar nuclei (50), ventral tegmental area (51) or thalamus (4,52), leads to widespread distribution of these enzymes in the CNS. Alternative targets, such as the external capsule (53) or lateral ventricles (8), take advantage of either interstitial fluid flow or CSF flow for distribution of lysosomal enzymes throughout the CNS. Successful translation of AAV-based (or lentivirus-based) approaches to humans will likely require targeting one or more of these structures to achieve therapeutic levels of these enzymes throughout the CNS. Pre-clinical studies in large animal models of some of these diseases are highly encouraging (54,55). Bone marrow transplantation with lentivirus-modified autologous HSCs has shown exceptional results in different mouse models of lysosomal storage diseases resulting in correction of pathologic findings throughout the CNS (18,56,57), as well as peripheral organs (58). This approach relies on genetically modified HSC-derived cells (macrophages in the case of CNS) trafficking to the sites of disease and becoming an in situ source of recombinant enzyme. Both of these gene therapy approaches are now being tested in human clinical trials for different lysosomal storage diseases.
AD is a neurodegenerative disorder characterized by severe memory loss and cognitive impairment with no available cure. Neuropathological correlates include extracellular amyloid-beta peptide deposition, intracellular neurofibrillary tangle formation, decreased synaptic integrity and neuronal loss. The basal forebrain cholinergic complex is significantly affected by AD-related neurodegeneration (59–63). To augment cholinergic function, Tuszynski and colleagues (64–68) delivered nerve growth factor (NGF), the prototypical neurotrophin with demonstrated neuroprotective properties, using retrovirus vector-transduced fibroblast grafts and demonstrated restoration and survival of cholinergic neurons in lesioned rodents and aged non-human primates. These initial studies set the stage for the first phase I clinical trial of ex vivo NGF gene therapy (ClinicalTrials.gov Identifier: NCT00017940). The rate of cognitive decline was slowed, and no apparent detrimental effects were observed arising from NGF expression 22 months post-engraftment (69). More recently, Ceregene has conducted phase I and II clinical trials using an AAV vector that expresses NGF (ClinicalTrials.gov Identifiers: NCT00087789 and NCT00876863, respectively). While the phase I trial demonstrated that stereotactic infusions of an NGF-expressing AAV vector is well tolerated, it is too early to know whether this gene therapy-based strategy will significantly impact the course and symptomology of the disease, given the phase II trial is currently ongoing.
During the past several years, promising efforts have focused on reducing the levels of neurotoxic Aβ peptide species within the brain through removal of pathogenic Aβ peptides or halting the proteolytic release of the amyloidogenic form of Aβ arising from pathogenic amyloid precursor protein (APP) processing. Such preclinical vector-based therapies include active and passive vaccines directed against specific epitopes of pathogenic Aβ peptides (70–75), expression of Aβ proteases (75–78), delivery of small inhibitory RNAs (RNAi) designed to specifically suppress the expression of APP processing enzymes (79–82) and delivery of a cholesterol degrading enzyme into the brain (83). Each experimental approach has exhibited strong preclinical efficacy in rodent models of AD, hence increasing enthusiasm for eventual translation of one or more of these strategies to clinical testing. The common challenge shared by gene therapeutic approaches for AD which require intraparenchymal delivery relates to sufficiency of brain tissue coverage. Given AD impacts a number of human brain sub-regions that are integral to learning and memory, viral vector strategies must include the means to monitor vector infusion in real time to widely and safely disseminate vector particles within afflicted networks. Recent advances in real-time monitoring of CED-mediated AAV vector infusions within the non-human primate brain indicate that this delivery hurdle can be overcome (84).
PD is a common progressive neurodegenerative disease. While many regions of the brain are affected, the primary motor symptoms are caused by the loss of dopaminergic neurons in the substantia nigra innervating the striatum in the basal ganglia (85). Although environmental toxins and aging are risk factors, genetic susceptibility is critical with 16 gene loci implicated in ~5% of cases, and multifactorial hereditary risk in many others (85,86). These risk factors point to oxidative stress, mitochondrial dysfunction and reduced ability to degrade abnormal proteins as etiologic factors. Although L-dopa has been used for decades to relieve motor symptoms of PD, it does not prevent degeneration and eventually ceases to be effective.
