A Safe and Reliable Technique for CNS Delivery of AAV Vectors in the Cisterna Magna

Abstract

Global gene delivery to the CNS has therapeutic importance for the treatment of neurological disorders that affect the entire CNS. Due to direct contact with the CNS, cerebrospinal fluid (CSF) is an attractive route for CNS gene delivery. A safe and effective route to achieve global gene distribution in the CNS is needed, and administration of genes through the cisterna magna (CM) via a suboccipital puncture results in broad distribution in the brain and spinal cord. However, translation of this technique to clinical practice is challenging due to the risk of serious and potentially fatal complications in patients. Herein, we report development of a gene therapy delivery method to the CM through adaptation of an intravascular microcatheter, which can be safely navigated intrathecally under fluoroscopic guidance. We examined the safety, reproducibility, and distribution/transduction of this method in sheep using a self-complementary adeno-associated virus 9 (scAAV9)-GFP vector. This technique was used to treat two Tay-Sachs disease patients (30 months old and 7 months old) with AAV gene therapy. No adverse effects were observed during infusion or post-treatment. This delivery technique is a safe and minimally invasive alternative to direct infusion into the CM, achieving broad distribution of AAV gene transfer to the CNS.

Cisterna magna injection of AAV gene therapy results in global distribution but is risky in humans. Taghian et al. describe a microcatheter-mediated technique for cisterna magna injection that results in widespread scAAV9-GFP distribution in sheep. This technique was then used to deliver AAV in two children with Tay-Sachs disease.

Introduction

Neurological disorders are among the most difficult diseases to treat because of limited access of therapeutic agents to brain structures. The anatomy of the brain and blood-brain barrier (BBB), which protects against pathogens and toxic agents, significantly inhibits the efficiency of drug delivery to the brain.¹ Gene therapy for the treatment of a variety of neurodegenerative disorders has shown great promise in rodent models.^2,3 However, reaching the same levels of vector distribution and therapeutic efficacy in large animal models has been challenging because of differences in the size and complexity of the CNS.⁴ This disparity has hindered translation of promising experimental therapies to human application. Therefore, customization of gene delivery techniques based on the disease phenotype and target structure is vital to the development of efficacious therapies.

Direct injection into the brain parenchyma is a gene delivery technique that provides therapeutic levels of transgene expression surrounding the site of injection.⁵ This method can be especially suitable for treating diseases that affect specific brain regions^6,7 and, depending on connectivity and/or therapeutic protein/adeno-associated virus (AAV) capsid characteristics, axonal transport can extend distribution.8, 9, 10 While this approach permits lower vector doses compared to other delivery methods, it is less desirable due to its invasiveness. Many neurological disorders affect cells in both the brain and spinal cord; hence, effective treatment will require delivery routes and techniques capable of broad gene transfer throughout the entire CNS. Systemic CNS gene delivery via intravenous (i.v.) injection became possible following discovery of the BBB crossing properties of AAV9.¹¹ Since then, intravenous AAV9 administration has been transformative for patients with spinal cord diseases, best illustrated with the recent approval of Zolgensma to treat spinal muscular atrophy.¹² Unfortunately, exceptional transduction appears to be limited to the spinal cord. Broad, but modest brain transduction after systemic delivery has been a consistent finding across species, except in mice treated neonatally, where diffuse transduction is observed.13, 14, 15, 16 High vector doses are a major drawback of i.v. delivery for neurological disorders, because of strong transduction and expression in peripheral organs and the subsequent risk for toxicity.^13,17,18 Moreover, application of systemic delivery is limited in patients with neutralizing antibodies against the AAV capsid.¹⁹

An alternative for safe gene delivery to the CNS is cerebrospinal fluid (CSF) administration. CSF continuously flows in cerebral ventricles, the subarachnoid, cisternal spaces, and the spinal canal, and its direct contact with the CNS makes it an ideal delivery route.²⁰ CSF delivery circumvents the challenges presented by the BBB and circulating neutralizing antibodies, while reducing the potential risk of toxicity associated with high vector doses administered intravenously.21, 22, 23 CSF is thought to penetrate the brain parenchyma via permeable pia mater, glia limitans, and Virchow-Robin spaces of the glymphatic system²⁴ and can be accessed via injection into cerebral ventricles, lumbar intrathecal space, or cisterna magna.²⁵ Intrathecal gene delivery via lumbar puncture is a convenient delivery strategy because of ease of access and minimal invasiveness. This method results in extensive spinal cord transduction in animal models and patients, with promising outcomes in an ongoing giant axonal neuropathy clinical trial (ClinicalTrials.gov: NCT02362438).^17,21,26,27 Preclinical data in large animals suggest that lumbar intrathecal delivery results in limited distribution to the brain because of distance to intracranial structures and leakage out of the needle track (unpublished data and Ohno et al.²⁸). As an alternative to lumbar delivery, suboccipital puncture into the cisterna magna has been used in animal models to attain intrathecal delivery closer to the brain. This method has resulted in widespread transduction of the spinal cord and brain in animal models.^21,28, 29, 30, 31 However, increased risk associated with medullary injury impedes the clinical translation of this method.^32,33

To circumvent these limitations, we developed a novel catheter-mediated cisterna magna delivery technique. In this study, we employ an intravascular microcatheter that is navigated in the spinal canal from the lumbar region to the cisterna magna for delivery of AAV vectors. We examined safety and biodistribution of this delivery technique by administering self-complementary (sc)AAV9-CB-GFP into the CSF of sheep. We also report the first-in-human use of this delivery technique in two patients receiving gene therapy for Tay-Sachs disease.

