Zeb2 is a negative regulator of midbrain dopaminergic axon growth and target innervation

Neural connectivity requires neuronal differentiation, axon growth, and precise target innervation. Midbrain dopaminergic neurons project via the nigrostriatal pathway to the striatum to regulate voluntary movement. While the specification and differentiation of these neurons have been extensively studied, the molecular mechanisms that regulate midbrain dopaminergic axon growth and target innervation are less clear. Here we show that the transcription factor Zeb2 cell-autonomously represses Smad signalling to limit midbrain dopaminergic axon growth and target innervation. Zeb2 levels are downregulated in the embryonic rodent midbrain during the period of dopaminergic axon growth, when BMP pathway components are upregulated. Experimental knockdown of Zeb2 leads to an increase in BMP-Smad-dependent axon growth. Consequently there is dopaminergic hyperinnervation of the striatum, without an increase in the numbers of midbrain dopaminergic neurons, in conditional Zeb2 (Nestin-Cre based) knockout mice. Therefore, these findings reveal a new mechanism for the regulation of midbrain dopaminergic axon growth during central nervous system development.

The establishment of neural pathways in the developing central nervous system (CNS) requires coordinated regulation of neuronal differentiation, axon growth and target innervation 1 . The signals that guide catecholaminergic innervation have been intensively studied in the peripheral nervous system (PNS), where the neurotrophic theory states that limiting amounts of target-derived neurotrophic factors maintain the survival of the required numbers of neurons, and promote the growth of innervating axons in their targets 2,3 . More recently, it has emerged that cell-autonomous signalling operates in these neurons to regulate axon growth and target innervation [4][5][6] . In the CNS, most catecholaminergic neurons are found in the brainstem, of which one of the largest populations are the midbrain dopaminergic (mDA) neurons of the substantia nigra that project axons to the striatum. Nigrostriatal mDA neurons are generated between embryonic day (E) 10 and E12 in mice (E12-E14 in rat) 7 . mDA axons extend rostrally from E11, reach the striatum at E12, and innervate the rostral part of striatum by E16 in mice (E18 in rat) 8 . Naturally-occurring cell death occurs in the postnatal period (~P0-P20) to prune the initially overproduced nigrostriatal mDA neurons, and depends on target-derived neurotrophic factors such as GDNF [9][10][11] . The nigrostriatal pathway is crucial for voluntary motor control 12 , as highlighted by the progressive loss of striatal mDA innervation in Parkinson's disease (PD) 13 . While the molecular basis of mDA specification and differentiation has been extensively characterised 7 , it remains largely unknown whether cell-autonomous mechanisms play a role in regulating axon growth and target innervation of mDA neurons in the CNS.
From a PCR screen to identify new regulators of mDA axon growth and innervation, we detected the expression of Zinc finger E-box-binding homeobox 2 (Zeb2) mRNA in the midbrain during the period of striatal innervation. Zeb2 (Smad-interacting protein-1; Sip1; and Zfhx1b) was identified as a DNA-binding transcriptional repressor that interacts with activated Smad1 14 , a mediator of bone morphogenetic protein (BMP) signalling. Within the embryonic CNS, Zeb2 controls hippocampal neuron development and is crucial for the correct timing of neurogenesis and gliogenesis in the neocortex in a non-cell-autonomous manner 15,16 . Zeb2 also acts cell-autonomously to control: guided migration of cortical GABAergic interneurons 17,18 ; neocortical axon growth 19 ; myelinogenesis in the CNS 20 ; and (re)myelination in the PNS 21,22 . In humans, mutations in one ZEB2 allele causes Mowat-Wilson syndrome (MOWS, OMIM#235730), which includes severe neurodevelopmental defects 23,24 . Here we investigate for the first time the role of Zeb2 in developing mDA neurons. We describe a new function for Zeb2 as a negative regulator of mDA axon growth and target innervation.