New gene/cell therapy strategies have been explored experimentally with some translated into clinical trials, including: delivery or upregulation of neurotrophic factors, generation of endogenous dopamine and alterations in neuronal circuitry, as well as implantation of supportive cells and dopaminergic neurons. Gene delivery has typically been carried out using AAV vectors with CED injection into the striatum. Vectors have delivered both glial-derived neurotrophic factor (GDNF; 87) and its analogue, neurturin (88), which serve as trophic factors for dopaminergic neurons. Since abnormally high GDNF can have toxic effects, efforts have gone into regulating its expression through a tetracycline analogue system (89) and by using highly specific zinc finger transcription factors to upregulate expression of the endogenous gene, with the latter imposing a physiologically controlled upper limit (90).
In terms of modulating neurotransmitter pathways, all three biosynthetic enzymes for dopamine have been included in the same lentivirus vector, including tyrosine hydroxylase (TH) and aromatic amino acid decarboxylase (AADC), as well as GTP cyclohydrolase for synthesis of the biopterin co-factor for TH, with the assumption that dopamine, as a paracrine neurotransmitter, can be made by any cell in the striatum (91). However, serotonergic or GABAergic neurons can also produce dopamine as a false transmitter leading to dyskinesia (92). AAV-mediated delivery of AADC alone has the advantage of making production of dopamine dependent on L-dopa intake, with ongoing phase I trials indicating continued expression of AADC for at least 2 years in the human brain (93). Investigators have also tried to block hyperactive neurotransmission in PD brains using an AAV vector to deliver the enzyme glutamic acid decarboxylase for synthesis of the inhibitory neurotransmitter GABA into the subthalamic nucleus (94). Cell therapies have included a number of cell types, such as human fetal mesencephalic tissue, which includes precursors of dopaminergic and serotonergic neurons, with the current focus on human dopaminergic neuroblasts generated from stem cells, in particular induced pluripotent stem (iPS) cells derived from adult tissue (95).
Strangely, few of these therapies address the etiology of the disease based on genetic insights, and in fact most animal models of PD employ lesioning of dopaminergic neuronal connections, with the exception of transgenic mice overexpressing alpha-synuclein and recently developed models of other PD genetic syndromes (96). Since three copies of the alpha-synuclein gene alone can cause PD (97) and upregulated alpha-synuclein inhibits neurotransmitter release (98), a logical approach would be to decrease alpha-synuclein synthesis with RNAi (84) or increase its degradation with the ubiquitin ligase, parkin (99). However, both the increased and decreased levels alpha-synuclein can cause neurodegeneration (84).
SMA is an autosomal recessive disease caused by the loss of function of the ‘survival of motor neuron' gene, SMN1, which leads to degeneration of motor neurons and infant mortality. One approach to gene therapy would be to replace the missing gene in multiple motor neurons, which may now be possible with i.v. administration of AAV9 vectors which were able to deliver the SMN1 cDNA to the spinal cord in a mouse model of SMA with marked correction of motor function and a dramatic increase in survival (100,101). However, scaling delivery from mice to humans remains a huge challenge. Other strategies have taken advantage of the presence of a second copy of this gene, SMN2, coding for an identical amino acid sequence in the human genome (102). The SMN2 gene is defective due to a point mutation in an intron which interferes with correct splicing, such that a truncated, non-functional protein is produced. In one therapeutic modality, an ASO homologous to the last translated exon in SMN2 is used to guide an exonic splicing enhancer sequence to the region as a binding platform for pre-mRNA splicing factors (103). This strategy gives increased SMN2 expression when infused into the lateral cerebral ventricles in an SMA mouse model (104,105). In another modality, transplicing is utilized in which the ASO binds to the defective intronic sequence and carries a splicing domain and an intact final exon from SMN1, with injection into the cerebral ventricles also extending survival in a mouse model of SMA (106).
