DNA Editing Enzymes as Potential Treatments for Heteroplasmic mtDNA Diseases

Abstract

Mutations in the mitochondrial genome are the cause of many debilitating neuromuscular disorders. Currently, there is no cure or treatment for these diseases, and symptom management is the only relief doctors can provide. Although supplements and vitamins are commonly used in treatment, they provide little benefit to the patient and are only palliative. This is why gene therapy is a promising research topic to potentially treat and in theory, even cure diseases caused by mutations in the mitochondrial DNA (mtDNA). Mammalian cells contain approximately a thousand copies of mtDNA, which can lead to a phenomenon called heteroplasmy, where both wild-type and mutant mtDNA molecules co-exist within the cell. Disease only manifests once the percent of mutant mtDNA reaches a high threshold (>80%), which causes mitochondrial dysfunction and reduced ATP production. This is a useful feature to take advantage of for gene therapy applications, as not every mutant copy of mtDNA needs to be eliminated, but only enough to shift the heteroplasmic ratio below the disease threshold. Several DNA editing enzymes have been used to shift heteroplasmy in cell culture and mice. This review provides an overview of these enzymes, and discusses roadblocks of applying these to gene therapy in humans.

Graphical Abstract

1. Introduction: mitochondria

Mitochondria are known as the “powerhouses” of the cell, producing most of the cell’s ATP through Oxidative Phosphorylation (OXPHOS). However, mitochondria are involved in a myriad of other functions outside of ATP production including but not limited to bioenergetic pathways (Krebs Cycle, ß-oxidation), biosynthetic pathways (synthesis of amino acids, nucleotides, cholesterol, glucose, heme), cell signaling, balancing redox equivalents (NADH/NAD+ levels), anti-oxidant defense, stress response, autophagy, and apoptosis [1–3].

Mitochondria are believed to have arisen from an alphaproteobacteria engulfed by a eukaryotic progenitor that later established an endosymbiotic relationship. Over time, most genes of the ancestral mitochondria genome were lost or transferred to the nucleus by lateral gene transfer, leaving the current mtDNA molecule encoding only 37 genes [4, 5]. Like their prokaryotic ancestor, mitochondria are comprised of two separate membranes: the inner mitochondrial membrane (IMM) and outer mitochondrial membrane (OMM) [2, 5]. In between the IMM and OMM is the Intermembrane Space (IMS). Oxidative Phosphorylation takes place in the complexes embedded in the inner mitochondrial membrane. In the mitochondrial matrix, tricarboxylic acid cycle (TCA) enzymes generate NADH and FADH₂, which act as electron carriers by donating electrons to the electron transport chain (ETC). The ETC is made up of four complexes (complex I-IV), which pump protons from the mitochondrial matrix to the IMM. The proton gradient generated is used to power ATP synthase (complex V), which phosphorylates ADP to make ATP [2].

The mitochondrial proteome makes up ~10% of the total cellular proteome. Therefore, most mitochondrial proteins are transcribed from nuclear genes, translated in the cellular matrix, and imported into the mitochondria. Optimal mitochondrial function depends on both nuclear and mitochondrial genomes, and mutations in either can affect mitochondrial function and health [6].

Separate from the nuclear genome, the 16.5 kb double-stranded circular mitochondrial genome (mtDNA) encodes 13 proteins protein subunits of the Oxidative Phosphorylation (OXPHOS) enzymes: 7 subunits of Complex I, 1 of Complex III, 3 of Complex IV, and 2 of Complex V; as well as the 22 tRNA’s and 2 rRNA’s needed for translation within the mitochondria [7, 8]. In mammals, mtDNA is almost always maternally inherited, and paternal mtDNA actively destroyed after fertilization [2, 9, 10].

