Volume and Infusion Rate Dynamics of Intraparenchymal Central Nervous System Infusion in a Large Animal Model

Introduction

Recent advances in adeno-associated viral (AAV) gene therapy have shown profound success in treating several rare diseases as illustrated by the recent approval of two AAV drugs, Luxterna and Zolgensma for treatment of Leber's congenital amaurosis and spinal muscular atrophy (SMA), respectively.^1,2 The success of these two programs is in part due to the accessibility of the target cells to AAV. In the case of Luxterna, subretinal injection was able to deliver the gene therapy directly into the affected retinal pigmented epithelial cells.³ With Zolgensma, intravenous AAV9 delivery has been well characterized to widely transduce motor neurons of the spinal cord, which is the primary site of cellular dysfunction in SMA.⁴

Although cerebrospinal fluid (CSF) and intravenous delivery routes are less invasive, preclinical data in large animals has shown limited transduction of deep brain structures, which may impact the overall therapeutic efficiency in neurodegenerative diseases affecting the thalamus and/or caudate-putamen.^5–7 Therefore, to achieve transformative outcomes in neurodegenerative diseases with pathologies in these regions, parenchymal injections may be required.^8–10

There are several ongoing clinical trials utilizing brain parenchymal injections in Parkinson's, Batten diseases, aromatic l-amino acid decarboxylase (AADC) deficiency, metachromatic leukodystrophy, and mucopolysaccharidosis type IIIA.^11,12 These trials utilize two different injection strategies to provide widespread delivery of therapeutic enzymes to the central nervous system: (1) intraputaminal injections for treatment of discrete pathologies associated with Parkinson's disease and interconnected structures, or (2) multiple cortical or white matter injections delivered across the cerebral cortex.^13–16 These strategies utilize brain connectivity to achieve broad delivery to distant affected areas through anterograde or retrograde transport of AAV and/or therapeutic proteins.^8,17

Similarly in preclinical studies using large animal models, bilateral injections into the thalamus provide widespread distribution throughout the cerebral cortex, due to extensive connectivity of the thalamus.^8,17–19 For this reason, thalamic injections are appealing in the delivery of AAV for neurodegenerative diseases that affect all brain areas, including deep brain structures. Tay-Sachs and Sandhoff diseases are global neurodegenerative disorders caused by a mutation in the alpha and beta subunits, respectively, of the enzyme β-N-acetylhexosaminidase (Hex;EC 3.2.1.52).²⁰ In these diseases, the thalamus is the most severely affected brain structure in patients with the earliest pathological changes notable on magnetic resonance imaging (MRI).²¹

Preclinical work has shown profound success in small and large animal models after combined bilateral thalamic and CSF delivery, thereby warranting a clinical translation of this delivery strategy.^18,22,23 However, thalamic injections have never been done in humans, let alone in the developing brains of pediatric patients. Although this strategy has been shown to be safe in mice, cats, sheep, and nonhuman primates, injection directly into the thalamus has several inherent risks, including thalamic bleeding and thalamic pain syndrome.^{8,18,22,24,25} In addition, the injection volumes that were efficacious in animals, when scaled up for the human brain, are large and would require prolonged anesthesia time based on the infusion rates used in preclinical studies.^{18,23,25–27}

Therefore, in this study we tested several infusion rates and volumes in sheep using a flexible infusion cannula to determine the optimal safe rate and volume of injection for clinical application.

Materials and Methods

Fiducial-based head registration

All animal procedures were conducted in accordance with the guidelines of the University of Massachusetts Medical School Institutional Animal Care and Use Committee (IACUC). Ten sheep (male, 3 months, 13–20 kg) were used for this study. Sheep were premedicated with buprenorphine (0.02 mg/kg), acepromazine (0.05 mg/kg), and glycopyrrolate (0.01 mg/kg) intramuscularly. Sheep were anesthetized using midazolam (0.3 mg/kg) and ketamine (10 mg/kg) intravenously, intubated, and anesthesia was maintained using isoflurane gas during the procedure. Meloxicam (0.25 mg/kg) was administered subcutaneously for pain management.

