Costorage of BDNF and neuropeptides within individual dense-core vesicles in central and peripheral neurons

Costorage of BDNF and Neuropeptides WithinIndividual Dense-Core Vesicles in Central andPeripheral Neurons

C. Salio,1 S. Averill,2 J.V. Priestley,2 A. Merighi1,3

1 Department of Veterinary Morphophysiology, University of Turin, Turin, Italy

2 Neuroscience Centre, Institute of Cell and Molecular Science, Queen Mary University of London,London, United Kingdom

3 Rita Levi-Montalcini Center for Brain Repair, University of Turin, Turin, Italy

Received 13 June 2006; accepted 23 September 2006

ABSTRACT: Some central and peripheral neurons

synthesize brain-derived neurotrophic factor (BDNF),

and, after anterograde transport, release it at synapses.

By immunocytochemistry, we examined, in rat and

mouse, the subcellular localization of BDNF and BDNF/

peptide coexistence, under normal conditions or after in-

trathecal infusion of nerve growth factor. In dorsal root

ganglion neurons and afferent terminals, and in the para-

brachial projection to amygdala, we show that BDNF is

costored in individual dense-core vesicles (DCVs) with the

neuropeptides calcitonin gene-related peptide (CGRP)

and substance P. At both locations, nerve endings costoring

all three peptides were fairly rare. Remarkably however,

costorage occurred in a stoichiometric ratio of 0.7

BDNF:1 CGRP:1 substance P, and DCVs contained 31

(spinal cord) �36 (amygdala) times the amount of BDNF

detected in agranular vesicles. This is the first direct

demonstration in peripheral and central neurons from

two different mammals, that a growth factor is selectively

packaged together with neuropeptide transmitters within

individual DCVs. It provides structural bases for differ-

ential release upon stimulation, and has important impli-

cations for understanding BDNF transmitter function.

' 2007 Wiley Periodicals, Inc. Develop Neurobiol 67: 326–338, 2007

Keywords: neuropeptide; amygdala; ultrastructure; dorsal

root ganglion; BDNF

INTRODUCTION

Although nerve growth factor (NGF) was originally

identified as a target-produced growth factor, which

is taken up by nerve terminals and then retrogradely

transported to the parent cell bodies, there is now

good evidence that several neurotrophic factors can

also be locally synthesized by neurons and/or endo-

cytosed at neuronal somatodendritic domains (trans-

cytosis) to be eventually targeted to terminals by an-

terograde axonal transport (von Bartheld et al., 2001;

von Bartheld, 2004). Neuronal synthesis and subse-

quent anterograde transport has been widely docu-

mented in the case of brain-derived neurotrophic fac-

tor (BDNF) in both the peripheral (PNS) and central

nervous system (CNS). In the latter, examples of neu-

rons synthesizing and anterogradely transporting

BDNF have been found in cerebral cortex, parabra-

chial nucleus, hippocampus, and locus coeruleus (Al-

tar et al., 1997; Conner et al., 1997; Kohara et al.,

2001). In the PNS, small diameter dorsal root gan-

glion (DRG) neurons synthesize and anterogradely

transport BDNF to their central terminals in the dor-

sal horn of the spinal cord (Michael et al., 1997).

Correspondence to: Prof. A. Merighi ([email protected]).

' 2007 Wiley Periodicals, Inc.Published online 12 January 2007 in Wiley InterScience (www.interscience.wiley.com).DOI 10.1002/dneu.20358

326

In cortical (Fawcett et al., 1997) and DRG (Michael

et al., 1997) neurons, ultrastructural localization or

subcellular fractionation studies indicated that the

neurotrophin is packaged within dense-core vesicles

(DCVs). Nevertheless it remains to be determined if

DCVs are the sole site of BDNF subcellular storage.

This issue is not trivial since: (i) it is unclear whetheror not newly synthesized neurotrophins may be pack-

aged in DCVs, since the latter appear to be very rare

in certain neurons that are believed to anterogradely

transport BDNF in vivo (Smith et al., 1997); (ii) thetransport vehicles for anterograde axonal transport of

internalized proteins have not been identified with

certainty, but it was suggested, although on circum-

stantial evidence, that DCVs may participate in this

function (von Bartheld et al., 2001); (iii) clear

vesicles in nerve terminals have also been indicated

to contribute to BDNF accumulation in central pro-

jections of DRG neurons (Luo et al., 2001).

In addition, many of the neurons that are capable

of anterogradely transporting BDNF also synthesize

neuropeptides (Block et al., 1989; Michael et al.,

1997) that are typically stored within DCVs (Merighi,

2002). Again, it remains to be determined with cer-

tainty whether or not the neurotrophin and the pepti-

des are stored and transported together within indi-

vidual DCVs, this being relevant to the mode of their

release at axonal terminals. Notably, a recent study

in DRGs has shown that BDNF and the neuropep-

tide substance P (SP) can be differentially released

according to the type of stimulus (Lever et al., 2001),

but it was not known whether and how this relates to

their vesicular storage.

Another issue deserving further attention, but still

incompletely clarified, is the functional significance

of BDNF anterograde transport. In locus coeruleus

and cortico-striatal projections, anterogradely trans-

ported BDNF was shown to regulate neuronal pheno-

type and survival (Altar et al., 1997; Fawcett et al.,

2000) in keeping with its well established role as a

trophic factor. However, BDNF is now being consid-

ered also an anterograde neuromodulator, with prop-

erties that are somewhat similar to that of a neuro-

transmitter. This dual action of BDNF was supported

by the observation in vitro of differential sorting

along the constitutive (when serving as a growth fac-

tor) or regulated (when acting as a neuromodulator)

secretory pathways (Mowla et al., 1999). Moreover,

studies on the intracellular processing pathways of

neurotrophins led the authors to conclude that

whereas internalized NGF is rapidly degraded in

lysosomes, other members of this growth factor fam-

ily, including BDNF, may escape intracellular degra-

dation and can be transported to axon terminals to be

released as neurotransmitters (von Bartheld et al.,

2001).

We have therefore utilized high resolution immu-

nocytochemistry to directly visualize the subcellular

site(s) of storage of BDNF and coexisting peptides in

peripheral neurons (DRG cells) and in the CNS (the

central nucleus of the amygdala, which receives a

BDNF projection from the parabrachial nucleus–Con-

ner et al., 1997). For additional validation, these stud-

ies were performed in two different species to obtain a

more comprehensive picture of the pattern of BDNF

subcellular localization in mammalian neurons.

METHODS

Animals

Eight adult male Wistar rats (light microscopy n ¼ 4, elec-

tron microscopy n ¼ 4) and 8 male CD1 mice (light micros-

copy n ¼ 4, electron microscopy n ¼ 4) were used in this

study. Rats were either controls (untreated) or had received

a 2 week intrathecal infusion of 12 �g NGF/day (Michael et

al., 1997). Under deep pentobarbital anesthesia (60 mg/100

g), animals were perfused with one of the following fixa-

tives: (i) 4% paraformaldehyde in 0.1 M phosphate buffer

(PB) for light microscopy; (ii) 4% paraformaldehyde

þ0.1% glutaraldehyde in 0.1 M PB for pre-embedding or

combined pre- and post-embedding electron microscopy;

(iii) 2% glutaraldehyde þ1% paraformaldehyde in 0.05 Msodium cacodylate buffer for post-embedding electron mi-

croscopy. All experimental procedures were approved by

the Bioethical Committees of the Universities of London

and Torino.

