Content uploaded by Muriel Koehl
Author content
All content in this area was uploaded by Muriel Koehl on Aug 07, 2015
Content may be subject to copyright.
Content uploaded by Muriel Koehl
Author content
All content in this area was uploaded by Muriel Koehl on Aug 06, 2015
Content may be subject to copyright.
Prenatal stress produces learning deficits associated
with an inhibition of neurogenesis in
the hippocampus
V. Lemaire*, M. Koehl*, M. Le Moal, and D. N. Abrous
†
Laboratoire de Psychobiologie des Comportement Adaptatifs, Institut National de la Sante´ et de la Recherche Me´ dicale Unit 259, Universite´ de Bordeaux II,
Domaine de Carreire, Rue Camille Saint Sae¨ ns, 33077 Bordeaux Cedex, France
Edited by Mark R. Rosenzweig, University of California, Berkeley, CA, and approved July 7, 2000 (received for review March 21, 2000)
Early experiences such as prenatal stress significantly influence the
development of the brain and the organization of behavior. In
particular, prenatal stress impairs memory processes but the mech-
anism for this effect is not known. Hippocampal granule neurons
are generated throughout life and are involved in hippocampal-
dependent learning. Here, we report that prenatal stress in rats
induced lifespan reduction of neurogenesis in the dentate gyrus
and produced impairment in hippocampal-related spatial tasks.
Prenatal stress blocked the increase of learning-induced neurogen-
esis. These data strengthen pathophysiological hypotheses that
propose an early neurodevelopmental origin for psychopatholog-
ical vulnerabilities in aging.
I
t is well documented from animal studies that during the perinatal
period, the development of an organism is subjected to complex
environmental influences. Deleterious life events during pregnancy
induce neurobiological and behavioral defects in offspring, some of
them involving the hippocampal formation (1–5). Indeed, prenatal
stress results in an enhanced production of stress hormones by the
mother during critical periods of fetal brain development and
provokes a definitively longer corticosterone response to stress in
the offspring associated with a reduction in the number of hip-
pocampal corticosteroid receptors (1, 3, 5). Behaviorally, the prog-
eny, from adulthood to senescence, exhibit memory deficits in a
hippocampal-dependent task (2, 4, 5).
Recently, it has been hypothesized that hippocampal-mediated
learning (6) may be related to the generation of new neurons in the
adult dentate gyrus (7, 8, 9). These newborn cells migrate in the
granule cell layer, and differentiate in granule neurons whose
projections, the mossy fibers, extend to the CA3 hippocampal
region (10, 11). Furthermore, the size of the mossy fibers’ projec-
tions correlates with variations in performances in spatial memory
tests (12, 13). Finally, glucocorticoid levels regulate de novo cell
proliferation in the dentate gyrus. Indeed, adrenalectomy per-
formed in young or aged rats increases neurogenesis, an effect that
is prevented by glucocorticoid treatment (14, 15, 16). These results
raised the critical question as to whether prenatal stress can impair
neurogenesis and, if so, whether it is related to learning ability.
To test this hypothesis, we first examined cell proliferation in
the progeny of stressed mothers with 5-bromo-2⬘-deoxyuridine
(BrdUrd), a thymidine analogue incorporated into genetic material
during synthetic DNA phase (S phase) of mitotic division. Cell-
specific markers were used to phenotype the newly born neurons
after longer survival times. We next examined whether the struc-
tural hippocampal defects resulting from prenatal stress had func-
tional consequences on learning abilities. Finally, we examined
whether the reduction in cell proliferation in the dentate gyrus had
an impact on spatial memory, a hypothesis supported by others (17).
Methods
Housing Conditions. Adult virgin Sprague–Dawley female rats (Iffa
Credo) weighing 240 g were housed, 10 days after arrival, in the
presence of a sexually experienced male Sprague–Dawley rat
weighing 400 g. Pregnant females then were randomly assigned to
prenatal stress (PS) or control (C) groups and individually housed
in plastic breeding cages. Animals were allowed ad libitum access
to food and water and were maintained on a constant light兾dark
cycle with constant temperature and humidity.
Prenatal Stress. Stress was performed each day of the last week of
pregnancy from day 15 until delivery. Pregnant females were
individually restrained for 45 min three times a day during the
light phase in plastic transparent cylinders (7-cm diameter,
19-cm long) exposed to bright light. Control pregnant females
were left undisturbed in their home cages. The offspring were
raised by their biological mothers until weaning (21 days after
birth). Only litters of 8–13 pups with similar numbers of males
and females were kept for the study, all other litters having been
eliminated to rule out additional stressors such as removal of the
pups. For each set of experiments, a maximum of two male pups
was taken from each litter to remove any ‘‘litter effects’’ (18).
After weaning, male rats from each experimental group (PS, n ⫽
39; C, n ⫽ 34) were housed in groups of three. One month before
each behavioral study, animals were individually housed.
Weight of Adrenal Glands. To have an index of chronic hypo-
thalamic–pituitary–adrenal (HPA) axis activity (19, 20), adrenal
glands were collected on the day of the animals were killed. They
were immediately weighed and results are expressed as a ratio of
body weight.
