Molecular mechanisms of dopaminergic subset specification: fundamental aspects and clinical perspectives

Abstract

Dopaminergic (DA) neurons in the ventral mesodiencephalon control locomotion and emotion and are affected in psychiatric and neurodegenerative diseases, such as Parkinson’s disease (PD). A clinical hallmark of PD is the specific degeneration of DA neurons located within the substantia nigra (SNc), whereas neurons in the ventral tegmental area remain unaffected. Recent advances have highlighted that the selective vulnerability of the SNc may originate in subset-specific molecular programming during DA neuron development, and significantly increased our understanding of the molecular code that drives specific SNc development. We here present an up-to-date overview of molecular mechanisms that direct DA subset specification, integrating our current knowledge about subset-specific roles of transcription factors, signaling pathways and morphogenes. We discuss strategies to further unravel subset-specific gene-regulatory networks, and the clinical promise of fundamental knowledge about subset specification of DA neurons, with regards to cell replacement therapy and cell-type-specific vulnerability in PD.

Introduction

Mesodiencephalic dopaminergic (mdDA) neurons in the ventral mesodiencephalon are involved in the control of voluntary movement and regulation of emotion, and dysfunction is associated with several psychiatric and neurodegenerative diseases, such as obesity, addiction, schizophrenia, and Parkinson’s disease (PD). A pathological hallmark of PD is the specific degeneration of neurons in the substantia nigra pars compacta (SNc), whereas neurons in the ventral tegmental area (VTA) remain unaffected [1, 2]. Despite the realization that this selective vulnerability of a specific DA cell group implies the existence of different subpopulations of mdDA neurons, it was only recently postulated that this specific vulnerability may be explained by specific molecular programming of the SNc. This increasingly popular hypothesis has lead to extensive research into the molecular code of developing mdDA subsets in the last decade and was initially reviewed by Smidt et al. [3, 4]. Ever since, major progress has been made in our understanding of developing mdDA subsets. Here, we review the molecular circuitry that defines mdDA subsets and present an up-to-date model of the molecular mechanisms that direct mdDA subset specification, integrating our current knowledge about subset-specific roles of transcription factors, signaling molecules, and morphogenes.

Early specification and patterning of the mdDA domain

In the developing brain, mdDA neurons are located in the neural tube area that later becomes the mesencephalon (midbrain) and diencephalon (forebrain). The early specification of the permissive region for mdDA neuron generation requires concerted action of signaling factors, such as Shh (sonic hedgehog), Fgf8 (fibroblast growth factor 8), Tgf-b (transforming growth factor beta), Wnt1 (wingless-type MMTV integration site 1), and (Wnt5a wingless-type MMTV integration site 1) on the one hand, and several key transcription factors (TFs) including Nurr1 (nuclear receptor related-1 protein), Pitx3 (paired-like homeodomain transcription factor 1), En1 (engrailed-1), En2 (engrailed-2), Otx2 (orthodenticle homeobox 2), Lmx1a (lim homeobox transcription factor 1a), Lmx1b (lim homeobox transcription factor 1b), Foxa1 (forkhead box protein a1), and Foxa2 (forkhead box protein a2) on the other hand [3–8] (Fig. 1).

Fig. 1.

Timescale of mdDA neuron development. Key factors in three subsequent steps of mdDA development are displayed, and regulatory roles are indicated for factors most clearly associated with mdDA subset specification (i.e., BDNF, RA, Dlk1, and En1, Otx2, Pitx3 (yellow boxes)). During the regional specification phase, several TFs and signaling molecules interact to establish the mdDA neuronal field. In the mdDA progenitor expansion phase, Otx2 controls proliferation of mdDA progenitors by regulating the expression of Lmx1a, Msx1, Ngn2, and Mash1, and during this phase, these precursors expresss a large set of genes from early mdDA progenitor specification (En1/2, Lmx1a/b, Foxa1/2). Around E10.5 neurons start to express Nurr1 that acts in concert with Pitx3 and En1 to activate the mdDA phenotype during mdDA differentiation (i.e., transition of postmitotic mdDA precursors to fully differentiated mdDA neurons). During this phase, complex interplay between Otx2, En1, Pitx3, and other factors defines mdDA subsets, as specified in more detail in Fig. 2

The mid-hindbrain border (MHB), or isthmus, is a critical signaling center in the establishment of the mdDA neuronal field, since it produces Fgf8 that acts together with the ventralizing factor Shh to specify the initial mdDA neuronal field [9]. Rostral to the isthmus, in the mdDA neuronal field, Otx2 is expressed, whereas caudal to the isthmus, in the presumptive serotonergic field, gastrulation brain homeobox 2 (Gbx2) is expressed, and manipulation of these factors affects the size of the early mdDA domain [3, 10–13]. Multiple other signaling molecules are involved in the early establishment of the mdDA neuronal field. Transforming growth factor beta (Tgf-b) induces early Shh signaling, and is therefore required for the induction of the ventral mdDA neuronal field [14, 15]. In human pluripotent stem cells, Tgf-b pathway mediators have critical upstream roles as regulators of Wnt1, Lmx1a, and Foxa2, and in complex interplay with Shh and Fgf8, promote mdDA neuron differentiation [16]. Wnt1, as well as Wnt5a, are critical factors in the establishment of the mid-hindbrain region. Among other functions, Wnt1 induces expression of engrailed genes, important in multiple stages of mdDA development [6, 17, 18]. The expression of Wnt1 and Fgf8 in the isthmic organizer (IsO) is in turn controlled by Lmx1b [19, 20]. Recently, an Lmx1b-miR135a2 regulatory circuit was identified that modulates Wnt signaling and determines the size of the mdDA progenitor pool [21]. Also, the first wave of retinoic acid (RA) signaling in the midbrain is important for positioning of the MHB [22, 23].

Once the permissive mdDA region has been established, multiple TFs are involved in the establishment of the size and specification of the mdDA progenitor domain. Neurogenin 2 (Ngn2) is important for the proper specification of mdDA precursors. Ngn2 activates Sox2+ progenitors that later develop into Nurr1+ mdDA neurons [24] and suppresses Nurr1-induced gene transcription in mdDA neurons. Achaete-scute homologue 1 (Ascl1/Mash1) is involved in the maintenance of progenitor populations in the developing mdDA area, and Mash1-null mice display a reduced neurogenesis in the mdDA system [25, 26]. The role of Mash1 in mdDA neuron development may, however, be rather permissive than instructive, and it is well possible that there is functional redundancy between Mash1 and Ngn2 [24] (Fig. 1).

Transcription factors involved in mdDA subset specification

After the early establishment of the permissive mdDA region and expansion of the mdDA progenitor domain, multiple TFs and signaling molecules act together to provide cellular diversity and subset specification in the developing mdDA system. From rostral to caudal, transverse domains of the Central Nervous System (CNS) are designated telencephalon, rostral diencephalon, prosomere 3–1, the midbrain and hindbrain, of which prosomere 1/2/3 and the midbrain comprises the mdDA neuronal field [3, 27–30].

The roles of key mdDA TFs in mdDA neuron development and maintenance have been extensively reviewed in two excellent recent reviews [7, 31]. In this review, we focus on subset-specific roles of TFs most prominently associated with mdDA subset specification: Pitx3, En1, Otx2, and Lmx1a/b (Figs. 1, 2, 3).

Fig. 2.

Transcription Factors in mdDA subset specification. a Expression domains of critical mdDA transcription factors at E14.5 in wild type. b Expression domains of Pitx3 and En1 in E14.5 saggital medial and lateral sections. c, d Expression domains of En1, Pitx3, and Nurr1 in En1- and Pitx3-deficient embryos at E14.5. e–g Expression domains of DA metabolism genes in E14.5 wild-type e, Pitx3-deficient (f), and En1-deficient (g) embryos. h During mdDA neuron development, En1 may initially activate Pitx3 expression (1) in a rostrolateral subset; subsequently, Pitx3 downregulates En1 in turn (2). i One option to infer subset-specific Pitx3-mediated repression of En1, is that active Pitx3 protein exceeds a certain repression treshold (Trepr). j Another option is that Pitx3 functionally modulates En1 protein activity, by subset-specific activation of genes encoding En1 modulatory proteins. k Both En1 and Pitx3 are critical for the induction of the RA-synthesizing enzyme Ahd2 in mdDA neurons, Pitx3 activates Ahd2 directly, whereas En1 may activate Ahd2 either directly, or indirectly, via regulation of Pitx3. In panel (a, c, e–g) a color gradient indicates expression in a midbrain and/or p1/2/3 subset, and reduced color intensity (opacity) indicates lower expression. A activator, R repressor, C caudal, R rostral, ko knockout, mdDA mesodiencephalic dopaminergic, p1 prosomere 1, p2 prosomere 2, p3 prosomere 3

Fig. 3.

