Neural stem cells restore cognitive function in Parkinsonian macaques

Cognitive decits as well as disorders of sleep and biological rhythms constitute non-motor symptoms that signicantly impact quality of life in Parkinson’s disease patients. Few studies have evaluated the impact of cell replacement therapy on such non-motor symptoms. Here we used a multidisciplinary approach to assess the therapeutic potential of bilateral grafts of neural stem cells in a macaque model of Parkinson’s disease on both motor and non-motor markers of functional recovery. Grafts led to varying degrees of functional recovery while sham experiments did not. We show unprecedented recovery from cognitive symptoms in addition to a clear clinical motor recuperation. Motor and cognitive recovery but not circadian rhythm recovery correlated with the degree of graft integration into the host environment and with in-vivo levels of striatal dopaminergic transporters and function. This study provides empirical evidence that neural stem cells transplantation eciently restore function at multiple levels in Parkinsonian non-human primates. We demonstrate the promising potential of multiple-sites neural stem cells grafts for Parkinson’s disease but furthermore underline the crucial importance of such multidisciplinary approaches for an effective clinical translation.

We implemented a multidisciplinary approach to assess the therapeutic potential of bilateral grafts of NSCs 15-17 in parkinsonian macaques (nine macaca fascicularis, including three histological controls) on both motor and non-motor markers of functional recovery (Extended Data Fig.1, Extended Data Table 1). We used repeat systemic injections of low-doses of MPTP (0.2 mg/kg every 3-4 days) that induce a parkinsonian syndrome in NHP (macaca mulatta and fascicularis) with a slow and typically dorso-ventral progression of the nigrostriatal denervation, together with the premotor expression of cognitive and circadian i.e. non-motor symptoms 18,19 . We performed in-vivo follow-up of the nigrostriatal lesion with Positron Emission Topography (PET) imaging of DA transporters (DAT) prior to and following the induction of stable motor symptoms 19 as well as following grafts of NSCs. Fluoro-DOPA (F-DOPA) imaging was additionally used to assess NSC effect on DA function 20 .
We derived NSCs from a Rhesus Embryonic Stem Cell line (LYON-ES1) stably expressing the uorescent marker tau-GFP 21,22 . We grafted NSCs neural stem cells (Extended Data Figs.2a-c, 3a, Extended Data Table 2) because of their inherent neuroprotective potential through interaction with host-cells 23 and, their higher survival rate as compared to mature DA cell grafts [24][25][26] . NSCs show both neuronal and astroglial fate in-vitro (Extended Data Fig.2d-f), allowing assessment of the in uence of the host-environment on the differentiation fate of transplanted cells. Importantly, no post-graft overgrowth was observed (Extended Data Fig.3b-c) and few CD68+ macrophages detected in the vicinity of transplanted sites despite the absence of immunosuppressive treatment (Extended Data Fig.3d-f).
We adapted cumulated MPTP-doses to individual clinical motor scores (Extended Data Table 3) in order to take into account the known variability in MPTP-sensitivity 12,18,19 . Clinical scores and behavioral measures were divided into quantiles (Q1-5, see Methods for details) in order to compare groups on the basis of the presumed DA-lesion rather than on the time spent in the premotor (MPTP), motor (post-MPTP/pre-Graft) and post-graft periods. During the pre-Graft period, all cases displayed a stable Parkinsonian syndrome with clinical motor scores ranging from mildly to severely symptomatic (Fig.1a, Extended Data Fig.1a). In two out of the ve grafted cases, we failed to observe recovery from motor symptoms and accordingly assigned cases to the recovered or non-recovered group based on clinical motor outcome following NSC grafts (Fig.1a, Extended Data Fig.1a).
