AAV Provides an Alternative for Gene Therapy of the Peripheral Sensory Nervous System

Recombinant adeno-associated virus (rAAV) is a promising vector for applications in the central nervous system (CNS). Clinical phase I/II studies have demonstrated safety of rAAV-expressing therapeutic transgenes for a variety of benign (i.e., noncancer) CNS disorders, namely, Canavan disease,¹ Parkinson's disease,^2,3,4 and Batten disease.⁵ Human studies were also recently reported targeting the retina.^6,8,9,9 From the early stages of preclinical development, rAAV was appealing because it was found to be neurotrophic and capable of long-term activity without immune interference or toxicity when injected into the CNS.¹⁰

But what about the potential of rAAV for the peripheral nervous system (PNS)? The PNS has a major role in a large number of neurological disorders, including a common one, chronic pain. The PNS is also, in principle, more easily accessible without major surgery, a potential advantage for any clinical gene therapy application. The PNS consists of three groups of neurons: primary sensory neurons, primary motor neurons, and autonomic neurons. Primary sensory neurons reside in the dorsal root ganglia (DRG)—hence their frequent designation as DRG neurons—and in the trigeminal ganglion. Primary sensory neurons are of clinical interest because of their role in sensory neuropathies and a wide range of chronic pain states.

Testing of rAAV in the PNS was initially reported as a sidekick of an adenovirus marker gene study,¹¹ followed by a report of its usefulness for analgesic treatments.¹² Both studies used rAAV consisting of inverted terminal repeats and capsid proteins derived from AAV serotype 2 (the original rAAV vectors developed in the mid-1990s) and administered the vectors directly into the DRG. In this issue of Molecular Therapy, Mason et al. rekindle this approach, reporting on the targeting of primary sensory neurons by injection of rAAV into the DRG.¹³ The authors packaged AAV2 recombinant genomes into seven different capsids—the original serotype 2 as well as the newer types 1, 3, 4, 5, 6, and 8—and compared their efficacy for expressing the marker gene eGFP. Their best-performing vector, AAV5, transduced >90% of DRG neurons—it can't get better than that. So is it prime time yet for a clinical trial targeting the PNS by direct rAAV injection into the DRG? A critical set of studies may help to light the way.

First, it would be of interest to repeat the study in other animal species to evaluate whether the marked performance difference among the various capsid serotypes reported by Mason et al.¹³ seems generalizable or is rodent-specific. The original reports on capsid pseudotyping of rAAV showed very marked differences between rodent muscle¹⁴ and rodent brain¹⁵ resembling or exceeding the differences between serotypes demonstrated by Mason et al. in the rodent DRG (e.g., in the muscle the difference was several log10 orders of magnitude¹⁴). Unfortunately, in the case of muscle- or brain-directed rAAV gene transfer, similarly systematic serotype comparisons seem not to have been performed in nonhuman primates or other large animals (dogs or pigs could be considered), but the “word on the street” (evaluating single-serotype studies and anecdotal findings in only a few animals) implies differences that are less than overwhelming. In the case of DRG gene transfer, any firmer guidance would certainly help clinical researchers to choose the best rAAV serotype when planning human trials in the PNS.