Gene-regulatory networks in mdDA subset specification. a In the rostrolateral mdDA neuronal subset that harbors neurons destined to form the SNc in adult, Pitx3 expression is initially induced by GDNF and En1 (1). Pitx3 then potentiates Nurr1 to drive the expression of Vmat2 and Dat [50]. Expression of Ahd2 is also under the combined transcriptional control of Nurr1 and Pitx3, and expression is restricted to the rostrolateral mdDA neuronal subset. Ahd2 synthesizes retinoic acid (RA) from its precursor retinaldehyde, and activates the expression of Th, D2R, and possibly BDNF, presumambly through binding to Retinoic Acid Receptors (RARs). Moreover, RA represses Nurr1-mediated expression of Dlk1 in the developing SNc. In rostrolateral mdDA neurons, Pitx3 is possibly engaged in a positive feedback loop with BDNF, augmenting survival of SNc neurons. Also, within this subset, Pitx3 antagonizes the caudal mdDA neuronal phenotype (as illustrated by the inhibition of Cck expression) through an RA-independent mechanism, possibly via direct or indirect inhibition of En1 expression within this subset. b In the developing caudal mdDA neuronal subset that eventually gives rise to the adult VTA, Pitx3 and En1 can both induce the expression of Dat and Vmat2, in an RA-independent manner, since Ahd2 is not expressed in this subset. Pitx3 is not required for Th expression within this subset, but partially depends on En1 expression. In caudal mdDA neurons, Pitx3 can not antagonize Cck expression, possibly due to the higher levels of En1. In the absence of RA, Nurr1-mediated expression of Dlk1 is not repressed, and this may suppress SNc properties in VTA neurons, although this remains to be established. In a subset of VTA neurons, Otx2 antagonizes the expression of Dat. These subset-specific GRNs illustrate the importance of RA-dependent and -independent aspects of the Pitx3-gene-regulatory networks, as well as the critical roles of Otx2, and subset-specific interplay of En1 and Pitx3 in mdDA subset specification

Pitx3

Pitx3 is expressed in the mouse ventral mdDA area from E11.5 onward and a terminal selector gene for mdDA neurons [32]. The spatiotemporal expression pattern of Pitx3 suggests that Pitx3 is critical for terminal differentiation of mdDA neurons, and not involved in earlier DA neuron development [33]. Indeed, many studies have demonstrated critical involvement of Pitx3 in the terminal differentiation and survival of an mdDA subset. In the absence of Pitx3, specific loss of mdDA neurons in the SNc is observed with concomitant loss of nigrostriatal projections to the dorsal striatum, whereas the neighboring neurons of the VTA are relatively unaffected. This was initially demonstrated in aphakia mice, a natural Pitx3-mutant that lacks Pitx3 transcript due to mutations in its promoter [34, 35], and later confirmed in the transgenic Pitx3 knockout, gfp knock-in mouse model [33, 36–38]. These early observations indicated a paradoxical role for Pitx3 in the mdDA system: whereas Pitx3 is expressed in most if not all mdDA neurons during late differentiation, in the absence of Pitx3 only a small subset is affected, and this subset comprises neurons destined to form the SNc. In Pitx3-mutant embryos, a rostrolateral subpopulation of mdDA neurons, that ultimately forms the SNc, is halted in its terminal differentiation, indicated by the lack of Th expression (Figs. 2d, f, 3) [33, 39–41]. In this rostrolateral subset, Pitx3 induces the expression of Ahd2/Aldh1a1 (aldehyde dehydrogenase 1 family, member A1), most likely by direct activation, since Pitx3 binds the Ahd2 promoter [40]. Ahd2 is selectively expressed in the mdDA subset that is critically dependent on Pitx3 (Figs. 2e, k, 3a, 4a–d). Ahd2 encodes an aldehyde dehydrogenase enzyme, that, in addition to the detoxification of DOPAL (dihydroxyphenylacetaldehyde) through the conversion to DOPAC (dihydroxyphenylacetic acid) [42], converts retinaldehyde to retinoic acid (RA), and therefore serves as a potent generator of RA in the developing mdDA system [40, 42]. Functional relevance of Pitx3/Ahd2-mediated RA signaling in a developmental mdDA subset has been demonstrated in vivo and ex vivo (discussed below), and RA-dependent and -independent gene-regulatory pathways of Pitx3 in mdDA neurons have been identified [40, 41]. At developmental stage E14.5, rostrolateral mdDA neurons display a specific dependence on RA to induce Th (tyrosine hydroxylase), and most likely D2R (Drd2; dopamine 2 receptor) expression (mainly expressed in the rostrolateral area at this stage). Expression of other key genes in DA metabolism, Dat (Slc6a3; dopamine transporter) and Vmat2 (Slc18a2; vesicular monoamine transporter 2), is not affected by RA (Figs. 3a, 4a–d).

Fig. 4.

Retinoic acid-dependent gene-regulatory pathways in mdDA development. a In wild-type mdDA neurons of the SNc subtype, Pitx3 cooperates with Nurr1 to induce the expression of the RA-synthesizing enzyme Ahd2. RA then induces the expression of Th and D2R. c In wild-type mdDA neurons of the VTA subtype, the expression of Th is RA-independent. In addition, expression of Vmat2 and Dat in these neurons is Pitx3-dependent, but RA-independent, as is the case in SNc neurons a. b Supplementation of RA to pregnant mice during the critical period for mdDA neuron differentiation restores the expression of Th and D2R in neurons of the SNc subtype, whereas expression of Vmat2 and Dat is not rescued. d In a passive diffusion model, retinal would be omnipresent in the mdDA system, only locally converted to RA by Ahd2 in the p1/p2 area, but diffusing freely into the caudal midbrain. e In an active degradation model, Cyp26 enzymes in the caudal midbrain could degrade RA, thus lowering the local RA levels. f When active degradation by Cyp26 enzymes is combined with caudal-specific expression of Dhrs3, that catalyzes the reversal of retinal into vitamin A, local retinal levels would be low in the caudal midbrain, which, if combined with local RA synthesis by Ahd2, could induce a relatively strict border of RA levels (high in the rostrolateral p1/2, low in the caudal midbrain). SNc substantia nigra pars compacta, VTA ventral tegmental area, RA retinoic acid, m midbrain, p1 prosomere 1, p2 prosomere 2

In addition to the subset-specific induction of Ahd2, other Pitx3-induced molecular programs may explain the subset-specific vulnerability observed in Pitx3-mutant mice. Pitx3 was recently demonstrated to activate the expression of brain-derived neurotrophic factor (BDNF) in rostrolateral SNc mdDA neurons specifically. In these experiments, it was elegantly shown that loss of BDNF expression correlates with the loss of SNc neurons in Pitx3-deficient mice [43]. Treatment of primary cultures of Pitx3-deficient mdDA neurons with BDNF augmented their survival, suggesting functional relevance for Pitx3-induced BDNF signaling in the development and maintenance of mdDA neurons of the SNc subset (Fig. 3a) [43]. Notably, specific dependency of SNc neurons on BDNF for their survival has been demonstrated in many studies (as discussed below). The same study reported that Pitx3 expression is induced by glial-derived neurotrophic factor (GDNF), another growth factor implicated in subset-specific vulnerability of mdDA neurons (see below) (Fig. 3).

Pitx3 regulates the conserved micro-RNA miR-133b that in turn suppresses Pitx3 expression [44]. Although the exact role of this Pitx3-miR-133b negative feedback loop is difficult to interpret, and mdDA development appears to be normal in miR-133b null mice [45], this miRNA may have more subtle roles in subset specification of mdDA neurons, especially given its recently established role in adipocyte specification [46, 47].

Regarding the function of Pitx3 in mdDA neurons, at least three important questions have remained unanswered: (1) is Pitx3 itself directly responsible for survival of mdDA neurons (for example, by activation of pro-survival and/or anti-apoptotic pathways), or is SNc neurodegeneration in Pitx3-deficient mice solely the result of mis-specification of SNc neurons? (2) at what stage do SNc neurons exactly start degenerating in Pitx3-deficient mice?, and (3) does dependency of SNc neurons on Pitx3 extend into adulthood? Regarding the latter, it would be of utmost interest to conditionally remove Pitx3 at postnatal and/or adult stages in future studies.

Interplay of Pitx3 and Nurr1

Nurr1 regulates Pitx3 expression in vitro [48], but Pitx3 expression appears unaffected in Nurr1-null mutants. Recent studies have unambiguously demonstrated that Pitx3 and Nurr1 coordinately regulate mdDA neuron specification. Pitx3 potentiates Nurr1 in terminal differentiation of mdDA neurons, most likely by releasing SMRT/HDAC-mediated repression of Nurr1 target genes [49–53]. Whereas most studied target genes were shown to be activated by both Pitx3 and Nurr1 (such as Vmat2 and Dat), at least two Nurr1 target genes have been described that are downregulated in the absence of Nurr1, but upregulated and ectopically expressed in the absence of Pitx3: cholecystokinin (Cck) and delta-like homologue 1 (Dlk1). In contrast to the genes activated by Nurr1 and Pitx3 in concert, Cck and Dlk1 expression is confined to the caudal mdDA area in wild type embryos at the terminal differentiation stage. The molecular mechanisms behind such subset-specific interplay of Nurr1 and Pitx3 remain to be determined (Figs. 2e, 3a, b).

En1

During early development, En1 and its paralogue En2 are critical for the proper patterning of the mdDA area. Initially, En1 and En2 are expressed at the MHB, and crucial for the formation of the isthmus [54, 55]. Although En1 was initially described to be broadly expressed in mdDA neurons starting at their early development and continuously into adulthood [56, 57], recent in-depth spatial analysis performed in our lab justifies a refinement of this model. In situ hybridization analysis of the E14.5 mdDA area clearly indicated substantially lower En1 expression in rostrolateral mdDA neurons (as compared to caudal mdDA neurons) (Fig. 2a, b). These findings have been confirmed by qPCR in purified mdDA neurons that indicate 2-3 fold higher levels of En1 in caudal mdDA neurons (JV Veenvliet, unpublished results). In line with subset-specific levels and roles of En1 is the recent observation that expression of En1 in more rostral diencephalic DA progenitors is lost early, and then only weakly reactivated in precursors [58]. Moreover, this study suggested that a small diencephalic Th+ population is generated independent of En1. However, conclusions were drawn from analyses at early timepoints (E10.5–E12.5) and not based on analysis of En1-null embryos, but on phenotypical consequences of En1 ablation in Fgfr1/2 conditional compound mutants, and should therefore be treated with caution.