During the pre-graft period, non-motor symptoms became manifest in all groups, and worsened during the premotor period i.e. before motor scores reached clinical thresholds (Fig.1a-c, Extended Data Fig.1). Alterations of circadian rhythms progressively appeared during the premotor phase (Fig.1c, Extended Data Fig.1c) to reach peak levels during the symptomatic period as described previously 12 . This pattern resembles circadian alterations of the sleep-wake cycle observed in PD patients and manifests as increased daytime sleepiness and fragmented sleep structure 27 . The premotor alteration of rest-wake locomotor rhythms in all cases exhibited reduced Light/Dark (LD) ratios and relative amplitude, increased rhythm variability and fragmentation (Fig.1c, Extended Data Fig.1c). Perturbations of the sleep-wake cycle varied greatly according to the level of motor impairments 12 and thus, were not all stably expressed during the motor period. In all cases cognitive performances gradually deteriorated during the premotor period (Fig.1b, Extended Data Fig.1b). Increased errors and reduced successes are mainly due to reduced control over motor planning and parallel executive dysfunctions as observed in PD patients 4,28 . Cognitive performance was dramatically and persistently altered in all cases during the whole motor period following MPTP intoxication offset (Fig.1b, Extended Data Fig.1b). Poor performances could not be imputed to motor impairment as motor-error trials were excluded from all periods. Hence, we can directly relate reduced cognitive performances to the nigrostriatal DA lesion and altered DA innervation of frontal cortex 29 . Nigrostriatal innervation decreased by more than 80% relative to control measures of striatal DAT levels, with the ventral striatum being the most preserved region (Fig.2a), as previously reported 19 .
We delayed NSC grafting by 3 months on average after the last week of MPTP injections (13 ± 6 weeks, see Methods). In order to potentiate graft e ciency on both motor and non-motor symptoms, we performed multiple bilateral transplantations (Extended Data  Transplantation of NSCs induced cognitive and circadian recovery in addition to the reduction of the clinical motor score (Fig.1a-c). Following NSC graft, the experimental group started to differ signi cantly from the non-recovered group as of the 2 nd quantile concerning clinical score whereas differences were signi cant after the 3 rd quantile concerning non-motor symptoms (Fig.1a-c). Clinical scores dropped below symptomatic thresholds on average 6 weeks after NSC-grafts, achieving maximal motor recovery 5 weeks later (Fig.1A, Fig.S1A). Recovery from non-motor de cits was more gradual, with cognitive performances and LD ratio not signi cantly different from control at an average of 18 weeks following NSC-grafts. Motor and non-motor symptoms persisted in the non-recovered group displaying no sustained improvements following NSC grafts ( Fig.1b-c, Extended Data Fig.1b-c). The initial LD ratio recovery after graft was followed by a relapse on the last two quantiles (Fig.1c). Non-parametric circadian rhythm analyses (NPCRA, see Methods) revealed high variability across cases withholding post-graft group comparison (Extended Data Fig.1c).
Markers of DA function and DAT increases following NSC graft correlate with behavioral changes (Fig.2jk, Extended Data Fig.5a). We show that PET-scan estimation of nigrostriatal innervation through DAT binding was signi cantly increased bilaterally in the ventral striatum as early as 15 weeks post-graft, suggesting a primarily protective effect of grafts on the remaining pool of nigrostriatal terminals (Fig.1d). F-DOPA scans following grafts con rm a selective effect of NSCs on DA synthesis with maximal motor recovery. Signi cant increases in DA function and DAT binding are observed in grafted parts of the striatum when compared to the same regions in non-recovered cases (Fig.4a). Notably, we observed no improvement of motor or non-motor symptoms after sham transplantation and functional markers displayed no difference between recovered vs. non-recovered cases in striatal DAT binding (Extended Data Fig.5b-c, see Methods for details).
