Midbrain Gene Screening Identifies a New Mesoaccumbal Glutamatergic Pathway and a Marker for Dopamine Cells Neuroprotected in Parkinson’s Disease

The ventral tegmental area (VTA) and substantia nigra pars compacta (SNc) of the midbrain are associated with Parkinson’s disease (PD), schizophrenia, mood disorders and addiction. Based on the recently unraveled heterogeneity within the VTA and SNc, where glutamate, GABA and co-releasing neurons have been found to co-exist with the classical dopamine neurons, there is a compelling need for identification of gene expression patterns that represent this heterogeneity and that are of value for development of human therapies. Here, several unique gene expression patterns were identified in the mouse midbrain of which NeuroD6 and Grp were expressed within different dopaminergic subpopulations of the VTA, and TrpV1 within a small heterogeneous population. Optogenetics-coupled in vivo amperometry revealed a previously unknown glutamatergic mesoaccumbal pathway characterized by TrpV1-Cre-expression. Human GRP was strongly detected in non-melanized dopaminergic neurons within the SNc of both control and PD brains, suggesting GRP as a marker for neuroprotected neurons in PD. This study thus unravels markers for distinct subpopulations of neurons within the mouse and human midbrain, defines unique anatomical subregions within the VTA and exposes an entirely new glutamatergic pathway. Finally, both TRPV1 and GRP are implied in midbrain physiology of importance to neurological and neuropsychiatric disorders.


Results
Unbiased gene-screening identifies 8 genes of interest as putative VTA or SNc markers. To compare gene expression profiles between the VTA and the SNc in the newborn mouse, the direct red fluorescence obtained from the reporter protein tdTomato (tdTom) was used to grossly outline the outer borders of the VTA and SNc which were dissected from Dat-Cre tdTom pups at postnatal day 3 (P3) (Fig. 1a top). Whole cell RNA was prepared and analyzed for gene expression by the GeneChip ® Mouse Gene 1.0 ST Arrays; by using the entire cell material, a bias towards DA neurons was avoided which allowed composite gene expression analyses in all cell types. A total of 28270 genes were analyzed (Supplementary Table S1) and by plotting adjusted p-value over expression difference, 21 genes displayed at least log2-fold higher expression level in the VTA and 36 in the SNc ( Fig. 1a and Supplementary Table S2). P-values were generally high so in order to sort the results, the top 50 genes according to p-value were chosen and thereafter sorted by log2-fold change (Fig. 1b bottom and c).
The histological screening revealed that neither Chrna2, Gsbs, Mc3r, Crisp1, Gpr83, Nmu, Zic2 nor Serpinf1 were detected in either the VTA or the SNc at P3 (Fig. 1d and Supplementary Fig. S1) and were therefore excluded from further analysis. In contrast, Six3 expression could be found in the SNc area while Grp, TrpV1, Nrp2, Ntf3, Tacr3, Calb1 and NeuroD6 genes were absent from the SNc, but showed various expression patterns within the VTA: Grp mRNA was found in the ventromedial VTA; TrpV1 mRNA was even more medially restricted than Grp mRNA; Ntf3, Tacr3 and Calb1 mRNAs appeared more caudal and lateral than Grp, while Fst and NeuroD6 were similar but not identical to Grp mRNA within the VTA ( Fig. 1d and Supplementary Fig. S1). The patterns of these 8 genes were of interest to map with anatomical detail and next, riboprobe high-resolution in situ hybridization was performed.