Place Navigation Task. Four-month-old rats (C, n ⫽ 10; PS, n ⫽ 10)
were used in this experiment. The apparatus consisted of a circular
swimming pool built of gray plastic (180-cm diameter ⫻ 60-cm
high), which was filled with water at room temperature and made
opaque by the addition of milk. In this task, animals are required
to locate a platform submerged (1.5 cm) in the pool by using only
the spatial cues available within the testing room. During the
habituation phase (no platform), rats (C, n ⫽ 5; PS, n ⫽ 5) were
given one trial per day (of 60 sec) over 3 days. During the testing
phase, the platform was hidden in a fixed location in one of four
quadrants halfway between the sidewalls and the center of the pool.
The animals were given four trials per day (90 sec) over 5 days. As
a control of training, rats of both groups (C, n ⫽ 5; PS, n ⫽ 5) were
handled in parallel with the trained rats but were not submitted to
the water maze task (manipulated groups).
This paper was submitted directly (Track II) to the PNAS office.
Abbreviations: PS group, prenatal stress group; C group, control group; HPA, hypothalam-
ic–pituitary–adrenal; IR, immunoreactivity; GFAP, glial fibrillary acidic protein; NeuN, neu-
ronal nuclei; n.s., not significant.
*V.L. and M.K. contributed equally to this work.
†
To whom reprint requests should be addressed. E-mail: Nora.Abrous@bordeaux.inserm.fr.
The publication costs of this article were defrayed in part by page charge payment. This
article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C.
§1734 solely to indicate this fact.
11032–11037
兩
PNAS
兩
September 26, 2000
兩
vol. 97
兩
no. 20
BrdUrd Injections. Rats were injected with BrdUrd dissolved in
phosphate buffer (21). Animals of different ages received four
injections: once 3 days before death, once 2 days before death,
and twice the day before death (15, 16). To determine the
phenotype of the newly born cells, 3-month-old control and
prenatally stressed rats were killed 2 weeks after the last BrdUrd
injection. For the behavioral study, 4-month-old animals trained
in the water maze were injected once a day, 2 min before the first
trial during the third, fourth, and fifth days of training. The
manipulated groups were injected in parallel. Animals were
killed the day after the last injection.
Histological Procedure. Rats were deeply anesthetized with chloral
hydrate (400 mg兾kg i.p.) and were perfused with 150 ml of
phosphate buffer saline (pH, 7.3) containing heparin (5 ⫻ 10
4
units兾ml), followed by 300 ml of 4% paraformaldehyde in 0.1 M
of phosphate buffer (pH, 7.3). After a 24-h postfixation period
of the brains in paraformaldehyde, 50-
m frontal sections were
cut on a vibratome and collected in PBS (0.1 M; pH, 7.4).
Free-floating sections were processed in a standard immunohis-
tochemical procedure (15, 16, 22). For BrdUrd labeling, sections
were treated with 2 M HCl (30 min at 37°C) and then rinsed in
borate buffer for 5 min (0.1 M; pH, 8.4). They were extensively
washed with PBS, preincubated 45 min with PBS containing
0.3% Triton X-100 and 3% of horse normal serum (blocking
solution), and incubated under agitation for 72 h at 4°C in mouse
monoclonal anti-BrdUrd antibody (1兾200; Dako) diluted in PBS
containing 0.3 Triton X-100 and 1% of horse normal serum. The
sections were then incubated under agitation for2hwitha
biotin-labeled horse anti-mouse IgG antibody (1兾200; Vector
Valbiotec, Paris). Sections from all animals were processed in
parallel and immunoreactivities (IRs) were visualized by the
biotin–streptavidin technique (ABC kit; Dako) with 3,3⬘-
diaminobenzidine as chromogen (10-min incubation).
The phenotype of newly born cells was examined by double-
labeled immunohistofluorescence according to a previously de-
scribed method (22). Briefly, sections were incubated with a rabbit
polyclonal anti-glial fibrillary acidic protein (GFAP) antibody (1兾
10,000; Dako) or with a mouse monoclonal anti-neuronal nuclei
(NeuN) antibody (1兾1,000; Chemicon). Bound anti-GFAP or
-NeuN antibodies were visualized respectively with an Alexa 488
goat anti-rabbit IgG antibody and a Alexa goat anti-mouse IgG
antibody (1兾1,000; Interchem, Montluc¸on, France). Then, sections
were incubated with a rat monoclonal anti-BrdUrd antibody (1兾
1,000; Accurate Scientific, Westbury, NY), and bound anti-BrdUrd
molecules were revealed by using a Cy3-labeled anti-rat IgG
antibody (1兾400; Jackson ImmunoResearch).
