Wnts are a highly conserved family of lipid-modified glycoproteins that work as morphogens to activate several signaling pathways, leading to remodeling of the cytoskeleton and the regulation of gene transcription. Wnt signaling regulates multiple cellular functions and cell systems, including the development and maintenance of midbrain dopaminergic (mDA) neurons. These neurons are of considerable interest for regenerative medicine because their degeneration results in Parkinson's disease (PD). This review focuses on new advances in understanding key functions of Wnts in mDA neuron development and using novel tools to regulate Wnt signaling in regenerative medicine for PD. Particularly, recent reports indicate that appropriate levels of Wnt signaling are essential to improve the quantity and quality of stem cell- or reprogrammed cell-derived mDA neurons to be used in drug discovery and cell replacement therapy for PD.
Wnts and Wnt signaling pathways
The name Wnt1 comes from the fusion of two words and two areas of research, developmental studies on the wingless gene that controls segment polarity during larval development in Drosophila (Nüsslein-Volhard and Wiechaus, 1980) and cancer studies on Int-1, a gene activated by integration of mouse mammary tumor virus proviral DNA in virally induced breast tumors (Nusse and Varmus, 1982). The realization that these two genes were homologues (Rijsewijk et al., 1987) started this fascinating field of research that blends development and disease. The identification of Int-1-related genes led to the standardization of the nomenclature and the birth of the Wnt family as such (Nusse et al., 1991). Moreover, with the advent of homologous recombination technology, Wnt1 was the first gene to be deleted in mice (McMahon and Bradley, 1990; Thomas and Capecchi, 1990).
Wnts form a large family of secreted lipid-modified glycoproteins that is largely conserved through evolution, from the first multicellular organisms such as the Hydra that has 13 Wnts, to mammals such as mice and humans that have 19 family members. Wnts are proteins of about 40 kDa, which contain 22 cysteine residues that form disulfide bridges and assemble into a globular structure. The structure of Xenopus Wnt8 was recently resolved and found to resemble ‘an index finger and a thumb’ (Janda et al., 2012). Notably, the thumb-like structure is of critical importance for the Wnt molecule, as it is the site of palmitoylation, a lipid modification that it is required for Wnt secretion, its interaction with the receptor, and its function (Willert et al., 2003; Takada et al., 2006).
Wnts are morphogens, molecules capable of forming concentration gradients and activating membrane-bound receptors in target cells, in a concentration-dependent manner. Wnts regulate multiple essential processes for animal development, including the determination of primary body axis, different aspects of polarity (anteroposterior, dorsoventral, and left-right), limb development, organogenesis, and patterning of different tissues (Gray et al., 2011; Wang et al., 2012). Moreover, within an organ or tissue, Wnts regulate multiple functions such as stem cell self-renewal, proliferation, cell fate specification, neurogenesis, differentiation, survival, cell polarity, migration, and neuritogenesis (Salinas and Zou, 2008; Wang et al., 2012). Wnts are also important for the maintenance of adult cell and tissue homeostasis, and deregulation of Wnt signaling is associated with multiple diseases including developmental abnormalities (Baron and Kneissel, 2013), different types of cancer (Anastas and Moon, 2013), and neurodegenerative disorders (Inestrosa and Arenas, 2010).
Wnt signaling pathways
Wnts are known to regulate three main different pathways, the Wnt/β-catenin (CTNNB1) signaling pathway (also referred to as canonical Wnt signaling), the Wnt/planar cell polarity (PCP) pathway, and the Wnt/calcium (Ca2+) pathway (both referred to as non-canonical or Wnt/β-catenin-independent pathways). It is important to note that the capacity of Wnts to activate distinct signaling pathways and its branches is cell type- or context-dependent, as described for Wnt5a (Schulte et al., 2005; Mikels and Nusse, 2006). Particularly, during development, Wnt gradients and Wnt signaling components are very dynamically regulated in a specific temporal and spatial manner, as described in the developing midbrain (Rawal et al., 2006; Fischer et al., 2007). Thus, Wnt signaling critically depends on the target cell type, the context of Wnt signaling components in the cell, including their post-translational modification and compartimentalization, the developmental stage, the position in space with respect to Wnt gradients, and the presence of extracellular modulators of Wnt signaling (Kikuchi et al., 2011; Niehrs, 2012).
In Wnt/β-catenin signaling, the best characterized pathway, Wnts bind to the cysteine rich domain of the seven-transmembrane receptor Frizzled (Fz, a family formed by 10 members) and the low-density lipoprotein receptor-related protein 5 or 6 (Lrp5/6). The binding event triggers the recruitment of Dishevelled (Dvl1–3) and Axin to the membrane and their interactions with the intracellular domain of Fz and the phosphorylated tail of Lrp5/6, respectively. This causes either the dissociation of the so-called destruction complex, which is formed by glycogen synthase kinase 3β (GSK3β), adenomatosis polyposis coli (APC), Axin, and casein kinase 1α (CK1α), or its recruitment to the membrane (Clevers and Nusse, 2012). This results in the inhibition or sequestration of GSK3β, thus preventing the phosphorylation of β-catenin, its ubiquitination by β-TrCP (β-transducin repeat-containing protein), and its subsequent proteasomal degradation. β-catenin in its active dephosphorylated state is accumulated, translocates to the nucleus, and binds to transcription factors, such as the T cell factor 4 and limphoid enhancer-binding factors, to displace the repressor Groucho/TLE and turn on the transcription of Wnt target genes, such as c-myc, cyclinD1, and Axin2 that regulate proliferation and differentiation in a variety of cell types.
The Wnt/calcium pathway is generally activated by Wnt5a or Wnt11 binding to the Fz receptor, followed by recruitment of Dvl. This pathway is thought to involve trimeric G proteins, the activation of phospholipase C, and the generation of diacylglycerol and Inositol 1,4,5 triphosphate (IP3). This leads to the release of calcium from intracellular stores, which in turn activates calcium-sensitive proteins such as protein kinase C, calcium-calmodulin-dependent kinase II (CamKII), calpain, and the phosphatase calcineurin, leading to the regulation of the nuclear factor of activated T cells (NFAT) (Gao and Chen, 2010; Kikuchi et al., 2011). Notably, NFAT can positively regulate neurogenesis by binding Dvl and inhibiting Wnt/β-catenin signaling (Huang et al., 2011), and CamKII regulates dendritic spines and synaptic strength (Ciani et al., 2011).