Whereas En1 and En2 were initially reported to be critical for the development and survival of all mdDA neurons in a dose-dependent manner [56, 59], recent research has indicated more refined and subset-specific roles of En1 (Figs. 2c, g, 3a, b). Initial studies were performed in En1 and En2 double mutants, and indicated massive loss of mdDA neurons by E14, due to caspase-dependent apoptosis [56, 60]. Simon et al. [60] analyzed various allelic combinations (wild-type, En1 ^−/−; En2 ^−/−, En1 ^−/−, and En1 ^−/−; En2 ^+/−) and concluded that mice mull for either En1 or En2 had no apparent reductions in mdDA neurons of the SNc and VTA subtype and may therefore compensate for each others loss partially or even entirely. It was reported that, at P0, distribution and density of mdDA neurons of the SNc were similar in En1-null mice as compared to wild type, and that the only detectable difference was a more loose arrangement of VTA neurons. In contrast, later research indicated that En1 is critical for the maintenance of mdDA neurons, since En1-heterozygous mice displayed progressive DA cell loss in both an En2-null and a wild-type background [61, 62]. However, the exact fate of mdDA neurons in the absence of En1 remained unclear in these studies, hampered by the perinatal lethality in En1-null mutant mice related to cerebellar deletions [59]. We and others have circumvented this impediment by transferring the original En1-null allele from the 129/Sv strain to the C57BL/6J background, thereby suppressing the cerebellar phenotype and facilitating the analysis of En1 knockout adult mice [63, 64]. Six-week-old En1-null mice have severe loss of both SNc and VTA neurons as assessed by the number of Th+ neurons, whereas En1 heterozygous mice display an intermediate phenotype, indicating a dose-dependent effect of En1 on the development and survival of mdDA neurons. Concomitantly, Th+ fibers are lost in the both SNc projection areas (dorsal striatum), as well as VTA projection areas (nucleus accumbens). In the same study, Nissl staining on adjacent sections revealed a clear reduction of DA cell density in both SNc and VTA of En1-null mice, indicating that these neurons were lost in the absence of En1 [64]. Similar to the Pitx3-knockout, where loss of SNc neurons in adult is preceded by developmental programming defects in a rostrolateral subset, the absence of En1 also leads to programming defects during mdDA neuron development. At the terminal differentiation stage, expression of all genes of the DA gene battery [Th, Dat, Aadc (aromatic l -amino acid decarboxylase), Vmat2, D2R] is affected in En1-null embryos (Fig. 2g). Especially Th and Dat displayed subset-specific dependence on En1, since expression is severely downregulated in rostrolateral mdDA neurons, but less affected in the caudal mdDA subpopulation [64]. Importantly, unaffected expression of early postmitotic mdDA precursor marker Nurr1 and mdDA progenitor marker Lmx1a was reported in En1-null embryos in the same study, suggesting that the loss of DA neuron marker expression is a direct consequence of En1-driven programming deficits, and not the result of cell death at this stage [64]. Notably, we have reported ectopic expression of multiple mdDA markers in rhombomere 1 of En1-deficient embryos (Fig. 2g).

Interplay of En1 and Pitx3

We have recently reported how extensive and subset-specific crosstalk of En1 and Pitx3 may be critically involved in DA subset specification (Fig. 3). Initial evidence for functional crosstalk of En1 and Pitx3 came from the observation that En1 expression is highly upregulated in Pitx3-null embryos [41]. In turn, the expression of Pitx3 is ablated in a rostrolateral subset in En1-null embryos [64], and the rostrolateral phenotype in En1-null embryos resembles that of Pitx3-deficient embryos of the same stage (Fig. 2c, d, f, g), suggesting that both En1 and Pitx3 are essential for the induction of the rostrolateral phenotype. Thus, En1 and Pitx3 clearly influence each others expression level, although it remains to be studied if this is mediated by direct or indirect effects. The ablation of Ahd2 in both Pitx3- and En1-deficient embryos, and the downregulation of multiple other factors involved in RA metabolism in En1-null embryos (Fig. 2k) [64] point to RA signaling as a convergence point of En1 and Pitx3 in rostrolateral DA subset specification.

Although En1 and Pitx3 have similar effects in the rostrolateral mdDA domain (i.e., induction of the mdDA phenotype), reciprocal regulation of target genes has been observed in the caudal mdDA domain using both genome-wide and in-depth expression analysis tools. The most striking example of such a reciprocally regulated gene is Cck (Fig. 3). In wild-type embryos, Cck expression is restricted to the caudal mdDA subpopulation. In Pitx3-deficient embryos, Cck is upregulated and expression is expanded into the rostrolateral region that, as a consequence of Pitx3 ablation, is devoid of Th expression, suggesting that Pitx3 antagonizes Cck in a rostrolateral mdDA subset. In sharp contrast, En1-null embryos completely lose Cck expression in the caudal mdDA domain, suggesting a critical role for En1 in the induction of Cck expression [41, 64]. Based on these data, and the evidence that birth of rostrolateral SNc neurons precedes that of caudal VTA neurons [65], we have proposed a model for dopaminergic subset specification based on functional and subset-specific interplay of Nurr1, En1 and Pitx3 [64]. In short, the induction of Nurr1 would generate a default DA neuron that could acquire the rostrolateral or caudal mdDA neuron phenotype based on differential interplay of En1 and Pitx3. A rather complicated negative feedback system, where En1 initially induces Pitx3 expression and is subsequently repressed by Pitx3 was proposed to program the rostrolateral mdDA phenotype (Figs. 2h, i, 3a) [64]. If such transcriptional regulation is executed directly (via promoter binding) or indirectly is unknown. Although the lower levels of En1 in rostrolateral mdDA neurons, and their upregulation in Pitx3-deficient embryos suggest that Pitx3 represses En1 at the transcriptional level, modulation of En1 protein function cannot be excluded, and may also be involved (Fig. 2j). Expression of multiple genes encoding En1-modulatory proteins (by binding to the En1 homology domain) of the Tle- and Pbx-family (Pbx1, Pbx3, and Tle3) is downregulated in Pitx3-deficient embryos and in wild type these genes display subset-specific expression that is maintained in adulthood [30, 64, 66]. After establishment of the rostrolateral phenotype, the molecular signature of the remaining caudal subpopulation could be established by a relatively simple and default program in which En1 and Pitx3 can (partially) compensate for each others loss (Fig. 3a, b).

Although subset-specific En1/Pitx3 crosstalk is insufficient to fully explain the differential molecular programming underlying DA subset specification, comparative analysis of Pitx3- and En1-deficient embryos lead to the identification of two mdDA subpopulations differentially dependent on En1 and Pitx3: the rostrolateral Ahd2+ population, and the caudal Cck+ population. At E14.5, it was demonstrated that both appear to be mutually exclusively expressed, and this is most likely maintained into adulthood, where Ahd2 expression is enriched in SNc neurons, whereas Cck expression is enriched in VTA neurons [30]. The subset-specific expression of Ahd2 and Cck may give rise to crucial physiological and functional differences. Although this remains to be assessed in future studies, Cck is associated with VTA functions, whereas Ahd2 is associated with SNc functions, and related to PD pathogenesis [67, 68]. Intriguingly, it has been described that Cck inhibits DA neurotransmission [69], but if increased levels of Cck in Pitx3-deficient mice affect DA neurotransmission remains to be established. Moreover, functional blockade of Cck may relieve PD symptoms, since Cck-B receptor antagonists potentiate L-DOPA effects in MPTP-lesioned monkeys [70].

Otx2

Otx2 belongs to the bicoid family of homeodomain TFs. During early mdDA neuron development, Otx2 has been suggested to activate the expression of Lmx1a, Msx1, Ngn2, and Mash1, either directly or indirectly, and thereby trigger proliferation and differentiation of mdDA progenitors [71, 72]. Pioneering work of Antonio Simeone and colleagues established Otx2 as one of the critical factors involved in mdDA neuron subset specification. Although Otx2 is expressed in all mdDA progenitors, it is selectively expressed in a subset of mdDA neurons of the VTA subtype at later developmental and adult stages, and required for the differentiation of this subset [73–76]. This pattern is conserved in primate and human, and maintained into adulthood, where Otx2 is selectively expressed in a subgroup of VTA neurons that expresses low levels of glycosylated Dat (glyco-Dat) and Girk2. Otx2 specificies this subset of VTA neurons by antagonizing molecular features of the dorso-lateral VTA, including Girk2 and Dat expression [74]. In the same study, it was demonstrated that Otx2 limits the number of neurons with efficient DA uptake, and that neurons with low levels of Otx2, and therefore high levels of glyco-Dat, are most vulnerable to MPTP toxicity. Moreover, ectopic expression of Otx2 in SNc neurons protected these neurons from MPTP-induced degeneration [74]. Otx2 gain-of-function experiments have been demonstrated to elevate expression of known VTA-enriched genes in vitro in both the MN9D cell line model and ventral midbrain primary cultures [76], including many factors involved in the formation of DA neuronal projections. Indeed, Otx2 conditional knockout mice (En1-cre/+;Otx2-flox/flox) loose VTA DA neuronal projections, whereas SNc neuronal projections (to the dorsolateral striatum) are not affected. In the same study, it was elegantly demonstrated that Otx2 knockdown reduces the number of mdDA neurons, and increases the vulnerability of mdDA neurons to MPTP [76]. Mild overexpression of Otx2 in SNc progenitors and neurons rescued the defects observed in case of En1 haploinsufficiency (i.e., progressive loss and increased MPTP sensitivity of SNc neurons, as discussed above). Based on these observations the author suggested that Otx2 and En1 share similar properties, controlling mdDA neuron development and maintenance in the VTA and SNc, respectively [73]. Although our observations that En1-heterozygous and En1-null adult mice display both SNc and VTA defects contradicts this theory to some degree, this is an intriguing possibility, and could be investigated by in-depth analysis of the downstream molecular pathways of En1 and Otx2 on a genome-wide scale. This is especially important, because the En1-cre driver is also heterozygous for endogenous En1, and therefore it is critical to establish to what extent defects observed in the Otx2 conditional knockout model are the result of En1 and Otx2 interplay.