Quantitative Evaluation of Staining. For each animal, starting ran-
domly, BrdUrd-IR cells were counted in the left dorsal hippocam-
pus, in one in five sections, 250
m apart. All BrdUrd-IR cells were
counted with a ⫻100 microscope objective in the subgranular layer
and in the granular layer of the dentate gyrus. The sections then
were counterstained and the surface of the granule cell layer was
measured with a Samba 2640 system (Alcatel system; TITN An-
sware, Grenoble, France). For each section, the numerical density
of BrdUrd-IR was calculated by dividing the number of BrdUrd-IR
cells by granule cell layer sectional volume. For each animal, the
mean numerical density of BrdUrd-IR was calculated. The total
number of BrdUrd-IR cells per dentate gyrus then was calculated
by multiplying the numerical density of BrdUrd-IR cells by the
reference volume. The volume of the reference space was calcu-
lated according to the Cavalieri estimation (23) applying the
following formula: V
ref
⫽ a ⫻ t ⫻ s, where a is the mean area of
the granule cell layer of the dentate gyrus, t is the thickness of the
vibratome section (50
m), and s is the total number of sections
through the defined regions of the dorsal hippocampus. The
number of pyknotic cells was determined on hematoxylin-
counterstained sections by using a similar procedure.
The absolute number of granule cells, as assessed morpho-
logically by hematoxylin staining, was determined by using a
Samba 2640 system. For each section, granule cells were counted
in 20 to 25 frames (20
m ⫻ 20
m) at evenly spaced x–y intervals
of 110
mby110
m. We disregarded cells that were in sharp
focus in the uppermost focal plane (optical disector principle).
The numerical density (number of cells兾mm
3
) was calculated
according to the following equation:
N
v
⫽
冘
Q
⫺
h ⫻ a共fra兲 ⫻
冘
P
, [1]
where (Q
⫺
) is the sum of cells, (P) the number of frame, (h) the
disector height, and a(fra) the area of the counting frame. By using
these sampling parameters, the number of optical disectors exam-
ined was 53.67 ⫾ 1.10 [CE(Fi), 13.67 ⫾ 0.22%], the number of cells
counted per dentate gyrus was 221.32 ⫾ 7.15 [CE(Qi), 13.99 ⫾
0.25%], and the CE(N
V
) was 13.69 ⫾ 0.22% (CE is the coefficient
of error). The total number of cells then was calculated by multi-
plying the numerical density by the volume of reference (ref. 23 and
see above).
General Procedures. First experiment: Prenatal stress and cell pro-
liferation. Juvenile, adult, middle-aged, and old rats (C, n ⫽ 20;
PS, n ⫽ 25) were killed the day after the last BrdUrd injection.
These rats were considered ‘‘naive’’ as they were not manipu-
lated except for injections and were not behaviorally tested. The
Fig. 1. Effect of prenatal stress on cell
proliferation in 28-day-, 3-month-, 10-
month-, and 22-month-old male rats. There
was a significant decline of cell proliferation
with increasing age in both control and pre-
natally stressed rats. Furthermore, for the
four ages investigated, prenatal stress consis-
tently decreased cell proliferation as com-
pared with control.
*
, P ⬍ 0.05 in comparison
to age-matched control group;
**
, P ⬍ 0.01.
Notice the different ordinal scale between
data for 28-day-old rats and the other ages
studied.
Lemaire et al. PNAS
兩
September 26, 2000
兩
vol. 97
兩
no. 20
兩
11033
NEUROBIOLOGY
total number of BrdUrd-IR cells per dentate gyrus was counted
on diaminobenzidine-immunostained sections. The absolute
number of granule neurons was determined on the same sections
after counterstaining. The total number of pyknotic cells was
determined on alternate counterstained sections.
Second experiment: Phenotype of the newly born cells. Adult
naive rats (C, n ⫽ 4; PS, n ⫽ 4) were injected four times with
BrdUrd and were killed 2 weeks after the last BrdUrd injection.
The total number of BrdUrd-IR cells per dentate gyrus was
counted on diaminobenzidine-immunostained sections. Double
immunofluorescence was performed and the identity of newly
born cells was examined with cell-specific markers.
Third experiment: Training in the water maze and cell prolifer-
ation.
Four-month-old rats were either trained in the water maze
(trained groups: C, n ⫽ 5; PS, n ⫽ 5) or handled in parallel with
the trained rats but not submitted to the water maze task
(manipulated groups: C, n ⫽ 5; PS, n ⫽ 5). Rats were injected
four times with BrdUrd (as described above) and were killed the
day after the last injection. The total number of BrdUrd-IR cells
per dentate gyrus was counted on diaminobenzidine-
immunostained sections. The absolute number of granule neu-
rons was determined on the same sections after counterstaining.