The Wnt/PCP pathway is the most complex Wnt pathway, which functions to establish asymmetrical distribution of proteins within a cell in order to organize the polarity of cells in a layer or plane. This pathway is generally activated by Wnt5a or Wnt11 binding to Fz2, 3, 6, or 7 alone or together with a co-receptor. Multiple Wnt/PCP co-receptors have been described, including the receptor tyrosine kinase-like orphan receptor 1/2 (Ror1/2), receptor-like tyrosine kinase (Ryk), protein tyrosine kinase 7 (PTK7), Syndecan, and Glypican 4/6. Core components of the Drosophila PCP pathway are conserved in mammals, including Fz, two transmembrane proteins, Van Gogh-like 1 and 2 (Vangl1/2) and the Cadherin, EGF, LAG seven-pass G-type receptors 1–3 (Celsr1–3), and cytoplasmic signaling components Dvl1–3, prickle1/2 (Pk1/2), and Diversin or Inversin (Gray et al., 2011; Kikuchi et al., 2011). In Drosophila, Fz directly interacts with core PCP components present in distal membranes (Dvl and Celsr-Inversin) rather than Vangl-Pk present in proximal membranes. However, in mammals, Wnt-Fz interacts indirectly with Vangl2 via Wnt co-receptors, such as Ror2 (Gao et al., 2011) or Ryk (Andre et al., 2012). The scaffolding protein Dvl has been found to interact with multiple signaling components including other core PCP components (Celsr, Inversin, Vangl, and Pk) and Wnt/PCP co-receptors (Ryk, Musk, PTK7, and Syndecan) (Gao and Chen, 2010). Thus a wide array of complexes can be formed and define different Wnt/PCP signaling branches. Moreover, it should be noted that the interaction of Wnt-Fz with some co-receptors (Ryk, PTK7, and Glypican) can lead to not only Wnt/PCP but also Wnt/β-catenin activation, depending on the context. Co-receptors that have been described to activate only Wnt/PCP signaling include Ror1/2 and Syndecan (Niehrs, 2012).
In Wnt/PCP signaling, Dvl activates small GTPases, such as Rac1, RhoA, and Cdc42, thus remodeling the actin cytoskeleton. The Dvl–Daam1–RhoA–Rock pathway, Daam1 (Dvl-associated activator of morphogenesis 1) becomes active after binding to Dvl. Both Dvl and Daam1 interact with a guanine nucleotide exchange factor (GEF) called weak-similarity GEF (WGEF) to form a Rho-GTP complex that in turn activates ROCK kinase and remodels the cytoskeleton (Habas et al., 2001). In the Dvl–Rac1–JNK pathway, Dvl interacts directly with Rac1 (Ras-related C3 botulinum toxin substrate 1) and the Rac1 GEF, Tiam1 (T-cell lymphoma invasion and metastasis 1) to form a Rac1-GTP complex (Cajánek et al., 2013), a process that requires β-arrestin and can be inhibited by Casein kinase 1δ/1ε/2 (Bryja et al., 2008). Rac1 in turn activates the downstream effector c-Jun N-terminal kinase (JNK) to regulate processes, such as dendrite growth (Rosso et al., 2005) and cell polarity movements during gastrulation (Habas et al., 2003). Wnt5a has also been reported via Ror2 and PI3 kinase to activate Cdc42 and the JNK signaling cascade, which via the transcription factors ATF2 and c-Jun regulate the expression of paraxial protocadherin and convergent extension in Xenopus (Schambony and Wedlich, 2007).
These Wnt pathways interact with one another at multiple levels, not only because they share ligands and the Fz-Dvl module, but also because there are multiple mechanisms to regulate each other. Wnt/β-catenin-independent pathways can block Wnt/β-catenin pathways via the Wnt5a-induced activation of Siah, a ubiquitin E3 ligase that degrades β-catenin (Topol et al., 2003) or the inhibition of Dvl by NFAT (Huang et al., 2011). Converse mechanisms to inhibit Wnt/β-catenin-independent pathways by Wnt/β-catenin-dependent pathways also exist. For instance, Lrp5/6 prevents the activation of the Wnt/PCP/Rac1 pathway and its deletion leads to neural tube closure defects (Bryja et al., 2009).
R-Spondins 1–4 are a family of secreted proteins that were identified as Wnt activators (Kazanskaya et al., 2004) and are currently known to work as positive regulators of not only Wnt/β-catenin but also Wnt/PCP signaling (Ohkawara et al., 2011). R-Spondins bind to three different types of receptors, Syndecan, Leucin-rich repeat-containing G protein-coupled receptor (LGR 4–6), and the RING finger proteins RNF43 and ZNRF3. R-Spondin binding to Syndecan or LGRs induces clathrin-mediated endocytosis. While Syndecan activates only Wnt/PCP signaling, LGRs can activate both Wnt/PCP and Wnt/β-catenin signaling (Carmon et al., 2011; de Lau et al., 2011; Glinka et al., 2011; Ohkawara et al., 2011). R-Spondin binding and internalization of RNF43 and ZNRF3 activates Wnt signaling by inhibiting the activity of these receptors as membrane E3 ubiquitin ligases involved in the degradation of Fz (Hao et al., 2012; Koo et al., 2012).
Wnt inhibitors interact with Wnts and Wnt receptor/co-receptors to modulate Wnt signaling (Cruciat and Niehrs, 2013). Secreted Frizzled-related protein 1–5 (sFRP1–5) and Wnt inhibitory factor (WIF) bind to Wnts and prevent their interaction with Wnt receptors, resulting in inhibition of Wnt/β-catenin-dependent and/or independent signaling. It has been recently suggested that while high levels of sFRPs inhibit Wnt signaling, low levels may enhance it by presenting Wnts to their receptors. For instance, the deletion of sFRP1 and sFRP2 phenocopies the midbrain dopaminergic (mDA) neuron differentiation defect observed in the Wnt5a null mice (Kele et al., 2012), indicating that low levels of sFRPs are necessary to enhance Wnt/PCP signaling in the midbrain.
