Abstract: Given improvements in viral vector design, production and efficiency of transduction in the central nervous system (CNS), as well as increased knowledge of neuropathological mechanisms in neurological disorders, success in treating a CNS disorder with gene transfer seems inevitable. Several different vector systems have been studied extensively and the adeno-associated viral vector system has been utilized in most early stage clinical trials in neurological disorders. Other vector systems, such as lentivirus, adenovirus, and herpes simplex virus are also viable vector platforms that should fill significant clinical niches based on their specific characteristics. In addition to the choice of the appropriate vector, the proper choice of transgene for the appropriate strategy to treat a neurological disorder is also critical. The example of glial cell line-derived neurotrophic factor ligands to treat Parkinson's disease is used to illustrate the importance of the interface between interpretation of pre-clinical data and consideration of the natural history of the disorder. This interface dictates the proper design of clinical trials that are capable of testing whether the treatment is actually successful.
The design, production, and efficiency of gene transfer vectors, especially for transduction of the central nervous system (CNS), has improved remarkably over time, leading to safer transduction and long-term and robust transgene expression in the brain. Considering these improvements, coupled with a progressive understanding of neurodegeneration and the underlying molecular mechanisms (i.e., neuropathology), it is reasonable to expect clinical progress in gene therapy for neurological disorders. However, despite a growing number of successful phase I CNS gene therapy clinical trials, clear clinical efficacy has been lacking, and few trials have reached phase 2 (Lim et al., 2010). The reasons for this apparent failure to successfully apply gene therapy to neurological disorders are elusive; however, some clues may be found while dissecting the progress of translational science, starting with pre-clinical animal studies and ending by re-examining the way these data were applied to design the clinical trials. Many disorders lack animal models that faithfully recapitulate the true etiology of the various disease states such as Parkinson’s Disease (PD), thus, any plausible clinical efficacy as well as potential safety issues cannot be fully predicted. This review will discuss general approaches and considerations one must take into account when considering to develop direct viral CNS gene therapy. Furthermore, we will specifically discuss the issues outlined here in the context of recent gene therapy clinical trials for PD, the second most common neurodegenerative disease, and why novel expression systems may be required to successfully accomplish trials such as these in the future.
Gene Therapy: Basic Experimental Design Choices
Current gene transfer approaches in the treatment of neurological disorders can be divided into four different conceptual modes of action: gene replacement, gene knockdown, non-specific (pro-survival) therapy, and cell suicide therapy (Figure 1B). Each of these gene therapy strategies have safety and efficacy characteristics that must be considered when performing preclinical translational research and planning clinical trials.
If the disease is the result of a loss of function mutation, a simple gene replacement approach may be suitable. For example, treatment of a recessive metabolic disorder such as late infantile neuronal ceroid lipofuscinosis (LINCL) has been tested clinically using gene delivery of a functional ceroid lipofuscinosis, neuronal type 2 (CLN2) gene which encodes the lysosomal protease tripeptidyl-peptidase I (Worgall et al., 2008). In the future, such therapies may hold promise in other neurodegenerative disorders caused by recessive mutations as well. For instance, virally mediated over-expression of parkin, a gene linked to a recessive form of PD, is protective in toxin induced rodent models of the disease (Manfredsson et al., 2007; Vercammen et al., 2006); however, to date, the function of many genes linked to neurological disorders are unknown.
In contrast to loss of function mutations, if the resultant neuropathology is due to a toxic gain of function, or the aberrant accumulation of a gene-product, removal of that gene, perhaps in conjunction with a corrected gene replacement, would be the option. Many neurological disorders are characterized by the intracellular or extracellular accumulation of proteins, including the Lewy Bodies seen in PD, amyloid plaques and neurofibrillary tangles seen in Alzheimer’s Disease, and neuronal inclusions in Huntington’s Disease. Although it is unclear whether or not these structures are pathological or protective in nature, they are putative targets for various gene therapy knockdown tools such as siRNAs and microRNAs (Manfredsson et al., 2006). Animal studies indeed suggest that removal of these accumulating proteins may be neuroprotective in certain conditions (Kim and Rossi, 2007). Furthermore, several genes linked to neurological disorders whose mutations are proposed to have caused a toxic gain-of-function are putative knockdown targets. Of course, it is possible that removal of the abnormal protein in question, albeit toxic, may in itself result in pathology due to loss of its potential normal protein function (Mandel et al., 2006). Thus, knockdown of the aberrant protein can be combined with the concomitant delivery of the corrected wild-type gene.