ALS is characterized by progressive muscle weakness resulting from the loss of motor neurons in the brain and spinal cord. Mutations in the SOD1 gene encoding cytosolic Cu/Zn superoxide dismutase were the first to be linked to familial forms of ALS (fALS), ~10% of all cases. In recent years, there has been a dramatic increase in the number of genes shown to be associated with fALS, such as ANG encoding angiogenin (107), TARDP encoding transactive response (TAR) DNA-binding protein TDP-43 (108), FUS encoding fused in sarcoma protein (also known as TLS for translocated in liposarcoma) (109,110) and, more recently, optineurin (111). Several other genes have been associated with rare forms of fALS (reviewed in 112). The disease mechanism(s) remains elusive. However, the fact that many fALS cases are autosomal dominant suggests that disease-associated mutations generate protein species with toxic functions instead of simple loss of function. An emerging picture in the field is that some of the fALS-associated proteins may also be involved in sporadic ALS (sALS) cases. As an example, recent work has shown that oxidized wild-type SOD1 shares a conformational epitope with fALS-associated mutant SOD1 (113), and can be found in spinal cord motor neurons of a subset of sALS patients (114). Moreover, oxidized wild-type SOD1 and mutant SOD1 share toxic properties to neurons by inhibition of kinesin-dependent fast axonal transport (114). Also, misfolded SOD1 mutants have been shown to directly bind and inhibit mitochondrial voltage-dependent anion channel (VDAC1) leading to mitochondrial dysfunction (115). Taken together, these results suggest that conformational changes in SOD1 (and possibly also other fALS-associated proteins), caused by mutations associated with some forms of fALS, and other as yet undetermined factors, may be the basis for fALS and sALS. Gene therapy approaches for ALS have centered on delivery of growth factors such as IGF-1 (116–118) and vascular endothelial growth factor (116,119), and RNAi-based silencing of mutant SOD1 alleles (120–123) in transgenic ALS mouse models. The most commonly used ALS mouse model carries a human SOD1 G93A transgene expressed at high levels with a mean survival of ~129 days. To date, many of the gene therapy studies have demonstrated a significant increase in median survival, but ultimately all treated mice have succumbed to disease progression. The reasons for the apparent failure of those interventions can be multifactorial, but an interesting aspect to consider is the fact that silencing mutant SOD1 expression in motor neurons delays disease onset but not progression, while silencing it in microglia does the opposite (124). This suggests that other CNS cell types contribute significantly to the disease phenotype. Development of new mouse models using mutant alleles from other ALS genes will enhance our understanding of common disease pathways and development of effective gene therapies for this disease. The recent findings on the toxicity of misfolded SOD1 proteins (normal or mutant) strongly suggest that either immunization (125), or passive immunization with conformation-specific antibodies (126) may also be viable approaches to develop effective therapies for ALS.
Malignant glioma tumors (glioblastoma; GBM) are the most common adult brain tumor and are virtually untreatable, with most patients dying within 2 years of diagnosis in spite of neurosurgery, radiation and chemotherapy. These tumors are very invasive in the brain and genetically heterogeneous with a cancer stem cell population that defies all current chemotherapies (127). Some benign tumors of the nervous system, e.g. vestibular schwannomas, regress in response to anti-angiogenic therapy (128), while for GBM tumors this treatment can enhance invasiveness without suppressing proliferation (129).
Treatment of malignant brain tumors presents a major therapeutic challenge requiring multicombinatorial approaches, with gene/cell therapy showing promise as adjunct therapies. Experimental therapies in brain tumor models have focused on direct elimination of tumor cells, including a ‘bystander' effect to confer selective toxicity to non-transduced tumor cells in the vicinity and on targeting invasive tumor cells. The basic strategy is to remove the bulk of the tumor mass, and during that intervention to inject virus vectors or cells into the resected region to kill remaining tumor cells over an expanded radius. Concepts employed in this arsenal include: (i) Prodrug activation (or suicide genes). In this case, viral vectors or cells are used to express enzymes, such as viral thymidine kinase (130), bacterial cytosine deaminase/uracil phosphoribosyltransferase (131) or mammalian cytochrome P450/carboxyesterase (132) within the brain to activate pro-drugs (ganciclovir, 5-fluorocytosine, cyclophosphamide/irinotecan, respectively) that can pass the BBB and be converted into active chemotherapeutic agents within the tumor. (ii) Viral oncolysis. In this approach mutant forms of viruses, typically HSV-1 and adenovirus, are used which replicate selectively in tumor versus normal brain based on cancer-related mutations or their increased proliferation rate (133). These oncolytic vectors can also be armed with therapeutic genes. An increasing number of different viruses are being tested in this context, including measles virus, reovirus, Newcastle disease virus (134), polio virus (135) and vaccinia virus (136). Although promising, this field is still wrestling with immune responses to the virus and an unfavorable tumor microenvironment which restrict virus spread within the tumor (134). (iii) Cellular delivery. Several studies have shown that some normal cells introduced into the brain, such as neural stem cells (137) and mesenchymal stem cells (138) migrate towards tumor foci. These cells can deliver agents which are selectively toxic to tumor cells, e.g. antibodies against tumor antigens (139,140), oncolytic virus vectors (141,142), apoptotic factors (143) and anti-angiogenic proteins (144), as well as prodrug activating enzymes. (iv) Immunotherapy to target tumor antigens. Several strategies have been used to increase recognition of tumor antigens and empower the immune system. These include vaccination with a common and unique GBM antigen, EGFRvIII (145) and co-treatment with oncolytic HSV vectors and cyclophosphamide to temporarily suppress the immune curtailment of virus spread, and inclusion of immune enhancing cytokines (146). In addition, T cells have been modified to express chimeric receptors targeting antigens, such as the IL13 receptor which is abundant on GBM tumors (147). (v) Zone of resistance. Other efforts have tried to increase the resistance of the brain to tumor invasion, including the use of AAV vectors to infect normal brain cells so they release interferon-beta which depresses tumor growth (148).