Dickkopfs (Dkk) are Wnt inhibitors that bind to Wnt receptors/co-receptors. Dkk1, 2, and 4 bind to Lrp5/6 and another type of receptors, Kremen 1 and 2 (Mao et al., 2002), forming a ternary complex that inhibits Wnt/β-catenin signaling (Ellwanger et al., 2008). It should be noted that Dkk3 modulates TGFβ signaling, and Dkk2, in the absence of Kremen, can activate Wnt/β-catenin signaling. At a functional level, Dkk1 null mutants lack head structures anterior to the mid-hindbrain boundary and Dkk1 is required for midbrain and limb morphogenesis (Mukhopadhyay et al., 2001; Ribeiro et al., 2011), emphasizing the importance of Wnt/β-catenin inhibition during development. Other inhibitors such as the Wise/Sclerostin (SOST) family bind to Lrp5/6 and compete for binding with Wnt (Semënov et al., 2005) and Insulin-like growth factor binding protein 1, 2, 4, and 6 (IGFBPs) bind to Fz8 and Lrp6 to block Wnt/β-catenin signaling (Zhu et al., 2008).
Wnt signaling in mDA neuron development
mDA neurons are localized in the ventral part of the midbrain and produce the neurotransmitter dopamine. They were first identified by the content of catecholamines (Dahlström and Fuxe, 1964) and subsequently by the expression of the rate-limiting enzyme in dopamine synthesis, tyrosine hydroxylase (TH) (Hökfelt et al., 1984). mDA neurons are mainly classified into two anatomically, molecularly, and functionally distinct subtypes, A9 substantia nigra pars compacta (SNc) and A10 ventral tegmental area (VTA). SNc neurons are positioned anterior and lateral to the VTA, express the G-protein-regulated inward-rectifier potassium channel 2 (GIRK2), project to the dorsal striatum (caudate and putamen) via the nigrostriatal pathway, and control motor function. VTA neurons express calbindin and cholecystokinin (CCK), project to the prefrontal cortex and ventral striatum (nucleus accumbens, amygdala, olfactory tubercle, and prefrontal cortex) via the mesocorticolimbic pathway, and control multiple other functions, such as emotions, motivation, and reward (Carlsson, 2001). SNc and VTA neurons are born in the midbrain floor plate (mFP). Birth-dating using 3H-thymidine and TH-immunostaining in mice has shown that while SNc mDA neurons are born between embryonic day (E)10 and E13, VTA mDA neurons are born from E11 to E14 (Bayer et al., 1995). Genetic fate-mapping experiments have indicated that both SNc and VTA DA neurons are generated by proliferating progenitors in the ventricular zone (VZ) of the mFP, which can be identified by their expression of two key morphogenic factors, Wnt1 (Brown et al., 2011) and another lipid-modified glycoprotein, Sonic hedgehog (Shh) (Joksimovic et al., 2009b; Blaess et al., 2011; Hayes et al., 2011). Interestingly, Wnt1-expressing progenitors labeled from E7.5 to E13.5 contribute to mDA neurons much more than other cell types in the ventral midbrain (VM), with a similar time-line and extent to SNc and VTA mDA subtypes (Brown et al., 2011). Progenitor cells in the VZ of the mFP, including radial glia-like cells (Bonilla et al., 2008), undergo neurogenesis and give rise to postmitotic migratory cells. These cells express the chemokine receptor CXCR4 and respond to meningeal-derived CXCL12 by migrating radially from the intermediate zone (IZ) towards the marginal zone (MZ) of the mFP (Yang et al., 2013b) where they become mature mDA neurons.
Wnt1 signaling in mDA neuron development
The VM is patterned by the coordinated action of two signaling centers regulated by transcription factors that produce morphogens and secreted factors that form signaling gradients, such as Shh in the mFP as well as fibroblast growth factor 8 (FGF8) and Wnt1 in the isthmic organizer (Acampora et al., 2001; Liu and Joyner, 2001; Wurst and Bally-Cuif, 2001; Placzek and Briscoe, 2005). The first factor, Shh, provides ventral identity to cells in the neural tube and its expression in the mFP, from E8.5 to E11.5, is required and sufficient for mDA neuron development in the midbrain-hindbrain region (MHR) (Hynes et al., 1995; Ye et al., 1998). The second factor, FGF8, is expressed by the hindbrain side of the MH boundary (MHB), contributes to antero-posterior identity, and its intersection with Shh creates the induction site for mDA neurons (Ye et al., 1998). The third factor, Wnt1, is expressed between E8 and E8.5 in the presumptive midbrain and extends to the anterior border of the MHB and the roof-plate. Between E10.5 and E12.5, Wnt1 is detected in two bands at both sides of the mFP (Prakash et al., 2006). Notably, analysis of Wnt1 null mice revealed that Wnt1 is required for the development of the MHR (McMahon and Bradley, 1990; Thomas and Capecchi, 1990) and the generation of mDA neurons by a mechanism involving the expression of the homeoprotein engrailed1 (En1) (Danielian and McMahon, 1996). Interestingly, FGF8 is required to induce or maintain Wnt1 expression and loss of FGF8 also results in a deletion of the MHR (Chi et al., 2003). Importantly, the capacity of FGF8 to induce ectopic mDA neurons (Ye et al., 1998) was found to require the expression of Wnt1 (Prakash et al., 2006), indicating that the inductive activity of FGF8 is actually mediated by Wnt1 and that the two key morphogens for mDA neuron development are thus Shh and Wnt1.
