Non-Viral Methods

Kotin described a novel approach to gene transfer that delivers only the DNA of interest (the gene), flanked by inverted terminal repeats from the adeno-associated virus (AAV), that forms closed-end, linear duplex molecules (CELiD). Compared to bacterial plasmids, which have also been used for non-viral gene transfer, CELiD DNA has no prokaryotic modifications that might trigger an innate immune response and no endotoxins, said Kotin. He and his colleagues have produced close-ended DNA (ceDNA) constructs with the gene for green fluorescent protein or nuclear-localized b-galactosidase, and showed that following transfer by hemodynamic injection into the tail vein of mice, gene expression was constant over 7 days (Li et al., 2013). Adding a liver-specific promoter to the construct resulted in fairly constant gene expression in liver tissue over 10 weeks and essentially no change in CELiD copy number, said Kotin. More recently, investigators have demonstrated efficient transduction in the central nervous system (CNS) using ceDNA constructs delivered directly to rat brain using convection-enhanced delivery. One advantage of the ceDNA approach is that the vectors are produced using a good manufacturing practice-compatible process, said Kotin. However, optimizing the chemistry for specific tissue targeting remains a challenge in developing ceDNA methodologies, and it will not be possible to target the CNS with systemic delivery, he said.

Small Molecules to Modulate Splicing

Small molecules that modify post-transcriptional processing of RNA represent yet another novel therapeutic approach for genetic disorders, according to Anuradha Bhattacharyya of PTC Therapeutics, which has developed a platform to discover and develop small molecule splicing modifiers that are orally available and have broad tissue distribution. They are currently applying this technology to the development of treatments for spinal muscular atrophy (SMA), familial dysautonomia (FD), and Huntington's disease (HD), said Bhattacharyya.

The SMA therapeutic in development targets endogenous SMN2 with a small molecule splicing modifier that promotes the inclusion of exon 7, said Bhattacharyya. Two other strategies are also being pursued, she said. Each of these approaches provides many potential druggable targets, said Bhattacharyya. One approach corrects splicing mutations in pre-mRNA; this strategy is being used to enable inclusion of an exon that is usually skipped in the disease setting of FD. The other approach being pursued as a treatment for HD activates a pseudo-exon in the huntingtin gene pre-mRNA, which creates a premature stop codon, resulting in degradation of the mRNA and protein. In HD mouse models, Bhattacharyya said they have demonstrated a dose-dependent decrease in huntingtin protein across all relevant brain areas following oral delivery of the compound.

Nanoplatforms for Brain-Targeted Genome Editing

CRISPR/Cas9 genome editing, mentioned in Chapter 3, is a powerful technique enabling gene insertion, deletion, and alteration, said Shaoqin Sarah Gong, Vilas distinguished professor in biomedical engineering at the Wisconsin Institute of Discovery, University of Wisconsin–Madison. As part of the Somatic Cell Genome Editing Consortium established last year by the National Institutes of Health (NIH), Gong and colleagues are developing nanoplatforms to deliver Cas9 protein/single guided RNA (sgRNA) ribonucleoprotein complex (RNP) packaged into a nanocapsule. This approach, she said, allows precise control over the ratio of Cas9 and sgRNA within the RNP and enables rapid editing with a low incidence of off-target effects. Like other non-viral vector approaches, it also offers a good safety profile and versatile chemistry, and is easy to scale up, said Gong.

The nanocapsule, composed of a covalently cross-linked yet intracellularly biodegradable polymer coating, has a surface that can be conjugated with a wide array of targeting ligands, said Gong. It is taken up by specifically targeted cells through receptor-mediated endocytosis, she said, and once inside the target cell, the nanocapsule disintegrates and releases the RNP into the cytosol. The RNP is then transported to the nucleus, where it can carry out its editing of the targeted gene, she said. The nanocapsules achieve high editing efficiency even when freeze dried, which provides benefits in terms of purification, long-term storage, transportation, and dosage control, said Gong. She added that the nanocapsule is much less cytotoxic than Lipofectamine, a commercially available reagent often used to deliver DNA or RNA into cells.

Gong and colleagues have tested the efficiency of a nanocapsule targeting the human Alzheimer's precursor protein gene (APP) in cell culture, building on the work of her collaborator Subhojit Roy and colleagues, who recently used a CRISPR/Cas9 strategy to selectively silence APP through gene editing, suggesting that this may be an effective treatment modality (Sun et al., 2019). Gong and colleagues are also exploring other types of nanoplatforms to target different types of cells and deliver a wide range of hydrophilic payloads, including DNA, mRNA, proteins, and small-molecule drugs.

Novel Delivery Methods

Returning to the challenge of delivery, Junghae Suh, associate professor of bioengineering at Rice University, and Viviana Gradinaru, professor of neuroscience and biological engineering at the California Institute of Technology, addressed from the bioengineering perspective how delivery vectors can be engineered to be more efficient, more specific, and more controllable and how features of gene expression can be controlled.

