Report

Activity-Dependent Transfer of Brain-Derived Neurotrophic Factor to Postsynaptic Neurons

See allHide authors and affiliations

Science  23 Mar 2001:
Vol. 291, Issue 5512, pp. 2419-2423
DOI: 10.1126/science.1057415

Abstract

Neurotrophins such as brain-derived neurotrophic factor (BDNF) are thought to be transferred from post- to presynaptic neurons and to be involved in the formation and plasticity of neural circuits. However, direct evidence for a transneuronal transfer of BDNF and its relation to neuronal activity remains elusive. We simultaneously injected complementary DNAs of green fluorescent protein (GFP)–tagged BDNF and red fluorescence protein into the nucleus of single neurons and visualized expression, localization, and transport of BDNF in living neurons. Fluorescent puncta representing BDNF moved in axons in the anterograde direction, though some moved retrogradely, and transferred to postsynaptic neurons in an activity-dependent manner.

Neuronal activity modifies the formation of neural circuits in developing cerebral cortex (1–3). Neurotrophins such as BDNF are attractive candidates for molecular signals that translate neuronal activity into such structural and functional changes in the cortex (4–7). Since the initial discovery of nerve growth factor, researchers have believed that neurotrophins are released or secreted from postsynaptic neurons or target cells (4, 8–10). However, recent studies suggested that BDNF may be supplied from presynaptic axons (11–18). Because these studies detected BDNF mainly with immunocytochemistry after fixation, dynamics of BDNF trafficking was not analyzed in living neurons, and the crucial question of whether anterograde, transneuronal transfer of BDNF—if it exists—is related to neuronal activity remains unanswered. In this study, we directly injected cDNAs of GFP-tagged BDNF and another fluorescence protein into the nucleus of neurons, which expressed both kinds of fluorescence within 24 hours.

A plasmid encoding BDNF tagged with GFP (19,20) was injected into the nucleus of cultured cortical neurons (21) through a micropipette under visual control (22). GFP-tagged BDNF resulting from this plasmid was confirmed to be biologically active and mimicked the releasing characteristics of untagged BDNF (20). Twenty to 30 percent of the injected neurons expressed fluorescent signal (23). Punctuated fluorescent signals were seen as clusters in neurites, whereas in the soma the fluorescent signal appeared dense and diffuse (Fig. 1, A and D). This pattern of signal distribution was quite similar to that of endogenous BDNF (Web fig. 1, 24), as reported previously (14,20, 25). We tested whether the fluorescent signal detected in the plasmid-injected neurons reflected the presence of BDNF. Neurons that expressed fluorescent signals were stained immunocytochemically with antibody to BDNF (Fig. 1B). The distribution of BDNF was almost identical to that of GFP fluorescent signal. This was confirmed by superimposition of the two pictures, although in the periphery of neurites there was a small number of red dots that probably reflected the presence of endogenous BDNF (Fig. 1C). Plasmid-injected neurons were functionally normal because they generated action currents in response to depolarization of the soma and spontaneous synaptic currents (26), as seen in noninjected neurons.

Figure 1

Distribution of BDNF-GFP in cortical neurons and its localization in axon branches. (A) BDNF-GFP expressed in a neuron 24 hours after an injection of plasmid cDNA encoding BDNF-GFP into the nucleus. Bar in (A) indicates unit of measure for (A) through (F), 10 μm. (B) Immunocytochemical image of the neuron shown in (A), stained with antibody to BDNF (anti-BDNF antibody). (C) Superimposed image of (A) and (B). (D) Distribution of BDNF-GFP expressed in another neuron. (E) Immunocytochemical image of the neuron shown in (D) stained with antibody to MAP2. (F) Superimposed images of (D) and (E). (G, H, and I) Magnified image of the boxed area in (D), each of which corresponds to (D), (E), and (F), respectively. Arrows in (G) and (I) show clusters of BDNF-GFP–positive puncta in the axon branch. Bar in (G) indicates unit of measure for (G) through (I), 10 μm.

