Introduction

One of the principal hallmarks of Parkinson’s disease is degeneration of dopaminergic neurons in the substantia nigra pars compacta (SNc), leading to striatal dopamine deficiency1. The onset of Parkinson’s disease motor manifestations is typically focal and highly correlated with selective loss of dopaminergic neurons in the SNc and dopamine depletion in the caudal putamen contralateral to the clinically affected body side2,3. The greatest loss of dopaminergic neurons is found in the ventral tier of the SNc (vSNc), whereas dopaminergic neurons in the dorsal tier of the SNc (dSNc) and ventral tegmental area (VTA) are less vulnerable4,5. More specifically, dopaminergic neurons contained within the so-called nigrosome region of vSNc appear to be the earliest and mostly affected4,6. Other dopaminergic populations are much less affected or remain intact until the neurodegenerative process is far more advanced many years after diagnosis4,5,7,8. This pattern of cell loss is essentially maintained regardless of etiopathogenic disease mechanisms (i.e., genetic, toxin-related, environmental factors, and infectious agents)9. This suggests that differential dopaminergic neuronal vulnerability is primarily related to intrinsic cellular properties and connectivity rather than to specific triggering factors of neurodegeneration10,11.

Work in rodents has shown significant molecular and genetic heterogeneity among dopaminergic neurons within the midbrain12,13. Transcriptomic studies comparing the dopaminergic systems of mice and non-human primates (NHPs) have revealed important inter-species differences14,15. Therefore, studies in NHPs are instrumental in ascertaining the onset and key mechanisms underlying the selective vulnerability of midbrain dopaminergic neurons16. Recent human and NHPs transcriptomic profiling studies have revealed distinct markers within midbrain neurons that correlate with vulnerability17,18,19,20,21, however, a precise anatomical correlation is lacking. Moreover, it is unclear whether different dopaminergic subpopulations correspond to the human nigrosome-matrix compartmentation13. This is particularly relevant, because neuronal vulnerability in the SNc has regional selectivity10,22,23.

This study addresses molecular vulnerability together with topography selectivity in the SNc of NHPs. We reveal distinct molecularly defined dopaminergic neuron subpopulations in the NHP brain and their differential loss in the MPTP (1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine) parkinsonian model. These subpopulations give rise to dopaminergic projections that, despite some overlap, target distinctively specific basal ganglia regions (i.e. putamen and globus pallidus) and show a differential degree of susceptibility to MPTP.

Results

Nigrostriatal damage following MPTP-treatment

The density of nigrostriatal terminals in control and MPTP-treated NHPs was first assessed by measuring the optical density within tyrosine hydroxylase (TH) stained coronal sections. A progressive reduction in TH immunoreactivity with increasing motor impairment was observed in the striatum of all MPTP-treated monkeys as previously reported in the same model (Supplementary Fig. 1a–c)24,25. Regarding intra-striatal changes, the pre-commissural putamen was relatively less affected than the post-commissural putamen and caudate nuclei (Supplementary Fig. 1d, e). There was also a clear immunoreactivity gradient in the dorsoventral axis, which was already apparent in the asymptomatic group. The nucleus accumbens showed the best preservation of TH immunoreactivity (Supplementary Fig. 1a). To elucidate the anatomical location and specific molecular markers underlying differential vulnerability of dopaminergic cellular phenotypes, double immunofluorescence for different pairs of antibodies was performed throughout the rostrocaudal axis of the NHP midbrain.

Patterns of calbindin expression in the ventral midbrain

Intense calbindin (Cb) neuropil staining was observed in the substantia nigra pars reticulata (SNr), together with minimal Cb staining in several patchy areas through the vSNc and the VTA (Fig. 1, Supplementary Fig. 2). These areas with minimal Cb staining formed a complex three-dimensional continuum and comprised densely-packed TH+ neurons, matching the so-called nigrosome region in the human midbrain26 (Fig. 1a). This region corresponds to the vSNc tier, and it is related to the densocellular region and the cell columns described by other authors27,28,29,30,31. The region enriched in Cb+ neuropil and Cb+ neurons is termed matrix in keeping with an analogous description reported for the human midbrain26, and includes the dSNc, part of the dorsal VTA and the SNr.

Fig. 1: Pattern of calbindin expression in the ventral midbrain.
figure 1

a Representative images of TH and Cb immunohistochemistry in the midbrain of control monkeys. Dashed insets at higher magnification of TH+ and Cb+ cells. b Percentage of TH+ neurons constraint within nigrosome and matrix in the control monkey’s midbrain (n = 2; 5978 ± 281 neurons). c Percentage of TH+ neurons co-expressing Cb in the control monkey’s midbrain (n = 2; 5724 ± 359). d, e Loss of TH+ neurons in the matrix and nigrosome compartments in control and MPTP-treated monkeys (n = 2/group; MPTP n = 6). f, g Loss of TH + /Cb+ and TH + /Cb- neurons in the control and MPTP-treated monkeys (n = 2/group; MPTP n = 6). Data represent mean + SEM. Symbols indicate significant differences compared with control *P < 0.05. Ctrl, control; Asym, asymptomatic; Rec, recovered; Park, parkinsonian; * symbol indicate nigrosome calbindin-poor region; dSNc: Dorsal Substantia nigra pars compacta; SNr: Substantia nigra pars reticulata; VTA: Ventral Tegmental Area; D: dorsal; V: ventral; M: medial; L: lateral. Scale bar: 1 mm (inset 50 μm).