To register the sheep head to the acquired magnetic resonance imaging (MRIs), an array of fiducial markers (Rogue Research, Quebec, Canada) was used. Six MRI-sensitive adhesive-backed fiducials were adhered to the circular portion of each fiducial peg. The fiducial markers were visualized as disk-shaped hyperintense marks on the MRIs, which were visualized with the Brainsight neuronavigation software (Rogue Research). To implant the fiducial array, the skin of the forehead was aseptically prepared. A midline skin incision ∼5 cm long was made over the frontal bone and a post was attached to the frontal bone using ceramic screws and bone cement (eSutures.com, Mokena, IL). A fiducial marker array was then attached to the post in the MRI.

Anatomical MRIs were acquired in a Phillips 3T scanner (Phillips Ingenia 3T; Philips Healthcare, Best, The Netherlands) using either a 16-channel knee coil (Philips Healthcare) or an 8-channel cardiac coil (Philips Healthcare). The imaging protocol included three-dimensional (3D) T1-weighted magnetization prepared-rapid gradient echo (MPRAGE) sequence (TR/TE 10/5 ms, flip angle [FA] = 8°, number of averages [NEX] = 8, matrix = 268 × 268, slice thickness = 0.75 mm, field-of-view [FOV] = 200 × 200 mm), and two-dimensional T2-weighted TSE sequence (TR/TE 3,000/80 ms, FA = 90°, NEX = 8, matrix = 220 × 179, slice thickness = 1.8 mm, FOV = 120 × 120 mm).

All fiducials were visible on 3D T1 MPRAGE images. After imaging, the fiducial array was removed and the post was left in place with skin closed above. Images were transferred to Brainsight neuronavigation software for target identification and trajectory planning.

Catheter placement in thalamus

Sheep were premedicated and anesthetized as described previously. Meloxicam (0.5 mg/kg) was used for pain management. The sheep were positioned in sternal recumbency with the head secured to a stereotaxic frame. The skin of the forehead was prepared aseptically, sutures were removed, and the fiducial array was reattached to the post. The fiducial array was used to register the patient using the Brainsight reference coordinate system, which utilizes an optical sensor camera and subject tracker with infrared reflective spheres mounted to the stereotaxic frame. Injection trajectories were planned to avoid the lateral ventricles and major blood vessels.

After calculation of the skull thickness, a 2–3 mm hole was hand-drilled in the skull as the entry points for each thalamus. The flexible catheter (REF:19772; Brainlab, Munich, Germany) containing a stylet was gently advanced along the calculated trajectory to the planned injection site. The flexible catheter was kept in place using the stereotactic arm and secured by seating the bone anchor (with the screw portion removed) against the craniotomy and sutured down to the skin.

The first sheep of this study was injected with 400 μL of phosphate-buffered saline (PBS) at 5 μL/min. Table 1 shows the volume and rate of infusion for the remaining animals in the study. Two thalami were assigned as control and received no infusion after catheter placement. After withdrawal of the stylet, PBS without calcium and magnesium (Thermo Fisher Scientific, Waltham, MA) was injected at the defined rate and volume simultaneously for each thalamus. Sheep were monitored during the surgery and no complications were observed.

After surgery the flexible catheter was removed, bone wax placed in the craniotomy, and skin was closed. Animals were recovered and managed postsurgery with buprenorphine for pain management along with enrofloxacin and/or ceftiofur for antimicrobial prophylaxis if deemed necessary by a veterinarian. Neurological examinations were performed postoperative day 1 and weekly after surgery. Animals remained within normal limits with the exception of one animal that developed lethargy secondary to meningitis. Two weeks and 4 weeks after infusion, 3D anatomical MRI and T2-weighted MRI were performed. Sheep were sacrificed 4 weeks postinfusion using an intravenous pentobarbital overdose (150 mg/kg).

