Coordinated Interaction of Neurogenesis and Angiogenesis in the Adult Songbird Brain

Abstract

Neurogenesis proceeds throughout life in the higher vocal center (HVC) of the adult songbird neostriatum. Testosterone induces neuronal addition and endothelial division in HVC. We asked if testosterone-induced angiogenesis might contribute importantly to HVC neuronal recruitment. Testosterone upregulated both VEGF and its endothelial receptor, VEGF-R2/Quek1/KDR, in HVC. This yielded a burst in local HVC angiogenesis. FACS-isolated HVC endothelial cells produced BDNF in a testosterone-dependent manner. In vivo, HVC BDNF rose by the third week after testosterone, lagging by over a week the rise in VEGF and VEGF-R2. In situ hybridization revealed that much of this induced BDNF mRNA was endothelial. In vivo, both angiogenesis and neuronal addition to HVC were substantially diminished by inhibition of VEGF-R2 tyrosine kinase. These findings suggest a causal interaction between testosterone-induced angiogenesis and neurogenesis in the adult forebrain.

Full text

Neuron, Vol. 34, 945–960, June 13, 2002, Copyright 2002 by Cell Press
Coordinated Interaction of Neurogenesis
and Angiogenesis in the Adult Songbird Brain
generated neurons by adult brain parenchyma. Cer-
tainly, brain endothelial cells may be targets of systemic
gonadal steroids (Goldman and Nottebohm, 1983; Hi-
Abner Louissaint, Jr., Sudha Rao,
Caroline Leventhal, and Steven A. Goldman
1
Department of Neurology and Neuroscience
Cornell University Medical Center dalgo et al., 1995), and testosterone in particular ap-
pears to induce endothelial cell division in precisely1300 York Avenue
New York, New York 10021 those regions of the neostriatum into which new neurons
are recruited (Goldman and Nottebohm, 1983). This
seemed unlikely to be coincidental; endothelial cells
(ECs) can serve as potent sources of secreted chemo-Summary
kines and trophic agents. These include FGF2 and IGF1
(Biro et al., 1994), which are potent mitogens for neuralNeurogenesis proceeds throughout life in the higher
vocal center (HVC) of the adult songbird neostriatum. progenitor cells (Drago et al., 1991; Gensburger et al.,
1987), and PDGF and IL-8, which may act as differentia-Testosterone induces neuronal addition and endothe-
lial division in HVC. We asked if testosterone-induced tion and survival factors, respectively, for newly gener-
ated neurons (Araujo and Cotman, 1993; Johe et al.,angiogenesis might contribute importantly to HVC
neuronal recruitment. Testosterone upregulated both 1996). Of particular interest, human brain endothelial
cells can secrete brain-derived neurotrophic factorVEGF and its endothelial receptor, VEGF-R2/Quek1/
KDR, in HVC. This yielded a burst in local HVC angio- (BDNF) (Leventhal et al., 1999) in quantities sufficient
to support neuronal differentiation from the adult ratgenesis. FACS-isolated HVC endothelial cells pro-
duced BDNF in a testosterone-dependent manner. In forebrain ventricular zone (Kirschenbaum and Goldman,
1995; Leventhal et al., 1999). This observation is particu-vivo, HVC BDNF rose by the third week after testoster-
one, lagging by over a week the rise in VEGF and VEGF- larly germane in the adult HVC, in which testosterone’s
neurotrophic effects may be mediated through a re-R2. In situ hybridization revealed that much of this
induced BDNF mRNA was endothelial. In vivo, both gional induction of BDNF, perhaps acting as a down-
stream effector (Rasika et al., 1994, 1999).angiogenesis and neuronal addition to HVC were sub-
stantially diminished by inhibition of VEGF-R2 tyrosine On the basis of these observations, we asked whether
the trophic effects of gonadal steroids in the adult HVCkinase. These findings suggest a causal interaction
between testosterone-induced angiogenesis and neu- might be mediated through the endothelial release of
neurotrophic cytokines, and if so, whether neuronal ad-rogenesis in the adult forebrain.
dition to the adult HVC might depend upon antecedent
endothelial cell activation and division. We report thatIntroduction
testosterone treatment of the adult female canary, which
can trigger primitive song in these otherwise nonsingingThe vocal control nucleus HVC of adult songbirds gener-
ates and recruits new neurons throughout life (Goldman birds, induces the rapid production of both vascular
endothelial growth factor (VEGF) and its receptor,and Nottebohm, 1983; Nottebohm, 1985). The integra-
tion and survival of these neurons is modulated by the VEGFR2/KDR/Quek1 in nucleus HVC. This leads to rapid
endothelial cell division, which in turn predicts the re-gonadal steroids testosterone and estradiol (Hidalgo et
al., 1995; Johnson and Bottjer, 1995; Nordeen and Nor- gionally restricted expansion of the HVC capillary vascu-
lature. The newly activated and expanded vasculaturedeen, 1989; Rasika et al., 1994), whose actions mediate
the seasonal hypertrophy of HVC in adult canaries (Not- substantially increases its production and release of
BDNF, whose induction is both spatially and temporallytebohm, 1981). Notably, estrogen and testosterone are
also associated with accentuated angiogenesis in the associated with the recruitment of new neurons to HVC.
By this means, gonadal steroids may act upon the localadult female canary HVC (Goldman and Nottebohm,
1983; Hidalgo et al., 1995). Testosterone treatment in microvascular bed to provide a permissive environment
for neuronal recruitment into the adult songbird brain.particular yielded a ⬎25-fold increase in the [
3
H]thymi-
dine-determined mitotic index of HVC endothelial cells.
This burst of endothelial cell division was transient and Results
self-limited; endothelial mitotic indices fell to baseline
within 2 weeks after testosterone implantation, despite Testosterone Induces Mitotic Angiogenesis
persistently elevated androgen levels during that period in the Adult Canary HVC
(Goldman and Nottebohm, 1983). Testosterone treatment substantially increases the pro-
We became intrigued about the concurrence of go- liferation rate of HVC endothelial cells (Goldman and
nadal steroid treatment with both neuronal recruitment Nottebohm, 1983). This finding suggested that testoster-
and angiogenesis in the adult HVC and sought to estab- one might directly induce mitotic angiogenesis in the
lish the nature of their coassociation. Specifically, we adult HVC. To test this possibility, we first set about to
asked whether androgen-associated endothelial prolif- confirm that testosterone could indeed induce endothe-
eration might contribute to the acceptance of newly lial cell division in the adult HVC (Figures 1A and 1B). A
cohort of six 1-year-old female canaries was implanted
with either testosterone or cholesterol-containing silas-
1
Correspondence: sgoldm@mail.med.cornell.edu

Neuron
946
Figure 1. Testosterone Increases Mitotic An-
giogenesis and Capillarization in the Adult
HVC
(A and B) Representative high-power images
of BrdU
⫹
(red)/laminin
⫹
endothelial cells
(green) in a testosterone-treated adult female
canary HVC. This bird was killed 7 days after
hormone-silastic implant and immediately
after 2 days of bidaily injections of BrdU. Dou-
ble-labeled cells in the testosterone-treated
HVC are indicated with arrows.
(C) This plot compares the difference in the
HVC endothelial labeling index, defined as the
fraction of BrdU
⫹
/laminin
⫹
cells among all
laminin
⫹
HVC endothelial cells, between tes-
tosterone-treated birds and their cholesterol-
treated controls.
(D) These plots compare the average area
and diameter of laminin
⫹
vessels in the HVCs
of testosterone- and cholesterol-treated fe-
male canaries. Each parameter was signifi-
cantly greater in the testosterone-treated
birds than in their controls.
(E) This graph plots the effect of cholesterol
(C), estradiol (E2), testosterone (T), and dihydrotestosterone (DHT), as well as that of VEGF and BDNF, on [
3
H]thymidine incorporation by
HVC/MCN endothelial cells in vitro. Neither androgens nor estrogen elicited any significant increase in [
3
H]thymidine incorporation by these
cells. In contrast, VEGF addition was associated with a statistically significant (p ⬍0.001), ⬎5-fold increase in net [
3
H]thymidine incorporation
over the 48 hr test period. Thus, VEGF can act directly on HVC endothelial cells to elicit their division, whereas the gonadal steroids are not
directly mitogenic for these cells. Scale ⫽50 ␮m.
tic implants (n ⫽3 each). This treatment has been shown of the HVC microvasculature. Specifically, we asked
whether androgen treatment was accompanied by an
to yield supraphysiological steroid levels within 24 hr
increment in HVC capillary number, cross-sectional area
of implantation (Legan et al., 1975). Beginning on the
and perimeter, or luminal diameter. Two additional
second day thereafter, the birds were injected daily for
groups of adult female canaries (n ⫽3 each) were im-
7 days with bromodeoxyuridine (BrdU; 100 mg/kg intra-
planted with either testosterone or cholesterol-con-
pectoral) and sacrificed on day 10, a day after the last
taining silastic implants and then sacrificed 21 days
BrdU injection. Their brains were cryosectioned at 14
later, and their HVCs were analyzed for these capillary
␮m and double-immunostained for BrdU and laminin,
morphometrics. From each of these brains, a set of ten
which is selectively expressed by endothelial cells (ECs)
coronal 12 ␮m cryostat sections, taken as every tenth
of the HVC capillary vasculature (Alitalo et al., 1982). In
section so as to span the HVC, were cut and stained
each HVC, the incidence of laminin⫹/BrdU⫹cells and
for capillary laminin. The number of capillary profiles as
the ratio of laminin⫹/BrdU⫹cells to total HVC laminin⫹
well as each vessel’s perimeter, cross-sectional area,
ECs were determined bilaterally in each of six sections and luminal diameter were then measured in each HVC.
spaced 200 ␮m apart. For total endothelial counts, indi- We found that both the mean capillary area and perim-
vidual ECs were defined by double-labeling with the eter increased significantly with testosterone. The mean
nuclear dye Hoechst 33258 and laminin. area of each laminin⫹capillary profile rose from 20.9 ⫾
Using this protocol, we found that testosterone- 1.6 ␮m
2
(n ⫽2037 capillary profiles in 42 fields) to 39.2 ⫾
treated birds had an HVC endothelial cell labeling index 4.1 ␮m
2
(n ⫽2085) (p ⫽0.007 by Student’s two-tailed
of 2.34% ⫾0.09% (n ⫽3 birds, at 12 HVC samples/ t test) while the mean perimeter increased from 21.0 ⫾
bird), which reflected 68 BrdU⫹/laminin⫹cells among a 0.5 ␮m to 28.4 ⫾0.8 ␮m(p⫽0.006). Similarly, the
total of 2805 laminin⫹/Hoechst⫹endothelial cells in the average diameter of each vessel was significantly larger
36 scored sections (Figure 1C). This was substantially in the testosterone-treated birds (6.1 ⫾0.5 ␮m) than in
greater than the 0.50% ⫾0.32% mitotic index of the the cholesterol controls (4.5 ⫾0.08 ␮m) (p ⫽0.02) (Figure
cholesterol-treated controls, in which 11 BrdU⫹/lami- 1D). In contrast, the number of vessel profiles/mm
2
did
nin⫹cells were found among a total of 2910 laminin⫹/not differ between the testosterone-treated and control
Hoechst⫹cells. The stimulatory effect of testosterone HVCs (n ⫽48.4 ⫾7.7 and 48.1 ⫾5.1 vessels/field,
on the mitotic index of HVC endothelial cells was highly respectively; p ⫽0.98; each field measured 7.3 ⫻10
4
significant (p ⫽0.0007 by Student’s two-tailed t test), ␮m
2
, so that the respective vessel counts/area were 663
confirming that testosterone stimulated endothelial cell and 659 vessels/mm
2
). Together, these data indicated
division in the adult female HVC. that testosterone induced both endothelial cell division
and expansion of the capillary microvasculature in the
Testosterone Treatment Results in the Expansion adult canary HVC. This occurred without any concomi-
of the HVC Microvasculature tant increase in capillary number, suggesting that tes-
We next asked whether testosterone-associated endo- tosterone induced capillary expansion without vasculo-
genic budding.thelial cell division was accompanied by an expansion