Results
Effect of Prenatal Stress on Cell Proliferation. We first examined, in
the progeny of stressed mothers, neurogenesis in the dentate gyrus,
a process known to be regulated by corticosterone levels (14, 15,
16). We also examined whether cell proliferation was affected at
specific times during the lifespan, including in senescent rats. Cell
proliferation was evaluated in rats injected with BrdUrd; cells that
incorporated BrdUrd (BrdUrd-IR cells) were revealed by immu-
nohistochemistry. BrdUrd-IR cell number was determined in ju-
venile (28-day-old), adult (3-month-old), middle-aged (10-month-
old), and old (22-month-old) naive rats (Figs. 1 and 2). There was
a significant decline of proliferation as age increased in both control
[H(3, 19) ⫽ 16.75, P ⬍ 0.001; 28 days, ⬎3 months, ⬎10 months, ⬎22
months at P ⬍ 0.05 with a nonparametric Kruskal–Wallis ANOVA]
and prenatally stressed rats [H(3, 25) ⫽ 22.09, P ⬍ 0.001; 28 days,
⬎3 months, ⬎10 months, ⬎22 months at P ⬍ 0.01]. Furthermore,
for the four ages investigated, prenatal stress consistently decreased
cell proliferation {⫺38.4% at 28 days [H(1, 11) ⫽ 4.03, P ⬍ 0.05];
⫺59.3% at 3 months [H(1, 11) ⫽ 7.51, P ⬍ 0.01]; ⫺42.3% at 10
months [H(1, 12) ⫽ 6.56, P ⫽ 0.01]; and ⫺55.2% at 22 months of
age [H(1, 10) ⫽ 4.69; P ⬍ 0.05]}. These differences were not related
Fig. 2. Illustration of BrdUrd-labeled cells in the dentate gyrus. Photomicrographs of BrdUrd immunoreactivity in frontal section of the dentate gyrus in a
control (a) and a prenatally stressed (b) rat . (Scale ⫽ 1.5 cm for 500
m.) (c) Confocal illustration of BrdUrd-labeled cells in the dentate gyrus showing that newborn
cells were mainly neurons. Indeed, BrdUrd-IR cells (red nuclear stain, Cy3) colocalized with the neuronal marker NeuN (green stain) as indicated by the yellow
nuclear staining. (d) Newly born cells are rarely astrocytes because BrdUrd-IR cells (red nuclear stain) is not stained for the astroglial marker GFAP (green stain).
11034
兩
www.pnas.org Lemaire et al.
to differences in the volume of the granule cell layer [28 days: H(1,
11) ⫽ 0.33, not significant (n.s.); 3 months: H(1, 11) ⫽ 2.13, n.s.; 10
months: H(1, 12) ⫽ 0.10, n.s.; 22 months: H(1, 10) ⫽ 0.64, n.s.].
Identity of Newly Born Cells. To examine the phenotype of Br-
dUrd-IR cells born in the dentate gyrus of control (Fig. 2a)or
prenatally stressed (Fig. 2b) rats, animals were allowed to survive
for 2 weeks after the last BrdUrd injection. Newborn cells, pheno-
typed by analysis of double-labeled immunofluorescence sections,
were mainly neurons and rarely astrocytes (Fig. 2 c and d). Indeed,
72 ⫾ 2% and 70 ⫾ 2% of BrdUrd-IR cells were immunoreactive for
NeuN in control and prenatally stressed rats, respectively. These
ratios were not significantly different among groups [H(1, 8) ⫽ 0.08;
n.s.]. On the other hand, 20 ⫾ 1.6% and 20 ⫾ 1.6% of BrdUrd-IR
cells were immunoreactive for GFAP in control and prenatally
stressed rats, respectively. These ratios were not significantly dif-
ferent among groups [H(1, 8) ⫽ 2.08, n.s.]. The double labeling was
verified with confocal microscopy. Exploration in the Z plane
revealed that bound anti-GFAP antibodies were present through-
out the section thickness, but bound anti-NeuN antibodies pene-
trate only a portion of the slice. Thus, the quantification of
BrdUrd-NeuN double-labeled cells was underestimated.
Effect of Prenatal Stress on the Total Number of Granule Neurons.
Because prenatal stress reduces cell proliferation in the dentate
gyrus early in life and in aged rats, it was of interest to evaluate the
consequence of this reduction on the total granule cell number.
Although the total number of granule neurons was not different
among different ages in control rats [H(3, 20) ⫽ 6.61, n.s.] (Fig. 3),
there was a progressive decline of total granule cell number with
increasing age after prenatal stress [H(3, 25) ⫽ 14.24, P ⬍ 0.01; 28
days, 3 months, ⬎10 months, ⬎22 months at P ⬍ 0.01]. Prenatal
stress decreased the total number of granule neurons beginning at
3 months of age [H(1, 11) ⫽ 4.80, P ⬍ 0.05] and continued until 10
[H(1, 12) ⫽ 4.33, P ⬍ 0.05] and 22 [H(1, 10) ⫽ 4.69, P ⬍ 0.05]
months of age.
The proportion of newborn cells in the dentate gyrus was
computed in percentage of total granule cells and represented
5.8% ⫾ 0.7 in 28-day-old control rats (the largest percentage) to
0.08% ⫾ 0.02 in 22-month-old prenatally stressed rats (the
lowest percentage). Thus, the low proportion of newborn cells
could explain why the deficit in cell proliferation is detectable
before the decrease in the total number of granule cells in
prenatally stressed rats.