Wnt1 conditioned media was found to promote VM progenitor expansion and allow for their subsequent differentiation into mDA neurons (Castelo-Branco et al., 2003), suggesting that Wnt1 is sufficient for the development of mDA neurons in vitro. The use of GSK3β inhibitors also allowed progenitor to differentiate into mDA neurons in VM cultures (Castelo-Branco et al., 2004). However, treatment of mDA progenitors with Wnt3a, a Wnt that also activates Wnt/β-catenin signaling in mDA neurons (Bryja et al., 2007), expanded VM progenitors but did not allow their differentiation (Castelo-Branco et al., 2003). Interestingly, deletion of the Wnt co-receptor Lrp6 transiently impaired mDA neuron differentiation at E11.5, but not progenitor proliferation (Castelo-Branco et al., 2010). Wnt1 overexpression under the control of the endogenous En1 promoter demonstrated that Wnt1 promotes the proliferation of mDA progenitors and their subsequent differentiation in vivo (Panhuysen et al., 2004). Further analysis of this transgenic mouse revealed a role of Wnt1 in establishing a mDA domain in the FP, as shown by the expression of orthodenticle homologue 2 (Otx2), which in turn represses Nkx2.2 and maintains Wnt1 in the hindbrain (Prakash et al., 2006). These events are followed at E10.5 by the expression of markers for mDA precursor, such as Aldh1a1, an enzyme involved in retinoic acid synthesis and Nurr1/Nr4a2 (Prakash et al., 2006), an orphan nuclear receptor required for the terminal differentiation and survival of all mDA neurons (Zetterström et al., 1997). Finally, ectopic mDA neurons are induced in the hindbrain FP (Prakash et al., 2006), as assessed by the subsequent expression of TH and the paired homeodomain transcription factor Pitx3, which is required for mDA differentiation and the survival of only SNc DA neurons (Hwang et al., 2003; Nunes et al., 2003; van den Munckhof et al., 2003; Smidt et al., 2004; Maxwell et al., 2005). On the other hand, deletion of Wnt1 resulted in decreased proliferation, but did not affect Otx2 expression, which, despite the segmental deletion of the midbrain, was present in the area where very few mDA neurons transiently arise (Prakash et al., 2006; Andersson et al., 2013; Yang et al., 2013a). Surprisingly, mDA proliferating progenitors, identified by the presence of the LIM homeobox transcription factor 1a (Lmx1a, Andersson et al., 2006a), are not found in the mFP but rather in the basal plate (BP) where they give rise to few ectopic Nurr1+ and TH+ postmitotic neuroblasts and neurons (Andersson et al., 2013; Yang et al., 2013a). Interestingly, the expression of Aldh1a1, Pitx3, or the dopamine transporter Slc6a2/Dat was not detected in Wnt1−/− mice. However, the few TH+ mDA cells that were detected at E11.5 were lost by E12.5, indicating that mDA neurons were misspecified and subsequently died (Prakash et al., 2006). In sum, these studies indicated an important role of Wnt1 in expanding and specifying DA progenitors in the mFP, promoting their differentiation, and finally maintaining the survival of mDA neurons (Figure 1).
More recently, β-catenin (Joksimovic et al., 2009a; Tang et al., 2009, 2010) and Wnt1 (Andersson et al., 2013) were found to inhibit Shh expression and promote mDA neurogenesis in vivo (Figure 1). Wnt1/β-catenin can regulate mDA neurogenesis in the mFP via four mechanisms: (i) inhibiting the expression of Shh in the mFP and decreasing progenitor proliferation (Joksimovic et al., 2009a; Tang et al., 2009); (ii) regulating the expression of Otx2 and inhibiting alternative fates (Prakash et al., 2006; Joksimovic et al., 2009a); (iii) regulating the expression of Lmx1a mRNA and protein in the mFP (Joksimovic et al., 2009a; Andersson et al., 2013; Yang et al., 2013a) and promoting the specification and neurogenesis (via Msx1) ) of mDA progenitors (Andersson et al., 2006a); and (iv) allowing the mFP expression of the proneural genes mouse achaete-schute homologue 1 (Mash1) and neurogenin 2 (Ngn2) (Joksimovic et al., 2009a; Andersson et al., 2013), which are required for mDA neurogenesis (Andersson et al., 2006b; Kele et al., 2006). Interestingly, Otx2 is also required for the expression of Lmx1a, Msx1, Mash1, and Ngn2 (Omodei et al., 2008), indicating that Otx2, Wnt1/β-catenin signaling and Lmx1a work in concert to regulate mDA neurogenesis. Moreover, conditional deletion of one allele or overexpression of Otx2 in mDA neurons was found to, respectively, increase or decrease the levels of GIRK2 and glycosylated DAT, two phenotypic features of SNc DA neurons (Di Salvio et al., 2010). Consistent with the importance of gene dosage, the Wnt1 hypomorphic mice, swaying (Wnt1SW/SW), showed a complete loss of VTA mDA neurons, which was phenocopied by an Fgf8 conditional knockout (En1Cre) in which Fgf8 is only detected from E8.5 to E9 (Ellisor et al., 2012). These results indicate that the actual level of Wnt1 signaling (regulated by FGF8) may control mDA neuron subtype specification.
Despite Wnt1 signaling controlling most developmental processes in the VM, neither Wnt1 nor β-catenin/Ctnnb1 deletion leads to any alteration in the expression of the forkhead box A2 (Foxa2) gene (Joksimovic et al., 2009a; Andersson et al., 2013). Foxa2 is downstream of Shh and together with the forkhead box A1 (Foxa1) gene, is dose-dependently required for DA neurogenesis and differentiation (Ferri et al., 2007), particularly for the acquisition of a SNc DA neuron phenotype (Stott et al., 2013). This finding indicates that subtype specification is controlled not only by Otx2 and Wnt1, but also by Foxa1/2, and possibly by the balance between these factors. Another gene that is not under the control of Wnt1/β-catenin is the LIM homeobox transcription factor 1b (Lmx1b) (Chung et al., 2009; Joksimovic et al., 2009a). Conversely, Lmx1b directly regulates Wnt1 (Chung et al., 2009) and maintains the expression of Wnt1 in the isthmic organizer (Adams et al., 2000). Surprisingly, Lmx1b null mice showed a relatively mild phenotype consisting of a reduced expression of Wnt1, Pitx3, and TH, followed by cell loss by E16, as Nurr1+ cells fail to initiate the expression of TH and Pitx3 (Smidt et al., 2000; Deng et al., 2011). However, double Lmx1a and Lmx1b null mice revealed a loss of Wnt1 expression in the mFP, a loss of proliferative progenitors, and a decrease in postmitotic Nurr1+, Pitx3+, and TH+ cells (Deng et al., 2011; Yan et al., 2011). This phenotype, which largely resembles that of Wnt1−/− mice, indicates that Lmx1a and Lmx1b redundantly regulate key aspects of mDA neuron development.
Recent ChIP experiments have allowed us to gain even further understanding of the developmental pathways regulated by Wnt1 during mDA neuron development. First, an autoregulatory loop in which Wnt1/β-catenin and Lmx1b directly regulate and are regulated by Lmx1a was described (Chung et al., 2009). In this network, β-catenin directly regulated Otx2 and importantly, Lmx1a/b controlled mDA neuron differentiation and survival at least in part by a direct regulation of Nurr1 and Pitx3. More recently, Foxa2 was found to directly regulate in a positive manner not only Shh, but also Lmx1a/b, Msx1, and Ferd3l/Nato3, a basic helix–loop–helix gene that can regulate Ngn2 (Metzakopian et al., 2012). In sum, all evidence currently available indicates that Wnt1 is absolutely required for the development of the MHR and regulates multiple steps in mDA development including patterning, specification, proliferation, neurogenesis, differentiation, and survival. Recent studies indicate that the development of mDA neurons greatly depends on the balance between two regulatory networks described above, the Shh-Foxa2 and the Wnt1/β-catenin-Lmx1a-Lmx1b-Otx2 network.