Full size image

In control monkeys, double immunofluorescence for TH+ and Cb+ revealed that 74% of TH+ neurons were located within the Cb- region defined as the nigrosome, whereas the remaining 26% of TH+ neurons were found within the matrix compartment (Fig. 1b). On average, colocalization of TH and Cb was observed in 14% of neurons; these dopaminergic cells were located in the dSNc and VTA (Fig. 1c). TH+ neurons located in the nigrosome lacked Cb immunoreactivity.

A non-homogeneous reduction in the number of TH+ neurons was observed after MPTP intoxication. Neuronal loss was higher in the nigrosome (54%) than in the matrix (41%) (Fig. 1d). Cell loss in the nigrosome (vSNc) was progressive and parallel to the degree of parkinsonism of MPTP-treated monkeys (Fig. 1e). In other words, monkeys with the greatest loss of TH+ neurons in the nigrosome (77%) were those that showed the most severe parkinsonism. In contrast, dopaminergic cell loss was roughly similar within the matrix compartment across all MPTP-treated groups (47% in the parkinsonian group) (Fig. 1e).

In MPTP-treated monkeys, dopaminergic neurons positive for Cb (TH + /Cb + ) were more resistant to MPTP treatment (15% of cell loss) than neurons lacking Cb (TH + /Cb-) (55% average cell loss). In the parkinsonian group, percentages of dopaminergic cell loss were 22% and 76%, respectively, for TH + /Cb+ and TH + /Cb− (Fig. 1f-g).

Patterns of Girk2 expression in the ventral midbrain

Double immunofluorescence for TH and G-protein-activated inward rectifying potassium type 2 (Girk2) revealed marked preferential expression of Girk2 in dopaminergic neurons of the vSNc (Fig. 2a), with variation in the intensity of staining (Fig. 2c). Girk2 expression followed a dorsoventral and mediolateral gradient, with neurons within the dSNc and VTA showing very low Girk2 expression (Fig. 2a; Supplementary Fig. 3). Girk2 staining was present in 85% of all midbrain TH+ neurons, with the remaining TH+ neurons (15%) displaying none or weak Girk2 immunoreactivity, the latter confined to the dSNc and VTA (Fig. 2d). Double immunofluorescence for Cb and Girk2 revealed that most of the neurons that showed the highest expression levels of Girk2+ were in the nigrosome compartment and lacked Cb expression (84%; Fig. 2b and Fig. 2e; Supplementary Fig. 4). Moreover, the percentage of double-labeled neurons with Girk2 and Cb was very low (5%; Fig. 2e). Girk2 expression was not restricted to dopaminergic cell somata; it was also observed in the dendritic trees of TH+ neurons located in the vSNc (Fig. 2a; Supplementary Fig. 3). Some Girk2+ fibers in the ventral SNr did not colocalize with TH (Fig. 2a; Supplementary Fig. 3). None or very faint Girk2 expression was found in other catecholaminergic nuclei of the brain (Supplementary Fig. 5). In terms of vulnerability, the Girk2+ population matches the Cb- population and is highly vulnerable to MPTP. This observation, together with the uneven intensity of Girk2 neuronal staining, led us to omit TH + /Girk2+ vs TH + /Girk2- neuronal quantification as part of the analysis. Instead, we focused on the Aldehyde Dehydrogenase 1 Family Member A1 (Aldh1a1) population.

Fig. 2: Pattern of Girk2 expression in the ventral midbrain.
figure 2

a Representative images of TH and Girk2 immunohistochemistry in the midbrain of control monkeys. Dashed insets at higher magnification of TH+ and Girk2+ cells. Solid inset in c. b Representative images of Cb and Girk2 immunohistochemistry in the midbrain of control monkeys. Dashed insets at higher magnification of Cb + /Girk2+ cells. c Magnified images of solid insets in a showing representative TH + /Girk2+ cells with varying intensities of Girk2 staining (i.e., weakly double-stained neurons). d Percentages of expression of TH + /Girk2+ in the control monkey midbrain (n = 2; 3809 ± 645 neurons). e Percentages of co-expression of Cb + /Girk2+ in the control monkey midbrain (n = 2; 2914 ± 304 neurons). dSNc Dorsal Substantia nigra pars compacta, vSNc: Ventral Substantia nigra pars compacta; SNr: Substantia nigra pars reticulata; VTA: Ventral Tegmental Area; D: dorsal; V: ventral; M: medial; L lateral. Scale bar: 1 mm (inset a: 100 μm; inset b: 50 μm).

