Randomized trial of intermittent intraputamenal glial cell line-derived neurotrophic factor in Parkinson’s disease

Introduction

Parkinson’s disease is the second most common neurodegenerative disorder (de Lau and Breteler, 2006). Current treatment options are symptomatic and do not prevent disease progression. Over time, patients accrue both motor and cognitive disability and develop complications of dopaminergic therapies (Rascol et al., 2011).

Glial cell line-derived neurotrophic factor (GDNF) is a potent neurotrophic protein for dopaminergic and other neurones (Lin et al., 1993; Airaksinen and Saarma, 2002) and shows robust neurorestorative and neuroprotective effects in nonhuman primate models of Parkinson’s disease when administered to targets within the CNS (Gash et al., 1996; Zhang et al., 1997; Grondin et al., 2002; Allen et al., 2013). In patients with Parkinson’s disease, intracerebroventricular administration of GDNF did not show clinical benefit (Kordower et al., 1999; Nutt et al., 2003). In two open-label studies using continuous intraputamenal infusion, GDNF substantially improved motor function at 6 and 12 months (Gill et al., 2003; Slevin et al., 2005), associated with a focal increase in ¹⁸F-DOPA uptake at the site of infusion in posterior putamen (Gill et al., 2003). Moreover, single-case reports suggested dopaminergic sprouting (Love et al., 2005) and clinical benefit years beyond end of treatment (Patel et al., 2013). The favourable clinical outcome, however, could not be replicated in a randomized, placebo-controlled study using a similar infusion scheme over 6 months (Lang et al., 2006).

The observed outcome discrepancies were possibly due to insufficient GDNF exposure across the putamen, since continuous low-rate infusions enable only diffusion-dependent, irregular (heterogeneous), spatially restricted distribution (Salvatore et al., 2006). Much wider, homogeneous distribution can be achieved with convection-enhanced delivery (CED) which, however, requires high infusion rates that in turn necessitate intermittent rather than continuous administration to avoid tissue ‘flooding’ (Gimenez et al., 2011). The above, together with recent reports on striatal pharmacokinetics and pharmacodynamics of GDNF (Hadaczek et al., 2010; Taylor et al., 2013), was the rationale for conducting a randomized, placebo-controlled, study of GDNF, administered on an intermittent (every 4 weeks) basis, in a manner to achieve CED across the putamen. The present study was the first clinical study worldwide to evaluate the effects of GDNF (or any other drug) when given via intermittent (every 4 weeks), bilateral, intraputamenal CED.

Dosing in the context of intermittent CED is complex and involves several dimensions, including infusion rate, infusion volume, drug concentration, dosing interval and total (or cumulative) dose over time. The infusion rate and volume were chosen to achieve intraputamenal CED. The intermittent dosing information available at the beginning of the study was mostly limited to reports on striatal pharmacokinetics and pharmacodynamics of GDNF in rats (Hadaczek et al., 2010; Taylor et al., 2013). The infusate GDNF concentration administered was 2-fold higher than in the Lang investigation (0.2 µg/µl versus 0.1 µg/µl) (Lang et al., 2006); however, moving from continuous to every 4-week intermittent dosing means the total dose delivered over 4 weeks (240 µg) is 3.5-fold smaller than the dose most widely used in the historic continuous dosing studies (840 µg). Previously, unexpected clinically silent cerebellar toxicity was observed when very high GDNF doses (2800 µg/4 weeks, translating to a human equivalent dose of 42 000 µg/4 weeks) were given via continuous intraputamenal infusion in a 6-month toxicity study in rhesus monkeys (Hovland et al., 2007). Therefore, to ensure patient safety, the GDNF dosing scheme used in the present study was purposefully low. The associated risk of underdosing was acknowledged but deemed acceptable. Additional dose groups were not feasible financially and logistically in this single-centre investigator-initiated study.