Results
Temporal-regulation of Zeb2 expression, and of BMP pathway components, in developing midbrain in vivo. Quantification of mRNA levels in the mouse ventral midbrain (VM) at intervals throughout the period of initial mDA axon growth and striatal innervation revealed significant differences in Zeb2 expression during development. Post-hoc testing showed that Zeb2 mRNA levels significantly increase from E10 to peak at E12, followed by a sharp decrease from E12 to E18 (p < 0.0001; Fig. 1A). Because Zeb2 represses BMP signalling during nervous system development 25 , we next investigated the expression of genes encoding components of the BMP pathway, which are subject to synexpression, autoregulation and feed-forward/-back regulation 14,[25][26][27] . We hypothesised that Zeb2 downregulation may be paralleled by modulated expression of some of these BMP pathway components. To test this, we quantified transcript levels of Gdf5, Bmp2, Bmpr1b and Sizn1 over the same time range. Significant age-dependent changes in expression were observed for Gdf5, Bmp2, Bmpr1b and Sizn1 (which protein levels in the mouse midbrain at E12 and E16. βIII tubulin was used as a loading control. (J) Immunohistochemistry showing p-Smad1/5 (red) co-localising with TH (green) in the E14 rat VM (indicated by arrow). (number of repetitions (N) = 3-6 from separate litters; mean ± SEM) (*p < 0.05, **p < 0.01, ***p < 0.001; E12 v E16 or E18; t-test or ANOVA with post hoc Tukey's test). Scale bar as indicated. encodes a transcriptional co-activator downstream of BMP signalling). These displayed different profiles to that of Zeb2, significantly increasing from E12 to E18/P1 (p < 0.05; Fig. 1B to E). Such changes in Zeb2 and Bmpr1b mRNA expression were also observed in the rat VM between E14 and E18 (data not shown and 28 ). In situ hybridisation confirmed the quantitative real-time PCR (RT-qPCR) data, and showed localisation of Zeb2 transcripts in the E11.5 mouse VM (Fig. 1F). Moreover, western blot was also used to demonstrate that Zeb2 and BMPR1b protein levels display these age-dependent expression changes in the mouse VM between E12 and E16 ( Fig. 1G to I). Finally, immunohistochemistry was used to examine phospho-Smad1/5 signalling in the VM at the time of mDA axogenesis. Triple _ labelled preparations in which mDA neurons were identified by tyrosine hydroxylase (TH) staining revealed that p-Smad1/5 were present in, but not limited to, mDA neurons in the E14 rat VM (Fig. 1J). This indicates that BMP-Smad signalling is active during mDA axonal growth in the developing midbrain.
Zeb2 is a cell-autonomous negative regulator of BMP-Smad1/5 signalling in cultured midbrain neurons. To determine if Zeb2 functions as a repressor of p-Smad1/5 signalling in primary VM neurons, we co-transfected E14 rat VM cells with a green fluorescent protein (GFP)-vector and either a control or Zeb2 expression vector. The latter resulted in increased Zeb2 expression in transfected (GFP + ) cells (p < 0.001; Fig. 2A). Upon vector-encoded Zeb2 overexpression, p-Smad1/5 levels in E14 VM cells were significantly reduced at 24 hour (h) post-transfection (p < 0.001; Fig. 2B). We also used SH-SY5Y cells, which are often used to model DA neurons and in which BMP-Smad signalling promotes neurite growth 29,30 , and found that Zeb2 overexpression reduced p-Smad1/5 levels (Fig. 2C).
We next transfected both cell types with scrambled or Zeb2-targeting siRNA, and found that knockdown of Zeb2 resulted in a significant increase in p-Smad1/5 levels in both SH-SY5Y and E14 VM cells (p < 0.05; Fig. 2D to F). Zeb2 knockdown increased p-Smad1/5 levels to levels comparable to those in cultures treated with 100 ng/ ml growth differentiation factor 5 (GDF5/BMP14), which was used as a positive control. To determine if this upregulation of p-Smad1/5 levels altered Smad-dependent transcription, SH-SY5Y cells were co-transfected with a reporter plasmid in which GFP expression is driven by a promoter containing Smad-binding elements (SBE-GFP) 28,31 , together with either scrambled or Zeb2-siRNA. Control cultures were transfected with the SBE-GFP reporter vector, and treated without or with GDF5, the latter resulting in p-Smad-dependent reporter transcription. Zeb2 knockdown also led to a significant increase in Smad-dependent transcription (indicated by GFP expression) at 48 h (p < 0.01; Fig. 2G and F). These data show that Zeb2 is a negative regulator of p-Smad1/5 signalling in VM cells, and suggest that decreased Zeb2 expression may facilitate Smad-promoted axon growth.