Immunofluorescence

Cryostat sections (12 �m) cut through the amygdala, lum-

bar spinal cord, and DRGs were stained using double or tri-

ple indirect immunofluorescence (IMF) or indirect tyramide

signal amplification (TSA). The rabbit anti-BDNF antibody

(1:2000–1:5000, TSA procedure) was combined with the

sheep anti-calcitonin gene-related peptide (CGRP-1:800) or

the rabbit anti-SP (1:2000) polyclonal antisera. Secondary

reagents included TRITC and AMCA-labeled anti-sheep

IgG for IMF, anti-rabbit IgG affinity-purified antisera

(1:400, Jackson ImmunoResearch, USA) and Vectastain

Elite peroxidase reagent (Vector, UK) followed by biotinyl

tyramide (TSA-indirect kit, Perkin Elmer, USA) and ExtrA-

vidin-FITC (1:500, Sigma, UK) for TSA labeling.

Electron Microscopy

Coronal sections from spinal cord and amygdala were cut

on a vibratome at a thickness of 50 �m (for pre-embedding

or combined pre-embedding and post-embedding) or

200 �m (for post-embedding).

BDNF and Neuropeptides in DCVs 327

Developmental Neurobiology. DOI 10.1002/dneu

Combined Pre-Embedding and Post-Embedding Immu-

nostaining. To simultaneously detect BDNF and CGRP

immunoreactivities in spinal cord (Michael et al., 1997)

sections were treated with 1% sodium borohydride in

PBS for 30 min and incubated with the rabbit anti-

BDNF antiserum (1:1000), followed by biotinylated goat

anti-rabbit IgGs (Vector) and Vectastain Elite peroxidase

reagent (Vector). Sections were developed with a solu-

tion containing 0.05% 3,30-diaminobenzidine, 0.04%

(NH4)2SO4.NiSO4, and 0.01% H2O2, contrasted in OsO4

(1%) and uranyl acetate (1%), dehydrated, and flat-em-

bedded in Durcupan (Fluka, Switzerland). After light mi-

croscopic examination of positively stained semithin

sections, ultrathin sections from areas of interest were

collected on nickel grids and stained for CGRP using the

post-embedding procedure. The simultaneous detection

of BDNF, trkB, and peptide immunoreactivities in spinal

cord (Salio et al., 2005) was carried out on free-floating

sections pre-incubated in PBS-NGS for 30 min and then

overnight with the chicken anti-full-length trkB (fl-trkB)

antibody (1:500). Sections were then incubated with an

anti-chicken IgY biotinylated secondary antibody (Vector)

followed by AlexaFluor 488-Fluoronanogold2-Streptavi-

din (1:100, Nanoprobes, USA). Counter-staining, dehy-

dration, and embedding was carried out as described

above.

Post-Embedding Immunostaining (Merighi and Polak,1993). Ultrathin sections were single, double, and triple

stained for BDNF, SP, or CGRP following a post-embed-

ding protocol. Briefly, sections were treated for 1 min with

a saturated aqueous solution of sodium metaperiodate

(Sigma), rinsed in 1% Triton X-100 in Tris buffered

saline (TBS) 0.5 M, and then incubated for 1 h in 10%

normal serum. Grids were then incubated overnight on

drops of primary antibodies (BDNFþSP, BDNFþCGRP or

BDNFþSPþCGRP) at optimal dilutions (chicken anti-

BDNF 1:20, rabbit anti-CGRP 1:500, rat anti-SP 1:20). Af-

ter rinsing in TBS, they were incubated in a mixture of the

appropriate gold conjugates (1:15), transferred into drops of

2.5% glutaraldehyde in cacodylate buffer 0.05 M, and

finally washed in distilled water. The sections were counter-

stained further with uranyl acetate and lead citrate before

observation.

Quantification of Immunostaining in Post-EmbeddingProcedures. After BDNF single immunostaining in mouse,

the number of gold particles/area over different subcellular

compartments (DCVs, small agranular vesicles, post-synap-

tic dendrites, cell nucleus) was calculated using the ImageJ

software (NIH, USA). The density of gold particles over

the nuclear compartment was chosen as index of back-

ground staining and subtracted from values obtained for the

other compartments. Hundred randomly selected immuno-

reactive terminals in spinal cord and 100 in the amygdala

were subjected to analysis.

Quantification of immunostaining in dual labeling

procedures (BDNFþSP or BDNFþCGRP) was carried

out in randomly chosen positive terminals from spinal

cord (257 DCVs for BDNFþSP and 264 DCVs for

BDNFþCGRP) and amygdala (267 DCVs for BDNFþSP

and 256 DCVs for BDNFþCGRP). The numbers of small

and large-sized gold particles in the DCV compartment

were calculated, and from these data the BDNF/peptide

as well as CGRP/SP ratios within individual DCVs were

estimated.

Normalization of Results in Multiple Post-EmbeddingProcedures. To calculate putative differences in the sensi-

tivity of immunolabeling for the three antigens, a test

model was developed to ensure equal labeling conditions

and comparable results. A 100 �L of 1.5% agarose in dis-

tilled water containing 0.5 �M of recombinant human

BDNF (PeproTech, UK), or synthetic SP (a kind gift of

Prof. JM Polak, London UK) or synthetic CGRP (a kind

gift of Prof. JM Polak, London, UK) were left for 1 h at

48C and subsequently fixed according to the same protocol

employed for tissue samples in 2% glutaraldehyde þ1%

paraformaldehyde for 2 h. After rinsing in PBS, the anti-

gen-containing agarose blocks were cut on a vibratome at

a thickness of 300 �m. Sections were then counter-

stained, dehydrated, and embedded using the same proce-

dure employed for post-embedding immunogold labeling

as described above. To ensure identical conditions of tis-

sue processing during all the steps of labeling and to allow

for easy identification, stacks of three agarose sections

each containing one of the antigens were prepared,

between which a piece of filter paper was interposed so

that each stack was composed as follows: paper/BDNF-

agar/paper/SP-agar/paper/CGRP-agar [Fig. 1(A)]. After

embedding, ultrathin sections were cut perpendicularly

through the stack and collected on uncoated nickel grids

[Fig. 1(B)].

The test sections, were employed in two sets of experi-

ments: (i) a simultaneous triple labeling procedure was car-

ried out using three appropriate 10 nm gold conjugates

(BDNF: anti-chick IgGs; SP: anti-rat IgGs; CGRP: anti-

rabbit IgGs) to calculate the labeling efficiencies for each

antigen under identical processing conditions; (ii) a simulta-

neous triple labeling procedure was carried out using the

10 nm gold anti-chick IgGs conjugate for BDNF and appro-

priate 20 nm gold conjugates for SP and CGRP, to estimate

possible variations in labeling efficiency as a consequence

of the use of larger colloidal gold particles in multiple

labeling procedures on tissue sections. A total of fifty elec-

tron micrographs (27500� magnification) for each layer of

the stacks corresponding to one of the three antigens were

randomly chosen for quantification from 10 different grids.