The mitochondrial genome organization is slightly different from the nuclear genome. Namely, vertebrate mtDNA codons AUA (codes methionine instead of isoleucine), UGA (codes tryptophan instead of STOP), and AGA and AGG (code STOP instead of arginine) [6]. mtDNA is not organized into paired chromosomes, but rather are present as compacted nucleoids (mtDNA and protein complexes), that proliferate independently of the nuclear DNA (nDNA) [11]. Though the mechanism is yet to be elucidated, it is known that mtDNA copy number is tightly controlled. [12]. For example, there appears to be an upregulation in mtDNA copy number during oocyte maturation, and a persistent copy number in primordial germ cells [13]. Mitochondria also have their own DNA repair machinery, which is thought to be less efficient than nDNA repair. This can be explained by the high mtDNA copy number, as deleterious mutations can be suppressed by copies of wild-type mtDNA [6].

2. Heteroplasmy, segregation, and disease

There are, on average, about 1000 copies of mtDNA in each human cell [14]. When only a subset of these molecules harbor a sequence variation, we have a so-called heteroplasmic condition. However, many people in the population carry low levels of mtDNA mutations without symptoms. This is because a certain threshold of mutant to wild-type mtDNA needs to be reached for a biochemical phenotype (usually >70–90%) depending on the mutation [15].

Symptoms of mitochondrial disease vary greatly, even between patients carrying the same mutation. This is because symptoms depend on level of heteroplasmy in affected tissues. The biochemical threshold is different for each tissue, and depending on which tissue reaches the threshold, the symptoms will manifest accordingly [16, 17].

Individuals affected by high mutant mtDNA levels often come from a family with no history of disease, and a mother that is healthy. If mtDNA is maternally inherited, how does this phenomenon occur? It is important to note that usually, the mother already carries a subset of mutant mtDNA molecules, but her heteroplasmy levels do not reach a biochemical threshold for disease manifestation. Contrarily, babies born with sub-threshold levels sometimes have heteroplasmy mutations that increase in levels over their lifetime, developing disease in adulthood. As explained below, the increase of mutant mtDNA can occur by one of several mechanisms, including; bottleneck transmission during oogenesis, purifying selection during germline transmission, vegetative segregation, relaxed replication, or by replicative advantage [6].

In mammals, mtDNA is maternally inherited, and a woman carrying a heteroplasmic mtDNA mutation can transmit variable amounts of mutant mtDNA to her offspring. This is due to a “bottleneck” that occurs in the maternal germline, in which mtDNA levels are greatly reduced first to a predicted ~200 copies per cell, and then to a predicted 100,000 copies in a mature oocyte [6, 18, 19]. In this manner, a small subset of mtDNA molecules can become over-represented in the mature oocyte. Also at the germline state, mtDNA can undergo molecular selection, a poorly understood post-fertilization section mechanism that limits offspring with certain types of mutations [20, 21].

Dividing cells randomly segregate mitochondria during division. If a heteroplasmic variant is present, each daughter cell may receive different proportions of wild-type and mutant mtDNA. The average percentage of mutant mtDNA may stay the same within the cell population, but if a cell with higher levels of mutant mtDNA proliferates faster, it will develop a “founder effect”, establishing a clonal line of cells with higher mutant mtDNA levels. This process is called vegetative segregation, and is applicable to segregation during oocyte formation, as well in dividing somatic cells in the body over time [6, 22].

The last method of mutant mtDNA selection is replicative advantage. In this mechanism, mtDNA molecules that have some sort of replicative advantage are copied more frequently than wild-type mtDNA molecules. This has been suggested for mtDNA molecules that carry large pathogenic deletions, several kilobases smaller than wild-type, where the deletion could, in theory, make replication of a smaller circle faster and available for replication more often. This mechanism is also a potential explanation to accumulation of mtDNA mutation levels over a lifetime of a patient, leading to progression of phenotype severity [6].

One difference between nDNA and mtDNA is that nDNA replicates only one time in a cell cycle, but mtDNA replicates continuously throughout the cell cycle, even in post-mitotic cells. This process is called relaxed replication. Throughout the cell cycle, mtDNA is constantly turning over, which can also cause shifts in heteroplasmy. Selection for or against mutant mtDNA (mentioned later) can drastically alter heteroplasmy over a lifetime [6, 22, 23]. This is one explanation of how heteroplasmy can change within an individual over time, and cause late presentation or progression of disease over a lifetime.