Because of a lack of commercially-available drug delivery devices that would facilitate intermittent infusions to the brain parenchyma over an extended time period, a novel in-house system was specified by the lead neurosurgeon (Lewis et al., 2016). Most importantly, the system included a skull-mounted transcutaneous port allowing for non-invasive repeat infusions via four separate microcatheters. Although a first-in-man device in its entirety, key device performance features and device attributes such as long-term catheter patency, controlled infusions and device safety had been developed previously (Bienemann et al., 2012; Barua et al., 2013, 2016; Gill et al., 2013).

Altogether, the study entered uncharted territory in several areas. The study results, while of primary relevance to GDNF and Parkinson’s disease, were considered potentially useful for other applications and indications where direct, targeted drug delivery to brain parenchyma could be beneficial. The primary hypothesis being tested was that GDNF, if administered in a manner to permit CED across the putamen, would achieve neurorestoration leading to clinically significant benefit.

Materials and methods

Results

Discussion

Reversing neurodegenerative disease remains one of the greatest medical challenges. Neurotrophic factors are among the most promising candidate therapies, with demonstrated neuroprotective and neurorestorative capabilities in animal models of Parkinson’s disease (Allen et al., 2013). One major difficulty in translating these findings into the clinic has been to find a way of delivering the agents across the blood–brain barrier to meaningful volumes within relevant brain targets and potentially over the remaining lifetime of the patient. Here we show for the first time that this challenge has been met with the development of a long-term implantable drug delivery system that facilitates intermittent intraparenchymal infusions for an extended time period.

However, in this double-blind trial of 41 randomized patients with moderate stage Parkinson’s disease, 40 weeks of fixed-dose GDNF intraputamenal infusions (120 µg GDNF in 600 µl artificial CSF to each putamen), administered every 4 weeks, did not produce a significantly larger percentage improvement than placebo in OFF state UPDRS motor score, the primary study outcome. In addition, no significant improvements over placebo were observed in any of the secondary or exploratory motor endpoints, nor in any of the non-motor or quality of life endpoints. In contrast to the main clinical findings, serial PET imaging revealed a significant increase in ¹⁸F-DOPA uptake in the GDNF group but not the placebo group, which was spatially extensive and suggested whole putamen target tissue delivery.

The essence of our clinical findings is consistent with an earlier randomized, placebo-controlled trial of GDNF in Parkinson’s disease (Lang et al., 2006), although the latter used continuous diffusion-dependent intraputamenal delivery resulting in spatially limited putamen delivery (Salvatore et al., 2006). In aggregate, therefore, these two studies raise the question as to whether the underlying growth factor hypothesis is flawed, or whether the hypothesis per se is correct but the clinical testing in both studies has been flawed and further evaluation is required, informed by this investigation and its follow-on open-label extension.

The regenerative effects of GDNF in Parkinson’s disease have been questioned by studies using a rat model with overexpression of human wild-type α-synuclein. In this model, targeted delivery of GDNF to the striatum failed to prevent loss of nigral dopamine neurones or their terminals in the striatum (Decressac et al., 2011). A subsequent investigation suggested that Nurr1, a regulator of neurotrophic factor signalling, as well as its downstream target, GDNF receptor component RET, are decreased in the presence of excess α-synuclein (Decressac et al., 2012). In other words, the intracellular signalling response to GDNF may be blocked in the presence of excess α-synuclein. A recent study, however, found that SNCA (α-synuclein) mRNA is not increased in sporadic Parkinson’s disease, and α-synuclein accumulation does not block GDNF signalling in either Parkinson’s disease or Parkinson’s disease models (Su et al., 2017). Consistent with these findings, the integrity of the GDNF signalling cascade in our patients is supported by the significant putamenal increase in ¹⁸F-DOPA uptake at Week 40.

Independent of the α-synuclein question, it is conceivable that mostly toxin-based, static preclinical models may not be representative of the progressive human disease. This is contradicted, however, by findings in a single patient in an early open-label phase I study who was treated with GDNF for 43 months via a unilateral continuous intraputamenal infusion and died from myocardial infarction after the study (Love et al., 2005). The patient had unilateral disease on the left and was therefore treated on the right, leaving the contralateral side as an intra-patient control. At 24 months, the patient showed appreciable clinical benefit on the left coupled with improved ¹⁸F-DOPA uptake in the right posterior putamen. This was contrasted by a small decrease in ¹⁸F-DOPA uptake on the left and new-onset motor symptoms on the right in the latter part of the treatment. Post-mortem findings showed dopaminergic sprouting in the posterior third of the right putamen and greater expression of tyrosine hydroxylase (TH) on the right than on the left. Consistent with the clinical laterality, the net loss of dopaminergic neurons in the substantia nigra was greater on the right than on the left. At the same time, presumably in response to treatment with GDNF, the nigral expression of GAP43 and synaptophysin was also greater on the right.

