|Home | About | Journals | Submit | Contact Us | Français|
BDNF plays a critical role in the regulation of synaptic strength and is essential for long-term potentiation (LTP), a phenomenon that underlies learning and memory. However, whether BDNF acts in a diffuse manner, or is targeted to specific neuronal sub-compartments or synaptic sites to affect circuit function, remains unknown. Here, using photoactivation of BDNF or syt-IV (a regulator of exocytosis present on BDNF-containing vesicles) in transfected rat hippocampal neurons, we discovered that distinct subsets of BDNF vesicles are targeted to axons versus dendrites and are not shared between these compartments. Moreover, syt-IV and BDNF-harboring vesicles are recruited to both pre and post-synaptic sites in response to increased neuronal activity. Finally, using syt-IV knockout mouse neurons, we found that syt-IV is necessary for both pre- and post-synaptic scaling of synaptic strength in response to changes in network activity. These findings demonstrate that BDNF-containing vesicles can be targeted to specific sites in neurons, and suggest that syt-IV-regulated BDNF secretion is subject to spatial control to regulate synaptic function in a site-specific manner.
Neurotrophins were initially discovered via their positive effect on neuronal differentiation and survival (Lewin and Barde, 1996). More recently, neurotrophins such as BDNF have been found to modulate synaptic function and plasticity (Poo, 2001); BDNF is essential for LTP (Korte et al., 1995), and for normal learning and memory (Linnarsson et al., 1997, Liu et al., 2004). A critical unresolved issue concerns the synapse specificity of neurotrophin targeting and action (Poo, 2001, Lu et al., 2007). Recent findings raise the hypothesis that neurotrophins might act in a site-specific manner. For example, BDNF-coated beads promote local nerve growth (Tucker et al., 2001), and potentiate presynaptic strength only at nearby synapses (Zhang and Poo, 2002). In addition, BDNF secreted from BDNF over-expressing neurons increases dendritic branching only in neighboring neurons (Horch and Katz, 2002), and induced BDNF release from single neurons causes phosphorylation of CREB only in neighboring neurons (Kuczewski et al., 2008). These studies demonstrate local effects of BDNF secretion. However, whether BDNF vesicles themselves are targeted to specific sites for local secretion, remains unexplored.
BDNF-containing vesicles have recently been found to harbor the integral membrane protein syt-IV (Dean et al., 2009), a member of the syt family of molecules (Koh and Bellen, 2003, Chapman, 2008) and an immediate early gene (Vician et al., 1995, Ibata et al., 2000) that is important for learning and memory (Ferguson et al., 2000, Ferguson et al., 2004). This isoform binds SNARE proteins, but fails to bind membranes in response to calcium (Chapman et al., 1998, Wang and Chapman, 2010), thereby inhibiting BDNF release from vesicles harboring syt-IV (Dean et al., 2009). In co-culture experiments, presynaptic strength is increased only in terminals contacting syt-IV knockout neurons (which have increased BDNF release), and not at nearby wild-type synapses (Dean et al., 2009). This further suggests local effects of BDNF secretion. However, whether neurotrophins are targeted specifically to synaptic sites in response to activity, whereby they could modulate synaptic function via local secretion (Poo, 2001, Lu et al., 2007) remains unknown. Timelapse imaging studies of BDNF vesicles tagged with BDNF-GFP or with syt-IV-GFP, indicate that these vesicles are highly mobile (Adachi et al., 2005, Dean et al., 2009). Thus, neurons have the potential to dynamically target BDNF vesicles to specific sites along neuronal processes, to modify synaptic function locally.
Here we examined the directed movement and specificity of syt-IV and BDNF vesicle targeting and the effect of syt-IV-regulated BDNF release on pre and post-synaptic function, in response to alterations in neuronal activity. We discovered that, 1) distinct subsets of syt-IV and BDNF-harboring vesicles are differentially sorted to axons and dendrites and are not shared between compartments, 2) syt-IV and BDNF-harboring vesicles are recruited to both pre and post-synaptic sites in response to neuronal activity, and 3) syt-IV is necessary for both pre and post-synaptic scaling of synaptic strength in response to changes in activity. Thus, BDNF secretion regulated by syt-IV is subject to extensive spatial control to regulate synaptic function.
Rat hippocampi were isolated from E18-19 rats as described previously (Banker and Cowan, 1977). Syt-IV knockout mice were provided by H. Herschman (University of California, Los Angeles), and hippocampi were isolated from P1-3 mice of either sex from syt-IV knockout heterozygote matings, where syt-IV knockout cultures could be directly compared to wild-type littermate cultures. Conditional BDNF knockout mice and CamKII-Cre mice (provided by Alexei Morozov, NIH, Bethesda) were crossed to generate homozygous forebrain specific BDNF knockouts, and these mice were crossed to syt-IV knockout heterozygotes for electrophysiology experiments comparing BDNF knockout to BDNF/syt-IV double knockout littermate hippocampal cultures. Hippocampi were treated with trypsin for 20 min at 37° C, triturated to dissociate cells, plated at 25,000-50,000 cells/cm2 on poly-lysine coated coverslips (Carolina Biologicals), and cultured in Neurobasal supplemented with 2% B-27 and 2 mM Glutamax (Gibco/ Invitrogen).
For HEK293 cell/neuron co-cultures, HEK cells were first cultured alone in DMEM with 10% fetal bovine serum and were transfected using Lipofectamine 2000 (Invitrogen) according to the manufacturer's instructions. After 24 hours, these cells were mechanically dissociated in Neurobasal culture medium by gentle trituration, and added to 12 DIV neuronal cultures at a density of 1000-2000 cells/cm2.
All procedures involving animals were performed in accordance with the guidelines of the National Institutes of Health, as approved by the Animal Care and Use Committee of the University of Wisconsin-Madison.