Full size image

Patterns of Aldh1a1 expression in the ventral midbrain

In the ventral midbrain of control monkeys, neurons which expressed Aldh1a1 were distributed within specific clusters (Figs. 3 and 4; Supplementary Fig. 6). Neurons co-expressing TH and Aldh1a1 (TH + /Aldh1a1 + ) comprised up to 71% of the total number of midbrain dopaminergic neurons, showing a sharp preference for location within the vSNc and the VTA (Fig. 4c). Catecholaminergic cell groups outside the midbrain dopaminergic neurons, including the olfactory bulb and locus coeruleus, did not show Aldh1a1 expression (Supplementary Fig. 7). Double immunofluorescence for Aldh1a1 and Cb revealed that virtually all Aldh1a1+ neurons are negative for Cb (97%) and indeed are preferentially located within the nigrosome compartment (Fig. 4a, d; Supplementary Fig. 8). Double immunofluorescence for Girk2 and Aldh1a1 showed up to three different neuronal phenotypes (Fig. 4b, e; Supplementary Fig. 9): (i) Aldh1a1 + /Girk2+ neurons, representing the most abundant phenotype (72%); (ii) Aldh1a1-/Girk2+ neurons, 15% of the total number of neurons and preferentially located in lateral territories of the SNc; and (iii) Aldh1a1 + /Girk2- neurons (none or weak Girk2 staining), being 13% of the total number of neurons, with an exclusive location in the VTA (Supplementary Fig. 9).

Fig. 3: Pattern of Aldh1a1 expression in the ventral midbrain.
figure 3

a–h Representative images of coronal sections immunostained for TH and Aldh1a1 from rostral (bregma -10 mm) to caudal (bregma -17 mm) (left to right, top to bottom). Note how the localization of Aldh1a1+ neurons is restricted to the ventral part of the SNc in the control monkey midbrain. Numbers in upper-right corners indicate millimeter distances from bregma from the atlas of Paxinos et al. 105. SNc: Substantia nigra pars compacta; SNr: Substantia nigra pars reticulata; VTA: Ventral Tegmental Area; D: dorsal; V: ventral; M: medial; L: lateral. Scale bar: 1 mm.

Full size image
Fig. 4: Pattern of Aldh1a1 expression in the ventral midbrain and vulnerability.
figure 4

a Representative images of Aldh1a1 and Cb immunohistochemistry in the midbrain of control monkeys. Dashed insets at higher magnification of Aldh1a1+ and Cb+ cells. b Representative images of Aldh1a1 and Girk2 immunohistochemistry in the midbrain of control monkeys. Insets at higher magnification of Aldh1a1+ and Girk2+ cells. c Percentages of expression of TH + /Aldh1a1+ in the control monkey midbrains (n = 2; 6661 ± 776 neurons). d Percentages of expression of Aldh1a1 + /Cb+ in the control monkey midbrains (n = 2; 2736 ± 365 neurons). e Percentages of expression of Aldh1a1 + /Girk2+ in the control monkey midbrains (n = 2; 3182 ± 459 neurons). f, g Representative images of double immunohistochemically labeled neurons for Aldh1a1 and TH in the midbrain of a control f and a parkinsonian g monkey at the level of the third cranial nerve. h, i Quantification of TH + /Aldh1a1+ and TH + /Alh1a1- neurons in control and MPTP-treated monkeys (n = 2/group; MPTP n = 6). j Percentages of TH + /Aldh1a1+ and TH + /Aldh1a1- neurons in symptomatic monkeys (Rec+Park; n = 4) with respect to the control group (n = 2). Data represent mean + SEM. Symbols indicate significant differences compared with control *p < 0.05. Ctrl, control; Asym, asymptomatic; Rec, recovered; Park, parkinsonian. SNc: Substantia nigra pars compacta; SNr: Substantia nigra pars reticulata; VTA: Ventral Tegmental Area; D: dorsal; V: ventral; M: medial; L: lateral. Scale bar: 1 mm (inset 50 μm).

Full size image

Aldh1a1+ staining was strikingly evident in dendritic processes with long trajectories arising from specific clusters of neurons: these dendrites extend dorsoventrally in the SNr (Fig. 5a, b). This was most evident in rostral sections (Supplementary Fig. 6). These long dendritic processes were also Girk2+ (Fig. 5c) and had a clear trend to be grouped into thick bundles that overlapped with dense cannabinoid receptor type 1 (CBR1) expressing fibers extending dorsally within the SNr, likely representing so-called dendron bouquets32 (Fig. 5d; Supplementary Fig. 10c). CBR1-immunoreactive bundles climbed along Aldh1a1+ dendrites and neurons (Fig. 5d; Supplementary Fig. 10c). Within the SNr, the CBR1 immunoreactive bundles did not overlap with Cb neuropil immunoreactivity (Fig. 5e; Supplementary Fig. 10d).

