Developing neurons form transient nanotubes facilitating electrical coupling and calcium signaling with distant astrocytes

Despite the well-documented cooperation between neurons and astrocytes little is known as to how these interactions are initiated. We show here by differential interference contrast microscopy that immature hippocampal neurons generated short protrusions towards astrocytes resulting in tunneling nanotube (TNT) formation with an average lifetime of 15 minutes. Fluorescence microscopy revealed that all TNTs between the two cell types contained microtubules but 35% of them were F-actin negative. Immunolabeling against connexin 43 showed that this gap junction marker localized at the contact site of TNTs with astrocytes. Using optical membrane-potential measurements combined with mechanical stimulation, we observed that ∼35% of immature neurons were electrically coupled with distant astrocytes via TNTs up to 5 hours after co-culture but not after 24 hours. Connexin 43 was expressed by most neurons at 5 hours of co-culture but was not detected in neurons after 24 hours. We show that TNTs mediated the propagation of both depolarization and transient calcium signals from distant astrocytes to neurons. Our findings suggest that within a limited maturation period developing neurons establish electrical coupling and exchange of calcium signals with astrocytes via TNTs, which correlates with a high neuronal expression level of connexin 43. This novel cell-cell communication pathway between cells of the central nervous system provides new concepts in our understanding of neuronal migration and differentiation.

Over the past decade intensive work focusing on the communication between neurons and astrocytes has led to a change in the long-standing view that astrocytes provide only structural and trophic support for neuronal cells. Astrocytes for example were shown to be crucial for the maintenance of the microenvironment of mature neurons by clearing neurotransmitter from within the synaptic cleft [1]. Furthermore, it was demonstrated that astrocytes modulate the formation and function of synapses and neuronal excitability by releasing neuroactive compounds such as apoE/cholesterol particles, thrombospondins, glutamate, ATP and D-serine [2], [3]. In murine postnatal hippocampal slice cultures, direct physical contact between astrocytes and pyramidal neurons was found to play a role in synapse formation [4]. Particularly, the employment of fluorescent membrane markers in conjunction with time-lapse imaging demonstrated the highly dynamic nature of the astrocyte-neuron interaction involving ultrafine astrocytic processes and dendritic protrusions of neurons [4], [5], [6]. However, little is known about the interaction of immature neurons and astrocytes during the early stage of embryogenesis, especially before the establishment of membrane excitability and synapse formation.

Nano-scaled membrane connections were recently observed during developmental processes of various organisms. In mouse embryo, ectoderm cells localized to juxtaposed folds, were connected by long thin cellular bridges during neural tube closure [7]. Similarly, thin membrane connections were also found between epiblast cells during gastrulation in zebrafish embryo [8]. In fact, these membrane tubes were described as the underlying structure of a previously unrecognized cell-to-cell communication pathway and referred to as tunneling nanotubes (TNTs) [9] or membrane nanotubes [10]. To date, TNT-like structures have been documented in numerous cell lines such as fibroblasts, epithelial cells and immune cells [11], [12], as well as for primary cells including neurons [13] and astrocytes [14], [15]. Furthermore, it was shown that TNT-like structures facilitate the transfer of diverse cellular components ranging from cytoplasmic molecules such as calcium ions to vesicles of endosomal origin and mitochondria [11], [12]. Also pathogens such as the human immunodeficiency virus (HIV) [16], [17] and prions [13], [18] were found to spread between cells through these tubes. Recently, depolarization signals were shown to diffuse through TNTs resulting in opening of low voltage-gated calcium channels in the receiving cells [19], [20]. Importantly, the electrical coupling depended on the presence of gap junctions interposed at the membrane interface between TNT and the connected cell. These TNTs were characterized by the absence of dye coupling and a small conductance compared to the normal gap junction connections. Most likely the unique biophysical character of the TNT determined these properties [19], [20].