Antibodies used were syt-IV, VGAT (Synaptic Systems), VGluT1, GluR2, synapsin, MAP2 (Chemicon), BDNF (Promega, Sigma), GFP, syt-17 (AbCam), synaptophysin (provided by R. Jahn, Göttingen, Germany), PSD-95 (Affinity Bioreagents) and bassoon (Stressgen). Cy2 and Cy3 secondary antibodies were from Jackson Immunologicals, and Alexa 647 secondary antibody was from Molecular Probes. Chemicals used were forskolin, glycine, TTX, APV (Sigma) and MG132 (Calbiochem). Mammalian expression constructs used were pEGFP-N1 (Clontech), mCherry (provided by R. Tsien, University of California, San Diego), and Nlg-GFP (provided by T. Dresbach, University of Heidelberg, Germany). Nrx-GFP was made by inserting PCR amplified GFP into a unique SalI site in VSV-tagged ß̃NRX-1 just after the LNS domain (constructed by H. Lee and provided by E. Isacoff, University of California, Berkeley). Nlg-IRES-GFP was made by inserting Neuroligin-1 (provided by P. Scheiffele, Columbia University, New York) into pIRES2-EGFP (Clontech). BDNF-PAGFP was constructed by replacing the GFP in BDNF-GFP previously described (Dean et al., 2009) with PAGFP (provided by J. Lippincott-Schwartz, NIH, Bethesda). PAGFP-sytIV was constructed by replacing the GFP in previously described GFP-sytIV (Dean et al., 2009) with PAGFP. GFP-syt17 was made by replacing the pHluorin in pHluorin-syt-1 (provided by T. Ryan, Weill Medical College of Cornell University, New York, NY) with GFP, and the syt-1 with syt-17 (provided by M. Craxton, MRC Laboratory, Cambridge, UK).
Neurons on 12 mm coverslips in 24-well plates were transfected with lipofectamine or calcium phosphate. For lipofectamine transfection (at 6-9 DIV), medium was removed, saved, and replaced with 500 μl fresh medium. 1 μl Lipofectamine 2000 (Invitrogen) in 50 μl Neurobasal medium and 0.5 μg DNA in 50 μl Neurobasal medium were incubated separately for 5 min, then mixed and incubated for 30 min at room temp. This mixture was added to the well of neurons, incubated for four hours at 37° C and 5% CO2, and then removed and replaced with half saved and half fresh medium.
Neurons were transfected by calcium phosphate at 3-4 DIV as described previously (Dresbach et al., 2003). Prior to transfection, medium was removed, saved, and replaced with 500 μl 37°C Optimem (Life Technologies) and incubated for 30-60 min. 105 μl transfection buffer (274 mM NaCl, 10 mM KCl, 1.4 mM Na2HPO4, 15 mM glucose, 42 mM Hepes, pH 7.06) was added dropwise to a solution containing 7 μg of DNA and 250 mM CaCl2, with gentle vortexing. This mixture was incubated for 20 min at room temp, 30 μl was added per well, and neurons incubated for 60-90 min. This medium was then removed, cells were washed 3× in 37°C Neurobasal medium, and saved medium added back to the transfected cells.
For photoactivation experiments, coverslips of neurons cotransfected with sytIV-PAGFP/mCherry or with BDNF-PAGFP/mCherry were transferred to a live imaging chamber (Warner Instruments) containing 119 mM NaCl, 2.5 mM KCl, 2 mM CaCl2, 2 mM MgCl2, 30 mM glucose, 25 mM HEPES, pH 7.4. Images were acquired on a Nikon A1Rsi 32 channel spectral imaging inverted confocal laser microscope equipped with a Perfect Focus System using 40× or 60× objectives and a hybrid galvano/ resonant scanner for simultaneous photoactivation and imaging (at the Nikon Imaging Center at the University of Heidelberg).
DIV 6-8 transfected cells were visualized by mCherry fluorescence, and chosen such that axons and dendrites could be clearly identified based on morphology (and verified by retrospective immunostaining for MAP2 to mark dendrites in all cases where axons and dendrites were difficult to distinguish). For each timelapse experiment, a baseline of GFP and mCherry fluorescence was collected for 10 seconds prior to photoactivation of the region of interest with 405 nm laser light. Regions to photoactivate in axons, dendrites, or cell bodies were chosen to include the largest region possible in the compartment of interest, without including any other regions. Fig. 1A shows the specificity of photoactivation within defined regions. Regions were photoactivated continuously with the 405 nm laser at 40% laser power and a pixel dwell time of 20 μsec during simultaneous acquisition of GFP and mCherry fluorescence, such that all vesicles transiting through the region during the timelapse were photoactivated. Laser intensity and pixel dwell time were selected to maximize photoactivation and minimize bleaching. The mCherry channel was imaged simultaneously and served as a control for focal drift.
Following image collection, cell bodies, axons and dendrites were selected as regions in NIS Elements software and fluorescence intensity was plotted versus time. Average fluorescence intensity at the end of the timelapse was calculated in 10 μm long × 3 μm wide regions of processes in axons and dendrites proximal to the cell body. For axon photoactivation experiments, fluorescence intensity was calculated in a region within the photo-activated area, in an area outside and just distal to the photoactivated area, in the cell body, and in proximal dendrites, at t=0 and 10 min. For dendrite photoactivation, fluorescence intensity was calculated in a region within the photoactivated area, in a different dendrite (i.e. the vesicles would have to travel from the photoactivated dendrite, into the cell body, and then into another dendrite), in the cell body, and in the proximal region of the axon, at t=0 and 10 min.
Cells were fixed with 4% paraformaldehyde, permeabilized and blocked in 10% goat serum and 0.1% Triton X-100, and immunostained with primary antibodies at room temperature for 4 h, or overnight at 4°C.
Images of immunostained cells were acquired on an Olympus FV1000 upright confocal microscope with a 60× 1.10 NA water immersion lens. 488 nm, 543 nm, and 635 lasers were used, with Olympus FV1000 automated dichroic/ emission settings for sequential scanning of Cy2, Cy3, and Alexa 647 secondary dyes. For quantitation, images were acquired with identical laser and gain settings and imported into Metamorph (Improvision) for analysis. For % colocalization of syt-IV at synapses (Fig. 1), thresholds were chosen for the VGluT channel (green), VGAT channel (blue), and syt-IV channel (red) such that all recognizable puncta were selected. Since the syt-IV antibody cross-reacts with other proteins in western blots of tissue isolated from brain, we used knockout mice as a control for specificity. Confocal gain in all experiments was set such that no syt-IV signal was visible in syt-IV knockout neurons. The average intensity of syt-IV signal within VGluT or VGAT puncta was determined using Metamorph for all conditions and expressed as % of control. For % colocalization, the fraction of syt-IV positive signal within VGluT or VGAT positive-puncta was calculated using Metamorph. Colocalization between VGluT and VGAT was 5% and a positive control double immunostain using an antibody against the luminal domain of syt I, and another antibody against the cytoplasmic domain of syt I showed 90% overlap. Cell bodies and regions 10 μm proximal to cell bodies were excluded from analysis because the syt-IV signal in the soma often overlapped with inhibitory terminals on cell bodies.