Pro-survival/symptomatic gene therapy
The third option, which has been utilized clinically in the treatment of several neurological disorders, is the use of non-specific pro-survival genes which, regardless of underlying molecular mechanisms, promote cell-survival during pathological conditions. One such approach has been the use of glial cell line derived neurotrophic factor (GDNF), or its close relative neurturin (cere-120®) (Marks et al., 2008), in the treatment of PD. Members of the GDNF family of ligands (GDFLs) have been shown to be absolutely required for the survival and maturation of dopaminergic neurons. This process is mediated through the akt and erk pro-survival and neurite outgrowth signal transduction pathways (Pascual et al., 2008). In addition to a neurotrophic approach, gene therapy aimed at producing symptomatic benefit without halting the disease process has also been attempted. In the case of PD, over-expression of key enzymes involved in the production of the neurotransmitter dopamine (DA) has been attempted in lieu of oral administration of L-DOPA (Lim et al., 2010; Mandel et al., 2006). Furthermore, rAAV mediated over-expression of glutamic acid decarboxylase (GAD) in the subthalamic nucleus was aimed at lowering the activity of this nucleus by increasing the production of the inhibitory neurotransmitter gamma-aminobutyric acid (GABA), an approach emulating that of the commonly used treatment deep brain stimulation (Lim et al., 2010; Mandel et al., 2006).
Cell suicide gene therapy
A final approach, utilized in the experimental treatment of CNS neoplasms such as glioblastoma, is a cell-suicide approach, whereby the delivered gene is responsible for the killing of cancerous cells. Clinical trials targeting these tumors have focused largely on delivering the Herpes Simplex virus (HSV) thymidine kinase gene together with the systemic administration of gancyclovir, resulting in cytotoxicity of infected cells (King et al., 2005; Lim et al., 2010). The treatment of glioblastoma multiforme (GBM) highlights a significant obstacle for CNS gene therapy. Although successful resection of the primary tumor can often be accomplished, a heterogeneous population of diffusely infiltrating tumor cells persists, often distally from the original tumor. As these cells are responsible for the occurrence of new tumors they are a crucial target for a gene therapy approach to be successful. However, the heterogeneous nature of these cancer cells and the absolute requirement for the transduction of very large areas of the brain have made them a difficult population to target with any form of therapy (King et al., 2005).
Gene Transfer Vectors
Once the appropriate gene therapy approach has been decided upon, the most appropriate viral vector must be chosen. Various viral vectors currently in use display varying affinities for different cell types as well as distinct anatomical sub-structures. For instance, in diseases where inflammation may play an important role in the pathology (i.e., Alzheimer’s Disease and PD), targeting inflammatory cells such as glia could be of importance. However, these cells are capable of further cell division, thus vector integration in the genome is required in order to sustain the therapeutically required levels of gene expression which otherwise would be lost due to dilution of any non-integrating viral genome over time. On the contrary, if the disease process originates in post-mitotic neurons, utilizing vectors with a strong neuronal tropism is necessary, and genome integration is not an absolute requirement (Lim et al., 2010; Mandel et al., 2006). Further consideration must also be placed upon the vector delivery method and the feasibility of covering large areas which is especially important for genetic defects resulting in a global disorder affecting all areas of the brain. Likewise, when utilizing a neurotransmitter replacement approach as has been performed in PD clinical trials, a large volume of the caudate/putamen would require being transduced in order to achieve a therapeutic benefit (Mandel et al., 2006). On the other hand, if highly targeted gene expression is required, the vector should display relative specificity for the cell population of interest.