Integration of the NSC-graft correlate with clinical, cognitive and functional recovery. Post-mortem immuno-histological examination shows that NSCs integrate into the nigrostriatal network and differentiate into mainly mature neurons and astrocytes ( Fig.2b-e). Furthermore, NSCs spontaneously differentiate into aminergic cells thereby potentially restoring a signi cant proportion of the lost DA pool (Fig.2b). In the recovered group the survival rate varied widely across cases and grafted sites, ranging from 2 -61% (corresponding to the number of surviving tau-GFP + cells per grafted site ranging from 3.1×10 3 to 1.8×10 5 respectively, Extended Data Table 4). Unilateral survival of grafted NSCs did not lead to a systematic difference in recovery across hemispheres (Extended Data Fig.6a-b) suggesting crosshemispheric compensation 30 , although we observed an asymmetric time-course of cognitive recovery with unilateral CdN survival (Extended Data Fig.6b, Extended Data Table 4) that could also be detected through a minor unilateral DAT binding decrease (Extended Data Fig. 6c, right). No difference in recovery could be attributed to NSC type grafted (although survival rate did vary, Extended Data Tables 2, 4).
Among surviving cells, astrocytic fate accounted for 34-52% (GFAP + /tau-GFP + , Fig.2b, e) and neuronal fate for 11-18% (MAP2 + /tau-GFP + , Fig.2b, d). Additionally, up to 13% of tau-GFP + cells in SN grafts differentiated into aminergic cells (TH + /tau-GFP + cells, Fig.2b-c) which corresponds to an estimated number of 1.5×10 4 cells and hence to a restoration of nearly 18% of the original pool of DA-containing midbrain neurons, as reported for the same NHP species 31 . Computerized reconstruction of tau-GFP + soma and processes showed that grafts are able to extend and project dendrites to distant locations (Extended Data Fig.7). We could further detect local processes (i.e. less than 1mm) directed toward surviving TH + cells from the host tissue, that formed synaptic contacts with host neuronal cells ( Fig.3a-b, d-f). At some graft sites, aggregates of host-cells (i.e. tau-GFP negative) were observed inside the graftcore with numerous DAT + processes ( Fig.3g-i), suggesting that differentiated NSCs provide a milieu which orients endogenous cells towards DA fate 32 , and include a relatively high proportion of GFAP + /tau-GFP + cells (Fig.2b, e). Immuno-labeling revealed that non-recovered cases presented no surviving tau-GFP + cells (Extended Data Fig.8). The nigrostriatal DA lesion was most severe in non-recovered cases (Fig.4b), which is considered deleterious for grafted cells and has been shown to decrease the functional impact of the graft 33 . Furthermore, immunological rejection of the grafted NSCs cannot be fully excluded as no immunosuppressive treatment was employed. The fact that non-motor symptoms initially improved in the non-recovered group and suddenly worsened at the 4 th Quantile (Fig.1b-c, Extended Data Fig.1b-c) suggests a rather chronic graft rejection in these cases.
As an additional control, we compared the time course of behavioral symptoms and functional markers to those following spontaneous recovery. In some cases, spontaneous recovery was initially induced by halting MPTP intoxication as described previously 18,19 . This allowed better differentiation between features accompanying spontaneous motor recovery from those observed following graft. Spontaneous recovery of clinical motor symptoms had faster dynamics compared to graft-induced recovery (Fig.2l). In addition there was an absence of cognitive improvements accompanying spontaneous motor recovery, contrasting with the progressive cognitive improvement following grafts (Fig.2l). We previously demonstrated that spontaneous motor recovery could occur despite an average 71.5±9.7% reduction in striatal DAT relative to control levels and that stable symptoms were observed when the lesion reached 82.3±7.9% of control 19 . We now show graft-induced recovery despite 78.04 ±12% average reduction in striatal DAT binding (Fig.4a). These results con rm the speci city of graft effects as compared to spontaneous recovery, especially concerning cognitive recovery (Fig.2l).
In sum, we demonstrate that multisite graft of NSCs allows recovery from cognitive and clinical motor symptoms while the effects on chronobiological rhythms were too variable to be attributed directly to grafts. In-vivo imaging con rmed positive effects on nigrostriatal innervation and function (Figs.1d, 2j-k, 3a). Post-mortem histology showed that a substantial proportion of DA fate was assumed by grafted NSCs (Fig.2b) thereby supporting their role in cognitive and motor recovery. Further, the large proportion of glial cells observed in the grafts (Fig.2b) is consistent with the protective effect of grafts on the remaining nigrostriatal innervation after lesion, as evidenced by functional ndings on striatal DAT levels ( Fig.1d) and as reported after human NSC grafts in MPTP-NHP model 34 .