Fig. 5: Dendritic processes with long trajectories extend dorsoventrally in the SNr.
figure 5

a Representative low-power images of TH + /Aldh1a1+ clusters of SNc ventral neurons and resultant ventrally extended dendrites in the SNr of a control monkey. a’ Note the presence of striking clusters of neurons in the SNc. b Higher power images of TH + /Aldh1a1+ dendrites extending into the SNr. c High-power images of Aldh1a1 + /Girk2+ dendrites extending into the SNr. d High-power Image of SNr CBR1-labelled dendrites extending dorsally into the adjacent sector of the SNc concurring with Aldh1a1+ ventrally extending dendrites. e High-power image showing the apparent dendron structures in the control monkey midbrain. Calbindin labels the fibers in the SNr and avoids the SNc and CBR1+ axons. For all low-power analogous images see Supplementary Fig. 10. Abbreviations: SNc: Substantia nigra pars compacta; SNr: Substantia nigra pars reticulata; VTA: Ventral Tegmental Area; D: dorsal; V: ventral; M: medial; L: lateral. Scale bar: 1 mm (insets 250 μm).

Full size image

To evaluate to what extent Aldh1a1 expression may be linked to dopaminergic cell vulnerability, the numbers of TH + /Aldh1a1+ and TH + /Aldh1a1- neurons were quantified in MPTP-treated macaques (Fig. 4f, g). An evident massive loss of TH + /Aldh1a1+ neurons was observed. This phenotype was significantly more susceptible to degeneration than TH + /Aldh1a1- neurons (Fig. 4h). A gradual TH + /Aldh1a1+ neuronal loss occurred in parallel to the development of parkinsonism, whereas no major differences were found in terms of TH + /Aldh1a1- cell loss across the different MPTP-treated groups (Fig. 4i). Interestingly, the parkinsonian motor phenotype was observed when the TH + /Aldh1a1+ neuronal cell loss was more than 60%. Moreover, in these monkeys there was a clear difference between the TH + /Aldh1a1+ (60%) vs TH + /Aldh1a1- (40%) cell loss (Fig. 4j). Remarkably, monkeys that exhibited net and stable parkinsonism showed almost complete loss of TH neurons and processes (Fig. 6).

Fig. 6: Loss of TH+ neurons and their dendrites located in the vSNc in parkinsonian monkeys.
figure 6

Representative confocal images of midbrain coronal sections at the level of the third cranial nerve immunostained for Cb and TH in Control a, Asymptomatic b, Recovered c, and Parkinsonian d monkeys. White arrows and insets show representative ventral dendrites. The greatest reduction of ventral dendrites occurs in monkeys with net and stable parkinsonism. D: dorsal; V: ventral; M: medial; L: lateral. Scale bar: 1 mm (inset a: 100 μm; inset b: 100 μm; inset c: 200 μm; inset d: 200 μm).

Full size image

Aldh1a1+ neurons define the nigrostriatal dopaminergic projection

Two injections into the caudate, three injections into the putamen, and two injections in the globus pallidus were selected for analysis after confirming the accurate placement of the tracers (Supplementary Fig. 11a–e). Neurons containing the tracer injected in the caudate and putamen were particularly abundant in the vSNc while neurons containing the tracer delivered in the globus pallidus appeared rather uniformly scattered in the SNc and VTA.

Among the neurons innervating the caudate nucleus and the putamen, 65% and 88% of cholera toxin subunit B (CTB) labeled neurons were Aldh1a1 + , respectively (Fig. 7a, b). By contrast, only a minority of nigropallidal-projecting neurons express Aldh1a1 (21% for the external division of the pallidus (GPe) and 3% for the internal division of the globus pallidus (GPi)) (Fig. 7c, d, g). To further determine if this distinct origin was associated with differential vulnerability, we compared the dopaminergic innervation of the striatum and the GPe and GPi in the same MPTP-treated monkeys. Dopaminergic projection was more profuse in the GPi than in the GPe in control animals. Although a marked reduction on TH+ terminals was observed in caudate and putamen nuclei (up to 87% of loss of dopaminergic terminals in the parkinsonian group) (Supplementary Fig. 1), both the length and density of TH+ axons remained unchanged at the level of the GPe or GPi nuclei (Fig. 7e, f, h, i).

Fig. 7: Aldh1a1+ neurons define the nigrostriatal dopaminergic projection.
figure 7

a Representative confocal images showing the expression of cholera toxin subunit B (CTB)-labeled midbrain neurons co-expressing Aldh1a1 after striatal tracer injection. White arrows show abundant Aldh1a1 + /CTB+ neurons. b Percentages of expression of Aldh1a1 + /CTB+ neurons in the monkey’s midbrain with injections in the caudate (n = 2; 78 neurons) and putamen respectively (n = 3; 623 neurons). c Representative confocal images showing the expression of CTB-labeled midbrain neurons expressing Aldh1a1 after globus pallidus injection. White arrows show abundant Aldh1a1-/CTB+ neurons. d Percentages of expression of Aldh1a1 + /CTB+ neurons in the monkey’s midbrain with injection in the external division if the globus pallidus (GPe) (n = 1; 127 neurons). e, f Quantification of TH-immunoreactive axonal length in the GPe in control and MPTP-treated monkeys (n = 2/group; MPTP n = 6). g Percentages of expression of Aldh1a1 + /CTB+ neurons in the monkey’s midbrain with injection in the internal division if the globus pallidus (GPi) (n = 1; 139 neurons). h, i Quantification of TH-immunoreactive axonal length in the GPi in control and MPTP-treated monkeys (n = 2/group; MPTP n = 6). Data represent mean + SEM. Ctrl, control; Asym, asymptomatic; Rec, recovered; Park, parkinsonian.