There are several distinct types of viral vectors that are commonly utilized for gene transfer to various organ systems throughout the body, including the CNS. Each approach has different drawbacks and benefits, and the choice of vector should be based on criteria discussed in this review. In addition to recombinant viruses, other approaches have also been utilized, including ex vivo gene transfer and synthetic vectors.
Viral vectors utilized for clinical CNS gene transfer studies include recombinant Adeno-Associated virus (rAAV), Lentivirus (LV), Adenovirus (Ad), and Herpes-Simplex virus (HSV). The various strengths and caveats for each vector are discussed below.
Of the different viral vectors, rAAV has been the most widely used and studied. Recombinant AAV supports a genomic/gene carrying capacity of roughly 6 kb, and the only viral elements remaining in the recombinant virus are the inverted terminal repeats in the distal ends of the genome, structures required for helper mediated replication and capsid packaging. Following infection, rAAV supports transgene expression in post-mitotic cells for the lifetime of the individual (Mandel et al., 2006). The only serotype utilized in CNS clinical trials so far has been rAAV type 2 which has displayed an excellent safety profile (Lim et al., 2010; Mandel et al., 2006). Recently, a plethora of additional serotypes have been identified from a variety of species. In addition, the mixing of viral genomes of one serotype with capsids from another serotype creating mosaic “pseudotypes” of rAAV have displayed a wide range of neuronal tropisms and efficacies (Figure 1A). Many of these “newer” rAAV vectors have displayed greater transduction efficiency and transgene expression than that of rAAV2 (Mandel et al., 2006). Moreover, efforts are underway to selectively alter the processing of the viral capsid as well as random shuffling of portions of the capsid between various serotypes in order to maximize transgene levels and viral distribution.
Historically, rAAV has only been a neurotrophic virus in the context of the CNS; however, with the identification of recent serotypes, as well as the production of higher titer vector preparations, increasing frequency of infection of non-neuronal cell-types of the CNS has been observed.
Although the use of rAAV results in a very acceptable safety profile, one of the major limitations to its use is the limited genomic carrying capability. In situations where large, or several, genes are required, other vectors such as Ad (8-30 kb) or HSV (40-150 kb) could be utilized. To date, Ad has been utilized exclusively in CNS clinical trials targeting brain tumors (Lim et al., 2010). Two types of Ad viral vectors have been extensively used. First generation Ad, based on Ad type 5, has the early genes E1a and E1b removed; however, infection with these first generation vectors result in significant toxicity and host immune response. Next generation Ad vectors, referred to as “gutless” or “high-capacity” Ad have all of the viral genome removed, resulting in greater transgene capacity. However, the active Ad infection remains toxic to cells due to structural elements on the capsid. Ad displays a rather promiscuous tropism when targeted to the CNS, infecting neurons and glial cells equally (Segura et al., 2008). However, the large carrying capacity allows for the use of cell-specific promoters, thus allowing for cell-specific expression of transgene(s).
Wildtype HSV is a naturally neurotrophic virus with a very large (152 kb) genome. HSV vectors based on HSV type 1 exist in two forms: recombinant vectors and amplicons. Recombinant vectors retain some of the normally very large HSV genome and are characterized by extensive cytotoxicity following infection. However, the natural oncolytic properties of these vectors have been utilized in clinical trials for CNS tumors. Amplicon vectors which contain a minimal portion of the HSV DNA genome, are replication deficient, and require helper functions for production. Developments in production methods now allow for the production of high titer vectors which are devoid of helper DNA (Tyler et al., 2006). So far the only clinical trials involving HSV has been the over-expression of preproenkephalin A in dorsal root ganglia in order to treat chronic pain (Lim et al., 2010). One limitation to the use of HSV based vectors is the absence of stable long-term expression, significantly reducing the utility of HSV in situations where persistent expression is required.
