The distribution and (patho-)physiological role of neuropeptides in the adult and aging brain have been extensively studied. Galanin is an inhibitory neuropeptide that can coexist with γ-aminobutyric acid (GABA) in the adult forebrain. However, galanin's expression sites, mode of signaling, impact on neuronal morphology, and colocalization with amino acid neurotransmitters during brain development are less well understood. Here, we show that galaninergic innervation of cholinergic projection neurons, which preferentially express galanin receptor 2 (GalR2) in the neonatal mouse basal forebrain, develops by birth. Nerve growth factor (NGF), known to modulate cholinergic morphogenesis, increases GalR2 expression. GalR2 antagonism (M871) in neonates reduces the in vivo expression and axonal targeting of the vesicular acetylcholine transporter (VAChT), indispensable for cholinergic neurotransmission. During cholinergic neuritogenesis in vitro, GalR2 can recruit Rho-family GTPases to induce the extension of a VAChT-containing primary neurite, the prospective axon. In doing so, GalR2 signaling dose-dependently modulates directional filopodial growth and antagonizes NGF-induced growth cone differentiation. Galanin accumulates in GABA-containing nerve terminals in the neonatal basal forebrain, suggesting its contribution to activity-driven cholinergic development during the perinatal period. Overall, our data define the cellular specificity and molecular complexity of galanin action in the developing basal forebrain.
Neuropeptides are coexpressed with major neurotransmitters in many brain systems (Hokfelt et al. 2003) and can modulate synaptic neurotransmission by influencing presynaptic neurotransmitter release, receptor trafficking and recycling, and postsynaptic signal coupling (Kupfermann 1991; Lundberg 1996; Salio et al. 2006). Galanin, a 29/30 amino acid long neuropeptide (Tatemoto et al. 1983), is one such peptide, which is widely distributed in the rodent brain (Skofitsch and Jacobowitz 1985; Melander et al. 1986; Perez et al. 2001) and acts via G protein-coupled galanin receptor subtypes 1–3 (GalR1–3; Lang et al. 2007). Galanin is under normal circumstances virtually undetectable in adult rat cholinergic forebrain neurons (Miller et al. 1998), but becomes up-regulated upon disruption of the fast axonal transport machinery (i.e. colchicine treatment) (Melander et al. 1985; Senut, Menetrey, et al. 1989; Cortes et al. 1990; Agoston et al. 1994). Galanin is also dynamically increased in brain pathologies stemming from neuronal hyperexcitability, particularly epilepsy (Lerner et al. 2010), likely strengthening endogenous inhibition of excitatory neurotransmission (Fisone et al. 1987). Accordingly, galaninergic “hyperinnervation” of cholinergic basal forebrain neurons in Alzheimer's disease is thought to protect these cells from noxious insults (Chan-Palay 1988; Mufson et al. 1993); particularly, since cholinergic cells with dense galanin afferentation retain prosurvival gene expression (Counts et al. 2010).
Despite the wealth of data available on galanin signaling in the adult nervous system in health and disease, significant caveats of knowledge exist regarding galanin's role during brain development. Considering the dramatic up-regulation of galanin expression on peripheral nerve injury (Hokfelt et al. 1987; Villar et al. 1989; Burazin and Gundlach 1998), and the positive impact of, in particular, galanin receptor 2 (GalR2) manipulation on the survival and regeneration of sensory neurons (Mahoney et al. 2003; Hobson et al. 2006; Suarez et al. 2006) and immortalized neuroblastoma cells (Berger et al. 2004; Tofighi et al. 2008), the prevailing hypothesis is that galanin signaling facilitates axonal growth. In adult sensory neurons, GalR2s recruit protein kinase C (PKC) and extracellular signal-related kinase 1/2 (Erk1/2; Hawes et al. 2006), as well as gp130 cytokines (Zigmond 2011), to induce growth responses.
Galanin and GalR1/2 mRNA expression in proliferative cell niches of the adult brain (Shen et al. 2003; Agasse et al. 2013) predict a developmental role for galanin overarching neuronal differentiation during fetal and postnatal periods. GalR2 is particularly interesting in a developmental context, since this receptor is developmentally regulated in the brain and spinal cord, with prominent mRNA expression reported in the basal forebrain and its cortical innervation targets during early postnatal stages (Burazin et al. 2000). Galanin signaling also appears important for the maintenance of cholinergic neuron identity, since a subset of these cells postnatally degenerates in galanin−/− mice (O'Meara et al. 2000). Nevertheless, neither galanin's anatomical localization nor its contribution to the control of the developmental programs of cholinergic neurons is known.
Here, we addressed the expression sites and subcellular distribution of galanin in the fetal mouse basal forebrain, and tested the hypothesis that galanin acts in a paracrine fashion by being focally released onto cholinergic neurons. Next, we dissected the reliance of cholinergic neurons on specific GalR subtypes and signal transduction cascades that convert galanin stimuli into ordered physiological responses. We show that cholinergic responsiveness to galanin is regulated by nerve growth factor (NGF; Planas et al. 1997), a classic neurotrophin implicated in the maintenance of cholinergic synaptic connectivity and survival (Honegger and Lenoir 1982; Seiler and Schwab 1984). Our results suggest that galanin is released from γ-aminobutyric acidergic (GABAergic) nerve endings juxtacellular to cholinergic perikarya and neurites, and regulates cholinergic neuritogenesis through GalR2-dependent Rho-family GTPase signaling, affecting cytoskeletal integrity and dynamics.