Numbers of gold particles/area and normalization coeffi-

cients (Table 1) were calculated using the ImageJ software

(NIH, USA).

Microscopy and Image Processing

Double or triple IMF was analyzed in three dimensions

with a Zeiss LSM 510 confocal microscope (Carl Zeiss

328 Salio et al.

Developmental Neurobiology. DOI 10.1002/dneu

MicroImaging, GE) in multitracking mode equipped with

three laser lines (AMCA, 345 nm; FITC, 488 nm; TRITC,

543 nm).

Electron micrographs were taken with a transmission

electron microscope (CM10, Philips, NL).

For all stainings, Photoshop version 7.0.1 (Adobe Sys-

tems, USA) was used to adjust contrast and brightness so

that individual pictures on the same composite figure

appeared similar.

Antibodies

Affinity purified antibodies against recombinant human

BDNF were from rabbit (Michael et al., 1997) or chicken

(Promega, USA; Salio et al., 2005), SP antibodies were

from rabbit (Michael et al., 1997) or rat (Merighi et al.,

1991), CGRP antibodies were from rabbit (Merighi et al.,

1991) or sheep (Michael et al., 1997); fl-trkB antibodies

were from chicken (Promega, USA; Salio et al., 2005).

RESULTS

Given the existence of some confusion in the termi-

nology used among authors to indicate the sub-cellu-

lar site of localization/storage of multiple transmit-

ters/modulators in nerve cells, we will employ here

the following definitions (Merighi, 2002): coexis-

tence indicates the concurrent presence of two or

more transmitters/modulators within a single neuron,

irrespective of the cellular compartment in which

they are localized; costorage indicates the concurrent

localization of two or more transmitters/modulators

within the same subcellular compartment. We will

avoid the use of colocalization, since this term is of-

ten employed to indicate either coexistence or co-

storage under different contexts.

Light Microscopy

DRG Neurons and Spinal Cord. It is well known

that intrathecal NGF treatment leads to an increase of

up to 80–90% of BDNF and its mRNA in a subpopu-

lation of DRG neurons (Apfel et al., 1996; Michael et

al., 1997). Therefore we have pretreated some rats

with NGF to augment BDNF expression in DRGs

and allow for an easier definition of BDNF/peptide

coexistence. Nonetheless it should be noted that

Figure 1 Semithin and ultrathin test sections of a stack

of agarose sections containing equimolar concentrations of

BDNF, SP and CGRP. (A) Semithin section showing the

arrangement of agarose-dissolved antigen slices and paper

sheets in stack. Note that to avoid misorientation only the

bottom layer of the stack is covered with a sheet of paper.

(B) Low power electron micrograph showing the stack

upon collection on a 200 mesh grid. The arrangement of the

stack is still very easily recognizable. Scale bars: 150 �m(A), 70 �m (B).

Table 1 Densities of Gold Particles Over Resin-Embedded Agarose-Dissolved Synthetic Peptides

Synthetic

Peptides

10 nm

Gold Particles

(N)

Total Area

(�2)

10 nm Gold

Particles/

Area (N/�2)

20 nm Gold

Particles (N)

Total

Area (�2)

20 nm Gold

Particles/

Area (N/�2)

Estimated

10 nm/20 nm

Ratio

BDNF (0.5 �M) 2088 353.64 5.9 – – – –

CGRP (0.5 �M) 2005 351.5 5.7 1056 353.17 2.99 1.9

SP (0.5 �M) 2008 351.82 5.7 1046 351.8 2.97 1.9

Densities of gold particles (number of gold particles/area) were calculated over resin-embedded agarose-dissolved synthetic peptides at the

same molar concentrations. The results of the first set of experiments—simultaneous triple labeling procedure with anti-chick IgGs (BDNF),

anti-rat IgGs (SP), and anti-rabbit IgGs (CGRP)—are reported in columns 2–4. There are no differences in labeling efficiencies when gold

conjugates of the same size are employed to label the three antigens. These results also confirm that only one single epitope is recognized by

the respective antibody for each peptide molecule. The results of the second set of experiments—simultaneous triple labeling procedure with

the 10 nm gold anti chick-IgGs (BDNF) and 20 nm gold conjugates for SP and CGRP—are reported in columns 5–7. Note the reduction in

labeling densities between 20 and 10 nm gold (columns 4 and 7). These data were used to calculate a normalization coefficient for BDNF

(10 nm) þ CGRP/SP (20 nm) double labelings in tissue sections, and this is expressed as the ratio of labeling densities obtained with 10 and

20 nm gold (column 8).

BDNF and Neuropeptides in DCVs 329

Developmental Neurobiology. DOI 10.1002/dneu

results obtained in control and experimental rats were

completely overlapping. At light microscopic level in

both rat and mouse, small to medium sized DRG neu-

rons were observed that were double or triple labeled

for BDNF and SP or CGRP. Labeling was clearly

visible in the form of a punctate staining within neu-

ronal cell bodies and processes, with individual fluo-

rescent puncta showing double labeling for BDNF

and one of the two neuropeptides [Fig. 2(A)]. A simi-

lar pattern could be visualized in the superficial lami-

nae of the spinal cord dorsal horn, where the central

projections of DRG neurons are known to terminate.

In this region, individual axon terminals showed dou-

ble or triple labeling [Fig. 2(B–D)].

Amygdala. In both species, the central nucleus of the

amygdala displayed a dense network of BDNF-im-

munoreactive axon terminals [Fig. 2(E)]. In the same

area, intense staining for SP [Fig. 2(F)] or CGRP was

also detected. Confocal analysis of BDNF plus SP

[Fig. 2(G,H)] or CGRP double labeled preparations

showed that the neurotrophin and peptides coexisted

in individual axons.

Electron Microscopy

Spinal Cord and Amygdala. High resolution electron

microscopic analysis, after either pre- [Fig. 3(A)] or

post-embedding [Figs. 3(B–E) and 4] procedures,

showed that BDNF immunoreactivity was concen-

trated in DCVs. Labeled DCVs were present in axon

terminals that consistently contained also variable

numbers of agranular synaptic vesicles. These termi-

nals formed axo-dendritic synapses or synaptic glo-

meruli (spinal cord only) with dendrites that were

devoid of DCVs, and thus unlabeled. In single post-

embedding procedures and after non-quantitative esti-

mation, DCVs appeared to be the sole site of BDNF

Figure 2 Coexistence of BDNF and peptides. BDNF coexists with SP (A) and CGRP in neurons

of DRGs, spinal cord (B–D) and the central nucleus (CN) of the amygdala (E–H). (A) Confocal

stack (5 � 0.4 � in the Z axis) showing dual color IMF for BDNF (rhodamine, red) and SP (fluores-

cein, green) in a DRG neuron from a rat that received an intrathecal infusion of NGF. The two mol-

ecules are stored in prominent granules within the cell body (asterisk) and the two axonal branches.