Lv belongs to a subclass of retroviruses that integrate into the host cell genome. Early Lv vectors, based largely on human immunodeficiency virus 1 (HIV-1), include components of the HIV genome, but most of these elements have been removed in the newest generations. The ability to integrate randomly into the genome is a cause of concern if one considers the potential disruption of regulatory elements such as transcription factors and cell-cycle proteins which could result in tumorigenesis. Recently developed vectors lack integrase. The resulting non-integrating lentiviral vectors (NIL) have been shown to lead to extra-genomic vector DNA being maintained either in a circular or linear fashion (Lundberg et al., 2008). Due to its natural ability to integrate, LV has been extensively utilized for ex vivo gene transfer, especially considering the strong tropism for neural stem and progenitor cells, effectively rendering the cell-line transgenic, allowing for transplantation and over-expression of the therapeutic gene. Lv shows a lot of promise in CNS application; however, when compared to rAAV, Lv results in fewer infected neurons.
The final decision when considering gene therapy is when to intervene in the disease process and the anatomical location of the administration of the vector. Transport of vectors or transgenes in long axonal tracts could lead to unexpected side-effects (Mandel et al., 2006). Our evaluation of GDNF over-expression in the nigrostriatal tract also illuminates an additional caveat for gene therapy in the CNS. When the rat substantia nigra (SN) is targeted with the viral injection, the resulting anterograde transport and release of GDNF throughout the unmyelinated medial forebrain bundle (Figure 2) result in the subsequent activation of specific nuclei within the paraventricular area of the hypothalamus. Downstream effects of this activation result in significant reduction of adipose deposits throughout the body and a rather remarkable weight-loss (Manfredsson et al., 2009d). Since these findings have been reproduced in the non-human primate (Su et al., 2009), the importance of choosing the appropriate anatomical target for gene therapy in PD is apparent. Specifically, when targeting patients with a significant portion of the nigrostriatal tract remaining, these findings likely preclude the SN from being a viable target. Importantly, retrograde transport resulting from injections targeting the terminal fields of the striatum (caudate/putamen) is much less pronounced, and in this injection paradigm no changes in body weight are observed (Manfredsson et al., 2009d).
Lessons Learned from GDFL Clinical Trials
Considering the vast collection of successful pre-clinical data concerning the use of GDFLs in the treatment of PD, the initiation of recent gene therapy clinical trials was met with great anticipation. However, thus far, the clinical outcome has been relatively disappointing. In the most recent phase 2 double-blind trial, rAAV2-neurturin was delivered to the putamen of PD patients. At the original 12-month endpoint, no primary efficacy outcome measures were significant. At 18 months following treatment, modest improvements were reported in the Unified Parkinson’s Disease Rating Scale (UPDRS) and secondary motor measurements (presented at the American Society for Gene Therapy Annual meeting) (Bartus, 2009). As has been previously demonstrated, the treatment was well tolerated and no major adverse events were reported.
One possible mechanism behind this trial’s efficacy endpoint failure is the misinterpretation of pre-clinical results. Two widely utilized models to study PD are the rat 6-hydroxydopamine (6-OHDA) and the mouse/non-human primate MPTP toxin-induced models which both cause relatively acute degeneration of nigral DA neurons (Manfredsson et al., 2009b). Both of these models rely on excess oxidative stress to DA neurons which may be one of the disease processes in idiopathic PD. Preclinical studies utilizing the 6-OHDA model have clearly shown an absolute requirement for the administration of GDFLs prior to, or immediately following, the toxic insult (Kirik et al., 2001). Thus, all data indicate that for a successful GDFL based clinical trial, the onset of treatment must occur at a time where a significant portion of the SN neurons remain intact (fig. 2). Nonetheless, in the Ceregene, and previous, clinical trials, late-stage patients were enrolled (Hoehn and Yahr stage 3-4), a stage of the disease where very little of the nigrostriatal pathway remains. Therefore, in these trials utilizing late stage patients, there were little to no target neurons remaining for the therapy to act upon. Patient selection was dictated by FDA-related safety concerns because early stage PD patients live for many years with the disease and symptomatic treatments are available (Manfredsson et al., 2009b). Thus, late stage patients may be a reasonable population in which to test high risk treatment strategies but also may not conform to the biological requirements of the treatment strategy.
