Clinical GDNF delivery methods for Parkinson's disease

Abstract

Intracerebral delivery of glial cell line-derived neurotrophic factor (GDNF) therapeutics continues to show promise, especially in Parkinson's disease (PD). However, randomized, placebo-controlled clinical trials using GDNF protein have been inconclusive to date. A major sham-surgery controlled trial using GDNF gene therapy commenced in 2024. In this review we aim to update the reader on the evolution and current state of the art of intracerebral delivery methods for GDNF protein and gene therapy in PD clinical trials. Our intent is to increase the awareness for the importance, subtleties, and pitfalls of intracerebral delivery when reviewing current available results for these therapies and their prospects going forward. We will compare and contrast GDNF protein infusion versus gene therapy strategies and define specific anatomical and physiological details in trial participants that continue to challenge clinicians and investigators attempting to maximize therapeutic coverage of the target putamen. Despite a growing consensus that convection-enhanced delivery (CED) is the optimal intracerebral infusion strategy for localized administration of therapeutics in general and gene therapy products in particular, there is less agreement on the need for related methods, such as the co-infusion of a gadolinium contrast agent and use of intraoperative magnetic resonance imaging (iMRI) to visualize the therapeutic distribution for optimizing target coverage. Whereas certain debates will continue, most investigators and clinicians respond positively to well-designed trials providing clean, conclusive results. Information included in this review is intended to provide additional insights to interested readers, allowing better assessment of details associated with upcoming GDNF therapeutic investigations.

Plain language summary title

Clinical Delivery Methods for Treatments Based on Glial Cell Line-Derived Neurotrophic Factor (GDNF) in Parkinson's Disease

Plain language summary

Treatments based on glial cell line-derived neurotrophic factor (GDNF), a naturally occurring growth factor, have been studied in humans for over 2 decades, especially in Parkinson's disease (PD). These treatments are hoped to stop or even reverse the natural progression of the disease. They can include GDNF as a protein or as gene therapy. Studies comparing GDNF protein head-to-head with a placebo have shown positive trends but no significant results to date. A major study comparing GDNF gene therapy head-to-head with a control in participants with PD commenced in 2024. Neither GDNF protein nor gene therapy can cross the so-called blood-brain-barrier. Therefore, to reach the putamen, a core brain structure affected by PD, both treatments need to be delivered directly into this structure via a neurosurgical intervention. We aim to update the reader on the evolution and current state of the art of GDNF delivery methods in PD. Our intent is to increase the awareness for the importance, subtleties, and pitfalls of putaminal delivery. We will discuss how specific anatomical and physiological details in trial participants impact the ability of clinicians to expose as much of the putamen to drug as deemed necessary. We will also discuss how delivery strategies for GDNF protein and gene therapy differ. Along with this, we will describe how intraoperative magnetic resonance imaging (iMRI) can be used to visualize drug distribution across the putamen when a suitable contrast agent is added to the therapeutic solution. A solid understanding of GDNF delivery is considered important to allow performing well-designed clinical trials that have a high likelihood of providing clean and conclusive results. Whereas certain debates will continue, most investigators and clinicians respond positively to such trials, especially if their assessment is also based on insights into at least some of the subtleties of the underlying methodology.

Introduction

It is now just over 30 years that glial cell line-derived neurotrophic factor (GDNF) was discovered and reported to enhance the survival of midbrain dopaminergic neurons. ¹ While GDNF was soon found to also support the development and maintenance of central noradrenergic neurons, spinal motor neurons, and a variety of peripheral neurons, ² it has been its influence on the dopaminergic system and, thus, as a potentially transformative drug candidate for the treatment of Parkinson's disease (PD) that has attracted most of the scientific interest in this compound over time.^3,4 Using different strategies and technologies for local intraparenchymal delivery of the protein itself or adeno-associated viral vectors serotype 2 (AAV2) carrying the GDNF transgene, this particular growth factor has been shown to protect target dopaminergic neurons from neurotoxic insults and to enable functional and anatomical recovery of previously lesioned, “sick-but-not-dead” dopaminergic neurons in various toxin-induced rodent and nonhuman primate (NHP) models of PD.^5–7 Despite the robust neuroprotective and neurorestorative effects seen in animals, however, clinical results in PD have been mixed, with promising data from early uncontrolled studies of both protein^8–11 and gene therapy,^12,13 followed by insignificant results in placebo-controlled protein infusion trials.^14–16 A similar disappointing pattern has been observed in clinical investigations testing an AAV2 vector carrying neurturin (NRTN), another member of the GDNF-family of ligands (GFL) that has been tested clinically.^17–19

While the consistency of this pattern underscores the undisputable need for adequately sized, robustly designed, placebo- and/or sham-surgery controlled trials, it is of interest that even the failed double-blind, placebo-controlled GDNF protein trials showed statistically significant increases in putaminal 18-flouro-DOPA (¹⁸F-DOPA) uptake, and nominally favored active treatment over placebo in terms of motor outcome.^14,16 Albeit based on small numbers, we interpret this co-directionality of results, which was observed despite substantial inter-study differences relating to core treatment modalities, such as delivery technique, dose, and length of treatment or follow-up, as being indicative of intrinsic drug activity. Based on this underlying assumption, we hypothesize that the full treatment potential of GDNF may not have been brought to fruition primarily because the target structure has not had sufficient exposure to GDNF (a) volumetrically, i.e., not enough target neurons have been exposed to drug-containing infusate, (b) temporally (contact time), i.e., target neurons have not been exposed to drug for a long enough period of time before the prespecified assessments were made, and/or (c) quantitatively, i.e., not enough target neurons have been consistently exposed to high enough, sustained drug concentrations for optimal efficacy. Interestingly, the importance and subtleties of optimized intracerebral delivery are often overlooked or ignored when considering the prospects of advanced therapies for PD, and sometimes even when discussing technical advances related to these therapies.^20,21 This is particularly remarkable given that convection-enhanced delivery (CED), guided by intraoperative magnetic resonance imaging (iMRI), has evolved as the gold standard for direct, localized drug administration in the treatment of PD and other central nervous system diseases. ²²

The clinical effect size of GDNF therapy largely depends on the availability and distribution of the protein within the putamen, and its level of retrograde transport to the substantia nigra.^23–25 More than 90% of the putamen's neuronal population is made up of dopamine receptor (D1- or D2-) expressing, gamma-aminobutyric acid (GABA)ergic medium spiny neurons (MSNs), that are largely preserved in idiopathic PD. ²⁶ Both dopamine receptor MSN populations receive dopaminergic input from the substantia nigra pars compacta (SNpc) via the nigrostriatal pathway, and project back to the substantia nigra pars reticulata (SNpr) through the direct and indirect striatonigral pathways. ²⁷ D1-MSNs are part of the direct pathway (promoting movement) and project to the SNpr via the globus pallidus internus, whereas D2-MSNs are part of the indirect pathway (inhibiting movement), also projecting to the SNpr but instead via the globus pallidus externus and subthalamic nucleus. ²⁷ In PD, the dopaminergic nigrostriatal projections progressively degenerate as a result of dopaminergic cell loss within the SNpc. ²⁸ Specifically, nigral cytoplasmatic aggregates of α-synuclein and other proteins (Lewy bodies) within idiopathic PD's dopaminergic neurons, are associated with functional early stage reductions in kinesin-dependent anterograde transport and eventually later-stage reductions in the dynein-dependent retrograde transport. ²⁹ In contrast, GABAergic striatonigral projections are largely unaffected by the dopaminergic neuronal α-synuclein pathology over the course of PD. ³⁰ This is a key difference versus the neuropathology of the parkinsonian subtype of multiple system atrophy which features early degeneration of both of the above MSN pathways. ³¹