Tissue preparation, immunostaining, and image acquisition

Sheep brains were harvested and sectioned transversely into 6 mm blocks, extending from the frontal lobe to most caudal aspect of the cerebellum/brainstem and further subdivided into hemispheres. Sections were then fixed in 10% neutral buffered formalin. Images of gross brain tissues were taken after fixation and then tissues were processed routinely, embedded in paraffin, and 5 μm thick sections were subjected to standard hematoxylin and eosin staining using a Dako autostainer (Dako Plus; Dako, Carpinteria, CA). Bright field images were captured either using a Leica DM5500 B upright microscope (Leica Microsystems, Buffalo Grove, IL) or an Olympus BX41 with an Olympus DP26 camera (Olympus, Tokyo, Japan).

MRI analysis

To evaluate and quantify the damage as a result of catheter placement and infusion, image analysis was performed using Amira (V2019.1; Thermo Fisher Scientific). The 3D T1-weighted MPRAGE images were reconstructed with the coronal plane at 0.1 mm thickness to capture the lesion region in as many slices as possible due to the small lesion size in several samples. The area containing the lesion was then selected and outlined using the transverse and coronal orientations of the images in the diencephalon region, which was then rendered into a 3D volume to calculate the lesion volume in each thalamus.

MATLAB (MathWorks, Natick, MA) was used to generate the 3D plots showing the lesion volumes from all the animals (Fig. 3A). For one sheep, 3D reconstruction was performed to illustrate the 3D image of diencephalon (Fig. 3B) and the whole brain (Fig. 3C) to depict the needle track and the lesion formation.

Statistical analysis

Analysis of variance (ANOVA) test for mixed models was used to determine if there was a significant effect of the infusion rate and volume on the size of the thalamic lesions. Before running the ANOVA test, the lesion data were normalized with a fractional rank correlation and tested for normality and homogeneity of variances with the Shapiro–Wilk and Levene's tests, respectively. Statistical analysis was performed using the SPSS statistical software package (IBM Corp. Released 2017. IBM SPSS Statistics for Windows, Version 25.0; IBM Corp., Armonk, NY).

Results and Discussion

Clinical trial

This study was conducted in preparation for thalamic injections in infantile and juvenile patients with Tay-Sachs disease.

Objectives and study design

The purpose of this study was to evaluate the effects of injection rate and volume after thalamic injection in sheep to determine the optimal delivery strategy for human patients. Ten sheep (wild type, male, 3 months, 13–20 kg) were used in this study. A total of 19 thalami from 10 sheep were injected using a 14G flexible cannula with phosphate-buffered saline (PBS) without calcium and magnesium at rates ranging from 0.5 to 5 μL/min, with total volumes extending from 0 to 800 μL as described in Table 1. MRI was performed before surgery, 14- and 28-days postsurgery to evaluate thalamic lesion development over time. Animals were sacrificed at 28 days for histopathological analysis.

Summary of data

As a proof of principle experiment, one sheep was injected with 400 μL of PBS containing 2 mM Magnevist (gadopentate dimeglumine) at a rate of 5 μL/min using either (1) a flexible infusion cannula with modified fixation to the skin (14G Brainlab: right side), or (2) a rigid cannula system fixed to a stereotactic frame similar to that used in preclinical animal studies (16G Smartflow Neuro Ventricular cannula, MRI Interventions: left side).²⁶ The rationale for modifying the fixation of the cannula was due to a lack of skull thickness (<5 mm) or stability necessary in young lambs, such as infants, for using the provided bone anchor.

This volume was selected as it was determined to be ∼1/3 the volume of the sheep thalamus, which is projected to be required for complete coverage of the thalamus after injection.^28,29 The thalamic injection site was identified using a presurgical MRI with a fiducial array (Supplementary Fig. S1A–F) and intraoperative neuronavigation (Supplementary Fig. S1D–F). Postoperative MRI revealed widespread fluid distribution bilaterally throughout the thalamus (Fig. 1A–C) and a small air bubble due to injection of the dead volume in the flexible cannula (Fig. 1C right side).

The sheep recovered from surgery uneventfully and had a normal neurological examination for the duration of the study. The sheep was sacrificed 1 month after surgery for histopathological assessment, and a thalamic lesion was identified on the right side of the brain (Fig. 1D–H). At the site of cannula implantation there was a grossly visible roughly linear defect that corresponded histologically to a focal area of necrosis and parenchymal loss accompanied by robust infiltration of lipid-laden macrophages (gitter cells) and, within the adjacent neuroparenchyma, a mild to moderate astrogliosis, spheroid formation, mild rarefaction/vacuole formation, vascular congestion and ectasia, endothelial hypertrophy, and rare perivascular hemosiderin-laden macrophages.

