Postsynaptic Induction of BDNF-Mediated Long-Term Potentiation

See allHide authors and affiliations

Science  01 Mar 2002:
Vol. 295, Issue 5560, pp. 1729-1734
DOI: 10.1126/science.1067766


Brain-derived neurotrophic factor (BDNF) and other neurotrophins are critically involved in long-term potentiation (LTP). Previous reports point to a presynaptic site of neurotrophin action. By imaging dentate granule cells in mouse hippocampal slices, we identified BDNF-evoked Ca2+ transients in dendrites and spines, but not at presynaptic sites. Pairing a weak burst of synaptic stimulation with a brief dendritic BDNF application caused an immediate and robust induction of LTP. LTP induction required activation of postsynaptic Ca2+ channels andN-methyl-d-aspartate receptors and was prevented by the blockage of postsynaptic Ca2+ transients. Thus, our results suggest that BDNF-mediated LTP is induced postsynaptically. Our finding that dendritic spines are the exclusive synaptic sites for rapid BDNF-evoked Ca2+ signaling supports this conclusion.

Neurotrophins promote neuronal survival and differentiation, but it has become increasingly clear that they also have essential roles in synaptic plasticity (1–3). Exogenous BDNF enhances transmission at the developing neuromuscular junction and at various central excitatory synapses (4–8). Furthermore, endogenous BDNF, via activation of the receptor tyrosine kinase TrkB, can regulate induction of hippocampal LTP (9–15). Although the induction of LTP is generally believed to be postsynaptic (16,17), the facilitating action of BDNF on LTP induction is assumed to involve presynaptic mechanisms (12,18, 19). This apparent contradiction resulted in the conclusion that “acute potentiation by neurotrophins cannot account for activity-induced LTP” in CA1 hippocampal pyramidal cells (2). Nevertheless, it has been suggested that BDNF regulates the induction of LTP by enhancing transmitter release during high-frequency synaptic stimulation (12, 20,21). Furthermore, it is thought that the enhanced synaptic transmission produced by application of exogenous BDNF is expressed as a sustained enhancement of transmitter release (5–7,19). However, because the synaptic potentiation by BDNF has been provoked by a protracted (20 to 60 min) treatment of all elements in the neural tissue (5,8), the specific cellular mechanisms of BDNF in inducing synaptic plasticity remain elusive. It has been difficult to distinguish between the role of the suppression of inhibitory synaptic transmission (22, 23) and the potential contribution of a general increase in neuronal excitability (24, 25). Moreover, the potentiating effect of a prolonged exogenous BDNF exposure may depend on the exposure time (6), and it is sometimes difficult to reproduce (3).

To overcome some of these difficulties and limitations, we applied brief and targeted pulses of BDNF to mature dentate granule cells in mouse hippocampal slices. We mapped the effect of focal BDNF pulses on different compartments of dentate granule cells (Fig. 1A) (26). Application of BDNF to a granule cell body evoked a rapid depolarization, causing action potential firing (27), which was associated with a transient increase in Ca2+ concentration (Fig. 1B). The rapidly evoked BDNF response in granule cells was also induced by neurotrophin-4/5 and was blocked by the tyrosine kinase antagonist K252a (26).

Figure 1

Ca2+ signaling in dentate granule cells evoked by focal application of BDNF. (A) Reconstructed image of a granule cell with the whole-cell pipette and the BDNF-ejection pipette shown schematically. (B) BDNF-evoked fluorescence signals (left column, ΔF/F) were recorded at four different locations in a granule cell, as indicated in (A) [in the axon (at sites 1 and 2), soma (at site 3), and distal dendrite (at site 4)]. Corresponding electrical recordings ofV m are shown on the right. Pulse durations were 10 ms (at site 4), 6 ms (at site 3), and 20 ms (at sites 1 and 2). Voltage scaling, 20 mV for sites 1, 2, and 4; 40 mV for site 3. (C) Pseudocolor image of a dendritic Ca2+ signal evoked by a BDNF pulse (10 ms). (D) BDNF-evoked Ca2+ transient (lower trace) from the active dendritic region shown in (C) and the associated electrical response (upper trace). (E) Block of the BDNF-evoked Ca2+transient (lower trace) in another cell dialyzed with D890 (1 mM). Data in (A and B), (C and D), and (E) are from different cells. TheV m was –70 mV; the arrowheads indicate the time of application of BDNF (50 ng/ml). Each trace is an average of three consecutive trials.