Detailed understanding of the architecture and functionality of the interconnections between putamen and substantia nigra is important for both types of GDNF therapies (protein and gene) being considered. Intraputaminally delivered protein is dependent on presynaptic dopaminergic terminal uptake and retrograde transport to the SNpc somata exclusively via the nigrostriatal pathway.^32,33 As the retrograde transport capacity decreases in parallel with disease progression, the clinical efficacy of this treatment is likely to diminish over time due to progressive loss of normally functioning nigrostriatal neurons. ²⁸ On the other hand, the AAV2 vector carrying the GDNF transgene (AAV2-GDNF), when delivered to putamen, has two options for influencing the SNpc. First, there is transduction of putaminal neurons that then locally produce and secrete GDNF protein, helping sustain putaminal dopaminergic terminals and their SNpc somata (via nigrostriatal retrograde transport, which is likely more available in mild/moderate stages of the disease). Second, the same putaminal AAV2-GDNF delivery allows the AAV vector and payload to undergo anterograde transport (via the largely preserved striatonigral pathway ³⁰ ) to the SNpr, where it transduces resident neurons, providing local production and secretion of GDNF, which then resuscitates and maintains the adjacent SNpc dopaminergic neurons.^30,34

Early trials delivering recombinant GDNF by intracerebroventricular (ICV) infusion and chronic continuous intraputaminal low-rate infusion taught the field that (1) more widespread exposure to GDNF within the CNS may be associated with significant clinical sequelae, including gastrointestinal side effects and sensorineural symptoms, ³⁵ and (2) diffusion-based tissue distribution of GDNF can be substantially limited due to binding to heparan sulfate/heparin molecules within the extracellular matrix,^36,37 especially with small doses being delivered.

With the aim of addressing these two challenges, subsequent work has focused on precise, direct targeted delivery of either recombinant GDNF, or AAV2-GDNF, administered by CED within the putamen. CED provides a pressure-driven extracellular bulk flow that optimizes volumetric distribution of large molecules such as GDNF or AAV2-GDNF within the brain parenchyma. ³⁸ In its current incarnation, CED involves administering the drug over a period of hours by means of a standard microinfusion pump and uniquely-designed catheters/cannulae, precisely placed and monitored using magnetic resonance imaging (MRI) during or immediately following the infusion. ³⁹ Clinical CED infusion rates currently range from 1–15 μL/min, and may be set according to static or variable rate protocols.

The average putaminal volume per hemisphere accounts for less than 0.3% of the total brain volume. ⁴⁰ The GDNF delivery challenge, therefore, appears to be twofold, i.e., consistently exposing large enough volumetric fractions of these target structures to drug to enable a meaningful therapeutic effect, while simultaneously minimizing the exposure of the surrounding non-target brain parenchyma and thereby reducing the probability of any unwanted side effects. Although this may sound simple and straightforward, the procedural complexities of intraputaminal delivery of GDNF are in fact very demanding, with the proverbial devil being in the details. Safe, accurate and efficacious intracerebral drug delivery requires a sophisticated combination of advanced stereotactic neurosurgical skills, including a detailed knowledge of the anatomical subtleties of the target regions, and specific devices and methods designed to optimize infusions of the therapeutic product, including an in-depth understanding of the underlying fluid dynamic principles relating to extracellular parenchymal drug distribution. More detailed considerations are presented in the following sections.

Putaminal targets: Considerations for delivery of GDNF protein and gene therapies

Challenge 1: Putaminal target size

Dissemination of an infusate using CED within a target volume typically follows fluid dynamic principles that differ significantly between diffusion and convection (directed bulk flow) within a target volume's extracellular space.^38,41 These principles, supported by experimentation in animals and humans, have provided an understanding of the relative infusate distributions within a variety of brain targets.^39,42–45 Although small as compared to total brain size, the human putamen is a relatively large structure (commonly ranging from 3.5 to 4.5 cm³ in parenchymal volume) when considering local direct parenchymal drug delivery. Such a large volumetric target size cannot be safely covered using conventional diffusion-based, non-CED infusion techniques. Even with CED, it not only requires longer infusion times, but also necessitates administration of a concomitantly larger infusion volume (Vi) to achieve adequate volumetric distribution and target coverage. Relative to diffusion-based delivery, CED reproducibly provides much greater and more homogenous distribution of therapeutic agents (and over a wider range of molecular weights), along with a steep concentration drop-off at the advancing margin of extracellular bulk flow.^38,39 However, clinical infusion rates $\gg 0.5\mu \mathrm{L}/\min$ are needed to reliably induce convective distribution within the brain's extracellular space.^38,46

In the majority of recent human putaminal CED infusions evaluated, for every 100 µL of Vi delivered within the target, the corresponding volume of distribution (Vd), assessed using MRI-visualized gadolinium contrast co-infusions, varied between 200 to 300 μL, thereby providing a Vd/Vi ratio in these examples of 2:1 or 3:1, respectively. ⁴³ Such a putaminal Vd/Vi ratio is fairly consistent for putaminal infusions in adult humans. The Vd/Vi ratio is greater for CED within highly myelinated target structures (e.g., cerebral white matter, thalamus, or pons) due to reduced extracellular volume associated with increased white matter volumes. Familiarity with such information is critical when considering and planning for putaminal and other target infusions via CED.

Historically speaking, a small Vi used for putaminal infusions has limited the degree of target coverage and, thereby, the potential therapeutic effect in PD. ⁴⁷ Using data from a limited group of PD patients undergoing MRI volumetric analyses (n = 11), baseline putaminal volumes ranged from 3.01 to 5.29 cm³, with a mean volume of 3.98 ± 0.15 cm³. ⁴⁰ There were no significant volumetric differences between right and left sides, and no additional differences noted when compared to a limited number of healthy individuals in their early seventies (n = 13). The mean putaminal volume in this example, therefore, represents 4.0 cm³, equaling 4.0 mL, 4000 μL, or 4000 mm³. If we assume again a Vd/Vi ratio of 3:1 or 2:1, the proposed minimum and maximum Vi required to achieve a theoretical 100% volumetric coverage would be (4000/3 = 1333 µL or 4000/2 = 2000 μL, respectively (assuming all of the Vi contributes to Vd within the putamen, with no leakage outside of the target). A lesser percent coverage requirement would necessitate a smaller Vi, while increased leakage outside the putaminal target would require a larger Vi to attain the target coverage.

We believe that the effects of putaminal GDNF therapies, both protein and gene therapy, are dose- or better: exposure-dependent, i.e., dependent on both coverage and drug concentration. However, this remains to be proven, as there is only limited data allowing direct comparison of different degrees of putaminal coverage, as well as different drug concentrations, and their respective impact on clinical or positron emission tomography (PET) imaging outcome.