Single labeled (arrows) and double labeled (double arrowheads) granules are indicated. (B–D)

Triple labeled (BDNF, CGRP and SP) axonal varicosities (double arrowheads) in lamina II of the

rat spinal cord. (E,F) Low magnification IMF showing that BDNF and SP display a partially over-

lapping distribution in the rat CN. (G,H) At higher magnification BDNF (fluorescein, green) and

SP (rhodamine, red) coexist in double labeled axons (arrows). Confocal stack (5 � 0.58 � in the

Z axis). Scale bars: 10 �m (A), 50 �m (B-D, G,H), 200 �m (E,F).

330 Salio et al.

Developmental Neurobiology. DOI 10.1002/dneu

immunoreactivity. In these experiments, isolated,

unevenly scattered gold particles were, at times,

observed over the small agranular vesicles, and other

subcellular compartments such as the nucleus, cytosol,

Golgi apparatus, lysosomes, and multivesicular bodies.

Quantitative analysis of the relative concentration of

BDNF in some of the above subcellular compartments,

corrected for background staining over the nuclear

compartment, demonstrated that the density of gold

particle labeling over the DCVs was 31.1 (401.5 par-

ticles/12.9 �2-spinal cord) or 36.0 (501.5 particles/

13.9 �2 - amygdala) times higher than over the small

agranular vesicles (Table 2). The ratios of gold particle

densities over the DCVs compared with that over the

post-synaptic dendrites were 308.8 (401.5/1.3) in spi-

nal cord and 146.6 (501.5/3.4) in amygdala.

By taking into consideration data obtained after im-

munolabeling of test sections (Table 1), and assuming

a linear relationship between BDNF concentration

and the degree of gold immunolabeling, the estimated

concentration of BDNF in DCVs is 34.07 �M in spi-

nal cord and 42.54 �M in amygdala.

In multiple post-embedding immunogold labeling

experiments, DCVs were observed that were single la-

beled for BDNF, SP or CGRP, together with others that

were double or triple labeled [Figs. 3(B–E) and 4].

Quantitative analysis of individual DCVs after double

labeling experiments (Table 3) showed that about 20%

of the BDNF-immunoreactive DCVs were also immu-

nopositive for SP/CGRP, whereas another 18% appeared

to contain the peptide only in both spinal cord and amyg-

dala. Interestingly, the percentages of BDNF single la-

beled vesicles in both BDNFþSP and BDNFþCGRP

immunolabeled terminals were different in spinal cord

(35.2) and amygdala (22.5). Consequently, unlabeled

vesicles were more abundant in amygdala than in spinal

cord. A second series of quantitative analyses was car-

ried out to estimate the relative densities of BDNF/pep-

tide immunogold labeling over the DCV compartment

considered in its entirety (Table 4).

The crude ratios were: BDNF/SP 1.50 (325/216-

spinal cord) or 1.28 (343/268-amygdala); BDNF/

CGRP 1.30 (335/258-spinal cord) or 1.38 (319/231-

amygdala); SP/CGRP 0.84 (216/258-spinal cord) or

1.16 (268/231-amygdala). When the values obtained

for spinal cord and amygdala are averaged, the fol-

lowing values were obtained: BDNF/SP 1.39; BDNF/

CGRP 1.34; SP/CGRP 1.00.

After normalization (Table 1), ratios were cor-

rected as follows: BDNF/SP 0.79 (325/410.4 - spinal

cord) or 0.67 (343/509.2-amygdala); BDNF/CGRP

0.68 (335/490-spinal cord) or 0.73 (319/439-amyg-

dala); SP/CGRP 0.83 (410.4/490-spinal cord) or 1.15

(509.2/439-amygdala). The new averaged values

obtained were: BDNF/SP 0.73; BDNF/CGRP 0.70;

SP/CGRP 0.99.

In a final series of multiple staining experiments

(spinal cord only), pre-embedding immunolabeling of

fl-trkB receptors was combined with post-embedding

immunolabeling of BDNF together with CGRP [Fig.

3(D)] or SP [Fig. 3(E)]. Interestingly, fl-trkB immu-

noreactive dendrites that were engaged in axo-dendri-

tic synapses with BDNF immunopositive axons were

completely devoid of BDNF immunoreactivity.

DISCUSSION

According to our current views on chemical neuro-

transmission, vesicular storage of neurotransmitters is

crucial to the function of synapses. In this study we

have for the first time provided direct ultrastructural

and quantitative evidence that BDNF is packaged in

DCVs, and that, within this subcellular compartment,

it may costore with the neuropeptides SP and CGRP.

In both rat and mouse this pattern has been observed

in two different neuronal populations: the DRG neu-

rons in PNS and the neurons of the parabrachial nu-

cleus projecting to the amygdala in CNS; moreover,

in rat, a similar pattern of costorage is seen when the

levels of BDNF and peptides in DRGs have been

experimentally manipulated by intrathecal adminis-

tration of NGF. Therefore, our study provides a defin-

itive demonstration that BDNF meets one of the

fundamental criteria to act as a neurotransmitter/neu-

romodulator, and offers a structural basis for the

corelease of the neurotrophin with one or more neuro-

peptide transmitters. In keeping with the present

observations, it was previously demonstrated, in cul-

tured primary neurons and neuronal cell lines, that

release of BDNF occurs along the regulated secretory

pathway, where signaling molecules released in

response to external stimuli, such as depolarization,

are specifically sorted by the cell (Thoenen, 1995;

Blochl and Thoenen, 1996).

Subcellular Storage of BDNF

The subcellular site of storage of BDNF has been the

subject of several investigations in the recent past, but

most of the work in this field has been done using con-

focal microscopy on isolated neurons and/or cell lines

(Mowla et al., 1999; Wu et al., 2004) or by a biochem-

ical approach (Fawcett et al., 1997; Berg et al., 2000).

High resolution immunocytochemistry at the electron

microscope level was clearly needed to provide an

unequivocal answer, but previous work at the ultra-

structural level (Michael et al., 1997; Luo et al., 2001)

BDNF and Neuropeptides in DCVs 331

Developmental Neurobiology. DOI 10.1002/dneu

did not allow to assess with certainty whether or not

BDNF, besides being localized to DCVs, is also stored

in agranular vesicles. This is because of the well

known limitations in the pre-embedding immunoper-

oxidase methodology (Priestley et al., 1992). On the

other hand, post-embedding immunocytochemistry has

often been subjected to criticism, because it is assumed

that only a small fraction of antigenicity survives the

embedding procedure in hydrophobic resins (Merighi

and Polak, 1993). We have therefore used the two

methods together to provide here an unequivocal dem-

onstration that BDNF is highly concentrated within

DCVs, which are well known to participate in neuro-

transmitter synthesis and storage. After extrapolation

of data obtained with immunolabeling in test slices of

agarose-dissolved recombinant BDNF, it was possible

to estimate that the density of gold particles over

DCVs corresponded to a concentration of 34.07 �M in

Figure 3

332 Salio et al.

Developmental Neurobiology. DOI 10.1002/dneu

spinal cord and 42.54 �M in amygdala. These figures

are well in line with the widely accepted notion that

neuropeptides are usually stored in DCVs at much

lower concentration that conventional transmitters

(Maxwell Cowan and Kandell, 2001), and comparison

with test models may lead to an under- rather than an

overestimation of antigen concentration in vesicles

(see Shupliakov et al., 1992 for further discussion).