Mutations accumulate over an individual’s lifetime, with an average of 100 mtDNA mutations seen per individual over the age of 70 [6, 24]. This accumulation of mutations can be affected by some of the abovementioned mechanisms to increase mutant mtDNA, and is also linked to the process of ageing and adult-onset disease [24].

Mutations in mtDNA have been implicated in neuromuscular and neurodegenerative disorders, as well as more complex diseases such as diabetes, cardiovascular disease, cancer, and aging [1, 2]. To date, over 300 mutations on the mtDNA have been associated with human disease, and affect 1/5000 individuals in the population [15, 24, 25]. Mutations in mtDNA mostly cause mitochondrial dysfunction via inhibition of OXPHOS biogenesis through mutations of OXPHOS complexes, or by impairment of mitochondrial translation through mutations causing instability in tRNAs and rRNAs. Reduction in ATP production- the biochemical change that occurs due to these mutations- can lead to the cell malfunction and clinical phenotypes [7].

3. Changing mtDNA heteroplasmy using DNA editing enzymes

Mitochondrial genetic diseases can severely affect quality of life. Patients can develop hearing and vision loss, muscle weakness, loss of coordination, strokes, and diabetes, to name a few [26–29]. Currently, there is no effective treatment for these patients, and rather than addressing the cause of disease, doctors can only manage patient symptoms to slow decline in health by prescribing various vitamins and supplements, exercise, speech therapy, physical therapy, occupational therapy, and respiratory therapy [30].

Scientists have been interested in changing the ratio of mutant to wild-type mtDNA, by eliminating mutant mtDNA in cells in order to rescue the biochemical defect. As the ratio for disease manifestation is >70–90% mutant mtDNA, it is possible to ameliorate symptoms by reducing mutant mtDNA levels by 20–30%. This would be a powerful approach, as there is no need to replace all or half of the mutant mtDNA copies, as would be the case for nDNA genes. In the last 20 years, there has been growing interest in gene therapy, and gene-editing enzymes have been found. Below we discuss the endonucleases used so far to alter mitochondrial heteroplasmy. Each architecture, which are depicted in Fig. 1 has its pros and cons, which are summarized in Table 1.

Figure 1. Different mitochondrial nucleases used to change mtDNA heteroplasmy.

Mitochondrial targeted restriction endonuclease, Zinc-finger nuclease, TALEN and Tev-TAL hybrids have been used in experimental models (cultured cells and mouse) to change mtDNA heteroplasmy.

Table 1:

Advantages and Disadvantages of Current Methods to eliminate mutant mtDNA

Method	Advantages	Disadvantages
RE	+ Small size	−Limited use for naturally occurring target sites
	+ Specific
ZFN	+ Specific	−Heterodimeric
	+ Can be engineered to recognize specific target sites	−Difficult to generate highly specific tools
TALEN	+ Specific	−Heterodimeric
	+ Can be engineered to recognize specific target sites	−Large
	+ Easy to generate highly specific tools
Tev-TALE	+ Specific
	+ Can be engineered to recognize specific target sites
	+ Monomer
CRISPR/Cas9	+ Small size	−Mitochondria may not have RNA import mechanism
	+ Specific
	+ sgRNA can be engineered to target any sequence

RE, Restriction Endonuclease; ZFN, Zinc-Finger Nuclease; TALEN, Transcription Activator-Like Effector Nucleases; CRISPR, Clustered Regularly Interspaced Short Palindromic Repeats

Mitochondria-targeted restriction endonucleases

Mitochondria do not have an efficient repair mechanism after double strand breaks (DSB) are generated. Moreover, recently it was shown that linearized mtDNA is rapidly degraded by the components of the replisome, including mitochondrial polymerase gamma (POLG) and the nuclease Mgme1 [31–33]. Mitochondria-targeted restriction endonucleases (mitoREs) are recombinant REs that are expressed in the nuclear-cytosolic compartments and reach the organelle due to mitochondrial localization signals (MLS), which are N-terminus polypeptides that help direct the protein into the mitochondria. Once imported, the mitoREs recognize specific sequences in the mtDNA and cause double-strand breaks. This method has been used to selectively cleave one type of mtDNA in a heteroplasmic cell or mouse line, causing shifts in heteroplasmy. A cybrid cell line containing both mouse and rat mtDNA was the first model to provide proof of principle for this mechanism. Mouse mtDNA contains two PstI sites, whereas rat mtDNA contains none. Expression of mito-PstI in this heteroplasmic cell line led to a significant shift in heteroplasmy, decreasing the mouse mtDNA [34].