This apparent disconnect between pre-clinical rationale and FDA safety regulatory mandates renewed the argument for a regulatable vector system to enable the use of earlier stage PD patients. Such a vector system would utilize the systemic administration of an exogenous “controlling agent” which in itself would be required to have an acceptable safety profile. In addition, this vector system would also be required to produce therapeutic levels of the transgene in an “ON” state, display complete shut-down of transgene expression in the “OFF” state, and show a dose-response to varying amounts of the controlling agent. Finally, novel genetic elements in the regulated vector itself would be required to adhere to strict safety guidelines (Manfredsson et al., 2009b). Despite tremendous advances in genetic expression systems, it is not until recently that such an expression cassette was constructed. We utilized a bicistronic single rAAV vector to express GDNF under the control of a tetracycline responsive promoter (Manfredsson et al., 2009a). Although this expression system resulted in lower GDNF expression than that seen when using strong constitutive promoters, protein levels were still above that required for therapeutic benefit (Manfredsson et al., 2009a). Equally important, no GDNF was detected during the administration of the controlling agent doxycycline, and the expression system showed high sensitivity, and a linear dose-response to levels of doxycycline far below that commonly used clinically (Manfredsson et al., 2009a). Thus, this vector system fulfills most of those requirements envisioned in order to translate a regulatable vector system into a clinical gene-delivery tool.
The lack of a measurable clinical efficacy in a number of clinical gene transfer trials in the CNS is disappointing. Nonetheless, with the advent of novel expression systems, more refined and efficient vectors, and a greater understanding of the etiology of neurological diseases, the future of gene transfer is promising. Whereas several gene therapy trials have been successful when targeting other organ systems throughout the body, treatment of the CNS is more problematic. The blood brain barrier makes systemic administration of the vector improbable (Manfredsson et al., 2009c), although in using new AAV serotypes, systemic administration of rAAV has resulted in limited CNS transduction (Foust et al., 2009; Manfredsson et al., 2009c). Secondly, due to the disparate nature of the various circuits in the CNS, and the way these circuits are affected in disease, gene transfer will often require the precise targeting of finite areas within the brain in order to minimize potentially disastrous “off-targeting” side-effects. Finally, considering that target cells are often post mitotic, gene therapy treatments will be considered permanent, thus, significant safety measures need to be in place. One key such safety mechanism now exists in the form of a regulatable vector, allowing for a treatment to be abolished should unsafe conditions arise.
Bartus RT. Scientific Symposium 113: Late stage industry clinical trials. (Platform presentation.) Mol Ther 17(suppl. 1):i, 2009.
Foust KD, Nurre E, Montgomery CL. Hernandez A, Chan CM, Kaspar BK. Intravascular AAV9 preferentially targets neonatal neurons and adult astrocytes. Nat Biotechnol 27:59-65, 2009.
Kim DH, Rossi JJ. Strategies for silencing human disease using RNA interference. Nat Rev Genet 8:173-84, 2007.
King GD, Curtin JF, Candolfi M, Kroeger K, Lowenstein PR, Castro MG . Gene therapy and targeted toxins for glioma. Curr Gene Ther 5:535-57, 2005.
Kirik D, Georgievska B, Rosenblad C, Bjorklund A. Delayed infusion of GDNF promotes recovery of motor function in the partial lesion model of Parkinson’s disease. Eur J Neurosci 13:1589-99, 2001.
Lim ST, Airavaara M, Harvey BK. Viral vectors for neurotrophic factor delivery: a gene therapy approach for neurodegenerative diseases of the CNS. Pharmacol Res 61:14-26, 2010.
Lundberg C, Bjorklund T, Carlsson T, Jakobsson J, Hantraye P, Deglon N, Kirik D. Applications of lentiviral vectors for biology and gene therapy of neurological disorders. Curr Gene Ther 8:461-73, 2008.