Considering the widespread occurrence of TNT-like structures and their multi-faceted functions in cell-to-cell communication, it is conceivable that they are also implicated in the crosstalk between astrocytes and neurons during brain development. To address this issue, we co-cultured immature neurons with astrocytes and probed for TNT-dependent electrical coupling. Our data show that TNTs indeed form between both cell types and facilitate electrical coupling and intercellular calcium signaling, which is accompanied by a high expression level of neuronal connexin 43 (Cx43).

To investigate if immature neurons can form TNTs with astrocytes, freshly prepared hippocampal neurons from fetal E18 Wistar rats were co-cultured with a low-density cell population of astrocytes. This ensured that a sufficient number of neurons grew in close proximity to individual astrocytes. TNT formation was monitored by differential interference contrast microscopy (DIC) 2 hours after start of the co-culture. This precluded the formation of neurites and their potential interference with the analyses. From the movies that were acquired we observed that occasionally non-polarized neurons, identified by their small diameter and phase brightness, generated short protrusions towards astrocytes resulting in TNT-like connections between the two cell types (movie S1 and selected frames in Figure 1A). In all observed cases (n = 12), TNTs were formed by neurons that grew in close proximity (less than 10 µm) to astrocytes. Furthermore, from 7 analyzed neurons with a distance of less than 10 µm to astrocytes, 6 grew protrusions specifically towards the astrocytes resulting in TNT formation (7 movies within a 2 hour time frame of co-culturing). These TNT structures formed transiently and displayed a lifetime of 14.8±7.9 min (mean ± SD, n = 12 TNTs, mean length = 7.1 µm). Their breakage always occurred at the end facing the astrocyte followed by retraction of the protruded end towards the neuron, and finally they disappeared. Furthermore, our data clearly showed that TNTs are formed in one direction only, because astrocytes were never observed to generate membrane protrusions towards developing neurons.

To address which cytoskeletal elements were present inside TNTs connecting neurons and astrocytes, the co-cultures were stained with fluorescently labeled phalloidin and a monoclonal antibody against α-tubulin to detect F-actin and microtubules, respectively. Subsequent fluorescent microcopy showed that all analyzed TNTs were positive for microtubules (n = 20, Figure 1B), which is different from the filopodial precursors of TNTs. However, 35% of TNTs did not show F-actin staining (Figure 1B). Interestingly, the TNTs lacking F-actin were longer (29.7±7.42 µm, mean ± s.e.m, n = 7) than TNTs containing F-actin (11.2±1.7 µm, mean ± s.e.m, n = 13) as shown in Figure 1C. Further analysis using a specific antibody against Cx43 revealed that a significant portion of TNTs (n = 7/20) displayed a punctate labeling of this marker, which was localized at the contact site of the TNT and the astrocyte (Figure 1D). In addition, all Cx43 positive TNTs (n = 7) contained F-actin and were shorter than TNTs without Cx43 (Figure 1E).

Previously we showed that the presence of Cx43 at one end of the TNT was essential for the transfer of electrical signals between distant cells [19]. To investigate if neurons and astrocytes are electrically coupled through TNTs, both cell types were co-cultured for 1 hour and then analyzed in the presence of the membrane potential sensitive dye bis-(1,3-dibutylbarbituric acid) trimethine oxonol (DiBAC4(3)). As a characteristic feature DiBAC4(3) increases its fluorescence signal with increasing membrane potential and its sensitivity to detect TNT-dependent electrical coupling was confirmed by patch-clamp measurements as described previously [19]. When an astrocyte (Figure 2A, “a1”) with a TNT-connection to a neuron (Figure 2A, “n”) was depolarized by mechanical stimulation, the DiBAC4(3) fluorescence of both astrocyte and neuron increased continuously over a period of ∼60 seconds (Figure 2A, pseudo-colored images). At the same time, a control cell (Figure 2A, “a2”) close to the stimulated cell but lacking physical connection did not display an increase in fluorescence excluding the possibility that depolarization was propagated via diffusion of molecular signals between cells. This indicated a TNT-dependent electrical coupling of the stimulated astrocyte and the neuron. To conclusively prove that the measured TNT-connected cell (“n”) was indeed a neuron, the analyzed cells were further co-cultured for a total of 24 hours and then probed with an antibody for the neuron-specific marker tau-1. The staining pattern obtained proved that this cell is of neuronal origin and had differentiated into a polarized neuron (Figure 2B). A subsequent analysis of 20 TNT-connected neuron/astrocyte cell pairs revealed that 7 pairs (35%) displayed electrical coupling. In comparison, the electrical coupling frequency of abutted cell pairs within 5 hours of co-culturing was also tested.