The transient expression of SmaI in a patient-derived cell line of Leigh’s disease carrying the heteroplasmic mt8993T>G mutation was achieved using mitochondria-targeted SmaI. The SmaI recognition sequence was specific for the point mutation, but not the wild-type mtDNA. This selective cleavage led to a reduction in mutant mtDNA levels in cell culture (from 98% mutant mtDNA to 60%). Restoration of total mtDNA levels, cellular ATP levels, and membrane potential were also observed, indicating that transient expression of mitoREs were not detrimental to the cells [35].

MitoREs were also effective in vivo using the NZB/BALB heteroplasmic mouse model. ApaLI has one recognition sequence on the BALB mtDNA, but not on NZB mtDNA. It was observed that using recombinant adeno-associated virus type-6 (rAAV6) and adenovirus type-5 (rAd5) BALB mtDNA was eliminated in heart and liver, respectively [36]. Similar results were observed using rAAV9 [37].

Mitochondria-targeted REs are effective at cleaving mutant mtDNA and shifting heteroplasmy, but they are limited by the paucity of unique recognition sites created by pathogenic mutations.

Mitochondria-targeted Zinc-Finger Nucleases

To overcome the barrier presented by mitoREs, scientists found ways to engineer endonucleases to recognize longer specific target sites of interest. Zinc Finger Nucleases (ZFNs) work as dimers that have both a Cys2His2 zinc finger DNA-binding domain and a FokI DNA-cleavage domain. The DNA-binding domain can be altered to recognize a relatively long and specific target sequence, and FokI cleavage domain, once dimerized, induces a double-strand break [38].

Patient-derived cybrid cells carrying the m.8993T>G mutation associated with neuropathy, ataxia, and retinitis pigmentosa (NARP) transiently transfected with mtZFN showed a reduction of mutant mtDNA, and restored mitochondrial respiratory function. mtZFN were also used to recognize the “common deletion,” a 4977-bp deletion associated with Kearns-Sayre and Pearson’s syndromes and effectively reduced mutant mtDNA load in cells [39].

More recently, mtZFNs have been used (in vivo) in a heteroplasmic mouse model carrying the m.5024C>T mutation in mt-tRNA^Ala [39]. This mouse model has a decrease in total mt-tRNA^Ala levels when mutation levels are high [40]. After systemic administration of rAAV9-mtZFN, animals showed significant decreases in mutant mtDNA in the heart, and an increase in tRNA^Ala, which showed that the treatment rescued the biochemical phenotype [39].

Mitochondria-targeted Transcription Activator-Like Effector Nucleases

Another family of programmable DNA-editing enzyme that has been engineered for gene therapy purposes are the Transcription activator-like effector nucleases (TALENs). Similarly to ZFNs, TALENs are made up of a DNA-binding domain fused to a FokI endonuclease domain, and function as dimers. The DNA-binding domain is made up of modules, each containing 34 amino acids, where the 12^th and 13^th amino acids recognize the targeted mtDNA sequence. These determinant amino acid residues are called repeated variable di-residues (RVDs) [41]. Because two DNA binding domains are needed, one of them was designed to bind specifically to the mutant mtDNA site, whereas the second monomer binds both wild-type and mutant mtDNA (Fig. 2A).

Figure 2. mitoTALEN recognition of specific mtDNA sequences.