Angiogenesis-Dependent Neurogenesis in Adult Brain
947
VEGF Is Produced by the Testosterone- was stable for at least the 2 weeks thereafter. Thus,
testosterone induced a burst of VEGF mRNA in HVCStimulated HVC
To investigate thehumoral basis fortestosterone-modu- that preceded the onset of androgen-associated mitotic
angiogenesis by 2–3 days.lated angiogenesis in the adult HVC, we next probed
the testosterone-stimulated female canary brain for vas- The rise in VEGF mRNA was attended by a parallel
increase in VEGF protein within HVC. ELISA revealedcular endothelial growth factor (VEGF). VEGF is an endo-
thelial cell mitogen that canbe secreted by most somatic that in adult female canaries given silastic testosterone
implants, HVC VEGF levels rose from 79.8 ⫾15.4 ngcell types (Leung et al., 1989). Androgen-induced VEGF
expression has previously been noted by prostatic fibro- VEGF/mg protein (4.8 ⫾0.2 ng/g tissue) at baseline to
106.0 ⫾10.8 ng/mg protein (8.3 ⫾1.0 ng/g tissue) at 1blasts (Levine et al., 1998) and by murine mammary
carcinoma cells (Jain et al., 1998). In the CNS, VEGF is week after androgen treatment, and to 115.2 ⫾21.2 ng/
mg protein (7.7 ⫾0.7 ng/g tissue) at 3 weeks. Two-waysynthesized and secreted by the developing neuroec-
toderm, which may utilize it to attract and expand capil- ANOVA revealed a highly significant overall effect of
testosterone on HVC VEGF as a function of time (p ⫽lary ingrowth (Breier et al., 1992). In the postnatal brain,
VEGF transcripts have been identified in both astrocytes 0.009, F ⫽5.63 [2, 27 degrees of freedom {d.f.}]) (Figure
2E). No significant changes in VEGF were noted in theand neurons (Ogunshola et al., 2000), by which means
brain VEGF—despite a decline after early develop- parahippocampus, archistriatum, or cerebellum of these
birds (data not shown). In addition, immunocytochemis-ment—is expressed throughout life. Yet despite VEGF’s
persistence in the brain (Robertson et al., 1985), and try revealed strong VEGF staining in the testosterone-
treated HVC 6 days after testosterone treatment begannotwithstanding several reports that its exogenous ad-
ministration can induce CNS angiogenesis (Rosenstein (Figure 2C), whereas matched sections taken from cho-
lesterol-treated control birds showed minimal VEGF-IRet al., 1998; Yancopoulos et al., 2000), the role of endog-
enous VEGF in the regulation of angiogenesis in the (data not shown).
normal adult CNS has never been studied.
To determine if testosterone induced VEGF in the VEGF Synthesis Was Selectively Localized to HVC
adult HVC, we treated six adult female canaries with To localize VEGF-expressing cells in the testosterone-
implants of either testosterone or cholesterol (n ⫽3stimulated HVC, in situ hybridization was used to local-
each) and assessed HVC VEGF mRNA levels thereafter. ize VEGF transcripts in birds given testosterone implants
Whereas only trace levels of VEGF mRNA were detected a week before sacrifice. VEGF cRNA probes were pre-
in cholesterol-treated control HVCs, VEGF mRNA was pared based upon the PCR products generated from
readily demonstrable by both RT-PCR and in situ hybrid- testosterone-treated canary brain mRNA that was re-
ization within 4 days of testosterone implantation (Fig- verse transcribed and subjected to PCR using rat VEGF
ures 2A and 2B). At that time point, even though little primers. The resultant 406 bp canary VEGF fragment
or no VEGF mRNA could be detected in cholesterol- was sequenced and found to correspond to positions
treated HVCs after 36 cycles of PCR, VEGF mRNA was 122–526 of quail VEGF cDNA (96% homology) and to
readily detected in testosterone-treated HVC after 24 positions 166–404 of human VEGF 121/165 cDNA (82%
cycles, and was pronounced by 30. The resultant differ- homology). The canary VEGF partial cDNA was then
ence in VEGF mRNA levels between testosterone-stimu- cloned into pGEM (Promega) and transcribed into an
lated and control HVC was consistent among the three RNA riboprobe using the T7 promoter. Using a P
32
-
pairs of treated birds and their matched controls (Figure labeled probe for in situ hybridization (ISH), we found
2B). Thus, testosterone treatment yielded the rapid in- that whereas the cholesterol-treated control canaries
duction of VEGF mRNA expression in HVC. exhibited only low levels of VEGF mRNA expression
throughout most of the caudal forebrain, their testoster-
one-treated counterparts exhibited significant VEGF
Testosterone-Induced VEGF Synthesis
mRNA in HVC. This androgen-induced VEGF mRNA was
Preceded Angiogenesis
strikingly circumscribed to HVC, with relatively little ex-
We had previously noted that endothelial cell mitotic
pression elsewhere in the caudal forebrain, at least
indices rose significantly within 6 days of testosterone
through the rostrocaudal extent subtended by HVC (Fig-
administration and peaked 9–10 days after the onset
ures 2A and 2C). Control sections, which included tes-
of treatment (Goldman and Nottebohm, 1983). In the
tosterone-treated canary HVC probed with sense RNA
present study, we predicted that the androgen-induced
as well as cholesterol-treated HVC probed with both
surge in VEGF would precede androgen-induced angio-
sense and antisense RNAs, revealed little detectable
genesis and hence be even more rapid in onset. To
forebrain VEGF mRNA by ISH and none in HVC.
define the time course of VEGF induction by testoster-
one, we prepared a set of 15 birds with testosterone
implants on day 0, and we sacrificed 3 birds each on VEGF, but Not the Gonadal Steroids, Was Mitogenic
for HVC Endothelial Cells In Vitrodays 0, 4, 8, 14, and 18. The HVCs were dissected from
each bird, and RT-PCR for VEGF was performed on the The testosterone-associated induction of endothelial
cell proliferation in the adult HVC suggested either thatextracted HVC mRNA. Semiquantitative PCR revealed
that HVC VEGF mRNA rose sharply within 4 days of the gonadal steroids—whether testosterone or its aro-
matized metabolite estradiol—were themselves mito-testosterone treatment (Figure 2D). The normalized ratio
of VEGF to G3PD mRNA was significantly higher on day genic or that they acted to induce endothelial cell divi-
sion through paracrine intermediaries such as VEGF. To4 than at day 0 (p ⬍0.01 by two-way ANOVA), and it

Neuron
948
Figure 2. Testosterone Induces the Produc-
tion of VEGF by the Adult HVC
(A) In situ hybridization using a P
32
-labeled
riboprobe revealed that VEGF mRNA was ap-
parent in HVC within 4 days of testosterone
silastic implantation. Within the area of the
caudal forebrain studied, VEGF mRNA was
sharply circumscribed to HVC.
(B) RT-PCR demonstrated the difference in
VEGF mRNA between HVCs taken from three
cholesterol and three testosterone-treated
birds, all sacrificed 6 days after hormone
implantation. The testosterone-associated
VEGF surge was rapid in onset; HVC VEGF
mRNA rose sharply within 4 days after testos-
terone implantation and peaked by 6–8 days.
(C) Immunolabeling revealed a distinct in-
crease in HVC VEGF immunostaining within
6 days after testosterone treatment; at this
time point, the increased VEGF immunoreac-
tivity was locally restricted to HVC. Scale ⫽
100 ␮m.
(D and E) Testosterone induced significant
increases in HVC VEGF protein as well as
mRNA. (D) This graph plots the normalized
relative level of HVC VEGF mRNA as a func-
tion of time after testosterone implantation.
These values were determined using semi-
quantitative PCR, with relative VEGF mRNA
levels estimated at each time point as the
normalized ratio of VEGF cDNA to G3PDH
cDNA. By day 4, HVC VEGF mRNA was signif-
icantly more abundant than that at baseline
(p ⬍0.01; see text). (E) This graph shows the
parallel effect of testosterone on HVC VEGF
protein, as measured by ELISA. The amount
of VEGF in HVC, paralleling that of its mRNA,
was significantly higher at 1 week than at
baseline (p ⬍0.01).
assess whether the gonadal steroids might act directly VEGF, the receptor tyrosine kinases VEGFR1/flt1 and
VEGFR2/flk1/KDR, are both expressed minimally in un-
as mitogens or whether they required paracrine interme-
stimulated endothelium. We therefore asked whether
diaries, we assessed their effects in cultures of canary
HVC endothelial cells express VEGF receptors and
neostriatal endothelia. These highly enriched endo-
whether VEGF receptor expression by these cells might
thelial cell cultures were prepared from the HVC and
be influenced by gonadal steroids. To this end, we as-
adjacent mediocaudal neostriata of adult canaries by
sessed the expression of the principal known endothe-
fluorescence-activated cell sorting (FACS) of DiI-LDL-
lial cell receptor for VEGF, VEGF-R2/Quek1, in hormone-
tagged dissociates of adult canary HVC (Leventhal et
treated female canaries in vivo as well as in isolated
al., 1999). The resultant endothelial cultures were raised
HVC endothelial cells in vitro.
for 2 days in media containing 1% steroid-depleted FBS,
Quek1 has been cloned from quail as the avian homo-
and then they were exposed to testosterone (100 ng/ml),
log of mouse flk-1 and human VEGF-R2/KDR, a VEGF
dihydrotestosterone (100 ng/ml), estradiol (10 ng/ml),
receptor tyrosine kinase that is one of the two known
cholesterol (100 ng/ml, as a negative control), BDNF (20
high-affinity endothelial receptors for VEGF, the other
ng/ml), or VEGF (20 ng/ml, as a positive control), all in
being flt-1 (VEGF-R1) (Wilting et al., 1997). Lower affinity
the presence of 3.5 ␮Ci/ml [
3
H]thymidine. The cultures receptors for VEGF include the neuropilins, coreceptors
were then extracted and net [
3
H]thymidine incorporation for VEGF that are also expressed by neurons as their
counted by scintillation counter. We found that whereas receptors for members of the semaphorin family of guid-
VEGF was a strong mitogen for these cells, none of the ance molecules (Soker et al., 1998). Whereas VEGF-R1/
gonadal steroids were (Figure 1D). Thus, whereas VEGF flt1 and the neuropilins may be expressed by many cell
directly induces HVC endothelial cells to divide, the go- types, VEGF-R2 is expressed by ECs, monocytes, and
nadal steroids are not directly mitogenic for these cells. hematopoetic stem cells (Millauer et al., 1993); impor-
tantly, it may also be expressed by some neural popula-
HVC Endothelia Upregulated VEGFR2/Quek1/ tions in development (Yang and Cepko, 1996, Ogun-
KDR in Response to Gonadal Steroids shola et al., 2002). Given its restricted expression, we
We next considered the possibility that gonadal steroids asked whether VEGF-R2/Quek1 was expressed by HVC
influenced not only VEGF production in HVC but also ECs, and if so, whether its expression—like that of its
ligand VEGF—might be modulated by testosterone.VEGF receptivity. The principal endothelial receptors for