Effect of Prenatal Stress on the Total Number of Pyknotic Cells. To
determine whether the observed alteration of granule neurons
number was related specifically to differences in cell proliferation,
we evaluated the consequence of prenatal stress on pyknotic cell
number (Fig. 4). The degenerating profiles were characterized by a
condensed chromatin and a light or absent cytoplasm. Prenatal
stress did not influence the total number of pyknotic cells at all ages
tested [3 months, H(1, 11) ⫽ 0.80, n.s.; 10 months, H(1, 11) ⫽ 0.38,
n.s.; 22 months, H(1, 10) ⫽ 0.11, n.s.]. In contrast, there was an
age-related increase in degenerating profile in both control [H(2,
12) ⫽ 9.9, P ⬍ 0.01; 3 months, ⬎10 months, ⬎22 months at P ⬍
0.01] and prenatally stressed [H(2, 19) ⫽ 22.09, P ⬍ 0.01; 3 months,
⬎10 months, ⬎22 months at P ⬍ 0.01] rats.
HPA Axis Activity. The weight of adrenal glands (expressed as a
ratio of the body weight) previously has been demonstrated to be
a reliable index of the chronic activity of the HPA axis because
a hypertrophy of the adrenal glands is observed in conditions of
chronic HPA axis hyperactivity (19, 20). As shown on Table 1,
this ratio was higher in prenatally stressed rats when compared
with control rats [H(1, 11) ⫽ 4.03, P ⬍ 0.05; H(1, 11) ⫽ 7.50, P ⬍
0.01; H(1, 12) ⫽ 3.69, P ⬍ 0.05)] indicating a chronic up-
regulation of the HPA axis.
Effect of Prenatal Stress on Spatial Learning. We next examined
whether such structural hippocampal defects resulting from
prenatal stress had functional correlates on spatial memory and
whether cell proliferation was influenced by training. We tested
4-month-old rats for spatial memory in a water maze.
Rats were injected with BrdUrd before the testing session on
days 3, 4, and 5 of learning. At 24 h after the last BrdUrd
injection, all animals were perfused and cell proliferation was
determined. There were significant differences in the rate of
acquisition between control and prenatally stressed rats (Fig. 5).
The latency to find the platform by the control group was
significantly less than that observed for the prenatally stressed
group [H(1, 10) ⫽ 6.82, P ⬍ 0.001]. The distance covered by
control animals was shorter than that covered by prenatally
stressed rats (data not shown, P ⬍ 0.05).
As shown in Fig. 6, acquisition of this spatial task increased
hippocampal neurogenesis in control rats [H(1, 10) ⫽ 4.81, P ⬍
0.05] but not in prenatally stressed rats [H(1, 10) ⫽ 0.88, n.s.]. As
previously shown, prenatal stress decreased the total number of
Fig. 3. Effect of prenatal stress on total number of granule cells in 28-day-,
3-month-, 10-month-, and 22-month-old male rats. Although the total num-
ber of granule cells was not different among different ages in control rats,
there was a progressive decline of total granule cell number with increasing
age after prenatal stress. Interestingly, this decline in prenatally stressed rats
was observable from 3 months of age continuing until 22 months of age.
*
, P ⬍
0.05 in comparison to age-matched control group.
Fig. 4. Effect of prenatal stress on number of pyknotic cells in 3-month-,
10-month-, and 22-month-old male rats. With increasing age, there was a
progressive increase in the total number of pyknotic cells. There was no effect
of prenatal stress.
Lemaire et al. PNAS
兩
September 26, 2000
兩
vol. 97
兩
no. 20
兩
11035
NEUROBIOLOGY
BrdUrd-IR cells. The number of BrdUrd-IR cells in prenatally
stressed rats was identical for manipulated and trained animals.
Furthermore, training did not influence the total number of cells
in control [H(1, 10) ⫽ 1.84, n.s.] and prenatally stressed [H(1,
10) ⫽ 0.53, n.s., data not shown] rats.
Discussion
The aims of the present study were to investigate (i) whether
prenatal events alter hippocampal plasticity, (ii) to which extent
these possible alterations may undermine behavioral perfor-
mances in spatial memory tasks depending on the hippocampal
formation, and (iii) whether neurogenesis in the dentate gyrus
and learning process can be related. The results show that aging
is accompanied by a progressive decline in cell proliferation. In
addition, prenatally stressed offspring exhibited a greater reduc-
tion in hippocampal cell proliferation by approximately 45% in
all age groups tested. These animals also displayed delayed
learning in a spatial memory task. Furthermore, learning a
spatial memory task stimulated cell proliferation in the dentate
gyrus of control rats whereas prenatal stress altered the ability
of progenitor cells to divide in response to a learning situation.
Prenatal stress induced structural abnormalities in the hippocam-
pal formation. Our results show for the first time that prenatal stress
affects cell proliferation from adolescence until aging. With in-
creasing age, there was a decline in cell proliferation in the dentate
gyrus of control male rats, an effect consistent with previous studies
(24, 25). Prenatally stressed animals displayed a greater reduction
in cell proliferation during aging, suggesting that the early stressful
experience accelerated the age-related decline in hippocampal
plasticity. It remains to be determined whether the age-related
decline in neurogenesis and the effects of prenatal stress are caused
by an alteration in total number of stem cells (or progenitors cells)
or differences in cell cycle length. Interestingly, these effects were
specific to neurogenesis because pyknosis was not modified by
prenatal stress at all ages tested. Because glucocorticoids inhibit
hippocampal cell proliferation (14–16), the increased HPA axis
activity of prenatally stressed animals (1, 3, 5), which is confirmed
here by increased adrenal mass, could explain their reduced neu-
rogenesis. The toxic effects of prolonged secretion of corticosterone
have been extensively studied on hippocampal pyramidal cells (26,
27) but it is not known whether the age-related increase in pyknosis
within the dentate gyrus is related to the activity of the HPA axis.