Three decades after the rst clinical trials of transplanting cells from fetal ventral mesencephalon in PD patients 35 , cell replacement therapy in PD is still not a proposed treatment. However, this therapeutic approach has recently been shown to be e cient at different levels 36-38 , leading to more recent trials in order to improve grafting procedures, and prepare for stem-cell based transplantation in humans 39 , while waiting for perspectives from pre-clinical investigations in animal models 40 . Here, we report for the rst time e cient recovery from cognitive and clinical motor symptoms following NSC grafts in the NHP model of PD. This study underlines the crucial importance of longitudinal and multifaceted approaches for a better translation to the clinic and appraisal of early-phase clinical trials.

Methods
In agreement with the 3Rs 41 , the rationale for the current study is that each case represents its own control through detailed follow-up of all the periods of the protocol i.e. CTR, MPTP, post-MPTP and Prevs. Post-Graft. We published precise and detailed reports for each clinical, behavioral and functional parameter followed as well as for the characterization of NSCs grafted in the present study. These precise and detailed control procedures are described in Extended Data Table 1.

Ethical statements
All procedures were carried out according to the 1986 European Community Council Directives (86/609/EEC) which was the o cial directive at the time of experiments, the French Commission for animal experimentation, the Department of Veterinary Services (DDSV Lyon, France). Authorization for the present study was delivered by the "Préfet de la Région Rhône Alpes" and the "Directeur départemental de la protection des populations" under Permit Number: #A690290402. All procedures were designed with reference to the recommendations of the Weatherall report, "The use of non-human primates in research". All monkeys were closely monitored on a regular basis throughout the day, by researchers and animal care staff, in order to ensure that levels of health and welfare were strictly maintained, particularly during the MPTP period. Adaptations to housing and feeding procedures were made in direct response to individual symptoms in the MPTP phase, for example adaptations of water provision to ensure monkeys were able to drink ad libitum. To alleviate physical suffering from motor symptoms progression (such as rigidity) the intoxication procedure was rst cautiously stopped when the PMRS-motor score was above ten for two consecutive days following one MPTP injection; further, antalgic and/or anti-in ammatory drugs were delivered to animals experiencing debilitating motor symptoms. Before being sacri ced, animals were rst tranquilized with a pre-anesthetic agent (chlorpromazine hydrochloride, Largactil) and anesthesia was induced with Ketamine before deep anesthesia was obtained by means of a large dose of pentobarbital sodium (Vibrac, 100 mg kg− 1, i.p.; lethal dose con rmed by complete loss of corneal re ex).