If, therefore, the growth factor hypothesis is still valid, the question about the limitations of the clinical studies becomes pertinent. These include: Can this potential treatment only be effective in early stage disease where innervation of the striatum with dopaminergic neurons is maintained above a critical threshold? Was the dose of GDNF selected for this study sufficient? Was the treatment duration long enough? Could patient phenotype, placebo effects or methods of assessment have played a part?

Ethical considerations limited recruitment in the present study to patients with moderate stage Parkinson’s disease. The failure to demonstrate clinical benefit at 40 weeks may be the consequence of recruiting patients too late in their disease. A recent post-mortem study showed that within 4 years post-diagnosis, patients with Parkinson’s disease had an almost complete loss of TH-positive terminals and dopamine transporter immunohistochemistry in the dorsal putamen (Kordower et al., 2013). However, these findings may reflect neuron hibernation and not just neuron death alone. Potentially in favour of hibernation, GDNF-treated patients in the present study, where the mean time from diagnosis was ∼8 years, showed a significant increase in putamenal ¹⁸F-DOPA uptake throughout the putamen with no or little differences in absolute improvement between the anterior, mid and posterior regions of interest. If the findings of the Kordower study were reflective of neurone death alone, the increase in the posterior third of the putamen, the locus of greatest reduction in TH in Parkinson’s disease, would perhaps have been expected to be smaller than in the anterior putamen. The hibernation hypothesis is also supported by the post-mortem improvements in the posterior putamen observed in the above phase I patient whose treatment commenced 5 years after diagnosis (Love et al., 2005).

Ethical considerations also limited the duration of the present study to 9 months to avoid an undue length of exposure to placebo in a surgical setting. However, 9 months may have been too short a period of repeated putamenal tissue GDNF exposure to achieve adequate functionality of reconnecting neurons. It is possible that there is a lag between a biomarker effect such as ¹⁸F-DOPA uptake and clinical improvement. Further insight into this question may, in part, be provided by the open-label extension study.

Other potential limitations are associated with drug delivery and dose. Drug distribution in the previous, continuous low-rate infusion clinical studies was diffusion-dependent, heterogeneous, and spatially restricted to less than 10% total putamenal coverage (Salvatore et al., 2006). This could explain the observed lack of clinical benefit in the earlier phase II double blind study (Lang et al., 2006). The intermittent CED dosing scheme used in the present study led to much wider distribution, supported by the gadolinium-evidenced coverage of putamenal volume of interest (67.1–78.5%) and total putamen (47.8–55.0%) over 40 weeks as well as the putamen-wide increase in ¹⁸F-DOPA uptake in GDNF-treated patients. In contrast, as the dosing information available at the beginning of the study was sparse, a conservative approach was taken to dose, so as to avoid any unnecessary safety risk in view of the cerebellar lesions that were unexpectedly observed in several rhesus monkeys treated with very high GDNF doses (Hovland et al., 2007). Due to the switch from continuous to intermittent dosing, the cumulative 4-week dose per putamen in the present study was 3.5-fold smaller than in the historic continuous dosing studies (120 µg versus 420 µg), albeit the infusate GDNF concentration was twofold higher (0.2 µg/µl versus 0.1 µg/µl).