Recently, the functional consequences of over- and ectopic expression of Otx2 were reported. When Otx2 was expressed throughout the En1 expression domain, increased numbers of glyco-Dat positive mdDA neurons and DA innervation were observed [72, 77]. This observation is contradictory to the Otx2 overexpression driven decrease of glyco-Dat levels reported in an earlier study [74]. An explanation for this discrepancy may be found in the use of different Cre-drivers. In the study by Tripathi et al. [77] En1-cre was used to drive Otx2 overexpression, a model previously shown to enhance mdDA neuron progenitor proliferation [72]. In contrast, Di Salvio et al. used Dat-cre as a driver, and Otx2 overexpression was therefore only induced after the mdDA progenitor stage. Thus, the effect of Otx2 overexpression on glyco-Dat levels in mdDA neurons appears highly timing dependent.

In the En1-cre driven Otx2 overexpression model, the number of glyco-Dat positive mesocortical DA fibers is increased, and correlated with increased density of parvalbumin-positive interneurons and hypolocomotion behavior [77]. Since hypolocomotion is observed in multiple murine PD models [78–80], it may therefore be questioned if Otx2 overexpression, despite its clear protective role against MPTP toxicity, can truly serve as a future treatment option in PD, since it might induce transdifferentiation of mdDA neurons from the SNc to the VTA subtype. It would therefore be critical to assess the true identity of neurons rendered less vulnerable to MPTP toxicity by Otx2 overexpression, for example, by analyzing their relative Ahd2 and Cck transcript levels. Additionally, the decrease of MPTP toxicity upon Otx2 overexpression, may well be related to lower uptake of MPTP in cells that express lower levels of Dat (expression of Dat is decreased upon Otx2 overexpression, as discussed above), since there is a quantitative relationship between the expression of Dat and the extent of cytotoxic effects of MPTP [81, 82].

Lmx1a and Lmx1b

LIM homeodomain transcription factors Lmx1a and Lmx1b are both involved in mdDA neuron development. Initial studies in chick suggested a critical role for Lmx1a in establishment of the mdDA neuron phenotype [83], supported by the observation that overexpression of Lmx1a in mouse embyronic stem cells (ESCs) induces the DA phenotype [83] which probably involves a Wnt1-Lmx1a regulatory loop [84]. However, results obtained in recent studies in mouse mutants for Lmx1a indicate that the initial observations in chick and ESCs must be treated with caution and suggest that Lmx1a may be less critical for mdDA neuron development than previously thought, at least in mouse. Several studies have described only mild mdDA phenotypes in Lmx1a mutant mice [85–88]. These mild phenotypes can be explained by functional redundancy of Lmx1a and Lmx1b in the developing mdDA system, but although indeed severe DA cell loss is observed in Lmx1a/Lmx1b double mutants, this hypothesis remains controversial [85, 87].

Intriguingly, some studies have described subset-specific roles of Lmx1a and Lmx1b in developing mdDA neurons. Perlmann and colleagues [87] have reported that Lmx1a and Lmx1b are most important for the specification of mdDA neurons in the medial and lateral progenitor domain, respectively. In Lmx1a-deficient embryos, generation of postmitotic mdDA neurons was limited to lateral positions at E11.5, suggesting that Lmx1a critically controls neurogenesis in the medial part of the mdDA progenitor domain. In contrast, lateral markers D2R (at E13.5) and Wnt1 (E11.5) are severely affected in Lmx1b-null embryos, but almost unaffected in Lmx1a-deficient embryos. However, Lmx1b may also be critical for differentiation of medial derived mdDA neurons, since Pitx3 and Th co-expression is not induced at terminal differentiation stages in the absence of Lmx1b [87, 89]. Since Lmx1b regulates Fgf8, Wnt1, and several isthmus-related TFs, and may therefore critically induce isthmic organizer (IsO) activity and specification of the mdDA progenitor domain [19, 90] it is also possible that the effect observed in Lmx1b-deficient embryos is secondary to its essential role for IsO activity. In support of this, specific inactivation of Lmx1b in mdDA progenitors does not affect mdDA neuron development [85].

More evidence for a role of Lmx1a in mdDA subset specification has come from a recent in-depth study of Lmx1a-deficient mice at embryonic and adult stages. This study indicated an esssential role for Lmx1a in the induction of the rostrolateral mdDA neuronal phenotype and using microarray analysis the molecular pathways underlying this phenotype were partially elucidated [86]. Downstream targets of Lmx1a included Nurr1 and Wnt/b-catenin signaling activator Rspo2. It was suggested that, within the rostrolateral mdDA domain, Lmx1a indirectly regulates the expression of Th, Ahd2, Aadc, and Vmat2 via the induction of Nurr1. Possibly in parrallel, Lmx1a induces Rspo2 expression that is in turn involved in the induction of Th, Ahd2, and Pitx3, since expression of these genes is downregulated in the rostrolateral mdDA domain in Rspo2-deficient mice [86]. However, it remains to be established if the observed defects reflect programming deficits, rather than migration defects or cell loss, although in Lmx1a-deficient embryos Th fiber outgrowth to the striatum appears unaltered, and in Rspo2-deficient embryos normal DAPI- and BGAL-staining (the Rspo2-mutant is an Rspo2-knockout;Lac-Z knockin model) suggests that there is no immediate neuronal loss in the mdDA region of these mice [86].

Other critical transcription factors

Nurr1 (Nr4a2) is an orphan nuclear receptor, for which no endogenous ligand has been identified to date. Nurr1 may function as a ‘master’ regulator of mdDA neuron development, since Nurr1 is critical for the development and maintenance of all mdDA neurons [91–93]. Although its role as a ‘master’ regulator strongly suggests that Nurr1 is not involved in mdDA subset specification, Nurr1 inactivation in already mature mdDA neurons induces a Parkinson-like phenotype, with SNc neurons being more severely affected than VTA neurons [94]. However, this may simply reflect the increased sensitivity of SNc neurons to environmental stress, especially since nuclear-encoded mitochondrial genes are massively affected upon conditional inactivation of Nurr1 [95]. Additionally, since Dat-cre was used as a driver in these studies, and Dat expression levels are typically higher in SNc than VTA neurons, lower Cre levels in the VTA could have lead to partial inactivation of Nurr1 in the VTA, thus explaining the ‘resilient’ phenotype of VTA neurons. Therefore, the most likely model, is that Nurr1 functions as a ‘master switch’ to induce mdDA neuron specification, and that subset specificity is determined by other TFs that modulate Nurr1 transcriptional activity by association with the Nurr1 transcriptional complex, such as En1 and Pitx3 [50, 64, 96].

Foxa1 and Foxa2 are broadly expressed in the mdDA neuron progenitor domain and critically involved in mdDA neuron development and maintenance at various stages [5, 97]. Regulatory mechanisms include their activation of Lmx1a and Lmx1b in mdDA neuron progenitors [98], induction of Nurr1 and En1 expression in immature mdDA neurons, and regulation of Pitx3, Th, Dat, Vmat2, and Aadc during terminal differentiation [5, 99, 100]. Foxa1 and Foxa2 appear to regulate mdDA neuron development in a dose-dependent manner, since failure to induce the mdDA phenotype is only observed when all four copies of Foxa1/2 are deleted [99]. As in the case of Nurr1, the only evidence for subtype-specific roles of Foxa1/2 comes from studies in adult mice. Foxa2 haploinsufficiency results in progressive degeneration of mdDA neurons of the SNc subtype, whereas VTA neurons remain unaffected, and is associated with Parkinson-like motor problems. Interestingly, Ahd2-positive neurons are specifically affected [101]. Moreover, conditional deletion of Foxa1 and Foxa2 in postmitotic mdDA neurons results in a severe decrease of Ahd2+ mdDA neurons of the SNc, whereas the number of Otx2+ mdDA neurons in the VTA is only mildly affected. In line with these observations, projections to the dorsolateral striatum are severely affected, whereas those to the nucleus accumbens are unaffected [97]. Although specific degeneration of SNc neurons in the absence of Foxa1/2 could again simply reflect the increased vulnerability of these neurons to environmental stress, this may not necessarily be the case, especially since Stott et al. have shown that mdDA neurons fail to maintain the expression of genes involved in DA metabolism, but continue to express Lmx1a, Lmx1b, and Nurr1 at E18.5 [97]. This suggests that mdDA neurons are not lost, but rather lose their mdDA neuronal phenotype. It is therefore possible that SNc and VTA neurons are truly differentially dependent on Foxa1 and Foxa2, and it should be investigated in more detail how specific allelic combinations of Foxa1 and Foxa2 loss-of-function mutants affect the development of mdDA neuronal subsets.

Engrailed-2 (En2) is expressed in a specific subset of the mdDA neuronal domain from early development into adulthood. Since the replacement of the En1 coding sequence with En2 in mouse has been described to completely rescue brain and skeletal muscle defects in the En1-null mutant, it has been suggested that the two paralogs are redundant, and that functional differences are the consequence of differences in spatiotemporal expression rather than different molecular properties [60, 102]. However, En2-null mice in an En1-heterozygous background have a less severe phenotype than En1-null mice in an En2-heterozygous background, and En2-null mice in an otherwise wild-type background have a less severe phenotype than En2-null mice in an En1-heterozygous background [103], suggesting that En2 is less critical for mdDA neuron development and/or maintenance than En1 [60]. Despite this interesting observation, in-depth analysis of mdDA neuron development and maintenance in single En2-null embryos and adult mice has not been performed to date, probably since initial studies reported no difference in distribution and number of mdDA neurons in adult En2-null mice [61, 103]. However, as discussed in the previous section, the same was initially reported for En1-null mice and recently contradicted based on more in-depth analysis [64]. It would therefore be of utmost interest to assess if the subset-specific expression of En2 is reflected by specific vulnerability or programming defects in case of its deletion. Moreover, not only En1, but also En2, is upregulated in Pitx3-deficient mice, and En2 is one of the most downregulated genes in En1-null embryos [41, 64]. Therefore, it is critical to investigate both convergence and divergence points of the molecular pathways of En1 and En2 during mdDA neuron subset specification in the near future.