Results

Comparison and Selection of CSF Delivery Route

In preparation for human clinical trials, we determined the maximal infusion volume that could be safely delivered without exceeding 1.5-fold of the CSF opening pressure in lambs, because this is the threshold for intracranial pressure-related clinical decompensation in patients. Infusion at a rate of 1 mL/min in sheep resulted in mild increases in CSF pressure until 15- to 20-mL infusion volumes were reached, after which pressures dramatically increased to a maximum of 76.5 cm of H₂O (∼4-fold greater than the normal opening pressure in humans; Figure S1). To weigh the risk/benefit of lumbar intrathecal versus cisterna magna delivery, we injected methylene blue dye by intrathecal puncture in either the lumbar intrathecal space (between the fourth and fifth lumbar vertebrae [L4-L5]) or cisterna magna (Figure S2). Distribution after lumbar delivery resulted in extensive distribution to the spinal cord and cerebellum, but almost no dye was found surrounding the cerebrum (Figure S2C). After cisterna magna delivery, dye distribution covered the occipital and frontal cortices, with the highest dye concentration along the ventral aspect of the brain, indicating superior spread by this method. Cerebellar and midbrain distribution was also improved after cisterna magna as compared to lumbar intrathecal delivery. However, spinal cord distribution decreased caudal to the cervical intumescence (Figure S2D). Combination of cisterna magna and lumbar intrathecal delivery yielded the best coverage of the brain and spinal cord (Figure S2E). Leakage of dye out of the needle track into the epidural and subcutaneous tissues was noted at both locations.

Effective but Damaging Application of an Epidural Catheter in the Cervical Intrathecal Space

Due to the superior delivery of dye to the brain after cisterna magna injection, we tested lumbar intrathecal placement of a catheter designed for epidural use followed by cranial advancement (Figure 1). At the high thoracic region, the catheter met resistance where movement was stopped. Gadolinium-based contrast agent was then infused (25 mL at 1 mL/min) while under MRI scanning to evaluate distribution. One hour after initiation of infusion, contrast was visualized in the brainstem, cerebellum, midbrain, and the most caudal aspect of the occipital cortex (Figure 1B). At 2 h, contrast agent extended around the dorsal most aspect of the occipital cortex and ventral aspect of the brain (Figures 1C and 1E). By 4 h, contrast enhancement was present in the superior and inferior sagittal sinuses as well as the parietal and frontal cortices (Figures 1D and 1F–1I). Post-mortem analysis of the spinal cord revealed contusions adjacent to nerve roots of the high thoracic region, near the end of the catheter (Figures 1J and 1K).

Figure 1.

Placement of an Epidural Catheter in the Intrathecal Space of Sheep

(A) T1-weighted MRI image before infusion of a gadolinium-based contrast agent. (B–D) Dynamic T1W MRI with images collected (B) 1 h, (C) 2 h, and (D) 4 h after contrast. Red arrows point to the distribution of gadolinium (white) and enhancement of brain regions at different time points. (E and F) Sagittal view of a 3D reconstruction of the whole brain and cervical region of spinal cord at (E) 2 h and (F) 4 h after contrast injection. (G) Ventral view of 3D reconstruction after 4 h. (H and I) Transverse (H) and coronal (I) views illustrating the distribution of gadolinium (white) and enhancement of the brain parenchyma. Contusions on the lateral (J) and ventral (K) aspect of the spinal cord from the catheter near nerve roots of the high thoracic region. The epidural catheter met resistance at the high thoracic region of the spinal cord. This location is not captured in the brain MR image.

Cisterna Magna Delivery Using an Intravascular Microcatheter Is Safe

Microcatheter advancement in the spinal canal was performed under fluoroscopic guidance by a neurointerventional radiologist trained in the delicate manipulations of these catheters. Smooth guidance of the microcatheter in the subarachnoid space of the spinal canal was achieved (Video S1) and the final placement of the microcatheter was in pre-medullary subarachnoid space or foramen magendie (Figures 2A and 2B; Video S2). Figure 2C shows the microcatheter tip placement, rostral to the cisterna magna (red arrow) during the infusion. To assess placement and immediate distribution in the cisterna magna, we infused iodinated contrast material in the cisterna magna and pericerebellar space (Figure 2D; Video S3) prior to administration of 15 mL of scAAV9-CB-GFP vector (1×10¹⁴ vector genomes [vg]) or PBS (n = 3/cohort). Approximately half of the sheep developed transient hyperthermia, tachycardia, and tachypnea, but they otherwise recovered well from anesthesia. Sheep were sacrificed 3 weeks following injection to assess GFP distribution in the CNS. No damage to the spinal cord or nerve roots was noted (data not shown).

Figure 2.