The results showed that both astrocyte/astrocyte and astrocyte/neuron cell pairs frequently displayed a strong coupling (Figure 2C, “a1-a2” and “a1-n1/2″, respectively). Analysis of 18 abutted neuron/astrocyte pairs revealed a coupling frequency of 55%, which is higher than that of the TNT-based coupling. This higher coupling frequency of abutted versus TNT-connected cells is in agreement with data previously obtained from other cell types and may reflect a higher number of gap junction channels in the case of abutted cells [19].

To investigate whether or not the degree of electrical coupling between neurons and astrocytes changes with the progressive differentiation of neurons, we tested the electrical coupling after 24 hours of co-culture. At this time point hippocampal neurons have been classified as “stage 3” neurons [21]. Evidently, also at this stage TNT-like structures were detected between neurons and astrocytes (Figure 3A, arrows). However, all analyzed TNT-connected neuron/astrocyte pairs (n = 8) were not electrically coupled as illustrated in Figure 3A (pseudocolored images) and thus contrasted to our results obtained within 5 hours of co-culturing. To address if the absence of electrical coupling was only evident for TNT-specific connections, abutted neuron/astrocyte pairs were analyzed in parallel. Interestingly, also at these large contact sites, a decrease in electrical coupling was observed, down to 18% (n = 33, Figure 3B) after 24 hours as compared to 55% within 5 hours. Taken together, these data suggest that with progressive neuronal differentiation, the degree of electrical coupling decreases for both abutted and TNT-connected cell pairs (Figure 3C).

In agreement with our previous observation that TNT-mediated electrical coupling between NRK cells is Cx43-dependent [19], we speculated that the reduction in neuron-astrocyte coupling with prolonged culture time could be a decline in connexin expression. To test this hypothesis, we performed Cx43 immunofluorescence analyses after 5 and 24 hours of co-culturing. In parallel, neuronal differentiation was analyzed using the antibody against tau-1. Cx43 was strongly expressed by most neurons after 5 hours of co-culturing but was absent in nearly all cells after 24 hours (Figure 4A). In contrast, after 5 hours only few neurons were positive for tau-1, whereas after 24 hours the vast majority of them displayed a strong tau-1 staining in soma and dendrites (Figure 4A). A detailed analysis revealed that 88% of the neurons (n = 68) were Cx43 positive after 5 hours and of these only 32% of them were tau-1 positive (Figure 4B, black bars). After 24 hours of co-culture only 10% of the neurons (n = 47) remained positive for Cx43, whereas 98% displayed the neuron-specific tau-1 marker (Figure 4B, white bars). These results demonstrate a reversed trend in the expression profiles for both markers with increasing culture time, indicating a reduction in Cx43 expression during neuronal differentiation in the first 24 hours. For astrocytes, which were maintained for several weeks in culture, no significant change in the amount of surface-exposed Cx43 was observed during the first day (Figure 4C). Taken together, these data suggest that the decrease in electrical coupling between neurons and astrocytes is paralleled by the decline of the Cx43 level during neuronal differentiation.