BDNF pulses failed to induce any response when applied to the axon (Fig. 1B). The lack of BDNF-evoked responses in the axon was found in all four cells tested. By contrast, pulse-like BDNF applications (6 to 20 ms) to the dendrites evoked subthreshold depolarizations (7.7 ± 1.6 mV, n = 8 cells), which were associated with large Ca2+ signals (Fig. 1B). These signals were spatially restricted to a small portion of the dendrite (less than 30 μm) near the site of application (Fig. 1, C and D). The dendritic Ca2+ signals had amplitudes of 119 ± 22% [relative increase in fluorescence (ΔF/F)], rise times of 82 ± 14 ms, and decay times of 536 ± 85 ms (n = 8 dendrites). When the cell was perfused with D890, a substance known to block voltage-gated calcium channels (28), the BDNF-evoked electrical response persisted (Fig. 1E), whereas the Ca2+ signal was abolished, even when long (20 ms) BDNF pulses producing an action potential were tested (n = 4 cells) (Fig. 1E). Furthermore, the BDNF-evoked Ca2+ transients were largely reduced (>70%) when the membrane potential (V m) was clamped to –80 mV (n = 3 cells).

We next examined whether BDNF also activates dendritic spines. Spines were functionally identified by their Ca2+ responsiveness after synaptic stimulation of the afferent perforant path. A single excitatory postsynaptic potential (EPSP) evoked a prominent spine Ca2+ response (ΔF/F = 39 ± 2%, n = 4 spines) (Fig. 2, A and B). A paired synaptic stimulus caused a larger spine Ca2+ transient (ΔF/F = 91 ± 6%, n= 4 spines) with a signal component in the parent dendrite (Fig. 2B, lower traces). Similarly, a brief pulse (3 to 10 ms) of BDNF, applied locally to a spiny dendrite (Fig. 2C), produced Ca2+signals in the dendrite and adjacent spines (Fig. 2D). These Ca2+ signals were associated with small somatic depolarizations (Fig. 2D, inset) of 3.2 ± 0.3 mV (n = 8 sites). An analysis of 19 spines at eight application sites (n = 6 cells) showed that BDNF consistently evoked larger Ca2+ signals in the spine than in the adjacent dendrite. On average, the amplitude of the spine signal was 1.8 ± 0.2 times that of the dendritic signal (Fig. 2E, inset).

Figure 2

BDNF-evoked and synaptically evoked Ca2+ signals in spines and dendrites. (A) Confocal image of a spiny dendrite and the schematically indicated positions of the pipette for synaptic stimulation. Inset, overview of the cell, reconstructed from confocal sections. The white box depicts the site of recording. (B) Ca2+ transients in the spine and in the adjacent dendrite from the regions marked in (A). The upper traces are the result of a single stimulus (arrow), and the lower ones resulted from a paired stimulation (20-ms interval) (pair of arrows). All traces are the averages of five consecutive trials. (C) Reconstructed image of a granule cell (left) and a spiny dendrite (right). The numbers correspond to three pairs of spino-dendritic regions of interest. The position of the BDNF-ejection pipette is as indicated. (D) BDNF-evoked (4-ms pulse) Ca2+ transients in three spines [red, regions from (C)] and in the adjacent dendritic regions [blue, regions from (C)]. The associated electrical signal is shown in the inset. (E) Plot of the spino-dendritic distribution of BDNF-evoked Ca2+ transients (n = 19 spino-dendritic pairs, six cells). The heavy line indicates the equal amplitude level in spine and dendrite. Inset, amplitudes in spines (red,n = 19 spines) normalized to the corresponding amplitudes obtained in the adjacent dendritic sites and the average dendritic reference amplitude (blue, n = 19 dendritic sites).