Materials and Methods
Animals and Tissue Processing
C57Bl6/N (n = 2–3/time point in the anatomy studies; Charles River) and glutamic acid decarboxylase 67-green fluorescent protein (GFP) (GAD67-GFP; n = 3) mice (Tamamaki et al. 2003), and Wistar rats (Charles River), were experimentally treated and processed according to standard protocols (Berghuis et al. 2007; Keimpema et al. 2010). M871 (Sollenberg et al. 2006) (0.06 or 0.6 mg/kg in saline), a GalR2 antagonist, was daily administered subcutaneously to neonatal mice from postnatal day (P)2 to P7. Embryonic tissues were immersion fixed in 4% paraformaldehyde (PFA) in 0.1 M phosphate-buffered (pH 7.4; PB) overnight. P7 and adult mice were transcardially perfused with 4% PFA in 0.1 M PB (20–100 mL/animal) and their brains postfixed in the same fixative overnight. Tissues were cryoprotected in 30% sucrose in 0.1 M PB for 48 h. Adult and fetal brains were sectioned on a cryostat microtome (Leica CV1850) at 50-µm thickness as free-floating sections collected in 0.1 M PB and 16 µm as glass-mounted specimens (SuperFrost+) (Berghuis et al. 2007), respectively. Experimental procedures were approved by the regional ethical committee (Stockholms Norra Djurförsöksetiska Nämnd; #N512/12). Particular effort was directed to minimize the number of animals and their suffering during the experiments.
Primary Cell Culture and Molecular Pharmacology
Fetal forebrains [embryonic day (E)16.5] were dissected in Hank's Balanced Salt Solution containing GlutaMAX (2 mM), penicillin (100 U/mL), streptomycin (100 µg/mL; all from Invitrogen/Gibco), and 0.5% glucose. Basal forebrains were isolated by a dorsal approach (Schnitzler et al. 2008). Cells were enzymatically dissociated by trypsin (0.1%; Invitrogen). Trypsin digestion (<5 min at 37 °C) was terminated by bovine serum albumin (BSA, 0.4%; Sigma-Aldrich) and DNAse (1000 U/mL; Promega). Cells were plated at a density of 50 000 cells/well in poly-d-lysine (PDL; Sigma-Aldrich)-coated 24-well plates. Primary basal forebrain cultures were maintained in dulbecco's modified eagle medium (DMEM)/F12 (1:1; Invitrogen/Gibco), B27 supplement (2%; Gibco), GlutaMAX (2 mM), penicillin (100 U/mL), and streptomycin (100 µg/mL) also containing galanin (100 nM; Bachem), AR-M 1896 (Mahoney et al. 2003; GalR2 agonist, 10 nM–1 µM; Bachem), M871 (Sollenberg et al. 2006; GalR2 antagonist, 100 nM–1 µM), M35 (nonselective GalR antagonist, 100 nM; Bachem; Mahoney et al. 2003), Y-27632 to inhibit Rho-associated kinase (Berghuis et al. 2007; 5 µM; Tocris), Gö 6976 (PKC inhibitor, 100 nM; Tocris), and NGF (50 ng/mL; Invitrogen). Ligands and growth media were replaced every other day. NGF was replenished daily. We have tested whether endogenous GalRs, including GalR2, undergo agonist-induced internalization in cultured basal forebrain neurons by exposure to galanin conjugated to saporin (GAL-SAP; 5 ng/mL for 8 h, Advance Targeting Systems) and using cell death as read-out 72 h later. Cell death was defined as the significantly decreased ratio of vesicular acetylcholine transporter (VAChT)+ neurons over Hoechst+ nuclei, the latter representing the total number of neurons and glia. All experiments were performed in triplicate with at least 2 independent observations (coverslips)/condition.
Target-Specific Isolation of Cholinergic Neurons
We have isolated cholinergic basal forebrain neurons from rat fetuses (E17.5) by a monoclonal antibody raised against the rat p75NTR (Clone: 192IgG, Millipore; Heckers et al. 1994) and coupled to paramagnetic beads coated with goat anti-mouse IgG (Invitrogen/Dynal; Berghuis et al. 2004). Dissociated cells were exposed to antibody constructs in suspension for 1.5 h at 4 °C. p75NTR-expressing cells were isolated by a magnetic particle concentrator (Invitrogen/Dynal). Neurons were freed from the antibodies by brief trypsination (0.1%, 5 min, 37 °C) and plated at a density of 50 000 (morphology or electrophysiology; Fig. 2E,F2) or 300 000 cells/well (western blotting; Fig. 3A) in PDL-coated 24-well plates (Berghuis et al. 2004). Cholinergic neurons were maintained in DMEM:astroglia-conditioned medium (1:1 mixture) (Berghuis et al. 2004), also containing B27 (2%), penicillin/streptomycin (100 U/mL), l-glutamine (2 mM), retinoic acid (1 nM), and glucose (0.5%).
Histo- and Cytochemistry, and Quantitative Morphometry
Multiple immunofluorescence labeling of fetal and adult mouse brains was performed by applying select cocktails of affinity-purified antibodies (Berghuis et al. 2007; Keimpema et al. 2010; Supplementary Table 1A). DyLight488/549/649- or carbocyanine (Cy)2/3/5-conjugated secondary antibodies (1:300; Jackson) were used to visualize primary antibody binding. Sections were coverslipped with Aquamount (Dako). Cells on coverslips were fixed in ice-cold 4% PFA in 0.1 M PB for 20 min and processed according to published protocols (Berghuis et al. 2004). F-actin was revealed by phalloidin-560 (1:500; Invitrogen). Hoechst 33 342 (Sigma) was routinely used as nuclear counterstain. Specimens were coverslipped by glycerol/gelatin (GG-1; Sigma).
Images were acquired on a 700LSM confocal laser-scanning microscope (Zeiss). Emission spectra were limited as follows: 405–480 nm (Hoechst), 505–530 nm (Cy2/DyLight488), 560–620 nm (Cy3/DyLight549), and 650–700 nm (Cy5/DyLight649). Colocalization of select histochemical marker pairs was verified by capturing serial orthogonal z image stacks at ×63 primary magnification (pinhole: 30 µm, 2048 × 2048 pixel resolution), resulting in image stacks usually containing 5–34 z levels. Z levels were continuous with their overlap automatically optimized by the ZEN2009 software. The coexistence of immunosignals, particularly for galanin and GAD67, was accepted if these were present without physical signal separation in ≤ 1.0-µm optical slices at ×40 (Plan-Neofluar ×40/1.30) or ×63 (Plan-Apochromat ×63/1.40) primary magnification, and overlapped in all 3 (x, y, and z) dimensions within individual cellular domains.