Angiogenesis-Dependent Neurogenesis in Adult Brain
949
terone may be aromatized to estradiol in the adult HVC,
its actions may be mediated through either androgen
or estrogen receptor systems. To distinguish which of
these might be necessary for testosterone-induced
VEGF-R2/Quek1 expression, we compared the effects
of testosterone, estradiol-17␤, and the nonnaromatiza-
ble androgen 5-␣-dihydrotestosterone (DHT) on Quek1
expression by isolated HVC endothelial cells in vitro. We
found that Quek1 mRNA was induced in cultured canary
neostriatal endothelial cells within 2 days of testoster-
one exposure; these cells expressed no detectable
Quek1 mRNA in the absence of gonadal steroid treat-
ment. This effect appeared specifically mediated by es-
trogen, in that estradiol-17␤induced endothelial Quek1/
VEGFR2 whereas the purely androgenic 5␣-DHT did not
(Figure 3C).
We have previously noted that estradiol, like testoster-
one, is associated with endothelial cell division in the
adult canary HVC, although the endothelial mitotic index
achieved is less than that yielded by androgen treatment
(Hidalgo et al., 1995). On this basis, we asked whether
estrogen might induce Quek1 in vivo, just as in vitro,
Figure 3. HVC Endothelial Cells Upregulate the VEGF-R2 Receptor,
thereby allowing ambient VEGF to activate local endo-
Quek1, in Response to Estrogen
thelial cells. We treated six adult females with silastic
(A) RT-PCR of mRNA isolated from adult female canary HVC (three
implants containing either estradiol-17␤or cholesterol
birds pooled/time point) revealed that VEGF-R2/Quek1 mRNA rose
(n ⫽3 birds each) and sacrificed them at various time
sharply within 4 days of testosterone silastic implantation. Testos-
points thereafter, subjecting their dissected HVCs to
terone (T)0, T4, and T14 refer to the 0, 4, and 14 day timepoints
RT-PCR for the Quek1/VEGFR2 receptor as well as for
after testosterone treatment.
VEGF itself. We found that estradiol induced HVC
(B) Treatment with estradiol yielded even more pronounced in-
creases in Quek1 mRNA in the adult female canary HVC. This gel
Quek1/VEGFR2 within 4 days of treatment (Figure 3B).
shows the RT-PCR products of mRNA derived from three pooled
Furthermore, VEGF mRNA also rose in response to es-
female HVCs 4 days after estradiol silastic implantation (right lanes)
tradiol, roughly following the same time course as
compared to cholesterol-treated control HVC (left). A small increase
Quek1/VEGFR2 (data not shown). The importance of
in VEGF signal was also noted in response to estradiol.
estradiol in the induction of VEGF is also suggested by
(C) Endothelial cells isolated from the mediocaudal neostriatum,
the finding that the VEGF promoter harbors an estrogen
which included HVC and the adjacent medial striatal wall, expressed
Quek1 mRNA in response to estrogen stimulation in vitro. From left
response element that binds estradiol-estrogen recep-
to right: reverse-transcribed Quek1 PCR products were not evident
tor complexes and is able to direct VEGF transcription
in endothelial cultures to which cholesterol (C) was added, but
(Mueller et al., 2000). Together, these observations sug-
Quek1 signals were noted in both estradiol (E)- and testosterone (T)-
gest that estrogen may be a critical intermediary in the
supplemented media. No signal was evident in media supplemented
stimulation of HVC angiogenesis by testosterone.
with the nonaromatizable androgen 5␣-dihydrotestosterone (DHT),
suggesting that endothelial Quek1 was selectively induced by es-
trogen.
The Adult Female HVC Produces BDNF mRNA
and Protein in Response to Testosterone
BDNF expression in the canary HVC is sexually dimor-To amplify canary Quek1/VEGF-R2, we used quail-
based primers to obtain a 405 bp fragment of canary phic, being higher in the adult male than the female, and
testosterone treatment has been shown to induce BDNFQuek1. This corresponded to bases 2537–2941 of the
quail Quek1 gene, with 89% homology to canary, and in the adult female HVC (Rasika et al., 1999). To better
understand the role of androgen-induced BDNF in HVCto bases 2519–2823 of human VEGFR2/KDR, to which
canary Quek1 proved 83% homologous. We then as- neuronal recruitment, as well as its potential relationship
to androgen-associated angiogenesis, we sought tosessed the effect of testosterone on Quek1/VEGFR2
mRNA by comparing Quek1 mRNA in testosterone- define the time course of testosterone-induced BDNF
expression. Using RT-PCR, in situ hybridization, andtreated and control HVC (Figure 3A). Using RT-PCR, we
found that Quek1 mRNA was scarcely detectable, if at ELISA, we found that HVC BDNF mRNA and protein both
rose significantly in response to testosterone, thoughall, in either the unstimulated or cholesterol-implanted
adult female canary HVC. In contrast, within 4 days of according to a notably delayed time course.
By PCR and ISH, BDNF mRNA levels in HVC appearedtestosterone treatment, HVC Quek1 levels rose substan-
tially and remained high through day 14 (Figure 3A). to rise only after a delay of at least 2 weeks following
initial testosterone treatment. Semiquantitative RT-PCRQuek1 signal intensity slowly fell thereafter (data not
shown), with a return to baseline level by 28 days. normalized against G3PD mRNA revealed no detectable
elevation in BDNF mRNA at 4 or 8 days after testosterone
treatment, but did detect a substantial increase in BDNFBoth Quek1 and VEGF Are Induced by Estradiol
The induction of endothelial VEGFR2/Quek1 by testos- signal by 14 and 18 days (Figure 4A). ELISA revealed
that BDNF protein levels remained roughly stable for atterone was dramatic and rapid. However, since testos-

Neuron
950
Figure 4. Testosterone Induces the Produc-
tion of BDNF in Both HVC and Its Endothelial
Cells
(A) RT-PCR revealed a sharp but delayed rise
in HVC BDNF mRNA after testosterone treat-
ment. This rise in BDNF mRNA was first mani-
fest, and achieved statistical significance
(see text), 14 days after silastic implantation.
(B) ELISA similarly revealed that the concen-
tration of HVC BDNF protein changed little
in the first week after testosterone, but rose
steadily thereafter. The delayed rise in BDNF
contrasted with the rapid elevation in HVC
VEGF (superimposed from Figure 2), which
achieved its maximum within a week of an-
drogen treatment. Like the level of BDNF
mRNA, the concentration of BDNF protein at
3 weeks was significantly higher than that
measured either at baseline or at 1 week; the
latter two values did not statistically differ
(see text).
(C–F) In situ hybridization for BDNF cRNA in
canaries treated with testosterone 18 days
prior to sacrifice compared to cholesterol-
treated controls. (C) A control section probed
with BDNF sense cRNA; (D) HVC of choles-
terol-treated controls revealed a low level of
baseline BDNF mRNA expression, which was
predominantly neuronal. (E and F) Sections
probed with BDNF antisense cRNA revealed
a marked increase in BDNF signal in the tes-
tosterone-treated HVC at 18 days. (F) A higher
magnification view shows that much of the
increment in BDNF mRNA was perivascular
and associated with capillary endothelial
cells (arrows). Scale ⫽30 ␮m.
least the first week after testosterone treatment, mea- al., 2000). However, the testosterone-treated birds were
also specifically distinguished by prominent BDNFsuring 40.3 ⫾9.4 ␮g BDNF/g protein (2.62 ⫾0.7 ng/g
tissue) on the day of testosterone implantation, and mRNA expression by HVC microvascular capillary cells.
In contrast, the control birds exhibited little or no such41.7 ⫾4.0 ␮g/g (3.3 ⫾0.6 ng/g tissue) at 7 days. Thereaf-
ter, the concentration of HVC BDNF rose to achieve a perivascular BDNF mRNA. Thus, testosterone treatment
was associated with a marked, but relatively delayedlevel of 70.2 ⫾23.2 ␮g/g (4.7 ⫾0.8 ng/g tissue) by 3
weeks after testosterone treatment. Thus, within 3 upregulation of BDNF in HVC. Importantly, the capillary
microvasculature contributed substantially to that in-weeks after testosterone implantation, HVC BDNF levels
rose by 79% (when expressed in ␮g/g protein, or by crement.
74% when expressed in ng/g tissue) (Figure 4B). The
relatively delayed accumulation of BDNF in the testos- Purified Cultures of HVC Endothelial Cells
Synthesize and Secrete BDNFterone-treated HVC contrasted sharply with the rapid
elevation in HVC VEGF and VEGF-R2/Quek1, each of We have previously noted that capillary endothelial cells
secrete BDNF protein and that they may do so in suffi-which rose significantly in the first week after androgen
treatment (Figures 2 and 3). cient amounts to support the migration and survival
of neurons arising from the adult rat ventricular zoneISH revealed that BDNF cDNA exhibited a substantial
elevation in signal intensity in the HVCs of testosterone- (Leventhal et al., 1999). To assess whether androgen-
induced angiogenesis might contribute to neuronal re-treated birds. The elevation in BDNF’s hybridization sig-
nal was first noted 14 days after androgen administra- cruitment in HVC through endothelial-derived BDNF, we
investigated the release of neurotrophins by microvas-tion. By 17 days after testosterone treatment—a time
point chosen on the basis of the PCR data—ISH revealed cular ECs derived from the adult HVC. These ECs were
harvested by FACS of DiI-LDL-tagged dissociates ofa marked increase in BDNF mRNA expression in HVC
in the testosterone-treated animals (Figure 4C). The cel- adult canary HVC and adjacent mediocaudal neostria-
tum, as previously described (Leventhal et al., 1999).lular basis for this increased BDNF mRNA signal proved
surprising; HVC BDNF mRNA was associated with both When raised in base media supplemented only with 2%
steroid-depleted FBS and hydrocortisone, the cells ex-morphologically-evident neurons and satellite astro-
cytes, as previously reported (Dittrich et al., 1999; Li et pressed measurable levels of both BDNF mRNA and

Angiogenesis-Dependent Neurogenesis in Adult Brain
951
Figure 6. BDNF Promotes Neuronal Outgrowth from the Cultured
Adult HVC VZ
This graph plots the number of neurons observed in the outgrowths
from adult HVC VZ explants, either BDNF-supplemented or not, as
a function of time in vitro. BDNF (20 ng/ml) was noted to significantly
increase both the numbers of neurons departing these explants and
their relative abundance thereafter through the first 24 days in vitro
(p ⬍0.001; see text).
Figure 5. Cultured HVC Endothelial Cells Exhibit Gonadal Steroid-
Dependent BDNF Production
(A) RT-PCR revealed that HVC/MCN endothelial cells, harvested by
terol-treated controls, which exhibited scarcely detectable
FACS of DiI-LDL-tagged dissociates of the adult canary mediocau-
BDNF message, and with estradiol- and 5␣-DHT-exposed
dal neostriatum, synthesized BDNF; this process was stimulated by
cultures, which showed lesser though consistent eleva-
testosterone, and less so by estradiol and DHT.
(B) ELISA confirmed that BDNF was secreted by the endothelial
tions in BDNF mRNA. Similarly, ELISA of endothelial
cells and that this too was strongly promoted by testosterone, the
culture supernatants revealed that the concentration of
addition of which more than tripled the rate of BDNF release from
BDNF protein in testosterone-treated cultures, 69.8 ⫾
these adult neostriatal endothelial cells. Asterisks denote statisti-
27 pg/ml (733 ⫾173 ng/g protein), was significantly
cally higher levels of BDNF release than those of cholesterol-treated
higher than in matched cholesterol-treated controls,
controls (p ⬍0.05; see text).
which had 23.8 ⫾4.4 pg BDNF/ml (258.3 ⫾62.2 ng/g)
(p ⫽0.025; F ⫽8.84 [1, 6 d.f.]). Both estradiol- and
5␣-DHT-exposed cultures exhibited lesser elevations in
protein (Figure 5). ELISA revealed that after 2 days in
BDNF protein; estradiol-treated endothelial cultures
vitro, these HVC ECs released a baseline level of 258 ⫾
harbored 46.5 ⫾9.4 pg BDNF/ml (546 ⫾169 ng/g), and
62 ng BDNF/g into the culture supernatant. This level
DHT-treated plates had 41.4 ⫾8.4 pg/ml (576 ⫾349
of endothelial BDNF secretion approximated that which
ng/g) (Figure 5B). Together, these findings indicated that
we previously noted in unstimulated cultures of human
gonadal steroids directly induced BDNF transcription
brain microvascular endothelial cells (Leventhal et al.,
and release by neostriatal endothelial cells.
1999). These observations suggested that the capillary
vasculature of the adult songbird HVC can serve as a
local secretory source of BDNF. BDNF Promotes the Migration and Recruitment
of Neurons from the HVC Ventricular Zone
In VitroTestosterone and Estrogen Induce BDNF Production
by Cultured HVC Endothelial Cells To establish whether the appearance of neurons arising
from the adult HVC ventricular zone (VZ) was directlyOn the basis of the twin observations that gonadal ste-
roids can induce BDNF production and that HVC ECs supported by BDNF, we assessed the outgrowth of neu-
rons from explants of adult canary HVC. To this end,might themselves serve as important sources of BDNF,
we next asked whether neostriatal capillary ECs modify five adult female canaries were sacrificed for HVC VZ
explant culture, as described (Goldman et al., 1992).their levels of BDNF secretion in response to gonadal
steroid activation. To this end, plates of FACS-purified These were raised in either base medium (DMEM/F12/
N2 with 10% platelet-depleted FBS) or the same supple-neostriatal endothelial cells were raised in serum-free
media supplemented with VEGF alone for 2 days, and mented with 20 ng/ml BDNF. Neurons arising from the
explants were scored using established morphologicalthen were exposed to estradiol, testosterone, 5␣-DHT,
or cholesterol for 48 hr ending at the end of day 4 in criteria (Goldman, 1990; Kirschenbaum and Goldman,
1995) and counted twice weekly for 3 weeks thereaftervitro. At that point, matched cultures were extracted for
either protein for BDNF ELISA or mRNA for semiquanti- (seven timepoints). We found thatthe numberof neurons
arising from these explants was significantly higher intative PCR.
We found that testosterone predictably elevated the BDNF-supplemented plates than in unsupplemented
controls (p ⬍0.001; F ⫽31.5 [1, 6 d.f.]) (Figure 6). BDNF’slevel of BDNF mRNA expressed by these endothelial
layers (Figure 5A). This contrasted with matched choles- support of neuronal differentiation and migration from