Second, any piece of evidence supporting the lack of immunogenicity of rAAV vectors would be reassuring. rAAV is notable for its inability to induce cellular immunity in rodents. However, administration of AAV2 into the portal vein in humans led to an apparent delayed immune response that ended a clinical trial prematurely.^16,17,18,19 rAAV has also been used as a platform for vaccine development, suggesting that it can induce strong immunity under some circumstances.²⁰ Perhaps the most interesting observation reported by Mason et al. relates to this problem of rAAV immunogenicity—expression of AAV6 (but not any of the other serotypes) declined 12 weeks after DRG injection and was accompanied by neuropathological tissue damage at the injection site. Interestingly, the authors found that only AAV6 among the vectors tested was able to transduce nonneuronal cells in the DRG, namely, satellite cells. Accordingly, the authors speculate that the less discriminate transduction profile of AAV6 may have led to AAV6 uptake by antigen-presenting cells triggering immunity, which consequently led to tissue damage and loss of transgene expression.¹³ Although these findings are intriguing, the present study leaves open many questions. For example, what happens to expression at later time points (loss of expression is only partial at 12 weeks)? Would neuropathological toxicity hold up in a blinded experiment (random variation in injection technique and tissue procurement remain a possibility)? Finally, what is the nature of the immune response (presumably cytotoxic), and is it directed against the transgene (eGFP), the virus capsid, or both? A careful follow-up study would be most welcome and certainly highly informative for future clinical research.

Third, a lingering concern with rAAV is whether it might be tumorigenic. Wild-type AAV has not been associated with carcinogenesis. But rAAV causes an increased incidence of liver tumors when administered to newborn mice of certain susceptible strains,²¹ in which vector integration was demonstrated in tumor tissues. In at least some of the cases, integration occurred in a micro-RNA locus on mouse chromosome 12. This led to regional transcriptional activation of many small RNA genes with known growth-regulatory properties, a plausible mechanism of carcinogenesis.²² Another study found an increased incidence of liver tumors only with a specific transgene but not with rAAV administration per se.²³ Such findings seem to be limited to rAAV administration in newborn (as opposed to adult) mice and might occur only in specific, susceptible strains (such as knockout mice with lysosomal storage disease) in which liver tumors can occur even spontaneously. They could also have been favored by relatively high vector doses (1.5 × 10¹¹ particles per newborn mouse) and have thus far not been noted in other organs or other species. It is therefore not surprising that no such findings were noted by Mason et al.¹³ Nevertheless, continued careful attention to even a low incidence of tumor formation will need to be an integral part of any preclinical testing.

Fourth, an obvious challenge is the development of a minimally invasive DRG injection technique resembling the method practiced by Mason et al. in rodents but applicable to humans. The human DRG is approximately 50 times larger than that of a rodent, e.g., approximately 150–200 mm³ in humans²⁴ as compared with 4–5 mm³ for the rat.²⁵ As a result, single-site injections will not work. Bankiewicz et al. tackled this problem in the CNS through the development of an administration technique that they refer to as convection-enhanced delivery, which they studied in different species, including primates.^26,27 They defined the technique in terms of biomechanical details such as flow and pressure,^28,29,30 tested it for safety with different vector types,^31,32,33 developed imaging for it,^34,35,36 and refined it down to such details as the needle to be employed.^37,38 From a neurosurgical perspective, the DRG is a structure that is routinely visualized in the scope of many spine operations. But spine surgery would ideally not be required for injecting rAAV. Instead, a minimally invasive technique would be desirable, particularly given that clinical syndromes involving the PNS would ultimately call for treatment of DRG at multiple spinal levels and possibly bilaterally. The absence of a clinically established procedure for injecting DRG in humans calls for some creative surgical research.

For the past decade, PNS gene therapy has been the domain of herpes simplex virus, a vector that transduces primary sensory neurons effectively in animal models when injected subcutaneously.^39,40 Many crucial questions, such as those asked above for rAAV, have been successfully resolved for herpes simplex virus, prompting ongoing clinical trials in patients with intractable pain.^41,42 For rAAV-based gene therapy of the PNS, it took several more years after the initial studies^11,12 for a more comprehensive picture to emerge that covered various administration techniques such as intrathecal delivery of various rAAV serotypes,^43,44 intraneural and systemic targeting of DRG neurons by rAAV6 (ref. 45), and long-term efficacy for the treatment of neuropathic pain.^46,47 The study by Mason et al.¹³ adds seminal data refocusing on rAAV injection directly into the DRG and further supports the candidacy of rAAV as an alternative for gene therapy of the PNS.