Mandel RJ, Manfredsson FP, Foust KD, Rising A, Reimsnider S, Nash K, Burger C. Recombinant adeno-associated viral vectors as therapeutic agents to treat neurological disorders. Mol Ther 13:463-83, 2006.
Manfredsson FP, Lewin AS, Mandel RJ. RNA knockdown as a potential therapeutic strategy in Parkinson’s disease. Gene Ther 13:517-24, 2006.
Manfredsson FP, Burger C, Sullivan LF, Muzyczka N, Lewin AS, Mandel RJ. rAAV-mediated nigral human parkin over-expression partially ameliorates motor deficits via enhanced dopamine neurotransmission in a rat model of Parkinson’s disease. Exp Neurol 207:289-301, 2007.
Manfredsson FP, Burger C, Rising AC, Zuobi-Hasona K, Sullivan LF, Lewin AS, Huang J, Piercefield E, Muzyczka N, Mandel RJ. Tight Long-term dynamic doxycycline responsive nigrostriatal GDNF using a single rAAV vector. Mol Ther 17:1857-67, 2009a.
Manfredsson FP, Okun MS, Mandel RJ. Gene therapy for neurological disorders: challenges and future prospects for the use of growth factors for the treatment of Parkinson’s disease. Curr Gene Ther 9:375-88, 2009b.
Manfredsson FP, Rising AC, Mandel RJ. AAV9: a potential blood-brain barrier buster. Mol Ther 17:403-5, 2009c.
Manfredsson FP, Tumer N, Erdos B, Landa T, Broxson CS, Sullivan LF, Rising AC, Foust KD, Zhang Y, Muzyczka N, et al. Nigrostriatal rAAV-mediated GDNF overexpression induces robust weight loss in a rat model of age-related obesity. Mol Ther 17:980-91, 2009d.
Marks W J, Jr, Ostrem JL, Verhagen L, Starr PA, Larson PS, Bakay RA, Taylor R, Cahn-Weiner DA, Stoessl AJ, Olanow CW, Bartus RT. Safety and tolerability of intraputaminal delivery of CERE-120 (adeno-associated virus serotype 2-neurturin) to patients with idiopathic Parkinson’s disease: an open-label, phase I trial. Lancet Neurol 7:400-8, 2008.
Pascual A, Hidalgo-Figueroa M, Piruat JI, Pintado CO, Gomez-Diaz R, Lopez-Barneo J. Absolute requirement of GDNF for adult catecholaminergic neuron survival. Nat Neurosci 11:755-61, 2008.
Segura MM, Alba R, Bosch A, Chillon M. Advances in helper-dependent adenoviral vector research. Curr Gene Ther 8:222-35, 2008.
Su X, Kells AP, Huang EJ, Lee HS, Hadaczek P, Beyer J, Bringas J, Pivirotto P, Penticuff J, Eberling J, et al. Safety evaluation of AAV2-GDNF gene transfer into the dopaminergic nigrostriatal pathway in aged and parkinsonian rhesus monkeys. Hum Gene Ther 20:1627-40, 2009.
Tyler CM, Wuertzer CA, Bowers WJ, Federoff HJ. HSV amplicons: neuro applications. Curr Gene Ther 6:337-50, 2006.
Vercammen L, Van der Perren A, Vaudano E, Gijsbers R, Debyser Z, Van den Haute C, Baekelandt V. Parkin protects against neurotoxicity in the 6-hydroxydopamine rat model for Parkinson’s disease. Mol Ther 14:716-23, 2006.
Worgall S, Sondhi D, Hackett NR, Kosofsky B, Kekatpure MV, Neyzi N, Dyke JP, Ballon D, Heier L, Greenwald BM, et al. Treatment of late infantile neuronal ceroid lipofuscinosis by CNS administration of a serotype 2 adeno-associated virus expressing CLN2 cDNA. Hum Gene Ther 19:463-74, 2008.
[Discovery Medicine, 9(46):204-211, March 2010.]