Clinical experience

A systematic overview of relevant infusion parameters across all clinical trials testing the intraputaminal administration of GFL members as protein (GDNF) or gene therapy (AAV2-GDNF and AAV2-NRTN) is provided in Table 1.

Table 1.

Infusion parameters in clinical trials testing the intraputaminal administration of GFL members.

Parameter/Study	Gill8,9	Slevin10,11	Lang 16	Marks 19	Marks 17	Olanow 18	Whone14,15	Heiss12,13	Van Laar 48
Publication year	2003	2005	2006	2008	2010	2015	2019	2019	2025
N	5	10	34	12	58	51	41	13	11
Investigational product	GDNF	GDNF	GDNF	AAV2-NRTN	AAV2-NRTN	AAV2-NRTN	GDNF	AAV2-GDNF	AAV2-GDNF
Trajectory (transfrontal vs. occipitoparietal)	T	T	T	T	T	T	T (n = 12)	T	O
							P (n = 3)
							O (n = 26)
No. of tracks per putamen	1	1	1	4	4	3	2	2	1
Cannula outer diameter (mm)	0.64	1.65	1.04	ND	ND	ND	0.6	1.65	1.65
Cannula type (end-hole/step vs. end-hole/no step vs. multi-port)	E/NS	M	E/NS	ND	ND	ND	E/S	E/S	E/S
Infusion strategy (fixed point vs. progressive)	F	F	F	F	F	F	F	F	Pr
Infusion mode (continuous vs. continuous/pulsed vs. intermittent vs. single-dose)	C	C/P	C	S	S	S	I a	S	S
Infusion rate (µL/min)	≤0.1	0.033/10.9 b	0.104	ND	2	3	3–5	≤5	≤20
Vi per putamen (µL)	≤144 (per day)	133 (per day)	150 (per day)	40 (2 x 5 per track)	40 (2 x 5 per track)	150 (50 per track)c	800 (400 per track)	450 (300 + 150)d	<=1800
Mean total putaminal coverage	ND	ND	ND	ND	ND	ND	48–55% e	26% f	63% f

AAV2 = adeno-associated viral vector serotype 2; C = continuous; C/P = continuous/pulsed; E/NS = end-hole/no step design; E/S = end-hole/step design; F = fixed point; GDNF = glial cell line-derived neurotrophic factor; GFL = GDNF-family of ligands; I = intermittent; M = multi-port design; N = sample size; ND = not determined; NRTN = neurturin; O = occipital; P = prefrontal; Pr = progressive; S = single-dose; T = transfrontal; Vi = infusion volume.

^a Repeat treatments every 4 weeks.

^b 117-s pulses given every 6 h.

^c The study employed both intraputaminal and intranigral delivery; delivery to the substantia nigra occurred via 1 track per side, with 30 (2 × 15) µL infused at 2 µL/min.

^d 300 µL were delivered to the anterior target, 150 µL to the posterior target.

^e Determined within 2 h post-end of infusion.

^f Determined immediately post-end of infusion.

Continuous intraputaminal delivery of GDNF protein using a sub-CED infusion rate (0.1 µL/min) and a GDNF concentration of 0.1 µg/µL was associated with very limited and highly heterogeneous drug distribution in the vicinity of the catheter tip, ⁴⁹ and led to a modest 23% median increase in ¹⁸F DOPA uptake in the posterior putamen at 6 months. ¹⁶ In contrast, intermittent protein administration into each putamen via CED (3–5 µL/min), using a Vi of 800 μL (400 μL through each of 2 catheters) at a GDNF concentration of 0.2 μg/μL every 4 weeks, resulted in a mean putaminal coverage of around 50% and a 100% mean increase in ¹⁸F-DOPA uptake in the targeted posterior putamen at 40 weeks. ¹⁴ In a parallel, non-GDNF gene therapy study in PD, increasing the Vi from ≤450 to ≤1800 μL per putamen, and the total dose of human aromatic L-amino acid decarboxylase (AADC) transgene-carrying vector dose from ≤7.5 × 10¹¹ vector genomes (vg) to ≤9.4 × 10¹² vg was not only associated with an increase in MRI-documented putaminal target coverage from ∼20% to ∼76%, but also with commensurate increases in ¹⁸F-DOPA PET signal and specified improvements in clinical signs.^47,50,51

The first-in-human AAV2-GDNF gene therapy Phase 1 trial was performed in advanced PD participants and tested escalating vector doses of 9 × 10¹⁰ vg (n = 6), 3 × 10¹¹ vg (n = 6), and 9 × 10¹¹ vg (n = 1) in a bilateral putaminal Vi of 450 μL, delivered by a minimum of two frontal trajectories per putamen. ¹² The treatment resulted in a mean putaminal coverage of 26%, with increasing putaminal ¹⁸F-DOPA PET signals from baseline to 6 months (median Ki increase by 36%) and further to 18 months (median Ki increase by 54%) across groups, and an apparent dose-dependent stabilization of motor scores over time. ¹² Most recently, promising 18-month data have been reported from a follow-up open-label Phase 1b trial evaluating mild and moderate stage PD cohorts treated with up to 1800 μL AAV2-GDNF per putamen, using the highest vector titer tested so far (3.3 × 10¹² vg/mL). ⁴⁸ Consistent with what was shown in the AAV2-AADC trial mentioned above, the larger AAV2-GDNF Vi administered with CED in the Phase 1b trial resulted in a substantial increase in mean putaminal coverage to 63% and more robust clinical responses than those described for the smaller Vi (450 μL/putamen) in the earlier dose-escalation study . Unfortunately, it is impossible to separate the effects of increased coverage from those of increased vector titer in this trial. Nonetheless, the outcome strongly suggests that GDNF effects in PD subjects are not only dependent on the stage of the disease but also impacted by the putaminal coverage and delivered dose as well. Importantly, both GDNF protein and AAV2-GDNF appeared to be well tolerated and safe at all doses tested in the above studies, and no drug-related serious adverse events were reported.

Challenge 2: Putaminal target shape

In addition to the size of the putamen and coverage by infusate, the non-spherical volumetric shape makes infusate distribution and volumetric coverage without concomitant leakage into adjacent brain regions more challenging. As it turns out, the putamen is shaped more like a kidney lying on its side, with a convex superior and lateral edge and concavities along the medial and ventral contours. The long axis of the putamen extends from a thicker anterior (frontal) pole to a thinner posterior (occipital) pole. The putamen's short axes extend from medial to lateral and dorsal (top) to ventral (bottom). Visualizing such a non-spherical volume on MRI clearly presents a target planning predicament for parenchymal infusions that try to maximize volumetric coverage while minimizing required cannula trajectories. Due to the size and shape of the putamen, therefore, a single-site CED infusion within the target volume is unable to adequately distribute infusate within the entire volume, without significant leakage outside of the target borders.