These data prompted a second study using flexible cannulas to define a rate/volume combination with minimal impact to the thalamus (Fig. 2). Similar to the first study, all animals recovered from anesthesia normally and showed no evidence of neurological symptoms from the injection except one sheep that developed lethergy secondary to meningitis. Each thalamus was allocated into an experimental group as shown in Table 1 and MRI visible lesions were noted in each injected thalamus and along the cannula trajectory with varying degrees of necrosis and chronic microhemorrhage (yellow-tan discoloration).

Grossly, lesions were visible in all thalami. The entire lesion was sectioned serially with assessments from in areas with the most severe pathology. Exact correlation of the cannula tip location on MRI with postmortem tissue was not possible due to contraction of the tissue during fixation and slight alterations of the orientation tissue blocks compared with slices on the MRI. However, histopathological examination of the thalami revealed that several animals had focally extensive areas of necrosis, in and around the injection site, with associated infiltrates of gitter cells.

In the adjacent neuroparenchyma, there was reactive gliosis, microhemorrhage, variable numbers of hemosiderin-laden macrophages, increased number and diameter of small caliber vessels, often lined by hypertrophied endothelium, formation of spheroids, scant perivascular cuffs of small lymphocytes, and occasionally mineralization of individual neurons (ferrugination), or the basement membranes of nearby small vessels. At the injection site, edema was noted on MRI and given the presence of perivascular cuffing, immune cell infiltration, and increased diameter of vessels edema is likely a contributor to pathology observed at the tip of the injection site.

In two cases, in the more superficial (cortical) regions along the cannula trajectory there was a variable amount of finely granular basophilic mineralized material admixed with eosinophilic amorphous material lining the cannula track surrounded by numerous multinucleated giant cell macrophages.

Quantification of these lesions based on MRI trended toward an association of larger lesion sizes with larger injection volumes and lower infusion rates. Although an analysis of variance test with mixed models did not show any statistically significant effects, this was likely due to the low sample sizes in the study (Table 1 and Fig. 3). In addition, the modified fixation technique of cannulas to thin skull of young lambs may have resulted in catheter movement during infusion, which likely contributed to lesion size to an unknown degree.

One thalamus injected at 1 μL/min with a total infusion volume of 800 μL may have been mistargeted near the third ventricle based on postoperative MRI, and, therefore, may not be representative of the effect of 800 μL infusion on brain parenchyma (Supplementary Fig. S2). During injection of one thalamus in the 600 μL at 0.5 μL/min group, the inner lumen of the cannula detached and ∼400 μL of injectate was not infused. To improve visualization, three-dimensional (3D) images containing the quantified lesion in the diencephalon (Fig. 3B and Supplementary Video S1), as well as the trajectory and needle track through the brain (Fig. 3C and Supplementary Video S2) are included. The effect of infusion volume, rate, and infusion time on lesion volume has been demonstrated in two-dimensional (2D) plots (Fig. 3D–F).

Conclusions

Infusion of large volumes, which may be required to achieve therapeutic benefit in human patients, will likely require long infusion times. The use of flexible cannulas allows MRI monitoring during infusion and/or awake infusion, thus minimizing the risk associated with prolonged anesthesia as well as the risks of thalamic bleeding, cavity formation, and development of thalamic pain syndrome. Preclinical experiments using AAV.rh8 vectors encoding the therapeutic protein and the appropriate injection volume (based on thalamic size) in sheep, cats, and nonhuman primates did not result in pathology at the injection site.^18,25–27 However, before injection of large volumes of AAV.rh8 in the human thalamus we explored the rate/volume relationship of larger volumes injected in a large brain (sheep).