Considering the combination of 5-fold larger putamenal coverage and 3.5-fold smaller cumulative 4-week doses, the resulting tissue GDNF concentrations in the exposed volumes were ∼18-fold lower in the present study. As retrograde transportation of GDNF from the putamen to the substantia nigra is known to be concentration-dependent (Aoi et al., 2000), it is therefore possible that whilst the tissue GDNF concentrations were sufficient to induce a PET-evidenced biological effect, they were too low to produce clinical benefit within 40 weeks. This would be supported by the fact that a post hoc analysis of nigral ¹⁸F-DOPA uptake did not show any differences between treatment groups (data not shown), whereas a noticeable increase was seen in the only continuous-dosing study that assessed nigral ¹⁸F-DOPA uptake (Gill et al., 2003). Clinically, the present study found numerical mean differences in favour of GDNF in all OFF state motor and Parkinson’s disease diary-based endpoints at all time points, and a post hoc responder analysis showed that significantly more patients on GDNF than on placebo had a moderate-to-large clinically important change in OFF state motor score (Shulman et al., 2010). While in the absence of statistically significant results for the primary and secondary endpoints, inferences are of course speculative, one potential explanation for these findings is that underdosing played a part.

It is also worth noting that the key preclinical studies that were used to derive the clinical dose demonstrated that both the pharmacokinetics and the pharmacodynamics of intrastriatally infused GDNF were dependent on infusate GDNF concentrations (Hadaczek et al., 2010; Taylor et al., 2013). In particular, the Taylor study showed that while a significant increase in striatal synaptogenesis (as determined by synaptophysin concentration) was observed at an infusate GDNF concentration of 0.2 µg/µl, maximal axonal sprouting only occurred at 0.6 µg/µl (Taylor et al., 2013). However, the latter concentration is very close to the concentration used in the rhesus monkeys that developed cerebellar lesions (0.67 µg/µl), and was therefore considered too high at the time (Hovland et al., 2007). Meanwhile, a further 9-month toxicity study testing the intermittent dosing paradigm in rhesus monkeys has established 0.67 µg/µl as the no-observed-adverse-effect level (Luz et al., 2018), thus opening the door to include this threefold higher dose level in future clinical studies.

A potential confounder of the present study was the magnitude of the placebo effect. While similar to that seen in other surgical studies in Parkinson’s disease (Olanow, 2005; Goetz et al., 2008; Marks et al., 2010), it was notably larger than in the earlier phase II double blind study that had been used as the main point of reference when estimating the sample size (Lang et al., 2006). Conceivably, the 4-weekly visit schedule, in concert with dedicated clinical care at a single site committed to maximize patient retention, the prospect of receiving active drug at the end of the study, and investigator bias may have contributed to the observed placebo response. However, the earlier phase II double blind study also included an optional open-label extension, and other structural differences between the studies are relatively subtle considering the magnitude and consistency of the placebo response across different clinical endpoints. Therefore, it is worth considering that putamenal tissue disruption as a result of catheter implantation and repeated high-pressure CED infusions may have played a part. It is known that striatal injury, primarily via a GDNF- and BDNF-dependent mechanism mediated by activated macrophages and microglia, leads to strong and potentially persistent stimulation of the nigrostriatal dopamine system (Batchelor et al., 1999, 2000; Liberatore et al., 1999). Associated with the amount of trauma that was appreciably larger in the present study than in the earlier phase II double blind study (two catheters per putamen versus one, longitudinal versus vertical putamenal trajectories, and continuous low-rate, low-pressure infusions versus intermittent high-rate, high-pressure infusions), these self-repair effects may have been more pronounced in the present study. Although they would presumably have been associated with an increase in ¹⁸F-DOPA uptake that was not noted in the placebo group, nevertheless it seems possible that an intermittent placebo CED arm produces effects beyond those of a traditional surgical placebo arm.

Two further limitations of the study include the residual effect of symptomatic medication and the sample size. Although we used the commonly agreed definition of the practically defined OFF state (all OFF assessments were performed at a similar time in the morning, following withholding of long-acting anti-parkinson medications the day before and all other anti-parkinson medications from 6 pm the evening before), we recognize that the true ‘wash out’ period for all dopaminergic drugs including long acting agonists exceeds the standard withdrawal period allowed by this paradigm. The sample size was small despite the study being adequately powered to detect significant change in the primary outcome, and we recognize this as a further limitation to interpreting the study results.

