|Home | About | Journals | Submit | Contact Us | Français|
Author contributions: K.L.T., K.L.B., M.B.E., F.W.K., and J.C.C. designed research; K.L.T., K.L.B., M.B.E., and A.G.T. performed research; C.E.N. contributed unpublished reagents/analytic tools; K.L.T., K.L.B., M.B.E., F.W.K., and A.G.T. analyzed data; K.L.T. and J.C.C. wrote the paper.
Significant migration cues are required to guide and contain newly generated rodent subventricular zone (SVZ) neuroblasts as they transit along the lateral ventricles and then through the anterior forebrain to their ultimate site of differentiation in the olfactory bulbs (OBs). These cues enforce strict neuroblast spatial boundaries within the dense astroglial meshwork of the SVZ and rostral migratory stream (RMS), yet are permissive to large-scale neuroblast migration. Therefore, the molecular mechanisms that define these cues and control dynamic interactions between migratory neuroblasts and surrounding astrocytes are of particular interest. We found that deletion of EphA4 and specifically ablation of EphA4 kinase activity resulted in misaligned neuroblasts and disorganized astrocytes in the RMS/SVZ, linking EphA4 forward signaling to SVZ and RMS spatial organization, orientation, and regulation. In addition, within a 3 week period, there was a significant reduction in the number of neuroblasts that reached the OB and integrated into the periglomerular layer, revealing a crucial role for EphA4 in facilitating efficient neuroblast migration to the OB. Single-cell analysis revealed that EPHA4 and its EFN binding partners are expressed by subpopulations of neuroblasts and astrocytes within the SVZ/RMS/OB system resulting in a cell-specific mosaic, suggesting complex EphA4 signaling involving both homotypic and heterotypic cell–cell interactions. Together, our studies reveal a novel molecular mechanism involving EphA4 signaling that functions in stem cell niche organization and ultimately neuroblast migration in the anterior forebrain.
SIGNIFICANCE STATEMENT The subventricular zone neurogenic stem cell niche generates highly migratory neuroblasts that transit the anterior forebrain along a defined pathway to the olfactory bulb. Postnatal and adult brain organization dictates strict adherence to a narrow migration corridor. Subventricular zone neuroblasts are aligned in tightly bundled chains within a meshwork of astrocytes; however, the cell–cell cues that organize this unique, cell-dense migration pathway are largely unknown. Our studies show that forward signaling through the EphA4 tyrosine kinase receptor, mediated by ephrins expressed by subpopulations of neuroblasts and astrocytes, is required for compact, directional organization of neuroblasts and astrocytes within the pathway and efficient transit of neuroblasts through the anterior forebrain to the olfactory bulb.
The rodent subventricular zone (SVZ) stem cell niche lies along the lateral walls of the brain's lateral ventricles and maintains its regenerative capacity throughout life (Gage, 2000; Braun and Jessberger, 2014; Conover and Todd, 2016; Lim and Alvarez-Buylla, 2016). Within this tightly regulated niche environment, neuroblasts migrate along the lateral ventricles as densely packed chains and then transit the anterior forebrain via the rostral migratory stream (RMS), en route to the olfactory bulbs (OBs) (Luskin, 1993; Lois et al., 1996; Capilla-Gonzalez et al., 2015; Liang et al., 2016). The SVZ and RMS contain a unique organization of densely packed astrocytes that form a meshwork around the neuroblasts (Doetsch and Alvarez-Buylla, 1996; Lois et al., 1996; Alvarez-Buylla and Lim, 2004), effectively restricting highly migratory cells to the SVZ/RMS core (Gengatharan et al., 2016). In human infants, newly described robust migration pathways (RMS and a unique medial migratory stream) located by the anterior portion of the lateral ventricles supply inhibitory neurons to the olfactory bulb and frontal lobe, including the prefrontal cortex (Sanai et al., 2007, 2011; Paredes et al., 2016). These pathways display an organization reminiscent of the rodent RMS with chains of migrating neurons flanked by astrocytes. The molecular players responsible for the dynamic organization of neuroblasts and astrocytes in postnatal and adult (in the rodent) neuroblast migration, and the cell–cell mechanisms that allow efficient migration of neuroblasts through the anterior forebrain are unclear. However, cell—cell-mediated signaling resulting in repulsion and/or attraction is likely to be a key regulator.
In the adult rodent SVZ, Eph–ephrin interactions have been implicated in the regulation of neuroblast chain formation (Conover et al., 2000), cell cycle rate and survival (Holmberg et al., 2005; Ricard et al., 2006), progenitor cell proliferation (Conover et al., 2000; Theus et al., 2010; del Valle et al., 2011), and neuronal fate commitment (Aoki et al., 2004; Yin et al., 2004). The Eph receptors comprise the largest known family of receptor tyrosine kinases. Along with their ligands, they are divided into A and B classes, based on homology and binding affinities. EphA receptors (1–8, 10 in mammals) typically bind glycosylphosphatidylinositol-linked ephrinA ligands (1–6), whereas EphB receptors (1–4, 6) prefer single-pass transmembrane ephrinB molecules (1–3) (Klein, 2001). Eph–ephrin signaling requires cell–cell contact and can proceed either unidirectionally through the receptor-bearing cell (“forward” or “classical” signaling) or the ligand-bearing cell (“reverse” signaling), or bidirectionally through both the receptor-bearing and ligand-bearing cells (Martínez and Soriano, 2005; Pasquale, 2008). Although it was previously thought that Eph–ephrin signaling mediated primarily contact-repulsive interactions, evidence suggests adhesive, contact-attractant mechanisms also occur (Holmberg et al., 2005; Pasquale, 2008; Cramer and Miko, 2016; Kania and Klein, 2016).