For quantitation of immunofluorescence signals on Nlg-IRES-GFP, Nlg-GFP, Nrx-GFP or GFP expressing HEK293 cells, transfected HEK cells were marked by thresholding the GFP signal. The average intensity of syt-IV signal within this thresholded region was then determined. Statistical significance for all quantitation was determined by a Student's t-test where *=p<0.05, **=p<0.01, and ***=p<0.001.
12-14 DIV hippocampal neurons treated with 200 μM glycine for 15 minutes or one hour, 50 μM forskolin for four hours, 50 μM MG132 for four hours, 1 μM TTX for 48 hours, or in control conditions growing in 10 cm dishes at a density of 50,000 cells/cm2, were washed twice in PBS, and 1 ml of hypotonic buffer (10 mM Tris-HCl pH 7.4) was added per plate for 5 min. Cells were harvested in 0.5 ml homogenization buffer (320 mM sucrose, 1 mM EDTA, 10 mM Tris-HCl pH 7.4) by scraping and passed through a 27-gauge needle ten times. A final concentration of 150 mM NaCl was added to the lysate. Cell lysates were centrifuged at 4000 rpm for 10 minutes to pellet nuclei and cellular debris. The supernatant was collected and the protein concentration determined by BCA assay. 12 μg of total protein per lane was resolved by SDS-PAGE and analyzed by immunoblotting for syt-IV (which runs as a 46 kDa band).
For imaging of syt-IV/synaptophysin or syt-I/ synaptophysin hippocampal neurons were fixed and immunostained with Atto565 and Dy-485XL secondary antibodies (Dianova, Hamburg, Germany). Pulsed excitation of Atto565 (Atto-Tec, Siegen, Germany) was achieved using a high-repetition rate laser diode source (PicoTA, Picoquant, Berlin, Germany) at a wavelength of λexc=532 nm, synchronized with the STED laser via a photodiode (Alphalas, Göttingen, Germany). Dy-485XL (Dyomics GmbH, Jena, Germany) was excited by the 470 nm line from a pulsed high-repetition rate laser diode source (Picoquant, Berlin, Germany). Fluorescence inhibition was accomplished at λSTED=647 nm using an actively mode locked Ar-Kr laser (Spectra Physics, Irvine, USA). The excitation and modulation beams were combined using acousto-optical tunable filters (AOTF) (Crystal Technology, Inc., Palo Alto, USA) and coupled into a microscope stand (DMI 4000B, Leica Microsystems GmbH, Mannheim, Germany) equipped with a three-axis piezo stage-scanner (PI, Karlsruhe, Germany). The AOTFs enabled blanking of the lasers and allowed the power of each laser beam to be controlled independently. They also provided a means for selecting counter-propagating fluorescence returning from the confocal microscope. The collected fluorescence passed through an additional band-pass filter (580/40 for Atto565 and Dy-485XL, AHF Analysentechnik, Tübingen, Germany) and was detected confocally with a photon counting module (SPCM-AQR-13-FC, PerkinElmer, Waltham, USA). The scanner fly backs were blanked via the line signal from the data acquisition software (Imspector, MPI Göttingen, Germany). A vortex phase plate (RPC Photonics, NY, USA) was used in the STED beam path to generate a donut shaped light distribution in the focal plane.
BDNF and Syt-IV were immunostained in fixed hippocampal neurons using Atto590 and Atto647N secondary antibodies (Dianova, Hamburg, Germany) and imaged with a custom-built two-color STED microscope as previously described (Buckers et al., 2011). The excitation wavelengths (λexc,Atto590=570 nm and λexc,Atto647N=650 nm) and the STED wavelengths (λSTED,Atto590=(720±10) nm and λSTED,Atto647N=(755±15) nm) were all provided by a single supercontiuum laser source (SC-450-PP-HE, Fianium Ltd., Southampton, UK). Fluorescence was detected with two separate confocal detection units covering the spectral ranges of λfluo,Atto590=(620±20) nm and in λflluo,Atto647N=(670±15) nm. The application of a pulse-interleaved acquisition scheme enabled the simultaneous recording of both imaging channels. Thus, no relative shift of the images, i.e. due to a drift of the sample, is expected. In addition, crosstalk, i.e. leakage of signal into the wrong imaging channel, was eliminated by linear unmixing.
Images were preprocessed using a single-step linear deconvolution (LD), i.e. Wiener filter, which was carried out with a theoretical effective point-spread-function of 40 nm FWHM.
Whole cell patch clamp recordings of mEPSCs were made from 15-21 DIV dissociated hippocampal cultures from syt-IV knockout and wild-type littermates (in control conditions and following treatment with 1 μM TTX, or 200 μM glycine/ 10 μM bicuculline for 48 hours), or BDNF knockout and syt-IV/BDNF double knockout littermate cultures. Neurons were voltage clamped at −70 mV in the presence of 1 μM TTX and currents recorded using an EPC-10/2 amplifier (HEKA Electronics, Germany) and Patchmaster software (HEKA). Cells were continuously perfused with extracellular solution consisting of, in mM: 140 NaCl, 5 KCl, 1.25 NaH2PO4, 26 NaHCO3, 2 CaCl2, 1 MgCl2, 10 glucose, 50 μM APV, bubbled with 95% O2 and 5% CO2 (pH 7.3, 300 mOsm). Intracellular solution consisted of, in mM: 130 K-gluconate, 10 EGTA, 10 HEPES, 5 Phosphocreatine, 2 Mg-ATP, 0.3 Na-GTP, (pH 7.3, 300 mOsm). For recording mEPSCs, 0.1 mM picrotoxin and 1 μM TTX were added to the bath solution. For recording mIPSCs, picrotoxin was replaced with 20 μM CNQX. Recordings were filtered at 2.9 kHz and digitized at 5 kHz. Events were detected with MiniAnalysis software (Synaptosoft) using a threshold of five times the RMS noise. Statistical significance was determined by a Student's t-test where * = p< 0.05, ** = p< 0.01, and *** = p< 0.001.