Identification of Calb1, Grp, Lipoprotein lipase (Lpl), NeuroD6
and TrpV1 gene expressions as restricted to VTA subnuclei. Overlays of images from high-resolution in situ hybridization of the 8 mRNAs (Grp, TrpV1, Nrp2, Ntf3, Tacr3, Calb1 and NeuroD6) with adjacent sections showing Th mRNA, used as a reference, guided the mapping in the adult midbrain ( Fig. 2) according to previous publications [44][45][46] . The Six3 riboprobe appeared non-selective and was therefore excluded. All other expressions are shown in Fig. 2a,b and summarized in Table 1. The VTA was structured into the following subnuclei: the parabrachial pigmented nucleus (PBP) and VTA rostral nucleus (VTAR) (PBP/VTAR; corresponds to only PBP in studies prior to Fu et al. 44 ); the paranigral nucleus (PN) and parainterfascicular nucleus (PIF) (PN/PIF; corresponds to only PN prior to Fu et al. 44 ); the interfascicular nucleus (IF) and the rostral linear nucleus (RLi). The caudal linear nucleus (CLi) was not analyzed while the ventrally located supramammillary nucleus (SuM) was used as a ventral reference structure. Strong Th gene expression was as expected found in the lateral areas corresponding to PBP/VTAR, PN/PIF and SNc, while a weaker expression was observed in the medial IF and RLi and was absent as expected from the GABA-rich SN pars reticulata (SNr). (a) Tissue obtained from P3 Dat-Cre tdTom mice (n = 3) was separated into VTA, SNc and RLi (top) and VTA and SNc subjected to microarray analysis. Scatterplot of genes analyzed in the microarray, plotted by expression differences and adjusted p-value (bottom). Positive values represent genes upregulated in the VTA, negative values upregulation in the SNc. (b) Overview of the 50 genes with highest adjusted p-value, sorted according to difference in expression in VTA and SNc. Genes in dark blue were used for further analysis. (c) Heatmap of genes with at least log2-fold difference between VTA and SNc (n = 3) (top) and within the individual samples (bottom). (d) Representative images of oligo in situ hybridization for different genes in the ventral midbrain of P3 C57BL/6 mice with putative borders of the Th-positive cell population (dotted lines) as guidance. Rows one to three and first three images in row four represent the caudal midbrain; last two images in row four depict more rostral levels. Row five represents overlays of candidate genes (blue) with adjacent Th slide (red) from the same mouse. Schematic illustrations were adapted from Paxinos, G. & Franklin, K. B. J. 42  Girk2 and Lpl, previously identified in microarray screens as elevated in the SNc over the VTA, and vice versa 6-8 , were included for histological validation. Girk2 mRNA appeared in all Th-expressing areas of the VTA and SNc, stronger laterally than medially, but not exclusive for SNc. Lpl mRNA was not detected in either the Riboprobe in situ hybridization analysis of candidate genes in adult C57BL/6 mice; row 2 onwards-expression pattern of adjacent Th slide in the same animal as red overlay (reduced intensity in red channel for visualization purposes) over expression pattern of the analyzed gene. All images were cropped from whole slides acquired with a digital high resolution slide scanner. (a'',b''') Schematic visualization of expression patterns of the candidate genes. Schematic illustrations were adapted from Paxinos, G. & Franklin, K. B. J. 42  SNc or the adjacent SNr but was restricted to the VTA, and was strongest within the PN/PIF area. The Calb2 gene was included for comparison with Calb1. In contrast to a previous study localizing protein distribution 16 , Calb1 and Calb2 mRNA showed almost opposite mapping patterns: Calb1 mRNA was most prominent in the medially located IF but was also detected within the laterally located PBP/VTAR and PN/PIF areas while Calb2 mRNA was strong in the RLi and lateral VTA as well as SNc and almost excluded from the IF. Similar to Calb2, the Ntf3, Tacr3 and Fst genes were also excluded, or almost so, from the IF but were present at various degrees in the laterally located PBP/VTAR and PN/PIF; Ntf3 was broadly expressed throughout the SNr and SNc, while Fst mRNA was very weakly detected in the SNr and Tacr3 mRNA in the SNc.