VAChT+ and vesicular GABA transporter (VGAT)+ profiles in the hippocampus (n = 8 fields of interest/mouse from n = 3 animals/treatment group) were defined as >5 pixel size in gray-scaled 2048 × 2048 pixel images, and automatically counted with fixed threshold settings using ImageTool v3.0. Intensities of somatic ChAT and VAChT immunofluorescence in cholinergic neurons in the horizontal diagonal band (HDB) of Broca were measured using ImageJ 1.45 s (n = 10 cells/mouse from n = 3 animals/group). Galanin+ fibers were classified as “en passant” if they traversed cholinergic territories without an identifiable postsynaptic target or “perisomatic” if they targeted the perikarya of cholinergic neurons (n > 10 neurons/observation). All measurements were performed on images acquired with equal intensities.
The topographic location of p75NTR+/ChAT+ neurons and galanin+ processes in the basal forebrain was determined in serial sections (140-µm intersection interval) of neonatal C57Bl6/N mice (P1, n = 3 brains). The distribution of cholinergic neuron subtypes was mapped onto generic brain overview plates. Images from in vitro experiments were processed using the Zeiss ZEN2011 software package to determine (Keimpema et al. 2010): (1) The length of the VAChT+ neurite (designated as the “primary neurite”/putative axon; µm), (2) the cumulative length of all neurites (µm), (3) the number of β-III-tubulin+ neurites emanating from cholinergic somata, (4) the surface of growth cones (µm2), and (5) the length of the F-actin+ neurite tip (µm). The brightness or contrast of confocal laser-scanning micrographs was occasionally linearly enhanced. Multipanel figures were assembled in CorelDraw X5 (Corel Corp.).
Whole-cell patch-clamp recordings were performed on isolated rat cholinergic neurons on days 3 (n = 6), 6 (n = 8), 12 (n = 5), and 16 (n = 5) in vitro (DIV). Neurons were superfused with oxygenated artificial cerebrospinal fluid (Berghuis et al. 2004) containing (in mM): 126 NaCl, 3.5 KCl, 10 mM glucose, 26 NaHCO3, 1.2 NaH2PO4, 2.5 CaCl2, and 1.3 MgCl2. Neurons were identified using differential interference contrast microscopy (Olympus BX51WI microscope) at ×40 magnification. The pipette solution contained (in mM): 125 K-gluconate, 4 ATP-Mg, 10 Na-phosphocreatine, 0.3 GTP, 20 KCl, 10 4-(2-Hydroxyethyl)piperazine-1-ethanesulfonic acid (pH 7.3; 310 mOsm/L). Microelectrodes were pulled from borosilicate glass (Harvard Apparatus) and had a resistance of 3.5–5 MΩ. Voltage and current-clamp recordings were obtained at 32–35 °C using a Cornerstone PC-One patch-clamp amplifier (Dagan). Electrical signals were filtered at 1 kHz and digitized at 20 kHz (ITC-18 converter; InstruTech). In current-clamp recordings, somatic current injections incrementing in 7 steps from 10 to 350 pA were delivered (500 ms) to visualize discharge wave patterns and to test spike accommodation. A hyperpolarizing step (−105 mV, 100 ms) was used to test whether cholinergic-like neurons produce spontaneous rebound action potentials (Wu et al. 2000) in vitro. Data were analyzed off-line using the IGOR Pro software (6.2, WaveMetrics) with custom-written routines.
Real-Time PCR of Cholinergic and GalR1-3 mRNA Transcripts
We used an RNAqueous micro kit (Life Technologies) to isolate total RNA from primary and cholinergic neuron-rich cultures. First-strand cDNA was produced by using SuperScript III reverse transcriptase (Invitrogen), and amplified with SYBR green PCR Master Mix (Applied Biosystems) also containing primer pairs (Supplementary Table 1B) in an ABI Prism 7000 Sequence Detector System (Applied Biosystems; 40 cycles). Relative mRNA levels were normalized to Gapdh and calculated as [relative change = 2−ΔΔCT], where CT is the point at which the amplification plot crosses the threshold and ΔΔCT = (ΔCT of the target mRNA − ΔCT of control mRNA). The ΔΔCT for control samples was calculated as ΔCT of control mRNA − ΔCT of sample mRNA and was set to 1 for all analyzed genes. Each sample was run in triplicate to avoid processing-related deviations.
GalR2-EGFP Expression In Vitro
The entire open reading frame encoding GalR2 was cloned into a pEGFP-N1 vector (Clontech) to produce GalR2-enhanced green fluorescent protein (EGFP) chimeras (Xia et al. 2004). Cholinergic neurons were transfected (5 µg DNA; 24 well format) with Lipofectamine 2000 (Invitrogen) in antibiotic-free medium for 30–45 min on 8 or 9 DIV. Cells were fixed 24 h posttransfection and analyzed.
