Glial cells are known to actively participate in neuronal development by shaping neuronal connections through axon pruning and by controlling dendritic spine morphology. These functions may in part be mediated by engulfment of neuronal structures and trans-endocytosis of neuronal material into glial cells. These processes are not well understood, and the molecular components that mediate these events have primarily been elusive. Here, we implicate the Eph/ephrin signaling system in trans-endocytosis events at the neuron-to-glia interface. Using time-lapse microscopy, we show that hippocampal neurons exogenously expressing EphB2 receptors release or pinch-off EphB2-containing vesicles at sites of neuron-to-glia contact. Cocultured glial cells endogenously express the corresponding ephrinB ligands and are able to trans-endocytose full-length EphB2 from neighboring cells. Although Eph/ephrin signaling often occurs in a bidirectional manner, the observed vesicle release from neurons to glia was only observed in a unidirectional manner, i.e., when the neurons expressed EphB2, but not ephrinBs. These findings suggest that Eph/ephrin signaling is involved in the glial cell-mediated fine sculpting of neuronal structures.
Traditionally, glial cells have been viewed to function as housekeeping cells and as scavengers that help to remove degenerating axons and neurons (Nedergaard et al., 2003). More recently, astrocytes are being appreciated as dynamic regulators of CNS synaptogenesis and synaptic transmission (Volterra and Meldolesi, 2005). In flies, glial cells engulf degenerating axons during developmentally regulated axon pruning (Awasaki and Ito, 2004). In vertebrates, during synapse elimination at the developing neuromuscular junction, withdrawing axons shed fragments into surrounding glial cells (Bishop et al., 2004). In the brain, astrocytes may also have an active role in developmentally regulated structural remodeling of neuronal processes. In the adult rodent hippocampus, double-membrane structures (spinules) were found through electron microscopy in axons and astrocytes derived from nearby dendritic spines, suggesting exchange of material among neurons and glial cells (Spacek and Harris, 2004). The molecular components that underlie these trans-endocytosis events, however, remained essentially unknown.
Eph receptor tyrosine kinases and their ephrin ligands have important roles in the establishment of neuronal and vascular networks during development and in the regulation of structural and physiological functions of excitatory synapses in the adult brain (Palmer and Klein, 2003; Davy and Soriano, 2005; Eichmann et al., 2005; Flanagan, 2006). Because of the nature of the membrane-anchored ephrins, signaling between Ephs and ephrins is bidirectional, consisting of forward signals into the Eph receptor-expressing cell and of reverse signals into the ephrin-expressing cell. Eph receptors are subdivided into A-type and B-type receptors, reflecting their binding affinities to either glycosylphosphatidylinositol-anchored ephrinAs (A1–A5) or transmembrane ephrinBs (B1–B3) (Kullander and Klein, 2002; Pasquale, 2005).
EphrinA3 is found on astrocytic processes near synaptic terminals, and the interaction between ephrinA3 and neuronal EphA4 stabilizes spine morphology in the hippocampus (Murai et al., 2003). In the context of cell-to-cell-mediated repulsion, ephrinB/EphB signaling leads to bidirectional endocytosis of the ephrinB/EphB protein complex (Marston et al., 2003; Zimmer et al., 2003). For these experiments, we used cells expressing EphB2 carrying in their cytoplasmic tail one variant of enhanced green fluorescent protein (GFP). On contact with a cell expressing ephrinB1, EphB2 receptors were rapidly clustered and internalized/endocytosed into both cells. Stimulation of ephrinB-expressing neuronal growth cones with cells expressing EphB2 caused rapid growth cone collapse and axon withdrawal. When EphB2 was C-terminally truncated and endocytosis into the stimulator cell was blocked, axon withdrawal was delayed. Thus, both cells actively remove the adhesive ephrin/Eph protein complex from the surface to allow repulsion to happen (Marston et al., 2003; Wilkinson, 2003; Zimmer et al., 2003).
Until now, we have shown that neurons can engulf ephrin/Eph complexes from other cells (Zimmer et al., 2003). It was not known whether neurons could also be the target of “attack” from other cells. Here we show that astrocytes trans-endocytose EphB2 receptors from cocultured hippocampal neurons. Using time-lapse microscopy, we demonstrate that hippocampal neurons release or pinch-off EphB2-containing vesicles from the neuron at sites of neuron-to-glia contact.