Access and infusion strategies

Initial infusion approaches to the putamen followed standard bi-frontal neurosurgical trajectories. Although such frontal approaches proved familiar to most neurosurgeons and were safely applied to PD study participants, infusate delivery (with or without CED) paralleled (at best) the dorso-ventral, short-axis of the putamen, provided limited volumetric coverage along such a trajectory (due to relatively short dorso-ventral distance), and typically required multiple (2 or 3) frontal trajectories per putamen to allow greater antero-posterior volumetric coverage (Figure 1). ⁴⁷ As stated above, the first AAV2-GDNF gene therapy trial in PD utilized such frontal delivery approaches, with at least two trajectories, delivering up to 450 μL Vi per putamen, and achieved a limited (26%) putaminal volumetric coverage. ¹² An occipitoparietal approach for intraputaminal delivery of GDNF was first utilized in a recent Phase 2 PD clinical study evaluating intermittent GDNF protein infusions, administered every 4 weeks over a period of 18 months. This trial used 2 catheter trajectories per putamen to deliver a total Vi of 800 µL per hemisphere, without iMRI infusion monitoring, and reproducibly yielded mean putaminal volumetric coverages of 47.8% to 55.0%.^14,15 Of note, however, coverage was to be assessed within 2 h post-end of infusion and may have increased during this time as a result of post-infusion diffusion. As diffusion-based tissue distribution of GDNF protein is limited,^36,37 the observed coverage (by contrast) may have overestimated the coverage by drug. For instance, assuming that the contrast front within the putamen would have advanced by 0.5–1.0 mm after the end of infusion without distributing the protein any further, coverage by drug would only have been around 32–44% instead of 50%.

Figure 1.

Frontal versus occipitoparietal “runways” within target putamen. Using a Frontal Approach , the left panel portrays two example delivery cannula trajectories passing through a dorsal burr hole, targeting the more anterior portions of the putamen (typically pre- and post-commissural). Such dorsoventral trajectories within the putamen are limited by the sagittal short axis “runway” distance (thin double-headed dashed arrows), that are approximately one third the length of the sagittal long axis (compare to right panel). In order to maximize delivery of the putaminal volume of distribution, without significant leakage outside the putaminal borders, at two cannula trajectories are depicted, providing dorsal followed by ventral infusion volumes in series (see numbered gray arrows) for each “trajectory runway”. Note that a third trajectory (not pictured) might be used to access the posterior aspect of the putamen, although likely requiring an additional burr hole and cortical entry site. The Occipitoparietal Approach in the right panel displays the single cannula trajectory entering the putamen via a posterior burr hole. This approach distributes infusate parallel to the long-axis of the putamen using the infuse-as-you-go method and accesses the longer “trajectory runway” (thin double-headed dashed arrow). This combined approach and method help maximize putaminal volumetric coverage with a single cannula trajectory per putamen. The panel portrays serially delivered volumes of distribution (numbered dotted circles) delivered, from posterior to anterior putamen. D-dorsal. V-ventral. A-anterior. P-posterior.

Since then, evolving infusion methods, paralleling the long-axis of the putamen in NHPs, have produced an optimized “infuse-as-you-go” technique.^52,53 This method takes advantage of iMRI guidance of the CED procedure, in order to maximize putaminal volumetric distribution using a single catheter trajectory per putamen and delivering spheroidal infusate volumes parallel to the putamen's long axis, in a stepwise fashion (Figure 1). In addition, the “infuse-as-you-go” technique provides several advantages over stationary fixed-point infusions (Figure 2). We have shown that Virchow-Robin spaces in the striatum may serve as conduits for non-convective fluid distribution,^54,55 making infusate leakage from such pathways from a fixed point source likely to result in diminished volumetric target distribution, even using a larger Vi, due to infusate egress along such low resistance perivascular spaces and not contributing to the intended Vd. Advancing the cannula tip during the infusion, away from such MRI-evident low resistance sumps, allows a greater Vi to effectively perfuse the extracellular space as Vd within the target. In addition, a posterior trajectory to the putamen, utilizing anterior advancement of the cannula tip during the infusion, takes advantage of a shape-conforming paradigm as previously demonstrated (Figure 3).^52,56 Note that retraction of the infusion cannula is not typically permitted during CED (only once the total infusion has been delivered). Cannula retraction creates a low-pressure space for additional infusate to fill and expand rather than promoting extracellular bulk flow distribution from the cannula tip, thereby defeating the goal of CED. When planning cannula advancements for long target trajectories (e.g., putamen), it is often recommended to advance the cannulae shorter distances along the target trajectory, confirm the location with iMRI, define the additional distance to target depth, and make the final depth adjustment, rather than overshooting, since cannula retraction is not advised. Such an evolution of a posterior CED approach to the putamen has allowed the Vi to safely and consistently increase, up to 1800 μL per putamen, with manageable and clinically asymptomatic extra-putaminal leakage volumes noted with iMRI. This posterior CED approach to the putamen was initially translated in the Phase 1 trial of AAV2-AADC gene therapy for PD,^47,57 and has been recently introduced in the Phase 1b AAV2-GDNF gene therapy trial in PD, where it provided a mean putaminal volumetric coverage of 63% at the end of infusion in 11 trial participants. ⁴⁸

Figure 2.

Fixed point versus progressive infusion strategies. The left panel presents a coronal T1-weighted MRI slice at the level of the mid thalamus, with the thalamic outline shown in one hemisphere (dashed white circle). In the opposite hemisphere a typical thalamic CED infusion is shown along with a Delivery Cannula schematic. The cannula tip is placed near the center of the thalamic target. Serial concentric infusion volumes (dashed black circles) are portrayed, previewing the fixed point (concentric) distribution to fill a thalamic target. With such a spherical target volume, little or no cannula movement, after initial placement, is required for maximal CED target coverage. The right panel presents an axial T2-weigthed MRI slice showing the putaminal outlines and a typical occipitoparietal approach to a putaminal target. This hemisphere shows a putaminal Delivery Cannula schematic and portrays the progressive, infuse-as-you-go method, providing sequentially delivered volumes of distribution (dashed line white circles), with each serial volume added with stepwise cannula advancements (denoted by the white arrows), from the posterior to anterior putamen. CED-convection enhanced delivery.

Figure 3.

Volumetric CED distribution rendering within a unilateral human putamen, using infuse-as-you-go methodology, overlayed on day of treatment MRI. The Pre infusion image portrays a single 3-dimensional (3D) volumetric putaminal rendering (in green), used to establish the baseline putaminal volume. Replicated MRI images (from left to right) portray serial ipsilateral CED distribution volumes that are either within the original putaminal volume (orange) or have extended beyond the original putaminal volume (red). Note the increasing total CED infusion volumes delivered to the putamen below each image and the fidelity with which the infuse-as-you-go method allows coverage of the putaminal volume.

Off-target delivery

As larger infusion volumes are used with the aim of increasing the volumetric coverage of putaminal targets, the possibility of leakage of infusate along low resistance pathways, associated with peri-vascular spaces ⁴⁵ or gray matter tracts, ⁵⁸ increases. Such leakage pathways have the potential to produce off-target clinical side effects; it is important, therefore, to assess such potential sequelae longitudinally, with serial MR imaging and clinical follow-up. One of our analyses, focusing on serial MR imaging over 5 years of the 13 participants in the AAV2-GDNF dose-escalation trial, has shown no evidence of significant parenchymal alterations attributable to either the AAV2-GDNF/gadoteridol infusate or the surgical intervention itself. ¹³ A subsequent analysis of a specific leakage pattern (putamen to caudate) in that same study set has provided evidence of infrequent, limited infusate leakage from putamen to caudate, using frontal trajectories to the putamen, a 450-μL Vi per putamen, and confirming no clinical sequelae attributable to vector extension from putamen to caudate. ⁵⁹ Future analyses are required to better appreciate leakage patterns commonly associated with use of the current occipitoparietal trajectories for AAV2-GDNF CED administrations that distribute up to 1800 μL Vi per putamen.