Because the EphA4 receptor is robustly expressed by cells in the SVZ (Conover et al., 2000) and has the unique ability to bind most ephrinA and ephrinB ligands, we proposed that it alone may have broad ability to control SVZ/RMS functions. We found that, in the absence of EphA4, and more specifically forward signaling through its kinase domain, populations of SVZ and RMS neuroblasts were loosely aligned, lacked compact bundling of fasciculated chains, and displayed aberrant migration into neighboring parenchymal brain structures. Disoriented neuroblast migration in EphA4−/− mice ultimately resulted in fewer neuroblasts reaching the periglomerular layer (PGL) of the OB. Single-cell analysis revealed a complex pattern of EPHA4 and EFN expression in which EFN′s were coexpressed with each other and/or with EPHA4 on both neuroblasts and astrocytes. This indicated that both homotypic (neuroblast–neuroblast, astrocyte–astrocyte) and heterotypic (neuroblast–astrocyte) cell–cell interactions could facilitate EphA4 functions in this region. In summary, our findings reveal a novel EphA4-dependent forward signaling mechanism essential for the regulated control of neuroblast and astrocyte organization within the SVZ and RMS. These results provide the first example that kinase-dependent EphA4 signaling is necessary for the unique organization of the postnatal and adult SVZ/RMS neurogenic niche and this strict organization supports efficient neuroblast migration.
EphA4+/+ and EphA4−/− (Helmbacher et al., 2000), a gift from Dr. Maria J. Donoghue, Georgetown University, were housed and bred in our vivarium. EphA4−/− mice exhibit complete atrophy of the peroneal muscles of the hindlimbs, resulting in a “club-foot” phenotype and abnormal gait. These deficiencies are due to absence of the peroneal nerve, caused by misrouting of axons emerging from L3, L4, and L5 spinal nerves and sciatic plexus in the absence of EphA4 (Helmbacher et al., 2000). In all other aspects, EphA4−/− mice mature normally and live through adulthood. Kinase-dead (KD)-EphA4eGFP mice, displaying a similar phenotype to EphA4−/− mice (Kullander et al., 2001), were received as a gift from Drs. Klas Kullander and Christiane Peuckert (Uppsala University). hGFAP:mRFP1 (Hirrlinger and Dringen, 2005), a gift from Dr. Frank Kirchhoff, B6-Tg(GFAP-Cre/ERT2)1Kdmc (Casper et al., 2006), a gift from Dr. Ken McCarthy (University of North Carolina at Chapel Hill), and B6.Cg-Gt(ROSA)26Sortm14(CAG-tdTomato)Hze/J (ROSA26-tdT; The Jackson Laboratory) mice were also housed and bred in our vivarium. Animal procedures were performed under protocols approved by the Institutional Animal Care and Use Committee of the University of Connecticut and conform to National Institute of Health and Association for Assessment and Accreditation of Laboratory Animal Care guidelines.
Male mice were anesthetized with isoflurane (Butler Schein) and perfused transcardially with 0.9% saline followed by 4% (v/v) PFA (Electron Microscopy Sciences) in 0.1 m PBS, pH 7.4. Brains were removed and postfixed in 4% PFA overnight (at least 12 h) followed by three 40 min washes in PBS. Brains were sectioned (50 μm) on a vibratome (VT-1000S; Leica) and then blocked for 1 h at room temperature with 10% (v/v) normal horse or goat serum (Thermo Fisher Scientific) in PBS with 0.1%–0.5% Triton X-100 (Sigma). Tissue sections were incubated overnight at 4°C with primary antibodies prepared in blocking solution using the following concentrations: goat anti-doublecortin (Dcx) 1:200 (Santa Cruz Biotechnology, RRID: AB_2088494), guinea pig anti-Dcx 1:1000 (Millipore, RRID: AB_1586992), rabbit anti-Dcx 1:200 (Santa Cruz Biotechnology, AB_2088488), rabbit anti-GFAP 1:1000 (DAKO, RRID: AB_10013382), mouse anti-GFAP 1:400 (Millipore, RRID: AB_2109815), rat anti-GFAP clone 2.2B10 1:250 (Thermo Fisher Scientific, RRID: AB_86543), goat anti-PECAM-1 1:50 (Santa Cruz Biotechnology, RRID: 632174), rabbit anti-EphA4 1:50 (Abcam, RRID: AB_304857), rat anti-EphA4 (Abcam, RRID: AB_2099346), mouse anti-EphA4 1:50 (Thermo Fisher Scientific, RRID: AB_2533301), goat anti-EphA4 1:20 (R&D Systems, AB_2099371), rabbit anti-ephrinA2 1:50 (Santa Cruz Biotechnology, RRID: AB_631409), rabbit anti-ephrinA3 1:50 (Santa Cruz Biotechnology, RRID: AB_2097777), rabbit anti-ephrinA5 1:50 (Santa Cruz Biotechnology, RRID: AB_2097911), or rabbit anti-ephrinB1/2/3 1:50 (Santa Cruz Biotechnology, RRID: AB_631413). Sections were then washed 3 times for 10 min each with 0.1% Triton X-100 in PBS and then incubated 1 h at room temperature with appropriate fluorescent secondary antibodies (1:1000 in blocking solution; AlexaFluor, Thermo Fisher Scientific). Following antibody treatment, tissue samples were washed 3 times for 10 min each in PBS, counterstained for cell nuclei with 10 mg/ml DAPI (Thermo Fisher Scientific, RRID: AB_2629482) for 5 min, mounted onto slides, and coverslipped using Aqua-Poly/Mount (Polysciences).