Zeb2 downregulation is necessary and sufficient for Smad-promoted neurite growth.
To test this hypothesis, SH-SY5Y cells and E14 VM neurons were co-transfected with a GFP-vector and either a scrambled or Zeb2-siRNA, or treated with 100 ng/ml GDF5, and neurite growth was examined 72 h later. Knockdown of Zeb2 resulted in a significant increase in neurite growth in both cell types (p < 0.001; Fig. 3A to C). This increase in neurite growth was entirely prevented by Smad4 knockdown (Fig. 3A and B), which inhibits p-Smad-dependent transcriptional activity 28,29 . To identify mDA neurons in GFP-transfected E14 VM cultures, cells were immunocytochemically stained for TH (Fig. 3D). Knockdown of Zeb2 resulted in a significant increase in neurite length in TH + (and GFP + ) mDA neurons at 72 h (p < 0.01; Fig. 3E and F). In contrast, vector-overexpressed Zeb2 did not affect basal levels of neurite growth in E14 VM neurons, whereas it completely prevented GDF5-promoted neurite growth (p < 0.05; Fig. 3G). We next examined whether Zeb2 modulates GDNF-induced neurite growth of E14 VM neurons, and found that Zeb2 overexpression did not inhibit the growth-promoting effects of GDNF (data not shown). These results demonstrate that high levels of Zeb2 maximally repress BMP-Smad-induced neurite growth. They also show that Zeb2 represses p-Smad1/5 activity in the basal state below a neurite growth-promoting threshold, and that downregulation of Zeb2 expression in mDA neurons facilitates Smad-promoted neurite growth.
BMPR signalling is necessary for axon growth induced by Zeb2 downregulation. We next determined whether the neurite growth promoted by Zeb2 downregulation involves extracellular and/or intracellular mechanisms. As the increase in neurite growth that occurred following RNAi-based knockdown of endogenous Zeb2, required Smad4 ( Fig. 3A and B), we focused on the canonical BMP-Smad signalling pathway. E14 VM neurons were co-transfected with a GFP vector and either a scrambled or Zeb2-siRNA, and then cultured with or without noggin (a secreted BMP antagonist) or dorsomorphin (a BMP receptor (BMPR) type 1 inhibitor) for 72 h (Fig. 3H). While knockdown of Zeb2 promoted a significant increase in neurite length (p < 0.01), noggin or dorsomorphin-mediated inhibition of BMPR activation significantly inhibited Zeb2-siRNA-induced neurite growth (Fig. 3I). We subsequently tested whether BMPR1b was required for neurite growth following Zeb2 downregulation by transfecting Zeb2-siRNA-treated VM cells with BMPR1b-siRNA 28,29 , and found that this completely prevented neurite growth (p < 0.05; Fig. 3J).
We hypothesised that if the mechanism involved in Zeb2-mediated repression of p-Smad-dependent neurite growth occurs at the transcriptional level, then knockdown of a key transcriptional co-factor and binding partner of Zeb2 may also increase neurite length. To investigate this, we used STRING analysis to identify the top predicted functional binding partner of Zeb2, which was C-terminal binding protein 1 (CtBP1) (Fig. 3J) 26,32 . Knockdown of CtBP1 in E14 VM neurons had the same neurite growth-promoting effect as knockdown of Zeb2 (p < 0.001; Fig. 3L), an effect also observed in SH-SY5Y cells (data not shown). This suggests that a Zeb2-CtBP transcriptional repressor complex acts to inhibit mDA axonal growth. Interestingly, treatment with noggin or dorsomorphin (Fig. 3I), or knockdown of Smad1, Smad5, or Smad7 (an inhibitory Smad that negatively regulates BMP-Smad signalling) (Fig. 3L), had no effect on neurite length. This further supports the conclusion that basal levels of Zeb2 repress p-Smad signalling below a growth-promoting threshold. Taking these findings together, Zeb2 appears to regulate mDA neuronal growth through a BMPR-Smad1/5 dependent mechanism, raising the possibility that Zeb2 may also act to control the extent of target innervation in vivo.