In summary, off-target extension of AAV2-GDNF CED is commonly noted and monitored in near-real time using gadolinium contrast co-infusions and iMRI. To date, these off-target extensions of infusate, to the caudate and structures adjacent to the putamina, have not been associated with clinical sequelae or MRI evidence of associated neuropathology over time.

Alternate targets: Midbrain/nigral/VTA delivery of GDNF gene therapy

A previous human gene therapy trial in PD tested the safety and feasibility of direct transduction with AAV2-NRTN of dopaminergic neurons within the substantia nigra, while attempting to supplement beneficial clinical effects suggested at 12 months following putaminal delivery alone. ⁶⁰ Although both safety and tolerability were re-confirmed in a subsequent sham surgery-controlled study using bilateral, small volume putaminal (3 delivery trajectories per putamen; 50-μL deposits per trajectory) and nigral (a single delivery trajectory per SNpc; two 15-μL deposits per trajectory) infusions of AAV2-NRTN, the resulting 24-month primary and most secondary endpoints showed no difference between the treatment and the control groups. ¹⁸

Additional midbrain infusions have been used for the treatment of “pediatric parkinsonism”, better known as AADC deficiency (AADC-d). ⁶¹ Such AAV2-AADC gene therapy approaches have successfully targeted bilateral nigral/ventral tegmental area (VTA) dopamine neurons, allowing restoration of intrinsic dopaminergic neuron populations in the midbrain that lack functioning AADC enzyme, and their otherwise intact nigrostriatal, mesolimbic, and mesocortical axonal projections. ⁶² The reported first 7 AADC-d participants treated with AAV2-AADC utilized separate frontal trajectories to each midbrain target, delivering mean Vi of 50 μL per substantia nigra (n = 13) and 30 μL per VTA (n = 14). The Vd/Vi ratios for these infusions were 3.0 in the SNpc and 3.4 in the VTA. Current approaches for treatment of AADC-d now use single bilateral trajectories, infusing Vi of up to 300 μL per trajectory between the two midbrain targets to cover both. ⁶³

Similar midbrain infusions as in AADC-d above are being evaluated for treatment of alcohol use disorder (AUD) but infusing AAV2-GDNF. ⁶⁴ In an animal model of AUD, bilateral MRI-guided CED of AAV2-GDNF/gadoteridol (n = 4) or vehicle/gadoteridol (n = 4) was delivered within the VTA of male NHPs previously conditioned to a chronic AUD state. ⁶⁵ The GDNF gene therapy-treated animals showed a significant mitigation of return-to-drinking behavior over a 12-month period of repeated abstinence-alcohol reintroduction challenges, compared to the vehicle-treated controls.

In summary, the safety and tolerability of surgically delivered gene therapies to the midbrain (SNpc/VTA region) have been confirmed in NHPs and humans for AAV2-NRTN and AAV2-AADC, and most recently AAV2-GDNF in NHPs, with Vi reaching up to 300 μL per trajectory. The efficacy for midbrain gene therapy delivery has not met clinical endpoints using AAV2-NRTN in PD but has shown significant clinical improvements of AADC-d with AAV2-AADC, and appears promising using AAV2-GDNF in an NHP model of AUD (and possibly similar addictive disorders, e.g., opioid abuse, obesity). Whether AAV2-GDNF directly delivered to the SNpc in PD might supplement recovery of the impaired nigrostriatal network, beyond what is possible with putaminal delivery and subsequent axonal transport to the SNpc (retrograde) and SNpr (anterograde), remains to be fully tested.

GDNF tissue distribution and pharmacokinetics

Endogenous GDNF concentrations in normal rat striatum have been determined to range between 0.04–0.05 ng/mg protein. ⁶⁶ Single dose striatal infusions of GDNF protein led to dose-dependent increases in GDNF tissue concentrations. The concentration-time profile showed a steep peak at the earliest measurement (Day 3: 13 and 60 ng/mg protein after infusion of 3 and 15 µg GDNF, respectively), ⁶⁶ followed by a rapid decline to modestly elevated tissue concentrations for 2 weeks (0.14–0.48 ng/mg protein) and 4 weeks (0.04–0.15 ng/mg protein) after infusion doses of 3–15 µg GDNF. ⁶⁶ Another study found matching results at 2 and 4 weeks for the dose range of 3–5 µg GDNF, but failed to show elevated tissue concentrations after lower doses (0.5–1 µg) at these time points, while there was evidence for increased synaptogenesis. ⁶⁷

Rat striatal transduction with escalating single vector doses of AAV2-GDNF per infusion (1.65 × 10¹⁰ vg; 9.07 × 10¹⁰ vg; or 1.65 × 10¹¹ vg) provided dose-dependent increases in GDNF tissue concentrations over approximately 30 days post-infusion, which plateaued at 3, 7.7, and 11.3 ng/mg protein, respectively, for up to 4 months of post-transduction. ⁶⁸

A study of bilateral CED transduction with AAV2-GDNF (9.9 × 10¹¹ vg; n = 8) or PBS (n = 7) in a chronic NHP hemi-parkinsonian model tested single putaminal doses delivered via frontal trajectories (75 µL Vi/putamen). ³⁴ Clinical rating scale (CRS) scores and [¹⁸F]fluoro-L-m-tyrosine (FMT)-PET imaging data were obtained in all animals at baseline and during at least 3 months following chemical induction of parkinsonism. Stable parkinsonian animals were then selected to receive either the test intracranial therapeutic or control infusions. Similar assessments were serially obtained following putaminal infusions. NHP necropsies occurred at 1, 6, 14, 20, and 24 months post-treatment and included brain processing for histology, immunohistochemistry (IHC), high-performace liquid chromatography (HPLC), and GDNF tissue assays. All macaques receiving GDNF gene therapy displayed considerable improvement in their CRS scores between 6–24 months post-treatment compared to controls. Bilateral FMT uptake was observed to increase in the AAV2-GDNF treated NHPs 6 months following infusions (18%-54%, respectively above baseline hemiparkinsonian levels), whereas control animals did not show significant change from baseline hemiparkinsonian levels. Two animals followed up to 22 months showed further increases in FMT uptake (with 37% increase on the moderate-lesioned side and 138% increase on the severly lesioned side). HPLC showed a threefold increase in dopamine following AAV2-GDNF treatment within the moderately lesioned putamina, almost completely restoring it to non-lesioned levels. Dopamine levels in severely lesioned putamina were not significantly improved. Extensive GDNF IHC expression was noted at sites of CED infusion and in association with axonal transport to globus pallidus, subthalamic nucleus and substantia nigra. Quantification of GDNF protein in the putamen and substantia nigra was obtained by ELISA of fresh tissue punches. GDNF levels in the putamen as high as 130 ng/mg total protein (mean, 24 ± 10 ng/mg total protein) were detected. No difference in GDNF expression was found between hemispheres. GDNF levels in the substantia nigra were similar for each hemisphere [left (moderately-lesioned), 1.7 ± 0.8 ng/mg protein; right (severely-lesioned) 1.5 ± 0.4 ng/mg protein]. No GDNF protein expression was detected in any of the PBS-treated controls. An overall increase in tyrosine hydroxylase (TH) IHC intensity was noted in the caudate-putamen of AAV2-GDNF treated NHPs, with large TH-positive histochemical structures noted, that did not differ in number, size or distribution at the 6-, 14-, and 24-month survival timepoints.