For whole-mount preparations, the entire lateral wall of both lateral ventricles was removed, as previously described (Doetsch and Alvarez-Buylla, 1996). Whole mounts of the lateral wall were washed, blocked, and then incubated with primary and secondary antibody solutions as described above.
Whole mounts were imaged with a Zeiss Axioskop 2+ microscope (Carl Zeiss) and Retiga 1300 EX digital camera (Q-Imaging System). Vibratome sections were imaged with a Leica TCS SP2 confocal system together with a Leica DMIRE2 microscope and Leica confocal software or with a Zeiss Imager.M2 microscope with ApoTome (Carl Zeiss) using Stereo Investigator software (MBF Bioscience) and Orca-R2 digital camera (Hamamatsu Photonics). Images were processed using ImageJ (National Institutes of Health) and Photoshop CS (Adobe).
Male mice were perfused transcardially with 0.9% saline, followed by 2% PFA/2.5% glutaraldehyde in 0.1 m phosphate buffer (PB), pH 7.4. Heads were postfixed by immersion overnight in 2% PFA/2.5% glutaraldehyde in 0.1 m PB; brains were removed and washed in PB three times for 40 min each. Three 300 μm sections of the anterior forebrain were cut with a vibratome, and the area surrounding the ventricle was removed from each. Briefly, sections were postfixed with 2% osmium tetroxide in 0.1 m PB for 1.25 h, unpinned, and dehydrated through a graded ethanol series. Sections were en bloc stained in 2% uranyl acetate at the 70% ethanol step for 1.5 h. After dehydration, sections were washed twice in propylene oxide and embedded in a SPI-PON 812 (SPI Supplies)/Araldite 506 (Ernest F. Fullam). After polymerization, thin sections were cut with a diamond knife, placed onto Formvar-coated slot grids, and heavy metal stained with uranyl acetate and lead citrate. Electron micrograph montages were constructed as previously described (Doetsch et al., 1997; Baker et al., 2006; Luo et al., 2006).
Fifty-micrometer coronal sections from adult male mice were immunolabeled for Dcx (Santa Cruz Biotechnology), and fluorescent images were taken between coordinates 1.75 mm and 3.00 mm anterior to bregma using a Zeiss Imager.M2 microscope (Carl Zeiss) and an Orca-R2 digital camera (Hamamatsu Photonics). The Dcx-immunoreactive region of the RMS from every other section was traced using stereology software (Stereo Investigator, MBF Biosciences), and contours were averaged. Six animals were used for each group (EphA4+/+, EphA4−/−, and KD-EphA4eGFP/eGFP). Statistical significance was evaluated using one-way ANOVA.
To examine the RMS cross-sectional area during the period of RMS reorganization in perinatal mice, 50 μm coronal sections were immunolabeled as described above, and fluorescent images were taken between the anterior edge of the lateral ventricle and the caudal end of the OB to include the entire RMS for mice of ages P6 and P12. Three mice were used for each group (EphA4+/+, EphA4−/−, and KD-EphA4eGFP/eGFP) at each age examined. The average cross-sectional area at each age was calculated by tracing the Dcx+ area of every other tissue section using Stereo Investigator, as described above, and statistically significant differences were evaluated using one-way ANOVA.
Equal numbers of adult male and female mice were injected once with 150 mg EdU/kg followed by a second injection of 100 mg EdU/kg 24 h later. At 21 d following the second injection, mice were perfused transcardially with 0.9% saline followed by 4% (v/v) PFA (Electron Microscopy Sciences) in 0.1 m PBS. Brains were removed and postfixed in 4% PFA overnight followed by three 40 min washes in PBS. Brains were sectioned coronally (50 μm), and EdU was visualized using the Click-It EdU AlexaFluor-647 Imaging Kit according to the manufacturer's instructions. Sections were imaged (Axioskop 2+, Zeiss), and EdU+ cells were counted in coronal sections of the OB, RMS, and SVZ using ImageJ (National Institutes of Health). Every fifth slice was analyzed beginning at 4.74 mm and ending at −0.14 mm relative to bregma. Six mice from each group (EphA4+/+ and EphA4−/−) were analyzed, and statistically significant differences were determined using an unpaired t test.
