The challenges of developing a gene therapy for amyotrophic lateral sclerosis

KEYWORDS:

• ALS
• amyotrophic lateral sclerosis
• gene therapy
• viral vectors
• gene delivery
• surgical delivery

Since Charcot’s description of the neuropathological characteristics of Amyotrophic Lateral Sclerosis (ALS) in 1869, our understanding of the disease has developed substantially. However, therapy development has lagged behind. Since ALS is linked to over 40 different mutations, gene therapy is of particular interest as an approach to attenuate disease progression [1]. ALS is broadly divided into familial cases (fALS) and sporadic cases (sALS) with respect to inheritance patterns and association with known mutations. Generally, the gene therapy strategy for sALS focuses on introducing neuroprotective agents like neurotrophic factors. The strategy for fALS focuses on addressing specific gene mutations [2,3]. On both fronts, preliminary research has shown progress, but successes are tempered by subsequent failures to scale up to translation. Here, we will discuss the challenges involved with developing a gene therapy for ALS and the concurrent potential solutions. Broadly, we will review the following areas: gene mutations, underlying pathogenesis, research models, vector selection, and surgical delivery paradigms. 

A major challenge in the development of a gene therapy for amyotrophic lateral sclerosis (ALS) is the continued idiopathic nature of the majority of cases. It is estimated that 10% of cases are familial ALS (fALS), of which 70% are linked to a known genetic lesion [4,5]. However, 90% of cases are sporadic ALS (sALS), of which 90% have an unidentified etiology [5]. In other words, of every 100 patients with ALS, only about 16 patients will have known genetic mutations. There are efforts to address this deficiency in knowledge of genetic etiology. For example, recent ALS associated mutations in the C9orf72 and NEK1 genes were discovered through the recruitment of large patient cohorts and advanced genetic analyses [6,7]. Given the rarity of ALS, discovery of genetic lesions will require large-scale collaborations across institutions and nations. One promising international initiative is Project MinE, which is collecting and analyzing 15,000 ALS patient samples to identify novel genetic lesions [8]. Projects like this will be the foundation of understanding the genetic component of ALS and therefore potential gene therapies, particularly for fALS.

Downstream of the genetic lesions, our understanding of the underlying pathogenesis is an ongoing challenge to gene therapy development. In this domain, current knowledge is surprisingly limited. For example, we do not understand the function of the native protein that is impacted by the C9orf72 mutation [9]. Even the well characterized SOD1 mutation does not have a definitive mechanism beyond a toxic gain of function [1]. To cloud the field even further, there are as many proposed pathological mechanisms as there are mutations. These include, but are not limited to, glutamate excitotoxicity, glial cell dysfunction, protein/RNA toxicity, and mitochondrial/cytoskeletal dysfunction [1]. While the reality of the disease pathogenesis may be multifactorial, the paucity of a clear mechanism limits therapeutic development progress. This knowledge gap is only addressable through experimental work with in vitro and animal models. One historical success has been SOD1 small-animal models. These models have provided a space to draw the crucial understanding of a toxic gain of function in SOD1-associated ALS. The result is that many strategies have focused on mitigating effects related to the SOD1 gain of function. Indeed, there have been successes in SOD1 gene-silencing approaches (e.g. various adeno-associated virus (AAV), siRNA, and shRNA constructs, and antisense oligonucleotides), showing delayed disease onset and extension of life-span [5]. For example, the use of viral vectors (particularly AAV9) with silencing constructs have shown delayed disease progression and improved survival in SOD1 mouse models [10,11]. Further, to address the proposed excito-oxidative stress mechanism, other groups have added a combination of neuroprotective and neurotrophic transgenes (e.g. EAAT2, GDH2, and NRF2) leading to improvements in survival, body weight and neurological findings [12]. While promising, it is not clear if these strategies will succeed outside of SOD1 mutants, meaning, at best, they would treat only 1% of ALS patients. In order to further address this challenge of underlying pathophysiology and the other 99% of patients, additional research models must be generated. Directly addressing these concerns, there have been strides in both the use of in vitro models and efforts to develop other gene-specific murine models (e.g. C9orf72 and NEK1) [3,13]. However, one great difficulty is that the symptomology and histology of ALS are not necessarily recapitulated in these models, meaning refinements are necessary [3].

A critical translational gap is that small-animal and in vitro models do provide a space for proof-of-principle, but have not adequately paralleled human disease phenotypes [3,14]. The end result is that few therapies have reached a pre-clinical state and that none have reached the clinic. A potential solution to this challenge may come in the form of new transgenic model systems that are possible with the advent of genetic tools such as the CRISPR/Cas9 system [15]. The efficiency and relative ease-of-use of such technologies may facilitate disease model generation especially with the poly-nucleotide expansions that occur in many ALS associated mutations. In terms of disease recapitulation, the anatomical and physiological differences between small-animal species and humans may account for differences in phenotype. Therefore, one possible approach is the use of large animal models with more anatomical and physiological similarities. For example, Yang et al. reported the development of SOD1 transgenic pigs that displayed histopathological and symptomatic findings more in parallel with humans [14]. Porcine models also have an established anatomic similarity to the human brain and spine, which is useful in the context of clinical and surgical translation [13]. Overall, the development of topical animal models is progressing rapidly and may provide a space for more robust pre-clinical studies and expanded understanding of the pathophysiology of ALS.