Full size image

Discussion

In this study, we have identified dopaminergic neuron subtypes in the SNc of the NHP brain and defined their distinct anatomical distribution, projection patterns, and differential vulnerability to MPTP toxicity. We demonstrate that Aldh1a1+ and Girk2+ neurons are highly enriched and intermingled in the vSNc, where cellular vulnerability is highest, in contrast with the more resilient dorsal neurons that express Cb. Aldh1a1+ neurons are the main source of the vulnerable dopaminergic nigrostriatal projection and receive CBR1 afferents. The loss of Aldh1a1+ neurons in the vSNc territory is in keeping with the onset of parkinsonism in our MPTP monkey model. Bearing in mind that Aldh1a1+ neurons were characterized as the dopaminergic cell phenotype predominantly affected in Parkinson´s disease patients, these findings support and expand the evidence of subtype-specific vulnerabilities underlying the pathophysiology of Parkinson´s disease.

Following progressive MPTP-driven toxicity, dopaminergic neuronal loss was consistently greater in the vSNc-nigrosome region than in the dorsomedial matrix region as described in Parkinson´s disease patients4. Of note, the Aldh1a1 + /Girk2+ and Cb+ subpopulations match fairly well with the nigrosome-matrix concept, respectively, in terms of anatomic localization and vulnerability13. They also align with the SOX6-Cb axis recently reported in humans17, macaques18 and mice33. Our findings confirm the relative sparing of dopaminergic neurons that express Cb in the SNc and the VTA34,35,36,37,38, while Aldh1a1 and Girk2 best correlate with the distribution of vulnerable neurons33,39,40,41,42. Several studies in rodents and humans have reported that Girk2 and Cb are not co-expressed in dopaminergic neurons and that expression of Girk2 can discriminate between vulnerable and resilient dopaminergic neurons in the vSNc and the dSNc respectively43,44,45,46,47. However, comparable information in monkeys is still very limited18,48. Our results indicate that Cb and Girk2 are expressed in different subpopulations and that vulnerable dopaminergic neurons within the vSNc express high levels of Girk2. Moreover, Girk2 expression follows a clear dorsoventral and mediolateral gradient, with the dSNc and VTA containing less Girk2 as described in mice and humans43,44,49.

Aldh1a1+ neurons appear to be a subpopulation of Girk2+ neurons that form marked clusters which are highly vulnerable to MPTP toxicity, highlighting a high inter-species analogy and specificity40,42. Interestingly, the group of Aldh1a1+ neurons located in the VTA expressed very low or no levels of Girk2. This subpopulation of VTA Aldh1a1 neurons is not especially vulnerable in Parkinson´s disease patients39, MPTP-treated mice41, or seemingly in our MPTP monkey model (Fig. 4g), suggesting that the sole expression of Aldh1a1 is not the cause of vSNc vulnerability. Indeed, Aldh1a1+ neurons in mice can be further divided into different sub-populations with additional functional heterogeneity50,51. Actually, a specific subpopulation of Aldh1a1+ neurons in the SNc that express angiotensin II receptor type 1_(AGTR1) is highly susceptible to neurodegeneration in Parkinson´s disease patients17.

Our data support and expand some phenotypic features that define highly vulnerable SNc neurons. Thus, those that are Cb-/Girk2 + /Aldh1a1+ exhibit the highest vulnerability. Several other cellular markers of vulnerability have been described in the midbrain11,13. To what extent these factors play a pathogenic role in the origin and mechanisms of neurodegeneration or are mere bystanders is not defined. Certainly, Girk2 and Aldh1a1 play an important role in functions that are altered in Parkinson´s disease patients such as calcium homeostasis, cholesterol modulation, and dopamine metabolism52,53,54,55. These neurons (but not those that are Cb + ) present autonomous pacemaking activity based on L-type voltage-dependent Ca2+ channels and show auto-receptor dependent regulation of dopamine release56,57. Importantly, Aldh1a1 is not expressed in any other catecholaminergic group42,58,59, and other catecholaminergic neurons do not express Girk2 or only exhibit low levels. Thus, Girk2+/Aldh1a1+ neurons identify a unique dopaminergic subtype within the vSNc that are especially relevant in Parkinson´s disease. However, it is noteworthy that NHPs (or any other animal) do not develop a Parkinson´s disease-like neurodegeneration spontaneously. Accordingly, these vulnerability factors as well as other putative mechanisms may be necessary but not sufficient for nigrostriatal degeneration.