Materials and Methods
The vector, pEYFP-mem, encoding for the membrane-tagged form (mem) of enhanced yellow fluorescent protein (EYFP) was purchased from Clontech (Mountain View, CA). The GFP–TrkB.T1 plasmid was from E. Hoppe (Merck KGaA, Grafing, Germany) and Y. A. Barde (Biozentrum, University of Basel, Basel, Switzerland) and corresponds to the previously described hemagglutinin (HA)–trkBΔICD (Bibel et al., 1999), only the N-terminal HA tag was replaced by EGFP (Clontech). The EphB2–YFP construct (pJK12) was described previously (Zimmer et al., 2003). To generate YFP–ephrinB1 (pJK31), the EYFP coding region of pEYFP–N1 (Clontech) was cloned in frame into SalI/BsRGI sites of pJP136 (Bruckner et al., 1999). The generation of the YFP–ephrinB2 expression construct is described in supplemental Materials and Methods (available at www.jneurosci.org as supplemental material).
Hippocampal neurons were dissected from embryonic day 18.5 rat embryos and plated onto coated glass coverslips (Marienfeld, Bad Mergentheim, Germany) or live-cell glass-bottomed dishes with or without grid (high density, P35G-0-14-C; low density, P35G-2-14-CGRD; MatTek, Ashland, MA) [1 mg/ml poly-d-lysine (Sigma, St. Louis, MO) and 5 μg/ml laminin (Invitrogen, Carlsbad, CA)] at densities of either 17,000 cells/cm2 in Neurobasal-B27 medium (Invitrogen) or 5000 cells/cm2 in DMEM–B27 (Goetze et al., 2003). EphB2–YFP release was blocked by bath application of 1–4 μg/ml unclustered ephrinB1–Fc (constant fragment of immunoglobulin heavy chain) or EphB2–Fc (R & D Systems, Minneapolis, MN) during live-cell imaging compared with control Fc (Jackson ImmunoResearch, West Grove, PA). Cultures were imaged for 35 min, then Fc proteins were added, and imaging continued for another 65 min. Trans-endocytosed fluorescent vesicles in 100 × 100 pixels2 regions were counted 30 min before and 30 min after Fc protein addition (at the end of the imaging period). Time-lapse imaging is described in supplemental Materials and Methods (available at www.jneurosci.org as supplemental material).
Primary cultures of astrocytes.
For astrocytic cultures (de Hoop et al., 1994), cerebral hemispheres were collected, trypsinized, and physically dissociated. Approximately 60,000 cells/cm2 were plated onto untreated flasks (Nalge Nunc, Rochester, NY) in MEM–horse serum (HS). After cell attachment, the medium was replaced. Glial cultures were split at a 1:3 ratio. For feeder layer preparation of low-density cultures, astrocytes were seeded onto glass coverslips (outer diameter of 10 mm) with paraffin dots in MEM–HS and cultured until ∼80% confluence, and then medium was replaced with neuronal MEM–B27. After 3 d, one to two feeder coverslips were flipped over into the live-cell dishes containing 3–5 d in vitro (DIV) hippocampal cultures, and half of the medium was replaced with the glial conditioned medium.