Panel A illustrates how one of the mitoTALEN monomers binds to a mutant-specific region, whereas the other monomer would bind both mutant and wild- type molecules. Panel B illustrates the mechanism of heteroplasmy shift. After a double-strand break triggered by the specific mitochondrial-targeted nuclease, linearized genomes are rapidly degraded. The reduced mtDNA levels lead to an increase in mtDNA replication of the residual molecules, thereby increasing wild-type mtDNA levels.

mitoTALENs have been designed in vitro to detect the common deletion (m.8483_13459del4977) and the Leber’s hereditary optic neuropathy mutation (m.14459G>A). Both constructs successfully shifted heteroplasmy towards wild-type in patient-derived cybrid cell lines [42]. MitoTALENs created for myoclonic epilepsy with ragged red fibers (MERRF) (m.8344A>G) and for the mutation in the ND5 gene associated with MELAS/Leigh syndrome (m.13513G>A) have also been described [43]. As described for other systems, the conceptual framework is that the DSB of the mutant mtDNA will lead to a transient depletion (mostly because of the high levels of mutant mtDNA), followed by a repopulation of mtDNA molecules with a higher percentage of wild-type molecules (Fig. 2B). This change in heteroplasmy would provide the essential OXPHOS component, previously missing, and restore OXPHOS function and a normal mitochondrial membrane potential (ΔΨm) (Fig. 2B).

rAAV9-mitoTALENs have been tested in the heteroplasmic mouse model harboring m.5024C>T mutation. After systemic administration, significant shift towards wild-type mtDNA was achieved in skeletal muscles after focal (intramuscular) and systemic injection. These animals also showed a recovery in total mt-tRNA^Ala levels, indicating a rescue in phenotype [44].

A factor that limits clinical application of mitoTALENs is that they are large heterodimers, which make them difficult to be packaged into a unique viral vector. As an alternative architecture that could circumvent this limitation, the monomeric GIY-YIG homing nuclease from T4 phage (I-TevI) targeted to mitochondria was explored (mitoTev-TALEs). The I-TevI catalytic domain, which is relatively non-specific, was fused to the TALE DNA binding domain. The TALE binding domain specifically targeted the anti-sense strain, where the point mutation m.8344A>G cite contains a cysteine. This was discriminated from the thymine in the same position in the wild-type strain. Two monomeric mitoTev-TALEs were made, differing in size of the TALE DNA binding domain RVDs. These molecules were tested in patient-derived cybrids harboring m.8344A>G point mutation, associated with myoclonic epilepsy and ragged-red fibers (MERRF). MitoTev-TALE effectively decreased mutant mtDNA and increased oxidative phosphorylation of the treated cybrid cells. Transient depletion of mtDNA was observed in treated cells, so further characterization would be required to assess the safety of the mitoTev-TALE architecture [45].

mitoTALENs have also been used to target the MELAS mutation in iPSCs. One report showed engineered mitoTALENs targeting the m.13513G>A mutation, and mutant mtDNA levels were decreased after transduction [46]. Another report showed an iPSC cell line that harbored the m.3243A>G mutation, and a mito-TALEN against this mutation successfully eliminated mutant mtDNA, both in iPSC cells and in porcine oocytes via direct injection of the mitoTALEN mRNA. It was suggested that this method could be used for the prevention of transmission of mutant mtDNA in advanced IVF [47].

Mitochondria-targeted CRISPR/Cas9

The Clustered Regularly Interspaced Short Palindromic Repeats (CRISPR)-Cas9 DNA-editing system has revolutionized gene editing. The CRISPR-associated (Cas) genes work with a single guide RNA (sgRNA) to edit sequence-specific DNA targets. However, for applications of mtDNA editing, the results are controversial and not conclusive. There is one report claiming that they were successful in using CRISPR to cleave mtDNA in HEK-293T cells. They expressed sgRNAs specific for the mtDNA genes Cox1 and Cox3, and hoped that enough Cas9 localized to mitochondria. They described almost 80–90% reduction of intact mtDNA in HEK-293T cells [48, 49]. Other studies, mostly unpublished, have not been able to get mitoCRISPR to work. Because there is no natural mechanism for the import of RNA into mitochondria, the mechanism involved was puzzling. Extensive discussions on the possibility of having a CRISPR-like system working in mammalian mitochondria has been discussed [49, 50]. Again, the main limitation is the ability to import RNA into mitochondria, which in itself, is a controversial topic [50]. The use of a CRISPR-like architecture in mitochondria would be very powerful, particularly because of the recent advances in DNA base editing tools [51].