Another challenge is selection of an appropriate viral-vector for delivery of the transgene for in vivo (direct application to the patient) or ex vivo gene therapy (transplantation of transduced cells). Current vector options include AAV, adenovirus, herpesvirus (HSV), and lentivirus. Adenovirus and HSV both have had major improvements in recent years, but are not ideal choices for application in ALS primarily due to transient gene expression [16]. Lentiviral vectors elicit broad tissue tropisms and mediate long-term gene expression. In addition, pseudo-typed lentivirus is efficient for retrograde axonal transport [16,17]. The capacity for genome integration makes lentiviral vectors a favorable choice for ex vivo gene therapy, a paradigm which has shown therapeutic efficacy in small animal models [18]. It is encouraging to note that this approach will be used next year in clinical trials for ALS at Cedars-Sinai in work coordinated by the Svendsen group [19]. Lentiviral vectors do have drawbacks that limit their use for in vivo gene therapy. These include limited volumetric spread from the site of infusion, risks of integrational mutagenesis, and only moderately-sized titers [16]. AAVs are considered the vector of choice for in vivo use due to their low-risk of integrational mutagenesis, diminished immunogenicity, high titers, and stable gene expression in non-dividing cells [16,20]. Promising serotypes of AAVs include: AAV9, AAV2, and Rh10, which are both neurotropic and useful in direct or indirect delivery paradigms for in vivo gene therapy. Overall, current vector choices are promising and further optimization is essential as we move toward clinical application.

A practical challenge for in vivo gene therapy is the selection of an appropriate surgical delivery paradigm that can be scaled to the clinic and/or operating room. Broad categories include indirect and direct delivery paradigms. Indirect delivery paradigms for ALS include both intraneural/muscular retrograde transport and intravenous (IV) injection. These are appealing, given their minimally invasive nature. While retrograde axonal delivery (with AAV2, Rh10, or pseudotyped-lentivirus) has shown success in small animal models, it has not scaled well to larger animal models and is not considered the first choice for translation [16,17,21]. In the domain of IV delivery, AAV9 is an option since it crosses the blood-brain barrier and demonstrates robust transduction of neuronal, glial, and endothelial cells [16,21]. The major limitations to IV delivery include off-target sequestration and, consequently, extremely high titer requirements [20]. Whether these limitations limit the efficacy and utility of this approach is currently being tested in an ongoing phase-1 clinical trial for spinal muscular atrophy (SMA) [22]. The existence of studies like this is encouraging, and they act as a representation of how close we are, broadly, to clinically translatable gene therapies.

Direct delivery to the spinal cord, either intraparenchymal (IP) or intrathecal (IT), may circumvent issues of the BBB, high dose requirements, and off-target sequestration. AAV9 and Rh10 are the leading candidates because of their neurotropic characteristics and spread beyond the infusion site via axonal transport [16]. Our group has conducted safety validation in pre-clinical porcine models and in a phase 1 clinical trial with patients, both demonstrating the safety of IP infusion paradigms [21,23,24]. One limitation of this approach is the segmental nature of transduction, which is not ideal for the diffuse motor neuron disease seen in ALS. For this reason, IT infusion may be more appropriate as shown by our group and others [21]. We are currently conducting further studies focused on the administration of neuroprotective and neurotrophic transgenes into the IT space in both rat and porcine models. While further evidence is required, we believe that IT administration will be the delivery paradigm of choice for viral vector delivery in ALS because of diffuse transduction and the lower dose requirements compared to IV administration.

There is often a tendency to dismiss prospects of developing a gene therapy due to past failures. Indeed, we acknowledge and have discussed numerous limitations and challenges (genetics, vectorology, surgical delivery) to the development of a gene therapy for ALS. However, each challenge that we have discussed is being overcome. Tangible progress continues even as of this publication, yielding new and promising perspectives in the field. We prefer a more forward thinking stance. By drawing on the multi-disciplinary insights in each of these domains, we believe that the beginning of a clinically translatable therapy is possible in the near future.

Declaration of interest

N. Boulis is a paid consultant for Agilis, MRI Interventions, Voyager, Oxford Biomedica, Q Therapeutics and Neuralstem Inc. He is a founder of Switch Bio Holdings and former employee of Above and Beyond LLC up to December 1st 2015. The authors have no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript apart from those disclosed.

Additional information

Funding

This paper was not funded.