In addition to the somatodendritic region, fluorescent signals of BDNF-GFP were also detected in neurites that seemed to be axon branches on the basis of their thin but constant caliber in the long distance (Fig. 1D and arrows in Fig. 1G). To confirm that these neurites were axons, the neurons were stained immunocytochemically with antibody against MAP2, which is known to exist in the somatodendritic region only (27). Neurites that contained the punctuated fluorescent signals (indicated by arrows in Fig. 1G) were spared from MAP2 staining (Fig. 1, E and H). This is seen in the superimposed pictures (Fig. 1, F and I). Such puncta arranged in line in MAP2-negative neurites were seen in all of the 24 neurons stained with antibody to MAP2. Four neurons were stained with antibody against tau, which is known to exist almost exclusively in axons (27). Fluorescent puncta were seen in tau-positive neurites in all of these neurons. These results suggest that BDNF exists not only in the somatodendritic region of neurons but also in the axon and its branches.

Fluorescent signals of BDNF moved quickly in axons of living neurons in the anterograde as well as retrograde directions. As shown in Fig. 2A, a long and thin neurite was spared from the staining with antibody to MAP2, indicating that it was an axon of a neuron whose soma was outside the frame of this picture. In this axon, a bright fluorescent spot was detected and was found to move in the anterograde direction toward the axon terminal (Fig. 2B; see also time-lapse Web movie 1, 24) (28). The movement velocity was 0.2 μm/s in this case. A few, other fluorescent puncta were seen to move in the retrograde direction and to pass by anterogradely moving puncta (see the second case in Web movie 1, 24). A similar movement of fluorescent puncta was observed also in dendrites of plasmid-injected neurons, but an exact quantification of movement was not feasible in dendrites because of complicated arbolizations. In Fig. 2C, we plotted against time the locations of four puncta that moved in the anterograde direction in three neurons. The mean velocity of the movement was 0.3 ± 0.1 (SEM) μm/s. This value is comparable to the reported value (mean ± SD, 0.69 ± 0.33 μm/s) of anterograde transport of a synaptic vesicle protein, synaptophysin, in axons of mouse dorsal root ganglion cells (29).

Figure 2

Movement of BDNF-GFP in an axon of a living neuron. (A) Superimposed image of an axon that expressed BDNF-GFP (green) with dendrites of another neuron stained with antibody to MAP2 (red). Parent soma of the axon is outside this image in the left direction. The image in the white square is magnified and shown in (B). Bar, 10 μm. (B) Anterograde movement of BDNF-GFP in the axon. Each image was taken at time point indicated at top left. Arrows indicate the same puncta representing BDNF-GFP. (C) Location of BDNF-GFP puncta expressed as the distance from the starting point was plotted against time. The data from the same puncta were connected with a line.