Erk42/44 Phosphorylation Assay
Experiments were performed on cholinergic neuron-rich cultures by acutely applying galanin, AR-M 1896, M871, or M35 alone or in combination at 100 nM concentration on 3 DIV. Pharmacological challenges were terminated by washing the cells in ice-cold PB saline (pH 7.2). Cells on coverslips were lysed using Tris–HCl (100 mM, pH 7.4), NaCl (150 mM), NP-40 detergent (1%, Sigma), Na-deoxycholate (0.5%), sodium dodecyl sulfate (0.1%), ethylenediaminetetraacetic acid (2 mM), phosphatase inhibitor cocktail (1%, P5726, Sigma), and protease inhibitor cocktail (1%, S8820, Sigma). Protein samples were denatured in the presence of 4× NuPAGE SDS sample buffer and 10× NuPAGE reducing agent (both from Invitrogen) at 70 °C for 10 min. Proteins (20 µg/lane) were separated on NuPAGE 4–12% Bis-Tris gradient gels (Invitrogen) and transferred onto nitrocellulose membranes using iBlot Western detection kits (Invitrogen). Membranes were blocked in Tris-buffered saline containing 0.1% Tween 20 and 4% BSA (TBST) and exposed to rabbit anti-phospho-Erk 42/44 (1:2000; Cell Signaling Technology) and rabbit anti-Erk 42/44 (Berghuis et al. 2007) (1:2000; Cell Signaling) antibodies diluted in TBST at 4 °C overnight. Membranes were incubated with horseradish peroxidase-conjugated goat anti-rabbit IgG secondary antibody (1:10 000; Santa Cruz) for 1 h at room temperature. Immunoreactivity was visualized by enhanced chemiluminescence (Millipore). Immunoreactive protein bands were analyzed using ImageJ and normalized to time point (t) = 0 (min). Experiments were performed in duplicate.
Chemotropic Growth Cone Turning Assay
Basal forebrain neurons were tested on 4 DIV. Experiments were performed at 30–32 °C in oxygenated DMEM/F12 (1:1) (Berghuis et al. 2007). A microscopic gradient was produced by puffing ligands (diluted in DMEM/F12) at a constant pressure of 10 mBar at 45° calculated from the growth direction of the distal neurite tip (Berghuis et al. 2007). Micropipettes were loaded with AR-M 1896 at a concentration of 100 µM since ligands become approximately 1000-fold diluted at the growth cone. For antagonist experiments, neurons were incubated with M871 (1 µM) 2 h prior to the assay followed by continuous superfusion at 100 nM concentration during the experiments. Control experiments were performed by exposing growth cones to equimolar vehicle released under constant positive pressure from the micropipette. Filopodial motility was monitored by collecting images at 0.016 Hz, and analyzed in ImageJ 1.45 s. Extension or collapse of the longest filopodium was measured from the center of the growth cone and normalized to t = 0 (min). Filopodial turning angles were plotted in relation to their initial vector toward the source of the ligand gradient. Image sequences of n > 9 growth cones were analyzed per condition.
Statistical analyses were performed using Student's t-test (independent samples design) to specifically test drug effects versus the drug-free (vehicle) group. A P-level of <0.05 was considered statistically significant. Data were expressed as means ± SEM.
Galanin-Containing Terminals Appose Cholinergic Neurons in the Developing Mouse Basal Forebrain
Cholinergic neurons are embedded in a moderately dense meshwork of galanin-containing(+) processes in the adult mammalian brain, including humans (Melander et al. 1986; Chan-Palay 1988; Herbert and Saper 1990; Kordower and Mufson 1990; Mufson et al. 1993; Perez et al. 2001). In Alzheimer's disease, hypertrophic galanin+ fibers are generated and thought to represent survival signals for cholinergic neurons (Chan-Palay 1988; Beal et al. 1990; Mufson et al. 2005). However, the localization of galanin during fetal development of the basal forebrain, and in relation to cholinergic neurons likely migrating from preoptic proliferative areas (Marin et al. 2000; Zhao et al. 2003), remains unknown. First, we used multiple immunofluorescence histochemistry aided by high-resolution laser-scanning microscopy to localize galanin afferents to fetal mouse cholinergic neurons. We designated fetal cholinergic basal forebrain neurons as “projection cells” (Fig. 1A) if they coexpressed the low-affinity NGF receptor p75 (p75NTR) and choline-acetyltransferase (ChAT; Fig. 1B; Mesulam et al. 1984; Tremere et al. 2000. ChAT+/p75NTR+ neurons colonizing the medial septum (MS), HDB of Broca, and the magnocellular basal nucleus (MBN) were detected earlier (E18; Supplementary Fig. 1) than the ChAT+ (but p75NTR−) cholinergic lineage of the striatum (Fig. 1C), giving rise to interneurons (Marin et al. 2000; Herzog et al. 2004).
At E18, galanin+ processes emanated from a cluster of neurons in the dorsomedial hypothalamic area, and coursed horizontally toward pallidal territories (Supplementary Fig. 1). The density of galanin+ innervation of the pallidum remarkably increased by birth (Fig. 1A), including galanin+ processes running vertically and perpendicular to the midline in both the MS and the MBN, respectively. By the neonatal period, galanin+ noncholinergic cells, likely sources of local galanin in the MS/HDB complex, lined the subpial surface at the base of the basal forebrain, and had broadly ramifying processes coursing in cholinergic territories (Fig. 1D,D1). Galanin+ fibers were of medium caliber with pearl lace-like varicosities apposing cholinergic projection neurons in the MS (Fig. 1E,E4). Putative nerve endings were in direct contact with cholinergic perikarya (Fig. 1E1,E3) or proximal dendrites (Fig. 1E2), forming discrete, putative release sites (Fig. 1E4). Neonatal cholinergic projection neurons did not express appreciable levels of galanin perisomatically (Fig. 1E).
GalR2 Antagonism Disrupts VAChT Expression In Vivo
The juxtaposition of galanin+ terminals to cholinergic neurons and their morphological specializations suggests the involvement of galanin in the molecular control of the development of cholinergic projection neurons, reminiscent of sensory neurons (O'Meara et al. 2000; Mahoney et al. 2003). We tested this hypothesis by injecting M871 (0.06 or 0.6 mg/kg), a GalR2 antagonist (Sollenberg et al. 2006), during P2–7, coincident with the terminal morphogenesis of cholinergic basal forebrain neurons. A 0.06 mg/kg daily dose of M871 robustly decreased VAChT expression (P < 0.05; n = 4 mice/group; Fig. 2A,A1), while not significantly depleting VGAT content (Fig. 2A,A2). The above data on VAChT deregulation were substantiated by applying an elevated dose of M871 (0.6 mg/kg), resulting in a 30.2% loss of this transporter. Galanin expression remained unchanged in the M871-treated basal forebrain (Fig. 2A). Quantitative immunofluorescence histochemistry confirmed a pronounced reduction in somatic VAChT (P < 0.05)—but not ChAT (P > 0.1)—content (Fig. 2B–D1; n = 3 mice/group). In contrast, M871 did not significantly affect the density of hippocampal cholinergic (VAChT+) innervation (Fig. 2E,E1). These data give rise to the possibility that disrupted GalR2 signaling could compromise cholinergic differentiation locally (i.e. neuronal polarity and the initiation of neurite outgrowth) in the basal forebrain, while leaving established cholinergic projection pathways unaffected.