A number of new genetic strategies are being explored to manipulate the endogenous cell genome in order to regulate expression or correct mutated gene transcripts. In one modality, zinc finger proteins (ZFPs) which recognize a typically 18 bp site in the promoter region of a gene are linked to a transcriptional activator (149) to selectively turn on a gene, as in the case of SMN2 (above; 90). This approach can also be used to stimulate gene correction by linking the ZFP to a nuclease causing double-stranded breaks in the DNA in combination with a correct extrachromosomal DNA sequence such that homologous recombination is stimulated (149). Another potent tool that can be linked to ZFPs are meganucleases which recognize >12 bp sequences and have been shown to inactivate HIV sequences (150). Other methods to facilitate gene correction have included triplex-forming oligonucleotides which can be used to modify sequence or inhibit gene expression (151). Gene correction has also been achieved using single-stranded DNA (73 bp; 152). In addition, there is the potential to deliver human artificial mini-chromosomes into the cell nucleus for gene-deficiency syndromes (153,154). These strategies include control of transplicing, for example in SMA therapies (see above). Naturally occurring and synthetic transposable elements, including the Sleeping Beauty transposon, have shown promise in stable integration and gene expression in cells, including neurons in vivo, and are discussed in more detail elsewhere in this issue. All these more ‘advanced' manipulations of the genome will depend on large vector capacity and more efficient delivery to the nervous system. A new arena will be modulation of miRNA levels which have such an important role in development and cancer, including decreasing levels of oncogenic miRNAs and increasing levels of tumor suppressor miRNAs (155). Epigenetic modulation of the genome is an expanding arena, with for example histone deacetylase inhibitors being used to slow down degeneration in mouse models of Huntington's disease (156).
The new breed of viral vectors that can reach the CNS after peripheral administration (14,16) also transduce peripheral organs at high efficiency, namely the liver. This potential limitation could be simply addressed with the use of CNS-specific promoters such as the synapsin-1, neuron-specific enolase or glial fibrillary acidic protein promoters. However, some of these promoters are relatively weak and some are rather large, thus reducing considerably the transgene capacity of the vectors. An alternative approach is to incorporate into the transgene expression cassette targets for miRNAs expressed at high levels in peripheral organs but absent or expressed at low levels in the CNS to prevent vector-mediated gene expression outside of the CNS (157, see below). This approach appears to be quite universal to de-target vector-mediated expression from specific cell lineages (57,158,159). Another approach to addressing the undesired targeting of peripheral organs consists in engineering the viral vector particles themselves. Lentivirus vectors can be pseudotyped with envelope proteins derived from other viruses, and the vesicular stomatitis virus glycoprotein is the most commonly used due to its pantropism. Previous studies have shown that incorporation of the rabies glycoprotein allows lentivirus vectors administered peripherally to reach the CNS via retrograde axonal transport (119,160). Non-integrating lentivirus vectors can also be used effectively in the neural cells, thus reducing the risk of oncogenic insertion events (161).