Local GDNF tissue concentrations have also been determined following continuous low-rate (0.1 µL/min, 0.1 µg/µL) intraputaminal infusion of GDNF over 7 days in normal rhesus monkeys. ⁴⁹ At the end of infusion, GDNF-treated animals showed a more than 4-fold variability in GDNF Vd and variability over two orders of magnitude in putaminal GDNF concentrations, with approximately 30 to 3000-fold increases when compared to vehicle-treated animals. The range of Vd observed in this study was similar to the limited Vd found in a study testing continuous midbrain infusion in MPTP-lesioned rhesus monkeys. ⁶⁹ Furthermore, the distribution pattern of GDNF was irregular. While the highest levels of GDNF were within a 2-mm radius around the catheter tip, sites with high concentrations of GDNF could be found directly adjacent to sites containing an order of magnitude less material. In addition, GDNF concentrations decreased logarithmically with increasing distance from the catheter tip, supporting Fick-like diffusion of GDNF. ⁴⁹ It is likely that these findings were the joint result of (1) diffusion-based fluid dynamics and drug distribution resulting from low-rate infusion, and (2) GDNF binding to heparan sulfate proteoglycan in the extracellular matrix.^37,49

Improved tissue distribution of GDNF during low-rate infusion (0.2 µL/min) was found when co-infusing heparin to prevent binding of GDNF to heparan sulfate in the extracellular matrix, ³⁷ however, the increased risk of intracerebral bleeding associated with this approach clearly outweighs its potential benefits.

More recently, the distribution characteristics of GDNF protein were assessed after single intraputaminal administration via MRI-guided CED, with infusion rates up to 1.5 µL/min, in Cynomolgus monkeys. ⁷⁰ Infusions were continued until real-time MRI eventually showed reflux along the outside of the infusion cannula. The study revealed a linear relationship between Vd and Vi, and the final Vd was significantly greater than that seen in the continuous low-rate infusion study reviewed above, ⁴⁹ which is likely to be a direct result of convection-based fluid dynamics overwhelming the local binding sites and allowing extended distribution of GDNF.^38,46

In summary, the tissue distribution and pharmacokinetic data available for both GDNF protein and AAV2-GDNF suggest that consistent meaningful putaminal target coverage can only be achieved by means of CED. Although the interpretability of the available data is limited, since GDNF tissue concentrations can only be measured post-mortem and are not currently attainable in living human subjects, it seems safe to accept the above conclusion, given the larger dimensions of the human putamen and brain providing a greater challenge to achieving adequate tissue distribution and target coverage.

The requisite high infusion rates necessitate that treatment be administered as single (AAV2-GDNF) or intermittent (GDNF protein) infusions so as to avoid “flooding” of the brain that would otherwise occur with continuous delivery. Intermittent drug delivery, in turn, is necessarily associated with substantial fluctuations in GDNF tissue concentrations. The potential influence of these fluctuations on the expected drug effect remains to be determined. Obviously, single-use AAV2-GDNF gene therapy delivery to the putamen provides a more stable and sustained local GDNF protein source and distribution to the putaminally-directed nigral dopaminergic axonal arborizations than intermittent delivery of GDNF protein. Such stable levels of intrinsically produced putaminal GDNF protein are anticipated to provide beneficial trophic support to prevalent “sick-but-not-dead” presynaptic dopaminergic terminals and axons associated with PD. In addition, anterograde (striatonigral) transport of putaminally distributed AAV2-GDNF vector provides support to nigral dopaminergic cell bodies, through GDNF transduction of SNpr neurons and subsequent secretion to influence adjacent SNpc cell bodies and dendrites.

Potential side effects/safety

Supporting the local delivery paradigm, potentially drug-related side effects have predominantly been observed in studies that consistently exposed, intentionally or inadvertently, the entire cerebrospinal fluid (CSF) space and/or larger parts of the brain to GDNF over an extended time period. In particular, ICV administration of GDNF over 6 months was commonly associated with gastrointestinal side effects (nausea, anorexia, and vomiting), weight loss, hyponatremia and sensorineural symptoms (paresthesias, Lhermitte sign) in PD subjects. ³⁵ In contrast, intraputaminal delivery was generally found to be safe. Amongst a total of 90 PD subjects receiving continuous (n = 49) or intermittent (n = 41) intraputaminal infusions over up to 43 months in 4 independent studies, the most commonly reported, potentially drug-related adverse events were transient paresthesias and Lhermitte's sign, and none of them considered serious.^{8–11,14–16} In no case was the study medication discontinued early due to a drug-related adverse event, and in the two intermittent intraputaminal delivery studies, compliance was high with 748 (98.4%) of 760 scheduled infusions being successfully delivered^14,15

In a 6-month toxicity study of continuous intraputaminal dosing of GDNF, multifocal cerebellar Purkinje cell loss was observed in 4 of 15 animals treated at the highest dose level (100 µg/day). ⁷¹ It was hypothesized that ongoing leakage of GDNF into the CSF during the infusion led to down-regulation of GDNF receptors on Purkinje cells as a result of persistent exposure to excessive GDNF concentrations. ⁷² As this was followed by abrupt withdrawal of drug, cells could become atrophic and ultimately die via a caspase-dependent non-mitochondrial pathway.^72,73 In contrast, no adverse findings were made in cerebellum or other parts of the brain in a subsequent 9-month toxicity study of intermittent intraputaminal dosing of GDNF. ⁷⁴ The stated hypothesis is also consistent with the complete absence of Purkinje cell findings in intermittent ICV dosing toxicity studies administering excessive doses of up to 10,000 µg/month. ⁷² Importantly, no cerebellar lesions have ever been found in humans treated with GDNF, despite a systematic post-study MRI volumetric and intensity analysis of the cerebellum including 9 of 10 participants in one of the two early, open-label studies testing continuous intraputaminal infusion of GDNF, ⁷⁵ and a systematic MRI follow-up in the 41 subjects receiving GDNF via intermittent intraputaminal CED.^14,15 Similarly, no indication of cerebellar toxicity was observed during serial postoperative (up to 5 years after infusion) MRI in the 13 subjects with advanced PD receiving bilateral co-infusions (52 infusions in total) of AAV2-GDNF and gadoteridol via intraputaminal one-time CED (Vi: 450 µL per putamen). ¹³

Formation of anti-GDNF binding antibodies was observed in approximately half of the participants in the continuous intraputaminal delivery program of GDNF, including 5 subjects with neutralizing antibodies. ⁷⁶ It has been hypothesized that the root cause for these surprisingly high rates was not the intracerebral delivery, but systemic exposure to GDNF during the periodic invasive refill process of the implanted infusion pumps in the abdominal wall.^71,72 This hypothesis is supported by the fact that no anti-GDNF binding serum antibodies were identified at any time point during the intermittent intraputaminal delivery studies which used a transcutaneous port for repeat access.^14,15 Increased anti-GDNF antibody titers in serum and/or CSF were again observed in approximately half of the participants in the Phase 1 study of AAV2-GDNF in patients with advanced PD. ¹² No antibody findings were reported in the subsequent Phase 1b study in patients with mild or moderate PD. ⁴⁸ The clinical implications of anti-GDNF antibody formation, if any, are currently unclear as all positive subjects remained asymptomatic.