Expression of endogenous ephrins and Eph receptors in astrocytes
Astrocytes of the CNS have been shown previously to express selected ephrins and to upregulate Eph/ephrin expression in response to injury (Conover et al., 2000; Bundesen et al., 2003; Murai et al., 2003; Goldshmit et al., 2004; Wang et al., 2005). Here we assessed the expression of endogenous ephrins and Ephs in mixed primary hippocampal cultures of neurons and astrocytes. Astrocytes were identified by their characteristic morphology and by immunostainings for glial fibrillary acidic protein (GFAP) (supplemental Fig. 1, available at www.jneurosci.org as supplemental material). We used soluble ephrin–Fc and Eph–Fc fusion proteins to stain for their respective binding partners on the surface of cultured cells as described previously (Harbott and Nobes, 2005 and references within). Application of preclustered ephrinB1–Fc, but not of control Fc, to the culture medium elicited many strongly fluorescent clusters on cells with astrocyte morphology, indicating the presence of endogenous EphBs (Fig. 1A,A′,B,B′). Similar responses to ephrinB1–Fc were observed on cocultured neurons (Fig. 1F,F′). A comparison with ephrinB2–Fc revealed similar signal intensities on neurons and astrocytes, suggesting that the endogenous EphB receptors had no binding preference to either ephrinB1 or ephrinB2 (Fig. 1I) (supplemental Fig. 2B,B′). A marked downregulation of EphB receptors [as reported for EphB2 (Henderson et al., 2001)] was not observed possibly because of the somewhat higher cell density of our cultures. Among the EphB–Fc fusion proteins, EphB1–Fc, EphB2–Fc, and EphB3–Fc elicited similar numbers of clusters on astrocytes, whereas neurons showed strong preference for EphB3–Fc (Fig. 1C,D,G–I) (supplemental Fig. 2, available at www.jneurosci.org as supplemental material). Regarding A-class ephrins, ephrinA2–Fc and ephrinA5–Fc induced similar numbers of clusters on hippocampal neurons, whereas astrocytes showed a stronger response to ephrinA5–Fc compared with ephrinA2–Fc (Fig. 1I) (supplemental Fig. 3, available at www.jneurosci.org as supplemental material). The responses to EphA–Fc proteins were generally weak on both astrocytes and neurons, indicating that the levels of endogenous ephrinAs were low in both cell types (Fig. 1I) (supplemental Fig. 3, available at www.jneurosci.org as supplemental material). Whereas the numbers of clusters varied fivefold between the stimuli (e.g., between EphB2–Fc and EphB3–Fc on neurons), the sizes of clusters did not vary as much (Fig. 1G–I).
Release of EphB2-containing vesicles from hippocampal neurons
To investigate whether EphB2 receptors were taken up by cells contacting neurons, we transfected an expression plasmid encoding an EphB2–YFP fusion protein (Zimmer et al., 2003) into 13–15 DIV hippocampal neurons. Expression levels of exogenous EphB2 were modestly higher compared with endogenous EphBs (supplemental Fig. 4, available at www.jneurosci.org as supplemental material). Two days later, transfected neurons in cultures containing glial cells often displayed filopodia-like protrusions at the tips of growth cones and along neurite shafts. These protrusions appeared to form contacts with other cells of the culture. During time-lapse imaging, we often observed small fluorescent clusters of EphB2–YFP at the tips of such protrusions. Occasionally, fluorescent clusters were released from the neuron or pinched off by another cell and transported away from the filopodia of the transfected neuron (Fig. 2A) (supplemental movie S1, available at www.jneurosci.org as supplemental material). Because the YFP tag was inserted into the cytoplasmic domain of EphB2, the presence of fluorescent clusters in neighboring cells indicated that the entire EphB2 molecule (including parts of the plasma membrane) had been transported.
During the release, which happened in a timescale of a few minutes, filopodia were highly mobile, possibly attributable to pulling forces executed by the engulfing cell. Small protrusions that did not undergo release remained rather motionless (supplemental movies S1, S5, available at www.jneurosci.org as supplemental material). Occasionally, the release of the EphB2–YFP cluster resulted in the shrinkage and complete disappearance of filopodia (Fig. 2B) (supplemental movies S2, S5, available at www.jneurosci.org as supplemental material). Analysis of a large number of time-lapse movies revealed that, in the majority of cases, EphB2–YFP clusters were being released from the neuron (Fig. 2C). In contrast, ectopic expression of YFP targeted to the inner leaflet of the plasma membrane (memYFP) rarely triggered such a response (1 per 18 movies) (Fig. 2C). Our interpretation of these events is that the EphB2-expressing neuron is contacted by a nearby cell (most likely a glial cell; see below) that expresses ephrinBs; this contact triggers bidirectional signaling events, resulting in the uptake of the EphB2-containing vesicles by the opposing cell possibly by trans-endocytosis (Zimmer et al., 2003).