For FM dye experiments, wild-type and syt-IV knockout mouse neuron cultures in control conditions, or treated with 1 μM TTX for 48 hours, were loaded with 10 μM FM1-43 in depolarizing buffer (100 mM NaCl, 45 mM KCl, 2 mM CaCl2, 2 mM MgCl2, 5.5 mM glucose, 20 mM HEPES, ph 7.3) for 2 min to maximally load all vesicles. Coverslips were then transferred to a 5 mM KCl solution (140 mM NaCl, 5 mM KCl, 2 mM CaCl2, 2 mM MgCl2, 5.5 mM glucose, 20 mM Hepes, ph 7.3) for 5 min to remove excess dye. Coverslips were then mounted in live imaging chambers (Warner Instruments) in the 5 mM KCl solution. FM-dye labeled boutons were visualized with a 484 nm excitation/605 nm emission filter and images acquired on a Nikon TE300 inverted microscope with a Roper Scientific Photometric Cascade II EMCCD camera, and a Lambda DG-4 high speed wavelength switcher, interfaced with Metamorph software. Images were collected at 1 sec intervals with 200 msec exposure times. A baseline was collected for 10 images before addition of depolarizing buffer (45 mM KCl) to destain neurons. Following image collection, FM dye-labeled boutons were selected as regions in Metamorph and fluorescence intensity was plotted versus time. Destain traces were normalized by setting maximal load to one and complete destain (disappearance of bouton to background levels) to zero, for comparison of rates. Rate of destain (tau) was determined by a single exponential fit of normalized destain curves.
Detection of epileptiform response by voltage imaging was performed as previously described (Chang et al., 2007). Five-week old mice were euthanized with CO2. Brains were removed into ice-cold cutting solution consisting of, in mM: 124 NaCl, 3.2 KCl, 26 NaHCO3, 1.25 NaH2PO4, 1 CaCl2, 6 MgSO4,10 glucose, and bubbled with 95% O2/5% CO2. Horizontal 350 μm thick hippocampal slices were cut with a tissue slicer (HR2, Sigmann Elektronik, Germany). Slices were incubated in bubbled artificial cerebrospinal fluid (ACSF), identical to the cutting solution but containing 2.5 mM CaCl2 and 1.3 mM MgSO4, for 1 h, stained with bubbled ACSF containing 0.02 mg/ml of the voltage-sensitive absorbance dye RH482 (NK3630, Hayashibara Biochemical Laboratories, Okayama, Japan) for 1 hr, and then returned to ACSF. During recording, slices were continuously perfused with bubbled ACSF at 29-31°C. Slices were stimulated with 200 μs current pulses delivered by a stimulus isolator (Model A365, World Precision Instruments, Sarasota, FL). The photodiode array instrumentation and software have been previously described (Chang and Jackson, 2006). Optical signals (average of four consecutive trials) are presented as the change in transmitted light divided by the resting light intensity (ΔI/I), and are proportional to the membrane potential change within the physiological range of voltages.
Syt-IV and BDNF-harboring vesicles in axons versus dendrites exhibit distinct intra-vesicular pH, and different modes of BDNF release (Adachi et al., 2005, Dean et al., 2009, Matsuda et al., 2009). These differences might be dictated by the cellular environment in these two compartments, or might be due to specific sorting of distinct subtypes of syt-IV and BDNF-harboring vesicles to either axons or dendrites. To address this issue, we performed photoactivation experiments using hippocampal neurons. Neurons were co-transfected with either sytIV-PAGFP/mCherry or BDNF-PAGFP/mCherry, where the mCherry allows visualization of transfected cells prior to photoactivation. The PAGFP (photoactivatable GFP) tagged to syt-IV or BDNF is non-fluorescent until photoactivated by 405 nm light, which causes it to emit green fluorescence when excited at 488 nm. This approach allows photoactivation of specific subcellular compartments, i.e. axons, dendrites, or the soma, and subsequent tracking of the fate of vesicles, visualized as green fluorescence, present in each of these locations.
We first tested the spatial accuracy of photoactivation (Fig. 1A) and two different methods for identification of axons versus dendrites in photoactivation experiments (Fig. 1B, C): in the majority of cases, axons versus dendrites could clearly be identified using prior low magnification imaging where axons extended long distances compared to dendrites (Fig. 1B). In cases where the two types of processes were difficult to distinguish morphologically, retrospective immunostaining for MAP-2, to mark dendrites, was used (Fig. 1C). We also verified, using high magnification of isolated processes where single puncta were clearly visible, that photoactivated (BDNF-PAGFP) and non-photoactivated (BDNF-GFP) vesicles (Fig. 1D) had similar velocities (0.60 ± 0.21 μm/ sec and 0.63 ± 0.24 μm/ sec, respectively, n=21) and traveled in both anterograde and retrograde directions (52.0 ± 4% anterograde and 48 ± 4% retrograde) indicating that photoactivation did not reduce vesicle mobility.