Cholinergic Neurons Preferentially Express GalR2 In Vitro
Next, we have defined in vitro conditions for cholinergic neurons, isolated from fetal mouse forebrain (E16.5) in mixed (Knusel and Hefti 1988; Fig. 3A–D3) or selectively enriched cultures (Schnitzler et al. 2008; Fig. 3E–G). The combination of tropomyosine receptor kinase A (TrkA; Sanchez-Ortiz et al. 2012), Islet-1 (Isl-1), a LIM homeobox gene (Elshatory and Gan 2008; Fig. 3A,A2), ChAT, and VAChT (Roghani et al. 1994; Fig. 3A2,B1) distinguished cholinergic projection neurons by 4 DIV. In vitro development of cholinergic neurons involved the molecular differentiation of their axon initial segment as revealed by ankyrin-G (Rasband 2010; Fig. 3B,B1). Vesicular glutamate transporter 3 (VGLUT3), identifying cholinergic or GABAergic interneurons in basal forebrain territories (Herzog et al. 2004), failed to colocalize with Isl-1 (n = 51 VGLUT3+ cells analyzed; Fig. 3C). We suggest that VGLUT3+ neurons in our cultures were noncholinergic, likely derived from the lateral septum (Riedel et al. 2008), which is coharvested with the MS/HDB complex during our dorsal dissection approach (Schnitzler et al. 2008). Other neurons in primary basal forebrain cultures expressed VGAT (Fig. 3D) or VGLUT1 (Fig. 3D1), recapitulating in vivo neuronal diversity (Manns et al. 2001; Gritti et al. 2006). In addition, neural progenitors and astroglia labeled for nestin (Fig. 3D2) and glial fibrillary acidic protein (Fig. 3D3), respectively. We then used selectively isolated cholinergic neurons (Supplementary Fig. 2A–C) to show that cholinergic neurons progressively acquired repetitive firing, hyperpolarization-evoked rebound action potentials, and synaptic coupling in culture (Supplementary Fig. 2C), reminiscent of cholinergic neurons probed ex vivo (Alreja et al. 2000; Wu et al. 2000).
Using a subtractive approach with VAChT and ChAT mRNAs as positive controls (Fig. 3E), we showed that selective isolation of fetal cholinergic projection neurons led to a significant enrichment in GalR2 mRNAs relative to GalR1/3 mRNAs found in primary (mixed) cultures. Notably, NGF, implicated in cholinergic differentiation (Hefti and Will 1987), significantly up-regulated GalR2 mRNA expression in cultured cholinergic neurons (53.91 ± 2.29 fold of control; P < 0.05, in triplicate).
Next, we addressed the subcellular targeting and recruitment of transiently overexpressed GalR2-EGFP chimeras to identify galanin's site of action. We found GalR2-EGFP accumulating in growth cones of cholinergic neurons in vitro (Fig. 3F,F1). Finally, we employed galanin fused to the ribosome-inactivating toxin saporin (Gal-SAP; Heckers et al. 1994) to assess if endogenous GalRs in cholinergic cells (Fig. 3G), including GalR2, are exposed on the cell surface, conferring signal competence. Gal-SAP induced cholinergic cell death [3.75 ± 0.40 (Gal-SAP) vs. 8.00 ± 1.24 (control) cells/field; P < 0.01, Fig. 3G], confirming that ligand-bound GalRs are endocytosed (Xia et al. 2004; Lemons and Wiley 2011). These data suggest that fetal cholinergic neurons can use surface-exposed GalR2s to drive their development. In particular, GalR2s in cholinergic growth cones may be poised to orchestrate neurite outgrowth.
GalR2 Activation Induces Cholinergic Neurite Outgrowth
Agonist-stimulated GalR2s have been implicated in the regeneration of sensory neurons by provoking neurite outgrowth (Mahoney et al. 2003; Hobson et al. 2006; Sanford et al. 2008). GalR2 activation used PKC and Erk1/2 signaling to couple to neurite outgrowth in sensory ganglia (Mahoney et al. 2003) and immortalized cell lines (Hawes et al. 2006). However, it is unknown whether GalR2 activation modulates morphological changes in cholinergic neurons and, if so, by which signal transduction mechanism.
First, we tested, using pharmacological challenges to selectively enriched cholinergic cultures, whether endogenous GalR2s induce second messenger signaling in cholinergic neurons by sampling Erk 42/44 phosphorylation (Hawes et al. 2006). Stimulation by AR-M 1896 (100 nM), a GalR2/3 agonist (Liu et al. 2001; Lu et al. 2005; Fig. 4A,A1), or galanin (100 nM) itself (Fig. 4A1) induced rapid Erk 42/44 phosphorylation reaching maximum by 15 min after agonist exposure. We verified GalR2 involvement by demonstrating the lack of Erk 42/44 phosphorylation upon coapplying M35 (100 nM), a nonselective GalR inhibitor (Mahoney et al. 2003), or M871 (100 nM), a peptide antagonist of GalR2 (Sollenberg et al. 2006; P < 0.05 vs. AR-M 1896; Fig. 4A2).