Release of fluorescent clusters requires neuronal expression of EphB2
Because cultured astrocytes express both ephrinBs and EphBs, we next reversed the experiment and asked whether exogenous expression of ephrinBs in neurons would also trigger release events. We transfected neuronal cultures with expression constructs encoding fluorescently tagged ephrinB1 or ephrinB2 (YFP–ephrinB1/B2). This N-terminal modification does not interfere with receptor binding or trans-endocytosis (Bruckner et al., 1999; Zimmer et al., 2003). Surprisingly, in only 1 of 15 cases, ephrinBs were released toward opposing cells in the culture (Fig. 2C), indicating that the Eph/ephrin system was not equally active in both directions. This suggested that the EphB2-expressing cell was more efficient in releasing clusters than the ephrinB-expressing cell or that the pulling force exerted by the ephrinB-expressing cell was stronger than the EphB-expressing cells. We next asked whether the presence of an unrelated transmembrane protein, a GFP-tagged natural splice variant of the TrkB neurotrophin receptor (GFP–TrkB.T1) (Klein et al., 1990; Bibel et al., 1999), would also initiate this process. In only 1 of 13 cases did we observe the release of a GFP cluster (Fig. 2C).
If Eph/ephrin interactions were responsible for the observed release events, we reasoned that the addition of soluble unclustered ephrinB1–Fc or EphB2–Fc fusion proteins should block this process. We imaged neuronal cultures expressing EphB2–YFP and quantified the number of released vesicles before and after addition of Fc fusion proteins (Fig. 3A–C) (supplemental movies S3, S4, available at www.jneurosci.org as supplemental material). Whereas addition of control Fc did not significantly change the number of released vesicles, addition of ephrinB1–Fc or EphB2–Fc massively blocked the release of EphB2–YFP containing vesicles. We conclude that the release of EphB2-containing vesicles from a neuron to an opposing cell is directly mediated by ephrinB/EphB interaction and not an indirect consequence of changes in surface protein content attributable to the expression of EphB2–YFP in the neuron.
Glial cells elicit the release of EphB2 clusters from neurons
We suspected that glial cells were responsible for this process because highly motile glial cells were often observed in close vicinity to the neurons undergoing EphB2 release, and filopodia growing close to or on top of glial cells often displayed highly dynamic movements and release events. To unambiguously identify the cell type, we imaged mixed low-density cultures and then fixed and stained the cultures for the glial marker GFAP. In Figure 3D–G, the contact between a glial cell and an EphB2–YFP-positive neurite was imaged during time lapse. Several fluorescent clusters were pinched off and moved away from filopodium-like protrusions toward the glia cell (Fig. 3G) (supplemental movie S5, available at www.jneurosci.org as supplemental material). In this case, the glial cell remained well spread and displayed highly mobile lamellipodia. In another case, we observed a fluorescent cluster attached to a glial cell lamellipodium pulling away from the neuron (data not shown). When neurons were in contact with astrocytes, in four of four cases, the EphB2 detachment process was associated with the glial cell. If neurons had no glial contact, in one of four cases, the EphB2 release was mediated by another neuron (data not shown).
Astrocytes take up fluorescent EphB2 clusters
To investigate whether the released fluorescent clusters were internalized by an endocytic route, we modified a previously developed cell-to-cell stimulation assay (Zimmer et al., 2003) consisting of a sparse monolayer of “recipient” astrocytes on glass coverslips to which we added “stimulator” HeLa cells. If we used HeLa cells that exogenously expressed EphB2–YFP, we observed an accumulation of fluorescent clusters near the site of cell–cell contact (Fig. 4A,Ca). Comparison of surface localization of EphB2–YFP (Fig. 4Cb, red label) with its intracellular distribution (Fig. 4Cb, green label) suggested that these clusters were partially endocytosed by the astrocyte (Fig. 4Cb, green signals and green arrowheads mark endocytosed EphB2–YFP). Immunostaining for the early endocytosis marker 1 (EEA1) (Fig. 4Ce, blue label) revealed partial colocalization with the YFP signal (Fig. 4Ce, green label), suggesting the internalization of EphB2 via an endocytic pathway (Fig. 4Ce, overlapping blue and green label, white arrowheads). We also performed the converse experiments and cocultured recipient astrocytes with stimulator HeLa cells exogenously expressing YFP–ephrinB1. Under these conditions and different from neurons expressing YFP–ephrinB1, astrocytes readily trans-endocytosed YFP (Fig. 4D). These findings indicate that naive cultured astrocytes are able to trans-endocytose EphB2–YFP from opposing cells.