Intravascular Microcatheter Placement in the Cistern Magna of Sheep

(A) MRI and computed tomography (CT) overlay of microcatheter (red line) location in pre-medullary subarachnoid space or (B) foramen magendie. Pink arrow points to the top of the microcatheter. (C) Fluoroscopy images of the microcatheter during infusion with the guidewire removed, anterior to the cisterna magna. The red arrowhead denotes the radiopaque microcatheter tip. (D) Iodinated contrast administered just rostral to the cisterna magna. Contrast is visualized surrounding the brainstem, caudal to the cerebellum, mesencephalic aqueduct, and a small amount in the third ventricle.

Cisterna Magna Delivery of AAV9 Transduces Cells throughout the Brain and Spinal Cord of Sheep

Microcatheter-based cisterna magna delivery of scAAV9-CB-GFP vector resulted in broad transduction of cells in the brain (Figure 3). The qualitative assessment of brain regions indicated that the frontal, occipital, and parietal cortices exhibited extensive neuronal and glial cell transduction, whereas in the motor cortex glial transduction was primarily observed. The transduced glial cells observed in cortices have the star-shaped astrocyte morphology. Transduction in the hippocampus was primarily neuronal in the dentate gyrus hilus and granular cell layer. Scarce positive neurons were present in the caudate, putamen, and thalamus; however, axons and glial cells with oligodendrocyte morphology located in internal capsule were broadly transduced. Few transduced neurons and glia in the midbrain were noted.

Figure 3.

GFP Expression in the Cerebrum 3 Weeks following Cisterna Magna Administration of scAAV9-CB-GFP

As shown in low magnification images (left), widespread transduction is present from the surface and throughout the deeper layers of the cerebral cortices. Two enlarged sections from each cortical region shown on the right exhibit GFP-positive neurons and glial cells in the cortices. Low-magnification (left) and high-magnification (right) images of hippocampus show widespread transduction of neurons and glia, which are distinguished based on morphology. Subcortical structures with less exposure to CSF show sparse GFP-positive neurons and glia in the thalamus and midbrain, while no GFP expression was detected in caudate and putamen. Subcortical structures with interconnections receiving projections from CSF-exposed cortical regions showed strong transduction as demonstrated in the axons of the internal capsule. Red and blue boxes on the low-magnification images on the left correspond to the first (red) and second (blue) high-magnification images on the right. GFP expression, brown; cell nuclei, blue. Scale bars, 25 μm.

Strong GFP staining was noted in the cerebellum, including Purkinje cells, deep cerebellar nuclei and adjacent axons in the white matter, cerebellar peduncles, and brainstem (Figure 4). Robust neuronal transduction was observed along the entire length of the spinal cord, from cervical to lumbar regions (Figure 4). Qualitative evaluation of spinal cord indicated that GFP staining of motor neurons in the ventral horn was greater than that of sensory neurons in the dorsal horn, with the most intense staining in the cervical intumescence. Strong GFP staining was also noted in dorsal root ganglia and nerve root axons, again with the greatest intensity in the cervical intumescence (Figure 4). Examination of H&E-stained slides showed normal histology of CNS tissue as illustrated in Figure S3. No significant non-specific binding of GFP antibody in untreated sheep tissue was observed (data not shown).

Figure 4.

GFP Expression in the Cerebellum, Brainstem, and Spinal Cord 3 Weeks following the Cisterna Magna Administration of scAAV-CB-GFP

Robust GFP expression was present in the cerebellum and throughout the entire length of the spinal cord. Low-magnification images (left) show widespread GFP expression in the cerebellum, deep cerebellar nuclei, cerebellar peduncles, and the brain stem. High-magnification images (right) show the strong GFP expression in Purkinje cells, neurons, and axons in white and gray matter. Low-magnification images (left) of spinal cord at the cervical and lumbar regions demonstrate that the transduction of ventral gray matter is stronger than in the dorsal region. As illustrated in high-magnification images (right), motor neurons of cervical and lumbar regions expressed high levels of GFP, with the highest expression in the cervical intumescence. Red and blue boxes on the low-magnification images on the left correspond to the first (red) and second (blue) high-magnification images on the right. GFP expression, brown; cell nuclei, blue. Scale bars, 25 μm.

Biodistribution of AAV9 in Peripheral Tissues 3 Weeks following Cisterna Magna Administration

qPCR was performed to assess the biodistribution of scAAV9-CB-GFP in select peripheral tissues at 3 weeks after vector administration. Figure S4 shows the differential biodistribution of AAV9 genome copies in peripheral tissues. The highest genome copies were detected in the uterus (∼3.7 × 10⁵ vg/100 ng of genomic DNA) of the pregnant sheep, followed by optic nerve (∼2.7 × 10⁴ vg/100 ng of DNA). To determine the ability of AAV9 to transduce autonomic and peripheral nerves after intrathecal delivery, qPCR was performed in the vagus nerve, sympathetic chain, vagosympathetic trunk, and phrenic and sciatic nerves (average values range from ∼2.7 ± 4.3 × 10³ vg/100 ng of DNA to 4.7 ± 3.9 × 10³ vg/100 ng of DNA). The average copy numbers in liver was 1.1 ± 0.7 × 10³ vg/100 ng of DNA. The lowest copy numbers were detected in cotyledon (fetal placenta). In sexual organs, testis (1.7 × 10³ vg/100 ng of DNA), ovary (1.2 ± 0.05 × 10³ vg/100 ng of DNA), and cotyledon (1.3 × 10² vg/100 ng of DNA) had fewer genome copies as compared to the uterus. AAV genome copies were below the level of detection in retina, placenta, and fetal liver.