Previously we reported that TNT-transmitted depolarization signals can activate low voltage-gated calcium channels resulting in transient calcium signals in the receiving cells [19]. To investigate whether depolarization of immature neurons is also accompanied by a rise in intracellular calcium ([Ca2+]i), we simultaneously measured changes in membrane potential with DiBAC4(3) and [Ca2+]i with the fluorescent calcium indicator X-rhod-1 in 5-hour co-cultures. To exclude that the increase of [Ca2+]i in TNT-connected neurons is the result of diffusible factors released from stimulated astrocytes, we used a cocktail of blockers (100 µM suramin to block purinergic receptors [22], 50 µM D-AP5 to block NMDA receptors [23] and 50 µM MCPG to block glutamate receptors [24]). Under these conditions, the mechanically stimulated astrocyte displayed depolarization and elevated [Ca2+]i (Figure 5A). Importantly, the neuron connected to this astrocyte via a TNT also exhibited both depolarization and elevated [Ca2+]i (n = 5, Figure 5A). A quantitative analysis revealed a simultaneous increase of the two signals in the neuron (Figure 5B). When astrocytes and neurons were not electrically coupled, no [Ca2+]i increase was observed in the TNT-connected neuron even though [Ca2+]i was strongly elevated in the astrocyte (n = 4, Figure 5C and D). Furthermore, we observed that membrane depolarization of a TNT-connected neuron did not always lead to changes in [Ca2+]i (n = 2, Figure S1). Taken together, our results suggest that immature neurons can receive both electrical and calcium signals from distant astrocytes through TNT connections.

Live-cell imaging demonstrated that immature neurons initiate the formation TNT-like connections with astrocytes by a filopodia-based mechanism. This appears to be a directed process since all thin membrane protrusions emanating from neurons were directed towards astrocytes. It implies that chemical guidance cues through so far unknown substances secreted by astrocytes could play an important role in the guidance of membrane protrusions. The formation of TNT-like connections observed here is distinct from neurite formation. Pioneering work on developing hippocampal neurons showed that neurites establish substrate-associated membrane protrusions after 24 hours of culturing [25]. These initially short protrusions grow longer with increasing culturing time and develop into a prominent growth cone. In contrast, TNT formation is a substrate-independent process and can be observed already after 1 hour of culturing, resulting in transient cell-cell connections lasting for only 10–15 minutes.

Although the nanotubes bridging neurons and astrocytes display typical TNT morphology, their cytoskeletal composition differs significantly from TNTs formed by other cell types. First, all TNTs were positive for microtubules, hitherto only found in a subclass of TNTs of very few cell types such as macrophages and some NK cells [26], [27]. Second, although F-actin is considered as a hallmark of TNTs, not all TNTs connecting neurons and astrocytes did contain F-actin. This raises the question as to how the F-actin-deficient TNTs are formed. The fast elongation/retraction dynamics of thin membrane protrusions extending from neurons during the course of TNT formation suggests that F-actin may be involved in the initial event of TNT formation, but disassembles locally at a later stage, while microtubules remain inside nanotubes. Such a view concurs with our finding that F-actin was only found in short but not long TNTs. Alternatively, considering the finding of Yang et al. that actin filaments is not directly involved in protruding filopodia [28], microtubules may be critical for TNT formation between neurons and astrocytes. This interpretation is supported by our finding that all analyzed TNTs contained microtubules.