4. Viral vectors used to deliver mitochondria-targeted nucleases

Adenoviruses

As mentioned briefly before, recombinant Adenovirus type 5 (rAd5) has been used in vivo in NZB/BALB mice to deliver mito-ApaLI to the mitochondria. rAd5 expressing mitoREs was successful in eliminating BALB mtDNA in heart [36]. Although rAd5 can carry a 7kb insert, and has been used in cancer gene therapies, it is not popularly used for in vivo applications due to its short expression window and promotion of a strong immunological response [52].

Adeno-Associated Virus

Adeno-associated virus (AAV) is the most commonly used viral vector for gene therapy mainly because of long-term expression of the transgene and a history of well-established clinical safety [53]. Recombinant AAV (rAAV) is mostly non-pathogenic, with reduced immunogenicity to patients [54]. Accordingly, wild-type AAV in a non-pathogenic virus and is not linked to any human disease or symptoms [55]. A wide range of tissue and cell subtypes can be transduced using different serotypes of AAV vectors [56, 57]. Scientists have worked on further altering the envelopes of these viruses to alter tissue specificity [58, 59]. Since the gene is expressed episomally, the low frequency of integration adds to the safety features [60, 61]. Episomal expression is also persistent in tissues, especially slow-dividing and post-mitotic tissues [61]. This is a helpful characteristic that can be used as an advantage to express therapeutic genes in muscles and nerves of patients suffering from neuromuscular disorders, such as those associated with mitochondrial dysfunction.

The viral vector does however have some setbacks. Namely, a limited packaging size of less than 5kb (including viral ITRs), which limits the size of the transgene that can be packaged and efficiently delivered [62]. Although the safety profile of AAV is good, there have been recorded immune response in patients exposed to AAV [63]. This is why in clinical trials and treatments using AAV, the patients are tested for previous exposure to the specific AAV serotype and/or treated with immunosuppressants [64].

Administration of AAV based therapies depends on the tissue in need of treatment and the disease. Either administration systemically (venous system) or by local injection directly in the tissue can be used. Several clinical trials have used systemic venous delivery to target the liver, as this organ is induced by a broad array of AAV variants. However, as mentioned before, systemic injection may induce immune reactions making it difficult, if not impossible to administer an effective gene therapy treatment more than once [62].

Transduction of post-mitotic tissues such as muscle overcomes the problem of needing multiple doses, as the transgene remains episomal, apparently expressing the transgene for years. With a single injection directly into the muscle, the transgene protein can act locally or systemically. This is the case for Glybera®, which is an rAAV1-based gene therapy for lipoprotein lipase deficiency (LPLD), directly injected into the thigh [62].

Direct administration of rAAV into the CNS has also been explored as an avenue, especially since it has been difficult to create therapies for the CNS that cross the blood brain barrier (BBB). rAAV2, rAAVrh10, and rAAV9 have all been used in clinical trials to achieve CNS transduction of the recombinant transgene. This route is especially necessary for diseases such as Alzheimer’s and spinal muscular atrophy (SMA) [65, 66]. Alternatively, rAAV9 has been shown to cross the BBB even with systemic delivery, which is currently being evaluated in clinical trials for the treatment of SMA (AVXS-101) [62].

Though there are many AAV serotypes to be used, many have broad tropisms that may transduce non-target tissues. This problem can be solved by altering existing AAV capsid proteins to more selectively transduce specific tissues, and decease off-target transduction [67]. Rational design, directed evolution, and in silico methods have all been used to engineer new viral capsids with more precise tissue specificities [48]. Tissue specificity can also be achieved by using tissue-specific promoters. The most common promoter used for mitochondrial nucleases so far is the cytomegalovirus (CMV) ubiquitous promoter [48]. However, tissue-specific promoters such as α1-antitrypsin (AAT) promoter for liver transduction [68], PGDF and NSE for neurons [69, 70], and desmin for skeletal muscle [71], may be required in human trials.