Neuron
952
these adult HVC explants was compatible with its role ment to HVC. Sixteen 1-year-old female canaries were
injected twice daily with bromodeoxyuridine (BrdU; 50in promoting the recruitment of new neuronal migrants
in vivo (Rasika et al., 1999; Zigova et al., 1998). mg/kg intrapectoral q12 hr) for 7 days beginning with
day 1 in order to label newly generated cells. On day 8,
they were all implanted with either testosterone-con-
VEGF Signaling Is Necessary for Testosterone-
taining or empty (control) silastic implants. Beginning
Induced Angiogenesis
on day 8, the canaries were also injected twice daily
On the basis of the temporal and spatial associations
with either VEGFR-TKI (2.5 mg/kg, or typically, 50 ␮g/
between BDNF production by the testosterone-stimu-
20 g bird) or a vehicle control (DMSO:PBS, 2:1). In this
lated HVC microvasculature and neuronal recruitment,
way, the 16 canaries were grouped into four treatment
we asked whether endothelial cell proliferation was nec-
categories (n ⫽4 birds/group): (1) no hormone/vehicle; (2)
essary for testosterone-dependent neuronal recruit-
no hormone/VEGFR-TKI; (3) testosterone/vehicle; and (4)
ment. To this end, we first assessed the effect of inhib-
testosterone/VEGFR-TKI. Treatment with VEGFR-TKI was
iting VEGF-R2 signaling upon testosterone-induced
continued from days 8–33, and the canaries were sacri-
HVC endothelial cell division, using a small molecule
ficed on day 40. Their brains were removed, sectioned
inhibitor of the VEGF-R2/KDR tyrosine kinase, 4-[(4⬘-
at 14 ␮m, and double-stained for BrdU and Hu, a neuron-
chloro-2⬘-fluoro)phenylamino]-6,7-dimethoxy-quinazo-
specific RNA binding protein that is uniformly and selec-
line (Calbiochem 676475). We chose this agent, which
tively expressed by neurons within the adult HVC (Bar-
we will designate VEGFR-TKI (VEGF receptor tyrosine
ami et al., 1995). For each of the 16 birds, the ratio of
kinase inhibitor), for several reasons. First, it is highly
Hu⫹/BrdU⫹cells to the total number of Hu⫹neurons
potent (IC
50
⫽0.1 ␮M for VEGF-R2/KDR/flk1/Quek1 and
within each cresyl violet-defined HVC was determined
2␮M for VEGF-R1/flt1), yet is more specific for VEGF-
bilaterally in each of five sections spaced 100 ␮m apart.
R2 than other currently available inhibitors of the KDR
These were chosen to span the stereotaxic coordinates
tyrosine kinase, with ⬍1% cross-inhibition of the PDGF,
AP0.0-P0.5 (Stokes et al., 1976), a region that includes
HGF, or FGF2 receptors (Hennequin et al., 1999). Sec-
most of the volume of the adult HVC.
ond, VEGFR-TKI is small, has a low hydrogen bond num-
Using this protocol, we found that the mitotic index
ber, and is highly lipophilic, permitting ready penetration
of Hu⫹HVC neurons in testosterone-treated birds not
of the blood-brain barrier (Pardridge, 1995). Third, in
injected with inhibitor was 1.91% ⫾0.1%. This included
both experimental and clinical trials, its similar but less
647 BrdU⫹/Hu⫹cells among a total of 33,766 Hu⫹neu-
blood-brain barrier-permeable analog ZD4190 has
rons in 40 scored sections. This incidence was 3.4-fold
shown little or no systemic toxicity at doses an order
greater than the 0.56% ⫾0.1% neuronal labeling index
of magnitude higher than those employed here (Wedge
(188 BrdU⫹/Hu⫹cells among 33,666 Hu⫹cells) found in
et al., 2000).
control birds given no inhibitor and an empty silastic
To determine the effect of inhibiting VEGF signaling
(Figure 7). This effect of testosterone on the neuronal
on HVC angiogenesis, we thus injected six testosterone-
labeling index was highly significant when assessed by
treated canaries with either VEGFR-TKI (2.5 mg/kg, or
two-way ANOVA (F ⫽382.9 [1, 1, 15 d.f.]; p ⬍0.0001),
typically, 50 ␮g/20 g bird) or vehicle (DMSO:PBS, 2:1)
confirming that testosterone promotes the recruitment
for 9 days, starting on the day of steroid implantation
of new neurons to the HVCs of adult female canaries.
(n ⫽3 each). Throughout the experiment, the birds were
However, when a matched cohort of testosterone-
also injected twice daily with bromodeoxyuridine (BrdU;
treated females was injected with VEGFR-TKI, the tes-
50 mg/kg intrapectoral q12 hr) to label newly generated
tosterone-induced increase in the Hu⫹neuronal labeling
cells. The canaries were sacrificed on day 10, and their
index was suppressed by 46%, to 1.04% ⫾0.1% (337
brains were removed, cryosectioned, and double-immu-
BrdU⫹/Hu⫹cells among a total of 32,451 neurons). When
nolabeled for BrdU and laminin. For each HVC, the inci-
assessed by ANOVA, this reduction in the labeling index
dence of laminin⫹/BrdU⫹cells and the ratio of laminin⫹/
was highly significant (F ⫽163.9; p ⬍0.01 after post
BrdU⫹cells to total HVC laminin⫹ECs were determined
hoc Boneferroni correction). The marked suppression
bilaterally in each of five sections spaced 100 ␮m apart.
of the Hu⫹/BrdU⫹neuronal labeling index by the VEGFR
For total endothelial counts, individual ECs were defined
tyrosine kinase inhibitor demonstrated that VEGF signal-
by double-labeling with the nuclear dye DAPI and lami-
ing was indeed required for testosterone-dependent
nin. Using this protocol, we found that testosterone-
neuronal recruitment to HVC.
treated birds injected with vehicle control had an HVC
To determine whether VEGFR-TKI influenced baseline
endothelial labeling index of 4.89% ⫾0.6% (221 BrdU⫹/
levels of neuronal recruitment, canaries not treated with
laminin⫹cells among a total of 4536 laminin⫹/DAPI⫹
testosterone were also injected with the inhibitor. The
cells). Strikingly, the endothelial labeling index of testos-
resulting labeling index of Hu⫹neurons in these inhibi-
terone-treated birds injected with VEGFR-TKI fell by
tor-treated hormonal nulls was 0.51% ⫾0.02%, not sig-
over 60%, to 1.94% ⫾0.4%, (66 BrdU⫹/laminin⫹cells
nificantly different than the 0.56 ⫾0.01% labeling index
among 3403 laminin⫹/DAPI⫹cells). Thus, pharmacologi-
found in birds given empty silastics and no inhibitor (F ⫽
cal inhibition of VEGF signaling substantially reduced
0.37, p ⬎0.05) (Figure 7).
testosterone-mediated HVC angiogenesis (p ⫽0.002 by
one-way ANOVA; F ⫽47.9 [1, 4 d.f.]).
Discussion
VEGF Signaling Is Necessary for Testosterone-
Induced Neuronal Recruitment In this study, we asked whether testosterone-induced
endothelial cell proliferation in the adult canary vocalWe next sought to determine if the inhibition of HVC
endothelial proliferation would reduce neuronal recruit- control nucleus HVC might be causally related to the

Angiogenesis-Dependent Neurogenesis in Adult Brain
953
Figure 7. Inhibition of VEGF Signaling Suppresses Testosterone-Induced Neuronal Recruitment
(A) This section shows the borders of a cresyl violet-stained HVC 1 month after administration of a testosterone silastic. Arrowheads identify
the border of HVC, delimiting the region scored.
(B) This plot compares the number of newly recruited BrdU
⫹
/Hu
⫹
neurons, expressed as a percentage of all Hu
⫹
HVC neurons, between birds
in four treatment groups. These include both testosterone-treated (T) and untreated (NS: null silastic) birds given either a VEGF-R2/Quek1/
KDR tyrosine kinase inhibitor (VEGF-TKI) or a DMSO vehicle control (vehicle). VEGF TK inhibition was associated with an almost 50% reduction
in the number of BrdU
⫹
/Hu
⫹
neurons recruited to HVC over the month after androgen administration.
(C–F) Examples of BrdU
⫹
/Hu
⫹
neurons in each of these groups: (C) an empty silastic null control subsequently injected only with vehicle (NS/
vehicle); (D) an empty silastic NS control given the VEGF tyrosine kinase inhibitor (NS/TKI); (E) a testosterone-treated bird otherwise given
only vehicle, as a positive control (T/vehicle); (F) a testosterone-treated bird given VEGFR TKI (T/TKI).
(G–I) Confocal validation of the double-labeling of a BrdU
⫹
(green)/Hu
⫹
(red) neuron noted in the HVC of a testosterone-implanted adult female
1 month after hormone administration. (G) A composite z stack of a series of six confocal images taken 0.5 ␮m apart (H), and (I) as viewed
orthogonally in the xz and yz planes. Scale: (A), 100 ␮m; (C–I), 10 ␮m.
induction and success of adult neurogenesis therein. ECs could be modulated by gonadal steroids. Steroid-
induced endothelial BDNF production appeared largelyWe found the following. (1) Testosterone treatment re-
sulted in the production of vascular endothelial growth androgenic, in that both testosterone and DHT yielded
substantial increments in endothelial BDNF mRNA andfactor, VEGF, within the adult HVC. (2) The resultant
VEGF surge was spatially localized to HVC, within which protein. (7) BDNF supported the emigration of neurons
from explants of the adult HVC ventricular zone, sug-it was temporally associated with the proliferation of
capillary endothelial cells. This led to an expansion of gesting that BDNF directly promotes neuronal recruit-
ment into the HVC parenchyma. (8) Both testosterone-the microvasculature within, but not beyond, HVC. (3)
ECs purified from the adult canary HVC responded to induced angiogenesis and neuronal addition to the adult
HVC could be substantially diminished by inhibition oftestosterone and estradiol by upregulating their expres-
sion of the VEGF receptor, VEGF-R2/Quek1. This in turn VEGF tyrosine kinase activity. Together, these findings
suggest a causal interaction of testosterone-inducedenabled the cells to proliferate in response to VEGF. (4)
Both BDNF mRNA and protein levels rose substantially angiogenesis and neurogenesis in the adult songbird
brain (Figure 8).in HVC after testosterone treatment, with a time course
that followed, by over a week, testosterone’s induction
of VEGF and its receptor. (5) HVC ECs synthesized and The Source and Role of Androgen-Stimulated VEGF
The induction of VEGF appeared to be an early andsecreted the neurotrophin BDNF, suggesting that the
VEGF-expanded vasculature increased BDNF availabil- cardinal event in the sequence of events leading to an-
drogen-associated neuronal recruitment to HVC. Cer-ity within the HVC. (6) The production of BDNF by HVC

Neuron
954
Figure 8. Our Model: An Orchestrated Series
of Paracrine Interactions Permits Neuronal
Recruitment
This schematic summarizes our understand-
ing of the means by which the gonadal ste-
roids may influence neuronal recruitment to
the adult HVC. The boxed-in area surrounding
the capillary at the lower left highlights the
role of testosterone-associated angiogenesis
in this process. These pathways are pre-
sented in the context of complementary re-
cent findings pertaining to the cellular basis
for neuronal recruitment in the adult HVC
(Barami et al., 1994, 1995; Bottjer and John-
son, 1997; Holzenberger et al., 1997; Jiang
et al., 1998; Goldman, 1998; Goldman and
Luskin, 1998).
Abbreviations: Ncad, N-cadherin; AR, andro-
gen receptor; ER, estrogen receptor-␣; IGF-1,
insulin-like growth factor-1; Hu, Hu protein;
3A7, a vimentin-associated antigen; VEGF,
vascular endothelial growth factor; Quek1,
the VEGF receptor VEGF-R2/Quek1/KDR;
BDNF, brain-derived neurotrophic factor.
tainly, the testosterone-associated increment in HVC VEGF May Act to Induce Local Vascular Permeability
VEGF protein from 79.8 ⫾15.4 to 115.2 ⫾21.2 ␮g/g, VEGF mobilized from the HVC interstitial matrix, as well
occurring concurrently with a sharp upregulation in as that produced locally thereafter, might mediate not
VEGF mRNA as measured by both PCR and ISH, argued only angiogenesis, but also local increases in local vas-
that androgen-induced angiogenesis was based upon cular permeability (Nag et al., 1997; Proescholdt et al.,
the antecedent induction of VEGF synthesis in HVC. Yet 1999; Yancopoulos et al., 2000; Zhang et al., 2000).
it may be instructive to note that even though no such Among other effects, such focal vasogenic edema might
VEGF mRNA was detected within the unstimulated HVC, increase the access of systemically bound agents to the
either by RT-PCR or ISH, the unstimulated HVC nonethe- testosterone-treated HVC. As such, one cannot rule out
less still harbored an average of almost 80 ␮g VEGF/g the possibility that markers of DNA replication and mito-
protein. This suggested that some VEGF protein may sis, such as bromodeoxyuridine and labeled thymidine,
reside within HVC as a relatively stable pool, perhaps might enjoy preferential access to HVC in testosterone-
matrix bound. This is of interest given recent observa- stimulated birds. To be sure, we have found no evidence
tions that proteolytic matrix metalloproteinases (MMPs), for such effects in these studies, and VEGF clearly in-
and MMP-9 in particular, may activate angiogenesis by duced endothelial cell mitotic labeling in vitro as well
releasing resident VEGF from stores bound within the as in vivo. In addition, BrdU in particular has wide avail-
interstitial matrix (Bergers et al., 2000). This suggested ability to normal brain tissue following systemic admin-
the possibility of a similar, androgen-evoked proteolytic istration and does not require tissue disruption for
liberation of VEGF from tissue stores sequestered within cerebral parenchymal access. Nonetheless, the andro-
the quiescent adult HVC. Together, these observations
gen-induced vascular permeability that may be associ-
suggest that while androgen-induced HVC angiogenesis
ated with angiogenic regions as such as HVC might serve
is causally preceded by the local synthesis of VEGF,
as an important example of hormonal regulation of
other, transcriptionally independent mechanisms, such
blood-brain barrier permeability in the adult brain. In-
as the androgen-elicited release of tissue-bound VEGF,
may also cooperate to induce angiogenesis. deed, future studies may prove the HVC a unique model