Our data show that a large proportion of the highly vulnerable nigroputaminal projection neurons express Aldh1a1 (88%). On the other hand, the majority of nigropallidal dopaminergic neurons were Aldh1a1- and were relatively spared. The small proportion of Aldh1a1 neurons projecting to the pallidum could correspond to the nigrostriatal collaterals that arborize in the pallidum and subthalamic nucleus60,61. This is consistent with previous studies that have suggested nigrostriatal and nigro-extrastriatal projections arise from different groups of midbrain dopaminergic neurons31,36,62,63,64 and show differential vulnerability, with the nigropallidal projection being much less affected62,64,65. In mice, Aldh1a1+ and Girk2+ neurons project principally and massively to the dorsolateral striatum, while Cb+ projections are biased toward more ventromedial striatal regions and extra-striatal regions42,45,47,66,67. The differential susceptibility to neurodegeneration of striatal and extra-striatal projections suggests that the projection pattern of diverse populations of dopaminergic neurons can be a relevant factor determining their vulnerability11,67,68. Likewise, different Aldh1a1+ subpopulations may have topographically different projections through the striatum, linking together vulnerable regions of the nigrostriatal system, both at origin and at destination42,66,67. Interestingly, Aldh1a1 is responsible for degradation of DOPAL, which triggers alpha-synuclein mediated neurotoxicity at the synaptic level, leading to neurodegeneration69. In the near future, further studies defining subpopulations of Aldh1a1+ and Girk2+ dopaminergic neurons and their connectivity will probably unravel further microcircuits more specifically linked to the most vulnerable SNc neurons. This in turn could facilitate the development of specific neurorestorative/protective therapies.

We have shown here that Girk2 is abundant in dopaminergic processes in monkeys, and more importantly, that the bundles of long dendrites that extend ventrally into the SNr region emerge specifically from clusters of Aldh1a1+ neurons and coincide with dense CBR1-expressing fibers that closely resemble the so-called dendron-bouquets in mice32,70,71,72. Interestingly, stable parkinsonism in our monkeys was paralleled with an almost complete loss of ventral neurons and dendrites. This concurs with recent data in mouse models which showed how overt parkinsonian features appeared not only in association with striatal dopamine depletion but with the onset of somatodendritic SNc dopamine release impairment73,74,75. Indeed, extensive Aldh1a1+ neuron loss coincided with the onset of motor manifestations in our MPTP monkey model, in keeping with their selective motor related activity42,51. We speculate that loss of Aldh1a1 neurons may lead to presynaptic dysregulation of nuclei that compensate and define basal ganglia neuronal output76,77,78,79,80. This could also involve a role for CBR1s in modulating the function of these clusters of Aldh1a1+ dopaminergic neurons70,78,81,82. Of relevance, CBR1 in the SNr is confined to striatonigral and subthalamonigral terminals83,84,85 and the major source of excitatory input to Aldh1a1+ neurons comes from the subthalamic nucleus42. Additionally, the vast gabaergic, glutamatergic, and cholinergic inputs to these neurons32,86,87 could make them more vulnerable or prone to energy failure88,89.

In sum, the available data indicate that Aldh1a1+ neurons may play a critically relevant function in dopamine release in both the substantia nigra and the striatum and, accordingly, in the regulation of motor functions and motivated behavior90,91. Moreover, the refined mechanisms that regulate vSNc neuron activity intuitively suggest a tight, possibly causal, relationship with their exquisite vulnerability for neurodegeneration in Parkinson´s disease.

The relatively low number of animals used here may be viewed as a potential limitation. However, the results were consistently comparable across monkey groups, showing minimal group variations in identified neuronal subpopulations, distribution, and intracellular location. Also, we had to optimize the study to minimize the number of monkeys used (e.g., opting for the lowest possible count and only males) due to the current worldwide shortage92 and the general trend to reduce experiments involving NHPs.

Unfortunately, we failed to generate reproducible and consistent immunohistochemical staining for other cellular markers of vulnerability with proven performance in rodents. This is open to further testing in the future. It would have been desirable to further define molecularly different subpopulations of Aldh1a1+ and Girk2+ dopaminergic neurons for more in-depth circuit studies51,90. Similarly, smaller tracer injections targeting more specific rostral, medial, and caudal territories of the caudate and putamen will be needed.

In conclusion, we have validated and expanded to the primate brain previous findings in Parkinson´s disease rodent models by describing in detail that Girk2 and Aldh1a1 are crucial markers sustaining cell-specific vulnerability of dopaminergic neurons. Additional efforts are required to establish detailed projection patterns for precisely defined dopaminergic cluster subtypes. Integrating connectivity studies, a more precise characterization of Aldh1a1+ or Girk2+ dopaminergic neuron subpopulations, and thorough circuit investigations could play a key role in uncovering the mechanisms supporting neuronal vulnerability in Parkinson´s disease. This could, in turn, guide development of new therapeutic approaches, including subtype-specific pharmacological interventions.