Signaling mechanisms involved in mdDA subset specification

The roles of various signaling mechanisms involved in general DA neuron development and maintenance have been addressed in multiple excellent reviews [6, 104–106]. In this section, we discuss the possible roles of such signaling mechanisms in mdDA subset specification, focussing on neurotrophic factors that may have protective, restorative and/or stimulatory effects in mdDA neurons [107].

Neurotrophins

Critical roles for many neurotrophins in mdDA neuron development and maintenance have been shown [3, 104, 107]. However, only for few neurotrophins subset-specific dependency in the mdDA system has been investigated, and the few studies that have addressed this issue have mainly focused on the adult system.

BDNF levels are reduced in the Parkinsonian SNc [108]. During mouse embryogenesis, BDNF expression is activated by Pitx3 in the rostrolateral mdDA neuronal subset that ultimately forms the SNc [43]. Further evidence for Pitx3-driven activation of BDNF comes from in vitro studies [109, 110]. In turn, in BDNF-deprived E13 mouse ventral midbrain primary cultures, a ~50 % reduction of Pitx3-expressing neurons was observed with concomitant loss of Th+ neurons, possibly due to death receptor and caspase activation. Similar effects were observed in GDNF-deprived cultures [111]. Interestingly, GDNF signaling may be upstream of Pitx3, since it has been demonstrated that expression of GDNF in the ventral midbrain induces Pitx3 transcription via NF-kB signaling [43]. The Pitx3/BDNF pathway may be controlled by PGC-1a (peroxisome proliferator-activated receptor gamma co-activator-1 alpha), a positive regulator of genes required for mitochondrial biogenesis and antioxidant responses, since AAV-mediated overexpression of PGC-1a in the adult SNc results in downregulation of Pitx3, increased susceptibility to MPTP, and reduced BDNF levels [112]. However, it should be noted that this particular finding is in contrast with other studies that have mainly suggested protective effects of PGC-1a against MPTP toxicity in SNc neurons [113, 114].

The conditional removal of GDNF receptor Ret by means of the Dat-Cre driver induces slow and progressive degeneration of SNc, but not VTA neurons, with concomitant loss of dorsolateral striatal innervation [115]. Although these findings not necessarily imply that GDNF is exclusively important for the maintenance of SNc neurons, especially given the profound degeneration of both SNc and VTA neurons upon tamoxifen-induced GDNF removal in adult mice [116], it suggests that the nigrostriatal pathway from SNc to dorsal striatum is more dependent on GDNF/Ret signaling than the mesolimbic pathway from VTA to ventromedial striatum. A similar specific dependency of SNc neurons on a neurotrophin has been described In Tgf-a knockout mice, where a 50 % loss of DA neurons in the SNc is observed, with a concomitant 20 % reduction of the dorsal striatum volume, whereas VTA neuron numbers remain unchanged [117].

Despite these critical roles in mdDA neuron development and, especially, maintenance, little is known about the roles of neurotrophins in mdDA neuron subset specification. Detailed analysis of developing mouse embryos in which neurotrophins have been (conditionally) removed is a prerequisite to understand the roles of these factors in mdDA subset specification and specific neuron vulnerability. Although such mechanisms have not yet been studied in mdDA subset specification, evidence for an instructive role of neurotrophin signaling in neuron subtype specification comes from two elegant studies in dorsal root ganglia (DRG). Expression of TrkC (the receptor for NT-3) from the TrkA (the receptor for nerve growth factor) locus causes a subset of DRG neurons to switch fate [118], and Ret is involved in differentiation and diversification of Low-Treshold mechanoreceptor neurons of the DRG [119].

FGF and Wnt family members

The roles of Wnts and FGFs in mdDA neuron development and maintenance are complex, and for in-depth coverage of Wnt- and FGF-regulated networks in mdDA neuron development and maintenance we refer to some excellent recent reviews [6, 20, 105, 120]. In this section, we mainly restrict ourselves to mechanisms that may confer subset specification of mdDA neurons.

Wnt1 activates the expression of engrailed genes that are critical for mdDA neuron development, maintenance and subset specification, as discussed in previous sections [6, 17, 18]. Additionally, a Wnt1-Lmx1a autoregulatory loop directly activates Otx2 expression and influences Nurr1 and Pitx3 expression through direct activation of Lmx1a [84]. In their review, Wurst and Prakash have recently proposed the existence of two different Wnt1-regulated genetic networks that may infer mdDA subset specification [20]. In a gene-regulatory network (GRN) specifying caudal mdDA neurons, Otx2 would initially induce Wnt1 expression. Subsequent activation of Wnt1 target gene Ccnd1 (cyclin D1) [121] then gives rise to mdDA progenitors that eventually differentiate into VTA mdDA neurons of the Otx2+, but glyco-Dat low subtype, a view supported by the increased and ectopic expression of Wnt1 and Ccnd1 in En1-cre induced Otx2 overexpressing mice in the posterior mdDA area [72], although the possibility that the apparent subset specificity is in fact the result of differential rostral versus caudal expression of En1 (and therefore cre) should not be excluded [64]. In the Wnt1 GRN in the rostrolateral mdDA area, Wnt1 would induce Lmx1b and Lmx1a in an autoregulatory loop. Subsequent activation of Pitx3 and Nurr1 by Lmx1a/b would then lead to the activation of Pitx3/Nurr1 gene-regulatory pathways, including the activation of Ahd2 and BDNF, and thus specification of mdDA neurons of the SNc subtype [20]. However, as acknowledged by the authors, also from this model it remains unclear how Wnt1 could function in a subset-specific manner in the generation of the SNc and VTA mdDA neuronal subsets. In this light, the recent discovery of an Lmx1b-miR135a2 regulatory circuit that modulates Wnt1 signaling is intriguing [21]. Although the data suggest that this regulatory circuit is involved in determining the size of the mdDA neuron progenitor pool rather then conferring subset specificity, it would be interesting to study this possibility in more detail. Recently, it was discovered that the Wnt/b-catenin signaling activator Rspo2 (r-spondin 2) is a marker for the rostral set of mdDA neurons that is encoded by Lmx1a and affected in Lmx1a knockout mice (discussed above). Although no mechanism was shown, Rspo2-deficient mice partly phenocopied the Lmx1a null mutant, since expression of Ahd2, Pitx3 and Th was affected in the rostrolateral mdDA domain, suggesting that modulation of Wnt signaling specifically within this subpopulation partially defines the specific molecular profile of this mdDA neuronal subset [86].

Several fibroblast growth factors (FGFs), as well as their receptors may be involved in mdDA neuron subset specification. Among these is Fgf2, one of the few factors of which deletion results in increased numbers of mdDA neurons in the SNc specifically, possibly due to increased or prolonged neurogenic production of Lmx1a in mdDA progenitor cells or decreased apoptosis at P0. In agreement with these observations, overexpression of Fgf2 in mdDA neurons during development results in decreased numbers of Th+ mdDA neurons [122, 123], and loss of Fgf2 results in increased mdDA fiber outgrowth during nigrostriatal wiring [124]. It should, however, be noted that the same group reported that in 6-OHDA-lesioned mice significantly less mdDA neurons survive in Fgf2-deficient mice, indicating that endogenous Fgf2 may actually protect mdDA neurons in case of neurotoxicity and that maintaining Fgf2 levels within certain constraints may be critical. The same study reported that mice heterozygous for Fgf receptor Fgfr3 display reduced mdDA neuron numbers in the SNc [123]. Although another study described that Fgfr3, as well as Fgfr2, are not required for early patterning of the mid-hindbrain region, and maintenance of mdDA neurons [125], this study was less detailed, and no quantification was performed, indicating that subtle differences may have been overlooked.

The knockout of another Fgf receptor, Fgfr1, has been studied extensively and shows an interesting phenotype. Upon conditional deletion of Fgfr1 (using En1-cre as a driver), a rhombomere 1-to-midbrain transition is observed, resulting in a posterior shift of the mdDA neuron population, with concomitant caudal expansion of Pitx3, Ahd2, and Otx2 expression at E11.5. At E15.5, the total number of mdDA neurons is not affected, but significantly more mdDA neurons are located in a caudal position, and consequently less in a rostral position [126, 127]. An excellent study by the same group described the contributions of the different Fgfrs to mdDA neuron development and maintenance using compound conditional mutant embryos for Fgfr1/2 (En1-cre driver) and a full Fgfr3 knockout. In single Fgfr1 mutants, mdDA neuron number was not affected, but SNc and VTA appeared disorganized. However, upon combined deletion of Fgfr1 and Fgfr2 mdDA neuron cell number in both SNc and VTA were severely reduced by E12.5, and deleted by E18.5. In Fgfr1/2 compound mutants Ahd2 and Pitx3 expression was ablated at E10.5 and E12.5, respectively [128]. Later, the molecular mechanisms behind this phenotype were studied in more detail. It was unveiled that the diencephalic mdDA domain contains not only DA neurons, but also non-DA Pou4f1+ cells, and suggested that in the absence of Fgfr1/2 postmitotic mdDA precursors in the mdDA area acquire a phenotype similar to these neurons [58].