Our high resolution post-embedding staining demon-

strates clearly on a quantitative basis that DCVs have a

BDNF content 31 (spinal cord)-36 (amygdala) times

higher than agranular vesicles. These results are con-

sistent with Western blot analysis of rat synaptosomes,

which have shown that BDNF is colocalized with the

synaptic marker synaptotagmin, although by this

approach it was not possible to ascertain whether the

neurotrophin was contained in agranular vesicles,

DCVs or both (Fawcett et al., 1997; Berg et al., 2000).

The observation that the density of gold particles over

the small agranular vesicle compartment, albeit much

lower than over DCVs, is anyway higher than back-

ground staining (13.4 versus 0.5 particles/�2 in spinal

cord and 14.4 versus 0.5 particles/�2 in amygdala) is

puzzling. It is generally accepted that high molecular

weight transmitters, such as the neuropeptides, are

packaged in DCVs, whereas low molecular weight

ones are stored in agranular vesicles (Merighi, 2002).

Under the experimental conditions employed here

(whole IgGs and 10 nm gold particles) the spatial reso-

lution of the technique is 21 nm (Merighi and Polak,

1993). Thus it remains impossible to ascertain with

certainty whether or not BDNF (a high molecular

weight species) is indeed contained within agranular

vesicles (that have a mean diameter around 40 nm) or

whether immunoreactivity is due to a limited number

of BDNF molecules that are localized in the cytosol

trapped among the agranular vesicles. This latter possi-

bility seems to be more realistic, and is supported by

the lack of immunostaining inside agranular vesicles

after immunoperoxidase localization [see insert of Fig.

3(A)]. If indeed some BDNF remains trapped among

the agranular vesicles, its presence can be explained,

albeit on purely speculative bases, by at least two dif-

ferent mechanisms. First, one cannot rule out that

some BDNF leaks out from the DCVs before the fixa-

tive has reached its target. Second, if BDNF, as it

appears to be the case for neuropeptides, can be

released by kiss and run of DCVs that undergo repeti-

tive exocytotic cycles (Tsuboi and Rutter, 2003; Rutter

and Tsuboi, 2004), one cannot exclude that some of

the neurotrophin leaks from the DCV core during the

process of vesicle emptying. On the other hand, it

seems highly improbable that BDNF is released from

agranular vesicles, since in cultured hippocampal neu-

Figure 3 DCV costorage of BDNF and peptides in spinal cord after pre- and post-embedding

immunolabeling. (A) An axon terminal in lamina II of the rat spinal cord is stained for combined

BDNF pre-embedding immunoperoxidase labeling and CGRP post-embedding gold labeling. The

insert images (1) are of a group of DCVs, with contrast adjusted to show that some BDNF peroxi-

dase labeled DCVs also display CGRP immunogold labeling. A DCV displaying BDNF labeling is

indicated by the arrow in the upper insert, another DCV is double labeled as indicated by the double

arrowheads. Note the absence of BDNF immunolabeling over agranular vesicles, despite the higher

sensitivity of the pre-embedding procedure. This result rules out the possibility that lack of BDNF

immunostaining in post-embedding procedures is an artifact due to reduction of antigenicity. (B,C)

Single, double, or triple labeled DCVs are visible in a simple axon terminal (B) and in a type I glo-

merulus (C) of mouse dorsal horn after post-embedding immunogold labeling. The simple terminal

(B) is in contact with an unlabeled dendrite. Note the lack of DCVs at the synaptic specialization.

Inserts show sample double (2) and triple (3) labeled DCVs (10 nm gold is not indicated by

markers, 20 nm gold is indicated by arrowheads, 30 nm gold by the arrow). (D–E) Combined pre-

embedding localization of fl-trkB receptors and post-embedding BDNF/peptide immunoreactivities

in axon terminals of mouse dorsal horn lamina II. BDNF/peptide immunostaining has been carried

out by conventional post-embedding procedures with dual size colloidal gold particles (see insert

4). Fl-trkB receptors have been localized by ultrasmall gold particles followed by gold intensifica-

tion, a procedure that gives rise to enlarged particles of irregular shapes that are easily distinguished

from colloidal gold (see inserts 5 and 6). Receptor immunoreactivity is clustered at the intracyto-

plasmatic aspect of the plasma membrane of post-synaptic dendrites, but not at synaptic specializa-

tions (inserts 5 and 6). The subcellular localization of receptor immunoreactivity is consistent with

the specificity of the primary antibody, directed against the intracytoplasmatic catalytic domain of

trkB (see Salio et al., 2005 for further discussion). Note the complete lack of BDNF immunostain-

ing at synapses and in the dendritic cytoplasm. Scale bars: 500 nm (A–E), 100 nm (inserts 1),

50 nm (inserts 2–6).

BDNF and Neuropeptides in DCVs 333

Developmental Neurobiology. DOI 10.1002/dneu

rons it was demonstrated that its rate of secretion is

more than 10 times slower than glutamate, and this has

been linked to the time needed for dissolution of the

peptide core in DCVs (Brigadski et al., 2005).

The issue of the subcellular localization of BDNF

in vivo has recently been exhaustively reviewed

(Lessman et al., 2003), and these authors have very

well outlined some difficulties in the interpretation of

ultrastructural studies, in particular in distinguishing

between synthesized and endocytosed BDNF. This

aspect is also quite important, since it was previously

demonstrated that BDNF can avoid degradation after

being internalized in the lysosomal compartment, and

thus enter a transcytosis pathway that enables it to

move across multiple synapses (von Bartheld et al.,

2001; von Bartheld, 2004). In keeping with such a

possibility, by using a pre-embedding approach it

was reported, on purely qualitative bases, that the

post-synaptic membrane was occasionally labeled af-

ter BDNF immunostaining (Luo et al., 2001). Analy-

sis of our post-embedding immunostained prepara-

tions failed to show significant labeling over the

Golgi apparatus, lysosomes, and multivesicular

bodies, i.e. the cellular compartments through which

Figure 4 DCV costorage of BDNF and peptides in amygdala. Post-embedding immunogold

labeling for BDNFþCGRP (A), BDNFþSP (B) and BDNFþSPþCGRP (C) in mouse CN. Inserts

show sample double (1 and 2) and triple (3 and 4) labeled DCVs (10 nm gold is not indicated by

markers, 20 nm gold is indicated by arrowheads, 30 nm gold by arrows). Scale bars: 500 nm (A–

C), 50 nm (inserts 1–4).

