GDNF was first characterized by Lin and coworkers in 1993 [1] as demonstrated to provide potent trophic effects on dopaminergic midbrain neurons in culture. Since that initial observation, GDNF has been found to protect and/or augment dopaminergic nigrostriatal function in normal animals, aged animals, and virtually every toxin animal model in multiple species [2]. These studies lead to open label and blinded clinical trial s testing initially the safety and tolerability and then efficacy of intraventricular and then intraputamenal GDNF protein infusions [3–5]. The intraventricular studies were based upon weak preclinical data and were found to not only be nonefficacious but also to have serious side effects. As an aside, given their potency and potentially widespread effects, putting any trophic factor into the ventricular system is probably not a good idea [3, 6]. The rationale for intraputamenal delivery is much stronger. However, this clinical trial studies failed to demonstrate efficacy and was stopped prematurely due to safety concerns although it is likely that most these safety concerns were not very serious [7]. At the time these trials were being planned and ongoing, Ceregene inc. attempted unsuccessfully to license the gene for GDNF from Amgen Inc., the stakeholder for GDNF, for a gene therapy trial testing whether viral vector delivery of GDNF would be safe and efficacious in patients with PD. Since there was no reasonable path to the clinic without a license to operate with GDNF, Ceregene approached Gene Johnson, Jeffrey Milbrandt, and Washington University and successfully licensed another member of the GFL family of ligands, neurturin. While my lab and many other demonstrated strong efficacy in rodent and nonhuman primate models with GDNF [2], there was far less data at this time testing neurturin with regard to distribution by gene delivery , efficacy in these same models, as well as safety and tolerability. Thus it was back to the drawing board and Ceregene’s inability to secure a license for GDNF put the field back a number of years.

Ceregene performed numerous experiments and demonstrated that neurturin was equipotent to GDNF in a number of rodent PD models [8, 9]. The distribution of neurturin was less than that of GDNF [9] although differences in the sensitivity between the GDNF and neurturin antibodies likely exaggerated GDNF’s superiority in distribution. Both were extremely well tolerated. With regard to efficacy, CERE-120 (the name for gene delivery of neurturin) increased dopamine markers in young rats, aged rats, 6-hydroxydopamine lesioned rats, normal monkeys, aged monkeys, and monkeys intoxicated with MPTP. The preclinical package for both efficacy and safety was extremely strong and Ceregene initiated a Phase I clinical trial in 2004 [10]. This trial consisted of 12 patients with half receiving a low dose (0.3 × 10¹¹ vector genomes (vg)/patient) and half receiving a high dose (5.4 × 10¹¹ vg/patient). The trial lasted 1 year post-treatment. Safety and tolerability was excellent as there were no serious adverse events due to the test article [10]. In the manuscript describing the open label trial, we reported that measures of motor function showed improvement at 1 year; for example, a mean improvement in the off-medication motor subscore of the Unified Parkinson’s Disease Rating Scale (UPDRS) of 14 points (SD 8; p = 0.000121 [36 % mean increase; p = 0.000123]) and a mean increase of 2.3 h (2; 25 % group mean increase; p = 0.0250) in on time without troublesome dyskinesia were seen. Interestingly, the expected increased in fluorodopa positron emission tomography was not observed. This was an open label trial that was not powered for efficacy. Open labeled trials, especially for PD, have a high potential for placebo effects and experimenter bias [11] making the efficacy data virtually meaningless. Indeed, the 36 % improvement in UPDRS is clearly within the range of placebo yet, and likely appropriately so, Ceregene continued on. But should we have continued on? At the time we had a signal, but the signal was not very strong. In an efficacy study, when the signal is not strong in an open label analysis it often does not pass muster when submitted to the more rigorous double blind assessment. In addition, perhaps the failure of CERE-120 to enhance fluorodopa uptake should have been a second red flag. If our hypothesis was that CERE-120 would enhance nigrostriatal function, we failed to do so on the one measure, albeit indirect, that we had. However, the fiscal realities of the biotech world demanded that we go forward. As a member of the scientific advisory board for Ceregene, I, as well as all my colleagues, fully supported that decision.

Following the success (safety and tolerability) of the Phase I trial, Ceregene initiated a Phase II efficacy trial. Fifty-eight PD patients were randomized to CERE-120 and placebo in a 2:1 ratio. Multiple injections into the putamen we made bilaterally using higher dose (5.4 × 1011 vg/patient) from the Phase I trial was employed. The primary endpoint was change from baseline to 12 months in the motor subscore of the unified Parkinson’s disease rating scale in the practically defined off state. The trial failed to meet this primary endpoint. So was this the end? No because another “signal” of efficacy was observed. When the trial ended 12 months after the last patient received surgery, most patients had been followed in a blinded fashion at 15 and 18 months. An ad hoc statistical analysis suggested that these points were significant at the 0.025 level. Multiple secondary analyses were significantly improved at 18 months relative to 12 months. However, these analyses were not corrected for multiple comparisons and thus do not have the appropriate scientific strength or rigor. However, they were still strong enough to for Ceregene to initiate a third clinical trial .