Although these data together suggest that Fgf signaling via Fgfr1/2 is primarily important for anterior-posterior patterning and possibly programming of early mdDA neuron progenitors, an elegant study where a dominant negative form of Fgfr1 was expressed from the Th-promoter, found that the roles of Fgfr1 in the SNc and VTA may not be similar. In these mice, the density of Th+ neurons is equally reduced in the SNc and VTA at birth, but in adult mice this decrease is only observed in the SNc. Concomitant loss of Dat expression in the striatum was observed, whereas Dat expression in the nucleus accumbens was even increased, supporting specific ablation of the nigrostriatal projections [129]. Moreover, these animals displayed a schizophrenia-like phenotype, possibly due to increased activity of VTA DA neurons. Altogether, these findings justify the study of Fgf receptors, in particular Fgfr1, in more detail at later developmental stages (i.e., after initial patterning and expansion of the progenitor pool) to assess possible roles in mdDA subset specification.

Fgf20 is a fibroblast growth factor that may be involved in mdDA subset specification. SNPs in Fgf20 have been associated with PD susceptibility [130]. Moreover, multiple experiments have indicated critical roles of Fgf20 in the development and maintenance of neurons of the SNc subtype [130, 131], but a clear role in mdDA subset specification has not yet been established.

Retinoic acid

A direct transcriptional target of Pitx3 is the retinoic acid (RA) synthesizing enzyme Ahd2 [40] that converts retinaldehyde into RA and is highly expressed in the developing mesodiencephalon [42]. In addition to Pitx3, Ahd2 is also transcriptionally regulated by Nurr1 (Fig. 3a) [132]. Although Ahd2 is expressed in a small ventral subset of VTA neurons [42], we [40, 41] and others [101, 133] have shown that Ahd2 is mainly expressed in the Substantia Nigra pars compacta (SNc) area at both developmental and adult stages. The finding that cultured embryonic midbrains selectively activate RAR-lacZ but not RXR-lacZ constructs, indicates the presence of endogenous atRA in the developing mesodiencephalon [134] and the functional relevance of RA signaling in mdDA neurons has been demonstrated in vitro and in vivo, in Pitx3-deficient stem cell cultures and embryos that lack Ahd2 expression and are therefore deprived of RA signaling in postmitotic mdDA neurons, respectively. RA increases the number of Th+ mdDA neurons in Pitx3-deficient stem cell cultures in vitro [135], and in vivo supplementation of RA during the window of embryonic development where Pitx3 activates Ahd2, bypassed the requirement for Pitx3 and Ahd2, and restored Th expression in the SNc [40]. In addition, RA restored expression of D2R [41] (Fig. 4a–c). Moreover, multiple in vitro studies support a role of retinoids in mdDA neuron development. RA critically induces cellular differentiation of the embryonic ventral midbrain derived DA cell line MN9D [136], retinoid receptor ligand DHA facilitates iPSC differentiation into Th-positive neurons [137], and RA activates Th expression in human neuroblastoma SK-N-BE(2)C cells, most likely through activation of RAR [138]. Although these results suggest a critical role for Nurr1/Pitx3 mediated RA signaling during final mdDA neuron differentiation and subset specification, little is known about the mechanisms by which RA directs subset specification in the developing mesodiencephalon. In vitro evidence, as discussed above, suggests that RA may act via binding to retinoic acid receptors (RARs) (Fig. 3a). Indeed, treatment of Pitx3-deficient explant cultures with the pan-RAR agonist TTNPB increases Th transcript levels [41]. However, despite the detection of RARα and RARβ protein in the adult midbrain [139], the presence of RAR proteins (and their subset specificity) in the developing mdDA area remains to be demonstrated in future studies.

Another outstanding question regarding RA signaling relates to its spatial constriction (Fig. 4d–f). Although Ahd2, and therefore RA production, is spatially constricted to the rostrolateral mdDA subset, free diffusion of RA would result in a RA gradient into the caudal mdDA area, resulting in a ‘transition zone’. Such local fluctuations in RA concentration would make it difficult to induce a sharp boundary between rostrolateral and caudal expression domains, since some cells could adapt an RA-induced molecular profile, whereas others would not, depending on whether or not exceeding an RA treshold [140]. Using the developing hindbrain as a model, where much more is known about the roles of RA in rhombomere specification, Zhang et al. have used an elegant combination of in situ experiments and computational modeling to address this issue. In the zebrafish hindbrain, RA initially induces expression of hoxb1a in rhombomere 4 (r4) and krox20 in r3 and r5. However, somewhat in analogy with Pitx3/En1 (as discussed above), hoxb1a/krox20 gene expression boundaries are not sharp, and such noise in hoxb1a/krox20 expression is required in an elegant feedback system that eventually results in the sharpening of rhombomere boundaries initially created by RA [141]. If a similar mechanism exists in the developing mesodiencephalon, and if this could be dependent on interplay of Pitx3, En1 and possibly other homeodomain TFs (e.g., Pbx-family members that are RA-inducible [142, 143]), remains to be investigated. Another possible mechanism to assure spatial confinement of RA signaling could involve the combined spatial expression of RA-synthesizing enzymes (e.g., Ahd2) and RA-degrading enzymes. In the developing hindbrain, such feedback is mediated by Dhrs3 (dehydrogenase/reductase member 3) that attenuates RA signaling by reducing retinal levels [144] and Cyp26a1 (cytochrome P450, family 26, subfamily A, polypeptide 1) that actively degrades RA [140, 145] (Fig. 4d–f). Indeed, Dhrs3 and Cyp26b1 (cytochrome P450, family 26, subfamily B, polypeptide 1), an RA hydroxylase qualitatively similar to Cyp26a1 [146], display enriched expression in caudal mdDA neurons, as well as surrounding non-mdDA cells at E14.5 (JV Veenvliet, unpublished results) and Cyp26b1 is expressed in a subset of the mdDA area in the adult stage (Allen Brain Atlas, http://mouse.brain-map.org/experiment/show/72081548).

Importantly, multiple lines of evidence suggest that the functional role of RA signaling in the SNc is not limited to developmental stages. Lower mRNA levels of the human Ahd2 homologue RALDH1 have been detected in the SNc of PD patients [147–149] and RA receptor stimulation using two different RAR agonists protected midbrain dopaminergic neurons from inflammatory degeneration in adult mice induced by lipopolysaccharide-activated microglia [139]. This rescue was accompanied by increased tissue levels of BDNF mRNA, and the neuroprotective effect of a RAR agonist was suppressed by TrkB inhibition and anti-BDNF neutralizing antibodies, suggesting that the neuroprotective effect of RAR activation is mediated via enhancement of BDNF expression. A later study by the same group further unraveled the RAR-BDNF molecular cascade and demonstrated an essential role for nNOS-derived nitric oxide (NO) in RAR signaling, by recruiting cyclic GMP and PKG, leading to ERK-dependent BDNF upregulation in mesodiencephalic DA neurons [150]. If similar molecular pathways downstream of RA are involved in mdDA subset specification in developing mdDA neurons remains an open question, although the confinement of both RA and BDNF signaling (as discussed above) to the rostrolateral subpopulation are suggestive of such a mechanism.

Dlk1

A recent study by our lab described how in vivo treatment of pregnant mother mice with RA suppresses the ectopic expression of Dlk1 mRNA and protein expression in the rostrolateral part of developing mdDA neurons in Pitx3-deficient embryos, the exact region that harbors the neurons that are devoid of Th in Pitx3 ablated mice. Dlk1 is a transmembrane protein that has mainly been described for its role in adipocyte and osteoblast differentiation [151, 152]. In adipocytes, Dlk1 inhibits differentiation [153–155]. At adult stages, Dlk1 has been reported to be expressed in Th-positive neurons of both SNc and VTA [156, 157]. However, at the terminal differentiation stage of mdDA development (E14.5), Dlk1 expression is limited to a caudal subpopulation of mdDA neurons [41, 49]. Initial in vitro studies indicated that treatment of VM-derived DA neuron precursors with Dlk1 protein increased DA precursor proliferation during primary culture expansion, whereas treatment during DA neuronal differentiation did not affect the number of Th-expressing neurons. However, interfering with Dlk1 expression during differentiation decreased the expression of some mdDA markers, suggesting a permissive role for Dlk1 during terminal differentiation [158]. In striking contrast with these data, analysis of the mdDA area in Dlk1-null mice in our laboratory suggested an inhibitory role for Dlk1 in the expression of Dat, since in the absence of Dlk1 Dat was ectopically and/or prematurely expressed. Notably, expression of other mdDA markers appeared unaffected in Dlk1-null mice [49]. In vivo, Dlk1 is therefore likely to play a suppressive role in terminal mdDA differentiation, in line with its function in many non-neuronal tissues [152–155]. The suppression of ectopic Dlk1 transcript and protein expression in the mdDA area of Pitx3-deficient embryos that have received embryonic RA treatment, suggests that the increased Dlk1 expression in Pitx3-deficient embryos is the consequence of RA signaling deprivation in the absence of Pitx3. Subset-specific Pitx3/Ahd2-mediated RA signaling may thus play an active role in the local suppression of Dlk1, allowing terminal differentiation of mdDA neurons in a specific subset (Fig. 3a, b). Notably, such interplay between RA signaling and Dlk1 may not be confined to mdDA neurons, since in neuroblastoma cells, both various doses of RA and Dlk1 knockdown induce differentiation [159]. However, although in multiple, both neural and non-neural tissues, critical roles for Dlk1 have been demonstrated in inferring molecular and functional subset-specific cellular characteristics [160–163], the exact role for, and downstream pathways of, Dlk1 in mdDA neuron development and subset specification remain poorly understood and should be studied in more detail. One important question is, if Dlk1 functions as a Notch ligand in DA cells. Studies in drosophila have shown roles for delta/notch singaling in early DA specification events [164, 165], and Dlk1 may act as a non-canonical Notch ligand [166], but if this is the case in vertebrate DA neurons remains to be revealed.

