Paradoxical effects of learning the Morris water maze on adult hippocampal neurogenesis in mice may be explained by a combination of stress and physical activity

Authors

  • D. Ehninger,

    1. Max Delbrück Center for Molecular Medicine (MDC) Berlin-Buch, and Volkswagenstiftung Research Group, Department of Neurology, Charité University Medicine Berlin, Berlin, Germany
    Search for more papers by this author
    • * 

      Present address: UCLA Medical Center, Departments of Neurobiology, Psychiatry and Psychology, 695 Charles Young Drive South, Los Angeles, CA 90095, USA.

  • G. Kempermann

    Corresponding author
    1. Max Delbrück Center for Molecular Medicine (MDC) Berlin-Buch, and Volkswagenstiftung Research Group, Department of Neurology, Charité University Medicine Berlin, Berlin, Germany
    Search for more papers by this author

G. Kempermann, Max Delbrück Center for Molecular Medicine (MDC) Berlin-Buch, Robert-Rössle-Str. 10, 13125 Berlin, Germany. E-mail: gerd.kempermann@mdc-berlin.de

Abstract

Studies in rats that assessed the relation of hippocampus-dependent learning and adult hippocampal neurogenesis suggested a direct regulatory effect of learning on neurogenesis, whereas a similar study in mice had not found such causal link. We here report a substantial decrease of BrdU-positive cells and other measures of adult hippocampal neurogenesis in mice trained in the hidden (HID) or cued version (VIS) of the Morris water maze as compared to untrained animals (CTR). Particularly, cells on advanced stages of neuronal development contributed to this decrease, whereas earlier progenitors (type 2 cells) were not diminished in HID, but were diminished in VIS as compared to CTR. The differential regulation of type 2 cells in HID and VIS may have been caused by a different degree of physical activity, given that a time-yoked control group did not differ from HID, and type 2 cells reportedly constitute the proliferative dentate gyrus population that primarily responds to physical activity. The decrease of hippocampal neurogenesis by water maze training was reversible by pre-exposing animals to the water maze prior to training, suggesting that stress associated with training may have caused the acute downregulation of adult neurogenesis. We propose that in mice the Morris water maze does not provide a pure enough learning stimulus to study the presumed effects of ‘learning’ on adult neurogenesis. In addition, however, our data show that physical activity that is intricately linked to many cognitive tasks in rodents might play an important role in explaining effects of learning on cellular hippocampal plasticity.

A link between adult hippocampal neurogenesis and learning as a hippocampus-dependent process has been suggested by many recent studies (Ambrogini et al. 2000, 2004; Döbrössy et al. 2003; Drapeau et al. 2003; Feng et al. 2001; Gould et al. 1999; Kempermann & Gage 2002; Kempermann et al. 1997; Leuner et al. 2004; van Praag et al. 1999a; Prickaerts et al. 2004; Schmidt-Hieber et al. 2004; Shors et al. 2001, 2002), and theoretical concepts trying to explain the functional relevance of adult hippocampal neurogenesis have been proposed (Chambers et al. 2004; Deisseroth et al. 2004; Kempermann & Gage 2002; Kempermann et al. 2004b).

In 1999, two unrelated studies examined the impact of hippocampus-dependent learning on hippocampal neurogenesis and came to contradictory conclusions (Gould et al. 1999; van Praag et al. 1999b). Whereas Gould et al. found that learning a hippocampal learning task increased neurogenesis, van Praag et al. did not find any quantitative effect of learning the Morris water maze on hippocampal neurogenesis. As discussed by William T. Greenough, comparing the two studies, one explanation of these incongruent results might lie in the study design (Greenough et al. 1999). Gould et al. labelled dividing cells prior to water maze training so that learning occurred during a late developmental state of the labelled population. In contrast, van Praag et al. used a prolonged training protocol and started labelling proliferating cells at the same time as training began. Consequently, the diverging results of the two studies could be explained by assuming a sensitive period of hippocampal neurogenesis to learning that starts only on relatively late developmental stages.

More recently, Döbrössy et al. reported that different phases of learning in the Morris water maze affected hippocampal neurogenesis differently (Döbrössy et al. 2003). Furthermore, Döbrössy et al. trained rats in the water maze at different time-points after BrdU injections. However, as BrdU application labels proliferating hippocampal cells of different maturity concomitantly (cells of different maturity are capable of proliferation) (Brandt et al. 2003; Filippov et al. 2003; Kronenberg et al. 2003; Steiner et al. 2004), this approach did not allow to dissect a potentially differential effect of learning the Morris water on progenitors of different maturity.