We hypothesized that GalR2 activation couples to neurite outgrowth in cholinergic neurons. The application of AR-M 1896 (1 µM) for 2 DIV [73.64 ± 3.45 µm (AR-M 1896) vs. 60.71 ± 3.65 µm (control), P < 0.05; Fig. 4B,C] or 4 DIV [114.87 ± 9.05 µm (AR-M 1896) vs. 85.52 ± 3.85 µm (control), P < 0.01; Fig. 4D] resulted in the progressive elongation of a single neurite whose enrichment in VAChT suggests prospective axon identity (Fig. 4B1; Keimpema et al. 2013). M871 abolished the AR-M 1896-induced neurite outgrowth in cholinergic cells (Fig. 4C,D). AR-M 1896 inhibited neurite formation by 4 DIV [number of neurites: 2.88 ± 0.38 (AR-M 1896) vs. 4.06 ± 0.31 (control), P < 0.05; Fig. 4D1], a finding that may be interpreted as GalR2-independent increase in neuronal polarity. AR-M 1896's effects on neurite outgrowth were dose-dependent (Fig. 4E). Notably, a tendency toward the pruning of the neurite arbor was only observed at an AR-M concentration of 1 µM (Fig. 4E1), confirming earlier data at 4 DIV (Fig. 4D1).
Next, we sought to identify the signal transduction mechanism linking GalR2 activation to cholinergic morphogenesis by inhibiting PKC (Mahoney et al. 2003) or Rho GTPase (Lang et al. 2007) signaling (Fig. 4F), the latter inducing rapid reorganization of the actin cytoskeleton. PKC inhibition by Gö 6979 did not significantly reduce neurite outgrowth as measured in post hoc identified cholinergic basal forebrain neurons in vitro. In contrast, Y27632, a Rho kinase inhibitor (Berghuis et al. 2007), abolished the AR-M 1896-induced elongation of the primary VAChT+ neurite [126.58 ± 7.24% (AR-M 1896) vs. 90.44 ± 8.00% (AR-M 1896 + Y27362), normalized to control, P < 0.05; Fig. 4F]. Similarly, Y27632 reduced the cumulative length of the entire neurite arbor of AR-M 1896-treated cholinergic neurons (Supplementary Fig. 3A). Nevertheless, Y27362 itself did not decrease total neurite length relative to controls (Supplementary Fig. 3A). This, together with the preserved morphology and F-actin network of cholinergic growth cones exposed to Y27632 (Supplementary Fig. 3B), excludes that Y27632 per se arrested neurite outgrowth by disrupting cytoskeletal integrity. In summary, these data suggest that GalR2 signaling may be particularly efficacious in modulating cytoskeletal dynamics in the motile neurite tip, including the growth cone.
GalR2 Activation Induces Growth Cone Remodeling
In the neonatal basal forebrain, galanin+ noncholinergic cells were neuroanatomically found interspersed with cholinergic neurons at the mediodorsal boundary of the globus pallidus (Fig. 5A). Galanin+ processes were frequently seen in the proximity of cholinergic neurites (Fig. 5A1), and fine-caliber galanin+ fibers surrounded cholinergic neurons (Fig. 5B,B1). The juxtapositioning of galanin+ varicosities and ChAT+ neurites suggests that cholinergic neurites can use galanin as a cue for directional growth. These morphological arrangements together with the efficacy of Y-27632 to modulate GalR2-mediated growth responses led us to suggest that GalR2-mediated Rho GTPase activation can lead to cytoskeletal reorganization affecting cholinergic growth cone motility and size.
AR-M 1896 dose-dependently modulated the surface area of cholinergic growth cones (Fig. 5C) in vitro. Furthermore, all AR-M 1896 concentrations tested increased the total length of the motile, F-actin+/β-III-tubulin− distal neurite segment at 4 DIV [47.29 ± 5.30 µm (AR-M 1896, 10 nM) vs. 27.78 ± 3.09 µm (control), P < 0.01; Fig. 5C1]. These data suggest that GalR2 activation is permissive for growth cone reorganization by mobilizing the actin cytoskeleton (Fig. 4F). Therefore, we have determined whether AR-M 1896 microgradients released onto the growth cones of basal forebrain neurons affect growth cone morphology, particularly filopodial dynamics (Fig. 5D,D1). GalR2 stimulation resulted in the extension of the lead filopodia facing the source of the AR-M 1896 source [20.7 ± 4.0% (AR-M 1896) vs. 1.3 ± 1.3% (control), P < 0.01; Fig. 5D,D3]. M871, a GalR2 antagonist, markedly reduced AR-M 1896-induced filopodiagenesis (Fig. 5D3). We attribute the residual filopodial growth upon combined AR-M 1896 and M871 treatments to nonselective GalR engagement, since the concentration of AR-M 1896 (i.e. 100 µM at the pipette tip) exceeded the inhibitory concentration (IC50) of this compound at both GalR2 (1.76 nM), GalR3 (271 nM), and GalR1 (879 nM). GalR2 activation did not affect growth cone steering decisions within 30–60 min of AR-M 1896 application (data not shown), recapitulating earlier findings in sensory neurons (Sanford et al. 2008). Thus, we conclude that GalR2 functions to modulate growth cone morphology (rather than directional growth) during neuronal differentiation.
GalR2 Dose-Dependently Modulates NGF-Induced Cholinergic Differentiation
NGF induces galanin (Planas et al. 1997) and GalR2 expression (above). Therefore, we studied whether NGF at saturating concentrations (50 ng/mL) affects GalR2-induced neurite outgrowth in mouse cholinergic basal forebrain neurons. Like GalR2 agonists, NGF induced elongation of the VAChT+ neurite [124.86 ± 9.16 µm (NGF) vs. 104.21 ± 7.37 µm (control), P < 0.05; Fig. 6A], and increased the length of the entire arbor emanating from cholinergic neurons on 4 DIV [306.77 ± 29.49 µm (NGF) vs. 229.01 ± 21.70 µm (control), P < 0.05]. AR-M 1896 treatment coincident with NGF failed to augment NGF-induced growth responses (Fig. 6A and Supplementary Fig. 3C).