Here we have shown the release of membrane material from healthy (non-injured) neurons to cocultured glial cells. Experiments with transfected HeLa cells and primary astrocytes suggest that astrocytes trans-endocytose this membrane material. These events require the direct interaction between EphB2 on neurons and its binding partners on astrocytes (ephrinBs). Trans-endocytosis among hippocampal neurons and with glia has been observed previously in the intact hippocampus through serial section electron microscopy (Spacek and Harris, 2004). Double-membrane structures called spinules were enriched in perisynaptic glia interacting directly with synapses on thin dendritic spines as opposed to mushroom spines. Our findings are consistent with a model that says that neuronal material including membranes is released from thin protrusions at sites of neuron-to-glia contact. We suggest that Eph/ephrin signaling is critically involved in this process.
What could be the general functions of trans-endocytosis from neuron to glia? Besides microglia, astrocytes have been known to phagocytose degenerating axonal boutons (Bechmann and Nitsch, 1997; Nedergaard et al., 2003; Volterra and Meldolesi, 2005). In Drosophila, glial cells are recruited to healthy neurons for development-appropriate axonal pruning (Awasaki and Ito, 2004; Watts et al., 2004). The type of trans-endocytosis observed in our mixed cultures does not resemble traditional astrocytic phagocytosis of degenerating axons. Rather, these events may be regulating structural remodeling between neuronal protrusions and glia under physiological conditions. At the mammalian neuromuscular junction, axonal inputs are eliminated under physiological conditions by a mechanism that shows some resemblance to the events described here. As axons disappear, they shed membrane-bound “axosomes” that are formed by engulfment of axon tips by glial (Schwann) cells (Bishop et al., 2004). Glia cells are active partners at CNS synapses, dynamically regulating synaptogenesis and synaptic transmission (Nedergaard et al., 2003). The observed exchange of membrane material may be part of an active information exchange that enables glial cells to regulate specific aspects of synapse formation, maintenance, and elimination.
What could be the specific function of EphB/ephrinB signaling and trans-endocytosis in perisynaptic glia? Activation of Eph/ephrin signaling at cell-to-cell interaction sites typically elicits some type of repulsive behavior (although adhesive responses are possible, too). Neuronal progenitor cells in the segmented hindbrain sort themselves into alternating Eph- and ephrin-positive segments by avoiding contact with cells expressing their binding counterparts (Sela-Donenfeld and Wilkinson, 2005); migrating growth cones expressing Eph receptors turn away from cells that present the correct ephrin (Pasquale, 2005). Mechanistically, Eph/ephrin signaling may be involved at two separate levels: (1) it may activate the cellular machinery necessary for membrane traffic leading to (macro-) pinocytosis and the execution of the pinching-off process; (2) it may instruct the cell to retract its processes away from its neighboring cell, thereby providing some of the pulling forces for the pinching-off process. It will be important in the future to functionally separate membrane trafficking from cell retraction and to identify the molecular machinery driving these processes.
In summary, our observations add to the increasing richness in glial cell biology and support the idea that glial cells act instructively in neuronal developmental processes. Moreover, the implication of the Eph/ephrin signaling system provides an entry point into the molecular pathways that drive neuron-to-glia interactions.
This work was in part supported by the Max-Planck Society, the European Union (ENDOTRACK), and Deutsche Forschungsgemeinschaft Grant SPP1172. We thank E. Hoppe and Y. A. Barde for the GFP–TrkB.T1 expression plasmid, A. Kossel, M. Zimmer, C. Eßmann, and B. Götze for valuable comments and experimental help, and L. Dehmelt, M. Sperling, and L. Meyn for experimental help. We acknowledge T. Mäkinen and K. Deininger for critically reading this manuscript.
- Correspondence should be addressed to Rüdiger Klein, Max-Planck Institute of Neurobiology, Am Klopferspitz 18, D-82152 Martinsried, Germany.