Distinct developmental steps of adult hippocampal neurogenesis can, however, be identified using sets of antigens that are differentially expressed by developing hippocampal neurons (Kempermann et al. 2004a). Importantly, different types of stimuli (physical activity and living under the conditions of an enriched environment) have been shown to differentially affect discrete steps of neuronal development in the context of adult hippocampal neurogenesis (Kronenberg et al. 2003).

We thus intended to use a complementary approach to that by Döbrössy et al. and exploit the stage-specific characteristics of adult-born hippocampal cells to examine whether learning the Morris water maze affects developmental stages of adult hippocampal neurogenesis differentially. For this purpose, we used Nestin-GFP reporter mice (Filippov et al. 2003; Fukuda et al. 2003; Yamaguchi et al. 2000) that, in combination with additional antigenic markers, allow to distinguish distinct developmental steps of adult hippocampal neurogenesis (Kempermann et al. 2004a). Following the hypothesis that learning may exert regulatory influences on adult hippocampal neurogenesis during advanced developmental stages, we set out with a study design closely resembling that of Gould et al. (1999). In particular, we examined the early postmitotic phase of granule cell development, which in the mouse is characterized by a transient expression of calretinin, because most of the quantitatively measurable regulation of ‘survival’ occurs on this stage (Brandt et al. 2003; Kempermann et al. 2004a).

Materials and methods

Animals and experimental design

The experimental design is outlined in Fig. 1.

Figure 1.

Experimental design. Experiment 1: mice (HID, VIS, CTR) were injected with BrdU at days −1 and 0, exposed to the Morris water maze at days 5–8 (six trials daily) and perfused at day 9 of the experiment. Experiment 2: mice (HID, HID yoked) were injected with BrdU at days −1 and 0, exposed to the Morris water maze at days 5–8 (six trials daily) and perfused at day 9 of the experiment. Experiment 3: at days −7 to −5 mice (HID, HID yoked, VIS, VIS yoked, CTR) were pre-exposed to the Morris water maze, with the platform being removed (three trials daily) prior to BrdU injections (days −1 and 0), Morris water maze training (days 5–8) and perfusion (day 9). HID, hidden version of the Morris water maze; VIS, visible version of the Morris water; HID yoked, time-yoked control to HID; VIS yoked, time-yoked control to VIS; CTR, standard-housed control group.

Experiment 1

Twenty-one female Nestin-GFP reporter mice (Yamaguchi et al. 2000) (C57BL/6, 10 weeks of age at the beginning of the experiment) were randomly assigned to one of the following experimental groups: Morris water maze hidden version (HID), Morris water maze cued version (VIS) and standard laboratory conditions (CTR). At day −1 and day 0 of the experiment, animals were injected intraperitoneally with 5-bromo-2-deoxyuridine (BrdU; Sigma, Munich, Germany, 50 µg/g body weight) resolved in sterile 0.9% NaCl solution twice daily.

Experiment 2

To address the possibility that differentially regulated hippocampal cell genesis in HID and VIS of experiment 1 was due to a different amount of physical activity, we assigned 14 female Nestin-GFP reporter mice (C57BL/6 background, approximately 10 weeks old) either to the hidden version of the Morris water maze (n = 7) or to a time-yoked control group (n = 7). BrdU (50 µg/g body weight) was administered twice daily at day −1 and day 0 of the experiment, analogous to the application scheme described above.

Experiment 3

To assess the possibility that decreases in hippocampal cell genesis in HID and VIS of experiment 1 were due to stress associated with water maze training, we conducted an analogous experiment this time including pre-exposure of animals to the Morris water maze setting prior to training with the rationale to decrease stress at the actual training. Forty female C57BL/6 mice (10 weeks old at the start of the experiment) were randomly assigned to one of the following groups (each n = 8): HID, time-yoked control to HID, VIS, time-yoked control to VIS, standard housed group (CTR). BrdU administration (50 µg/g body weight) was twice daily as described above at day −1 and day 0 of the experiment. Animals were pre-exposed to the Morris water maze at days −7 to −5 of the experiment. Pre-exposure took place in the same room using the same apparatus as during the actual training and consisted of three trials per day (each lasting 2 min, intertrial interval about 2 h). During these trials, the platform was removed from the pool to prevent learning the platform location prior to training.