We then photoactivated the cell bodies of neurons co-transfected with sytIV-PAGFP and mCherry, by continuous scanning of the cell body at 405 nm, while simultaneously imaging GFP fluorescence with 488 nm excitation. We found that sytIV-PAGFP vesicles trafficked rapidly out of the soma into both axons and dendrites, reaching distal processes within five minutes of activation (Fig. 2A, G). Cytoplasmic PAGFP, activated in the cell body, diffused to distal processes much more quickly, at a rate of approximately 35 ± 3 μm2/ sec, as previously reported (Bloodgood and Sabatini, 2005). Thus, the translocation of sytIV-PAGFP likely occurs via directed movement of vesicles to which it is attached, and not via diffusion of cytoplasmic sytIV-PAGFP. We then specifically photoactivated regions of axons in sytIV-PAGFP/mCherry transfected cells, by continuous scanning using pixel dwell times optimized for maximal activation and minimal bleaching, to photoactivate all vesicles transiting through the region during the ten minute period of image acquisition. Following the timelapse, the fluorescence intensity in the soma, dendrites, and axonal regions both within and outside of the photoactivated region, were quantified (Fig. 2G). We found that the vesicles within axons traveled from the photoactivated region both away from and toward the cell body (where a significant increase in fluorescence was observed). Surprisingly, despite trafficking back to the cell body within minutes, these vesicles did not enter dendrites to an appreciable extent over tens of minutes (Fig. 2B, G). We next performed the converse experiment, and selectively photoactivated vesicles present in dendrites and followed their trafficking post-activation. In this case, we analyzed the fluorescence within the photoactivated dendrites, within the cell body, in the axon, and within other dendrites, that were not photoactivated (i.e. if vesicles trafficked to other dendrites they would have to move from the photoactivated dendrite, back through the cell body, and into the second dendrite). We found that the vesicles in dendrites were in fact transported to the cell body and between different dendrites, but were excluded from axons (Fig. 2C, G).
We also performed similar photoactivation experiments on cells transfected with BDNF-PAGFP/mCherry. Like sytIV-PAGFP, photoactivated BDNF-PAGFP trafficked rapidly into axons and dendrites from the cell body (Fig. 2D, H). BDNF-PAGFP vesicles in photoactivated axons trafficked anterogradely and retrogradely back to the cell body, but could not enter dendrites to a significant extent (Fig. 2E, H). Likewise, the BDNF-PAGFP in photoactivated dendrites could traffic back to the cell body and into other dendrites, but failed to enter axons (Fig. 2F, H). We observed a constant increase in fluorescence in the cell body following photoactivation of axons or dendrites (Fig. 2I), indicating that degradation within the cell body is not a significant parameter to prevent movement of vesicles between compartments, on the timescale of our experiments. We note that more sytIV-PAGFP than BDNF-PAGFP was trafficked back to the cell body, following photoactivation of axons or dendrites. This could be a result of release of some of the BDNF-PAGFP from axons or dendrites prior to or during retrograde transport; we occasionally observed an increase in BDNF-PAGFP fluorescence in neighboring cells, indicating that the BDNF-PAGFP can be released and taken up by other neurons, as previously reported (Kohara et al., 2001) (Fig. 3).
BDNF, and the regulation of its release by syt-IV (Dean et al., 2009), is important for modulating synaptic function during LTP (Poo, 2001). We therefore next tested whether syt-IV and BDNF vesicle trafficking is affected following incubation with 200 μM glycine for 15 minutes to activate NMDA receptors (Musleh et al., 1997, Lu et al., 2001). Incubation with glycine did not alter the rate of export of syt-IV or BDNF vesicles out of the cell body (Fig. 4A) or the speed of the fluorescence wave (corresponding to multiple vesicles) leaving the cell body; speed (axons) = 1.6 ± 0.2 (syt-IV), 1.5 ± 0.3 (syt-IV + glycine), 1.5 ± 0.2 (BDNF), 1.7 ± 0.3 (BDNF + glycine); speed (dendrites) = 0.5 ± 0.1 (syt-IV), 0.9 ± 0.3 (syt-IV + glycine), 0.5 ± 0.2 (BDNF), 0.7 ± 0.2 (BDNF + glycine). However, following photoactivation of cell bodies after glycine treatment, we observed an accumulation of photoactivated BDNF-PAGFP at “hotspots” along axons and dendrites (Fig. 4B). We found, using retrospective immunostaining that these hotspots correspond to synaptic sites (Fig. 4C, arrowheads). Following photoactivation, BDNF-PAGFP increased first at proximal synaptic sites, and later at more distal sites (Fig. 4D), and accumulated at synapses in greater amounts in glycine treated neurons compared to controls; the intensity of BDNF-GFP overlapping with synaptophysin was 17.6 ± 1.7 (control) and 23.3 ± 2.6 (glycine-treated), p = 0.033.
We next tested the effects of increased neuronal activity on the localization of endogenous syt-IV and BDNF. Treatment of hippocampal neurons with 200 μM glycine for 15 minutes or one hour induced the recruitment of syt-IV to both excitatory and inhibitory synapses, with a slightly greater increase at inhibitory synapses (Fig. 5A-C). VGluT, VGAT, and synaptophysin puncta area was unchanged by glycine treatment and thus could be used to quantify syt-IV accumulation at synapses (VGluT: cntrl = 8.2 ± 0.8, 15′ gly = 8.6 ± 1.4, 1h gly = 8.7 ± 0.8, fsk = 8.2 ± 0.9, MG132 = 8.6 ± 1.1; VGAT: cntrl = 9.8 ± 0.7, 15′ gly = 10.2 ± 0.9, 1h gly = 9.6 ± 0.7, fsk = 10.1 ± 0.8, MG132 = 10.4 ± 0.9; synaptophysin: cntrl = 9.3 ± 0.5, 1h gly = 9.2 ± 0.5). The average intensity of the syt-IV signal at synapses increased following glycine treatment (Fig. 5F). However, the intensity did not increase uniformly at all synapses in response to activity. Instead, syt-IV was present at relatively high levels at a small subset of synapses in control conditions (arrow in Fig. 5A), and was recruited to additional synapses by activity. Therefore we also assessed the fraction of syt-IV-positive synapses. This fraction increased significantly following glycine treatment (Fig. 5G). Syt-IV may therefore fill a subset of synaptic sites under normal conditions and be recruited to fill additional sites during activity.
Treatment with glycine for 15 minutes or one hour resulted in the recruitment of syt-IV and BDNF-harboring vesicles to synapses, but did not increase overall syt-IV protein levels (Fig. 5F, inset) suggesting a rapid translocation of pre-existing syt-IV to synaptic sites. In contrast, four-hour treatment with forskolin increased accumulation of syt-IV at synapses (Fig. 5D-G), but also increased syt-IV protein levels (Ibata et al., 2000) (Fig. 3F, inset). Since syt-IV is rapidly degraded, with a half-life of two hours (Ferguson et al., 1999), we also tested the effects of proteasome inhibition on syt-IV. Treatment with the proteasome inhibitor MG132 resulted in a dramatic increase in syt-IV at synapses, and in syt-IV protein levels (Fig. 5E-G), suggesting that proteasome-mediated degradation may play a role in regulating the level of syt-IV in neurons.