The nine 10-point responders in the GDNF group are a potential focus of interest; however, as this a post hoc finding we would not wish to over-interpret its meaning. As shown in Fig. 2B, absolute changes in UPDRS motor score demonstrated a fairly even spread across subjects in the active arm from minimal worsening in score to greater than 10-point improvements. In other words, a bimodal distribution of absolute responders versus absolute non-responders was not seen. Furthermore, post hoc covariate analyses investigating phenotypic characteristics such as age, disease duration, disease severity, tremor predominance etc. did not identify a subtype of patients predicting an enhanced benefit. The nine 10-point responders did not differ in surgical approach, drug delivery or maintenance of their delivery systems, infusate volumes of distribution on MRI scans or magnitude of ¹⁸F-DOPA PET response.

At this point, therefore, we are not able to identify a priori a particular subgroup of Parkinson’s patients that are either more or less likely to respond to GDNF therapy. That said, it may be that some patients require longer duration of GDNF exposure to experience clinical benefit than others, therefore creating the appearance of responders versus non-responders with a 40-week infusion programme.

In contrast to our clinical outcomes, serial ¹⁸F-DOPA PET imaging revealed a significant increase in radioligand uptake in the GDNF group. This increase was meaningful in effect size, present throughout the putamen and notably higher than in prior Parkinson’s disease trophic factor studies (Lang et al., 2006; Marks et al., 2008, 2010). The earlier phase II double blind study used continuous infusions via abdominal pumps with subsequent diffusion-dependent putamenal distribution and therefore, showed increased PET signal uptake predominantly around the catheter tips (Lang et al., 2006). With the intermittent CED approach, however, significant change was seen throughout the putamen, with a percentage gradient of increased improvement from posterior to anterior putamen, but not affecting the non-infused caudate, in keeping with a true biological effect of the treatment.

The increase in ¹⁸F-DOPA signal could indicate terminal sprouting, reawakening of hibernating terminals, an upregulation of aromatic amino acid decarboxylase, or a combination of all three (Moore et al., 2003). It is important to appreciate that ¹⁸F-DOPA is trapped by all monoaminergic neuron types, and hence, the contribution from nigral dopaminergic versus raphe serotonergic terminal change cannot be differentiated (Moore et al., 2003). Moreover, it cannot be confirmed that such change results in restoration of dopamine neurone function. For this, any potential future investigation would need to include additional imaging outcomes, such as using dopamine neuron-specific radiotracers and raclopride displacement assessments (Piccini et al., 1999; Kaasinen and Vahlberg, 2017).

Comparing with prior investigations, it is worth noting that the biological effect was sufficient to move absolute ¹⁸F-DOPA PET Kocc uptake values from those typically associated with moderate to advanced disease to those seen in mild Parkinson’s disease (Sossi et al., 2003). This contrasts with the prior clinically negative viral vector neurturin studies in Parkinson’s disease, where no PET signal benefit was observed (Marks et al., 2008, 2010). Despite the marked relative percentage increase, the absolute caudal putamen uptake values at Week 40 remain approximately half that of normal control and perhaps to achieve clinical benefit the absolute level of PET signal improvement needs to be yet higher (Sossi et al., 2003).

Previous foetal graft studies, which also showed a mismatch between ¹⁸F-DOPA uptake improvement and negative clinical outcomes, are not necessarily a mirror of our findings, as in these studies it was the grafted tissue that accounted for the enhanced radiotracer trapping rather than biological alteration within the endogenous terminal plexus per se (Olanow et al., 2003). Indeed, a recent post-mortem examination performed 16 years after foetal grafting showed TH innervation indistinguishable from normal and yet clinical benefit was not observed (Kordower et al., 2017). We fully acknowledge that the history of disease-modifying studies in Parkinson’s disease contains similar examples of achieving significant ¹⁸F-DOPA imaging improvements while failing to achieve clinical outcomes (Cochen et al., 2003; Whone et al., 2003; Mittermeyer et al., 2012). This study, we believe, is yet further evidence that for disease-modifying studies the field to-date does not have an adequate clinical outcome correlating biomarker.