Gradinaru provided some concrete examples from the field of protein engineering, where directed evolution has emerged as a means to engineer natural products for novel functionality. Working with Frances Arnold, who in 2018 won a Nobel Prize in chemistry (along with co-winners Gregory Winter and George Smith) for her development of the directed evolution approach, Gradinaru and colleagues generated thousands of variants of light-responsive proteins called opsins and then put them through very stringent selection criteria to get opsins that emit light in response to voltage, effectively turning them into voltage sensors (Flytzanis et al., 2014).

The same principles have also been applied to the development of improved viral vectors, said Gradinaru. Using directed evolution, she and her colleagues refined the properties of the AAV9 capsid to enable the virus to cross the blood–brain barrier in rodents, thus creating the PHP.B variant mentioned in Chapter 3 (Chan et al., 2017; Deverman et al., 2016). However, translating this from the rodent to non-human primate brain has proved difficult, in part because of the increased volume of the primate brain and cross-species variations, including different mechanisms to cross the blood–brain barrier, said Gradinaru. It becomes a numbers problem, she said, because meeting these challenges with directed evolution requires the generation of many offspring—as many as 1 billion or more depending on the molecule being targeted. Because no biological screen can fully sample this space, computational biologists, statisticians, and machine learning approaches are essential, she said. The concept is simple, but sounds like magic, she said: You feed training data into the computer, which spits out sequences that can be tested for the presence of desired characteristics.

Gradinaru said that once they had solved the challenge of crossing the blood–brain barrier in rodents, they encountered another challenge that required better opsins. Application of directed evolution and machine learning approaches to this problem yielded improved opsins for systemic delivery, which can be used for minimally invasive optogenetics (Bedbrook et al., in press). Now, in a parallel approach, they are applying directed evolution with deep sequencing to create variants that can achieve biodistribution to specific cell types such as endothelial cells in the vasculature or neurons.

Although the Parkinson's disease (PD) trial failures discussed by Jeffrey Kordower in Chapter 2 may have doomed glial-derived neurotrophic factor and neurturin as treatments for PD, they informed another approach that Kordower and colleagues are now pursuing: using intracellular antibodies (intrabodies) delivered with an AAV vector to clear α-synuclein (Chatterjee et al., 2018). The intrabodies are designed to bind to both monomeric and fibrillar forms of α-synuclein, prevent aggregation, and promote clearance. In a rat model of PD, direct delivery of the AAV-intrabody construct into the substantia nigra has resulted in improved motor behavior and increased intrastriatal dopamine, said Kordower. Next, they are hoping to use better vectors that will enable delivery throughout the brain. Steven Paul, chief executive officer of Karuna Pharmaceuticals, commented that a similar approach has been used to reduce tau pathology in mutant tau transgenic mice, and that this approach seemed to have superior efficacy in reducing tau-dependent neurodegeneration (Liu et al., 2016b).

Developing Gene-Targeted Therapies of Psychiatric and Neurodevelopmental Disorders

Although scientists have begun to tackle complex polygenic disorders with gene therapy, they have not yet begun to apply this therapeutic approach to psychiatric disorders, said Hyman. He asserted that the time has come to think about this challenge while recognizing that clinical applications are a long way off.

Although neuropsychiatric disorders are highly heritable, all common neuropsychiatric disorders are highly polygenic, with phenotypes resulting from myriad small genetic nudges rather than a large genetic shove, said Hyman. He noted that this complicates efforts to identify which genes could be targeted. Moreover, he said, highly penetrant alleles that substantially elevate risk for early-onset behavioral disorders are quickly selected out of the gene pool because these highly disabling illnesses reduce the likelihood that an affected person will have children (Power et al., 2013). Large-effect alleles can occur de novo, but are rarely transmitted, while common and rare variants with low effect sizes can be readily transmitted, said Hyman.

Nonetheless, scientists have created animal models that replicate the human phenotype and have been able to reverse the phenotype by targeting specific genes, suggesting that gene therapy may be feasible. For example, Hyman cited the work of Adrian Bird and colleagues, who were able to reverse neurological defects in a mouse model of Rett syndrome (Guy et al., 2007).

Schizophrenia, however, is far more complex, said Hyman, with hundreds of genome-wide significant loci identified and extreme phenotypic heterogeneity (Huckins et al., 2019). He described a tool that could help scientists roughly stratify affected individuals at a genetic level by creating, for each individual, a weighted sum of risk alleles across the entire genome to produce a polygenic risk score (PRS). The PRS enables stratification of subjects by severity of genetic loading and permits identification of shared common variant risk across phenotypes, said Hyman. It can also help identify important genetic pathways and potentially identify targets for gene therapy interventions, he said. For example, about 70 percent of genome-wide association study (GWAS) hits in AD studies are expressed in microglia and are thought to be involved in inappropriate synapse elimination, said Hyman. He speculated that in schizophrenia, although none of the GWAS hits are expressed in microglia, there may be similar biological processes reached through different pathways such as genes encoding synaptic proteins and complement proteins that signal to microglia, adding that useful target selection can sometimes emerge from analysis of these pathways. What is important, he said, is to understand the biology. Transgenic animal models would require vastly improved methods of multiplexing using gene editing technologies.