MPTP-intoxication and study design
Six late middle-aged -11-13 (13-17) years old at protocol onset (end) female macaque monkeys (Macaca fascicularis, 4-5kg) were intoxicated with low-dose 6-methyl-1-Methyl-4-phenyl-1,2,3,6tetrahydropyridin injections (MPTP, 0.2mg.kg -1 , i.m). Animals were housed in a room dedicated to MPTP experiments, with free access to water and received food twice a day. The neurotoxin was delivered according to two different regimes: 1) chronically each 3-4 days and 2) acute intoxication (daily injections) as described previously 19,42 . In cases 1, 3 and 4, the rst session of MPTP injections was suspended as soon as the clinical score reached symptomatic threshold (Extended Data Fig.1), allowing investigation of dopaminergic (DA) neurotransmission after spontaneous behavioral recovery 18,19 and thus providing critical parameters to compare with potential recovery following transplantation. For the last-MPTP intoxication i.e. ending with persistent motor-symptoms in all cases, injections were cautiously stopped when the PMRS-motor score was above ten for two consecutive days following one MPTP injection. For results presentation, animals were grouped according to their clinical state after transplantation (see Fig.1) i.e. 1 st group -recovered (n=3, cases 1, 4, 5) and, 2 nd group -non-recovered (n=2, cases 3 and 6). Delays between last MPTP injection and i) transplantation were 4-23 weeks (n=6, all cases), ii) sham-grafts were 6-16 weeks (n=2, cases 1 and 6). Data are presented according to the following 5 periods of the protocol: 1-MPTP (during MPTP intoxication period); 2-post-MPTP Recovery (following arrest of rst MPTP injections, clinical motor score returned below 5 throughout this period i.e. non-symptomatic); 3-post-MPTP Symptomatic (following arrest of last MPTP injections, clinical motor score remained stably above 5 during this period i.e. motor-symptomatic); 4-post-Sham (following sham grafts, details in Surgical procedures); 5-post-Grafts (following transplantation, details in Surgical procedures). Case 4 had a high-score post-MPTP (motor symptoms>15), characterized by tremor at rest and freezing. This motor-symptomatic state persisted during the 4 weeks prior to transplantation (Extended Data Fig.1a). Case 6 had a high-score post-MPTP (motor symptoms>15), characterized by severe rigidity. Musculoskeletal pain is a recognized source of pain syndromes and discomfort in Parkinson's disease patients. We thus delivered regularly ketoprofen (Profenid), from week 5 post-MPTP, in order to relieve suspected apparent musculoskeletal pain. Subsequently the clinical score for this animal progressively improved, due to amelioration on the rigidity scale and ability to manipulate food (from week5 to week10 post-MPTP, Extended Data Fig.1a), but stabilized at week10 and remained largely symptomatic (score >12) despite continuous treatment and disappearance of apparent musculoskeletal to frame injections (maximum total score of 23): a score below 5 de ned the non-symptomatic state in premotor and recovery periods, and a score above 5 de ned symptomatic periods.
Cognitive behavior task -detour task Cognitive performance was monitored 3-5 days a week using a previously described behavioral test 18,19 . Brie y, performance on the 'detour task' was evaluated by the percent of successes (retrieval of reward on the rst reach, over the total number of trials) and errors (barrier hits, over the total number of responses observed -there could be several responses per trial except in the case of success). Errors due to motor impairments were classi ed as such and further discarded from the performance evaluation, so as to avoid confounds due to the symptomatic state of the animal. The performance on the 'detour task' depends on the integrity of frontal cortex, the dopaminergic system and DA innervation of frontal cortex 44 and has been used to assess the impact of lesion of the nigrostriatal axis on cognitive performance as well as potential therapy for cognitive impairments [45][46][47] . For inter-individual comparison, percent errors and success were, for each case, rst converted in percent change by the average control value and normalized between 0 and 1 by overall maximum and minimum score respectively.
Circadian rhythm follow-up Animals were maintained under an LD cycle (12h light: 12h dark) of approximately 450-500 lux during the light phase and 0 lux during the dark phase. Locomotor activity was continuously recorded in 1 min intervals throughout the entire duration of the study using passive-infrared motion detectors mounted above each animal's cage. The motion captors were connected to a computerized data acquisition system (Circadian Activity Monitoring, INSERM, France 48 ). Data were analyzed using the Clocklab software package (Actimetrics, Evanston, IL, USA).
We previously used Chi-squared periodogram method (Extended Data Table 1) to calculate the period of behavioral rhythm which is de ned as the time elapsed for one complete oscillation or cycle (the distance in time between two consecutive peaks or troughs of a recurring rhythm). The stability and the consistency of the rhythm are re ected by the value of the periodogram amplitude which corresponds to the zenith of a periodogram curve. The subjective day and the subjective night are de ned as the segment of a circadian cycle during the free-running state that corresponds to, respectively, the light and dark segments during entrainment by a LD cycle.