BDNF increased at synapses concomitantly with syt-IV following induction of activity with glycine (Fig. 6A), in terms of both intensity (Fig. 6B) and % of puncta at synapses (Fig. 6C). Blockade of NMDA receptors with 10 μM APV 20 minutes prior to and during glycine treatment, abolished the recruitment of syt-IV and BDNF to synapses (Fig. 6D). Treatment of cultures with 45 mM KCl for 10 minutes also induced a recruitment of syt-IV to synapses (Fig. 6E). The effects of glycine were specific to syt-IV and BDNF-containing vesicles, since syt-17 vesicles, which traffic constantly and rapidly in neuronal processes (Fig. 6F), were not recruited to synapses by glycine treatment (Fig. 6G).
Whether syt-IV itself is localized to synaptic vesicles (Osborne et al., 1999, Ting et al., 2006), or not (Berton et al., 2000, Ibata et al., 2002, Dean et al., 2009) has been an ongoing subject of debate. Although our previous data suggest otherwise (Dean et al., 2009), it remains possible that the syt-IV that is recruited to synapses by activity is localized to synaptic vesicles and not to separate neurotrophin-containing vesicles. To distinguish between these two possibilities we used STED microscopy (Willig et al., 2006) with 40-50 nm resolution, which allows resolution of single vesicles (Fig. 7A). Using dual color STED, we found that following glycine treatment for 15 minutes, the syt-IV signal present at synapses was indeed on vesicles that are distinct from synaptophysin-labeled synaptic vesicles (Fig. 7B, C). This contrasts STED images of neurons co-immunostained with synaptophysin and syt-1 (Fig. 7D), where these two proteins largely colocalize on synaptic vesicles. BDNF co-localized with a subset of syt-IV positive puncta (Fig. 7E). The absence of BDNF on some syt-IV vesicles could reflect either an inability to detect low levels BDNF, or the absence of BDNF on these vesicles, perhaps due to prior release of BDNF cargo (Dean et al., 2009).
The site of BDNF release and action is an ongoing subject of debate (Kovalchuk et al., 2002, Zakharenko et al., 2003, Gartner et al., 2006), with recent evidence both that post-synaptic release affects presynaptic function (Jakawich et al., 2010), and that presynaptic release affects post-synaptic function (Amaral and Pozzo-Miller, 2007, Jia et al., 2010, Li et al., 2010). To determine whether syt-IV is recruited to synapses from the pre or post-synaptic side, i.e. from axonal or dendritic vesicle pools, we made use of a system in which hippocampal neurons are co-cultured with HEK cells expressing the presynaptic adhesion molecule neurexin (Nrx), or its post-synaptic binding partner, neuroligin (Nlg). Nlg-expressing HEK cells, when added to hippocampal cultures and incubated for 24 to 48 hours, induced the formation of presynaptic terminals, marked by the integral synaptic vesicle protein synaptophysin, in contacting axons (Fig. 8A) as shown previously (Scheiffele et al., 2000, Dean et al., 2003). These terminals were not apposed to postsynaptic compartments, as is evident from the lack of GluR2 signal. GFP-transfected HEK cells did not induce a significant syt-IV signal in contacting axons, while a ring of syt-IV was clearly present in presynaptic boutons contacting Nlg-expressing cells (Fig. 8B). Following induction of activity for one hour with 200 μM glycine, the syt-IV signal in these boutons increased in intensity. Since HEK cells have no NMDA receptors, glycine must exert its effects on the neurons in these co-cultures, driving the recruitment of syt-IV to presynaptic boutons that contact the neuroligin-expressing HEK cells.
We next tested the converse experiment. Neurexin is the presynaptic receptor of neuroligin, and can induce recruitment of post-synaptic proteins to sites of adhesion with contacting dendrites (Graf et al., 2004). When we co-cultured Nrx-expressing HEK cells with hippocampal neurons, we observed a ring of post-synaptic sites, marked by PSD-95, in contacting dendrites which lacked presynaptic terminals, marked by synapsin (Fig. 8C). A ring of syt-IV was observed in post-synaptic sites contacting Nrx-expressing cells (Fig. 8D). This syt-IV signal increased following induction of activity with glycine.
The intensity of bassoon, a presynaptic cytoskeletal active zone protein, contacting Nlg-expressing HEK cells was unchanged by glycine treatment (Fig. 9A), indicating that the presynaptic increase in syt-IV was not due to an increase in synapse formation potentially caused by glycine. Similarly, the intensity of PSD-95 contacting Nrx-expressing HEK cells was also unchanged by glycine treatment (Fig. 9B), indicating that the increase in syt-IV was due to recruitment to existing post-synaptic sites, and not to an increase in synapse number.
As expected from the syt-IV immunofluorescence, BDNF was also present at both pre and post-synaptic sites in a ring surrounding both Nlg-GFP and Nrx-GFP expressing HEK cells. This BDNF signal increased in response to activity induced by glycine (Fig. 9C).
The recruitment of syt-IV and BDNF-harboring vesicles to synaptic sites in response to neuronal activity (Fig. 5, ,6)6) prompts the hypothesis that the synaptic recruitment of syt-IV-harboring BDNF vesicles may act as a homeostatic mechanism to scale synaptic strength (via syt-IV-regulated BDNF release) in response to changes in network activity. If this were the case, syt-IV would not only be upregulated by long-term increases in network activity (such as forskolin treatment for 4 hours), but also down-regulated by decreases in network activity. Blockade of activity with TTX for 24-48 hours indeed resulted in a reduction in syt-IV protein, to barely detectable levels (Fig. 10A), in comparison to control, or to forskolin-treated cultures where syt-IV protein levels increased (Ibata et al., 2000) (Fig. 10A). A decrease in syt-IV expression (and subsequent increase in BDNF release) could act as a “homeostatic” mechanism to increase network excitation, by adjusting the strength of synapses, to compensate for the global activity blockade induced by TTX (Turrigiano et al., 1998, Burrone and Murthy, 2003, Turrigiano and Nelson, 2004).