Then we asked whether BDNF in axons moves transsynaptically to postsynaptic neurons. We simultaneously injected two kinds of plasmid into the nucleus: a plasmid containing GFP-tagged BDNF and one containing DsRed fluorescent protein. An expression of these two kinds of protein in a single neuron was detected by changing the wavelength of the fluorescent excitation. DsRed was expressed together with BDNF-GFP in soma and neurites of plasmid-injected neurons (Fig. 3). As visualized with DsRed, an axon of a neuron whose soma was outside this frame terminated, so as to form many punctuated spots surrounding the soma of another, unstained neuron (Fig. 3A). A superimposed image of the DsRed with an image of the staining with antibody to MAP2 showed that the DsRed-positive puncta represented terminals of the axon surrounding the soma of the postsynaptic neuron (Fig. 3D). Such punctuated spots were confirmed to form presynaptic sites by immunocytochemical staining of a synaptic vesicle protein, synapsin I, in part of the experiments (30). In neurons illustrated in Fig. 3, the distribution of BDNF-GFP shown in Fig. 3B was almost identical to that of DsRed, except that the soma of the postsynaptic neuron had the fluorescent signal of BDNF-GFP only (arrow). The superimposed image in Fig. 3C shows that the soma of the postsynaptic neuron remained green, indicating the existence of BDNF-GFP in the postsynaptic neuron. Figure 3F confirms that the soma of the postsynaptic neuron contained the BDNF-GFP signal. These results suggest that BDNF was transferred from the presynaptic axon to the postsynaptic neuron because only the presynaptic neuron received plasmid injection. To quantify these findings, neurons were stained immnunocytochemically with antibody to GFP, and the fluorescence intensity was measured in somata adjacent to DsRed-positive axon terminals and in other somata that did not make contact with those terminals as a control (see insets of Fig. 4I) (31). Because the somata of control neurons had some background fluorescence, the intensity of this fluorescence was expressed as 100% and that in the former somata was normalized to this value. The mean fluorescence intensity of seven neuronal somata, which were contacted by DsRed-positive axon terminals, was 153 ± 18% of that of another seven control somata (Fig. 4I). Next, we addressed the question of whether the transfer of BDNF-GFP is mediated through BDNF receptors, TrkB (4). We stained neurons with antibody to TrkB and found that almost all the neurons including postsynaptic neurons contained TrkB. Then, we tested if a blocker of TrkB function, TrkB–immunoglobulin G (IgG) (32), can affect the transfer of BDNF-GFP. In all of the five neurons treated with TrkB-IgG, we did not see any visible signal of BDNF-GFP in postsynaptic neurons. The mean intensity of fluorescence in the somata contacted by DsRed-positive terminals was 92 ± 10% of control (Fig. 4J). This value was significantly [analysis of variance (ANOVA), P < 0.05] smaller than that without TrkB-IgG, suggesting that the transfer of BDNF-GFP was mediated by TrkB.

Figure 3

Transfer of BDNF-GFP to a postsynaptic neuron. (A) Horizontally running axon and its terminals containing DsRed. Its parent soma is outside this frame. Bar, 10 μm. (B) BDNF-GFP image of the same frame as in (A), visualized by antibody to GFP. (C) Superimposition of (A) and (B). Arrows in (B) and (C) indicate the soma of the postsynaptic neuron that is not seen in (A). (D) Superimposition of (A) and another image visualized by antibody to MAP2 (blue). (E) MAP2 image of the same frame as the others, stained by antibody to MAP2. (F) Superimposed image of (C) and (E).

Figure 4

Activity-dependent transfer of BDNF-GFP to postsynaptic neurons. (A) Representative records from a neuron before (left) and 5 min after (right) TTX treatment. (B) BDNF-GFP image of neurons treated with TTX for 48 hours and visualized by antibody to GFP. Bar, 10 μm. (C) Superimposition of (B) and DsRed image of the same frame. Arrow indicates the soma of the postsynaptic neuron, which was not visible in this image. (D) Superimposition of DsRed image (red) and MAP2 image (blue). Arrow indicates the soma of the postsynaptic neuron. (E) Representative records from a neuron before (left) and 5 min after (right) picrotoxin treatment. (F) BDNF-GFP image of neurons treated with picrotoxin for 48 hours and visualized by antibody to GFP. (G) Superimposition of (F) with DsRed image of the same frame. Arrow indicates the soma of the postsynaptic neuron. (H) Superimposition of DsRed image (red) and MAP2 image (blue). (I) Mean fluorescence intensity of BDNF-GFP in neuronal somata that did not contact DsRed-positive terminals (solid) and contacted such terminals (open). Vertical bars indicate SEM. Asterisk indicates statistical significance of the difference from the control at P < 0.02 (t test). (J, K, andL) Mean fluorescence intensity of BDNF-GFP in somata of neurons treated with TrkB-IgG, TTX, and picrotoxin, respectively. Double asterisks in (L) indicate statistical significance of the difference from the control at P < 0.01 (t test). Other conventions are the same as in (I).