There appears to be a linear relationship between volume injected and lesion size in this study, where the larger infused volume generates the larger lesion as shown in Fig. 3D; however, it seems to be an upper limit (∼600 μL) where beyond that volume, the generated damage becomes larger than that estimated by linear extrapolation. The largest lesion (94 mm³) has a size of unknown significance compared with size of the thalami in sheep (∼3,000 mm³),²⁹ with the thalamic lesion occupying ∼6.3% of the thalamus after an infusion volume (800 μL) that corresponds to 53% of the thalamic volume (∼1,500 μL). It remains unclear if this lesion size is clinically significant in humans, although the sheep displayed no apparent neurological side effects.

Direct thalamic injection has its pros and cons. Although this method carries potential risk of thalamic bleeding and/or thalamic pain syndrome, it is an efficient method for treating Tay-Sachs and Sandhoff disease in animals,^8,18,24,25 and is where pathology first occurs in humans.^30–32 In addition, direct injection of the thalamus is an attractive delivery location because it is highly interconnected to most of the brain where axonal transport markedly extends distribution, as evidenced by preclinical studies.^8,18,25,27

Animals have smaller brains than humans; therefore, translation requires a scale-up in injection volume and carries aforementioned inherent risks (cavity formation and/or prolonged anesthesia time). For example, in sheep, the thalamus is 1,500 μL and we injected 175 μL per thalamus with a ratio of 12% injectate to thalamic volume.¹⁸ Similarly in nonhuman primates we injected 150 μL per thalamus with a thalamic size of ∼1,200 μL with a ratio of ∼12.5% injectate to thalamic volume (unpublished data).³³ Based on these studies we anticipate that infants with a thalamic size of 4,500 μL will require injection of ∼550–600 μL to replicate successes in preclinical studies.³⁴

An alternative to a single bolus injection is multiple injections of smaller volumes. Stepwise injections along a single needle tract are appealing but are not possible using this flexible cannula system because the rigid stylet must be removed before injection. The lack of rigidity and bending of the cannula during skull fixation make movement of the cannula (either superficially or deep) in the parenchyma during injection challenging. Therefore, multiple injection sites may require multiple needle tracks, increasing surgical risk and brainshift that could result in mistargeting. Therefore, the focus of this study was to determine if there is an optimal safe volume and rate using the flexible cannula for delivery at a single location. The overarching goal was to determine if this methodology might be applicable for awake infusion in pediatric patients.

One aspect that was noted in the use of these flexible cannulas with a modified fixation technique was catheter movement during infusion. This likely contributed to lesion size to an unknown degree, as represented by the presence of lesions larger than the tip of the cannula in the noninfusion controls.

The relevance of this finding will likely be mitigated in pediatric patients as the depth of catheter placement (∼6 cm) will provide a greater distance between the brain surface and the rigid distal section (25.7 mm) of the catheter, which is likely to minimize tip displacement due to small movements of the flexible portion of the catheter outside the skull. In the sheep studies, there was only a small portion of the flexible cannula in the brain parenchyma due to smaller brain size (i.e., shorter depth to the thalamic target). Modification of this anchoring system to be compatible with thin skulls would alleviate the concern of catheter movement and expand the utility of this system to the young pediatric population.

There may be a relationship between lesion volume and infusion rate or infusion time, where higher rates (5 μL/min; total volumes of 200 and 600 μL), corresponding to shorter infusion times, resulted in smaller lesion volumes (Fig. 3E, F). The different rates of infusion (0.5–5 μL/min) used in this study to infuse 200 and 600 μL did not result in large thalamic lesions; however, our study design did not include infusion of the largest volume (800 μL) at the higher rates.

Overall, the results of this study provide insight into assessing well-tolerated upper limits of infusion rate and volume for thalamic infusion in sheep. In AAV gene therapy, initial results in patients suggest that earlier treatment leads to better outcomes. In the case of many lysosomal storage diseases, including infantile Tay-Sachs disease, pre- or early-symptomatic intervention would require AAV infusion before 10 months of age; therefore, adaptation of this infusion system to this young population may be warranted.

Author Disclosure

No competing financial interests exist.

Funding Information

The authors acknowledge financial support from the University of Massachusetts Medical School.

Supplementary Material

Supplementary Figure S1

Supplementary Figure S2

Supplementary Video S1

Supplementary Video S2