Microcatheter Navigation under Fluoroscopic Guidance Safely Delivers AAV Vectors via the Cisterna Magna in Patients

After evaluating safety and distribution of our microcatheter delivery method in the ovine CNS, two patients with Tay-Sachs disease were treated with a 1:1 formulation of two separate AAV vectors encoding the alpha (HEXA) and beta (HEXB) subunits of hexosaminidase. The first treated patient was a 30-month-old child. Intrathecal access was attained by placement of a Tuohy needle at L4-L5 level of the patient’s spinal canal under fluoroscopic guidance by a combination of expertise from the neurosurgery and neurointerventional radiology teams (Video S4). Intracranial opening pressure was 41 cm H₂O, which was >2-fold above normal for this age (∼10–12 cm H₂O). Increased CSF opening pressure in Tay-Sachs patients has been previously reported.³⁴ The pressure decreased to 21 cm H₂O after removal of 14 mL of CSF and remained unchanged with AAV delivery. The microcatheter was threaded through the Tuohy needle and advanced cranially (Video S5). The softness and flexibility of the microcatheter during the navigation is notable, illustrated by delicate redirection after getting caught on nerve roots. The guidewire was then removed (Figures 5A and 5B). Final placement of the microcatheter was at the premedullary cistern as illustrated by overlaying a fluoroscopic image of the microcatheter onto the patient’s brain MRI (Figure 5C). The AAV vector formulation was injected (∼1 mL/min) at the premedullary cistern (9 mL) and then the microcatheter was lowered to the lumbar spinal cord (Figure 5D) where a second injection (3 mL) at the L2 level was performed (Video S6). No adverse effects were noted from the infusion procedure.

Figure 5.

Cisterna Magna and Lumbar Intrathecal Microcatheter Delivery in a 30-Month-Old Patient

(A) Microcatheter with guidewire in the intrathecal space terminating at the premedullary cistern (red arrowhead). (B) Microcatheter without the guidewire during infusion with a radiopaque terminus (red arrowhead). (C) Overlay of fluoroscopy and T1-weighted MRI images of the patient illustrating microcatheter location (pink arrow indicates terminus of the microcatheter). This overlay may not illustrate the exact location of the microcatheter due to spatial differences between imaging modalities (X-ray versus MRI). (D) lumbar intrathecal placement of the microcatheter (L2 level) during the lumbar infusion (radiopaque terminus is highlighted by the red arrowhead).

A second patient treated with this method was a 7-month-old child diagnosed with Tay-Sachs disease. Under fluoroscopic guidance an 18G Tuohy needle was placed in the spinal canal of the patient at the L2–L3. The opening pressure was 28 cm H₂O. Following removal of 5 mL of clear CSF, the microcatheter was advanced into the cisterna magna under fluoroscopic guidance (Figure 6A; Video S7). The catheter location at the cisterna magna was confirmed following injection of 1 mL of Omnipaque 180 (Figure 6B; Video S8). Final placement of the microcatheter is displayed in Figure 6C, in which a fluoroscopic image of the microcatheter is overlaid onto the patient’s brain MRI. AAV vector formulation (3.75 mL) was injected at the cisterna magna at ∼0.25 mL/min. The Tuohy needle was then removed followed by retraction of the microcatheter to the T12–L1 interspace spinal level, and an additional 1.25 mL of vector was injected over 5 min (Figure 6D; Video S9). The microcatheter was then withdrawn. A sterile bandage was placed at puncture site. The patient tolerated the procedure well without complication.

Figure 6.

Cisterna Magna and Lumbar Intrathecal Microcatheter Delivery in the 7-Month-Old Patient

(A) Microcatheter without the guidewire in the intrathecal space terminating at the cisterna magna (red arrowhead). (B) Intrathecal space following injection of the contrast (black) in the cisterna magna space. Arrowhead points to the tip of the microcatheter, and the arrow shows the contrast distributed around the cerebellum. (C) Overlay of fluoroscopy and T1-weighted MRI images of the 7-month-old patient. The perceived increased volume of CSF in infant’s brain is due to benign extraaxial fluid accumulation of infancy and is considered a normal finding due to delayed maturation of arachnoid villi and CSF absorption, which normalizes at 2–2.5 years of age. The pink arrow points to the terminus of the microcatheter. (D) The lumbar intrathecal placement of the microcatheter (T12-L1) during the lumbar infusion (radiopaque terminus is highlighted by the red arrowhead).