The failure of AR-M 1896 to facilitate cholinergic neurite growth in the presence of NGF might be due to NGF's ability to induce premature growth cone differentiation, restricting neurite elongation (Keimpema et al. 2013). Therefore, we measured growth cone size in the presence of NGF, AR-M 1896, or both. Our data uncovered a dose–response relationship, with AR-M 1896 and NGF interaction prevailing at 100 nM AR-M 1896 concentration [187.78 ± 42.34 µm2 (NGF) vs. 256.67 ± 67.69 µm2 (NGF/AR-M 1896); Fig. 6B1], as opposed to antagonism when coadministering 1 µM AR-M 1896 [80.76 ± 16.51 µm2, P < 0.05 vs. NGF; Fig. 6B,B1]. We noted AR-M 1896-induced cholinergic cell death at 1 µM concentration (data not shown) corroborating previous findings on PC12 cells (Tofighi et al. 2008). Therefore, we attribute the additive NGF/AR-M 1896 responses as physiologically relevant. If NGF induces axonal differentiation and cytoskeletal stability, then it must reduce the length of the F-actin-rich distal neurite tip, contrasting GalR2 effects. Indeed, we found that NGF eliminated the GalR2-induced dissociation of F-actin or β-III-tubulin-rich neurite domains (Fig. 6B2), where β-III-tubulin suggested prevailing NGF-dependent cytoskeletal stability. These data suggest that NGF can use galanin to amplify prodifferentiation signals in developing cholinergic neurons.
The Neurochemical Identity of Galanin-Containing Presynapses
Galanin can coexist with multiple neurotransmitters in the adult rodent brain, including a subset of rat cholinergic neurons (Melander et al. 1985; Senut, de Bilbao, et al. 1989). Therefore, we deployed high-resolution neuroanatomy and imaging tools to identity galanin+ afferents onto cholinergic neurons of the MS/HDB complex in neonates and adult mice. We found, albeit sparsely, galanin in a subset of en passant ChAT+ fibers (Fig. 7A,B2), likely recurrent cholinergic collaterals, recapitulating previous data from adult rodents (Melander et al. 1986). Nevertheless, the majority of galanin was present in GABAergic processes, identified by their expression of GAD65/67, the enzyme rate-limiting GABA synthesis (Fig. 7C,C3). ChAT and GAD can colocalize in the adult forebrain (Brashear et al. 1986; Bayraktar et al. 1997). However, we did not find detectable galanin levels in ChAT+/GAD65/67+ processes at the developmental periods studied (data not shown). Next, we probed whether the neurochemical specificity of galanin+ terminals is maintained during postnatal life. We showed that galanin continues to coexist with GAD65/67 in the basal forebrain (Fig. 7D,D1), including in a GAD67-GFP reporter line (Tamamaki et al. 2003) used as a genetic control (Fig. 7E,E1). Upon determining the number of galanin+ boutons that colabeled for GAD65/67, we find that perisomatic (63.35 ± 7.55%) rather than en passant boutons (35.44 ± 3.63%) were double-labeled (Fig. 7F). Focal galanin accumulation in varicose structures (Fig. 7D1–F) led us to histochemically verify that galanin localizes to presynapses. By using vesicle-associated membrane protein 2 (VAMP2), a component of the presynaptic vesicle docking/fusion machinery (Berghuis et al. 2007), as presynaptic marker we identified galanin in synaptic terminals along cholinergic dendrites (Fig. 7G,G2) and perikarya (Fig. 7H,H2). High-resolution, 3-dimensional reconstruction of select synapses confirmed galanin's presynaptic localization (Fig. 7I,I2). In summary, these data lend support to a developmental switch of GABAergic neurotransmission (Ben-Ari 2002) rather than a change in galanin signaling per se to determine the physiological sign of galanin action if coreleased during synaptic activity.
Galanin+ innervation of the rodent forebrain is neurochemically heterogeneous (Melander et al. 1986; Gabriel et al. 1995; Sherin et al. 1998; Xu et al. 1998), producing diverse behavioral and metabolic impairments upon galanin deprivation. The colchicine or lesion induced up-regulation of galanin in rodent basal cholinergic forebrain neurons (Melander et al. 1985; Senut, Menetrey, et al. 1989; Cortes et al. 1990), and the reorganization of galanin+ afferents in the human basal forebrain in Alzheimer's disease (Chan-Palay and Asan 1989; Mufson et al. 1993) increased interest in this neuropeptide during the recent past. Even though the contribution of galanin signaling to the development, as well as postnatal maintenance of cholinergic neurotransmitter systems, is highlighted by the loss of cholinergic projection neurons in aging (Senut, de Bilbao, et al. 1989) and in galanin−/− mice (O'Meara et al. 2000), the distribution and function of galanin signaling in the formation of the cholinergic basal forebrain remained hitherto unexplored. Here, we show the progressive increase of galanin innervation of the basal forebrain by birth, with galanin+ processes terminating on cholinergic neurons.
We discerned 2 subclasses of galanin+ fibers: medium caliber, varicose fibers were primarily GABAergic in the MS and HDB. Alternatively, a dense meshwork of fine galanin+ processes of as yet unknown phenotype populated the MBN and associated pallidal regions. Galanin accumulation in GABAergic afferents is not entirely unexpected, since most of the galaninergic neurons in the ventrolateral preoptic area are GAD+ (Sherin et al. 1998), and GABAergic interneurons in laminae I–III of the adult spinal dorsal horn also contain this neuropeptide (Simmons et al. 1995). Varicose galanin labeling in GAD65/67+ axons was confined to subcellular domains enriched in VAMP2, a presynaptic marker, suggesting activity-dependent vesicular galanin release onto cholinergic neurons. Nevertheless, galanin systems are known to exhibit evolutionary differences in their topography (Le Maitre et al. 2013) calling for the functional assessment of developmental galanin action in higher-order mammals.