Angiogenesis-Dependent Neurogenesis in Adult Brain
955
for reversible hormonal disruption of the normal adult testosterone induced an increase in endothelial VEGF-
blood-brain barrier. R2 mRNA within 48 hr. Importantly, dihydrotestosterone
elicited no such induction of VEGF-R2/Quek1, even
Potential Collaborative Angiogenic Agents though DHT did elicit BDNF production in matched cul-
It is important to note that whereas VEGF mRNA and tures of neostriatal endothelial cells. Thus, the induction
protein both remained elevated after testosterone silas- of Quek1 receptor by gonadal steroids appeared to be
tic implantation, mitotic angiogenesis in the testoster- mediated by estrogen rather than androgen receptor-
one-treated adult HVC is relatively short-lived. The en- dependent pathways. A role for this pathway is also
dothelial mitotic indices of the androgen-stimulated suggested by the high level of aromatase, which con-
HVC fall within 2 weeks of the initial burst of hormone- verts testosterone to estradiol, in the mediocaudal neo-
induced angiogenesis (Goldman and Nottebohm, 1983). striatum of adult songbirds (Balthazart et al., 1996; Shen
Thus, androgen-associated angiogenesis appears to et al., 1995) and by the regulation of aromatase expres-
follow a parabolic time course, despite the persistent sion by testosterone (Fusani et al., 2001). A direct estro-
elevation of HVC VEGF. Rather than representing local genic induction of endothelial VEGF-R2/Quek1 is partic-
endothelial tachyphlaxis to VEGF, it is likely that VEGF’s ularly attractive in light of the widespread vascular
role changes during persistent endothelial activation expression of both the ER␣and ER␤families of estrogen
from inducing mitogenesis to supporting capillary main- receptors (Iafrati et al., 1997; Lindner et al., 1998; Mos-
tenance (Hanahan, 1997). This may be a function of other selman et al., 1996). ER␣is expressed heavily in HVC
angiogenic factors released in the androgen-stimulated (Bernard et al., 1999; Gahr, 1990) and may modulate
tissue (Suri et al., 1996). The latter possibility is sug- both the endothelial expression of, and response to,
gested by the observation that VEGF’s actions may shift humoral cytokines (Caulin-Glaser et al., 1996). ER␤is
from inducing endothelial division to supporting vascu- expressed by both vascular endothelial and smooth
lar modeling once the new endothelial cells are stimu- muscle cells in mammals (Andersson et al., 2001) and
lated by angiopoetin-1 (Suri et al., 1998; Thurston et al., is a potential mediator of the estrogenic induction of
1999). The latter is serially activated during angiogenesis Quek1, although recent studies have not identified ER␤
and may induce endothelial cells to recruit smooth mus- in HVC (Bernard et al., 1999). Pending further study of
cle pericytes to the nascent capillary wall (Yancopoulos the phenotypic specificities and regulation of these re-
et al., 2000). Thus, VEGF’s actions ma yvar y as a function of ceptors in the canary HVC, each remains a candidate
the local coexpression of other angiogenic growth factors to mediate the effects of gonadal steroids on endothelial
released in the androgen-stimulated HVC, whether in par- production of neuronal cytokines. Such vascular estro-
allel with or downstream to VEGF. gen receptors might thereby prove important to the gen-
esis of sexually dimorphic neural architecture and be-
Sequential Activation of VEGF-R2/Quek1, VEGF, haviors alike.
and BDNF in the Testosterone-Stimulated HVC The induction of endothelial KDR/Quek1 by gonadal
Our results suggest that gonadal steroid-modulated an- steroids appears to be rate limiting for VEGF’s activation
giogenesis in HVC depends not only upon the release of ECs in the adult HVC. This process appears likely to
of VEGF by androgen-receptive cells in the region, but be triggered by circulating testosterone acting through
also on the concurrent induction of the endothelial re- an estrogenic intermediary, presumably after its in situ
ceptor for VEGF. Previous studies have found that both aromatization to estradiol. Together, these results indi-
neurons and astrocytes express androgen receptor cate that testosterone stimulates angiogenesis in the
mRNA (Nastiuk and Clayton, 1995); either of these cell adult HVC by concurrently inducing the expression of
types might be potential sources of testosterone-asso- VEGF and its principal receptor VEGF-R2/Quek1 in neu-
ciated VEGF. In HVC, the transcriptional activation and ral and endothelial cells, respectively. This in turn sug-
synthesis of VEGF after testosterone exposure was gests a tight coordination of androgenic and estrogenic
swift; within 4 days of silastic implantation, HVC VEGF actions in mediating steroid-induced angiogenesis in
rose to levels readily detectable by RT-PCR, immuno- the adult songbird brain.
labeling, and ELISA. Moreover, testosterone’s induction
of VEGF’s endothelial receptor, VEGF-R2/Quek1, was BDNF and the Direction of Neuronal Recruitment
similarly rapid and even more marked in degree; within to the Testosterone-Stimulated HVC
2–4 days of androgen exposure, a sharp upregulation Our results indicate that the testosterone-stimulated
in Quek1 mRNA was observed in both HVC tissue and
HVC vascular bed may constitute a major local source
isolated endothelial cells. Indeed, VEGF-R2/Quek1’s rel-
of BDNF. The effects of this regional burst in tissue
ative absence in the unstimulated HVC, its rapid expres-
BDNF on HVC neurogenesis are likely manifold; besides
sion in response to testosterone, its importance to
its actions as a survival factor for neurons generated in
VEGF-activated vascular expansion, and the impor-
adulthood (Kirschenbaum and Goldman, 1995; Rasika
tance of the latter in BDNF release suggested that the
et al., 1999; Zigova et al., 1998), BDNF has been shown
induction of the VEGF receptor might represent a critical
to be a chemotactic cytokine for fetal neurons in tissue
checkpoint in the process of testosterone-modulated
culture (Behar et al., 1997). As such, HVC BDNF may
neuronal recruitment.
act as a chemotactic factor for newly migratory neurons
upon their departure from the VZ. This is particularly
Androgens and Estrogens Collaborate to Induce
intriguing since dividing endothelial cells of the testos-
HVC Angiogenesis
terone-stimulated adult HVC are spatially delimited to
The induction by testosterone of VEGF-R2/Quek1 in
HVC was reproduced in neostriatal ECs in vitro, in which HVC, and the highest density of new neurons added to

Neuron
956
the androgen-treated mediocaudal neostriatum is within testosterone-induced VEGF, and that only then does
the HVC capillary bed begin to provide a significantHVC (Alvarez-Buylla and Nottebohm, 1988). Together,
these observations suggest that the HVC microvascula- source of BDNF.
ture, by secreting BDNF into the local parenchyma,
might establish a vectorial gradient of BDNF sufficient Potential Physical Interactions of New Neurons
to attract neuronal migrants to sites of endothelial cell and the HVC Vascular Bed
activation, including those of active or recent angiogen- The provision of an “angiogenic niche” for local neuro-
esis. As such, endothelial BDNF might serve as both a genesis has been reported in the dentate gyrus of the
chemoattractant and humoral neurotrophin for neuronal hippocampus, in which mitotic neurogenesis may occur
migrants departing the VZ, directing them to the HVC concurrently with angiogenesis in pseudo-glomerular
capillary bed, wherein they may integrate and survive structures, in which dividing neuronal and endothelial
under the influence of the high local concentrations of precursor cells are directly contiguous (Palmer et al.,
BDNF. 2000). In the adult avian brain, this concept may now
be extended significantly to include the interaction of a
Gonadal Steroids Act at Several Levels neurogenic germinal matrix, the ventricular zone, with a
to Regulate HVC BDNF noncontiguous angiogenic bed into which new neurons
Our observations suggest that two serially activated derived from that germinal zone migrate. In this case,
pathways collaborate to provide a permissive environ- the angiogenic niche provided by the activated micro-
ment for gonadal steroid-induced endothelial BDNF pro- vascular bed includes both a spatial target for neuronal
duction. First, gonadal steroid-induced endothelial migrants as well as a source of trophic support upon
VEGF-R2/Quek1 expression sensitizes the capillary bed their arrival. It is worth noting in this regard that in the
to respond to VEGF. Next, testosterone-induced VEGF, developing telencephalon, after an initial period of radi-
generated as a paracrine angiogenic factor by andro- ally directed migration, new neurons may leave their
gen-receptive cells in HVC, activates Quek1 to stimulate radial cell substrate to migrate tangentially to their tar-
local HVC angiogenesis. The remodeled vessels then gets (Walsh and Cepko, 1992; Walsh and Doherty, 1992).
express BDNF, and they do so in an androgen-depen- In the adult HVC, radial cells may terminate frequently
dent manner. As a result, the androgen-induced expan- on capillary endothelial cells (Goldman, 1995). Thus, it
sion of the HVC capillary vasculature is followed, roughly is possible that new neurons arriving at the angiogenic
2 weeks later, with a geographically discrete upregula- bed of the adult HVC might leave their radial cell partners
tion of BDNF mRNA and protein. The rise in HVC BDNF to migrate upon the adluminal surfaces of the capillary
is thus a function of both the expanded vascular bed in microvasculature, as has been frequently observed with
HVC and the gonadal steroid-enhanced production of transplanted neural progenitor cells. Such migrants
BDNF by the cells therein. might have a ready source of neurotrophic support in
BDNF secreted adluminally by the testosterone-acti-
The Importance of Endothelial BDNF vated endothelium. Furthermore, BDNF-secreting endo-
The increased endothelial expression of BDNF mRNA thelial cells have already been observed to support the
in the testosterone-stimulated birds, coupled with the migration and survival of neurons derived from the adult
robust secretion of BDNF by HVC ECs in vitro, sug- rodent VZ in a trkB-dependent manner (Leventhal et
gested that the capillary endothelium might be a major al., 1999). Together, these observations suggest that
contributor to total BDNF in the testosterone-stimulated endothelial cells of the adult HVC may serve to provide
adult canary HVC. This in turn suggests that ECs might both humoral direction of neuronal chemotaxis and sur-
be a major source of BDNF-associated neurotropism in vival as well as the physical scaffolding for local paren-
this system. In a recent study of adult-derived human chymal migration and dispersal.
brain endothelial cells (HBECs), we noted that cultured
HBECs secreted ⬎1 ng BDNF/10
6
cells/24 hr (Leventhal
Experimental Procedures
et al., 1999). In the present study, we similarly found
RNA Preparation and RT-PCR
that canary brain ECs secreted an average of 1 ng BDNF/
Total RNA was extracted from dissected HVC tissue or HVC endo-
10
6
cells/24 hr under our culture conditions. These con-
thelial cell cultures, and equal amounts of each sample were sub-
centrations are more than sufficient to promote and sup-
jected to RT-PCR as previously described (Leventhal et al., 1999).
port neuronal migration and survival (Leventhal et al.,
Each reaction was standardized against a G3PDH control to permit
comparison between samples in each PCR. cDNA was generated
1999).
using MLV reverse transcriptase (GIBCO, Rockland, ME) and ampli-
The functional importance of this endothelial contribu-
fied by the following gene-specific primers based on chicken ho-
tion to the total testosterone-associated increment in
mologs:
HVC BDNF is suggested by the observation that HVC
The primers used for RT-PCR were as follows: vegf,5⬘-CAT
BDNF levels first rose during the third week after testos-
GAACTTTCTGCTCTCTTG-3⬘and 3⬘-TCTTTCTTTGGTCTGCATTCA
terone administration, only after the expansion and sta-
CAT-5⬘;quek1,5⬘-GGGCGTGGAGCTTTTGGTC-3⬘and 3⬘-CGTCAC
TCAGGGATCGCTCTTC-5⬘;bdnf,5⬘-AGCCTCCTCTGCTCTTTCTG
bilization of the HVC capillary microvasculature. Thus,
CTGGA-3⬘and 3⬘-CTTTTGTCTATGCCCCTGCAGCCTT-5⬘;G3PDH,
relative to the very rapid time course of androgen-
5⬘-CCATGTTCGTCATGGGTGTGAACCA-3⬘and 3⬘-GCCAGTAGAG
induced angiogenesis, the androgen-associated eleva-
GCAGGGATGATGTTC-5⬘.
tion in BDNF was delayed in onset. This suggests that
PCR was carried out for 35 cycles in a Perkin-Elmer 2400 thermal
the HVC vascular bed becomes receptive to VEGF by
cycler (94⬚C, 45 s; 55⬚C, 45 s; 74⬚C, 45 s). The PCR products were
virtue of an initial gonadal steroid-mediated induction
cloned into pGEM-T (Promega) and sequenced by the Cornell Bio-
technology facility in Ithaca, NY. BLAST analysis (NCBI) was per-
of VEGF-R2/Quek1, that it next expands in response to