334 Salio et al.

Developmental Neurobiology. DOI 10.1002/dneu

the neurotrophin should transit along a presumptive

transcytotic pathway. Moreover, we have calculated

that dendrites post-synaptic to BDNF-immunostained

terminals have a content of BDNF of 1/223 (spinal

cord) or 1/128 (amygdala) that of DCVs, and that

such a content is only slightly higher than background

staining. Pre-embedding immunocytochemical stud-

ies need to be interpreted critically, since relocation

of antigens and/or secondary reagents can occur

(Priestley et al., 1992). The use of post-embedding

staining is therefore highly recommended, since by

such an approach one can easily distinguish between

DCVs and endocytotic vesicles by their ultrastruc-

tural features, and we have not observed BDNF im-

munostaining within endocytotic vesicles. In addi-

tion, if transcytosis of BDNF occurs at synapses, it

should follow exocytosis into the synaptic cleft and

subsequent endocytosis upon binding to the high af-

finity receptor trkB. However, in multiple labeled

preparations whereby BDNF and fl-trkB receptors

were simultaneously localized at individual synapses,

there was no BDNF immunoreactivity at synaptic

specializations, as well as within the cytoplasm of fl-

trkB immunolabeled dendrites. Therefore, from all

the above considerations, one is led to conclude that

transcytosis of BDNF is unlikely to be of relevance

in vivo in either spinal cord or amygdala.

Costorage of BDNF and Peptides in DCVs

At light microscopy level, coexistence of BDNF and

neuropeptides has been the focus of several studies in

PNS (Michael et al., 1997; Lever et al., 2001; Luo

et al., 2001), but has never specifically been examined

in CNS, although neuropeptides have been reported in

neuronal systems that are known to contain BDNF

(Schwaber et al., 1988). In this study, we have shown

that the previously reported coexistence of BDNF and

SP/CGRP in rat DRG neurons (Michael et al., 1997;

Lever et al., 2001; Luo et al., 2001) also occurs in

mouse, and that, in both species, the same pattern of

coexistence occurs in axon terminals within the cen-

tral nucleus of the amygdala. We did not establish the

origin of the BDNF-immunoreactive terminals in our

preparations, but previous studies have clearly estab-

lished that terminals in the central nucleus expressing

BDNF (Conner et al., 1997), SP (Block et al., 1989)

or CGRP (Schwaber et al., 1988) originate from the

parabrachial nucleus. Although we only examined this

one population of CNS neurons, coexistence of BDNF

and peptides is likely to be a general phenomenon,

given the widespread occurrence of BDNF (Conner

et al., 1997; Yan et al., 1997) and neuropeptides (Mer-

ighi, 2002) in CNS axon terminals.Table2

BDNFDistributionin

DifferentSubcellularCompartm

entsofMouse

SpinalCordandAmygdala

AfterSinglePost-EmbeddingIm

munogold

Labeling

Cell

Nuclei

Total

Area(�

2)

Gold

Particles

in

Cell

Nuclei(N

)

Gold

Particles/

Nuclear

Area

(N/�

2)

DCVTotal

Area(�

2)

Gold

Particles

inDCVs

(N)

Gold

Particles/

DCV

Area

(N/�

2)

Small

Agranular

Vesicle

TotalArea

(�2)

Gold

Particles

in

Small

Agranular

Vesicles(N

)

Gold

Particles/

Small

Agranular

VesicleArea

(N/�

2)

Post-Synaptic

Dendrite

Total

Area(�

2)

Gold

Particles

in

Post-Synaptic

Dendrites(N

)

Gold

Particles/

Post-Synaptic

Dendrite

Area(N

/�2)

Spinalcord

110.4

60

0.5

6.74

2712

402

12.7

171

13.4

39

72

1.8

Amygdala

109.5

53

0.48

3.84

1930

502

8.09

117

14.4

13.2

51

3.9

Densities

ofgold

particles

(number

ofgold

particles/area)

werecalculatedin

differentsubcellularcompartm

entsofthemouse

spinal

cord

andam

ygdala.Gold

particlenumbersandareasoftheDCV,

smallagranularvesicleandpost-synapticdendritecompartm

entswerecalculatedin

atotalof200labeled

term

inalsengaged

inaxo-dendriticsynapses,randomly

selected

from

50differentultrathin

sec-

tionsfrom

atleast25differentgrids.Gold

particlenumbersover

nuclei,andnuclearareaswerecalculatedfrom

thesamegrids.ThetotalDCV

area

was

calculatedbyaddingtheareasofindividual

DCVs,whereasthetotalsm

allagranularvesicle

area

was

calculatedbydelim

itingtheclustersofvesicleswithin

each

term

inal.Thepost-synapticdendrite

area

was

calculatedbydelim

itingthepost-

synapticdendriticprofileateach

axo-dendriticsynapse.

BDNF and Neuropeptides in DCVs 335

Developmental Neurobiology. DOI 10.1002/dneu

In addition to coexistence at the cell level, we have

shown that BDNF, SP and CGRP are costored within

single DCVs. The methodology involved was techni-

cally demanding and this is the first time such a direct

demonstration is given, although several recent re-

views have speculated that BDNF/peptide costorage

may occur (Fawcett et al., 1997; Michael et al., 1997;

Berg et al., 2000; Luo et al., 2001; von Bartheld et al.,

2001; Lessman et al., 2003; von Bartheld, 2004).

In addition, we show here on quantitative grounds

that the ratio of BDNF/peptides within DCVs is

remarkably constant in both locations examined. This

figure has been calculated by demonstrating that, in

test model experiments, the number of gold particles/

area is related to the molar concentration of antigens,

and sensitivity of detection for the three different

antigens is identical when 10 nm gold probes are

employed, in accord with literature on post-embed-

ding multiple immunostaining procedures (Merighi

and Polak, 1993). Moreover, we have calculated that

labeling efficiency for neuropeptides is reduced by a

1.9 factor when 20 nm gold probes are employed,

and thus we have normalized the BDNF/peptide con-

tent in our double labeling experiments. Results of

quantitative analysis led to a remarkable uniformity

of data in both spinal cord and amygdala. Since we

could estimate that the two peptides are costored in a

ratio of about 1:1, and assuming that no other pepti-

des with neurotransmitter function are contained in

DCVs together with BDNF, one can conclude that in

the two neuronal systems examined, the BDNF con-

tent of DCVs is about 25%. Taken together these

data indicate that BDNF contained in individual

DCVs should be more than sufficient to elicit a bio-

logical response upon release.

Costorage of BDNF and neuropeptides strongly

suggests that similar mechanisms control the release

of both types of molecules, and that they may there-

fore be released together. This interpretation is con-

sistent with the existing literature on neuropeptide

coexistence (Bean et al., 1994; Merighi, 2002) and

with recent studies on BDNF release (Lessman et al.,

2003; Brigadski et al., 2005). Such studies have

clearly shown that many parameters associated with

neuropeptide secretion, such as the lack of physical

docking at synaptic sites, the virtual lack of fusion-

competent DCVs, the need for prolonged intracellular

Ca2þ elevations in the release compartment, and the

slow emptying of peptide content from DCVs, also

apply to BDNF released from neurons (Balkowiec

and Katz, 2000; Lessman et al., 2003; Brigadski et al.,

2005). Recent studies have shown that DCVs can

release some of their cargo by kiss and run, raising the

possibility of differential release of low versus high

molecular weight components (Tsuboi and Rutter,

2003; Rutter and Tsuboi, 2004). However, simultane-

ous capacitance measurements and confocal imaging

has shown that peptide release by this mechanism is

negligible, whereas complete vesicle fusion is usually

required (Barg et al., 2002). Moreover, most of the

studies on neuropeptide release have been carried out

on isolated neurons in culture, and their relevance

in vivo remains to be established.