Development of CNS-targeted AAV vectors has followed three different routes: (i) in recent years hundreds of new AAV capsids have been identified from humans and quite a few other species, and systematic screening for new CNS-targeting properties may yet yield new AAV pseudotypes more powerful than AAV9; (ii) chimeric AAV2 capsids carrying CNS-targeting peptides identified from in vivo phage display experiments can increase delivery. This approach had met with limited success until recently Chen et al. (17) showed that it can be used to develop AAV vectors highly specific for brain microcapillary endothelial cells that are part of the BBB; (iii) molecular evolution of AAV vectors has been propelled by using new AAV capsid libraries developed by DNA shuffling of existing AAV capsid genes (162–164). This approach has been used to generate new AAV vectors with improved transduction properties for particular targets, such as lung epithelium (165), myocardium (166) and more recently to a particular subset of neurons in the piriform cortex in the brain following seizure-induced temporary disruption of the BBB (167). This selection is performed in live animals, and one of the highly significant findings has been newly identified chimeric AAV capsids with dramatically reduced transduction of liver upon systemic infusion. The work by Kumar et al. (168) where a rabies virus glycoprotein-derived peptide was used to target an artificial siRNA to the CNS suggests that the distinction between artificial/synthetic and viral vectors is likely to become increasingly blurred as both sides borrow from each other to achieve the ultimate goal. AAV-phage (AAVP) vectors are another example of this melding of components with target peptide-displaying M13-derived phage capsids carrying genomes with mammalian expression cassettes flanked by AAV2 inverted terminal repeats (169). These AAVP vectors appear to be quite effective at targeting brain tumor vasculature (170).
The recent developments in engineering new CNS-targeted vectors give great optimism that within the next decade we will finally achieve the long-standing goal of global gene delivery to the post-natal CNS via the vasculature. This will undoubtedly revolutionize neuroscience research, and mark a new era in neurology with the genetic tools existing to develop effective therapies for many conditions that remain untreatable today. But before then, we already have exceptional tools at hand to change the outcome of many neurological diseases using viral vectors for focal gene delivery to particular structures in the CNS. Most work thus far has relied on the use of strong constitutive promoters, but in many instances it may be necessary to regulate gene expression. This has been accomplished to a large extent by incorporating into the gene transfer vectors a combination of drug-responsive transcription factors and respective promoters, thus being able to regulate transgene expression in the CNS by peripheral administration of a particular drug. The most commonly used and investigated systems for this purpose are the tetracycline-responsive (89,171–173) and rapamycin-responsive systems (174,175). These drug-regulated viral vector systems have yet to be tested in the CNS of patients.
Neuronal cell death is associated with most of the major diseases that afflict the CNS. Stem cell therapeutics holds promise for replacing degenerating or ablated neurons to ultimately restore neuronal network integrity. While embryonic stem cell-based therapy has been approved for clinical trial testing in patients with spinal cord lesions (176), the ethical implications and funding policy inconsistency associated with the use of stem cells isolated from human embryos for CNS therapy have led to the rapid development of new cell sources, including iPS cells (Fig. 3). iPS cells require the controlled ‘reprogramming' of adult somatic cells through careful introduction of key molecular cues and re-derivation of viable clones that lack tumorigenic potential (177,178). Viral vectors, including those derived from retrovirus and lentivirus, have proved useful for efficient delivery of genes encoding these molecular cues (Oct3/4, Sox, Klf, Myc, Nanog and LIN28) to adult human fibroblasts (179–181). The challenges facing the clinical implementation of iPS cells for neurodegenerative diseases relate to achieving an optimal balancing of reprogramming factors [reviewed by Lewitzky and Yamanaka (182)], lowering risk of integrating vector-mediated gene disruption (183,184) and prevention of teratoma formation, which is an inherent risk of iPS cell derivation (185,186).
In the past 20 years, genetic therapy has moved from a dream to a near medical reality for some diseases, with marked beneficial effects on immune function in X-linked SCID cases (187) and restoration of some vision in Leber's optic atrophy (188). There have been many hurdles along the way, including toxicity of vectors, oncogenic insertions into the genome, immune rejections and difficulties in scaling up from mouse models to humans. The burgeoning successes have depended in large part on genetics, including identifying disease genes and types of mutations, design of genome-interacting elements, manipulation of splicing events and novel genetic elements, such as microRNAs and regulatory non-coding RNAs. Critical factors lie in tailoring the mode and content of the genetic vehicle to the specifics of the disease, and effectively utilizing the broad range of targeting modalities and vector types available.
Support for X.O.B. comes from NIH/NCI CA069246 and NIH/NINDS NS242791, for W.J.B. from NIH R01-AG026328 and for M.S.-E. from NIH/NINDS R01NS066310, U01NS064096 and NIH/NICHHD R01HD060576.
We thank Ms Suzanne McDavitt for skilled editorial assistance, and Ms Emily Mills at Millstone Design (http://www.millstone-design.com) for preparation of figures.
Conflict of Interest statement. None declared.