AAV2-GDNF gene therapy

As previously detailed by Barker et al., ³ human gene therapy trials for PD testing GFL members (apart from GDNF itself, only NRTN to date) have solely used non-pathogenic AAV2 gene transfer vectors driven by a constitutive cytomegalovirus (CMV) promoter. Such AAV2 vectors appear to provide a favorable safety profile with direct brain delivery of neurotrophic factor gene therapy,^{12,13,17–19,48} at least in part related to AAV2 having an exclusive neuronal tropism and, therefore, restricted distribution when directly delivered within brain parenchyma. ⁷⁷

Delivery of intraputaminal AAV2-GDNF gene therapy requires an understanding of the same issues described above for the distribution of GDNF protein within the target putamen. Binding of the AAV2 vector to the primary neuronal population, MSNs, rather than glia, ⁷⁷ allows the GDNF transgene to enter the neurons and initiate the local production and release of the GDNF protein into the extracellular environment. Remaining terminal dopaminergic arborizations of the nigrostriatal pathway have synaptic contacts with MSNs and would be optimally positioned to benefit from the newly released neurotrophic factors. ⁷⁸ In these cases, GDNF would be taken up in functioning presynaptic dopaminergic terminals, supporting various axon terminal functions, and be retrogradely transported back to the nigral somata to provide additional neurotrophic effects.

AAV2-GDNF putaminal distribution using CED with iMRI easily allows optimal therapeutic distribution and coverage of a significant putaminal volume (Figure 3). Such optimized target coverage, therefore, allows widespread MSN transduction and significant local production of GDNF protein. Availability in the putamen of trophic factor at sufficient levels is able to positively influence significant numbers of the remaining viable or “sick-but-not-dead” nigrostriatal dopaminergic terminals and their host neuron cell bodies in the SNpc, via retrograde transport during early disease stages. An additional benefit of using the AAV2-GDNF vector is that it is also anterogradely transported from the putamen to the substantia nigra, via the mostly monosynaptic direct and multisynaptic indirect striatonigral pathways, eventually transducing a substantial number of SNpr neurons to produce and release GDNF protein, which then binds to and positively influences the remaining viable dopaminergic neuron cell bodies within the adjacent SNpc.^30,34

Of note, based on natural AAV vector serotype versus recombinant vector production differences, AAV2 has shown a natural capacity to preferentially transport in an anterograde manner, ⁷⁹ while novel AAV2 variants (AAV2 retro) have been shown to efficiently transport in a retrograde manner.^80,81 For intraputaminal GDNF gene therapy, it is of critical importance, therefore, to select AAV2 + transgene production methods yielding specific AAV2-GDNF particles that efficiently utilize anterograde transport from the putamen (to the SNpr) but not retrograde transport (to the cortex). The criticality of this AAV2 selection becomes evident when looking at autopsy findings from 4 prior participants in either of the first 2 NRTN gene therapy studies, which utilized an AAV2 vector with unknown retro properties.^17,19 The immunohistochemical evaluation of the brain tissues showed no detectable NRTN in the SNpr, whereas numerous NRTN-immunoreactive neurons were seen in the cerebral cortex. ⁸² By contrast, there is convincing evidence from hemiparkinsonian models of both rats ⁸³ and NHPs, ³⁴ that intraputaminally delivered AAV2-GDNF distributes via direct and indirect striatonigral connections using anterograde transport to the SNpr, and also transducing intervening nuclei of the globus pallidus, entopeduncular nucleus, and subthalamic nucleus. As a result of the above, previous and ongoing AAV2-GDNF clinical trials utilize preclinical rodent and NHP studies to confirm anterograde transport of clinical AAV2 vectors prior to clinical use in PD (Bankiewicz KS, personal communication).

In addition to the known plethora of AAV serotypes, ⁸⁴ the evolution of AAV engineering has yielded numerous capsids tailored for cell-type specificity, target cell infectivity, and/or transport properties. ⁸⁵ However, most of these capsids are at an early preclinical stage, and discussing their potential role relative to the carefully selected, purpose-built, and robustly tested AAV2 vectors that are in current clinical development is beyond the scope of this article. The future value of such variants as next-generation vectors will depend especially on the outcome of the ongoing clinical program of AAV2-GDNF in PD.

Initial impressions from using intraputaminal AAV2-GDNF gene therapy in the Phase 1 study, including 13 subjects with advanced PD, have provided evidence of safety and tolerability, with no drug-related serious adverse events or brain MRI abnormalities noted up to 5 years post-dosing.^12,13 This early GDNF gene therapy investigation supports the concept of time-dependent target engagement assessment via ¹⁸F-DOPA PET scanning. ¹² Specifically, ¹⁸F-DOPA Ki parametric brain maps showed bilateral putaminal Ki increases from baseline to 6 months post-GDNF gene therapy in 10 of 13 study participants (median: 36%). At the 18-month post-surgery timepoint, bilateral putaminal Ki increases were seen in 12 of 13 participants (median: 54%). Importantly, there was excellent overlap noted when comparing regions of intraoperative gadoteridol distribution (co-infused with AAV2-GDNF), and the regions of increased ¹⁸F-DOPA Ki values. The majority of participants in this study maintained (low dose group) or numerically improved (medium dose group) their clinical levels of motor function over a 5-year period post-treatment, which is particularly remarkable given that the mean putaminal coverage was only 26%. Very recently, an autopsy evaluation of putaminal gene therapy sites in both hemispheres from a study participant that died 3.5 years following AAV2-GDNF gene therapy from an unrelated cause provided valuable insight at the cellular level. ⁸⁶ While the treatment sites displayed no histological evidence of tissue toxicity or untoward reactions, they showed continuing production of GDNF protein, increase in local dopaminergic metabolites, and robust TH-immunoreactive dopaminergic fiber sprouting. ⁸⁶ All of these findings were consistent with what had previously been documented in similarly treated NHPs.^30,34 As they contrast markedly with the expected degeneration of the dopaminergic nigrostriatal projections in PD, ²⁸ they are thought to be reflective of GDNF's regenerative potential.