Angiogenesis-Dependent Neurogenesis in Adult Brain
957
formed to confirm the homology of each derived sequence to known VEGF
Immunoreactivity was localized using rabbit anti-VEGF165 (Neo-sequences.
Markers). Slides were placed in 0.1% Triton X-100 for 30 min,
washed in 0.1 M PB, exposed to 10% normal rabbit serum for 30
Semiquantitative RT-PCR
min, then anti-VEGF IgG at 1:200 overnight at 4⬚C. After washing,
RNA was extracted from pooled bilateral HVC samples at each of
the slides were incubated in biotinylated goat anti-rabbit IgG (1:200;
six time points, which included 0, 4, 8, 14, 21, and 28 days post-
Vector Laboratories) at 25⬚C for 90 min, washed, incubated in avidin-
silastic hormone implantation (n ⫽3 birds/time point). RNA was
biotin-HRP complex (Vectastain ABC; Vector) for 60 min, washed
extracted from pooled bilateral HVC samples, and equal amounts
again, then developed with diaminobenzidine (0.2 mg/ml) and
were subjected to [
32
P]dCTP-reported RT-PCR, with each reaction
0.003% hydrogen peroxide for 5 min. The slides were then washed,
standardized against a glucose 3-phosphate dehydrogenase
cleared through ascending alcohols and xylene, mounted in Per-
(G3PD) internal control. After extraction by RNA-STAT-20, 50–200
mount, and photographed using an Olympus IX70 photomicro-
ng of each sample was reverse transcribed using Superscript II
scope. Control slides were treated similarly except for the substitu-
RNase H-reverse transcriptase (GIBCO/Life Technologies) and then
tion of 1:1000 normal rabbit serum for the primary antibody; no
digested with RNase H (1 ␮l) for 20 min at 37⬚C. The cDNAs thereby
cellular labeling was observed in these sections.
generated were amplified in 50␭PCR reactions that contained 2
Laminin
mM MgCl
2
,1⫻buffer, 0.2 ␮M primers, 0.2 ␮M dNTP mix, 50 ng (10
Visualization of the vasculature was achieved using a rabbit anti-
␮l) RT product, 1 U red Taq polymerase, and 2.5 ␮Ci [
32
P]dCTP.
laminin IgG (Sigma). Sections were permeabilized by 0.1% saponin
Reactions were carried out for 24 cycles (G3PDH), 28 cycles (VEGF),
for 15 min, blocked with 10% goat serum for 1 hr, and exposed to
or 30 cycles (BDNF) in a Perkin-Elmer 2400 thermal cycler, with
anti-laminin IgG (1:100) overnight at 4⬚C. After washing, the primary
denaturing at 94⬚C, 30 s; annealing at 55⬚C, 1 min; and synthesis at
anti-laminin antibody was detected with Alexa Fluor 566-conjugated
72⬚C, 2 min. These conditions produced amplicons within the linear
goat anti-rabbit IgG (Molecular Probes) (1:400). Endothelial cell divi-
exponential phase of the PCR amplification curve. The resultant
sion was detected by staining for BrdU (see below) prior to laminin
PCR products were then separated using PAGE and exposed to a
immunolabeling.
phosphoimager cassette. The volume-density of each VEGF and
BDNF band was then quantified using ImageQuant software and
normalized against G3PD by dividing that time-point band density Endothelial Cell Selection, Separation, and Culture
Adult canary brain endothelial cells (CBECs) were obtained fromby that of its matched G3PD PCR.
the dissected neostriatum of adult canaries. The birds were perfused
with cold HBSS via a transcardiac approach following terminal anes-
In Situ Hybridization
thesia by pentobarbital overdose. The tissue was collected in
Upon sacrifice, brains were fixed by intracardiac perfusion with
HEPES-buffered DMEM/F12 and minced into 2–4 mm
3
cubes. Isola-
cold PBS followed by 4% paraformaldehyde in PBS, followed by
tion and preparation of endothelial cell cultures was performed using
postfixation for 2 hr at 4⬚C and saturation with 30% sucrose over-
our previously described modification (Leventhal et al., 1999) of
night. 40 ␮m sections were then cut on a freezing microtome, col-
established protocols (Biegel et al., 1995). Briefly, the HVC endothe-
lected in cold 0.1 M phosphate buffer (PB), cleared through alcohols
lial cells were identified by their cobblestone-like morphology and
to chloroform, and rehydrated in 0.1 M PB. Sections were transferred
binding of DiI-conjugated acetylated LDL (1,1⬘-dioctacecyl-3,3,3⬘3,
to cold 2⫻SSC. Prehybridization and hybridization were carried out
tetramethyl indocarbocyanine-acetylated low density lipoprotein)
in buffer containing 2⫻SSC, 50% formamide, 10% dextran sulfate,
(Intracel, Rockville, MD). The DiI-labeled endothelial cells were then
5⫻Denhardt’s solution, and 1 mg/ml denatured salmon sperm DNA.
enriched to purity using FACS. To this end, the cultures were incu-
Probe DNA was prepared using partial cDNA sequences for ca-
bated at 37⬚C for 1 hr with diI-acetylated LDL (1 ␮g/ml), washed
nary vegf,quek1, and bdnf, each obtained by reverse transcription
three times in PBS, trypsinized, and resuspended in 2 ml phenol
of RNA from intestine (vegf and quek1) and brain (bdnf). The probes
red-free medium. The endothelial dissociates were then subjected
were labeled with [S
35
]dCTP using random primer labeling (Amer-
to FACS using a Becton-Dickinson FACS Vantage system, and the
sham). Denatured probe was added to the prehybridization buffer
DiI-labeled fraction was separated at 2000–3000 cells/sec. We have
and hybridized overnight. Posthybridization washes were performed
previously verified, by immunolabeling for the endothelial-specific
in 2⫻SSC, 1⫻SSC, and 0.5⫻SSC at 42⬚C followed by a 25⬚C wash
antigens factor VIII and PECAM, that ⬎99% of DiI-LDL-bound cells
in 0.05 M PB. Sections were mounted on gelatin slides and either
sorted by this protocol are endothelial (Leventhal et al., 1999).
dipped in NTB2 emulsion (Kodak) or exposed to X-ray film for autora-
Upon isolation, the cells were raised in microvascular endothelial
diography.
basal medium (MEBM; Sigma) supplemented with Endothelial Sup-
plement (1⫻, Sigma), 20 ng/ml VEGF (Genzyme), and 20 ng/ml FGF2
Immunohistochemistry (Sigma). The cells were plated on 24-well plates (Falcon Primaria)
For VEGF and Hu immunohistochemistry, brains were fixed by intra- coated with human fibronectin (GIBCO, 1 ␮g/cm
2
)at5⫻10
4
cells/cm
2
.
cardiac perfusion of the animal with cold HBSS followed by 4%
paraformaldehyde and postfixation and saturation with 30% su-
[
3
H]Thymidine Incorporation Assay
crose. Each brain was sectioned transversely at 14 ␮m on a Hacker
Endothelial cells were enriched from the HVC region by FACS of DiI-
cryostat and stored frozen prior to immunohistochemistry.
LDL-tagged dissociates of adult canary HVC and adjacent medio-
Hu/BrdU
caudal neostriata. The cultures were raised for 2 days in MEBM
Newly generated HVC neurons were detected by staining for Hu
medium containing 1% steroid-depleted FBS, then switched for 2
protein, a neuronal marker (Barami et al., 1995), and BrdU using
days into test media supplemented with either testosterone (100
double-immunofluorescence. Hu antigen was detected using a
ng/ml), ␤-estradiol (10 ng/ml), dihydrotestosterone (100 ng/ml), cho-
mouse monoclonal anti-Hu IgG, 16A11 (Molecular Probes, Eugene,
lesterol (100 ng/ml), or VEGF (20 ng/ml), in the presence of 3.5 ␮Ci/
OR), which recognizes the Hu/Elav family members, HuC, HuD, and
ml [
3
H]thymidine. Each steroid was added either as a cyclodextran-
Hel-N1, and specifically labels neurons. The antibody was used at
encapsulated water-soluble formulation (Sigma: testosterone,
a concentration of 10 ␮g/mL. Sections were rehydrated with PBS
CT5035; estradiol, E4389) or presolubilized in dimethylsulfoxide
and treated with 10 mM sodium citrate (pH 6.0) for 40 min. Sections
(cholesterol, 5␣-dihydrotestosterone). After 2 days in test media,
were then permeabilized with 0.1% saponin for 15 min, blocked
the cultures were extracted, and [
3
H]thymidine incorporation was
with 10% goat serum for 1 hr, and exposed to anti-Hu IgG overnight
counted by a Packard 1900CA scintillation counter.
at 4⬚C. After washing, the primary anti-Hu antibody was detected
with biotinylated goat anti-mouse IgG Fab (1:200), followed by a
Texas red-conjugated avidin (Vector, 1:50). The sections were dena- ELISA
Tissuetured in 2 N HCl for 30 min, then stained for BrdU using rat anti-
BrdU IgG antibody at 1:200 (Harlan Sprague Dawley, Indianapolis, To quantify BDNF and VEGF protein levels, dissected tissue samples
were first frozen in liquid nitrogen and then stored at ⫺80⬚C. At theIN), followed serially by Alexa Fluor 488-conjugated anti-rat IgG
(Molecular Probes). time of assay, the samples were thawed and lysed in 2 ml/g of tissue