The lack of effect of prenatal stress on pyknosis suggests the
existence of a complex relationship between corticosterone and cell
death. The maintenance of the effect of prenatal stress on neuro-
genesis during aging can thus be related to the ‘‘glucocorticoid
cascade’’ hypothesis, which proposes that stressful experiences,
through an excessive release of corticosterone, are responsible for
alterations in the structure and function of the hippocampal for-
mation in senescent animals (28).
The effect of aging on total cell number was complex. Indeed,
a decline in cell number was observed between 3 and 10 months
of age but this tendency was reversed later in life. The lack of an
overall effect of aging on granule neuron number was not
consistent with the fact that a net decline in neurogenesis and an
increase in cell death was observed in the same animals. There
is conflicting evidence on the effects of aging on the number of
granule neurons, and the reasons for which variations in this
number may or may not have occurred remain unknown (29–32).
In prenatally stressed rats, despite the early impact of the
manipulation on neurogenesis, an alteration of the total granule
cell number could be observed beginning only at 3 months of age.
This delayed decline in total granule cell number could be
related to the fact that the percentage of granule neurons
produced in adulthood is relatively small as compared with the
1 or 2 millions mature granule neurons. Thus, it is reasonable to
suppose that a relatively long period of neurogenesis reduction
is necessary to affect the total granule cell number and that the
deficits in newborn cells between the beginning of life and 28
postnatal days is negligible as compared with the total pool of
granule neurons. Later in life, deficits in total granule cell
number appear in prenatally stressed rats; this result may be a
consequence of an alteration in cell proliferation, a hypothesis
reinforced by the lack of influence of prenatal stress on pyknosis.
It has been shown that prenatal stress decreased synaptic density
in the CA3 area, the projection site of the granule neurons (33)
and hippocampal density of nitric oxide-producing neurons (34).
Taken together, these data indicate that early experiences alter
Fig. 5. Spatial learning in a water maze. The latency to find the platform by
the control group was significantly less than that observed for the prenatally
stressed group. (Inset) The means over the 5 days of testing.
**
, P ⬍ 0.01 in
comparison to control group.
Fig. 6. Effect of training in the water maze on cell proliferation in control
and prenatally stressed rats. Acquisition of a spatial task increased the rate of
hippocampal neurogenesis in control rats but not in prenatally stressed rats.
Prenatal stress decreased the total number of BrdUrd-IR cells and, in this
group, the number of newly born cells was identical for manipulated as well
as for trained animals. (Inset) Correlation between cell proliferation and mean
latency to reach the platform.
*
, P ⬍ 0.05 as compared with manipulated
control group; °, P ⬍ 0.05 and °°, P ⬍ 0.01 as compared with control group.
Table 1. Effect of prenatal stress on adrenal weight
Group 28 Days 3 Months 10 Months
Control 260.49 ⫾ 7.74 105.41 ⫾ 8.51 64.66 ⫾ 1.49
Prenatal stress 297.88 ⫾ 12.91* 159.36 ⫾ 6.10** 101.00 ⫾ 14.24*
Data are expressed as a ratio of body weight (mg兾kg of body weight).
*
, P ⬍
0.05 and
**
, P ⬍ 0.01 as compared to controls.
11036
兩
www.pnas.org Lemaire et al.
the structure of the hippocampal formation. This finding is of
particular interest because the dentate gyrus is associated with
spatial learning and memory (35, 36), which clearly are affected
in prenatally stressed rats (present data and refs. 2, 4, 5).
We show that training in the water maze enhances cell
proliferation in control rats. Previous studies have reported that
survival of newly born cells is enhanced by exposure to an
enriched environment, which also increased spatial learning in
the water maze (37), and by exposure to training in the water
maze (8). So far, the only condition that seems to increase cell
proliferation is voluntary exercise on a running wheel (38). We
show here that training in the water maze enhances cell prolif-
eration in control rats; however, this effect was not seen in a
previous study (38). This discrepancy could be related to the
timing of BrdUrd injections. In this experiment, BrdUrd was
injected at the end of the learning phase, when control rats begin
to reach asymptotic level of performance, instead of throughout
testing. These days were chosen because they correspond to a
period of learning consolidation in which the hippocampal
formation is activated (39). The stimulatory effect of training on
cell proliferation observed in control rats is not caused by an
increased motor activity because these animals cover a shorter
distance than prenatally stressed rats do. Training by itself is not
a sufficient condition to increase cell proliferation because no
such stimulatory effect of training was observed in prenatally
stressed rats. Thus, low cell proliferation is associated with poor
behavioral performance whereas high cell proliferation is asso-
ciated with good learning capabilities. Taken together with the
decreased cell proliferation observed after prenatal stress, the
behavioral deficits induced by prenatal stress suggest an enabling
role for granular cell proliferation in the dentate gyrus to
facilitate spatial memory performances. These results confirm
the importance of the network into which these cells are inserted,
because newborn neurons in the granule cell layer establish
connections with the CA3 pyramidal neurons (10), which are
known to be implicated in spatial memory (35, 36). Reciprocally,
information processing influences brain structures, and spatial-
learning-induced hippocampal plasticity takes place in the GCL-
CA3 network.