GABA is thought to be excitatory in the developing brain (Cherubini et al. 1991), and required for activity-dependent developmental processes, such as neuronal migration and synaptogenesis (Manent et al. 2005). GABAB receptor stimulation has been demonstrated to increase ChAT activity in developing neurons (Kenigsberg et al. 1998), suggesting maturation, while GABA application stimulates axonal elongation (Ageta-Ishihara et al. 2009). Since GABAergic interneurons populate the basal forebrain earlier than cholinergic differentiation commences (Lauder et al. 1986), they can be positioned to provide activity-dependent trophic support to GalR2+ cholinergic projection neurons by coreleasing excitatory GABA and galanin, facilitating cholinergic cell migration, differentiation, and neurite outgrowth. However, GABA is inhibitory in the adult brain, and the maintenance of excitation/inhibition balance is critical for cholinergic neurons to survive under disease conditions associated with neuronal hyperexcitability (e.g. Alzheimer's disease). Therefore, we propose that hypertrophic galanin+ processes in Alzheimer's disease (Chan-Palay 1988; Mufson et al. 1993; Counts et al. 2009) might reflect inhibitory synapse reorganization. If hypertrophy is reflective of the reduced ability of neuropeptide release, rather than of increased galanin availability as generally viewed, then the lack of inhibitory GABA/galanin corelease might be expected to perpetuate cholinergic demise. This concept is compatible with earlier data identifying galanin to inhibit cholinergic neurotransmission (Fisone et al. 1987), and being neuroprotective in models of nerve injury (Hobson et al. 2006).
The GalR subtype expressed by a specific cell type and in a discrete cellular context can diversify galanin actions, such as morphogenic versus survival signals (Lang et al. 2007). We found GalR2 expression in fetal cholinergic neurons, the GalR receptor subtype implicated in neurite outgrowth also from PC12 cells and regenerating adult sensory neurons (Mahoney et al. 2003; Hawes et al. 2006; Hobson et al. 2006). GalR2 is widely expressed during the late embryonic/early postnatal period in mouse forebrain, with highest mRNA levels in cortical and thalamic territories (Burazin et al. 2000). GalR2 is different from GalR1 in that its mRNA expression is reduced in adult, suggesting its predominant contribution to developmental processes. Further support for the temporal control of GalR2 expression is provided by cDNA microarray technology (D'Onofrio et al. 2011), showing GalR3 (but not GalR1 or GalR2) mRNA transcripts in the forebrain of adult mice. Here, we demonstrate the significant contribution of GalR2 to cholinergic differentiation, particularly axonal growth. This is not entirely unexpected since galanin and GalR2 mRNA are up-regulated upon peripheral nerve injury (Hokfelt et al. 1987; Villar et al. 1989; Burazin and Gundlach 1998). Yet a critical difference from in vitro systems employing adult (sensory) neurons or immortalized cells (Hawes et al. 2006) is that GalR2 in fetal cholinergic neurons increases cytoskeletal dynamics by activating Rho-family GTPases, rather than coupling to PKC (Mahoney et al. 2003), poised to control microtubule elongation during neurite outgrowth and growth cone motility. Although the molecular basis of differential signal transduction remains unclear, we interpret the recruitment of Rho-family GTPases as a response adequate to allow GalR2-mediated filopodiagenesis vital to control growth cone steering decisions.
We found that NGF induces GalR2 mRNA expression in cholinergic neurons. This is significant, since enhanced signaling at GalR2s can augment either the differentiation promoting (Mahoney et al. 2003; Hobson et al. 2006) or apoptotic (Tofighi et al. 2008) (or both) effects of galanin. The predominance of GalR2 in modifying cytoskeletal stability in cholinergic neurons is supported not only by the dynamics of filopodia, but also by the NGF-induced elimination of F-actin from distal growth domains of cholinergic neurons. The importance of these observations is that a galanin–NGF interplay during cholinergic neuron development utilizes a parallel pathway to induce neurite outgrowth and is unlikely to be additive. This interaction can persist at the level of the growth cone, which is particularly appealing in view of the Rho GTPase-dependence of cytoskeletal instability. While NGF-induced differentiation of the growth cone relies on cytoskeletal stability, this could be modulated by GalR2 signaling, suggesting that GalR2 signaling can be permissive for the maintenance of cytoskeletal dynamics to promote the forward advance of undifferentiated growth cones.
In summary, our results advance present knowledge on the molecular regulation of galanin signaling by neurotrophins, the neurochemical identity of galanin+ afferents innervating cholinergic neurons, the developmental mechanisms and their sites of action regulated by galanin. Our findings also emphasize that pharmacological interventions efficacious on fetal cholinergic neurons might compromise adult cholinergic neurotransmission and inadvertently exacerbate pathological processes due to a developmental shift in the physiological sign of the fast neurotransmitter system containing galanin. Future studies are warranted to explore the origins of galanin+/GABA+ afferents to cholinergic neurons, since these might unravel a novel level of organizational complexity of the basal forebrain in relation to sleep control, arousal (Sherin et al. 1998), and other complex behaviors (Semba 1991).
T.Hö. and T.Ha. conceived the general ideas of this study; E.K., K.Z., T.Hö., and T.Ha. designed experiments; E.K., K.Z., S.B., P.B., M.B.D., and J.M. performed experiments; R.S., P.G.M.L., Z.D.X., J.R., Ü.L., and B.L. contributed unique reagents and technologies; E.K., T.Hö., and T.Ha. wrote the manuscript. All authors commented on the manuscript and approved its submission.
This work was supported by the Swedish (Medical) Research Council (T.Ha. and T.Hö.), National Institutes of Health Grant DA023214 (T.Ha.), Hjärnfonden (T.Ha. and T.Hö.), the Novo Nordisk Foundation (T.Ha. and T.Hö.), the Marianne and Marcus Wallenberg Foundation (T.Hö.), and the Knut and Alice Wallenberg Foundation (T.Hö.).
We thank M. Zilberter for assistance with microgradient assays, E. Theodorsson for rabbit anti-galanin antibodies, Y. Yanagawa for GAD67-GFP mice, and the “CLICK” Imaging Platform of Karolinska Institutet for infrastructural support. Conflict of Interest: None declared.