Mice were injected with 300 mg BrdU/kg 2 h before perfusion. BrdU immunostaining of 50 μm sections was conducted as described previously (Cameron and McKay, 2001). Briefly, sections were mounted on slides, heated in 0.1 m citric acid, pH 6.0, for antigen retrieval, and treated with trypsin and 2 m hydrochloric acid for 1 h. Sections were then incubated with mouse anti-BrdU 1:100 (BD Biosciences, RRID: AB_10015222) and rabbit anti-Ki67 1:1000 (Novocastra; Leica Biosystems, RRID: AB_442102) followed by the appropriate fluorescent secondary antibodies (1:1000 in blocking solution; AlexaFluor, Thermo Fisher Scientific). Sections were imaged (Axioskop 2+, Zeiss), and the BrdU+ and Ki67+ cells were counted in coronal sections of the SVZ using Openlab 3.1.5 imaging software (Improvision) in 34 anterior forebrain sections (50 μm), from coordinates 0.5–1.4 anterior, relative to bregma. Three mice were used for each group, and statistical analyses were performed using an unpaired t test.
Fixed coronal sections, 50 μm thick, were immunostained with anti-caspase-3 (R&D Systems, RRID: AB_2243952) to label apoptotic cells. Caspase-3+ cells in the SVZ were counted in 22 sections between coordinates 0.5–1.4 anterior/posterior, relative to bregma. In the RMS and OB, caspase-3+ cells were counted in every section within the entire region for both structures. At least three mice were used for each group. Statistical analyses were performed using an unpaired t test.
The RMS was dissected from two male mice, minced, and combined followed by RNA extraction using the RNeasy Mini kit (QIAGEN) according to the manufacturer's instructions. cDNA was prepared with the SuperScriptIII First-Strand cDNA synthesis kit (Thermo Fisher Scientific) using random hexamer primers according to the manufacturer's instructions. qRT-PCR was performed on a Bio-Rad C1000 thermal cycler with CFX96 Real-Time system using the following mouse specific TaqMan assays (20×; Thermo Fisher Scientific): Gapdh (Mm99999915_g1), EPHA4 (Mm00433056_m1), EFNA1 (Mm01212795_m1), EFNA2 (Mm00433011_m1), EFNA3 (Mm01212723_g1), EFNA4 (Mm00433013_m1), EFNA5 (Mm01237700_m1), EFNB1 (Mm00438666_m1), EFNB2 (Mm00438670_m1), and EFNB3 (Mm00433016_m1). Data were analyzed with the Bio-Rad CFX Manager software. A total of 6 mice were analyzed, and statistically significant differences were evaluated using one-way ANOVA. Gapdh served as a reference gene for each sample.
Single-cell suspensions of neuroblasts and astrocytes were isolated per experiment from two male P25 hGFAP:mRFP1 mice (Hirrlinger and Dringen, 2005) and two male P13 B6-Tg(GFAP-Cre/ERT2)1Kdmc × ROSA26-tdT mice (all GFAP+ cells fluoresce red), respectively. Cre was induced in B6-Tg(GFAP-Cre/ERT2)1Kdmc × ROSA26-tdT mice by daily intraperitoneal injection (P6-P8) of 0.2 mg 4-hydroxytamoxifen (Sigma). The region of the anterior forebrain containing the RMS was dissected, minced, and digested in 2.5 mg/ml protease XXIII (from Aspergillus melleus; Sigma) in high sucrose dissociation solution (30 mm Na2SO4, 2 mm K2SO4, 10 mm HEPES, 185 mm sucrose, 10 mm glucose, 0.5 mm CaCl2, 6 mm MgCl2, and 5 mm ascorbic acid, pH 7.4) for 10 min at 34°C. Digested tissue was collected by centrifugation and resuspended in high sucrose dissociation solution with 1% BSA and 1 mg/ml trypsin inhibitor from chicken egg white (TI; Sigma). This was repeated twice with chilled high sucrose dissociation solution with 1% BSA and 1 mg/ml TI. Suspensions were then sequentially triturated using a series of flame-polished glass pipettes of progressively decreasing bore diameter (~300–100 μm) and then passed through 70 and 40 μm mesh filters (BD Biosciences) to remove debris and cell clumps. Myelin was removed using magnetically conjugated anti-myelin removal beads (Miltenyi) according to the manufacturer's instructions. Isolation of single cells via FACS was performed on a BD FACSAria II system operated by FACSDiva software, and single cells were sorted directly into a 96-well plate (one cell/well) containing lysis buffer with the following contents in each well: 0.25 μl IGEPAL CA-630 (5%; Sigma), 0.25 μl Rnasin Plus (Promega), 1 μl PBS, and 1 μl of a TaqMan pooled assay (0.2× of each assay). An unbiased sort was conducted when isolating neuroblasts, and astrocytes were isolated based on tdTomato fluorescence. In all cases, debris was excluded using a sort gate based on forward and side scatter amplitude, and doublet events were excluded using sort gates of forward and side scatter pulse width versus height.
The 96-well plates containing isolated single cells in lysis buffer were denatured for 2 min at 37°C followed by a 5 min incubation at 4°C. Reverse transcription (RT) was performed as previously described (Gibson et al., 2009) with slight modifications. RT reagents were added to each sample [0.1 μl dNTP's (25 mm; Thermo Fisher Scientific), 0.25 μl MMLV Reverse Transcriptase (Promega), 1 μl MMLV 5× buffer, 0.6 μl MgCl2, 0.05 μl Rnasin plus (Promega), and 0.5 μl water], and reactions were incubated at cycling temperatures (37°C 2 min, 42°C 1 min, 50°C 1 s for 40 cycles, then 85°C 5 min, and 4°C hold). Preamplification reactions consisted of 2 μl cDNA (from previous RT), 2.5 μl 0.2 TaqMan pooled assay, 5 μl Takara Premix TaqDNA Polymerase (Clontech), and 0.5 μl water. Reactions were amplified: 95°C 3 min, 55°C 2 min, 72°C 2 min; then 95°C 15 s, 60°C 2 min, 72°C 2 min for 16 cycles followed by a 4°C hold.