Zeb2 knockout mice have dopaminergic hyperinnervation of the developing striatum. To assess the physiological relevance of Zeb2 in mDA axon growth in vivo, we quantified the numbers of mDA neurons, and DA innervation density of the developing striatum, in conditional Zeb2-deficient mice (Nestin-Cre:Zeb2 fl/KO ; Zeb2 cKO), and in mice previously demonstrated to produce normal levels of Zeb2 (Nestin-Cre:Zeb2 fl/WT ; control) 16 . There were no significant differences in the numbers of TH + mDA neurons between the groups at E12.5 ( Fig. 4A and B) or E16.5 ( Fig. 4C and D). In the mouse, mDA axons extend from E11, reach striatum from E12 and innervate the rostral part of striatum by E16 (for review, see 7 ). Therefore, we assessed striatal mDA innervation density by quantifying TH + fibres in the striatum at E16.5, a stage at which mDA innervation of the striatum is well underway. TH + innervation density in the striatum of E16.5 Zeb2 cKO was significantly greater than in controls (p < 0.01; Fig. 4E and F). To rule out the possibility that this increase in TH + innervation density was due to increased TH expression, rather than increased DA innervation, we analysed the expression of TH following Zeb2 knockdown in vitro and in vivo. We found no differences in TH expression in mDA neurons following Zeb2 knockdown (data not shown). These findings suggest that Zeb2 plays a physiological role in regulating mDA axon growth and target innervation in vivo.

Discussion
We report that Zeb2 is a novel negative regulator of mDA axon growth and target innervation during embryogenesis (Fig. 5). Downregulation of Zeb2 expression in embryonic VM neurons resulted in promotion of axon growth via Smad signalling, while its absence in Zeb2-deficient mice led to mDA striatal hyperinnervation. In the mouse VM, Zeb2 expression increased from E10 to sharply peak at E12, suggesting that Zeb2 may contribute to mDA neurogenesis 33 , as it does in the hippocampal anlage and the embryonic neocortex 15,16 . However, no changes in numbers of TH + mDA neurons were observed in the Zeb2 cKO mice at E12.5 or E16.5. Interestingly, Wnt5a is expressed in the mouse VM at E11.5 and is important for the initiation of neurite outgrowth, but by E14.5 causes neurite retraction 34 . Zeb2 has been identified as a candidate downregulator (either directly or indirectly) of Sfrp1, which encodes a secreted Wnt antagonist 15,33 . Thus, high levels of Zeb2 at E12 may facilitate Wnt-promoted initiation of mDA axon growth, which may then become more responsive to BMP-Smad signalling-induction as Zeb2 expression declines from E12-E18, due to a combination of increasing inhibition of Wnt signalling (via increased Sfrp1) and decreasing Zeb2-mediated inhibition of BMP-Smad signalling. Indeed, axonal growth requires integration of multiple signals, as well as the changing of responses to cues along their trajectory to targets 1 . This proposed mechanism is supported by our finding that Zeb2 negatively regulates BMP-Smad dependent gene transcription, in agreement with other studies 14,20,25 . Moreover, Zeb2 regulates Wnt and BMP signalling output during neural crest development, via a mechanism that may depend on Smads 25,35 . In addition to this, TGF-β superfamily members, such as GDNF and TGF-βs, are expressed during nigrostriatal pathway development and regulate its development [9][10][11][36][37][38] . While we found that Zeb2 did not inhibit the neurite growth-promoting effects of GDNF in vitro, it would be interesting in future work to disentangle any potential cross-talk between Zeb2 and other molecular pathways that promote neurite growth in mDA neurons, such TGF-βs and Wnts.