5. Non-viral vectors

The main drawback of using viral vectors are the immunogenic effects they can induce. Using non-viral vectors would circumvent this problem. Non-viral vectors, such a nanoparticles, are used to transfer large plasmid DNA molecules, small DNA molecules in the form of oligodeoxynucleotides, and RNAs. Nanoparticles for chemical non-viral nucleic acid delivery usually come in the form of lipoplexes (DNA/catatonic lipid), polyplexes (DNA/catatonic polymer), and lipopolyplexes (DNA/catatonic polymer/catatonic lipid). Inorganic nanoparticles can be made of silica or gold, while organic particles come in forms of lipid emulsions, lipid nanoparticles, and a wide variety of organic polymers. Delivery of these molecules is not always straight-forward; requiring physical methods such as direct injection, electroporation, sonoporation, and magnetofection to induce gene delivery [72, 73]. The use of nanoparticles for gene therapy of mitochondrial diseases have not yet been reported.

6. Challenges for human application of mitochondria-targeted DNA editing enzymes

DNA editing enzymes have successfully transduced and specifically shifted mtDNA heteroplasmy in cell culture and animals, but they each present their distinctive obstacles for mitochondrial gene therapy in patients. REs are only applicable if they recognize a unique mutation in the mtDNA, which is rare. Although ZFNs and TALENs can overcome this obstacle since they can be programmed to recognize almost any desirable sequence, their dimeric nature and large size make it difficult to compact both halves of the enzyme into a single viral vector. Using two independent viral particles raise costs and technical issues, as both halves of the enzyme complex would need to be transduced into the same cell to function. The search for highly specific monomeric enzymes is certainly going to continue to overcome this hurdle.

Immunological barriers exist for AAV based gene therapy. As mentioned earlier, AAV can give an immune response, and patients that receive AAV-based treatment are tested for previous exposure to the AAV capsid or treated with immunosuppressant drugs. Because of this, AAV-based gene therapies must also be given in a single dose, as subsequent administration will be neutralized by the immune system. Further engineering of new AAV serotypes with reduced immunological response would broaden availability of therapies to patients. Altering the viral capsid or reducing the number of viral genes are both methods that can be used.

It is challenging to produce AAV viral particles in high enough titers that are needed for human patient application, though with further clinical use and research this limitation will be overcome [55]. However, the highest dose may not always be the most effective. A recent paper analyzed the effects of an AAV9-mtZFN in decreasing amounts of mutant mtDNA in a mouse model carrying the m.5024C>T point mutation in the tRNA^Ala gene in three separate doses; low, intermediate, and high titer. They found that the highest titer had severely depleted total mtDNA, due to off-target cutting of the wild-type mtDNA. Furthermore, the intermediate titer showed no significant mtDNA depletion, and did just as good or better job at shifting heteroplasmy towards wild-type, and rescuing the biochemical phenotype in mouse heart [74].

One of the most important hindrances of gene therapy application to humans is the cost to perform clinical trials and consequently the treatment itself. Therapies approved by the FDA can cost millions of dollars (such as Glybera®, which costs US$1.2 million, and Luxterna, which costs US$425,000 per single eye dose). This high price is mostly a reflection of the cost of developing the reagent into a drug. Reducing cost of production by finding more efficient ways to produce high-titer virus in amounts needed can be a first stepping stone to increase the ease of the process and make them more affordable [75].

7. Other current gene therapy approaches to mitochondrial diseases

Gene therapy for nuclear genes coding for mitochondrial proteins have been attempted in mouse models. Torres-Torronteras and colleagues restored normal nucleoside pools in plasma, small intestine, skeletal muscle, brain and liver of Tymp−/− mice, a model of Mitochondrial neurogastrointestinal encephalopathy (MNGIE). In this experiment, they used lentiviral particles expressing thymidine phosphorylase [76]. Flierl and colleagues restored ANT1 to a mouse model of Mitochondrial myopathy associated with muscle weakness and progressive external ophthalmoplegia (PEO) caused by mutations in heart–muscle isoform of ANT1. Recombinant AAV carrying the mouse Ant1 cDNA was injected in the gastrocnemius muscle of the mouse KO for the muscle isoform of ANT1, and improvements in the histopathology were detected [77]. Recombinant rAAV9-tafazzin resulted in phenotypical improvement in a Barth syndrome mouse model [78].