Given that BDNF and peptides are costored,

fusion of a DCV containing, for example, both BDNF

and SP would therefore result in the release of both

molecules. However Lever et al. (2001) have shown,

in the dorsal horn, that one pattern of afferent stimu-

lation releases SP alone, while another pattern

releases SP together with BDNF. Possible mecha-

nisms of differential release are suggested by our

quantitative analyses. Costorage of BDNF and SP/

CGRP was seen in only a fraction of the total popula-

tion of labeled DCVs. Thus, the differential release of

BDNF and neuropeptides that has been reported in

some studies may simply reflect the exocytosis of dif-

Table 3 Costorage of BDNF/Peptides in DCVs of Mouse Spinal Cord and Amygdala

Immunostaining

Protocol (location)

Terminals

(N)

Total Number

of DCVs (N)

Double Labeled

DCVs (N)

BDNF Single

Labeled

DCVs (N)

Peptide Single

Labeled

DCVs (N)

Unlabeled

DCVs (N)

BDNFþSP 85 1186 257 406 201 322

(spinal cord) 100% 21.7% 34.2% 16.9% 27.2%

BDNFþSP 99 1350 267 301 255 527

(amygdala) 100% 19.8% 22.3% 18.9% 39%

BDNFþCGRP 81 1267 264 459 211 333

(spinal cord) 100% 20.8% 36.2% 16.7% 26.3%

BDNFþCGRP 93 1252 256 284 227 485

(amygdala) 100% 20.5% 22.7% 18.1% 38.7%

Numbers and percentages of labeled vesicles in dual labeling experiments using combinations of BDNFþSP or BDNFþCGRP antibodies.

A total of 358 dual labeled terminals were randomly selected from spinal cord and amygdala (2nd column). DCVs were considered to be la-

beled if they contained at least two gold particles of the size corresponding to BDNF or to peptide.

336 Salio et al.

Developmental Neurobiology. DOI 10.1002/dneu

ferent subpopulations of mature DCVs containing ei-

ther BDNF, or the peptides, or both.

However, after bulk analysis, BDNF, SP and

CGRP were shown to be stored in the DCV compart-

ment in a stoichiometric ratio of 0.7:1:1. We cannot

exclude the possibility to have slightly underesti-

mated the BDNF content in DCVs in simultaneous

labeling experiments on tissue sections. In test model

experiments the anti-BDNF antibody was demon-

strated to label only one epitope on the molecule, and

equimolar concentrations of BDNF, SP, and CGRP

yielded identical labeling densities. In the real tissue,

an epitope must be exposed at the section surface to

be available for labeling, and BDNF, being much

larger than the peptides, has a lower probability to

give rise to a positive stain (see Merighi and Polak,

1993 for further discussion). If we have indeed under-

estimated BDNF/peptide costorage, and all DCVs

contain a cocktail of the three molecules, the differen-

tial release of BDNF and costored peptides could

more likely rely on differences in the relative rate of

their dissolution from the DCV core, since this

appears to be the critical determinant of the speed of

peptide/neurotrophin secretion in vitro (Brigadski

et al., 2005).

Altogether, our results support the view that

BDNF and neuropeptides are released together from

DCVs in proportion to their degree of costorage, but

that separate release of BDNF and neuropeptides may

also occur.

We gratefully acknowledge support from the Italian

MIUR, Regione Piemonte, Compagnia di San Paolo, Well-

come Trust, and generous provision of rabbit BDNF antise-

rum by Dr. Qiao Yan (Amgen) and recombinant NGF by

Genentech.

REFERENCES

Altar CA, Cai N, Bliven T, Juhasz M, Conner JM, Acheson

AL, Lindsay RM, et al. 1997. Anterograde transport of

brain-derived neurotrophic factor and its role in the brain.

Nature 389:856–860.

Altar CA, DiStefano PS. 1998. Neurotrophin trafficking

by anterograde transport. Trends Neurosci 21:433–

437.

Apfel SC, Wright DE, Wiideman AM, Dormia C, Snider

WD, Kessler JA. 1996. Nerve growth factor regulates the

expression of brain-derived neurotrophic factor mRNA

in the peripheral nervous system. Mol Cell Neurosci

7:134–142.

Balkowiec A, Katz DM. 2000. Activity-dependent release

of endogenous brain-derived neurotrophic factor from

primary sensory neurons detected by ELISA in situ.

J Neurosci 20:7417–7423.Table4

RelativeDensitiesofBDNF/PeptideIm

munogold

Labelingin

DoubleLabeledDCVsFrom

Mouse

SpinalCordandAmygdala

Immunostaining

Protocol

(Location)

DCVs

(N)

DCV

Total

Area(�

2)

Small

Sized

Gold

Particles

(BDNF)in

DCVs(N

)

BDNF

Particles/

DCVArea

(N/�

2)

Large

Sized

Gold

Particles

(SP)in

DCVs(N

)

Uncorrected

SPParticles/

DCVArea

(N/�

2)

Norm

alized

SPParticles/

DCVArea

(N/�

2)

Large

Sized

Gold

Particles

(CGRP)in

DCVs(N

)

Uncorrected

CGRP

Particles/

DCVArea

(N/�

2)

Norm

alized

CGRP

Particles/

DCVArea

(N/�

2)

Uncorrected

BDNF/

Peptide

Ratio

in

DVCs

Norm

alized

BDNF/

Peptide

Ratio

in

DVCs

BDNFþS

P

(spinalcord)

257

2.90

944

325

628

216

410.4

(216�

1.9)

––

–1.50

0.79

BDNFþS

P

(amygdala)

267

2.60

894

343

699

268

509.2

(268�

1.9)

––

–1.28

0.67

BDNFþC

GRP

(spinalcord)

264

2.60

871

335

––

–671

258

490

(258�

1.9)

1.30

0.68

BDNFþC

GRP

(amygdala)

256

3.01

963

319

––

–696

231

439

(231�

1.9)

1.38

0.73

Densities

ofgold

particles

(number

ofgold

particles/area)

over

theDCV

compartm

entin

dual

labelingexperim

entsusingcombinationsoftheBDNFþS

PorBDNFþC

GRPantibodies.Thetotal

DCVarea

was

calculatedbyaddingtheareasofdoublelabeled

individualDCVsin

immunoreactiveterm

inals.Only

doublelabeled

DCVswereconsidered

inthistypeofanalysisin

order

tofirstextrap-

olatetheuncorrectedratioofBDNF/peptidecontent(astheratioofvalues

given

incolumns5,7,and10)within

individualDCVs.After

calculationofthenorm

alizationcoefficientforpeptidelabeling

using20nm

gold

particles

(see

Table

1)thenorm

alized

densities

ofpeptideim

munolabelingarereported

incolumns8and11,andthenorm

alized

ratioofBDNF/peptidecontentwas

determined

(as

theratioofvalues

given

incolumns5and8,11).

BDNF and Neuropeptides in DCVs 337

Developmental Neurobiology. DOI 10.1002/dneu

Barg S, Olofsson CS, Schriever-Abeln J, Wendt A, Gebre-

Medhin S, Renstrom E, Rorsman P. 2002. Delay between

fusion pore opening and peptide release from large dense-

core vesicles in neuroendocrine cells. Neuron 33:287–299.