Downregulation of Zeb2 also facilitated axon growth through a BMP-BMPR-Smad mechanism. In support of this, canonical BMP-BMPR1b-Smad signalling induces mDA neuronal growth 28 . Furthermore, we found that increased Zeb2 levels represses GDF5-promoted, BMPR-Smad-mediated neurite growth. Our findings also suggest that such repression by Zeb2 is dependent on the CtBP co-repressor, similar to Zeb2-mediated neural induction in the amphibian embryo 26 . These data indicate that downregulation of Zeb2 during mDA axonal growth and target innervation may facilitate BMP-Smad-mediated induction of this developmental process. In support of this, BMPs and BMPRs are present in both the midbrain and striatum, as well as in mDA neurons, at key time-points for nigrostriatal pathway development 28,36 . Thus, BMPs may act locally in the midbrain, and/or in the striatum as target-derived neurotrophic factors, to induce mDA axonal growth and target innervation in a process that, at least in part, is negatively regulated by Zeb2 levels in a cell autonmous manner. Indeed, axon growth and target innervation of peripheral catecholaminergic neurons is controlled by cell-autonomous signalling [4][5][6] , and is also regulated by Wnt5a and BMP-Smad signalling 6,31 . In future studies, it will be important to further characterise the role of BMP ligands in mDA innervation, and to determine whether Zeb2 functions in a similar manner in peripheral catecholaminergic neurons.
We also report for the first time mDA hyperinnervation of the striatum in the Zeb2 cKO mouse embryos at E16.5. This may recapitulate some developmental defects in MOWS, in particular the enlarged basal ganglia 24 . However, whether this mDA hyperinnervation persists throughout the lifespan is an important topic for future investigation. Nigrostriatal mDA neurons are initially overproduced during neurogenesis, and subsequently undergo naturally-occurring cell death in the postnatal period (~P0-P20) 9 . Future studies should analyse striatal mDA fibre density and mDA neuronal number in the Zeb2 cKO mouse at ages following this period of postnatal developmental programmed cell death. Finally, progressive degeneration of striatal mDA innervation causes debilitating motor dysfunctions in PD 13 . There is robust evidence that mDA axonal degeneration in the striatum occurs early in PD pathogenesis, and precedes neuronal loss 39,40 . Thus, transcription factors that control mDA axon growth may be novel candidate targets in future therapeutic approaches for PD 41 . In this regard, the present study suggests that Zeb2 downregulation may be a viable neurotrophic strategy to restore striatal mDA innervation for the treatment of PD. In summary, our work provides new evidence for a Zeb2-controlled signalling mechanism that operates in central catecholaminergic neurons to regulate axon growth and appropriate target innervation during embryonic development. signalling, which promotes neurite growth in mDA neurons. In future work it will be interesting to disentangle any potential cross-talk between Zeb2 and other molecular pathways that promote neurite growth in mDA neurons, including TGF-βs and Wnts. (B) Proposed model of the contribution of Zeb2 to the embryonic development of mDA target innervation based on current data. During peak mDA neurogenesis at ~E12 in mice, high levels of Zeb2 repress BMP-Smad signalling to limit mDA neurite growth. As mDA neurons extend axons and innervate the striatum from E12-E16, a reduction in Zeb2 levels leads to an increase in BMP-Smaddependent transcriptional activity which promotes neurite growth in mDA neurons. (C) Knockdown of Zeb2 removes a cell-autonomous inhibitory mechanism that results in excessive mDA neurite growth through BMPR1b-Smad-dependent signalling, which causes hyperinnervation of the striatum in vivo. Whether this mDA hyperinnervation persists throughout the lifespan is an important topic for future investigation. In summary, these data provide new evidence for a Zeb2-controlled signalling mechanism that operates in mDA neurons to regulate axon growth and target innervation during embryonic CNS development.