We next tested whether dendritic application of BDNF modifies glutamatergic synaptic transmission. Taking advantage of the rather strict laminar organization of the perforant path (29), we positioned the stimulation and BDNF-ejecting pipettes within the same beam of afferent fibers in the middle third of the molecular layer (Fig. 3A). This procedure allowed the ejected BDNF to reach the synapses that were activated by electrical stimulation. Application of BDNF alone did not modify the efficacy of synaptic transmission (Fig. 3B). However, a prominent synaptic potentiation was induced when the BDNF application was paired with an afferent burst of just five or six EPSPs (Fig. 3, C and D) (30). This weak synaptic stimulation was, by itself, insufficient to produce any potentiation (Fig. 3, C and D). The pairing-induced potentiation developed rapidly and peaked within 1 to 2 min to about 280% of the control (Fig. 3, C and D). The amplitude decayed toward a stable level of about 150% of the control within 30 to 40 min after the induction. We next established that BDNF mediates a strong facilitation of normal, tetanus-evoked LTP (31,32) rather than evoking a new form of LTP. Indeed, tetanus-evoked LTP completely occluded BDNF-mediated LTP (Fig. 4, A and B), and vice versa (Fig. 4, C and D). We also determined the time window during which the pairing must occur. There is no LTP induction if the electrical stimulation precedes the BDNF pulse by 200 ms (Fig. 4E). Instead, LTP is reliably induced during the standard “simultaneous” pairing protocol and, to a slightly decreased extent, if the burst stimulation follows the BDNF pulse within about 1 s (Fig. 4F). This poststimulation interval corresponds exactly to the period during which the postsynaptic cell is depolarized by BDNF. As soon as the V m reaches its resting value (Fig. 4H, bottom panel), the pairing fails to induce LTP (Fig. 4G).

Figure 3

LTP induced by pairing a brief BDNF application with a burst of EPSPs. (A) Reconstructed image of a granule cell and a schematic drawing of the experimental arrangement of the stimulation and the BDNF-ejection pipettes. The horizontal lines indicate the orientation of the medial perforant path fibers (MPP). (B) BDNF application alone has no effect on EPSCs. At the time indicated by the arrowhead, the amplifier was switched to the current-clamp mode and BDNF was ejected to the site of synaptic activation for 1 s. Inset (circled b), train of action potentials produced by the BDNF application. EPSCs (circled a and circled c, top row) were obtained at the time points indicated in the lower part of the graph and are each averages of eight consecutive responses. (C) Induction of LTP (at time point 0, open arrowhead) by pairing six EPSPs (50 Hz) with a BDNF pulse of 300 ms. Pairing was applied twice (at an interval of 20 s). A burst of six EPSPs alone [delivered at the time point marked with an arrow (boxed inset b)] did not change the amplitudes of the test EPSCs. The top row depicts EPSCs before (circled a and c) and after (circled e) induction of LTP. The traces are averages of eight consecutive responses. Inset b, burst of six consecutive EPSPs (stim alone) delivered at the time point marked with an arrow. Inset d, voltage response evoked during pairing a BDNF pulse with an EPSP burst (BDNF + stim). (D) Pooled data from eight experiments; the data points represent the average value of four consecutive responses.