The combination of intravenous (IV) and intracerebroventricular (ICV) injection of rAAV2/9-hNDUFS4 in Ndufs4-KO newborn and young mice, show partial improvements in disease phenotypes [79]. Our group has recently shown that a mitochondrial complex I subunit (NDUFS3) deletion in mouse skeletal muscle was associated with to a severe myopathy with mitochondrial proliferation. rAAV9-Ndufs3 systemic injections led to a reversion of the myopathy [80].

Clinical trials are currently ongoing for LHON, using rAAV-mediated allotopic expression of ND4, which consists in the expression of a re-engineered, normally mtDNA-coded gene in the nucleus [81]. Because the eye is an immune-privileged tissue, in theory multiple doses can be given to the patient over a lifetime without worry of immune response [62]. Phase III Clinical Trials for allotopic rAAV2-ND4 for Leber’s Hereditary Optic Neuropathy (LHON), is now underway. We believe the pre-clinical work on this model did not fully explore the difficulties of allotopic expression of ND4. The ND4 protein is extremely hydrophobic which makes it difficult to be imported through the mitochondrial import pores [82]. Moreover, the allotopic expressed protein would have to compete with the endogenous one for assembly into complex I. The results so far have not been promising, with both experimental and control eyes behaving similarly. Therefore, the results from the phase III clinical trial did not meet end points, although it was found to be safe [83]. LHON clinical trials for allotopic mitochondrial gene replacement therapy are summarized in Table 2.

Table 2:

Current Ongoing Clinical Trials for Mitochondrial Allotopic Gene Replacement Therapy for LHON

Clinical Trial	Study	Disease	Gene/Treatment	Phase	Endpoint	Sponsor
NCT02652767	Efficacy Study of GS010 for the Treatment of Vision Loss up to 6 Months From Onset in LHON Due to the ND4 Mutation (RESCUE)	LHON	rAAV2/2-ND4, intravitreal injection	Phase 3	Visual Acuity, High Resolution Spectral Domain Optical Coherence Tomography to Measure Optic Nerve RNFL Thickness/Volume, Color Hue Vision Test, Immune Response	Gensight Biologics, Paris, France
NCT03293524	Efficacy & Safety Study of Bilateral IVT Injection of GS010 in LHON Subjects Due to the ND4 Mutation for up to 1 Year (REFLECT)	LHON	GS010	Phase 3	Visual acuity, Adverse Effects	Gensight Biologics, Paris, France
NCT02064569	Safety Evaluation of Gene Therapy in Leber Hereditary Optic Neuropathy (LHON) Patients	LHON	GS010 (rAAV2/2-ND4)	Phase 1	Safety and Tolerability	Gensight Biologics, Paris, France
NCT03406104	RESCUE and REVERSE Long-term Follow-up (RESCUE/REVERSE)	LHON	GS010	Phase 3	Adverse Events, Visual Acuity, GCL Thickness/Volume	Gensight Biologics, Paris, France
NCT02161380	Safety Study of an Adeno-associated Virus Vector for Gene Therapy of Leber’s Hereditary Optic Neuropathy (LHON)	LHON	scAAV2-P1ND4v2 (various titers)	Phase 1	Toxicity	University of Miami, Miami, FL USA
NCT03153293	Safety and Efficacy Study of Gene Therapy for The Treatment of Leber’s Hereditary Optic Neuropathy	LHON	rAAV2-ND4	Phase 2/3	BCVA, RNFL Thickness/Volume, Liver and Kidney Function in Plasma	Huazhong University of Science and Technology, Wuhan, China
NCT03153293	Safety and Efficacy Study of Gene Therapy for The Treatment of Leber’s Hereditary Optic Neuropathy	LHON	rAAV2-ND4	Phase 2	Visual acuity, RNFL Thickness/Volume, VEP, Liver and Kidney Function in Plasma	Huazhong University of Science and Technology, Wuhan, China

LHON: Leber’s Hereditary Optic Neuropathy; RNFL: Retinal Nerve Fiber Layer; GCL: Ganglion Cell Layer; BCVA: Best Corrected Visual Acuity; VEP: Visual Evoked Potential