Another Wnt capable of activating the Wnt/β-catenin pathway is Wnt2, which binds and activates Lrp5/6 resulting in the phosphorylation of Lrp5/6 and Dvl-2/3, as well as increased levels of dephosphorylated active β-catenin in mDA cells (Sousa et al., 2010). Administration of purified Wnt2 protein to VM progenitor culture increased the proliferation, the number of Nurr1+ cells and the proportion of mDA neurons (%TH/Tuj1) in the culture. Moreover, deletion of Wnt2 decreased progenitor proliferation in the VM resulting in a reduction in the number of Nurr1+ and TH+ cells (Sousa et al., 2010), indicating a role of this Wnt in VM and mDA neuron development, which was similar to that of Lrp6 (Castelo-Branco et al., 2010).
Wnt5a in mDA neuron development
Wnt5a, the second best studied Wnt in the developing VM, does not activate Wnt/β-catenin signaling in DA cells, but rather Wnt/PCP/Rac1 signaling (Schulte et al., 2005; Andersson et al., 2008; Bryja et al., 2008; Cajánek et al., 2013). Wnt5a is expressed in this region from E9.5, being progressively restricted to the midline, and expressed in few TH+ cells in the VTA by E13.5. Notably, Wnt5a is expressed not only in VZ progenitors of the mFP and BP, but also in Nurr1+ and TH+ cells in the mFP at E12.5 (Andersson et al., 2008), suggesting a specific function of this Wnt in mDA neuron development. Wnt5a is also expressed by other cell types, such as midbrain radial glia and astrocytes (Castelo-Branco et al., 2006) as well as meningeal cells (Hayashi et al., 2008). Gain-of-function experiments have revealed that Wnt5a decreases the proliferation of mDA progenitors, enhances the differentiation of Nurr1+TH− cells into Nurr1+TH+ neurons and promotes neuritogenesis of TH+ cells in vitro (Castelo-Branco et al., 2003; Schulte et al., 2005; Bryja et al., 2007). Accordingly, deletion of Wnt5a results in enhanced progenitor proliferation, which is accompanied by an overproduction of Nurr1+TH− postmitotic cells and a nearly normal number of TH+ cells (Andersson et al., 2008). Moreover, Wnt5a mutant mice exhibit morphogenesis defects, manifested by a lateral expansion of the Shh, Foxa2, Lmx1a, and Ngn2 expression domains in the VM, a shortening of the engrailled1 domain, a broadening of the Pitx3 expression domain, and a lateral expansion and anterior-posterior shortening of TH+ mDA neurons due to convergent-extension defects (Andersson et al., 2008) (Figure 1). It also regualtes DA axon growth and guidance both in vitro and in vivo (Blakely et al., 2011). Moreover, deletion of the Wnt receptor Fz3, the Wnt/PCP co-receptor Celsr3 or Vangl2 (Looptail mice) results in decreased rostral and aberrant caudal DA projections, as well as lack of innervation of the striatum in Fz3 and Celsr3 mutants (Fenstermaker et al., 2010). Additionally, Fz3 and 6 are also required for midbrain morphogenesis (Stuebner et al., 2010) and Ryk was recently found to mediate some of the effects of Wnt5a on mDA neurons in vitro and regulate mDA neurogenesis as well as axon morphogenesis in vivo (Blakely et al., 2013).
Interactions between Wnt1 and Wnt5a
Wnt1 and Wnt5a are known to activate Wnt/β-catenin and Wnt/PCP signaling pathways, respectively in mDA neurons. Individual deletion of Wnt1 or Wnt5a revealed that these two Wnts mainly regulate different functions during mDA neuron development (Inestrosa and Arenas, 2010). One clear example is the regulation of mDA progenitor proliferation, which is increased by Wnt1 (Andersson et al., 2013) and decreased by Wnt5a (Andersson et al., 2008), and as expected, the loss of proliferation in Wnt1 null embryos can be compensated by the additional deletion of Wnt5a in double knockout mice (Andersson et al., 2013) (Figure 1). However, analysis of double Wnt1 and Wnt5a knockout mice also revealed an unexpected level of cooperation between these two Wnts to potentiate each other in a manner that was not anticipated by the analysis of individual null mutant mice. For instance, deletion of Wnt1 and Wnt5a revealed that the loss of Wnt1 worsened the Wnt5a/PCP morphogenesis phenotypes observed in Wnt5a−/− mice, resulting in increased mediolateral and dorsoventral distribution of DA neurons as well as A–P shortening of the TH+ domain in double null mutants (Andersson et al., 2013). Similarly, while loss of Wnt5a did not affect neurogenesis or the number of TH+ mDA neurons (Andersson et al., 2008), deletion of both Wnt5a and Wnt1 worsened the DA neurogenesis and differentiation phenotypes found in Wnt1−/− mice (Andersson et al., 2013). Combined, these results indicate that Wnt1 and Wnt5a interact and cooperate in a more extensive and complex manner than anticipated, to regulate mDA neuron development (Figure 1). This cooperation can be explained by some shared signaling components between the Wnt/β-catenin and Wnt/PCP pathways, such as Fz and Dvl, but also by their interaction with common targets involved in the regulation of the cytoskeleton, cell cycle, or neurogenesis. Future experiments should focus on the identification of such molecular targets that are common to both pathways and key for the development of mDA neurons.