Neuroblasts were isolated from unbiased single-cell preparations following prescreening for Dcx expression via qPCR using preamplified cDNA and the TaqMan gene expression assay for Dcx (Mm00438400_m1) according to the manufacturer's instructions. Dcx+ cells and cells sorted for tdTomato fluorescence were subjected to further analysis using the Flex Six Gene Expression Integrated Fluidic Circuit (Fluidigm) on the Biomark HD platform according to the manufacturer's instructions, using the following TaqMan gene expression assays: Dcx (Mm00438400_m1), Gfap (Mm01253033_m1), Pecam 1 (Mm01242584_m1), Aif1 (Mm00479862_g1), Aldh1l1 (Mm03048957_m1), Aqp4 (Mm00802131_m1), Gapdh (Mm99999915_g1), EPHA4 (Mm00433056_m1), EFNA1 (Mm01212795_m1), EFNA2 (Mm00433011_m1), EFNA3 (Mm01212723_g1), EFNA4 (Mm00433013_m1), EFNA5 (Mm01237700_m1), EFNB1 (Mm00438666_m1), EFNB2 (Mm00438670_m1), and EFNB3 (Mm00433016_m1). Gene expression data were analyzed using the Fluidigm Real-Time PCR Analysis software and gplots package in R.
To investigate EphA4's functional role in the SVZ, we examined whole-mount preparations of the anterior lateral ventricle wall from adult EphA4+/+ wild-type (EphA4+/+) and null mutant EphA4−/− mice. Typically, Dcx+ neuroblasts within the SVZ formed tightly fasciculated chains oriented parallel to one another in arrayed bundles (Fig. 1A); however, in EphA4−/− mice, whole-mount preparations of the anterior lateral ventricle wall showed disorganized neuroblasts that lacked ordered, parallel alignment and bundling. This disorganization and lack of bundling along the SVZ resulted in diffuse Dcx labeling when the lateral ventricle wall was viewed en face (Fig. 1A). Diffuse Dcx labeling correlated with the absence of parallel alignment and fasciculation of chains resulting in easily observable immunolabeling; Dcx+ neuroblasts were not reduced, just not properly aligned.
Transmission electron microscopy (TEM) of the anterior forebrain of EphA4+/+ mice, used to examine the cytoarchitecture of the SVZ, showed the typical monolayer of ependymal cells along the length of the lateral ventricle wall and an underlying SVZ stem cell niche of neuroblast clusters surrounded by astrocytes and a scattering of transit amplifying progenitors (Fig. 1B). In contrast, TEM of the anterior forebrain of EphA4−/− mice revealed that neuroblasts were more dispersed and many breached SVZ boundaries and populated the corpus callosum (Fig. 1B). This atypical distribution of neuroblasts was corroborated by immunolabeling of coronal sections of the SVZ and neighboring structures, where Dcx+ neuroblasts were found in both the striatum (Fig. 1C) and corpus callosum (Fig. 1D). Together, these findings indicate that EphA4 functions in neuroblast bundling, orientation, and confinement to establish SVZ boundaries.
We then assessed the functional importance of EphA4–ephrin signaling in the RMS. Eph–ephrin interactions can result in three possible outcomes: classical “forward” signaling through the Eph receptor, “reverse” signaling through ephrin ligands, or a combination of both resulting in bidirectional signaling (Fig. 2A). We examined anterior forebrain coronal sections from adult EphA4−/− mice, in which both forward and reverse signaling was ablated, and KD-EphA4eGFP/eGFP mice (Kullander et al., 2001) in which only forward signaling was blocked. The RMS is typically characterized by dense, parallel bundles of migratory Dcx+ neuroblasts that transit through the anterior forebrain (EphA4+/+ mice; Fig. 2B,C). In EphA4−/− and KD-EphA4eGFP/eGFP mice, Dcx+ neuroblasts were more loosely organized, with many orientated perpendicular to the RMS when viewed in both cross and sagittal section (Fig. 2B,C). Area measurements taken along the length of the RMS confirmed that both EphA4−/− and KD-EphA4eGFP/eGFP mice had significantly larger average cross-sectional areas compared with EphA4+/+ mice (Fig. 2D). These findings confirm that EphA4 forward signaling is critical for proper alignment of neuroblast migration tangential to the lateral ventricles and for the overall compactness of the RMS.
Organization of the postnatal rodent RMS into the mature adult RMS occurs from birth until ~21 d of age (Peretto et al., 2005). During the early stages of this process (up to and including postnatal day 3), the RMS consists solely of loosely associated neuroblasts, which migrate en masse toward the OB (Nie et al., 2010). By P6, astrocytes begin to infiltrate the RMS core; and by P12, a complex meshwork of astrocytes exists within the RMS (Peretto et al., 2005). To determine when EphA4–ephrin interactions become necessary, we examined the organization of RMS neuroblasts in sagittal and coronal sections at P6 and P12 (Fig. 3). Representative sagittal images of the RMS (Fig. 3A) at P6 show that GFAP+ astrocytes are beginning to enter the RMS core in both EphA4+/+ and EphA4−/− mice. At P12, astrocytes in EphA4+/+ mice are oriented parallel to migrating neuroblasts, whereas astrocytes in EphA4−/− mice lack directionality and are disorganized and hypertrophic.