Bean AJ, Zhang X, Hokfelt T. 1994. Peptide secretion:

What do we know? FASEB J 8:630–638.

Berg EA, Johnson RJ, Leeman SE, Boyd N, Kimerer L,

Fine RE. 2000. Isolation and characterization of sub-

stance P-containing dense core vesicles from rabbit optic

nerve and termini. J Neurosci Res 62:830–839.

Blochl A, Thoenen H. 1996. Localization of cellular storage

compartments and sites of constitutive and activity-depend-

ent release of nerve growth factor (NGF) in primary cultures

of hippocampal neurons. Mol Cell Neurosci 7:173–190.

Block CH, Hoffman G, Kapp BS. 1989. Peptide-containing

pathways from the parabrachial complex to the central

nucleus of the amygdala. Peptides 10:465–471.

Brigadski T, Hartmann M, Lessmann V. 2005. Differential

vesicular targeting and time course of synaptic secretion of

the mammalian neurotrophins. J Neurosci 25:7601–7614.

Conner JM, Lauterborn JC, Yan Q, Gall CM, Varon S.

1997. Distribution of brain-derived neurotrophic factor

(BDNF) protein and mRNA in the normal adult rat CNS:

Evidence for anterograde axonal transport. J Neurosci

17:2295–2313.

Fawcett JP, Alonso-Vanegas MA, Morris SJ, Miller FD,

Sadikot AF, Murphy RA. 2000. Evidence that brain-

derived neurotrophic factor from presynaptic nerve ter-

minals regulates the phenotype of calbindin-containing

neurons in the lateral septum. J Neurosci 20:274–282.

Fawcett JP, Aloyz R, McLean JH, Pareek S, Miller FD,

McPherson PS, Murphy RA. 1997. Detection of brain-

derived neurotrophic factor in a vesicular fraction of

brain synaptosomes. J Biol Chem 272:8837–8840.

Kohara K, Kitamura A, Morishima M, Tsumoto T. 2001.

Activity-dependent transfer of brain-derived neurotrophic

factor to postsynaptic neurons. Science 291:2419–2423.

Lessmann V, Gottmann K, Malcangio M. 2003. Neurotro-

phin secretion: Current facts and future prospects. Prog

Neurobiol 69:341–374.

Lever IJ, Bradbury EJ, Cunningham JR, Adelson DW,

Jones MG, McMahon SB, Marvizon JC, et al. 2001.

Brain-derived neurotrophic factor is released in the dor-

sal horn by distinctive patterns of afferent fiber stimula-

tion. J Neurosci 21:4469–4477.

Li H, Waites CL, Staal RG, Dobryy Y, Park J, Sulzer DL,

Edwards RH. 2005. Sorting of vesicular monoamine

transporter 2 to the regulated secretory pathway confers

the somatodendritic exocytosis of monoamines. Neuron

48:619–633.

Luo XG, Rush RA, Zhou XF. 2001. Ultrastructural local-

ization of brain-derived neurotrophic factor in rat pri-

mary sensory neurons. Neurosci Res 39:377–384.

Maxwell CW, Kandell ER. 2001. A brief history of synap-

ses and synaptic transmission. In: Maxwell Cowan W,

Sudhof TC, Stevens CF, editors. Synapses. Baltimore:

John Hopkins University Press, pp 1–87.

Merighi A. 2002. Costorage and coexistence of neuropepti-

des in the mammalian CNS. Prog Neurobiol 66:161–190.

Merighi A, Polak J. 1993. Postembedding immunogold

staining. In: Cuello AC, editor. Immunohistochemistry

II. New York: Wiley, pp 229–264.

Merighi A, Polak JM, Theodosis DT. 1991. Ultrastructural

visualization of glutamate and aspartate immunoreactiv-

ities in the rat dorsal horn, with special reference to the

co-localization of glutamate, substance P and calcitonin-

gene related peptide. Neuroscience 40:67–80.

Michael GJ, Averill S, Nitkunan A, Rattray M, Bennett DL,

Yan Q, Priestley JV. 1997. Nerve growth factor treatment

increases brain-derived neurotrophic factor selectively in

TrkA-expressing dorsal root ganglion cells and in their

central terminations within the spinal cord. J Neurosci

17:8476–8490.

Mowla SJ, Pareek S, Farhadi HF, Petrecca K, Fawcett JP,

Seidah NG, Morris SJ, et al. 1999. Differential sorting of

nerve growth factor and brain-derived neurotrophic fac-

tor in hippocampal neurons. J Neurosci 19:2069–2080.

Priestley JV, Alvarez FJ, Averill S. 1992. Pre-embedding elec-

tron microscopic immunocytochemistry. In: Polak JM and

Priestley JV, editors. Electron Microscopic Immunocyto-

chemistry. Oxford: Oxford University Press, pp 89–121.

Rutter GA, Tsuboi T. 2004. Kiss and run exocytosis of

dense core secretory vesicles. Neuroreport 15:79–81.

Salio C, Lossi L, Ferrini F, Merighi A. 2005. Ultrastructural

evidence for a pre- and postsynaptic localization of full-

length trkB receptors in substantia gelatinosa (lamina II)

of rat and mouse spinal cord. Eur J Neurosci 22:1951–

1966.

Schwaber JS, Sternini C, Brecha NC, Rogers WT, Card JP.

1988. Neurons containing calcitonin gene-related peptide

in the parabrachial nucleus project to the central nucleus

of the amygdala. J Comp Neurol 270:416–419.

Shupliakov O, Brodin L, Culheim S, Ottersen OP, Storm-

Mathisen J. 1992. Immunogold quantification of gluta-

mate in two types of excitatory synapses with different

firing patterns. J Neurosci 12:3789–3803.

Smith MA, Zhang LX, Lyons WE, Mamounas LA. 1997.

Anterograde transport of endogenous brain-derived neu-

rotrophic factor in hippocampal mossy fibers. Neurore-

port 8:1829–1834.

Thoenen H. 1995. Neurotrophins and neuronal plasticity.

Science 270:593–598.

Tsuboi T, Rutter GA. 2003. Insulin secretion by kiss-and-

run exocytosis in clonal pancreatic islet �-cells. BiochemSoc Trans 31:833–836.

von Bartheld CS. 2004. Axonal transport and neuronal

transcytosis of trophic factors, tracers, and pathogens.

J Neurobiol 58:295–314.

von Bartheld CS, Wang X, Butowt R. 2001. Anterograde

axonal transport, transcytosis, and recycling of neurotro-

phic factors: The concept of trophic currencies in neural

networks. Mol Neurobiol 24:1–28.

Wu YJ, Kruttgen A, Moller JC, Shine D, Chan JR, Shooter

EM, Cosgaya JM. 2004. Nerve growth factor, brain-

derived neurotrophic factor, and neurotrophin-3 are

sorted to dense-core vesicles and released via the regu-

lated pathway in primary rat cortical neurons. J Neurosci

Res 75:825–834.

338 Salio et al.

Developmental Neurobiology. DOI 10.1002/dneu