Three genes, namely NeuroD6, Grp and TrpV1, in addition to Calb1 and Lpl described above, showed expression exclusively within subnuclei of the VTA and were not detected at all in either SNr or SNc. Of these, NeuroD6 mRNA appeared as weak as Lpl mRNA in the IF but was detected in the PBP/VTAR. Both NeuroD6 and Lpl mRNAs were at their strongest in the PN/PIF area. An almost opposite pattern was seen with Grp which was expressed mostly within the IF and less in PN/PIF and PBP/VTAR areas. TrpV1 mRNA was detected in only few cells distributed within the IF and PN/PIF. Taken together, these histological analyses demonstrate several new expression patterns within the analyzed midbrain areas. Of these, five patterns are exclusive for subnuclei within the VTA: NeuroD6 and Lpl mRNAs are stronger in the lateral VTA while Calb1, TrpV1 and Grp mRNAs are stronger in the medial VTA of the adult mouse. Calb1, Grp, NeuroD6, Ntf3 and Tacr3 expression mainly in dopaminergic neurons. To validate the neurotransmitter identity of the VTA cells, mRNA-selective double-fluorescent in situ hybridization (sdFISH) analysis of Calb1, Grp, NeuroD6, Ntf3, Tacr3, and Trpv1 mRNAs was performed. Probes towards Th and Dat mRNAs were included for detection of dopaminergic neurons, Vglut2 mRNA for glutamatergic neurons and the Vesicular inhibitory amino acid transporter (Viaat) mRNA for inhibitory neurons. TrpV1 mRNA analysis gave no conclusive results due to the scarcity of positive cells, but Calb1 (Fig. 3a), Tacr3 (Fig. 3b), Grp (Fig. 3c), NeuroD6 (Fig. 3d,e) mRNA all co-localized with Th and/or Dat mRNA with no or very little co-localization with either Vglut2 or Viaat. Ntf3 mRNA co-localized mainly with Dat but some overlap with Viaat mRNA was also seen (Fig. 3e,f). Comparing Th and Dat mRNA confirmed previous reports 37,47,48 of higher level of Dat mRNA within laterally distributed Th-expressing cells of the VTA and SNc (Fig. 3d). Accordingly, the laterally expressed NeuroD6 gene co-localized entirely with both Th and Dat mRNA, Tacr3 and Ntf3 mRNA overlapped substantially with both, and the medially expressed Grp gene overlapped only partially with Dat mRNA (Fig. 3d,e). NeuroD6 and Ntf3 mRNA also showed some overlap with each other while Grp and Calb1 mRNA showed very little overlap (Fig. 3f). Overall, the Grp, Calb1, Tacr3, Ntf3 and NeuroD6 genes were demonstrated as mainly expressed in DA cells, with only Nft3 mRNA as a putative marker for some GABA or GABA/DA co-releasing cells, while none of the mRNAs co-localized with Vglut2 mRNA.
TrpV1-Cre cells in the VTA show low I h -currents and are mostly glutamatergic. As TrpV1 gene expression was strong in the medial VTA at P3 but only detected in few cells in the adult VTA, this gene was likely identified in the current microarray screen as a result of the approach taken; non-DA-biased and performed in the newborn mouse, as opposed to previous DA-selective approaches performed in adult mice, none of which reported TrpV1 6,7 . The TRPV1 molecule was originally identified as the capsaicin receptor and is strongly linked    to endocannabinoid, heat and pain signaling 49,50 . Detection of TrpV1 mRNA within the VTA was therefore of particular interest and a TrpV1-Cre transgenic mouse tool 51 was selected for further analyses of this discrete VTA-population. sdFISH analysis on midbrain sections derived from newborn TrpV1-Cre tdTom mice showed that mRNA of the tdTom reporter was confined within the VTA while it was absent from the SNc, i.e. expression of the TrpV1-Cre transgene represented a similar spatial distribution pattern as the endogenous TrpV1 mRNA (Fig. 4a). Within the VTA, both endogenous TrpV1 and tdTom mRNAs were confined to scattered cells of the IF and PN/PIF areas, but tdTom mRNA was also detected within the RLi and in the medial aspect of the PBP/VTAR. Quantification revealed that 73% of all cells expressing the endogenous TrpV1 gene in the VTA co-expressed the tdTom-reporter gene, while only 36% of the tdTom-expressing cells in the VTA area co-expressed the endogenous TrpV1 gene (Fig. 4a).   Curiously, within the DA neuron-rich PBP area was a group of TrpV1-Cre-driven tdTom-expressing cells not co-expressing either Th or Dat, but that instead co-expressed the Vglut2 gene; this particular PBP subarea was here given the name "subzone of the PBP" (szPBP) to distinguish its strongly glutamatergic nature from the surrounding DA cells (Fig. 4b,f).