Dcx+ neuroblasts at P6 (before astrocyte invasion) show a similar loose organization in EphA4+/+ and EphA4−/− mice when viewed in sagittal (Fig. 3A) and coronal section (Fig. 3B), with no significant difference in relative RMS cross-sectional area (Fig. 3C). By P12 when the RMS begins to compact, the core of neuroblasts is more loosely arranged and exhibits a larger cross-sectional in the RMS of EphA4−/− mice compared with EphA4+/+ mice (Fig. 3A–C), with many neuroblasts exhibiting radial migration perpendicular to the normal direction of tangential chain migration (Fig. 3A). PECAM-1 immunolabeling of endothelial cells revealed that RMS blood vessels do not display any organizational differences in P6 or P12 between the EphA4+/+ and EphA4−/− mice (Fig. 3A). Together, our data suggest that, as the RMS develops from a loose array of tangentially migrating neuroblasts to a more condensed structure of fasciculated neuroblast chains surrounded by an astroglial meshwork, EphA4 signaling becomes critical for restriction and containment of neuroblasts and organization of astrocytes within a narrow RMS corridor, as found in the late postnatal and mature adult anterior forebrain.
Our finding that EphA4 is essential for astrocyte organization in the developing RMS prompted us to investigate the role of EphA4, and more specifically EphA4 forward signaling, in adult RMS astrocytes. The RMS astroglial meshwork consists of a dense grouping of astrocytes with elongated, bipolar morphology and an orientation parallel to chains of migrating neuroblasts (EphA4+/+; Fig. 4). Similar to the disorganization of RMS neuroblasts in adult EphA4−/− mice (Fig. 2), RMS astrocytes were significantly disorganized and the cell bodies appeared hypertrophic in the absence of EphA4 (Fig. 4). In many cases, astrocytic processes lacked parallel projection alignment and displayed a tangled appearance (Fig. 4, red arrowheads). The astrocyte disorganization was phenotypically mimicked in KD-EphA4eGFP/eGFP mice, indicating that EphA4 exerts its effects on astrocyte organization via forward signaling. These findings suggest that either neuroblast–astrocyte heterotypic EphA4 signaling or astrocyte–astrocyte homotypic EphA4 interactions would appear to be required for RMS astroglial meshwork organization.
Based on the above findings showing neuroblast and astrocyte disorganization in EphA4−/− mice, we next examined whether EphA4 absence affected the efficiency of neuroblast migration to the OB and in particular to the PGL. We administered the thymidine analog EdU and analyzed the number of EdU+ cells that remained in the SVZ or had migrated to the RMS or OB 21 d later. Twenty-one days allows sufficient time for newly formed neuroblasts to migrate from the SVZ along the RMS and to integrate within the existing OB circuitry (Luskin, 1993; Lois et al., 1996). We found no difference in the total number of EdU+ cells in the SVZ of EphA4+/+ and EphA4−/− mice 21 days after injection (Fig. 5A), indicating that labeled neuroblasts had left the SVZ and that only slow-cycling SVZ neural stem and progenitor cells remained (Codega et al., 2014; Zhang et al., 2015). However, evaluation of the total number of EdU+ cells that reached the OB in EphA4−/− mice revealed a significant decrease in number (~22%), with an even more dramatic decrease in the number of EdU+ cells integrating into the PGL (~49%) compared with EphA4+/+ mice (Fig. 5A,B). These findings indicate that EphA4 is required for the efficient transit of neuroblasts to their final site of differentiation in the OB.
To assess whether there were any differences in proliferation or cell cycle dynamics, we examined the percentage of BrdU+ cells out of the total population of Ki67+ cells in the SVZ of EphA4+/+ and EphA4−/− mice. We found no difference, indicating that cell cycle dynamics were similar in EphA4+/+ and EphA4−/− SVZ (Fig. 5C). Additionally, we found no significant differences in the number of caspase-3+ cells in the SVZ, RMS, and OB of EphA4+/+ and EphA4−/− mice, indicating no differences in cell death levels (Fig. 5D).