Methods
Cell Culture. All experiments were carried out in accordance with relevant guidelines and regulations, in compliance with law, and approved by the UCC Animal Ethics Experimentation Committee and the Health Products Regulatory Authority. For E14 rat VM cultures, E14 embryos were obtained by laporatomy from timemated female Sprague-Dawley rats following decapitation under terminal anaesthesia induced by the inhalation of isoflurane (Isoflo ® ). Dissected VM tissue was centrifuged at 100 g for 5 minutes at room temperature. The tissue pellet was then incubated in a 0.1% trypsin-Hank's Balanced Salt solution for 5 minutes, at 37 °C with 5% CO2. Fetal calf serum (FCS; Sigma) was then added to the tissue followed by centrifugation at 100 g for 5 minutes at 4 °C. The resulting cell pellet was resuspended in 1 ml of differentiation media (Dulbecco's modified Eagle's medium/F12, 33 mM D-glucose, 1% L-glutamine, 1% FCS, supplemented with 2% B27; all Sigma) using a P1000 Gilson pipette, and then carefully triturated using a sterile plugged flame-polished Pasteur pipette, followed by a 25-gauge needle and syringe. Cell density was estimated using a haemocytometer. Cells were plated on poly-Dlysine (Sigma)-coated 24-well tissue culture plates at a density of 5 × 10 4 cells per well in 500 μl of differentiation media at 37 °C with 5% CO 2 .

Electroporation of Cultured Cells. Electroporation of cultured cells was carried out using the NeonTM
Transfection System (Invitrogen). Cell suspensions were prepared for counting (as outlined above), and the required volume of cells to give 200,000 cells per well was centrifuged at 4 °C at 100 g for 5 minutes. The cell pellet was washed twice with 10 mM phosphate buffered saline (PBS) (without CaCl 2 and MgCl 2 ) (Sigma), and then resuspended in the required amount of the manufacturers resuspension buffer (12 µl per transfection/plasmid) (Invitrogen). 0.5 µg of GFP plasmid; 1 µg of desired plasmid DNA (Zeb2 overexpression 42 or Smad4 siRNA 29 vectors); and/or 500 nM of Zeb2, CtBP1, Smad1, Smad5, or Smad7 siRNA (Life Technologies) was added to the resuspended cells. 10 µl of the cell/plasmid mixture was then electroporated according to the manufacturer's protocol under specific parameters (1100 V; 30 ms; 2 pulses).
Immunohistochemistry. Brains were dissected and washed in ice-cold PBS and fixed overnight with 4% paraformaldehyde (wt/vol) followed by progressive alcohol-assisted dehydration and paraffin embedding. 10-μm-thick serial sections were cut on a Leica microtome and then processed for immunohistochemistry using an automated platform (Ventana Discovery, Ventana Medical Systems) and mounted in DAPI-supplemented Mowiol. The following primary antibodies were used: p-Smad1/5 (1:50), TH (Novocastra; 1:500), and GFP (Novocastra; 1:500). For secondary antibodies, CY2-, CY3-or biotin-conjugated donkey antibodies (all at 1:600, Jackson ImmunoResearch) were used. Sections were photographed using a confocal radiance microscope connected to a spot camera (Visitron Systems) and data analysis was done with ImageJ software. 5 control and 5 knockout brains were analyzed, and 10 sections per brain were quantified. In each brain and time point, TH-positive cell bodies or TH-positive neurites/fibres were quantified in midbrain/striatum of the same coronal plane. TH-positive striatal fibres were analysed through 400 µm of the striatum, spanning from the rostral striatum to the region between the caudal striatum and diencephalon. Brain regions were identified using an atlas of the embryonic mouse brain and landmarks, such as the lateral ventricles, after which 10 µm serial sections (1:10 series) were counted and every fourth section was sampled from the striatum. For the striatal innervation experiments, TH-positive fibres were quantified in the E16.5 mouse striatum using ImageJ.
Quantitative real-time PCR (RT-qPCR). The midbrains of E10 to P1 CD-1 mice, obtained from time-mated pregnant dams as outlined above, were micro-dissected and total RNA was then extracted and purified. Midbrain samples were disrupted and homogenised in 1 ml of QIAzol Lysis Reagent (Qiagen). After the addition of 200 µl chloroform, homogenates were separated into aqueous and organic phases by centrifugation at 10,000 g for 15 minutes. The upper aqueous phase was mixed with an equal volume of 70% ethanol, to precipitate