Further, while endogenous TrpV1 mRNA within the midbrain was almost uniquely detected within the medially located VTA subnuclei (IF and PN/PIF), tdTom mRNA was not only broader within the VTA (IF, PN/PIF, RLi, medial PBP/VTAR) but was also detected medially within the dorsal aspect of the midbrain, confined to the periaqueductal grey (PAG). In addition, tdTom mRNA was found in a more rostral location than the endogenous TrpV1 mRNA, and was, as also described for a different TrpV1-Cre transgene 52 , detected in the posterior hypothalamus (PH) (Fig. 4c).
The TrpV1-Cre activity in the VTA was analyzed in further detail. Medial VTA neurons have been ascribed an electrophysiological profile of low hyperpolarization (I h ) currents, as opposed to the higher currents displayed by lateral VTA neurons 27 . Basic electrophysiological properties of TrpV1-Cre positive cells in the VTA were therefore investigated by patch-clamp analyses in TrpV1-Cre tdTom mice. Under current-clamp conditions, the patched TrpV1-Cre tdTom cells, identified through direct red fluorescence, possessed a resting potential of − 54.12 ± 2.75 mV and responded to stimulation with fast spiking activity up to 19.13 ± 2.44 Hz at 200 pA injected current (n = 10) (Fig. 4d). Whole-cell voltage-clamp experiments revealed that 73% of the patched cells (n = 11) were I h -positive with a low-I h phenotype (32.88 ± 6.63 pA), while the remaining cells (n = 4) were I h -negative (Fig. 4d) and, thereby, the TrpV1-Cre tdTom neurons represented an electrophysiological phenotype typical of the medial VTA.
Next, sdFISH protocols were implemented to pinpoint the extent of neurotransmitter cell types present in the heterogeneous VTA area that are represented by the TrpV1-Cre transgene. Co-expression of mRNA from the TrpV1-Cre-driven floxed reporter gene tdTom with mRNA for Viaat for GABA-signaling neurons, Vglut2 for glutamatergic neurons and Th as well as Dat for dopaminergic neurons was analyzed ( Fig. 4e-h).
Many tdTom-expressing cells were found in a location rostral of the IF and caudal of the PH, an area devoid of name in current atlases and indicated to contain only the retromamillary decussation 46 . This area, which similarly to the new szPBP was characterized by cells showing a high degree of co-localization of the TrpV1-Cre-driven tdTom mRNA and Vglut2 mRNA, is from here onwards referred to as the rostromedial VTA (rmVTA) and was included in the quantification (Fig. 4e).
Equally selective as endogenous TrpV1 mRNA for VTA subnuclei, sdFISH demonstrated a mixed neurotransmitter phenotype of the TrpV1-Cre cells (Fig. 4f). Quantification was performed throughout all VTA subnuclei in both newborn and adult TrpV1-Cre mice ( Fig. 4g and Supplementary Table S5). Pooling all VTA subnuclei together, the majority (61.6 ± 3.0%) of the TrpV1-Cre-driven tdTom-cells in the adult mice co-expressed Vglut2, followed by co-expression with Viaat (22.8 ± 3.0%) Th (17.3 ± 2.4%) and Dat (6.5 ± 0.5%) (Fig. 4g) and similar results were found in the newborn mice (Supplementary Table S5). The lower co-expression with Dat than Th was in accordance with the more medial than lateral distribution of the TrpV1-Cre-driven tdTom-expressing cells within the VTA (Fig. 4f). When analyzing subnuclei separately, it was clear that the glutamatergic phenotype was prominent in the medial areas, including the IF, RLi, rmVTA and szPBP of the VTA as well as the SuM (Fig. 4h). Within the IF, PBP/VTAR and PN/PIF areas, the total amount of Th-and Vglut2-co-expressing tdTom-cells, respectively, added up to more than 100%, suggesting that some of these TrpV1-Cre-driven tdTom-cells thereby might be DA/glutamate co-releasing neurons. TrpV1-Cre expression therefore defines a mixture of glutamatergic, dopaminergic and GABA-signaling neurons within the VTA and possibly also represents various co-releasing neurons that co-exist within the VTA. The results further show that the majority of TrpV1-Cre cells are glutamatergic.