Translational value of fundamental knowledge about subset-specific molecular programming

Obtaining knowledge about molecular programs that regulate mdDA neuron subset specification is not only interesting from a fundamental point of view. It also harbors great translational value, since it may (1) increase our understanding of molecular mechanisms of neuronal vulnerability in PD and (2) offer novel leads for the generation of SNc-subtype-specific neurons from stem cells (SCs).

Translational value for the understanding of PD pathogenesis

The potential translational value of developmental genetics for the understanding of PD pathogenesis is exemplified by the many studies that have found associations between PD and SNPs in genes that are critically involved in mdDA neuron development. Polymorphisms in EN1, PITX3, NURR1, LMX1A, and LMX1B are associated with PD [167–171]. Also, AHD2 expression is downregulated in the SNc of PD patients and may serve as a peripheral biomarkers for diagnosing PD [147, 172]. Likewise, polymorphisms in, and/or downregulated expression of growth factors critically involved in mdDA neuron development and maintenance, such as BDNF, TGF-b, and FGF20 has been observed in PD patients [130, 173–175].

An outstanding question is if these polymorphisms affect mdDA development and subset specification, and if developmental deficits in the specification of mdDA subpopulations are observed in PD patients and related to these SNPs. This is impossible to study in vivo because this would require studying human SNc and VTA ante mortem. However, an interesting alternative would be the generation of iPSCs from patients with such polymorphisms, differentiate them toward the mdDA neuronal lineage, and investigate if these polymorphisms bias iPSC-derived mdDA neurons toward an SNc or VTA phenotype, which would suggests that these polymorphisms result in developmental defects in mdDA subset specification that may precede SNc degeneration in PD.

Do developmental TFs influence the terminal phenotypes of SNc neurons that put them at risk to neurodegeneration?

Could polymorphisms in critical TFs and signaling factors affect later survival of mdDA neurons without affecting subset specification and/or terminal differentiation of mdDA neurons? An excellent recent review by Doucet-Beaupré and Lévesque has suggested that developmental TFs contribute to mdDA neuron survival and maintenance in the adult system by transcriptional and/or translation regulation of nuclear-encoded mitochondrial genes and genes involved in mitochondrial metabolism [31]. Mitochondrial dysfunction has long been suggested to be at the heart of PD pathogenesis [176]. Indeed, several TFs that are implicated in mdDA neuron development and subset specification have recently been shown to regulate mitochondrial function. En1/2 protects adult mdDA neurons from mitochondrial complex I insults by regulating the translation of subunits of this complex, Ndufs1 and Ndufs3 [177]. In fetal muscle progenitors, Pitx2/3 regulates Nrf1, a TF that regulates mitochondrial biogenesis, and in the absence of Pitx2/3 excessive levels of reactive oxygen species (ROS) are observed [178]. In the case of Otx2, more than half of Otx2 target genes are nuclear-encoded mitochondrial mRNAs [179].

Mitochondrial dysfunction in the pathogenesis of PD is closely intertwined with another terminal phenotype of SN neurons that makes them more susceptible to degeneration in PD: the reliance of SN neurons on Ca²⁺ for their autonomous pacemaking. SNc neurons use l-type Ca²⁺ channels [180], as opposed to VTA neurons that rely on sodium with minimal Ca²⁺ channel contribution [181]. The resulting continuous influx of Ca²⁺ puts a strain on mitochondria, that produce ATP to keep intracellular levels of Ca²⁺ within limits, and are involved in Ca²⁺ buffering themselves [182]. Thus, metabolic stress as a result of sustained Ca²⁺ entry in SN, but not VTA neurons, may underly the specific vulnerability of these neurons in PD. In addition, it has been suggested that Ca²⁺ boosts DA synthesis from l-DOPA in SNc, but not VTA neurons, which would imply Ca²⁺ driven increase of toxic DA metabolites in SNc neurons specifically [183], further increasing metabolic and proteostatic stress in these neurons. In addition to its role in RA synthesis, Ahd2 is also involved in the detoxification of DA metabolites, suggesting that misregulation of Ahd2 as a result of reduced (functional) levels of these TFs can further increase oxidative stress in SNc neurons. In line with the Ca²⁺ hypothesis for SNc-specific vulnerability, mdDA neurons that express the Ca²⁺ buffer, calbindin (most VTA neurons, but only a small subset of SN neurons), are relatively spared from neurodegeneration in PD [184].

Another channel that appears to be involved in selective vulnerability of SN neurons is the K-ATP channel. This channel is selectively activated in SN neurons upon toxic challenges and this prevents action potential generation. Studies in K-ATP knockout mice confirmed the instrumental role of these channels in SNc neuron selective vulnerability, since these mice were resistant to MPTP-induced neurodegeneration [185]. Quantitative single-cell analysis has revealed that mRNAs coding for these channels are enriched in SN neurons as compared to VTA neurons. It would be highly interesting to study if genes affecting mitochondrial function and channel composition are differentially expressed in SN as compared to VTA neurons, if such a distinction is already visible during development, and if these possible differences are encoded by subset-specific roles of critical TFs and signaling molecules.

Translational value for cell replacement strategies

In the last decade, many excellent neurodevelopmental studies have focused on the generation of a good cell replacement model by reprogramming of inducible pluripotent SCs (iPSCs) and/or embryonic SCs (ESCs) [186–191]. In most of these studies resemblance of SC-derived neurons with their in vivo counterpart was assessed based on mainly morphological and functional properties, as well as mdDA-specific gene sets [189–193]. Such scarce characterization of iPSC/ESC-derived DA neurons is problematic, since classical differentiation protocols for mouse and human iPSCs and ESCs typically yield a heterogeneous population of DA neurons [194, 195], that is associated with problems that limit the clinical application of cell replacement therapy in PD, such as poor graft survival and tumor formation [195]. These are critical issues that need to be overcome before iPSCs and ESCs can be considered as a safe and successful treatment for PD. Recently, some studies have therefore focused on the resemblance of SC-derived DA neurons with their in vivo counterpart. Roessler et al. [196] performed a detailed analysis of the transcriptomic and epigenetic signature of iPSC-derived DA neurons. Using microarray and Reduced Representation Bisulfite Sequencing (RRBS) technology, the genetic and epigenetic profile of iPSC-derived FACS (fluorescence-activated cell sorting)-purified Pitx3(gfp/+) DA neurons was compared with their in vivo counterpart at various developmental stages. Although the iPSC-derived DA neurons had largely adopted characteristics of their in vivo counterparts that are generally used to assess the resemblance of SC-derived mdDA neurons with true mdDA neurons (e.g., morphology, DA production, functionality in behavior models, and expression of mdDA markers), major relevant deviations in genome-wide gene expression and CpG island methylation profiles were observed. Many genes involved in neurodevelopment were hypermethylated and consequently showed reduced expression levels in iPSC-derived DA neurons as compared to their in vivo counterparts. Moreover, residual expression and a permissive methylation state of fibroblast markers and enriched expression of genesets involved in mesodermal lineage specification was reported, indicative of epigenetic memory of the cells of origin, in line with previous reports that iPSCs are prone to differentiate along their lineage of origin [197, 198]. To what extend the observed epigenetic memory and expression profile deviations interfere with in vivo functionality of SC-derived DA neurons after grafting remains to be investigated in future studies. However, such findings stress the importance of genome-wide screens of SC-derived DA neurons to evaluate their true potential for stem cell therapy in PD.

More studies have focused on global transcriptome analysis of iPSC- and ESC-derived DA neurons. Momcilovic et al. [199] described changes in gene expression during mdDA differentiation of ESCs and iPSCs using bead microarrays at four stages during ESC- and iPSC- differentiation toward mdDA neurons (undifferentiated, neural stem cells, DA precursors, DA neurons). An excellent study by Studer and colleagues [190] described global transcriptome analysis of progeny from Hes5:GFP, Nurr1:GFP, and Pitx3:YFP ESC reporter lines and identified expression of many known and novel DA genes in these ESC-derived DA neurons. Unfortunately, these studies lack a comparison of SC-derived mdDA neurons with their in vivo counterpart. In this regard, Salti et al. [194] took an interesting, although not genome-wide, approach. They performed quantitative comparison of gene expression profiles during DA differentiation in vitro and in vivo (E11.5–E13.5) at five stages of differentiation [ES cells, embryoid bodies (EBs), neural selection, neural patterning, and neural differentiation], and used such profiles to relate gene expression milestones at various stages of differentiation to the efficiency of DA differentiation at a protocol’s endpoint. In this particular study, levels of DA metabolism genes (Th, Vmat2) and critical mdDA TFs (Otx2, En1, and Foxa2) were comparable with E11.5 dissected midbrain tissue. Such in-depth characterization studies that compare the molecular signature of SC-derived mdDA neurons with that of their in vivo counterpart are of critical importance to reveal the true potential of protocols that generate SC-derived mdDA neurons to use in cell replacement strategies. However, all studies that have used cross-comparison approaches, have not taken the existence of multiple mdDA neuronal subsets into account, and this should be addressed in future studies.