Wnt inhibitors are also capable of regulating mDA neuron development. Surprisingly, deletion of Dkk1 or sFRP1/2 did not result in Wnt gain-of-function phenotypes, but rather loss-of-function-like phenotypes. Deletion of Dkk1, which results in the loss of all brain structures anterior to the midbrain, caused a very severe loss of mDA neurons and morphogenesis defects by E17.5 (Ribeiro et al., 2011). Moreover, heterozygous embryos exhibited differentiation defects at E13.5 resulting in a reduced number of mDA neurons that persisted at E17.5. These results show that Dkk1 is required for mDA differentiation and morphogenesis. Similarly, double sFRP1;sFRP2 null mice phenocopied the differentiation and the morphogenesis phenotypes of the Wnt5a null mutant (Kele et al., 2012), indicating that the low endogenous levels of these two sFRPs in the developing midbrain are actually required for Wnt/PCP signaling. Accordingly, in vitro treatment of primary midbrain progenitors with low concentrations of sFRP2 promoted mDA neuron differentiation and only in the presence of very high concentrations of sFRPs did they act as Wnt inhibitors (Kele et al., 2012). These results emphasize the importance of the so-called Wnt inhibitors to actually maintain and modulate Wnt signaling and promote proper mDA neuron development.
Parkinson's disease (PD) is the most common motor neurodegenerative disorder. At clinical level, PD is characterized by tremor, rigidity, akinesia, postural instability, and gait disturbances. The cause of PD is unknown in the majority of the cases, being <10% familial genetic forms and the rest considered as sporadic PD. It is currently thought that environmental factors, e.g. pesticides (rotenone) or toxins (MPTP), together with genetic factors (mutations, haplotypes, and polymorphisms) play a role in PD (Thomas and Beal, 2007; Singleton et al., 2013). Genetic forms of PD are rare, but they have provided very valuable information about the genes and signaling pathways involved. Mutations have been identified in genes encoding either structural cellular components (α-synuclein) or signaling components including kinases, such as leucine-rich repeat kinase 2 (LRRK2) and PTEN-induced putative kinase 1 (PINK1), ubiquitin ligases (Parkin), mitochondrial proteins (DJ1), and lysosomal proteins, such as glucocerebrosidase (GBA1). While some of these mutations are associated with autosomal dominant PD (α-synuclein and LRRK2), others are recessive (PINK1, Parkin, and DJ1) or risk (GBA1) alleles (Singleton et al., 2013). The penetrance of these mutations is variable, for instance, increasing with age in the case of LRRK2. These mutations, as well as environmental factors, seem to affect common cellular mechanisms leading to mitochondrial dysfunction, oxidative stress, protein misfolding, and impairment of the ubiquitin-proteasome and autophagy-lysosomal systems. Interestingly, several disease genes are also linked to Wnt signaling (Inestrosa and Arenas, 2010) and directly interact with Wnt signaling components, such as Parkin, a ubiquitin ligase for β-catenin (Rawal et al., 2009), LRRK2 that interacts with Lrp6, Axin1, GSK3β, and β-catenin (Berwick and Harvey, 2012), as well as Dvl 1–3 (Sancho et al., 2009).
Pathologically, PD is characterized by the degeneration of neurons, in particular SNc DA neurons, and the presence of abnormal fibrillar cytoplasmic inclusions rich in α-synuclein and ubiquitin, called Lewy bodies, present in surviving neurons of both central and peripheral nervous system (Jellinger, 2012). Subcellularly, degeneration appears first at the synapse and then extends in a retrograde fashion to the rest of the neuron. In the nigrostriatal system, this process leads to a severe loss of the neurotransmitter dopamine. The first motor symptoms of PD appear when the loss of dopamine in the striatum exceeds 60% (Jellinger, 2012). Thus, by the time of diagnosis, a significant level of neurodegeneration has already occurred, which limits the efficacy of current therapeutic interventions. PD is currently treated by increasing dopaminergic transmission with dopaminergic agonists or L-DOPA, a precursor of dopamine, or alternatively, by inhibiting overactive pathways by surgically implanting electrodes for deep brain stimulation (Lees et al., 2009). The most common and efficacious symptomatic treatment today is L-DOPA. However, its long-term use leads to adverse effects, such as dyskinesias, and decreasing efficacy due to the progressive loss of DA neurons remaining in the brain, thus prompting the development of regenerative medicine strategies for treating PD.
Wnt signaling in regenerative medicine strategies for PD
Fetal tissue transplantation
Studies of fetal tissue transplantation in PD patients have provided proof of principle that cell replacement therapy is a viable therapeutic option by showing that mDA neurons morphologically and functionally integrate in the striatum and the circuitry of patients. However, clinical outcomes have been variable due to multiple factors including the way tissue was prepared for transplantation, the amount of cells, the position of the graft, the duration of immunosuppression, and patient selection (Lindvall and Björklund, 2004, 2011). Notably, younger patients with motor forms and pathology restricted to the striatum benefit most, and under optimal transplantation conditions, symptomatic relief lasted for up to 16 years without L-DOPA. However, 15% of the patients developed graft-induced dyskinesias that have been associated with the presence of serotonin neurons in grafts and can be blocked by 5HT1A agonists (Politis et al., 2010). Lewy bodies, the hallmark of PD, were found in fetal cells one decade post-transplantation but did not hamper long-term improvements in PD patients (Dunnett et al., 2001; Lindvall and Björklund, 2004, 2011; Arenas, 2010). The main obstacle for the therapeutic application of human fetal VM tissue is the poor tissue availability and the difficulties to standardize fetal tissue prior to grafting. It is currently thought that neural stem cells derived from human VM tissue or pluripotent stem cells are better sources for cell replacement therapy, since they are expandable and allow for standardization and quality control.
Stem cell-derived mDA neurons and transplantation
As Wnts are key building blocks in neural development, in particular the development of mDA neurons, they have thus become very attractive basic tools to create stem cell-derived mDA neurons. One strategy targeting the expansion of midbrain stem/progenitor cells present in the VM tissue and their subsequent differentiation was first developed for rodent VM tissue grown as neurospheres (Parish et al., 2008) and recently successfully applied to improve the quantity and quality of human mDA neurons derived from human fetal midbrain tissue (Ribeiro et al., 2012). Briefly, bFGF in combination with Shh and FGF8 were used to increase the overall cell number by 3-fold during 2 weeks. Subsequent differentiation in the presence of Wnt5a and survival factors, such as brain derived neurotrophic factor (BDNF), glial cell-line derived neurotrophic factor (GDNF), and transforming growth factor (TGFβ3), for 3–15 days resulted in a 6- to 7-fold increase in the number of TH+ cells. Importantly, Wnt5a increased per se the differentiation of mDA neurons by 2-fold, enhanced the expression of typical midbrain markers, such as Lmx1a, Lmx1b, Foxa2, and Nurr1 in TH+ cells, promoted neuritogenesis, and increased dopamine release as well as the electrophysiological properties of the cells. This strategy reduces the number of human fetal tissues required for transplantation and may thus contribute to a more rapid development of cell transplantation in PD patients. Human fetal tissue from the MHR has also been used for deriving other immortalized (Courtois et al., 2010) or non-immortalized cell lines (Tailor et al., 2013), but Wnts have not yet been applied to these cells.