The above findings indicate that EphA4 is required for the organization of both neuroblasts and astrocytes in the SVZ and RMS. Although neuroblasts still reach the OB in the absence of EphA4, there is a marked decrease in the overall number of neuroblasts integrating into the OB and specifically the GL. To examine cell-specific expression patterns and to define potential cell–cell interactions between EphA4 and its ephrin-binding partners, we initially analyzed total RMS lysate (control mice) for transcripts. This analysis revealed robust expression of EPHA4, EFNA2, EFNA3, EFNA5, EFNB1, EFNB2, and EFNB3 genes (Fig. 6A). To delineate cell-specific EPHA4/EFN expression, we examined gene expression at the single-cell level. Isolation of single neuroblasts was validated by expression of Gapdh and Dcx in the absence of Pecam1 (platelet endothelial adhesion molecule 1; endothelial cells), Aif1 (allograft inflammatory factor 1; microglia), Aqp4 (aquaporin 4; astrocytes), and Aldh1l1 (aldehyde dehydrogenase 1 family, member L1; astrocytes) (Fig. 6B). EPHA4 was expressed in 39.8% of neuroblasts. The most frequently expressed EFN in neuroblasts was EFNB2 (75.7%) followed by EFNA2 (45.6%), EFNB3 (35%), EFNB1 (25.2%), EFNA5 (6.8%), and EFNA3 (2.9%). The most frequently coexpressed EFNs were EFNA2 and EFNB2 (35.9%), and EPHA4 was most often coexpressed with EFNB2 (35%) followed by EFNA2 (20.4%). In total, 89.3% of neuroblasts expressed at least one EFN, and 93.2% of neuroblasts expressed EPHA4, EFN(s), or a combination thereof. Total percentages of EPHA4/EFN expression and coexpression in neuroblasts are shown in Table 1. Euler diagrams, positioned below the heat map, show relative gene expression and coexpression of EFN genes with one another and with EPHA4 in neuroblasts (Fig. 6B).
In a similar experiment, we isolated single astrocytes and found that EPHA4 was expressed in 40% of all SVZ/RMS astrocytes (Fig. 6C). EFNs were present at a lower frequency in astrocytes (62%) than in neuroblasts (89.3%), and the most abundant ligand was EFNA2 (34.5%) followed by EFNB2 (21.8%), EFNA5 (20%), EFNB1 (14.54%), EFNB3 (3.6%), and EFNA3 (1.8%). The most commonly coexpressed EFNs were EFNA2 (14.5%) and EFNA5 (14.5%), and the most commonly coexpressed with EPHA4 was EFNA2. In total, 70.9% of astrocytes express EPHA4, EFN(s), or both. Total percentages of EPHA4/EFN expression and coexpression in astrocytes are shown in Table 1. Euler diagrams show relative gene expression and coexpression of EFN genes with one another and with EPHA4 in astrocytes (Fig. 6C).
The complex gene expression patterns of EphA4 and its ephrin binding partners were confirmed at the protein level in Figure 7. EphA4 was expressed by approximately half of RMS Dcx+ neuroblasts, and colocalized with multiple GFAP+ RMS astrocytes (yellow arrowheads). EphrinA2 was also expressed on approximately half of neuroblasts and, to a lesser degree, on astrocytes. In agreement with gene expression data, the ephrinB ligands were expressed on a majority of neuroblasts and a smaller fraction of astrocytes.
The long-distance migration of newly generated SVZ neuroblasts along the RMS, and through mature forebrain structures, to the granule and periglomerular layers of the OB necessitates directed and tightly constrained navigation. To achieve this, neuroblasts use a unique method of transit in which they travel along each other as tightly fasciculated chains within a surrounding astroglial meshwork (Belvindrah et al., 2007; Kaneko et al., 2010). We found that forward signaling through the tyrosine kinase receptor EphA4 is essential for the formation of compact, aligned neuroblast bundles within an SVZ/RMS astroglial meshwork, an organization that facilitates neuroblast transit through the anterior forebrain. Forward signaling through EphA4 is known to be involved in downstream alterations in actin filament dynamics leading to axonal growth cone collapse and repulsion, events integral to proper axonal guidance (Shamah et al., 2001; Sahin et al., 2005). EphA4–ephrin signaling can alter cytoskeletal arrangement directly or with the help of effector proteins, including ephexins, leading to the formation of cell–cell and cell-matrix focal adhesions and migration (Puschmann and Turnley, 2010). Interestingly, it was recently shown that forward signaling through Eph receptors is necessary and sufficient for the Rho kinase-dependent generation of an actin differential that drives adhesive/repulsive signaling events leading to directed movement of embryonic neuroepithelial cells (O'Neill et al., 2016). These studies linking forward signaling through Eph receptors to adhesion and migration, specifically through cytoskeletal modulation, directly support our findings that EphA4 is essential for the maintenance of RMS neuroblast fasciculation and coordinated, directed migration to final sites of differentiation in the OB.
Interestingly, we found that EphA4 is not necessary for RMS organization at P6, a time point when astrocytes are just beginning to infiltrate the RMS. However, by P12, when the astroglial meshwork in the RMS is forming, the absence of EphA4 resulted in disorganized RMS neuroblasts that did not form fasciculated bundles, as well as misaligned and hypertrophic RMS astrocytes. This resulted in a loosely organized, chaotic RMS (Fig. 8). Our findings suggest that the transition of astrocytes from the RMS perimeter to the RMS core is essential to facilitate neuroblast bundling and alignment to generate the compact, mature adult RMS; this is likely due to heterotypic astrocyte-neuroblast EphA4-mediated interactions.