The effect of prenatal stress on spatial learning may be the
consequence of a deficit in neurogenesis, which can itself result
from the dysfunction of the HPA axis. Indeed, cognition is
modulated by corticosterone in a complex way (40), and high
levels of corticosterone impair learning and memory (27, 41).
Exposure to the water maze increases corticosterone secretion
(42) and prenatally stressed animals show a delayed habituation
of the corticosterone response to repeated exposure to stress
(43). Thus, it seems reasonable to hypothesize that the prenatal
stress-induced cognitive impairments may result from a pro-
longed corticosterone secretion that inhibits cell proliferation.
Taken together, the results of this study allow us to propose a
pathophysiological path where the effect of prenatal stress on
spatial learning may be the consequence of a deficit in neuro-
genesis that could results from HPA axis dysfunction. Although
a direct effect of corticosterone on neurogenesis is possible,
another hypothesis is that corticosterone acts through serotonin
to reduce neurogenesis. Indeed, serotonin stimulates neurogen-
esis (44), but serotonin levels are reduced by corticosterone (45).
Thus, the decreased serotonin levels observed in prenatally
stressed rats (33) provide further evidence that corticosterone
could be at the origin of the effect of prenatal stress that we
report here. It can be hypothesized that the present results
obtained following maternal restraint stress could be generalized
to other procedures. Indeed, other types of maternal gestational
stress were shown to have similar effects on endocrine and
behavioral parameters (1).
In conclusion, the major impact of our data lies in the
demonstration that deleterious environmental conditions occur-
ring early in life have profound effects on neurogenesis in the
dentate gyrus, an index of hippocampal plasticity, and that these
alterations are associated with impaired performances in a
spatial memory task. Neurogenesis, cell number, and cognitive
capabilities are altered from adulthood to senescence in prena-
tally stressed rats. The fact that stressful experiences during
development could have a long-term effect on hippocampal
function by directly altering its structure recently has been
suggested by Gould and Tanapat (17). Thus, it may be hypoth-
esized that elevated corticosterone levels throughout the lifespan
of these animals could lead to cognitive deficits through devel-
opmental inhibition of neurogenesis (27). Our results reinforce
the hypothesis that many psychopathological affections have
their origin in early developmental influences. More generally,
they show the heuristic value of accurate animal models to better
understand the mechanisms by which early stress and epigenetic
risk factors promote learning disabilities in children and in
age-related disorders, such as Alzheimer’s disease (46).
We thank Drs.G. Rougon, W. Mayo, and P. V. Piazza for helpful comments,
and Dr. M.-F. Montaron and Mrs. C. Brechenmacher for their help with the
confocal analysis. The technical help of Mrs. M. C. Donat, Mrs. J. M.
Claustrat, and O. George is acknowledged. Supported by Institut National
de la Sante´ et de la Recherche Me´dicale, Centre National de la Recherche
Scientifique (98.72.017), and University of Bordeaux II.
1. Weinstok, M. (1997) Neurosci. Biobehav. Rev. 21, 1–10.
2. Szuran, T., Zimmerman, E. & Welzl, H. (1994) Behav. Brain Res. 65, 153–155.
3. Henry, C., Kabbaj, M., Simon, H., Le Moal, M. & Maccari, S. (1994) J. Neuroendocrinol. 6,
341–345.
4. Lordi, B., Poroais, P., Mellier, D. & Caston, J. (1997) Physiol. Behav. 62, 1087–1092.
5. Valle´e, M., Maccari, S., Dellu, F., Le Moal, M., Simon, H. & Mayo, W. (1999) Eur.
J. Neurosci. 11, 2906–2916.
6. Squire, L. R. (1992) Psychol. Rev. 99, 195–231.
7. Altman, J. & Das, G. D. (1965) J. Comp. Neurol. 124, 319–336.
8. Gould, E., Beylin, A., Tanapat, P., Reeves, A. &Shors, T. J. (1999) Nat. Neurosci. 2, 260–265.
9. Gould, E., Tanapat, P., Hastings, N. B. & Shors, J. T. (1999) Trends Cognit. Sci. 3, 186–191.
10. Stanfield, B. B. & Trice, J. E. (1988) Exp. Brain Res. 72, 399–406.
11. Cameron, H. A., Wooley, C. S., McEwen, B. S. & Gould, E. (1993) Neuroscience 61, 203–209.
12. Schwegler, H. & Crusio, W. E. (1995) Behav. Brain Res. 67, 29–41.
13. Scho¨pke, R., Wolfer, D. P., Lipp, H. P. & Leisinger-Trigona, M. C. (1991) Hippocampus 1,
315–328.