Wnt signaling has been reported to play a role in the maintenance and early differentiation of embryonic stem (ES) cells (Merrill, 2012; Holland et al., 2013). However, here we will focus on the role of Wnt signaling in mDA neuron development. A combination of Wnt5a and FGF20 treatment was found to promote the in vitro differentiation of parthenogenetic primate ES cells into mDA neurons (Sanchez-Pernaute et al., 2008). Notably, when these cells were transplanted into a rodent model of PD, they survived better and induced greater functional recovery, as assessed by motor function, than control ES cells. It is important to note that in vitro treatment of mouse ES (mES) cells with inhibitors of the Wnt/β-catenin pathway similarly promotes mDA differentiation. Several studies have shown that the Wnt/β-catenin and Wnt/PCP pathways are in balance and inhibition of the former pathway leads to activation of the latter (Topol et al., 2003; Veeman et al., 2003; Tahinci et al., 2007; Bryja et al., 2009). Consistently, Dkk1 treatment, deletion of Wnt1 or Lrp6, or treatment with low concentrations of sFRP1 or sFRP2 was found to increase mDA differentiation of mES cells (Cajánek et al., 2009; Kele et al., 2012). Interestingly, higher or lower concentrations of sFRP1 or sFRP2 were less effective or completely ineffective (Kele et al., 2012), indicating that a very precise level of Wnt signaling is required for proper mDA differentiation.
For what we know from developmental studies, the balance between Wnt/β-catenin and Wnt/PCP signaling changes at different stages of mDA neuron development. It is thus conceivable that similar variations might be necessary to improve current protocols for generating mDA neurons from stem cells. In a recent attempt to improve stem cell differentiation, a sequential treatment with Wnt3a first to activate Wnt/β-catenin signaling during mDA progenitor proliferation, and then Wnt5a to activate Wnt/PCP signaling and promote mDA differentiation, was tested in both mouse VM neurospheres and mES cells (Andersson et al., 2013). While Wnt5a alone increased the number of TH+ cells, Wnt3a alone did not. However, sequential Wnt treatment significantly increased the proportion of TH+/TuJ1+ cells in both VM neurosphere and mES cultures by 80% than controls without Wnt treatment, as well as the number of TH+ cells compared with either Wnt5a or Wnt3a treatment alone (Andersson et al., 2013). It also increased the number of mES-derived TH+ cells expressing typical midbrain markers (Lmx1a, Foxa2, Nurr1, and Pitx3), compared with control cells treated with Shh, FGF8, and bFGF, followed by ascorbic acid, BDNF, and GDNF, but without Wnts. Moreover, double Wnt1 and Wnt5a knockout mice showed a greater loss of mDA neurons, compared with each individual knockout (Andersson et al., 2013). These results indicate that the use of Wnts in developmentally-based protocols has the potential to improve both the quantity and quality of mDA neurons. However, this dual Wnt protocol has not yet been implemented in human stem cells or animal models of PD, both of which are essential for determining the function and therapeutic potential.
In recent years, GSK3β inhibitors have become one of the most commonly used tools to activate Wnt signaling. Tang et al. (2010) used a protocol combining Shh and the GSK3β inhibitor CT99021 to mimic midbrain floor plate development and promote the differentiation of mES cells into mDA neurons. Activation of Wnt/β-catenin signaling in mES cells was found to be useful, but not critical because of its high endogenous level of Wnt/β-catenin signaling. However, administration of GSK3β inhibitors to human pluripotent stem (PS) cells was essential for the correct specification of mDA neurons (Kriks et al., 2011; Kirkeby et al., 2012; Xi et al., 2012). This protocol used Smad inhibitors for neural induction, Shh and FGF8 for ventralization and midbrain patterning, CHIR99021 to activate Wnt/β-catenin signaling and induce a midbrain floor plate identity, and continued with survival and differentiation factors. It resulted in 70%–80% mDA neurons with a very high degree of co-localization of Lmx1a and Foxa2, indicating optimal activation of the Wnt1/Lmx1a and Shh/Foxa2 pathways. Kirkeby et al. (2012) also found that low and high concentrations of GSK3β inhibitors induced diencephalic and hindbrain cells, respectively, while intermediate concentrations resulted in cells with midbrain identity. These two studies represent an important breakthrough for the advancement of cell replacement therapies and regenerative medicine for PD, as mDA neurons produced by these protocols co-expressed key midbrain transcription factors, exhibited appropriate electrophysiological properties, were capable of releasing dopamine, survived and innervated the host striatum, induced behavioral motor recovery in diverse animal models of PD, and did not lead to neural overgrowth (Kriks et al., 2011; Kirkeby et al., 2012). Two other studies also reported that GSK3β inhibitors can improve the mDA neuron differentiation of pluripotent cells in vitro (Xi et al., 2012; Denham et al., 2012). However, it should be noted that GSK3β inhibitors have off target effects (Bain et al., 2007) and can modulate other pathways (Hur and Zhou, 2010). In addition to β-catenin and Lrp6, GSK3β also phosphorylates Gli (Smelkinson et al., 2007), Smad (Eivers et al., 2008), nuclear receptors (Burns and Vanden Heuvel, 2007), secretases involved in Notch signaling (Ly et al., 2013), and transcription factors, such as basic helix–loop–helix (Ma et al., 2008) and CREB, all of which are important for mDA neuron development. Thus, those observed effects of GSK3β inhibitors could also be attributed to Shh, BMP/TGFβ, or neurotrophin signaling. Future studies should aim at using more specific Wnt/β-catenin activators and explore the involvement of Wnt-independent GSK3β-regulated pathways. These studies may contribute to develop more effective protocols for the generation of mDA neurons from stem cells and improve their applications in cell replacement therapy or drug development (Figure 2).