If a homeostatic increase in excitation following TTX treatment is caused by down-regulation of syt-IV, one would expect that 1) knockout of syt-IV would mimic the effects of TTX-treatment and increase excitation, and 2) TTX would have no effect on syt-IV knockout neurons, where syt-IV is already down-regulated, i.e. absent. To test this we first measured synaptic function in syt-IV knockout neurons. Knockout of syt-IV increased the amplitude and frequency of mEPSCs, as reported previously (Dean et al., 2009), and decreased the frequency of mIPSCs, also as reported previously (Wardle and Poo, 2003) with no change in their amplitude (Fig. 10B-D). Thus, loss of syt-IV causes a net increase in excitation, in terms of mPSCs, by affecting both excitatory and inhibitory synapses. If these effects were due to syt-IV acting specifically through BDNF, one would expect no difference in minis in syt-IV/BDNF double knockouts compared to BDNF knockouts (because in both cases there would be no BDNF for syt-IV to act upon). We tested this, and found no change in mEPSC or mIPSC amplitude or frequency in syt-IV/BDNF double knockouts compared to BDNF knockouts (Fig. 10E), indicating that syt-IV acts via BDNF. Note that the mIPSC frequency in BDNF knockouts was lower than wild-type, most likely because BDNF is necessary for the formation and maturation of inhibitory synapses (Abidin et al., 2008, Hong et al., 2008).
We next tested the effects of TTX treatment on wild-type and syt-IV knockout littermate cultures. Activity blockade caused a homeostatic increase in the amplitude of mEPSCs in wild-type cultures, as expected (Turrigiano et al., 1998), but had no effect on syt-IV knockout littermate cultures (Fig. 10F). The amplitude of wild-type mIPSCs was homeostatically decreased by activity blockade, also as expected, but was unchanged in syt-IV knockouts (Fig. 10F). In the case of mIPSCs, syt-IV knockout prevented the reduction in amplitude caused by activity blockade. This observation is consistent with previous findings that excess BDNF (as in syt-IV knockouts (Dean et al., 2009)) prevents the effects of activity deprivation on GABAergic synapses (Rutherford et al., 1997, Swanwick et al., 2006). Thus, loss of syt-IV prevents post-synaptic scaling of synaptic strength.
We additionally tested the effects of chronic increases in activity by treatment with 200 μM glycine/ 10 μM bicuculline (Watt et al., 2000, Kennedy et al., 2010), but did not observe a homeostatic response in wild-type neurons with respect to scaling of mIPSC amplitude (wt = 26.6 ± 2.0 pA, wt + gly/bic = 22.2 ± 1.7 pA) or frequency (wt = 2.6 ± 2.0 Hz, wt + gly/bic = 1.0 ± 0.3 Hz), or mEPSC amplitude (wt = 13.7 ± 1.2 pA, wt + gly/bic = 12.5 ± 0.8 pA). We did observe a homeostatic decrease in mEPSC frequency in wild-type neurons following glycine/ bicuculline treatment, and found that syt-IV knockout neurons, by contrast, are unable to scale mEPSC frequency in response to increases in activity (wt = 4.1 ± 1.1, wt + gly/bic = 2.0 ± 0.5, syt-IV ko = 8.7 ± 2.5, syt-IV ko + gly/bic = 8.2 ± 2.3), consistent with the TTX results indicating an inability of syt-IV knockout neurons to scale their strength.
We also tested presynaptic homeostatic scaling in syt-IV knockouts by analyzing FM dye destaining kinetics following activity blockade. Synaptic vesicles were loaded with the lipophilic dye FM1-43 and the kinetics of dye release from presynaptic terminals examined during depolarization with 45 mM KCl. Chronic blockade of activity in wild-type neurons (prior to loading with FM1-43) resulted in a homeostatic increase in the rate of FM dye destaining (Wierenga et al., 2006) (Fig. 10G). Syt-IV knockout neurons had a fast FM dye destaining rate to begin with, similar to that of wild-type neurons treated with TTX, and the destaining rate in knockout neurons was unchanged following treatment with TTX. The homeostatic increase in FM dye destaining rate most likely reflects scaling of excitatory synapses, since ~70-80% of synapses in our hippocampal cultures are excitatory, and because the mIPSC frequency in wild-type neurons was homeostatically decreased in response to TTX (Fig. 10H); in syt-IV knockout neurons, mIPSC frequency was unchanged by TTX treatment.
In summary, activity blockade induced by TTX causes a down-regulation of syt-IV, which increases excitation; knockout of syt-IV mimics the homeostatic effect following TTX treatment, to increase excitation (via increased BDNF release). In addition, homeostatic responses of mEPSC amplitude, mIPSC amplitude and frequency, and FM dye destaining rate were absent in syt-IV knockout neurons. Thus, the homeostatic response induced by TTX to increase excitation appears to be caused by down-regulation of syt-IV to affect BDNF release.
Although EPSCs are unchanged in syt-IV knockout mice compared to wild-type, syt-IV knockouts exhibit an increase in LTP (Ferguson et al., 2004, Dean et al., 2009). This could be due to minis themselves altering the homeostatic threshold for LTP (Sutton et al., 2006) where syt-IV knockout circuits are more “excitable” than wild-type circuits. To test this idea, we examined electrically evoked epileptiform responses in syt-IV knockout and wild-type hippocampal slices using voltage imaging. These evoked discharges show spatiotemporal patterns of spread similar to those reported previously in rat hippocampal slices (Chang et al., 2007). We first determined the threshold to generate an epileptiform response by increasing the stimulus strength in 1μA increments beginning at zero. Syt-IV knockouts had a threshold of 5.4 ± 0.6 μA for epileptiform response, while the threshold for wild-type slices was twice that, 11.5 ± 0.5 μA (n=5). The magnitude of epileptiform response to threshold stimulation was also greater in syt-IV knockouts compared to wild-type (Fig. 11A, B; in wt, ΔI/I = 1.49 ± 0.04 × 10-3, in syt-IV ko, ΔI/I = 2.10 ± 0.04 × 10-3). In addition, the spatial spread of response was larger in syt-IV knockouts, with the response quickly spreading to the entorhinal cortex (Fig. 11C). In contrast, no significant signal was seen in the entorhinal cortex of wild-type slices. This increase in seizure susceptibility in syt-IV knockouts provides an indication that the absence of syt-IV increases the excitability of hippocampal circuits.