Patient safety was reviewed with regard to both drug and drug delivery device. As in previous studies using continuous infusion schemes (Gill et al., 2003; Slevin et al., 2005; Lang et al., 2006), GDNF appeared to be well tolerated and safe. There were no serious adverse events related to the study drug. GDNF-related adverse events included dyskinesia and ON/OFF phenomena, but without the problematic diphasic dyskinesias reported in a previous foetal graft trial (Olanow et al., 2003). It cannot be excluded that Lhermitte’s phenomena and paraesthesia that occurred at higher rates in GDNF-treated patients may have led to partial unblinding. In contrast to the earlier phase II double blind study (Lang et al., 2006), no GDNF-binding serum antibodies in GDNF-treated patients were found in the present study.

Regarding the device, it is important to note that it was an in-house system developed for this trial. The 140 micro-catheters implanted into the primary stage population were delivered into the putamen safely with an operational accuracy of 0.6 ± 0.5 mm (range: 0.0–2.0 mm). In total, 417 infusion cycles amounting to 1668 individual catheter infusions were delivered in the study.

The majority of device-related adverse events were port site-associated, most commonly local hypertrophic scarring or infections, amenable to antibiotics. The frequency of these declined during the trial as surgical and device handling experience improved. No confirmed intracranial infections occurred. Test infusions to demonstrate adequate infusate delivery, prior to randomization, produced a minor stroke in one patient that subsequently fully resolved, and an asymptomatic haemorrhage in another leading to a persistent change in the implantation technique. Neither subject was subsequently randomized.

Treatment infusions post randomization were consistently asymptomatic during administration. Catheter systems remained patent during the 9-month treatment period except for one blocked infusion channel in each treatment group. This was thought to be due to filter occlusion and was compensated for by doubling the infusion volume through the paired catheter on the same side.

In conclusion, we have conducted the first randomized trial in Parkinson’s disease to use CED to administer a trophic factor to the putamen on an every-4-weeks basis via a skull-mounted port. Recruiting patients from across the UK and delivering study treatment on an outpatient basis, this trial shows that, independent from conclusions on GDNF, attending for monthly putamenal infusions of a putative neurorestorative therapy is feasible and tolerable. This study, therefore, marks a potential paradigm shift in direct-target delivery of future novel therapies as they become available, for a host of neurological conditions. As evidenced by increased ¹⁸F-DOPA PET signal, we have shown that this method of administration affords a spatial delivery of GDNF sufficient to achieve a biological effect across the entire putamen. At the 40-week point, however, we have not shown clinical benefit despite this putamen-wide tissue engagement. Future GDNF investigations will need to address potential reasons why our clinical primary endpoint was not reached despite apparently optimizing putamenal therapeutic delivery. The open-label extension study, to be reported separately, providing a further 40 weeks of therapy, may offer evidence on the impact of longer time-duration tissue exposure on clinical outcomes.

Supplementary Material

Funding

The study was funded by Parkinson’s UK (Grant J-1102), with support from The Cure Parkinson’s Trust and was sponsored by North Bristol NHS Trust. Study drug, additional project resources and supplementary funding was provided by MedGenesis Therapeutix, who in turn received program funding support from the Michael J. Fox Foundation for Parkinson’s Research. Renishaw plc manufactured the CED device on behalf of North Bristol NHS Trust and provided additional technical and analytical support. The Gatsby Foundation provided a 3T MRI scanner.

Competing interests

M.L., L.L., C.B., G.A.J., H.C.F. and E.M. are employed by and have shares and/or share options with MedGenesis Therapeutix Inc., owners of the license for GDNF. C.L.B. is a consultant to MedGenesis Therapeutix, Inc. M.W., R.H., O.L., G.P., M.H., C.I., D.J., S.K. and P.S. are employees of Renishaw plc, contract manufactures of the drug-delivery system. S.S.G. is the Medical Director of Renishaw plc and is the inventor of the drug delivery system from which he may have a future royalty share. S.S.G. is on the scientific advisory board of MedGenesis Therapeutix Inc. for which he is reimbursed with share options. No other potential conflict of interest relevant to this article is reported.