Non-parametric circadian rhythm analyses (NPCRA) were used to estimate the strength and fragmentation of rest-activity rhythms 12,18 . These included measurements of inter-daily stability (calculated as the ratio between the variance of the average 24-hour pattern around the mean and the overall variance and gives an indication of the consistency of day to day activity or the strength of coupling to the LD cycle), intra-daily variability (calculated as the ratio of the mean squares of the difference between successive hours and the mean squares around the grand mean and gives an indication of the frequency of transitions between rest and activity periods, corresponding to the fragmentation of the rhythm) and, relative amplitude (ratio between acrophase and nadir of the rhythm, representing the ratio between activity amplitude in light and dark phases). When the rest-activity rhythm is stable, inter-daily stability is high, the intra-daily variability is low and relative amplitude is high.
Additionally, we used images from PET scans using L-3,4-dihydroxy-6-( 18 F) uorophenylalanine ([ 18 F]-FDOPA, an analog of the DA precursor L-DOPA) to evaluate the central dopaminergic function of presynaptic neurons, i) just before grafts (during stable expression of typical motor parkinsonian symptoms i.e. symptomatic period) and, ii) after grafts at different delays (see Extended Data Fig.1).
We performed [ 11 C]-PE2I and [ 18 F]-FDOPA PET scans with an ECAT Exact HR+ tomograph (Siemens CTI), in 3D acquisition mode, covering an axial distance of 15.2cm. The transaxial resolution of the reconstructed images was about 4.1mm full-width and half maximum in the center. We acquired transmission scans with three rotating 68Ge sources to correct emission scans for the attenuation of 511 keV photonrays through tissue and head support. Full procedure for PET-scans acquisition, modeling and ROIs de nition for regional assessment of [ 11 C]-PE2I changes have been published 19 . Fluoro-DOPA was labeled with 18 F-uoride (cyclotron-produced isotope, half-life = 109 min). Speci c radioactivity was 3.27 ± 1.3 mCi/μmol. Radiochemical and chemical purity of produced [ 18 F]-FDOPA (as determined by HPLC) was above 99%. After anesthesia was induced and head secured with a MRI-compatible stereotaxic frame (Kopf, CA, USA) to reduce variability in the measure, a cannula was inserted in the femoral vein.
[ 18 F]-FDOPA was injected as a bolus over a 4 s period followed by a saline ush. Radioactivity was measured in a series of 26 sequential time frames of increasing duration (from 30 sec to 5 min; total time 90 min).
PET modeling and regional assignment of PET changes were done as previously described 19 and applied to F-DOPA measurement 52 . Statistical positive (shown in Fig.1d) and negative (shown in Extended Data

Transplanted cells and Surgical procedures
We derived neural stem cells (NSCs) from a rhesus embryonic stem cell line stably expressing tau-GFP (LYON-ESC line) 21 . NSCs were obtained either i) as described in 53 where multipotent NSCs were ampli ed by mild trypsinization and cultured in the presence of EGF and FGF2 (Extended Data Fig.2a, c), or ii) after MS5 induced-neural differentiation followed by early midbrain differentiation, as described in 54 (Extended Data Fig.2b, c, f). The latter cell type corresponds to early DA midbrain neural precursors. For simplicity and because the two cell types lead to similar recovery when grafted in different hemispheres (Extended Data Fig.6), we used the generic term of NSCs. NSCs express NSC speci c markers (Extended Data Figs.2a-c, 3a) and showed in-vitro potential to fully differentiate along glial and neuronal pathways (Extended Data Fig.2e) including mature DA neurons (Extended Data Fig.2f). Grafted NSCs retained the capacity to differentiate into glial and neurons in vivo, depending on the local environment of the graft (Figs.2b-e, i, 3 and Extended Data Fig.4a).
We performed all transplantations bilaterally using 5µL Hamilton syringe and small trepanations over target sites. Stereotaxic coordinates were calculated from individuals T1 and T2 structural MRI images. We undertook sham-grafts (trepanation and injection of 5µL PBS without NSCs) in cases 1 and 6 in order to control for behavioral and clinical effects due to the surgical act in itself. We placed sham-grafts (PBSonly) in posterior caudate and anterior putamen, using the same protocol as for NSCs transplantation.