Discussion

Despite overwhelming preclinical evidence showing superior AAV delivery by cisterna magna over lumbar intrathecal administration, to date, lumbar intrathecal delivery has been the only CSF route used in clinical application of gene therapy.²⁷ Recently, two clinical trials have started that use cisterna magna infusions (RGX-111 gene therapy in patients with MPS I, and RGX-121 gene therapy in patients with MPS II; ClinicalTrials.gov). Until now, patient risk has outweighed the potential benefit of delivering therapeutic vectors to a location with the most efficacious outcome. In this study, we have developed in sheep a novel intravascular microcatheter-based technique that overcomes the safety challenges associated with off-the-needle cisterna magna injections in humans. While catheter delivery in the spinal canal has been previously reported, human translation to date has yet to be documented.35, 36, 37 The lack of advancement to clinical application may be due to the risk of bruising and damage to the spinal cord caused by rigid epidural catheters (Figures 1J and 1K); however, this damage could potentially be prevented by use of a guidewire during the advancement of the rigid catheter in the spinal canal. The ability to direct the trajectory is unique to intravascular microcatheters, making them an excellent delivery tool via the spinal canal. The delivery technique reported here is applicable in both the laboratory setting for large animal studies, as well as for investigational use in clinical practice. Compared to the lumbar intrathecal route alone, combined delivery to the cisterna magna and lumbar space using a single microcatheter enables widespread vector distribution to the cerebral cortex, cerebellum, and spinal cord. This approach may be useful in the treatment of neurological disorders that could benefit from CNS-wide gene delivery, while bypassing the BBB. This method may allow for multi-site delivery along the spinal cord, which may more effectively address global diseases of the CNS.

Cortical or spinal cord distribution of AAV vectors can be achieved by intracerebroventricular (i.c.v.) or lumbar injection, respectively. Nonetheless, cisterna magna injection results in improved penetrance to both the brain and spinal cord in non-human primates.^21,28,29,38 These data support the hypothesis that the degree of exposure to CSF determines the level of transduction.³⁹ The relatively poor transduction observed in deep brain structures in this study (i.e., the caudate, putamen, and thalamus) could be attributed to their relatively limited exposure to CSF compared to the cortex. As such, we hypothesize that the GFP-positive axons of the internal capsule likely originated from cortical neuronal projections. Direct parenchymal injection results in superior transduction of deep brain structures,8, 9, 10^,40 and it may be required for effective treatment of aggressive neurodegenerative diseases, primarily affecting the caudate-putamen or thalamus. Our observation that substances injected into the cisterna magna first accumulate on the ventral aspect of the brain followed by relatively slow distribution to the cortices is consistent with a previous report;²⁸ therefore, the lack of transduction in structures in close proximity to the ventral aspect of the brain (e.g., ventral thalamus) is surprising. In addition, antisense oligonucleotides (ASOs) infused intrathecally display a similar brain distribution profile with limited delivery to deep brain structures.⁴¹ The fact that AAVs and ASOs, which are entities of different sizes and chemical composition, share the same limitation suggests that distribution to deep brain structures is governed by a fundamental aspect of CSF physiology, which remains to be elucidated. Further investigation is required to rule out other causes, including promoter specificity and/or reduced capsid affinity of these regions. However, similar phenomena are documented across capsids and different chemical entities.^21,41, 42, 43

In agreement with previous studies, vector delivery solely to the cisterna magna results in robust transduction of the entire spinal cord.^21,28,29 However, our results show that motor neuron transduction in the lumbar region was reduced as compared to the cervical region. This is likely because of the length of the sheep spinal cord (∼56 cm), which is considerably longer than that of non-human primates, but more in line with that of humans (∼47 cm).⁸ For more uniform coverage of the spinal cord in the Tay-Sachs patients, we delivered 75% of the AAV dose to the cisterna magna and then retracted the microcatheter to the lumbar region and injected the remaining 25%. It is worth noting that the postural differences in sheep (quadrupedal) versus human (erect posture) may affect AAV residence time in cisterna magna before it distributes to the CNS and subsequently result in species differences in distribution of vector.

In agreement with previous studies, biodistribution in peripheral tissues indicates that gene transfer by AAV delivery to CSF is not confined to the CNS.44, 45, 46 The highest genome copies were noted in the uterus, and we hypothesize that pregnancy may have contributed to this high level. The relevance of these data is unknown, as nonhuman primates receiving cisterna magna injection of scAAV9.CBA.GFP demonstrated that vector distribution often does not correlate with transgene expression or mRNA levels in peripheral tissues.⁴⁴ Biodistribution in the liver was modest, recapitulating data from non-human primates receiving AAV9-GFP via cisterna magna injection with sparse GFP expression in the liver and other peripheral tissues.⁴⁵ Serotype likely plays a role in peripheral tissue transduction, where comparison of AAV6 and AAV9 vectors after intracerebroventricular (i.c.v.) injection in mice showed that AAV6 is more restricted to the CNS while AAV9 showed high levels of transgene copies in liver.⁴⁶ Although intravenous injection of AAV9 in adult mice has been reported to result in transduction of all layers of the retina, despite the presence of a mature blood-eye barrier, in our study, we did not observe any genome copies in retina following cisterna magna delivery of AAV9 in sheep.⁴⁷ Ruminants have epitheliochorial plastination, with the most restrictive transport of all species, and therefore this unique barrier may explain non-detectable or low levels of genome copies in placenta, cotyledon, and fetal liver.

The rate of CSF delivery has been reported to have a profound effect on vector distribution, with rapid infusion superior to slow infusion.²⁸ With i.c.v. delivery, the rate is largely determined by the size of the ventricles, with higher rates resulting in reflux up the needle track (our unpublished data and Ohno et al.²⁸). With cisterna magna delivery, a much faster infusion rate is possible due to the increased size and distensibility of the space. In this study, AAV was delivered at the highest rate that maintained an intracranial pressure below 1.5-fold of the opening pressure. This parameter is clinically accepted by approximation, and intracranial pressure should dictate the rate of delivery in patients. Similar to a previous report,³⁴ higher than normal CSF opening pressure was documented in both Tay-Sachs patients, so appropriate amounts of CSF should be removed prior to AAV administration.