The findings reported here lead to three main conclusions: 1) distinct subsets of syt-IV and BDNF-harboring vesicles are differentially sorted to axons and dendrites and are not shared between these compartments. 2) Syt-IV and BDNF-harboring vesicles are recruited to both pre and post-synaptic sites in response to glycine-induced increases in neuronal activity, and 3) syt-IV is necessary for both pre and post-synaptic scaling of synaptic strength in response to changes in network activity. Thus, neurons have the ability to dynamically control BDNF release, via regulation by syt-IV, at specific sites along neuronal processes, to modify synapse and circuit function in response to alterations in network activity.
Apical versus basolateral trafficking has been well-described in epithelial cells (Simons and Fuller, 1985, Folsch et al., 1999) where these sorting events are essential for maintaining a selective barrier between the inside and outside of organisms (Cereijido et al., 2004). In neurons, distinct subtypes of syt-IV and BDNF-harboring vesicles may be trafficked to axons versus dendrites to insure that BDNF signaling occurs with specific spatial and temporal dynamics at inputs versus outputs of cells in a network.
Interestingly, BDNF vesicles exhibit distinct modes of depolarization-induced BDNF release assayed by BDNF-pHluorin, in axons versus dendrites. Dendritic release events are larger and faster than axonal events, and are consistent with full fusion and complete release of BDNF-pHluorin (Dean et al., 2009, Matsuda et al., 2009). Syt-IV reduces the magnitude of these events (Dean et al., 2009). In axons, exocytotic events correspond to partial release of BDNF-pHluorin from vesicles, and syt-IV inhibits secretion by increasing the proportion of “non-productive” fusion events corresponding to transient fusion pore opening without significant release of BDNF. In addition, the intraluminal pH of syt-IV and BDNF-harboring vesicles is different in axons and dendrites; dendritic vesicles have a higher pH than axonal vesicles (Dean et al., 2009).
The segregation of subtypes of syt-IV and BDNF-harboring vesicles to axons versus dendrites indicates that post-synaptic exocytosis of BDNF might have different effects than presynaptic BDNF release. Indeed, although both pre and post-synaptic BDNF release affect synaptic function, distinct pre and post-synaptic effects have been found in co-cultures of wild-type and syt-IV knockout neurons: post-synaptic release of BDNF, which occurs via full fusion (Dean et al., 2009, Matsuda et al., 2009), increases pre and post-synaptic strength, while presynaptic release, which occurs via partial release of BDNF from vesicles, increases only presynaptic strength (Dean et al., 2009). In addition, postsynaptic release of BDNF is needed to establish the correct ratio of glutamatergic and GABAergic neuron terminals (Singh et al., 2006, Kohara et al., 2007), while presynaptic BDNF, and not post-synaptic BDNF, has been found to be essential for LTP (Zakharenko et al., 2003).
Selective transport, via motor proteins that can distinguish axonal versus dendritic microtubules (Nakata and Hirokawa, 2007, Lewis et al., 2009), or exclusion of some syt-IV and BDNF motor/cargo complexes from axons by an axonal selectivity filter (Song et al., 2009), and retention of others, would insure that vesicles with distinct cohorts of proteins, that provide specific fusion properties, are targeted to axons versus dendrites. Our results provide direct evidence that these axonal and dendritic syt-IV and BDNF-harboring vesicles can be specifically targeted to synapses in response to increased network activity. A highly mobile reservoir of neurotrophin-containing vesicles, transiting throughout axons and dendrites, may allow for their rapid recruitment to synapses by activity. This phenomenon has been observed at the Drosophila NMJ, where a mobile pool of peptidergic vesicles are present in axons, and these vesicles are captured at synaptic boutons in response to increases in neuronal activity (Shakiryanova et al., 2006). These observations suggest that neurons can precisely control the site of neurotrophin release and further suggest that this is accomplished via increased capture at active sites, rather than increased synthesis in the soma or increased trafficking of vesicles to distal sites. This would ensure the rapid recruitment of neuropeptide and neurotrophin-containing vesicles specifically to active sites. Thus, the synaptic recruitment of syt-IV and BDNF-harboring vesicles, and the regulation of BDNF release by syt-IV, could potentially modulate synaptic strength on relatively fast time scales in a site-specific manner.
Our data demonstrate that syt-IV affects both pre and post-synaptic function via BDNF, and affects excitatory versus inhibitory synapse function differentially, consistent with the effects of exogenous BDNF (Schinder et al., 2000). In addition, we found that syt-IV is necessary for scaling of pre and post-synaptic strength in response to changes in activity. Synaptic scaling has recently been found to occur not only globally but also at specific subcellular sites (Yu and Goda, 2009). Syt-IV is a likely candidate to mediate this local scaling of synaptic strength, via regulation of BDNF release.
In summary, while synaptic vesicle exocytosis acts as a fast unidirectional channel of communication between neurons in a network, syt-IV regulated neurotrophin exocytosis represents a slower bidirectional channel of communication that can be invoked to modulate the first channel in a temporally and spatially specific manner in response to perturbations in activity. This modulation occurs at several levels including the dynamic regulation of syt-IV expression/ degradation, the incorporation of syt-IV into BDNF vesicles to affect their release, the trafficking of distinct subsets of vesicles with unique fusion properties to axons or dendrites, and the targeting of these vesicles to specific synaptic sites. Thus BDNF secretion mediated by syt-IV is subject to extensive temporal and spatial control to modulate the spatial pattern of synaptic efficacy in networks of neurons.
We thank Ulrike Engel and Christian Ackermann at the Nikon Imaging Center at the University of Heidelberg for assistance with photoactivation experiments, Thomas Dresbach for the use of his lab while in Heidelberg, Jon Audhya for the use of his NIS Elements software for analysis, Akhil Bhalla for assistance with western blot experiments, and Mark Dunning for assistance with molecular biology. This work was supported by NIH grant MH61876 to E.R.C., and Epilepsy Foundation, NIH NRSA, Sofja Kovalevskaja Alexander von Humboldt and ERC grants to C.D. E.R.C. is an investigator of the Howard Hughes Medical Institute.