Our data show that TNT-like structures bridging neurons and astrocytes frequently led to electrical coupling during the first few hours of co-culturing and disappeared after 24 hours, when neuronal development was already significantly advanced. In agreement with our results, dye transfer assays and electrophysiological measurements demonstrated that coupling of neurons from E18–E20 embryonic rats to the astrocytic monolayer was most frequent between day 2 and 3 and declined over the next 4 days [29]. This led to the concept that junctional communication between neuronal and astrocytic networks occurs predominantly during early embryonic development. The disappearance of heterocellular coupling with progressive cell differentiation is thought to be crucial for proper neuronal function since it prevents leakage of action potential from mature neurons to astrocytes. Subsequent work demonstrated that the underlying mechanism for the change in electrical coupling is the developmental regulation of connexin expression [30]. During the embryonic phase, a moderate expression of Cx43 was found in migratory neurons during neocortical brain development of E16–E19 rats and in conditionally immortalized mouse hippocampal multi-potent progenitor cells (MK31) [31], [32]. In addition, it was reported that Cx26 is also expressed in immature neurons [31], [33]. We therefore cannot exclude the possibility that Cx26 contributes to the documented electrical coupling. With progressive differentiation, hippocampal neurons were shown to shut down Cx43 expression and to form only gap junctions with themselves but no longer with astrocytes [34]. This is consistent with our data showing that expression of Cx43 is apparent for immature neurons and declines with progressive neuronal differentiation.

Thus, TNTs failed to facilitate heterocellular electrical coupling between neurons and astrocytes, most likely due to the absence of interposed gap junctions at later developmental stages.

We observed that TNTs were sufficient to transmit simultaneous depolarization and calcium signals to connected neurons. Importantly, we demonstrate that calcium signaling in neurons was absent when they were not electrically coupled with astrocytes despite the high level of [Ca2+]i in the stimulated astrocytes. On one hand, these data do not support a simple model of calcium diffusion through TNTs but rather suggests a local influx of calcium ions into neurons through activated low-voltage calcium channels. In support of this, a special L-type calcium channel was found in hippocampal neurons of E18 embryos, which could be activated at low depolarization thresholds [35], [36]. On the other hand, the lack of a calcium signal in some depolarized neurons may be due to cell-to-cell variation of the expression levels of low-voltage calcium channels in our hippocampal cultures. Alternatively, the necessary threshold potential to activate the respective channels may not be always be reached through the TNT-mediated depolarization.

Calcium signals have been considered as important regulators of proliferation, migration and differentiation of neurons [37]. In particular, migration of interneurons is controlled by calcium influx through activation of voltage-sensitive calcium channels regulated by the potassium/chloride exchanger [38]. Since neuronal progenitor cell migration is accomplished by functional cooperation between neuronal and astrocytic networks [39], [40], it is conceivable that neurons may receive instructive cues for their migration via TNT-dependent signaling from astrocytes. Besides calcium signaling, it has been reported that Cx43-specific gap junctions of newborn neurons function as adhesion sites along radial glia fibres during the migration of neurons to the cortical plate [31]. Because radial glial cells share many characterizers with astrocytes [41], [42], immature neurons may also form connexin-positive TNTs to probe for the anchorage sites on radial glia cells and thereby increase the efficiency of the migratory process. Given that TNTs were shown to facilitate the intercellular exchange of molecules and organelles [20], morphogens or other factors relevant in development may also transfer via TNTs between astrocytes and neurons. These considerations suggest that the novel long-range transient interaction between neurons and astrocytes described here could shed new light on the mechanism of the documented roles of astrocytes in neurogenesis and neuronal migration during morphogenesis.

Primary whole-brain astrocytes and hippocampal neurons from fetal E18 Wistar rats were prepared as described in accordance with the European Community Council Directive of 24 November 1986 (86/609/EEC) and approved by the Norwegian Committee for Animal Research [21], [43]. Astrocyte cultures were maintained in DMEM supplemented with 10% fetal bovine serum (Invitrogen, Carlsbad, CA, USA) for up to 4 weeks by passaging the cultures 1–2 times a week. For co-culture experiments, astrocytes (4.2×103 cells/cm2) were first plated out in 35 mm glass bottom dishes (MatTek Corp., Ashland, MA, USA) coated with 0.1 mg/ml poly-l-lysine (Sigma-Aldrich Co., St. Louis, MO, USA) and 10 µg/ml laminin (Sigma-Aldrich). One hour after plating the astrocytes were stained with 10 µM CellTracker™ Blue CMAC (Invitrogen) for 30–45 min and washed with medium. Freshly prepared hippocampal neurons (3.3×103 cells/cm2) were then plated directly to the stained astrocyte cultures in N-MEM supplemented with B27 neurobasal supplement (Invitrogen) and PSN antibiotic mix (Invitrogen) conditioned on astrocyte cultures for 3 days.