Neuron
958
of Protein Lysis Buffer (137 mM NaCl, 20 mM Tris, 1% NP-40, 10% for VEGF-R2 and 2.0 ␮M for VEGF-R1/flt1) (Hennequin et al., 1999).
The substituted halogens and its methoxy groups at C6 and C7glycerol, 1 mM PMSF, 10 ␮g/ml aprotinin, 0.5 mM sodium vanadate).
The mixture was homogenized and centrifuged at 2500 rpm for 15 made this particular anilinoquinazoline both more potent and the
most likely of the series to cross the blood brain barrier based onmin; the supernatant was then collected and assayed either for
BDNF, using Promega’s BDNF ELISA, or VEGF, usingR&DSystem’s lipophobicity.
Quantikine ELISA, both according to the manufacturer’s protocols.
Cells Confocal Imaging
To measure BDNF protein, confluent layers of endothelial cells in In sections double-stained for BrdU and Hu, single BrdU
⫹
cells that
T75 flasks were homogenized in 1.5 ml buffer (50 mM Tris-HCl [pH appeared double-labeled with Hu were further evaluated by confocal
7.4], 600 mM NaCl, 0.2% Triton X-100, 1% BSA fraction V, 200 KIU/ imaging. Using a Zeiss LSM510 microscope, images were acquired
ml aprotonin, 0.1 mM benzethonium chloride, 1 mM benzamidine, in both red and green emission channels using an argon-krypton
and 0.1 mM PMSF; all from Sigma or GIBCO). Lysates were centri- laser. The images were then viewed as series of single 0.9 ␮m opti-
fuged at 10,000 ⫻g for 30 min at 4⬚C, and their supernatants frozen cal sections and as merged z-dimension composites thereof. The
at ⫺80⬚C. Conditioned media from each sample were also frozen z-dimension reconstructions were observed in profile, as every BrdU
⫹
/
for ELISA. BDNF levels were then determined using a two-site ELISA Hu
⫹
cell was observed orthogonally in both the vertical and hori-
(Promega) (Mizisin et al., 1997). The mouse capture antibody specifi- zontal planes. Cells were scored as double-labeled, newly generated
cally recognized BDNF and did not cross react with other members neurons only when observed to have central BrdU immunoreactivity
of the neurotrophin family at concentrations up to 10,000 times that surrounded by Hu-immunoreactivity in every serial optical section
used for the standard curve. The reporter antibody was a biotinyl- and in each rotated orthogonal view, as previously described (Ben-
ated rabbit polyclonal anti-BDNF. The dynamic range of the ELISA raiss et al., 2001).
was 10 pg/ml–20 ng/ml for undiluted samples; all samples were
diluted in assay buffer to bring them into the linear range of the Scoring
assay. Sample values were extrapolated from the standard curve, Capillary Parameters
and each represented the average of triplicate samples. Total pro- In each sampled HVC, every laminin
⫹
profile within the cresyl-
tein levels were assessed by DC protein assay (Bio-Rad). defined borders of HVC was traced and recorded manually into
BioQuant image analysis software. A profile was defined as the
Adult Canary VZ Explant Cultures net surface area of laminin immunolabeling. Capillary counts were
Culture Preparation derived from these profiles, and no correction was made for tangen-
For neuronal outgrowth and survival studies, explants were obtained tial cuts. This was based upon an assumption of randomly distrib-
from 1-year-old female canaries (Serinus canarius, American Singer uted capillary orientations in HVC coronal sections. Each traced
str., n ⫽5). Cultures were prepared from the neostriatal VZ, both outline was then analyzed for its area, perimeter, and diameter, and
overlying and medial to nucleus HVC, using previously reported the average numbers per HVC and per unit area were determined.
methods and media (Goldman, 1990; Goldman et al., 1992). In brief, Statistical analysis between treatment groups was then accom-
to phenol red-free DMEM/F12 with 15 mM HEPES, we added gluta- plished by ANOVA followed by post hoc Boneferroni t tests.
mine (2 mM), glucose (6 mg/ml), an insulin (6 ␮g/ml)-transferrin (6 Endothelial Labeling Indices
␮g/ml)-selenium (6 ng/ml) mix supplemented with linoleic acid-rich For the evaluation of testosterone-induced angiogenesis, the brains
(5 ␮g/ml) albumin (1.25 mg/ml; ITS
⫹
, Collaborative Research), hydro- of six birds were cryosectioned at 14 ␮m and double-immunostained
cortisone (300 ng/ml; water-soluble preparation, Sigma H0396), pro- for BrdU together with laminin. In each nucleus HVC, the incidence
gesterone (60 fg/ml), putrescine (16 ␮g/ml), and tri-iodothyronine of laminin
⫹
/BrdU
⫹
cells and the ratio of laminin
⫹
/BrdU
⫹
cells to total
(30 ng/ml). To this base, we added charcoal-stripped fetal bovine HVC laminin
⫹
ECs were determined bilaterally in each of six sections
serum (Hyclone) and castrated rooster (capon) serum (Cocalico), spaced 200 ␮m apart. Total HVC endothelial cells were counted
each at 10% v/v. The charcoal-stripped serum was assayed to con- by double-labeling laminin-stained sections with the nuclear dye
tain 330 pg/ml T3 and 1.3 ␮U/ml insulin, each representing an ap- Hoechst 33258 so as to visualize and score individual endothelial
proximate 80% reduction in their initial concentration (Hyclone nuclei.
Labs). Estradiol, testosterone, and androstenedione were all unde- Neuronal Labeling Indices
tectable in both the stripped sera and resultant depleted medium For the VEGF TKI inhibition study, five 14 ␮m sections were chosen
(Mayo Clinic Labs) (see Hidalgo et al. [1995] for assay description). from each of 14 birds. These were taken at intervals of 100 ␮m
Each bird’s explants were apportioned into treatment (BDNF, 20 ng) spanning the nucleus HVC through the antero-posterior coordinates
and control groups. Once an explant displayed cellular outgrowth, A0.0 to P0.6 in the stereotaxic plane (Stokes et al., 1976). Each
its neurons were identified using morphological criteria that we have sampled slide included both hemispheres, roughly symmetrical in
established and verified immunocytochemically, ultrastructurally, the coronal plane at the level of HVC. HVC was delineated in each
and functionally in this system (Goldman et al., 1992). Neurons within hemisphere by cresyl violet Nissl staining in adjacent sections. The
each explant outgrowth were counted every third day beginning total number of Hu
⫹
neurons and double-positive Hu
⫹
/BrdU
⫹
neu-
on day 5; one-third volume media changes were performed twice rons were scored in the ten HVC samples analyzed per bird. The
weekly, with BDNF supplementation as appropriate. percentage of HVC Hu
⫹
neurons that were double-labeled for BrdU
Statistical Analysis wasdetermined for each bird, and the group means, standard devia-
At each time point, the mean number of neurons per explant were tions, and errors were then calculated. The group means were then
calculated for each treatment group (BDNF and null control). Explant compared by two-way ANOVA, with post hoc comparisons by Bon-
growth patterns were then analyzed using two-factor analysis of ferroni t tests. Comparisons were considered significant for p ⱕ
variance (ANOVA). The time points (days in vitro) at which counts 0.05.
were made constituted the repeated measures factor, whereas the
type of treatment (BDNF versus vehicle) was the main effect of the Acknowledgments
ANOVA. For statistically significant ANOVA results of p ⬍0.05, post
hoc pairwise comparisons were made between treatments using a Supported by the G. Harold and Leila Y. Mathers Charitable Founda-
Boneferroni adjustment for multiple comparisons. tion, NINDS R37/R01NS29813 and R01NS33106, the National Multi-
ple Sclerosis Society, and Project ALS. A.L. was supported by NIH
MSTP grant GM07739 and the Friedman Fellowship Endowment.VEGF Tyrosine Kinase Inhibition
4-[(4⬘-Chloro-2⬘-fluor o)p heny lamino] -6,7 -di meth oxyquin azoline We are grateful to Drs. George Yancopoulos, Stan Wiegand, and
Anne Acheson of Regeneron Pharmaceuticals for BDNF protein,(VEGFR Tyrosine Kinase Inhibitor, Calbiochem No. 676475) was
used to inhibit VEGF-dependent angiogenesis according to the ex- cDNA, and for sharing their ELISA protocols. We thank Dr. Marcus
Reidenberg for advice on the use of VEGF tyrosine kinase inhibitors,perimental design described above in the Results. This inhibitor is
the fourth of a series of anilinoquinazolines synthesized as potent Drs. Francis Barany and Abdellatif Benraiss for advice on relative
PCR, and Donna Roh for technical assistance.inhibitors of VEGF receptor tyrosine kinase activity (IC
50
⫽0.1 ␮M

Angiogenesis-Dependent Neurogenesis in Adult Brain
959
Received: March 20, 2001 Fusani, L., Hutchison, J., and Gahr, M. (2001). Testosterone regu-
lates the activity and expression of aromatase in the canary neostria-Revised: April 15, 2002
tum. J. Neurobiol. 49, 1–8.
References Gahr, M. (1990). Delineation of a brain nucleus: comparisons of
cytochemical, hodological, and cytoarchitectural views of the song
Alitalo, K., Kurkinen, M., Virtanen, I., Mellstrom, K., and Valheri, A. control nucleus HVC of the adult canary. J. Comp. Neurol. 294,
(1982). Deposition of basement membrane proteins in attachment 30–36.
and neurite formation of cultured murine C-1300 neuroblastoma Gensburger, C., Labourdette, G., and Sensenbrenner, M. (1987).
cells. J. Cell. Biochem. 18, 25–36. Brain basic fibroblast growth factor stimulates the proliferation of
Alvarez-Buylla, A., and Nottebohm, F. (1988). Migration of young rat neuronal precursor cells in vitro. FEBS Lett. 217, 1–5.
neurons in adult avian brain. Nature 335, 353–354. Goldman, S.A. (1990). Neuronal development and migration in ex-
Andersson, C., Lydrup, M.-L., Ferno, M., Idvall, I., Gustafsson, J.-A., plant cultures of the adult canary forebrain. J. Neurosci. 10, 2931–
and Nilsson, B. (2001). Immunocytochemical demonstration of oes- 2939.
trogen receptor-beta in blood vessels of the female rat. J. Endocri- Goldman, S.A. (1995). Neuronal precursor cells and neurogenesis
nol. 169, 241–247. in the adult forebrain. Neuroscientist 1, 338–350.
Araujo, D., and Cotman, C. (1993). Trophic effects of interleukin-4, Goldman, S.A. (1998) Adult neurogenesis: from canaries to the clinic.
-7 and -8 on hippocampal neuronal cultures: potential involvement J. Neurobiol. 36, 267–286.
of glial-derived factors. Brain Res. 600, 49–55. Goldman, S.A., and Luskin, M.B. (1998). Strategies utilized by mi-
Balthazart, J., Absil, P., Foidart, A., Houbart, M., Harada, N., and grating neurons of the postnatal vertebrate forebrain. Trends Neu-
Ball, G. (1996). Distribution of aromatase-immunoreactive cells in rosci. 21, 107–114.
the forebrain of zebra finches: implications for the neural action of Goldman, S.A., and Nottebohm, F. (1983). Neuronal production, mi-
steroids. J. Neurobiol. 31, 129–148. gration, and differentiation in a vocal control nucleus of the adult
Barami, K., Kirschenbaum, B., Lemmon, V., and Goldman, S.A. female canary brain. Proc. Natl. Acad. Sci. USA 80, 2390–2394.
(1994). N-cadherin and Ng-CAM/8D9 are involved serially in the
Goldman, S.A., Zaremba, A., and Niedzwiecki, D. (1992). In vitro
migration of newly generated neurons into the adult songbird brain.
neurogenesis by neuronal precursor cells derived from the adult
Neuron 13, 567–582.
songbird brain. J. Neurosci. 12, 2532–2541.
Barami, K., Iversen, K., Furneaux, H., and Goldman, S.A. (1995). Hu
Hanahan, D. (1997). Signaling vascular morphogenesis and mainte-
protein as an early marker of neuronal phenotypic differentiation by
nance. Science 277, 48–50.
subependymal zone cells of the adult songbird forebrain. J. Neuro-
biol. 28, 82–101. Hennequin, L.F., Thomas, A.P., Johnstone, C., Stokes, E.S., Ple
´,
P.A., Lohmann, J.J., Ogilvie, D.J., Dukes, M., Wedge, S.R., Curwen,
Behar, T., Dugich-Djordjevic, M.M., Li, Y.X., Ma, W., Somogyi, R.,
J.O., et al. (1999). Design and structure-activity relationship of a new
Wen, X., Brown, E., Scott, C., McKay, R.D., and Barker, J.L. (1997).
class of potent VEGF receptor tyrosine kinase inhibitors. J. Med.
Neurotrophins stimulate chemotaxis of embryonic cortical neurons.
Chem. 42, 5369–5389.
Eur. J. Neurosci. 9, 2561–2570.
Hidalgo, A., Barami, K., Iversen, K., and Goldman, S.A. (1995). Estro-
Benraiss, A., Chmielnicki, E., Roh, D., and Goldman, S.A. (2001).
gens and non-estrogenic ovarian influences combine to promote
Adenoviral BDNF induces both neostriatal and olfactory neuronal
the recruitment and decrease the turnover of new neurons in the
recruitment from endogenous progenitor cells in the adult forebrain.
adult female canary brain. J. Neurobiol. 27, 470–487.
J. Neurosci 21, 6718–6731.
Holzenberger, M., Jarvis, E., Chong, C., Grossman, M., Nottebohm,
Bergers, G., Brekken, R., McMahon, G., Vu, T., Itoh, T., Tmaki, K.,
F., and Scharff, C. (1997). Selective expression of insulin-like growth
Tanzawa, K., Thorpe, P., Itohara, S., Werb, Z., et al. (2000). Matrix
factor II in the songbird brain. J. Neurosci. 17, 6974–6987.
metalloproteinase-9 triggers the angiogenic switch during carcino-
genesis. Nat. Cell Biol. 2, 737–744. Iafrati, M.D., Karas, R.H., Aronovitz, M., Kim, S., Sullivan, T.R., Jr.,
Lubahn, D.B., O’Donnell, T.F. Jr., Korach. K.S., and Mendelsohn,
Bernard, D., Bentley, G., Balthazart, J., Turek, F., and Ball, G. (1999).
M.E. (1997). Estrogen inhibits the vascular injury response in estro-
Androgen receptor, estrogen receptor alpha, and estrogen receptor
gen receptor ␣-deficient mice. Nat. Med. 3, 545–548.
beta show distinct patterns of expression in forebrain song control
nuclei of European starlings. Endocrinology 140, 4633–4643. Jain, R., Safabakhsh, N., Sckell, A., Chen, Y., Jiang, P., Benjamin,
L., Yuan, F., and Keshet, E. (1998). Endothelial cell death, angiogen-
Biegel, D., Spencer, D.D., and Pachter, J.S. (1995). Isolation and
esis, and microvascular function after castration in an androgen-
culture of human brain microvascular endothelial cells for the study
dependent tumor: role of VEGF. Proc. Natl. Acad. Sci. USA 95,
of blood-brain barrier properties in vitro. Brain Res. 692, 183–189.
10820–10825.
Biro, S., Yu, Z., Fu, Y., Smale, G., Sasse, J., Sanchez, J., Ferrans,
Jiang, W., McMurtry, J., Niedzwiecki, D., and Goldman, S.A. (1998).
V., and Casscells, W. (1994). Expression and subcellular distribution
Insulin-like growth factor-1 is a radial cell-associated neurotrophin
of basic fibroblast growth factor are regulated during migration of
that promotes neuronal recruitment from the adult songbird ventric-
endothelial cells. Circ. Res. 74, 485–494.
ular zone. J. Neurobiol. 36, 1–15.
Bottjer, S., and Johnson, F. (1997). Circuits, hormones, and learning:
Johe, K.K., Hazel, T.G., Muller, T., Dugich-Djordjevic, M.M., and
vocal behavior in songbirds. J. Neurobiol. 33, 602–618.
McKay, R.D. (1996). Single factors direct the differentiation of stem
Breier, G., Albrecht, U., Sterrer, S., and Risau, W. (1992). Expression cells from the fetal and adult central nervous system. Genes Dev.
of vascular endothelial growth factor during embryonic angiogen- 10, 3129–3140.
esis and endothelial cell differentiation. Development 114, 521–532.
Johnson, F., and Bottjer, S. (1995). Differential estrogen accumula-
Caulin-Glaser, T., Watson, C., Pardi, R., and Bender, J. (1996). Ef- tion among populations of projection neurons in the higher vocal
fects of 17␤-estradiol on cytokine-induced endothelial cell adhesion center of male canaries. J. Neurobiol. 26, 87–108.
molecule expression. J. Clin. Invest. 98, 36–42.
Kirschenbaum, B., and Goldman, S.A. (1995). Brain-derived neuro-
Dittrich, F., Feng, Y., Metzdorf, R., and Gahr, M. (1999). Estrogen- trophic factor promotes the survival of neurons arising from the
inducible, sex-specific expression of BDNF mRNA in a forebrain adult rat forebrain subependymal zone. Proc. Natl. Acad. Sci. USA
song control nucleus of the juvenile zebra finch. Proc. Natl. Acad. 92, 210–214.
Sci. USA 96, 8241–8246.
Legan, S., Coon, G., and Karsch, F. (1975). Role of estrogen as
Drago, J., Murphy, M., Carroll, S.M., Harvey, R.P., and Bartlett, P.F. initiator of daily LH surges in the ovariectomized rat. Endocrinology
(1991). Fibroblast growth factor-mediated proliferation of central 96, 50–56.
nervous system precursors depends on endogenous production of
insulin-like growth factor I. Proc. Natl. Acad. Sci. USA 88, 2199–2203. Leung, D., Cachianes, G., Kuang, W., Goeddel, D., and Ferrara, N.