Figure 4

(A to D) BDNF-induced LTP (LTPBDNF) and tetanic stimulation–evoked LTP (LTPtet) occlude each other. (A) LTPtetoccluded the induction of LTPBDNF. Five min after the induction of LTPtet, the stimulation intensity was reduced to obtain the EPSCs with amplitudes corresponding to the control values (stim down); at t = 30 min, the stimulation intensity was set to its initial value (stim up). Each data point reflects the amplitude of a single EPSC. (B) Mean change in EPSC amplitude (n = 4 experiments) measured at several relevant time points. The black bars correspond to the potentiation of EPSCs produced by tetanic stimulation just after the tetanus (3 min) and at t = 35 min. The white bars correspond to the relative change measured just after the attempt to induce LTPBDNF (13 min) and at t = 25 min; “30 min control” indicates the potentiation by the LTPtet protocol at 30 min in control experiments without any attempt to induce LTPBDNF (n = 6 experiments). LTPtet was induced by five bursts of 20 stimuli at 100 Hz delivered at intervals of 20 s, and the amplifier was switched to the current-clamp mode. (C and D) Similar experiments as in (A and B), but with a reversed order of LTP induction protocols. (E to H) The timing requirements for the induction of LTPBDNF. (E to G) Electrical stimulation precedes the BDNF pulse by 0.2 s (E), follows it by 1 s (F), and follows that by about 3 s (G). The standard pairing protocol (control, Δt = 0) was applied to test the viability of the synaptic inputs (E and G). (H) The extent of potentiation (normalized to the control value, measured 3 min after LTP induction) for various time intervals Δt, as indicated. The lower graph shows the corresponding mean value of theV m at the time just before the burst stimulation.

One mechanism by which BDNF might facilitate the induction of LTP is to enhance the synaptic response during high-frequency synaptic activation (12, 18, 19, 21) by increased neurotransmitter release. We thus compared the synaptic burst response with and without simultaneous BDNF application (Fig. 5A). BDNF application did not enhance the burst-evoked synaptic response. On the contrary, both the amplitude of the first excitatory postsynaptic current (EPSC) and the overall charge mediated by the synaptic burst were reduced to a similar extent during the BDNF pulse (Fig. 5B). Thus, BDNF-mediated LTP induction is not likely to involve a presynaptic enhancement of the synaptic response during conditioning. We then tested the presynaptic BDNF responsiveness more directly in conditions in which both the afferent axons and the postsynaptic cells were loaded with a membrane-permeable Ca2+ indicator dye (33). (Fig. 5C). Although application of BDNF to the cell body of a granule cell produced a large BDNF-induced Ca2+ transient (Fig. 5D), BDNF had no effect on afferent axons and terminals (Fig. 5E). Both the postsynaptic (Fig. 5F) and presynaptic sites (Fig. 5G) displayed the expected action potential firing behavior. In addition, the presence of BDNF had no effect on burst stimulation–evoked presynaptic Ca2+signals (Fig. 5E).

Figure 5

Postsynaptic induction of BDNF-mediated LTP. (A) Comparison of an EPSC burst (five events at 50 Hz) obtained in the control and during the application of BDNF (50 ng/ml) for 1 s (solid bar). The pipette solution contained 2 mM D890. The traces are averages of four consecutive responses. (B) Summary results (n = 8 cells) of the BDNF-mediated reduction of the amplitude of the first EPSC of each burst (EPSC) and the charge of the total response (charge). (C toG) Post- and presynaptic Ca2+ signals in response to different stimuli. (C) Experimental arrangement. For indicator-dye delivery, the tip of an Oregon BAPTA1-AM–containing pipette was placed in the dendritic region of the dentate gyrus. The boxed region within the granule cell layer (GC) is the site of postsynaptic recording (post), and perforant path fiber (PP) activity was determined in the region marked “pre.” (D and E) A BDNF pulse (20 ms) evoked a Ca2+transient in the cell body of a granule cell (D) but failed to evoke a Ca2+ response presynaptically (E, left). A fiber-mediated Ca2+ transient (E, middle) produced by afferent stimulation (burst, five pulses at 100 Hz) was not affected by the presence of BDNF (50 ng/ml) (E, right). (F) Ca2+ transient produced by antidromic single-shock stimulation in granule cell bodies in which BDNF-mediated responses were also recorded (D). (G) Ca2+transients recorded from axons and terminals in response to single-shock (single) and burst-afferent stimulation (burst). Ca2+ transients were not affected by the combination of 6-cyano-7-nitroquinoxaline-2,3-dione (10 μM) andd,l-2-amino-5-phosphonovaleric acid (50 μM). The data in (D to F) are from the same experiment. In (D to G), the traces are averages of three or four trials, and postsynaptic recordings are averages from three neighboring cell bodies. (H) Pooled data from six experiments showing block of the BDNF-mediated LTP by D890 (1 mM). (I) Pooled data from five experiments on the effect of CPP (10 μM) on BDNF-mediated LTP. In (H and I), the traces represent averages of eight consecutive EPSCs obtained from individual experiments. (J) Summary of the effect on synaptic transmission of BDNF application alone (BDNF), synaptic stimulation alone (stim), pairing of BDNF and synaptic stimulation (BDNF + stim), pairing after treating the slice with TrkB-IgG (5 ng/ml) for at least 1 hour (TrkB-IgG), pairing while the recording pipette contained D890 (D890), and pairing in the presence of 10 μM CPP (CPP). The amplitude values were obtained 30 min after the conditioning.