Single-cell gene expression analysis in the RMS showed, for the first-time, intricate patterns of EphA4 and ligand ephrin expression on discrete subpopulations of both neuroblasts and astrocytes, with coexpression of multiple ligands, receptors, or both within the same cell (Fig. 6). More than 90% of RMS neuroblasts and 70% of RMS astrocytes expressed EPHA4, an EFN ligand, or both, pointing toward the importance of Eph–ephrin signaling in this region. EphA4, specifically, was expressed on 40% of both neuroblast and astrocyte populations, which supports our findings that not all neuroblasts and astrocytes are affected organizationally in EphA4−/− and KD-EphA4eGFP/eGFP mice, and a proportion of neuroblasts appear to be unaffected in their transit to the OB in EphA4−/− mice. Our single-cell studies indicate the heterogeneity of RMS neuroblasts and astrocytes, highlighting the existence of multiple subpopulations that may have discrete functional roles in maintaining neuroblast migration and RMS organization (Merkle et al., 2007; Lledo et al., 2008; García-Marqués et al., 2010). In addition, our results are consistent with studies showing EPH/EFN expression as heterogeneous and complex, with cell-specific, location-specific, and development-specific patterns and coexpression patterns involving multiple ligands and/or receptors (Goldshmit et al., 2006; Kania and Klein, 2016). Together, these findings provide a conceptual framework to understand Eph–ephrin interactions in the RMS, in which frequent receptor and ligand expression on neighboring cells or coexpression within the same cell could allow unique and simultaneous forward, reverse, or bidirectional signaling capacities as has been reported in motor axons, the retinotectal system, and in cancer pathogenesis (Marquardt et al., 2005; Dudanova and Klein, 2011; Falivelli et al., 2013). If receptors and ligands are expressed in the same membrane compartment, there is also the potential for cis-interaction on the cell surface, which has been linked to topographic mapping of retinal axons in the tectum (Yin et al., 2004; Carvalho et al., 2006). We suggest that simultaneous activation of multiple Eph–ephrin signaling cascades in subpopulations of cells creates combinatorial signals (neuroblast–neuroblast, astrocyte–astrocyte, neuroblast–astrocyte) that support both attractive and repulsive cues in a uniquely regulated migration pathway for newly generated neuroblasts to transit the mature anterior forebrain.
Eph–ephrin signaling works in concert with other molecular mechanisms that contribute to the dynamic interactions among neuroblasts and astrocytes and are critical for cell migration and organization within the RMS. RMS neuroblasts secrete Slit1, which associates with transmembrane Robo receptors on neighboring astrocytes and promotes repulsive morphological changes in astrocytes so as to create a pathway for neuroblast migration (Kaneko et al., 2010). The RMS of Slit1−/− mice is also characterized by disturbed neuroblast migration and disorganized RMS astrocytes, similar to what was found in EphA4−/− mice. Additionally, the formation of neuroblast chains requires signaling between laminin and β1 integrins that promotes cell–cell adhesion, and disruption of β1-laminin signaling results in dispersed RMS neuroblasts and disarrayed RMS astrocytes (Belvindrah et al., 2007). Neuroblast migration is also facilitated by homotypic interactions among cell adhesion proteins, such as PSA-NCAM (Hu et al., 1996; Wichterle et al., 1997). We propose that EphA4 forward signaling works collaboratively with PSA-NCAM and integrins in the control of homotypic neuroblast–neuroblast adhesive interactions that maintain neuroblasts in migratory chains, whereas contact-repulsive events via EphA4 result in the saltatory migration of neuroblasts along the RMS to the OB. Simultaneously, heterotypic EphA4–ephrin interactions between neuroblasts and the surrounding astroglial meshwork may help maintain neuroblasts within the RMS or alternatively help to keep astrocytes tightly associated with neuroblast chains. A lack of EphA4 results in fewer neuroblasts efficiently reaching the OB, specifically the periglomerular layer, following a 3 week chase period, suggesting that EphA4-mediated adhesive and/or repulsive homotypic and heterotypic signaling mechanisms that maintain RMS organization and coordinated neuroblast migration are essential for efficient population of interneurons within the OB. Our findings support a model in which EphA4 signaling acts in concert with integrin-laminin and PSA-NCAM as a critical cell–cell adhesion process to reinforce the compact nature of the RMS, and together with Slit-Robo signaling, supports neuroblast migration through the dense astroglial meshwork to the OB. Because EphA4 is not expressed on all of RMS neuroblasts and astrocytes, other Eph receptor signaling likely facilitates neuroblast fasciculation and directed migration in the subpopulations that are unaffected EphA4−/− mice.
The SVZ stem cell niche and the RMS migration pathway form an extraordinary system that facilitates large-scale movement of newborn neurons through mature tissue through the use of unique cytoarchitectural and cell–cell signaling modalities, including Eph–ephrin signaling. Our findings show, for the first time, the distinct importance of EphA4–ephrin signaling in establishing RMS organization and regulated control of neuroblast migration. A better understanding of neuroblast migration and the mechanisms that govern attractive and repulsive events leading to effective neuroblast movement to target destinations extends far beyond the rodent RMS. It is unclear how the infant human RMS, medial migratory stream, and recently characterized Arc migratory stream around the anterior lateral ventricles are organized or the molecular mechanisms that facilitate these distinct neuroblast migration pathways (Sanai et al., 2011; Paredes et al., 2016). Detailed information about the molecular players involved in dictating neuroblast organization and migration in the rodent RMS will inform further studies on neuroblast migration in the developing human brain and ultimately advance therapies associated with fetal brain development.
This work was supported by National Institutes of Health/National Institute of Neurological Disorders and Stroke R01NS50338. We thank Joshua Klein for illustrating our theoretical model; Patrick Malloy for assistance with creation of the Euler diagrams in Figure 5; and Stephen Daniels for assistance with electron microscopy as presented in Figure 1.
The authors declare no competing financial interests.