TrpV1-Cre cells project exclusively to the NAc shell within the striatal complex and release glutamate upon optogenetic stimulation. Based on the striking location of the TrpV1-Cre neurons exclusively within the VTA, their exclusion from the SNc and their strongly glutamatergic nature, it was of particular interest to functionally ascertain where these neurons project and if they there release measurable amounts of glutamate. To this end, a floxed AAV-ChR2-EYFP virus was injected into the VTA of TrpV1-Cre mice (Fig. 5a). Post-mortem histological analyses showed EYFP-positive cell bodies exclusively located within the medial part of the VTA of TrpV1-Cre::ChR2-EYFP mice (Fig. 5b). Thus, by implementing stereotactic virus injections, TrpV1-Cre transgenic mice can be used to selectively target the medial VTA neurons while leaving both lateral VTA and SNc structures as well as the dorsal PAG and rostral PH structures untargeted. In this way, the TrpV1-Cre transgene is spatially more restricted within the VTA than the Th-Cre, Dat-Cre, Vglut-Cre and Gad-Cre transgenic tools that have been used in previous optogenetic analyses of the VTA (reviewed in refs 17,42,43).
EYFP-positive fibers were sparse but possessed similar distributions as VGLUT2-positive terminals within the lateral septum and the NAc medial shell (NAcSh) while no EYFP fibers detected were in other parts of the NAc or the dorsal striatum which are target areas for dopaminergic neurons (Fig. 5c,e). To assess the functional properties of the newly identified pathway between the rostromedial VTA and the NAcSh, glutamate release was measured in the target area of TrpV1-Cre::ChR2-EYFP mice by in vivo amperometric recordings in which a glutamate-sensitive microelectrode was coupled in a complex with an optical fiber (Fig. 5d,e). Basal glutamate levels of 0.7 ± 0.3 μ M were determined in the medial NAcSh. In control animals lacking ChR2-expression in the VTA and NAc, no detectable glutamate release was present upon optic stimulation. TrpV1-Cre::ChR2-EYFP expressing mice on the other hand exhibited reliable glutamate release of 0.10 ± 0.04 μ M (n = 5 mice, average of five consecutive stimulations each) upon optical stimulations of 1 s at 40 Hz (Fig. 5f). In summary, by demonstrating that TrpV1-Cre cells of the rostromedial VTA project to the medial NAcSh region and there release glutamate upon activation, this study has identified the so far anatomically most restricted mesoaccumbal glutamatergic pathway.