Lastly, we asked whether the transfer of BDNF-GFP to postsynaptic neurons is modified by activity of neurons. Tetrodotoxin (TTX) was applied to neurons at a concentration of 1 μM, which abolished their spontaneous activities (Fig. 4A) (33). After TTX treatment, the soma of the neuron adjacent to the DsRed-positive terminals did not show any GFP signal (Fig. 4C). In seven neurons tested, the mean intensity of fluorescence in neuronal somata adjacent to DsRed-positive axon terminals was 98 ± 15% of control (Fig. 4K). This value was significantly (ANOVA, P < 0.05) smaller than that without any drug. These results indicate that the transfer of BDNF was dependent on neuronal activity. To further confirm this, we applied picrotoxin, a GABAA (γ-aminobutyric acid) receptor antagonist, at 50 μM, which increased the discharge rate of neurons under observation (Fig. 4E). After picrotoxin treatment the neuronal soma adjacent to DsRed-positive axon terminals had a strong green signal (Fig. 4G), indicating that BDNF-GFP was transferred markedly to the postsynaptic neuron. In seven neurons thus tested, the mean intensity of fluorescence in somata adjacent to DsRed-positive terminals was 197 ± 32% of that of the control (Fig. 4L).

To our knowledge, the direct injection of plasmid cDNA into the nucleus has not successfully been applied to neurons in the central nervous system, probably because of its technical difficulty. However, in the present study we found that it is applicable to cultured cortical neurons and has notable advantages over conventional methods such as viral vectors. The expression of BDNF-GFP and DsRed after direct intranuclear injection was very rapid. Usually the fluorescent signal emerged in the whole dendritic field within 24 hours after injection. Furthermore, only the neuron that has received the intranuclear injection of plasmid is expected to express the signal. Therefore, other neurons in which fluorescence signal of BDNF-GFP is detected must have received the tagged BDNF through transneuronal transfer. On the basis of such a simple but direct logic, we could demonstrate that BDNF was transferred to other neurons. This transfer occurred in the anterograde direction, because we detected BDNF-GFP in the soma of postsynaptic neurons that were surrounded by DsRed-positive axon terminals. There is a possibility that axon terminals of the postsynaptic neuron might have contacted the soma of the plasmid-injected neuron and BDNF-GFP might have been transported retrogradely to the soma of the postsynaptic neuron. This possibility seems unlikely, however, because we did not detect any fluorescent signal in axons originating from postsynaptic neurons, despite the fact that we could clearly see axons of presynaptic neurons that expressed DsRed and BDNF-GFP.

The present results are inconsistent with the traditional view that neurotrophins are released or secreted from postsynaptic neurons or target cells. Previous reports also suggested anterograde transport of BDNF, but they were based on indirect evidence in specific nervous systems: structures that lacked mRNA for BDNF production or changes in phenotype or receptor activity in target neurons after activation of presynaptic afferents in the cortico-striatal or noradrenergic projection system (11–18). In a study using radio-iodinated neurotrophins, only exogenously applied NT-3 was suggested to be transported anterogradely in the developing retino-tectal pathway of the chick (34). In none of these studies was the dynamics of BDNF observed in living neurons.

Neurotrophins such as BDNF are suggested to be involved in activity-dependent neural plasticity and, thus, implicitly supposed to be released in an activity-dependent manner (2–7). However, there has been no direct evidence for this, although BDNF assayed with the methods of immunoprecipitation or enzyme immunoassay in the perfusion medium of cultured neurons or tissue slices was reported to increase by high K+- or electrical stimulation–induced activation (35–37). The results presented here indicate that the transneuronal transfer of BDNF is dependent on neuronal activity. We have also demonstrated that the transfer of BDNF-GFP is not due to general transneuronal movement of protein molecules because DsRed, which was co-expressed with BDNF-GFP, was not detected in the postsynaptic neurons. This suggests that BDNF is released or secreted through presynaptic secretion mechanisms. The existence of BDNF associated with secretory vesicles was suggested previously (14), but these results did not show actual movement of BDNF in living neurons. Thus, the co-expression of two fluorescent proteins in this study has made it possible to directly observe the activity-dependent, transneuronal transfer of BDNF.

  • * To whom correspondence should be addressed. E-mail: ttsumoto{at}nphys.med.osaka-u.ac.jp

REFERENCES AND NOTES

View Abstract

Stay Connected to Science

Navigate This Article