Other methods have been used to increase vector distribution after vascular delivery, including co-administration of mannitol^48,49 and/or use of novel AAV capsids that can cross the blood-brain barrier.^13,50 However, i.c.v. administration of the current generation of AAV capsids results in negligible penetration into the brain parenchyma and, thus far, only modest improvements are achieved by engineered capsids (e.g., AAV2.5).²¹ Another potential strategy is to transduce ependymal cells (AAV4 and AAV5), which can produce and secrete therapeutic proteins into the CSF,^51,52 but clinical trials based on this approach have yet to start. Until new AAV capsids are discovered or engineered with superior CNS transduction profiles (distribution and percentage of transduced cells) after CSF delivery, intracranial administration may still be required to achieve transformative outcomes in diseases that affect deep brain structures.

In this study, we describe a novel microcatheter-based cisterna magna infusion technique developed from preclinical studies and its translation to a first-in-human AAV gene therapy clinical trial for Tay-Sachs disease. Our preclinical studies in sheep and clinical trial results in two patients to date suggest that intravascular microcatheters can be successfully adapted for CSF delivery of AAV to the cisterna magna and lumbar intrathecal space. By decreasing the risk to patients, this new approach could facilitate broad implementation of cisterna magna-like intrathecal AAV gene delivery to the CNS. Such fundamental advances may positively influence the success of AAV gene therapy for neurological diseases.

Materials and Methods

AAV Vectors

The GFP vector, scAAV-CB-GFP, consisted of an AAV serotype 9 capsid carrying an expression cassette comprising a version of the chicken β-actin promoter, a GFP cDNA, and an artificial intron.¹⁶ The GFP vector was produced by triple transfection as previously described and purified by cesium chloride gradient. Formulation was then dialyzed in Dulbeccos’s PBS without calcium and magnesium (Thermo Fisher Scientific, Waltham, MA, USA).^53,54 The vector was injected at a dose of 1 × 10¹⁴ vg in a total volume of 15 mL per sheep. The vectors used in the patients were based on a two-vector system. Two separate monocistronic AAVs (AAV.rh8) were utilized in this treatment, with each vector packaged with either the HEXA or the HEXB transgene, purified, and dialyzed in PBS without calcium and magnesium. Both vectors were mixed and co-injected at a 1:1 ratio, at a total dose of 1 × 10¹⁴ vg for the first patient and 5 × 10¹³ vg for the second patient. The AAV dose was administered to the patient based on the brain weight. The dose was determined, scaled up from preclinical experiments performed in mice, non-human primates, cats, and sheep.^8,9,55, 56, 57

Animals and Surgery

All animal procedures were conducted in accordance with the guidelines of University of Massachusetts Medical School Institutional Animal Care and Use Committee (IACUC), Auburn University IACUC, or the South Australian Ethics Committee (AEC).

Intrathecal Pressure Testing

A sheep was anesthetized using midazolam (0.4 mg/kg) and ketamine (10 mg/kg), intubated, and anesthesia was maintained using isoflurane gas during the procedure. A 22-gauge ½-inch spinal needle was placed in the intrathecal space between L4 and L5. Opening pressure was measured using a manometer as shown in Figure S1. PBS without calcium and magnesium was injected at a rate of 1 mL/min using a Medfusion syringe pump (Smiths Medical, Dublin, OH, USA). Using a three-way stopcock, CSF pressure was measured every 5 min for a total of 25 min. The animals recovered normally and were euthanized 24 h later using a pentobarbital overdose (150 mg/kg i.v.), and brains and spinal cords were removed for post-mortem analysis.

Epidural Catheter Placement in the Intrathecal Space and MRI

Sheep were anesthetized (n = 2) using a mixture of diazepam (1 mg/kg) and ketamine (10 mg/kg; Ceva Animal Health, Glenorie, NSW, Australia), intubated, and anesthesia was maintained using isoflurane gas during the procedure. The skin was aseptically prepared, a Tuohy needle was then placed in the lumbosacral junction (L7–S1), and the epidural catheter was placed intrathecally using iohexol contrast agent (Omnipaque 240, GE Healthcare). One catheter coiled on itself and could not be advanced past the lumbar region. The second catheter was advanced into the thoraco-cervical region and stopped once resistance was encountered at the high thoracic region. The catheters were used without guidewire. The sheep was recovered from anesthesia and the catheter remained implanted for 48 h. The sheep was re-anesthetized for MRI, and a T1 MPRAGE dynamic scan was performed (Siemens Healthcare, Malvern, PA, USA). Gadolinium-based contrast was injected (25 mL) at a rate of 1 mL/min. Contrast distribution was measured over 4 h. After the infusion, the sheep were euthanized by pentobarbital overdose (150 mg/kg) intravenously and the spinal cord was removed for post-mortem evaluation.