The high degree of polygenicity seen in schizophrenia is also seen in common forms of neurodevelopmental disorders such as autism (Gaugler et al., 2014), said Hyman. Indeed, schizophrenia is considered to be neurodevelopmental in origin (Rund, 2018). A recent study by the Autism Sequencing Consortium (ASC) examining 36,000 exomes from affected and unaffected individuals identified 102 genes that are strongly associated with autism, said Joseph Buxbaum, director of the Seaver Autism Center for Research and Treatment at the Icahn School of Medicine at Mount Sinai in New York. Many of the gene mutations that have been identified are highly penetrant with serious deleterious impact and can be parsed into those that are autism specific and less autism specific, Buxbaum said.

Buxbaum summarized data from ASC and the Deciphering Developmental Disorders Study in the United Kingdom, which sequenced the genomes and phenotyped more than 4,000 individuals with severe developmental disorders (Deciphering Developmental Disorders Study, 2015). Together, he said, these studies and others have implicated multiple pathways. The genes discovered to date are overwhelmingly contributing to dominant disorders. The importance of a dominant disorder, said Buxbaum, is that there is a second allele that can be manipulated. He said most mutations identified to date result in a loss of function and are dose sensitive, typically displaying a U-shaped curve indicating that problems arise when there is either too much or too little gene expression. Most of the mutations result in severe conditions, he said, and in some cases re-expression of the gene later in life can be beneficial. For example, Guoping Feng and colleagues reported that in mouse models, mutations in SHANK3 result in synaptic dysfunction and autistic-like behaviors, but that re-expressing the gene in adult mice improved neural function and behavior, said Buxbaum (Mei et al., 2016).

Among the genes identified in these studies, those with the strongest findings are relatively common and highly penetrant, said Buxbaum. For example, SHANK3 haploinsufficiency explains about 0.5 percent of autism cases and 0.5 percent of intellectual disability cases, according to some studies (Betancur and Buxbaum, 2013), while other studies suggest an even higher association of SHANK3 with autism, he said. Loss of SHANK3 is called Phelan-McDermid syndrome (PMS), said Buxbaum. Taking PMS as one example, Buxbaum made the following points. Given that genetic testing can readily identify PMS; that it is thought to be fully penetrant with many potential biomarkers, rodent models, and even a primate model; and that there is an effective and engaged family foundation, Buxbaum suggested that this disorder may be an excellent target for gene therapy. Moreover, he said, a successful gene therapy for PMS might be repositioned to treat other forms of autism where there is no SHANK3 mutation, but where overlapping pathways are disrupted via other mechanisms. This is but one example, he said, suggesting there are already several dozen other known genes that are natural targets for gene therapy.

Investing in Natural History Studies

To reach more diseases, particularly rare and ultrarare diseases, a collaborative platform approach starting with a natural history cohort is critical, added Petra Kaufmann. Such an approach could allow for the application of advanced and novel technologies and other resources to rare populations and provide recruitment registries and control populations, she said. Holly Tabor added that engaging with patients and patient communities can enable the building of natural history study databases. Rare disease groups and patients have long advocated for paying more attention to the natural history of their disorders, she said. Story Landis noted that academic groups and patient groups may not be aware of what is needed to make a natural history study useful for those who are conducting clinical trials.

Natural history studies are essential to developing compelling clinical trials, said Landis. Many single-gene defects are ripe for therapy development, she said, but in the absence of the natural history, selecting appropriate endpoints is not possible. Kaufmann agreed, noting that natural history studies should include “regulatory actionable” outcome measures. Kathleen Reape also noted that in ultrarare diseases, natural history data may be used as the control data. Indeed, in Chapter 2, Kaufmann described how natural history data were used in this way in the AVXS-101 trial, which led to the approval of a gene replacement therapy for SMA.

Landis said the National Institute of Neurological Disorders and Stroke (NINDS) has a clinical readiness funding program to support partnerships between academics and patient groups for natural history studies and identification of clinical endpoints. Walter Koroshetz, director of NINDS, reiterated Landis's comment, noting that NIH has funded several programs relevant to these proceedings, including not only natural history studies for nine different diseases, but other programs for biomarker validation and clinical trial readiness as well. They are also funding industry and disease collaborations to conduct clinical trials, he said. Frank Bennett added that having industry participation in the design of natural history studies is critical.