The subsequent Phase 1b study, using a 40-fold higher vector dose than the Phase 1 study, included 11 participants with mild (n = 6) or moderate (n = 5) PD, all of whom have completed at least 18 months of follow-up. ⁴⁸ None of the 6 serious adverse events observed in the study and only one nonserious event of perioperative headache was considered possibly drug-related. Three participants presented with small unilateral T1 hypointensities at 6 and 18 months post-treatment which remained asymptomatic and were considered related to the surgical procedure. At the 18-month primary follow up, participants with mild PD were found to be numerically stable in terms of their Movement Disorder Society-Unified Parkinson's Disease Rating Scale (MDS-UPDRS) Part II and III scores and their self-reported motor diary outcomes (excluding a subject with a heterozygous TH mutation). Participants with moderate PD showed mean improvements of 20.4 ± 4.5 and 10.6 ± 3.6 points in the MDS-UPDRS Part III OFF and ON scores, respectively, paralleled by changes of −1.7 ± 1.1 and 2.2 ± 1.0 h/day for OFF and Good ON time, respectively, in their motor diaries. At the same time, their levodopa equivalent daily dose was reduced by −257.6 ± 162.2 mg. When compared with baseline, dopamine transporter (DaT) binding at 18 months, as assessed by the ratio of DaT specific binding in the striatum to background binding in a control area in the brain, was stabilized within the (treated) putamina, but showed progressive decline in the (untreated) caudate nuclei in both PD cohorts.

What are the delivery-based conclusions for the way forward?

Based on the aggregate experience published to date, the following key conclusions can be drawn in regard to localized brain delivery of GDNF:

• ICV delivery of GDNF: obsolete—no/insufficient distribution to target structures; unacceptable side effect profile; lack of efficacy.

• Continuous intraputaminal infusion of GDNF: obsolete—insufficient distribution across and coverage of target structure; acceptable side effect profile, but high rates of antibody formation and small potential for development of cerebellar lesions; no/minimal efficacy.

• Intermittent intraputaminal CED of GDNF: remaining viable mode of administration for GDNF protein—sufficient distribution across and coverage of target structure; acceptable side effect profile, no antibody formation and no cerebellar risk; limited (retrograde) transport from the putamen to the SNpc; limited efficacy relative to control may be enhanced by (1) further increasing Vi/coverage, (2) increasing the GDNF concentration in the infusate, (3) reducing the length of the dosing interval (from monthly); (4) increasing the double-blind treatment period. Downside (not currently discussed in this review): need for (improved/simplified) chronically implanted delivery system.

• Single-dose intraputaminal CED of AAV2-GDNF: undoubtedly the most promising treatment strategy—adequate distribution across and coverage of target structure; delivery technology available; excellent side effect profile; limited (retrograde) transport of locally expressed protein from the putamen to the SNpc via nigrostriatal projections, plus available striatonigral (anterograde) transport of the AAV2-GDNF vector from the putamen to the SNpr, thereby releasing GDNF to SNpc; promising efficacy (only open-label data available so far). Downside: current need for iMRI monitoring.

Key questions around the emerging options, intermittent intraputaminal CED of GDNF and single dose intraputaminal CED of AAV2-GDNF, are summarized in the following concluding sections.

What are the key questions with intermittent intraputaminal delivery of GDNF?

As stated above, dosing with intermittent CED is determined by Vi/coverage, GDNF concentration in the infusate, and length of the dosing interval. For obvious reasons, tissue pharmacokinetics cannot be determined in humans. ¹⁸F-DOPA uptake is considered a relevant pharmacodynamic measure, which may indicate terminal sprouting, reawakening of hibernating terminals, an upregulation of AADC, or a combination of all three, ⁸⁷ demonstrating target engagement by induction or modification of the dopaminergic innervation of the putamen. Whether and how the degree and timing of changes in ¹⁸F-DOPA uptake correlate with clinical benefit remains to be established. Measurement of clinical benefit is, however, very challenging in view of the previously observed, large placebo response and the extended trial duration needed to establish meaningful clinical effects. Indeed, participants in the 2 intermittent intraputaminal delivery studies were found to develop increasingly prevalent dyskinesia in the practically-defined OFF state over time, suggesting that treatment periods of greater than 40 weeks may be required for certain functional effects of GDNF to develop. ⁸⁸

As the intermittent intraputaminal delivery studies have not yielded any safety concerns, and the total dose given per 4 weeks in these studies was substantially smaller than in prior continuous dosing intraputaminal delivery studies (240 µg vs. approximately 806–2419 µg),^{8–11,14–16} it is suggested that raising the dose close to the lower end of the range tested with continuous delivery would be justified. Due to practical restrictions on catheter number, infusion frequency, and infusion volume, this could most easily be achieved via increasing the GDNF concentration in the infusate (from 0.2 to 0.6 µg/µL, thereby increasing the total dose per 4 weeks from 240 to 720 µg). Justification for such an increase is also provided by the absence of local or remote lesions in a 9-month toxicity study that tested the intermittent intraputaminal delivery of GDNF at 0.67 µg/µL in rhesus monkeys. ⁷⁴ However, as significant perivascular leakage has been observed in putaminal infusions, stationary cannula tip infusions of GDNF may be challenging.^54,55

What do we know about intraputaminal infusion volume and vector dose with CED?

Table 2 summarizes the available infusion volume and vector dosing information from the two open-label studies testing AAV2-GDNF in subjects with PD.^12,13,48

Table 2.

AAV2-GDNF dosing in clinical trials.

Study/Dosing	Vector titers	Maximum Vi/Putamen	Maximum dose/Putamen	Total patient dose	Mean volumetric putaminal coverage
AAV2-GDNF Phase 112,13	9.9 × 1010 vg/mL a	0.450 mL	4.5 × 1010 vg	9.0 × 1010 vg	26%
	3.3 × 1011 vg/mL a	0.450 mL	1.5 × 1011 vg	3.0 × 1011 vg
	9.9 × 1011 vg/mL a	0.450 mL	4.5 × 1011 vg	9.0 × 1011 vg
AAV2-GDNF Phase 1b 48	3.3 × 1012 vg/mL a	1.800 mL	5.94 × 1012 vg	11.9 × 1012 vg	63%

AAV2 = adeno-associated viral vector serotype 2; GDNF = glial cell line-derived neurotrophic factor; Vi = infusion volume.

^a Originally approved titers for use in Phase 1 dose-escalation trial.

The volumetric putaminal coverage at the end of infusion observed in the Phase 1b study exceeded the protocol-defined threshold of 50% in 21 of the 22 putamina infused, providing robust evidence for the reliability and replicability of the iMRI-guided “infuse-as-you-go” CED technique employed. The associated clinical results through 18 months, especially within the moderate stage PD cohort, are promising and serve as the basis for a Phase 2 study of approximately 87 subjects with moderate stage PD, which uses the same delivery technology and vector dose as the Phase 1b and commenced recruitment in the US in 2024 and in the UK and EU in 2025 (REGENERATE-PD; NCT06285643; EUCT2023-506519-16-00). We believe that this and future studies will shed more light on several important aspects of potential GDNF treatment in PD such as: a) what is the relationship between putaminal coverage and PET imaging and/or clinical outcome; b) is there a clinically meaningful threshold value for coverage; c) should treatment with AAV2-GDNF target a specific anatomical aspect of the putamen to provide for the best possible clinical outcome, as opposed to maximizing total putaminal coverage; and d) how significant is the use of iMRI monitoring of putaminal CED for providing safe and efficacious AAV2-GDNF gene therapy to patients.

There remains challenging work ahead in trying to confirm the benefits (or lack thereof) of advanced therapies, such as AAV2-GDNF gene therapy, for the treatment of PD and other conditions, to ourselves, the medical community, patients, their families, and society at large. We remain encouraged and accept this challenge, especially on behalf of our patients and the PD community.