Delivery Route Selection

Sheep were anesthetized using midazolam (0.4 mg/kg) and ketamine (10 mg/kg), intubated, and anesthesia was maintained using isoflurane gas during the procedure. The skin was aseptically prepared and a 22-gauge ½-inch spinal needle was placed in the intrathecal space at between L4 and L5 and/or at the cisterna magna (atlanto-occipital junction). Methylene blue dye (1%) (Sigma-Aldrich, St. Louis, MO, USA) or eosin dye (Sigma-Aldrich, St. Louis, MO, USA) at a total volume of 5 mL was injected at a rate of 1 mL/min. After 5 min, animals were euthanized and brains and spinal cords were removed.

Intravascular Microcatheter

The intravascular microcatheter (SL-10, Stryker Neurovascular, Fremont, CA, USA) used in this study is a single-lumen device with a lubricious outer surface coated with Hydrolene. It is composed of a highly flexible tip (length, 6 cm; inner diameter, 0.42 mm; outer diameter, 0.60 mm), semi-rigid proximal section (outer diameter, 0.80 mm), and a dead space of 0.29 cm³. It can be used with a steerable guidewire (diameter, 0.36 mm) that enables accessing distal vasculature and, in our case, allows navigation of the spinal canal and reaching to the cisterna magna. To control the proper introduction, movement, positioning, and removal of the microcatheter during the interventional procedure, the angiographic and fluoroscopic guidance should be employed. A radiopaque tip facilitates fluoroscopic visualization. This device should be used by physicians trained in performing endovascular procedures. Compatibility study of AAVrh8 vectors with this intravascular microcatheter showed that no losses in titer occurs during the injection of vector through the microcatheter. AAVrh8-HEXA vector (0.5 mL, 9.9 × 10¹² vg/mL) was mixed with AAVrh8-HEXB (0.413 mL, 1.2 × 10¹³ vg/mL) to generate a 1:1 equimolar formulation with a total measured titer of 1.51 × 10¹³ vg/mL in sterile Dulbecco’s PBS without Mg²⁺ or Ca²⁺ (DPBS) (Invitrogen, Carlsbad, CA, USA). The catheter should be saturated with ∼1 mL of saline prior to vector injection to prevent AAV binding; therefore, the catheter was filled with 300 μL of saline, corresponding to all its void volume, using a 1-mL syringe. AAVrh8-hHEXA/B was loaded into a 1-mL syringe, attached to the catheter through the Luer-Lok system, and injected into the catheter as a bolus. The AAV was then chased with 300 μL of saline to ensure that nothing was left in the catheter, using another 1-mL syringe. Everything was collected for later titration. The titers were 1.5 × 10¹³ vg/mL before passage through the microcatheter and were measured to be 1.23 × 10¹³ vg/mL; therefore, data suggest that the AAVrh8 capsid does not interact with the SL-10 microcatheter material or plastic syringes.

Intrathecal Placement of Intravascular Microcatheter

Three (two females, one male) sheep (12 months, 30 kg) were used for this study. The skin was aseptically prepared, and the sheep were anesthetized using midazolam (0.4 mg/kg) and ketamine (10 mg/kg), intubated, and anesthesia was maintained using isoflurane gas during the procedure. Prior to surgery, anatomical MRI images (Philips 3T Ingenia) were acquired. During the injection, animals were placed in left lateral recumbency with lower lumbar spine flexion. The lumbar puncture was performed by using a 17G Tuohy spinal needle at the lumbosacral intervertebral space (L6/L7-S1) and 5–12 mL of CSF was collected. A 1.7-French outer diameter (OD) intravascular microcatheter with a 0.014-inch guidewire (SL-10, Stryker Neurovascular, Fremont, CA, USA) was introduced into the subarachnoid space via the Tuohy needle. Under fluoroscopic guidance, the microcatheter was navigated into the cisterna magna or premedullary cistern with the “J” shape wire tip advanced slightly distal to the catheter. Cone beam computed tomography (Allura Xper FD20 system) was used to confirm adequate catheter position relative to bony and neural structures. After confirming the catheter position, 1 mL of iodinated contrast (Omnipaque 240 mg/mL, GE Healthcare) was slowly injected to identify the distribution pattern of contrast material prior to vector injection. The Tuohy needle was removed over the wire and a total amount of 15 mL of scAAV-CB-GFP vector was infused at 1 mL/min. Animals were sacrificed 3 weeks post-injection (n = 3).

Tissue Preparation

Sheep were perfused with ice-cold saline. Brains, spinal cords, livers, sexual organs (testis, uterus, and ovary), nerves (sciatic, phrenic, vagus, vagosympathetic trunk, sympathetic chain, optic) and retina were collected. One of the sheep was pregnant during the procedure, and placenta, fetus liver, and cotyledon were collected for that sheep. Peripheral tissues were frozen in a dry ice-isopentane bath and stored at −80°C. Brains were sectioned transversely into 6-mm blocks, extending from the frontal lobe to most caudal aspect of the cerebellum/brainstem and further subdivided into hemispheres. Representative segments of the spinal cord at the rostral cervical, mid-cervical, cervical intumescence, thoracic, thoracolumbar, mid-lumbar, and lumbar intumescence were collected and fixed in 10% neutral buffered formalin.