Scientific RepoRts | 6:35203 | DOI: 10.1038/srep35203 GRP and TRPV1 expression patterns identified in the human SNc. Next, human tissue sections were analyzed using oligo in situ hybridization probes to human TRPV1, GRP, NTF3 alongside TH and VGLUT2. As expected, TH expression was found throughout the SN (subdivided into a matrix area surrounding the five nigrosomes; the Substantia nigra pars dorsalis (SNpd); Substantia nigra pars lateralis (SNpl)), A8 and the Medial (M) and Medioventral (Mv) areas, while no TH-expression was seen in the red nucleus (Fig. 6a top left). According to literature, the SNpd corresponds to the PBP in rodents, while Mv is similar to other parts of the VTA, including most prominently the IF and PN [53][54][55] . NTF3 mRNA was not detected at all in the human midbrain (Fig. 6a top right). In contrast, GRP expression was found mainly in areas corresponding to similar areas in the rodent: most expression was seen in the SNpd (mouse PBP) but also some in the A8 and Mv (Fig. 6a top middle left). TRPV1 and VGLUT2 showed similar expression patterns in the human midbrain and were both faintly detected in the matrix region and in A8 and Mv (mouse VTA) (Fig. 6a top middle right). Most overlap was seen between TH and GRP, possibly because these were the most strongly expressed genes (Fig. 6a bottom row). In the brains derived from PD patients, none or very little VGLUT2 or TRPV1 expression was detected (Fig. 6b, middle right and right). TH expression was considerately less detectable than in control brains and only remained  in few cells (Fig. 6b, left). Strikingly, GRP expression was almost as prominent as in the control brains (Fig. 6b,  middle left). This was curious as DA cells are the primary cell type to degenerate in PD, and for this reason, the GRP-expressing cells were analyzed in more detail.
GRP is mostly expressed in non-melanized neurons and are spared in PD. The cause of the selective death of SNc over VTA DA neurons in PD is far from understood, but several studies have proposed the involvement of the pigment neuromelanin (NM) as DA neurons with high amounts of NM are more susceptible to degeneration [54][55][56] . High-resolution cellular analysis, in which the NM can be directly seen due to its dark color, revealed that GRP in the healthy brain is most strongly expressed within non-melanized TH neurons (Fig. 7a) which show a more restricted distribution than melanized TH cells. The non-melanized GRP-expressing neurons were localized within the SNpd and A8 while TH, which was detected within both melanized and non-melanized neurons, was seen throughout the SN, A8 and M, MV regions (Fig. 7b). Some weak GRP expression was, however, also detected within few melanized neurons in the Mv. In PD brains, melanized cells were mostly lost, and hence most of the TH-expression. The remaining TH expression was seen in non-melanized cells and the few remaining melanized neurons (Fig. 7c, left and d). In brain sections obtained from PD cases of different range of pathology, GRP expression was exclusively detected in non-melanized neurons in the PD brain. Further, with almost equally strong intensity between control and PD brains, this shows that non-melanized neurons expressing GRP are spared in PD (Fig. 7c right and d). Taken together, the histological analyses of human post-mortem brains demonstrate that while TRPV1 and VGLUT2 expression is low in both control and PD midbrain, the strong GRP expression remains in the PD midbrain: GRP can thereby be used as a new and selective marker for the non-melanized DA neurons that are neuroprotected in PD.

Discussion
In contrast to the bias imposed towards DA neurons in previous expression analyses in rodents [6][7][8][9][10]16 , the current study implemented a crude mass-dissection approach of the VTA and SNc areas to obtain microarray-based gene expression data from all the different neuronal cell types that inhabit either the SNc or the VTA. Such gene expression patterns are essential in order to improve the current anatomical resolution in animal-based research centered around midbrain-disorders, but also to enhance the understanding of the etiology as well as to promote searches for new biomarkers of value for human therapies.
Summarizing the current findings, this study has shown that: (i) Calb1, TrpV1 and Grp gene expression patterns represent medial VTA subnuclei while NeuroD6 and Lpl represent lateral VTA subnuclei in the mouse; (ii) Grp, Calb1, Tacr3, Ntf3 and NeuroD6 are mainly expressed in various DAergic subpopulations while TrpV1-Cre-expressing neurons in the VTA show a mixed neurotransmitter phenotype, representative of the heterogeneous VTA; (iii) Co-localization of TrpV1-Cre-expression with the glutamatergic Vglut2 mRNA marker enabled the identification of two new subnuclei in the medial VTA, i.e. the szPBP and rmVTA; (iv) Optogenetics in TrpV1-Cre-active VTA neurons identified a new glutamatergic mesoaccumbal pathway, a finding which makes this Cre-driver the most selective transgenic mouse tool for pre-clinical targeting within the VTA; (v) Human TRPV1 was sparse in the midbrain but GRP was readily detected in non-melanized DA neurons within the SN: GRP expression persisted in PD where melanized DA neurons were lost, a finding which demonstrates GRP as a marker for cells that are neuroprotected in PD.