Methods

Experimental animals

Thirteen adult naive male Macaca fascicularis NHPs were used in this study (R.C. Hartelust BV, Tilburg, The Netherlands). The monkeys were housed in the animal vivarium under standard conditions and handled in full keeping with European and Spanish animal care regulations (2010/63/UE and RD53/2013, respectively). The experimental protocols were approved by the Ethical Committee for Research of the Fundación de Investigación HM Hospitales (CEEA-01/2015; CEEA-06/2015) as well as by the department of Animal Welfare of the Government of Madrid (PROEX 269/2015; PROEX 040/2016).

Two monkeys were used as controls (Control group) and received systemic administration (i.v.) of saline (NaCl 0.9%, vehicle of MPTP solution) every 2 weeks, whereas six monkeys were treated with MPTP (Sigma, Saint Louis, MO, USA) by systemic intravenous administration. under light anesthesia (ketamine 10 mg/kg; i.m.) using a dose regimen of 0.5 mg/kg every 2 weeks to obtain partial and slowly-progressive degeneration of the nigrostriatal dopaminergic system as described previously24,25,93,94,95. Motor status was assessed with a clinical-rating validated scale96. As each monkey has a different susceptibility to MPTP97, the number of injections varied depending on the response to the toxin (i.e. individual sensitivity) until the motor score was reached. Two monkeys appeared to be more susceptible and displayed an evident and persistent parkinsonian syndrome after the first MPTP injection and therefore did not receive any further injection; these two monkeys form the Parkinsonian group. Two monkeys required two injections of MPTP to exhibit a clear motor phenotype, however they later showed substantial recovery and remained minimally affected (Recovered group)24,98. The Asymptomatic group was composed of two monkeys that did not show any motor phenotype even after three MPTP injections. All monkey records are shown in Supplementary Table 1. Throughout the duration of the study, none of the monkeys received dopaminergic medication. Analysis was performed on all MPTP-treated monkeys, both grouped together (n = 6) and per group (n = 2/group).

The retrograde neuroanatomical tracer unconjugated CTB (List Biological Laboratories, Campbell, CA) was pressure-delivered into the caudate nucleus of one hemisphere and the putamen of the contralateral hemisphere in four monkeys (bilateral injections); whereas in one additional monkey, CTB was delivered into the GPi of one hemisphere and into the GPe of the contralateral hemisphere.

Surgical anesthesia was induced by intramuscular injection of ketamine (5 mg/kg) and midazolam (0.5 mg/kg). Local anesthesia was implemented just before surgery with a 10% solution of lidocaine. Analgesia was achieved with a single intramuscular injection of flunixin meglumine (Finadyne®, 5 mg/kg) delivered at the end of the surgical procedure and repeated 24 and 48 h post-surgery. A similar schedule was followed for antibiotic coverage (ampicillin, 0.5 ml/day). After surgery, monkeys were kept under constant monitoring in individual cages with ad libitum access to food and water. Once they showed complete post-surgical recovery (24 h), they were returned to the animal vivarium and housed in groups for a follow-up of two weeks until final sacrifice.

Stereotaxic coordinates for the caudate and putamen were calculated from the atlas of Szabo and Cowan99. Coordinates for the caudate nucleus were 20 mm rostral to the interaural plane, 12 mm ventral to the cortical surface, and 5 mm lateral to the midline. Coordinates for the pre-commissural putamen were 20 mm rostral to the interaural plane, 16 mm ventral to the cortical surface and 19 mm lateral from the midline, whereas post-commissural putaminal injections were performed 13 mm rostral to the interaural plane, 15 mm ventral to the cortical surface and 12 mm lateral to the midline. Stereotaxic coordinates for GPe and GPi were calculated from the atlas of Lanciego and Vazquez100. Coordinates for the GPi were 4 mm caudal to the anterior commissure (ac), 1 mm ventral to the ac-posterior commissure (ac-pc) plane and 13 mm lateral to the midline; coordinates for the GPe were 4 mm caudal to ac, 1.5 mm dorsal to ac-pc plane and 14 mm lateral to the midline (Supplementary Fig. 11f, g). CTB deliveries were performed through a Hamilton® microsyringe in pulses of 0.1–1 μl/2 min for a total volume of 100 nl–5 μl/injection (5 mg/ml in 0.01 M PB, pH 7.5). Once completed, the needle was left in place for 15 min before withdrawal to minimize CTB reflux through the injection tract.

Necropsy and histological processing

Monkeys were deeply anesthetized with a terminal dose of sodium pentobarbital (200 mg/kg, i.p.) and perfused through the ascending aorta with an infusion pump. The perfusates consisted of a saline solution, followed by 4% paraformaldehyde in phosphate buffer and an ascending series of buffered sucrose solutions (5–10–20%). The brains were blocked in the coronal stereotaxic plane (plane 0 corresponding to the interaural plane) and sectioned (40 μm-thick sections) on a freezing microtome.