Our results suggest an exclusive postsynaptic site of Ca2+signaling in response to BDNF pulses and a direct role of this site in LTP induction. When the postsynaptic cell was dialyzed with the Ca2+ channel blocker D890, the conjunction of BDNF application and the brief afferent burst failed to induce any potentiation (Fig. 5, H and J). In the presence of theN-methyl-d-aspartate (NMDA) receptor antagonist 3-(2-carboxypiperazinyl) propyl-phosphate (CPP) (34), the conjunction resulted only in a transient potentiation and not in LTP (Fig. 5, I and J). Dialyzing the postsynaptic cell with the Ca2+ chelator 1,2-bis(2-aminophenoxy)ethane-N,N,N′,N′-tetraacetic acid (BAPTA) also prevented the induction of LTP (35). We verified the specificity of the BDNF action by showing that the BDNF scavenger molecule TrkB-IgG prevented the induction of BDNF-mediated LTP (12). A similar LTP-blocking action was also exerted by K252a, which suggests that LTP induction required phosphorylated TrkB receptors. Finally, we performed the present experiments in the presence of the γ-aminobutyric acid type A receptor antagonist bicuculline, excluding the possibility that the effects of BDNF were mediated through a modulation of synaptic inhibition (22, 23).

BDNF-evoked dendritic excitation may serve as an instructive postsynaptic signal for the induction of associative LTP. Such a role is also supported by recent experiments with a caged function–blocking antibody to BDNF (36), although the site of BDNF release is not clear yet. Although an autocrine action of postsynaptically released neurotrophin-4/5 has been proposed for the neuromuscular junction (2), there is evidence for an activity-dependent release of BDNF from presynaptic terminals in cortical neurons (37). A recent report (38) provides evidence for activity-dependent postsynaptic release of BDNF from cultured hippocampal neurons. However, because this BDNF release has not yet been directly correlated with LTP recordings, more work is required to determine whether, during LTP induction, BDNF is released from either pre- or postsynaptic compartments, or even from both compartments together.

Whatever the site of BDNF release, our results show that a brief BDNF pulse and a weak afferent burst robustly elicited LTP. Like tetanus-induced LTP (16, 17), the induction of BDNF-mediated LTP required postsynaptic Ca2+signaling (39) and NMDA receptor activation, as well as a close temporal association of pre- and postsynaptic activity. The mutual occlusion between BDNF-mediated and tetanus-evoked LTP suggests shared expression mechanisms. However, our results do not exclude a presynaptic expression of the BDNF-mediated LTP and are, therefore, not necessarily in contradiction with earlier evidence stressing the contribution of presynaptic mechanisms. An important conclusion of the present study is that spiny dendrites of mature dentate granule cells represent a highly responsive compartment for the rapid BDNF action. The block of this responsiveness prevented the induction of BDNF-mediated LTP. Our results reveal a critical mechanism underlying the surprisingly rapid, LTP-inducing action of BDNF and support an instructive role for BDNF in the induction of LTP in the mature brain.

  • * Present address: Department of Physiology and Pharmacology, Göteborg University, 40530 Göteborg, Sweden.

  • To whom correspondence should be addressed. E-mail: konnerth{at}


View Abstract

Navigate This Article