By taking the finding of TrpV1 expression within the rostromedial VTA at P3 through to functional transgenics-and optogenetics-based analyses in adult mice, the current study identified TrpV1 mRNA as spatially confined within the VTA and also exposed a group of VTA neurons, joined by their expression of the TrpV1-Cre transgene, with unique properties: This population of VTA neurons shows a restricted innervation pattern within the striatum exclusively to the NAcSh, a pattern which bears no resemblance in any other reported Cre-driver and further, although mainly glutamatergic, it is of mixed neurotransmitter phenotype, thus representing the heterogeneity of the VTA itself. Additionally, by characterizing the neurotransmitter identities of the TrpV1-Cre subpopulation within the VTA, it was possible to distinguish two novel subnuclei within the VTA that were characterized by Vglut2-and TrpV1-Cre-co-expression, the herein named szPBP and the rmVTA. While previous transgenics-and optogenetics-based studies have implemented neurotransmitter-selective Cre-drivers, including for example the broadly expressesd Vglut2-Cre transgene to successfully unravel target areas, signaling properties and behavioral roles of midbrain glutamatergic neurons [27][28][29][30][33][34][35] , implementation of anatomically restricted but heterogeneous Cre-drivers like TrpV1-Cre possess the spatial advantage of targeting neuronal populations that co-exist within subregions of the SNc and the VTA. The findings presented here strongly highlight the importance of identifying and mapping unique expression patterns and implementing these in functional assays for the value of increasing the anatomical detail in pre-clinical studies aimed dissecting out brain circuitries.
While GRP expression in the human midbrain was similar, but not identical, to that in the mouse, the situation was strikingly different for NTF3 with ample expression in the VTA of the mouse and no detectable expression at all in the human midbrain. Mice are often used to model human conditions, but this finding shows that there is no complete overlap in gene expression patterns between mouse and human midbrains, which is important to bear in mind when translating data between species. Post-mortem human material can, however, vary in terms of conservation which might cause differences in the limits of detection in histological assays, and based on ethical considerations, the use of human tissue was restricted here.
Previous histological studies have shown that medium-sized to large DA neurons express higher NM levels and are more susceptible to degeneration, and that within the VTA areas (SNpd, Mv), only half of the TH-positive population contains NM while the other half is non-melanized [54][55][56][57] . Curiously, in addition to enabling the use of GRP analysis for identification of those DA cells that are spared in PD, which might prove clinically useful, there is the possibility that GRP itself might be of therapeutic value as a neuroprotective factor. The implication of GRP in pathophysiology is far from new; both GRP analogues and GRP receptor antagonists are important targets in anti-cancer therapies already 58,59 and both have been proposed as putative targets in neuropsychiatric disorders [60][61][62] . Indeed, GRP was listed as elevated in the VTA over the SNc in two separate "VTA vs SNc" screens over a decade ago 6,7 and when PC12-cells over-expressing the alpha-synuclein gene were exposed to GRP, it was shown that these cultured cells were less vulnerable to the DA cell toxin MPP + 6 . Thus, almost 12 years after the identification of GRP as a neuroprotective factor in cell-based models of PD, the current study contributes with human mapping data that supports those previous findings.