For detection of the different dopaminergic neuron subpopulations, combinations of double-labeling experiments were performed. Free-floating coronal sections were washed in tris-buffer and treated with citrate buffer (pH 6) for 30 min at 37 °C for antigen retrieval. Normal serum was applied for 3 h to block non-specific binding sites. The sections were incubated at 4 °C for 72 h with primary antibodies directed against TH, Cb, Girk2, Alhd1a1, CBR1, and CTB. The sections were washed in tris-buffered solution and transferred for 2 h to a solution containing the corresponding secondary fluorescence antibodies.

For optical density quantification in the striatum and stereological counting in the GPe and GPi, free-floating immunohistochemistry was performed using a primary monoclonal anti-TH mouse antibody and a secondary biotinylated polyclonal anti-mouse antibody. Development of the immunoreaction was done using diaminobenzidine as substrate.

The sections were then dehydrated through graded ascending ethanol solutions, cleared in xylene, and cover slipped with DPX. All sections were processed simultaneously in all monkeys. For control purposes, all immunohistochemical stains included sections incubated without the corresponding primary antibody.

A complete list of antibodies, together with incubation concentrations and commercial sources is provided in Table 1.

Table 1 List of antibodies
Full size table

Assessment of terminal densities in the striatum

The relative optical densities of TH+ fibers in the caudate nucleus, putamen, and ventral striatum were quantified in the diaminobenzidine developed sections using computer-assisted image analysis techniques (ImageJ 1.41o, National Institutes of Health, USA). Images were captured in black and white 8-bit monochrome using a digital camera (AxioCam HRc, Zeiss, Germany) attached to a Nikon Multiphot microphotography system. Digital images were captured under the same exposure settings for all experimental cases. Seven rostrocaudal sections, equally spaced at intervals of 2400 μm, were examined for each monkey. Three sections were more rostral, and four sections were more caudal to the midline decussation of the anterior commissure. The optical density of a 0.080 μm2 region of white matter in the corpus callosum in the same section was subtracted as background. To quantify the dorsoventral dopaminergic gradient of the striatum, both the caudate and putamen were delineated into three equivalent dorsal, intermediate, and ventral subdivisions.

Midbrain sections sampled and neuronal quantification

Five to seven coronal sections, regularly spaced 1200 μm apart, covering the whole rostrocaudal extent of the midbrain were scanned under the microscope. Immunofluorescence sections were inspected using a Leica SPI TCS SP5 confocal laser-scanning microscope. To ensure appropriate visualization of the labeled elements and to avoid false positive results, the emission from the argon laser at 488 nm was filtered through a band pass filter of 505–530 nm and color-coded in green. The emission following excitation with the helium laser at 543 nm was filtered through a band pass filter of 560–615 nm and color-coded in red. Due to the large size of the region of interest, automated images were taken to obtain a mosaic of each midbrain with a 40× objective, where each quadrant was acquired in 5 sections in the z-axis with a resolution of 512 × 512 ppi. Co-localization and quantification were analyzed by blinded researchers for each marker using ImageJ (2.1.0/1.53c) software in 5 to 7 sections.

To investigate the expression of Girk2 and Aldh1a1 in the different catecholaminergic populations (TH + ), sections were selected based on Sanchez-Gonzalez et al. 101, including the olfactory bulb and the locus coeruleus.

Stereological quantification of TH axons in the globus pallidus

The density of axons immunoreactive for TH was estimated in the GPe and GPi by means of the optical fractionator method by a blinded researcher using 3D hemispherical probes (StereoInvestigator® software, MicroBrightField, USA) as previously described95,102. Three TH immunostained coronal sections regularly spaced 2,400 μm apart were studied in each monkey. Virtual hemispheres (radius = 10 μm) were projected along the Z-axis, spaced in a sampling grid of 250 × 250 μm. A guard-zone 2 μm high was used on the top of the section. These parameters were set to reach an error coefficient below 0.10 (Gundersen, m = 0) and 0.05 (Gundersen, m = 1). The intersections of TH axons with the boundaries of the hemispheres, observed with a 100× immersion objective, were quantified. Total axonal length was calculated. To correct for potential differences in brain sizes, the volume of each nucleus was calculated by the Cavalieri method103. Axonal length density was defined as the total axonal length per volume of each nucleus104.

Statistics and figures

Statistics were performed between data from the control group and the MPTP-treated groups using the Statistica 12 software (StatSoft Inc., Dell Software). Data normality was checked with the Kolmogorov–Smirnov test. Differences of neuronal numbers and axonal length density among groups were analyzed by one-way repeated measure analysis of variance (ANOVA) followed by the Bonferroni post-hoc test. When comparing the means of the control group with the MPTP-treated group, a two-sample non-parametric unpaired t-test (Mann-Whitney U test) was performed. All data are expressed as the mean ± standard error of the mean (SEM). The significance threshold was P < 0.05. Graphs were made with GraphPad Prism 6.01 for Windows (GraphPad® Software).

Images for figures were acquired on a Leica MICA microscope at 20× (high-power) and 63× (insets) magnification. Figures were made with Inkscape 1.1.