For time-lapse differential interference contrast (DIC) imaging, freshly prepared neurons were seeded on pre-plated astrocyte cultures. When the neurons were completely attached (1 hour after plating), the co-culture was imaged at 1 frame/2 min over a period of 2 hours using an Olympus IX70 microscope with a 40×/1.40 NA oil-immersion objective (Olympus Europa GmbH, Hamburg, Germany).

Co-cultures were fixed with 4% paraformaldehyde/4% sucrose in PBS 5 or 24 hours after plating. Immunofluorescence was performed according to standard procedures using the following primary antibodies: anti-tau-1 (Chemicon International, Temecula, CA, USA), anti-connexin-43 (Sigma-Aldrich) and anti-α-tubulin (Sigma-Aldrich). Actin was fluorescently labeled by Alexa Fluor 488™-coupled phalloidin (Invitrogen). Nuclei were stained by Hoechst 33342 (Invitrogen). Confocal imaging was performed on a Leica TCS SP5 confocal microscope (Leica Microsystems GmbH, Mannheim, Germany) equipped with a 40×/1.25 NA oil-immersion objective.

Mechanical stimulation experiments were performed 1–5 or 20–24 hours after the start of the co-culture. Co-cultures were pre-stained in DMEM supplemented with 2 µM of DiBAC4(3) (Sigma-Aldrich) for 30 min at 37°C and then replaced with pre-warmed fresh DMEM. Astrocytes and neurons were identified morphologically by DIC microscopy CellTracker Blue staining. Mechanical stimulations were applied to astrocytes in a chamber incubated at 37°C and 5% CO2, by microinjection using a FemtoJet/InjectMan NI 2 system (Eppendorf AG, Hamburg, Germany). Fluorescence images (16 bit) were acquired before and after mechanical stimulation with a 60×/1.40 NA oil-immersion objective, a Polychrome V monochromator (T.I.L.L. Photonics GmbH, Gräfelfing, Germany), and an Andor DU-885 camera controlled by IQ 7.0 software (Andor Technology, Belfast, Northern Ireland). CellTracker Blue and DiBAC4(3) were excited at 400 nm and 488 nm, respectively. Upon membrane depolarization cells display increased fluorescence of DiBAC4(3). The level of depolarization was expressed by pseudo-colored intensity images acquired with ImageJ software (http://rsbweb.nih.gov/ij/).

For simultaneous recordings of membrane potential and [Ca2+]i, 30 min co-cultures were loaded with 0.2 µM calcium indicator X-rhod-1 AM for 20 min, followed by a 30-min incubation in conditioned medium at 37°C supplemented with 4 µM DiBAC4(3), 100 µM suramin (Sigma-Aldrich), 50 µM D-AP5 (D(−)-2-Amino-5-phosphonopentanoic acid, Sigma-Aldrich) and 50 µM MCPG ((+)-α-Methyl-4-carboxyphenylglycine, Sigma-Aldrich). For two-wavelength imaging, the medium was replaced by colorless DMEM (Invitrogen) containing the three blockers mentioned above. The dye-loaded cells were excited at 488 nm and 560 nm and monitored under the same conditions as applied to single membrane potential measurements. Quantitative image analysis was performed with Image J. For quantitative analysis of DiBAC4(3) and X-rhod-1 images, the selected regions of interest (ROIs) and the mean fluorescence intensity of ROIs were calculated. The change of fluorescence intensity of cells (ΔF/F) was calculated as ΔF/F = (Fn − F0)/F0, where Fn is the mean fluorescence intensity at time frame n after mechanical stimulation, F0 is the original mean fluorescence intensity before mechanical stimulation.