Neuron
960
(1989). Vascular endothelial growth factor is a secreted angiogenic Robertson, P., Du Bois, M., Bowman, P., and Goldstein, G. (1985).
Angiogenesis in the developing rat brain: an in vivo and in vitromitogen. Science 246, 1306–1309.
study. Brain Res. 23, 219–223.
Leventhal, C., Rafii, S., Rafii, D., Shahar, A., and Goldman, S.A.
Rosenstein, J., Mani, N., Silverman, W., and Krum, J. (1998). Patterns
(1999). Endothelial trophic support of neuronal production and re-
of brain angiogenesis after vascular endothelial growth factor ad-
cruitment from the adult mammalian subependyma. Mol. Cell. Neu-
ministration in vitro and in vivo. Proc. Natl. Acad. Sci. USA 95,
rosci. 13, 450–464.
7086–7091.
Levine, A., Liu, X., Greenberg, P., Eliashvili, M., Schiff, J., Aaronson,
Shen, P., Schlinger, B., Campagnoni, A., and Arnold, A. (1995). An
S., Holland, J., and Kirschenbaum, A. (1998). Androgens induce the
atlas of aromatase mRNA expression in the zebra finch brain. J.
expression of VEGF in human fetal prostatic fibroblasts. Endocrinol-
Comp. Neurol. 360, 172–184.
ogy 139, 4672–4678.
Soker, S., Takashima, S., Miao, H., Neufeld, G., and Klagsbrun, M.
Li, X., Jarvis, E., Alvarez-Borda, B., Lim, D., and Nottebohm, F.
(1998). Neuropilin-1 is expressed by endothelial and tumor cells as
(2000). A relationship between behavior, neurotrophin expression,
an isoform-specific receptor for VEGF. Cell 92, 735–745.
and new neuron survival. Proc. Natl. Acad. Sci. USA 97, 8584–8589.
Stokes, T., Leonard, C., and Nottebohm, F. (1976). The telencepha-
Lindner, V., Kim, S., Karas, R., Kuiper, G., Gustafsson, J., and Men-
lon, diencephalon and mesencephalon of the canary, Serinus cana-
delsohn, M. (1998). Increased expression of estrogen receptor ␤
ria, in stereotaxic coordinates. J. Comp. Neurol. 156, 337–374.
mRNA in male blood vessels after vascular injury. Circ. Res. 83,
224–229. Suri, C., Jones, P., Patan, S., Bartunkova, S., Maisonpierre, P., Davis,
S., Sato, T., and Yancopoulos, G. (1996). Requisite role of angiopoe-
Millauer, B., Wizigmann, S., Schnurch, H., Martinez, R., Moller, N.,
tin-1, a ligand for the TIE2 receptor, during embryonic angiogenesis.
Risau, W., and Ullrich, A. (1993). High-affinity VEGF binding and
Cell 87, 1171–1180.
developmental expression suggest FLk-1 as a major regulator of
vasculogenesis and angiogenesis. Cell 72, 835–846. Suri, C., McClain, J., Thurston, G., McDonald, D., Zhou, H., Old-
mixon, E., Sato, T., and Yancopoulos, G. (1998). Increased vasculari-
Mizisin, A.P., Bache, M., DiStefano, P.S., Acheson, A., Lindsay, R.M.,
zation in mice overexpressing angiopoiein-1. Science 282, 468–471.
and Calcutt, N.A. (1997) BDNF attenuates functional and structural
disorders in nerves of galactose-fed rats. J. Neuropathol. Exp. Neu- Thurston, G., Suri, C., Smith, K., McClain, J., Sato, T., Yancopoulos,
rol. 56, 1290–1301. G., and McDoanld, D. (1999). Leakage-resistant blood vessels in
mice transgenically overexpressing angiopoietin-1. Science 286,
Mosselman, S., Polman, J., and Dijkema, R. (1996). ER␤: identifica-
2511–2514.
tion and characterization of a novel human estrogen receptor. FEBS
Lett. 392, 49–53. Walsh, C., and Cepko, C.L. (1992). Widespread dispersion of neu-
ronal clones across functional regions of the cerebral cortex. Sci-
Mueller, M., Vigne, J., Minchenko, A., Lebovic, D., Leitman, D., and
ence 255, 434–440.
Taylor, R. (2000). Regulation of VEGF gene transcription by estrogen
receptors ␣and ␤. Proc. Natl. Acad. Sci. USA 97, 10972–10977. Walsh, F., and Doherty, P. (1992). Second messengers underlying
cell-contact dependent axonal growth stimulated by transfected
Nag, S., Takahashi, J., and Kilty, D. (1997). Role of VEGF in blood-
N-CAM, N-cadherin, or L1. Cold Spring Harb. Symp. Quant. Biol.
brain barrier breakdown and angiogenesis in brain trauma. J. Neuro-
57, 431–440.
pathol. Exp. Neurol. 56, 912–921.
Wedge, S., Ogilvie, D., Dukes, M., Kendrew, J., Curwen, J., Hen-
Nastiuk, K., and Clayton, D. (1995). The canary androgen receptor
nequin, L., Thomas, A., Stokes, E., Curry, B., Richmond, G., and
mRNA is localized in the song control nuclei of the brain and is
Wadsworth, P. (2000). ZD4190: an orally active inhibitor of VEGF
rapidly regulated by testosterone. J. Neurobiol. 26, 213–224.
signaling with broad-spectrum antitumor efficacy. Cancer Res. 60,
Nordeen, E., and Nordeen, K. (1989). Estrogen stimulates the incor- 970–975.
poration of new neurons into avian song nuclei during adolescence.
Wilting, J., Eichmann, A., and Christ, B. (1997). Expression of the
Brain Res. Dev. Brain Res. 49, 27–32.
avian VEGF recptor homologues Quek1 and Quek2 in blood-vascu-
Nottebohm, F. (1981). A brain for all seasons: cyclical anatomical lar and lymphatic endothelial and non-endothelial cells during quail
changes in song control nuclei of the canary brain. Science 214,embryonic development. Cell Tissue Res. 288, 207–223.
1368–1370.
Yancopoulos, G., Davis, S., Gale, N., Rudge, J., Wiegand, S., and
Nottebohm, F. (1985). Neuronal replacement in adulthood. Ann. N Holash, J. (2000). Vascular-specific growth factors and blood vessel
Y Acad. Sci. 457, 143–161. formation. Nature 407, 242–248.
Ogunshola, O., Stewart, W., Mihalcik, V., Solli, T., Madri, J., and Yang, X, and Cepko, C. (1996). Flk-1, a receptor for vascular endo-
Ment, L. (2000). Neuronal VEGF expression correlates with angio- thelial growth factor (VEGF), is expressed by retinal progenitor cells.
genesis in postnatal developing rat brain. Brain Res. Dev. Brain Res. J. Neurosci. 16, 6089–6099.
119, 139–153.
Zhang, Z., Zhang, L., Jiang, Q., Davies, K., Powers, C., van Bruggen,
Ogunshola, O., Antic, A., Donoghue, M.J., Fan, S.Y., Kim, H., Stewart, N., and Chopp, M. (2000). VEGF enhances angiogenesis and pro-
W.B., Madri, J.A., and Ment, L.R. (2002). Paracrine and autocrine motes blood–brain barrier leakage in the ischemic brain. J. Clin.
functions of neuronal VEGF in the CNS. J. Biol. Chem. 277, 11410– Invest. 106, 829–838.
11415.
Zigova, T., Pencea, V., Wiegand, S., and Luskin, M. (1998). Intraven-
Palmer, T., Willhoite, A., and Gage, F. (2000). Vascular niche for tricular administration of BDNF increases the number of newly gen-
adult hippocampal neurogenesis. J. Comp. Neurol. 425, 479–494. erated neurons in the adult olfactory bulb. Mol. Cell. Neurosci. 11,
Pardridge, W. (1995). Transport of small molecules through the 234–245.
blood-brain barrier: biology and methodology. Adv. Drug Deliv. Rev.
15, 5–36.
Proescholdt, M., Heiss, J., Walbridge, S., Muhlhauser, J., Oldfield,
E., and Merrill, M. (1999). VEGF modulates vascular permeability
and inflammation in the rat brain. J. Neuropathol. Exp. Neurol. 58,
613–627.
Rasika, S., Nottebohm, F., and Alvarez-Buylla, A. (1994). Testoster-
one increases the recruitment and/or survival of new high vocal
center neurons in adult female canaries. Proc. Natl. Acad. Sci. USA
89, 8591–8595.
Rasika, S., Alvarez-Buylla, A., and Nottebohm, F. (1999). BDNF medi-
ates the effects of testosterone on the survival of new neurons in
an adult brain. Neuron 22, 53–62.