Generation of specific mdDA neuron subtypes from stem cells

Given the specific degeneration of SNc neurons in PD, generating neurons of the correct (SNc) subtype from iPSCs/ESCs is an important focus of SC research, and has been proposed as a critical criterium for assuring pluripotent SC quality for cell therapy [200]. In support of this, it was shown that in intrastriatal grafts the SNc component is of critical importance for recovery in a rodent model of PD [201]. Also, it has been demonstrated that, in intrastriatal grafts, distinctive features are retained after transplantation, and can be used to distinguish SNc and VTA subpopulations, that innervate distinct projection areas, providing evidence that axonal outgrowth from these subpopulations is differentially regulated in the grafts. Indeed, fiber outgrowth upon grafting in PD models appears to reflect, at least to some extend, normal innervation patterns, as SNc neurons extend their fibers toward the dorsolateral striatum after transplantation, whereas VTA neurons project to the frontal cortex and other forebrain areas [201–203]. Since the axon terminal network is critical for the release of DA in the dorsolateral striatum, grafted SC-derived mdDA neurons need to have the capacity to innervate the host dorsolateral striatum for functional improvement in PD patients, and this appears to be a unique feature of SNc neurons. Distinct molecular features of SNc and VTA neurons with respect to axon guidance factors, as observed in normal developing mdDA neurons, are likely to form the basis for this different dorsostriatal innervation capacity [8]. Notably, even if this innervation issue could be circumvented, it is not sure whether grafting SC-derived neurons of the VTA subtype would be functionally beneficial, since their distinct molecular makeup is, at least to some extent, reflected by functional diversity, and it is not sure to what degree SC-derived VTA neurons would be flexible enough to adapt SNc neuron functionality upon grafting (for detailed reviews about molecular versus functional diversity in the adult mdDA system we refer to [204, 205]).

Two different strategies can be applied to obtain a homogeneous population of SNc-subtype mdDA neurons (Fig. 5). The first possibility is using a protocol that allows for the specific generation of SNc DA neurons from SCs. Multiple studies have claimed enriched and/or specific generation of SNc-subtype DA neurons using a variety of differentiation strategies [133, 206–210]. Some of these approaches involve over-expression of critical TFs, such as Pitx3 [133], Lmx1a [207] and Lmx1b [208]. However, in virtually all of these studies the successful conversion to the SNc subtype is solely based on the observation that SC-derived DA neurons express the G protein-gated inwardly rectifying K⁺ channel Girk2. Although Girk2 has indeed been reported to exclusively mark SNc neurons [202, 211], later studies have reported that Girk2 is expressed in both SNc neurons, as well as 50–60 % of VTA neurons in humans and mice [74, 212, 213], and therefore, the true potential of all reported approaches in the specific generation of SNc-subtype DA neurons is questionable. Although more in-depth knowledge of the transcriptional programs that direct mdDA subset specification may eventually lead to a protocol to drive mdDA subset specification in a dish (for example by careful titration of TFs, such as Pitx3, En1, and Otx2, and/or application of retinoids), it is conceivable that such an approach is treacherous, since the process of mdDA subset specification appears to be extremely complex, and requires not only detailed knowledge of the transcriptional and signaling machinery, but also detailed knowledge about timing and the critical timeframe during which induction of SNc and VTA subsets can be achieved in vitro.

Fig. 5.

Strategies for the generation of SNc-subtype-specific SC-derived DA neurons. a One option to obtain a pure population of SC-derived mdDA neurons of the SNc subtype is to use a differentiation protocol that specifically generates SNc neurons (‘Directed Differentiation’). However, this requires detailed knowledge about the Gene-Regulatory Network that underlies the specific programming of these neurons, and may involve a complicated strategy, using a combination of transduction of many TFs, and multiple signaling molecules. b–d″ A much easier approach to obtain a purified population of SC-derived SNc neurons is to use a molecular sort strategy. In vivo, neurons of the SNc and VTA can be roughly separated based on their anatomical position (c). However, after in vitro differentiation of SC toward mdDA neurons a heterogeneous pool of both SNc and VTA subtype mdDA neurons is obtained, and it is not possible to distinguish between both subtypes (c′, d′). However, if a SNc-specific surface marker can be identified, live cells can be stained, and SC-derived DA neurons of the SNc subtype can then be purified using FACS (c″, d″). Such an approach is greatly facilitated if it is combined with a fluorescent reporter that marks all DA neurons (e.g., Pitx3-gfp) (c–c″). However, in the absence of such a reporter, a combination of a DA surface marker that stains all SC-derived DA neurons, and an SNc-subtype-specific surface marker would suffice to obtain a purified population of SC-derived SNc DA neurons (d–d″). SC stem cell, TF transcription factor, DA dopaminergic

A second, more straightforward approach is the generation of a heterogeneous pool of SC-derived DA neurons, followed by FACS-purification of mdDA neurons of the SNc subtype. Such an approach requires the identification of SNc-specific cell-surface markers that can be used in flow cytometry. Although SNc-specific cell-surface markers have not been identified yet, such a strategy harbors great promise, since it circumvents the necessity for laboreous optimization of the current protocols that generate heterogeneous populations of DA neurons. Moreover, if an SNc-specific cell-surface marker is combined with an mdDA-specific surface marker, this additionally circumvents the need for mdDA-specific fluorescent reporters in SC-lines (such as Pitx3-GFP) that are relatively easily implemented in vitro and in murine models, but problematic when translated to human cells. A list of such markers has been suggested based on the global expression profiling of Nurr1:GFP and Pitx3:YFP ESC reporter lines by Studer et al. [190], including Chrna6 and Chrnb3 that are expressed in developing mdDA neurons in vivo [51]. Several strategies can be applied to identify SNc- and VTA-subtype-specific cell-surface markers. One possibility is genome-wide expression profiling of SNc versus VTA neurons after purification by laser capture microdissection (after staining for a DA-specific gene), and subsequent mining of these data for subtype-specific surface markers. Such data have been obtained in adult mouse and rat [66, 214, 215], and these data could potentially be used. However, given the typical similarity of SC-derived DA neurons with embryonic DA neurons [196], genome-wide expression data of developing SNc versus VTA neurons may be favorable. Unfortunately, such a resource is currently lacking.

Conclusions and future directions

Although the specific vulnerability and degeneration of SNc neurons in PD was acknowledged long ago [1, 216, 217], the molecular mechanisms that underly such specific vulnerability are still largely unclear. In their recent perspective, Fishell and Heintz argue that, although studying the immense diversity of cell types in the mouse CNS provides many harsh challenges for researchers, the ultimate identification and molecular characterization of neuron subtypes will offer a great potential for therapeutic intervention in neurodegenerative diseases that affect a certain neuronal subtype, such as PD [218]. They stated that “It seems apparent that detailed molecular profiling of the affected cell types during development and disease progression is a necessary step in understanding the molecular consequences of destructive genetic or environmental events”, but that “these studies cannot be pursued without comprehensive information regarding CNS cell types” [218].

The last decade has brought enormous advances in our understanding of the molecular programming underlying the development of ‘default’ mdDA neurons, but relatively few in-depth understanding has been acquired about mechanisms regulating mdDA subset specification in the developing DA system, although important progress has been made, as reviewed here. However, to truly understand the molecular cascades that underly mdDA subset specification, a mixture of clever fate-mapping designs, -omics approaches, single-cell expression analysis, and in-depth spatiotemporal analysis of the expression of critical factors, should be used to identify subset-specific marker genes and pathways.

In this respect, some pioneering work has come from the laboratories of Roeper and Stuber [219, 220]. Lammel et al. first mapped the anatomical distribution of retrogradely labeled Th+ neurons in the mdDA area for different projection areas of the mesocorticolimbic mdDA system, and then used a combination of electrophysiological and molecular techniques to unravel the functional and molecular diversity of these subpopulations. They demonstrated that in the adult system at least two functionally and molecularly distinct types of mdDA neurons exist. A particular unique subset consisted of the mdDA neurons that project to the medial prefrontal cortex and was shown to lack functional D2R receptors. Also, Dat expression levels were not equal among mdDA subsets, with low absolute Dat and relative Dat/Th and Dat/Vmat2 expression ratios observed in mesocorticolimbic mdDA neurons (with more than half of these neurons not expressing Dat protein), but high levels in mesostriatal mdDA neurons and neurons projecting to the mesolimbic lateral shell [220]. This is suggestive of differential DA metabolism in distinct subsets of mdDA neurons. Moreover, at the level of DA synthesis, Aadc and Th may be differentially post-translationally modified as a result of Ca²⁺ influx in SNc as compared to VTA neurons [183]. The study of Lammel et al. also clearly demonstrated that the sole distinction of the mdDA system in SNc and VTA neurons is not exhaustive. Also within the SNc and VTA, multiple subsets exist with distinct molecular profiles and electrophysiological properties [212, 220] that reflect functional diversity. For example, Stamatakis et al. [219] recently described a subset of VTA mdDA neurons that co-release GABA (Gamma-aminobutyric acid) from their terminals, and express GABA neuron markers. Further form-to-function profiling, by relating molecular diversity with functional diversity is therefore critical in future research, and it will be of utmost interest to determine to what extend this further segregation of the mdDA area in distinct subsets is reflected in the developing mdDA system.

As a final remark, we note that unveiling the degree of conservation of subset-specific programming is critical to select for those mechanisms that hold translational value. Eventually, increased understanding of mdDA subset specification may find its way to the clinic, given the potential importance of such knowledge to generate SC-derived DA neurons that can safe and soundly replace those neurons lost in PD.

Abbreviations

CNS

Central nervous system

DA

Dopaminergic

DOPAC

Dihydroxyphenylacetic acid

DOPAL

Dihydroxyphenylacetaldehyde

ESCs

Embryonic stem cells

FACS

Fluorescence-activated cell sorting

GABA

Gamma-aminobutyric acid

GFP

Green fluorescent protein

iPSCs

Inducible pluripotent stem cells

IsO

Isthmic organizer

mdDA

Mesodiencephalic dopaminergic

miRNA

microRNA

GRN

Gene-regulatory network

MHB

Mid-hindbrain border

P1/2/3

Prosomere 1/2/3

PD

Parkinson’s disease

R3/5

Rhombomere 3/5

RA

Retinoic acid

SCs

Stem cells

SNc

Substantia nigra, pars compacta

TF

Transcription factor

VTA

Ventral tegmental area