Immunostaining and Image Acquisition

Paraffin-embedded blocks were cut into 5-μm sections for immunostaining by a Dako autostainer (Dako Plus, Dako, Carpinteria, CA, USA) for GFP (1:2,000, ab290, Abcam, Cambridge, MA, USA) in Dako antibody diluent (S0809, Dako, Carpinteria, CA, USA) at room temperature. Sections were washed with PBS and incubated with secondary antibody (K4063, Dako, Carpinteria, CA, USA) for 45 min at room temperature. Slides were counterstained with hematoxylin. Standard H&E staining was performed for histopathologic analyses. Bright-field images were captured using a Leica DM5500 B upright microscope (Leica Microsystems, Buffalo Grove, IL, USA). The tiled images of whole sections were acquired using a Leica Aperio CS slide scanner (Leica Microsystems, Buffalo Grove, IL, USA). Virtual slides (tiled images) were viewed with Aperio image scope software (Leica Microsystems, Buffalo Grove, IL, USA). To ensure that GFP staining on sheep tissues were specific, brain and spinal cord tissue sections of naive sheep control tissue were processed and stained in parallel.

Biodistribution Analysis

Vector genome copy numbers from various peripheral tissues (liver, nerves, sexual organs) were determined by qPCR after extraction of total DNA using a DNeasy blood and tissue kit (QIAGEN, Hilden, Germany). Vector genome content in each tissue was determined using 100 ng of total DNA by qPCR method using primers and probes for the BGH poly(A) in the vector (forward primer, 5′-CCTCGACTGTGCCTTCTAG-3′; reverse primer, 5′-TGCGATGCAATTTCCTCAT-3′; probe, 56-AM/TGCCAGCCA/ZEN/TCTGTTGTTTGCC/3IABkFQ). The lower limit of detection in the qPCR assay was 100 genome copies (gc)/100 ng of DNA, and any sample below this limit was considered non-detectable. Experimental results are shown as mean ± SD. A Kolmogorov-Smirnov test was performed using GraphPad Prism version 8.2.0 (GraphPad Software, San Diego, CA, USA) to evaluate the statistical difference.

Patient Treatment

The study was approved by the Institutional Review Board at the University of Massachusetts Medical School. Written informed consent was obtained from parents of the children. Two patients with Tay-Sachs disease were injected either at the premedullary cistern or cisterna magna and lumbar spinal cord using the technique described above. Patients were positioned at left lateral decubitus. For treatment of the 30-month-old child (12.1 kg), propofol (Fresenius Kabi, Bad Homburg, Germany) infusion 200 μg/kg/min and isoflurane (Piramal Critical Care, Bethlehem, PA, USA) at 1.5% was used to induce and maintain anesthesia. The Tuohy needle was placed at L4-L5. A total of 14 mL of CSF was removed by passive flow, and the procedure was followed by administration of AAVrh8 vector formulation in a total of 12 mL at ∼1 mL/min, 9 mL at the premedullary cistern and 3 mL at the L2 level of the spinal cord. For treatment of the 7-month-old child (10.8 kg), anesthesia was induced using propofol (50 mg) and maintained by 2% sevoflurane. The Tuohy needle was introduced to the spinal canal at L2-L3 and a total of 5 mL of CSF was removed. Final placement of the catheter at the cisterna magna level was confirmed by injection of 1 mL of Omnipaque 180. A total of 5 mL of AAVrh8 vector at ∼0.25 mL/min was administered, 3.75 mL at the cisterna magna level and 1.25 mL at the T12-L1 level of the spinal cord.

Author Contributions

T.T. performed the immunohistochemistry and microscopy, assisted in running the MRI scanner, performed sheep necropsies in flexible microcatheter sheep experiments, extracted DNA for biodistribution study, and was the primary author of the manuscript. M.G.M. co-developed the microcatheter delivery technique and performed the sheep injections. A.S.P. placed and navigated the microcatheter in the spinal canal of two patients and injected the vector for the first patient. O.I.C. consulted on the development of the microcatheter delivery technique, performed lumbar puncture in two patients, and injected the vector for the second patient. R.M.K. ran the fluoroscopy and computed tomography in flexible microcatheter sheep experiments. E.B.D. and P.D.G. assisted in intrathecal pressure measurement experiments. A.S.M. assisted in intrathecal injections for pressure measurement experiments. D.R.M. collaborated in CSF pressure measurement and dye injections in sheep experiments. D.F. coordinated the experiments and assisted in sheep necropsies. A.R.B. prepared the vector delivered in sheep and manufactured the vector delivered in patients and performed qPCR for biodistribution and compatibility study. T.K. and C.C. placed the rigid catheter in the spinal canal in the gadolinium injection experiments. R.P. ran the MRI scanner in gadolinium injection experiments. S.C. collaborated in intrathecal catheter placement experiments. S.B. provided daily care for the sheep used for study and coordinated with D.F. to provide sheep for injection at UMMS. T.R.F. is the principal investigator of the trial (expanded access). D.M.-Y. is the clinical nurse. P.W.L.T. performed the deep sequencing analysis for the vector delivered into the patients. N.A. provided the vector for sheep experiments. M.J.G. co-developed the intrathecal catheter delivery technique and is the director of New England Center for Stroke Research where all sheep injection experiments were performed. M.S.-E. sponsored expanded access clinical trial and manufactured the vector for the human clinical trial. H.L.G.-E. co-developed the intrathecal catheter delivery technique and oversaw the sheep experiments, downstream analysis, and manuscript preparation.

Conflicts of Interest

The authors declare no competing interests.