Direct reprogramming and transplantation
The initial work by Takahashi and Yamanaka (2006) on induced PS (iPS) cells has paved the way for the direct induction of neurons by reprogramming. In direct reprogramming, neurons are generated from other somatic cells, such as astrocytes or fibroblasts, by the expression of key master transcription factors and/or microRNAs relevant to the cell type being generated (Ladewig et al., 2013). While astrocytes can be reprogrammed into cortical neurons by overexpression of the transcription factor Pax6 (Heins et al., 2002), a combination of Brn2, Ascl1, and Mytl1 (BAM) was needed to reprogram fibroblasts into electrophysiologically functional cortical neurons (Vierbuchen et al., 2010). More recently, rodent and human fibroblasts were first reprogrammed into induced DA (iDA) neurons by a combination of transcription factors relevant for reprogramming (Ascl1) and mDA neuron development, such as Lmx1a and Nr4a2/Nurr1 (Caiazzo et al., 2011). Other studies have used BAM, Lmx1a, and Foxa2 (Pfisterer et al., 2011), or Ascl1, Lmx1a, Foxa2, Nurr1, En1, and Pitx3 (Kim et al., 2011). Notably, transplantation of rodent iDA cells led to clear host innervation and behavioral recovery in an animal model of PD (Kim et al., 2011).
The GSK3β inhibitor CHIR99021 was recently used in combination with Smad inhibitors and two proneural transcription factors, Ascl1 and Ngn2, in direct reprogramming of fibroblasts into neurons (Ladewig et al., 2012). This method resulted in a remarkable increase in neuronal yield of 150%–200% and a final purity of 70%–80%. However, whether GSK3β inhibitors can also increase the number, phenotypic properties, and transplantability of directly reprogrammed iDA neurons remains to be determined.
Finally, direct in vivo reprogramming of endogenous brain striatal cells in situ by BAM lentiviruses, was recently tested as an alternative to cell transplantation (Torper et al., 2013). So far the efficiency of direct in vivo reprogramming is low, but the full potential of this attractive method remains yet to be explored (Figure 2).
Disease modeling and drug discovery
It is worth mentioning that reprogrammed cells obtained from PD patients and differentiated into mDA neurons are also ideal tools for disease modeling and drug discovery (Figure 2). Indeed, human iPS cells derived from patients with different forms of PD are currently used to investigate the pathogenesis of human PD and screen for drugs that impact on defined pathogenic mechanisms. In this context, Wnt signaling may improve protocols for the differentiation of these iPS cells, and thus PD modeling in vitro. Moreover, several PD-associated proteins, including LRRK2 and Parkin, interact with Wnt signaling components and modulate Wnt signaling (Rawal et al., 2009; Sancho et al., 2009; Inestrosa and Arenas, 2010; Berwick and Harvey, 2012). IPS cells derived from PD patients may be very interesting tools to understand the pathogenesis of PD and develop novel drugs combating this devastating disease.
Within the field of regenerative medicine and PD, the development of small molecules capable of selectively and efficiently modulating different branches of Wnt signaling is likely to have a profound impact on our capacity to regulate stem cell behavior and reprogramming, thus contributing to cell replacement and drug development strategies. During the last years, many small molecules capable of modulating Wnt signaling have become available (see details in Supplementary material). However, molecules capable of specifically and selectively activating certain branches of Wnt signaling are still lacking. This is particularly true for Wnt/PCP signaling since only a Wnt5a-mimetic small peptide was described (Säfholm et al., 2006). There is thus a need to develop small molecules or biologicals, such as antibodies, to activate Wnt/PCP. Potential strategies include: (i) inhibitors of syndecan-mediated internalization of R-spondins; (ii) activators of Wnt co-receptors such as Ror1/2; or (iii) selective modulators of Dvl sites involved in Wnt/PCP signaling. It is also necessary to identify more selective small molecule activators of Wnt/β-catenin signaling. Commonly used GSK3β inhibitors have off target effects (Bain et al., 2007) and affect other pathways regulated by GSK3β (Hur and Zhou, 2010). Alternatives to GSK3β inhibitors could be: (i) Tiki1 inhibitors, since Tiki1 inhibits the proteolytic cleavage of Wnt (Zhang et al., 2012); (ii) inhibiting Dkk1 binding to Lrp5/6 and Kremen 1–2; or (iii) inhibiting the sequestration of Lrp5/6 away from Fz.
For cell transplantation, it is currently thought that human PS cells are the optimal source of mDA neurons because of their capacity to engraft and induce recovery of motor function in animal models of PD (Kriks et al., 2011; Kirkeby et al., 2012). Compared with iPS cells, ES cells are preferred for cell replacement therapy, since they are not genetically modified and their epigenetic state is more stable and homogenous (Cahan and Daley, 2013). Currently available stem cell-derived mDA neurons contain both SNc and VTA mDA neurons. As discussed before, several factors including Wnt signaling control the specification for SNc or VTA subtype. In the future, it will be very important to develop protocols that would promote differentiation towards SNc mDA neurons. In addition, shorter robust protocols for the generation of SNc mDA neurons from ES cells are required, as the current ones are relatively long. One option could be to use PS cell-derived neuralized cells that could be used as a platform for mDA differentiation in shorter time. Potential candidates are PS cell-derived long-term neuroepithelial stem cells, which can be differentiated into mDA neurons (Koch et al., 2009; Falk et al., 2012). However, to date, Wnt-based protocols and in vivo functionality in animal models of PD have not yet been tested with these cells. Finally, it remains to be explored whether Wnt small molecules can be used in vivo, to regulate the function of transplanted or in situ directly reprogrammed mDA cells, or diseased endogenous mDA neurons. Such applications, which will require the development of drugs that can be administered systemically, may open up new exciting possibilities (Figure 2).
In sum, Wnt signaling plays a critical role in mDA neuron development and its implementation in protocols for in vitro differentiation of stem cells and reprogramming cells have had a major impact on the development of cell replacement therapy for PD. The incessant growth of the Wnt field will undoubtedly continue to contribute to a better understanding of mDA neuron development and to promote the development of regenerative medicine and drug discovery for PD.
This work was supported by grants from the Swedish Foundation for Strategic Research (CEDB and SRL Program), Swedish Research Council (VR2008:2811 and 3287, VR2011:3116, and DBRM), Karolinska Institutet (SFO Thematic Center in Stem cells and Regenerative Medicine), and European Commission (NeuroStemCellRepair and DDPD).
Conflict of interest: none declared.
The author thanks members of the Arenas lab, in particular Enrique Toledo, Daniel Gyllborg, and Carmen Saltó, for critical reading of the manuscript and Carlos Villaescusa for help with preparation of the manuscript and Figure 2.