14. Gould, E., Cameron, H. A., Daniels, D. C., Woolley, C. S. & McEwen, B. S. (1992) J. Neuroci.
12, 3642–3650.
15. Rodriguez, J. J., Montaron, M. F., Petry, K. G., Aurousseau, C., Marinelli, M., Premier, S.,
Rougon, G., Le Moal, M. & Abrous D. N. (1998) Eur. J. Neurosci. 10, 2994–3006.
16. Montaron, M. F., Petry, K. G., Rodriguez, J. J., Marinelli, M., Aurousseau, C., Rougon, G.,
Le Moal, M. & Abrous, D. N. (1999) Eur. J. Neurosci. 11, 1479–1485.
17. Gould, E. & Tanapat, P. (1999) Biol. Psychiatry 46, 1472–1479.
18. Chapman, R. H. & Stern, J. (1978) J. Comp. Physiol. Psychol. 92, 1074–1083.
19. Akana, S. F., Shinsako, J. & Dallman, M. F. (1983) Endocrinology 113, 2232–2237.
20. Lemaire, V., Taylor, G. T. & Morme`de, P. (1997) Psychoneuroendocrinology 22, 563–573.
21. Nowakowsky, R. S., Lewin, S. B. & Miller, M. W. (1989) J. Neurocytol. 18, 311–318.
22. Lemaire, V., Aurousseau, C., Le Moal, M. & Abrous, D. N. (1999) Eur. J. Neurosci. 11, 4006–4014.
23. Howard, C. V. & Reed, M. G. (1998) Unbiased Stereology. Three-Dimensional Measurement
in Microscopy (Springer, New York).
24. Kuhn, H. G., Dickinson-Anson, H. & Gage, F. H. (1996) J. Neurosci. 16, 2027–2033.
25. Seki, T. & Arai, Y. (1995) NeuroReport 6, 2479–2482.
26. Woolley, C., Gould, E. & McEwen, B. S. (1990) Brain Res. 531, 225–231.
27. Sapolsky, R. M., Krey, L. C. & McEwen, B. S. (1985) J. Neurosci. 5, 1222–1227.
28. Sapolsky, R. M. (1992) Stress, the Aging Brain, and the Mechanisms of Neuron Death (MIT
Press, Cambridge, MA).
29. Rapp, P. R. & Gallagher, M. (1996) Proc. Natl. Acad. Sci. USA 93, 9926–9930.
30. Rasmussen, T., Schliemann, T., Sorensen, J. C., Zimmer, J. & West, M. J. (1996) Neurobiol.
Aging 17, 143–147.
31. Landfield, P. W., Baskin, R. & Pitler, T. (1981) Science 214, 5881–5884.
32. Boss, B. D., Peterson, G. M. & Cowan, W. M. (1985) Brain Res. 338, 144–150.
33. Yahashi, A., Nagaoka, M., Yamada, K., Ichitani, Y., Miake, Y. & Okado, N. (1998) Int. J.
Dev. Neurosci. 16, 209–216.
34. Vaid, R. R., Yee, B. K., Shalev, U., Rawlins, J. N., Weiner, I., Feldon, J. & Totterdell, S.
(1997) J. Neurosci. 17, 5599–5609.
35. Jung, M. W. & McNaughton, B. L. (1993) Hippocampus 3, 165–182.
36. McNaughton, B. L., Barnes, C. A., Meltzer, J. & Sutherland, R. J. (1989) Exp. Brain Res. 76, 485–496.
37. Kempermann, G., Kuhn, G. F. & Gage, F. (1997) Nature (London) 386, 493–495.
38. Van Praag, H., Kempermann, G. & Gage, F. H. (1999) Nat. Neurosci. 2, 266–270.
39. Bontempi, B., Laurent-Demir, C., Destrade, C. & Jaffard, R. (1999) Nature (London) 400, 671–675.
40. De kloet, E. R, Oitzl, M. S. & Joe¨ls, M. (1999) Trends Neurosci. 22, 422–426.
41. De Quervain, D. J.-F., Roozendaal, B. & McGaugh, J. L. (1998) Nature (London) 394, 787–790.
42. Sandi, C., Loscertales, M. & Guaza, C. (1997) Eur. J. Neurosci. 9, 637–642.
43. Fride, E., Dan, Y., Feldon, J., Halevy, G. & Weinstock, M. (1986) Physiol. Behav. 37, 681–687.
44. Brezun, J. M. & Daszuta, A. (1999) Neuroscience 89, 999–1002.
45. Chaouloff, F. (1955) Fundam. Clin. Pharmacol. 9, 219–233.
46. Leckman, J. F. (1999) Am. J. Psychiatry 156, 1495–1448.
Lemaire et al. PNAS
兩
September 26, 2000
兩
vol. 97
兩
no. 20
兩
11037
NEUROBIOLOGY