Based on the findings presented here, by defining a restricted glutamatergic pathway between the VTA and NAcSh, TRPV1 might serve as an interesting target in future VTA-based studies, while the identification of GRP as a marker for non-melanized DA neurons that are neuroprotected in humans with PD could be of clinical value when exploring new therapies. For all animal experiments, ethics approval was obtained from the Uppsala Animal Ethical Committee. For analysis of human brain material, human brain tissue samples were obtained in a Brain Donation Program of the Brain Bank "GIE NeuroCEB" run by a consortium of Patients Associations: ARSEP (association for research on multiple sclerosis), CSC (cerebellar ataxias), France Alzheimer and France Parkinson. Informed consents were signed by the patients themselves or their next of kin in their name, in accordance with the French Bioethical Laws. The Brain Bank GIE NeuroCEB has been declared at the Ministry of Higher Education and Research and has received approval to distribute samples (agreement AC-2013-1887). Kit (Qiagen, Sweden). RNA expression was analyzed by microarray on GeneChip ® Mouse Gene 1.0 ST Arrays by the Uppsala Array Platform, Uppsala University as previously described 64 . The raw data was normalized with Expression Console (Affymetrix, USA) using the robust multi-array average (RMA) method 65 . Subsequent analysis of the gene expression data was carried out in R (http://www.r-project.org) using Bioconductor packages (www.bioconductor.org) 68 . An empirical Bayes moderated paired t-test was applied 66 using the 'limma' package 67 and p-values were adjusted 71 . Principal component analysis revealed that one of the 5 SN samples clustered further away than all other samples. This and the matching VTA sample were removed and the statistical analysis re-performed. The genes were ranked according to adjusted p-value and the 50 most significant genes ordered by log 2 -fold change between the SN and VTA.

Animals. Dat-Cre
In situ hybridization on mouse and human tissue. Oligonucleotide and riboprobe sequences are listed in Supplementary Table S3. Radioactive in situ hybridization was performed on coronal cryosections derived from mouse and human brains as previously described 69 . For cellular mRNA expression analysis, slides were dipped in NTB emulsion, revealed after an exposition of 6 weeks and sections counterstained with toluidine blue. For non-radioactive in situ hybridization (colorimetric and fluorescent; sdFISH) analysis performed with riboprobes, cryosections were air-dried, fixed in 4% paraformaldehyde and acetylated in 0.25% acetic anhydride/100 mM triethanolamine (pH = 8). Sections were hybridized for 18 h at 65 °C in 100 μ l of formamide-buffer containing 1 μ g/ml digoxigenin (DIG) labelled probe for colorimetric detection or 1 μ g/ml DIG and 1 μ g/ml fluorescein-labelled probes for fluorescent detection. Sections were washed at 65 °C with SSC buffers of decreasing strength, and blocked with 20% FBS and 1% blocking solution. For colorimetric detection, DIG epitopes were detected with alkaline phosphatase-coupled anti-DIG fab fragments at 1/500 and developed with NBT/ BCIP. For fluorescent detection, sections were incubated with HRP-conjugated anti-fluorescein antibody at 1/1000. Signal were revealed with the TSA ™ Kit (Perkin Elmer) using Biotin-tyramide at 1:75 followed by incubation with Neutravidin Oregon Green conjugate at 1:750. HRP-activity was stopped by incubation of sections in 0,1 M Glycine and 3% H 2 O 2 . DIG epitopes were detected with HRP conjugated anti-DIG antibody at 1:1000 and revealed with TSA ™ Kit (Perkin Elmer) using Cy3 tyramide at 1:200. All slides were scanned and analyzed on NanoZoomer 2.0-HT Ndp2.view (Hamamatsu). The sdFISH method allows co-expression analysis of different mRNAs and thus enabled the use mRNA for the vesicular transporters, Viaat and Vglut2, to identify GABA-and glutamate-signaling neurons, respectively. As their protein products are localized in the synapses, only the use of mRNA-based methodology is useful for this purpose, while immunohistochemical approaches are not. Midbrain sections at the SN level from control (n = 2; case number 3549, 8401) and PD patients (n = 3; case number 3392, 3490, 5874) (reference: GIE Neuro-CEB BB-0033-00011) were used. The pathology reports stated Diffuse Lewy body disease (DLB) for PD patient case number 3392 and 5874, and DLB + Alzheimer's disease: Braak IV and Thal amyloid phase 1 + grain disease for PD case number 3490.