The amount of injected NSCs varied between 3.10 5 -10 6 cells for SN sites and 1.5.10 5 -10 6 for striatum sites (Extended Data Table 2). The syringe was lowered 500µm further than the target position and then retracted to target; cells were then injected at a rate of 1µL per min after a 5min delay. After an additional 3 minute delay after injection completion, the syringe was retracted slowly by 2mm; another 2 minute delay was observed before full retraction. We assessed cell viability by trypan blue exclusion before injection and counted the remaining cells after injection. Viability at both time points was found to be between 95 and 100%. Cell mortality due to syringe uptake was estimated at 2%.
For immuno uorescent staining on free-oating brain slices, sections were washed six to ten times in Tris buffered saline (TBS) and permeabilized in Triton X-100 (0,5%; SIGMA, #T9284). Nonspeci c binding was blocked with 10% normal goat serum or donkey serum (Goat: Invitrogen; #16210064; Donkey: SIGMA #D9663) for 45 min at room temperature. Sections were incubated for three days at 4°C, with primary antibodies diluted in Dako antibody diluent (Dako #S3022) supplemented with 0,5% Triton X-100. They were then exposed to secondary antibodies and DAPI (1/10 000; Invitrogen #D1306) for 2 hours at room temperature. Sections were mounted on slides and examined using a confocal microscope (Leica TCS SP5). Sections were regularly (every ve sections i.e. every 250µm) processed for GFP immunostaining to detect grafted sites extent.
We determined the length of processes emanating from the graft core by measuring the length of tau-GFP positive projections using the ImageJ software (NeuronJ plugins, http://imagej.nih.gov/ij/), presented in Extended Data Fig.5d. We estimated the graft volume through a 3D reconstruction process using contours traced and compiled with the Mercator software (ExploraNova, http://www.exploranova.com/products/mercator/), presented in Extended Data Table 3, Fig.2f Immunostaining and RT-PCR analyses of cultured cells were done according to previously published protocols 21 .
Semi-quantitative immunohistochemistry Three age and weight matched Macaca fascicularis served as controls. We evaluated the extent and amplitude of the nigrostriatal lesion through semi-quantitative immunohistochemistry of striatal structures, as previously described 19, 55 . The mean optical density (O.D.) of TH and DAT in immunopositive regions were computed on at least 5 sections per animal (5-20) from 8bit images, normalized and compared with two-sample t-test separately for each ROI (ttest2, Matlab). In Fig.2a, composite CTR and MPTP images has been enhanced for brightness and contrast, for illustration purpose.

Statistics
We segmented data for MPTP, post-MPTP and post-Graft periods into ve equal epochs and grouped variables into each of these epochs. This method of segmentation, described in 18,19 , thus reveals normalized stages of the progression of processes and allows for comparison of different parameters across an equivalent number of epochs for all subjects, called Quantiles (Fig.1, 27 ± 3 days for the CTR period; duration per quantile for premotor 10 ± 5 days , motor 18 ± 8 days and post-Graft 42 ± 12 days).
Results are presented as means ± standard errors.
Parameters were compared between groups by a treatment-contrast test using one group as the rst level and Bonferroni corrected for multiple comparisons and, within groups across periods by a treatmentcontrast test using control measures as the rst level (estimated standard errors and z-ratio were computed using GLM t and contrast result was given by two-tailed p-values corresponding to z-ratio based on a Student t-test). Signi cance was considered at p<0.05. Statistical analyzes were computed using R software (R Foundation for Statistical Computing, Vienna, Austrian http://www.R-project.org) and the summary function (Chambers, J. M. and Hastie, T. J. (1992) Statistical Models in S. Wadsworth & Brooks/Cole).

Data availability
The data that support the ndings of this study are available from the corresponding author upon reasonable request.