Use of the animals in this study followed all applicable local and federal regulations. All mice lived in regular cages with food and water ad libitum and 12/12 h light/dark cycle (light: 0600 to 1800 hours). After arrival at the animal colony, mice were allowed to adapt to the new environment for 2 weeks. Prior to experimental interventions, animals were handled by the experimenter for five consecutive days (for 1–2 min each). Because there is a small, but significant, effect of time-of-day on the activity-dependent regulation of adult neurogenesis (Holmes et al. 2004) and stress responses show circadian fluctuations, we attempted to perform water maze testing exactly at the same time every day. Because the total testing period is longer on the first trials than on consecutive days, a small influence of circadian mechanisms on the results cannot be excluded. To minimize these, the order in which the mice were tested was changed every day.

Morris water maze

Animals were subjected to Morris water maze training at days 5–8 of the experiments (days 5–8 after the last BrdU injection). We followed the widely used protocol devised by Wolfer et al. (1998). Six trials of training each maximally lasting for 2 min were given each day. The circular pool was 1.88 m in diameter and contained water (temperature 23–25 °C) that was made opaque with non-toxic white paint. Latencies to reach the platform and swim paths were recorded with an automatic video tracking system (Ethovision; Noldus, Freiburg, Germany).

Animals trained in the hidden version of the task (HID) were exposed to the water maze that contained a platform that was submerged 1.5 cm below the water surface and kept at a constant location within the pool during the first 3 days of training (day 5–7). At the fourth day of training (experimental day 8), the platform was removed from that location and invisibly placed in the quadrant opposite to the former (target quadrant). The animals were introduced to the pool using a randomly alternating sequence of six defined entry positions. The experimenter introducing the animal to the pool remained at the entry position during the entire trial, thus being in a variable spatial location to the platform and not allowing any conclusions about the location of the platform to be drawn. Each trial lasted until the animal had either located the hidden platform or was stopped after 2 min, which ever occurred first. In each case, the animal was left for 20 seconds on the hidden platform before it was removed from the pool and exposed to several minutes of infrared radiation to avoid hypothermia. To evaluate learning of the spatial location of the platform, latencies to reach the platform (in seconds)/total length of swim path (in cm) were compared between trials. Additionally, the time spent in the target quadrant on the probe trials was used as an indicator of targeted searching for the platform.

In the cued version of the task (VIS), the platform visibly loomed over the water surface (about 2 cm) and its location was altered from trial to trial to avoid learning of a platform location and force the animals to use local cues to find the platform. For the cued trials, the platform was coloured dark and thus contrasted to the white coloured water, facilitating its recognition. Mice were trained on the cued version for 4 days, six trails a day. The further procedure was analogous to the one described above for the hidden version of the task.

In experiments 2 and 3, time-yoked control groups were used. Each individual of the HID and VIS group was matched to a time-yoked animal that was exposed to the water maze for the same amount of time. For these trials, however, the platform was removed from the pool prior to training. Otherwise treatment did not differ from the respective experimental group. In experiment 3, the mice were pre-exposed to the Morris water maze as described above.

Tissue preparation

Animals were killed with an overdose of ketamine and transcardially perfused with 0.9% NaCl solution followed by 4% paraformaldehyde in cold 0.1 m phosphate buffer (pH 7.4). For postfixation, brains were kept in 4% paraformaldehyde at 4 °C for 24 h and then transferred in 30% sucrose. For the generation of coronal section series (40 µm), brains were mounted on a dry-ice-cooled copper block that was attached to a sliding microtome (Leica, Bensheim, Germany). Sections were stored in a cryoprotectant solution (25% ethylene glycol, 25% glycerine and 50% 0.1 m phosphate buffer; v/v) at −20 °C. The sections were stained using free-floating immunohistochemistry and prepared for BrdU detection by incubation in 2N HCl for 30 min at 37 °C and subsequent washing in 0.1 m borate buffer (pH 8.5) for 10 min.

Antibodies

All antibodies were diluted in Tris-buffered saline (TBS; pH 8.4) containing 0.1% Triton-X-100 and 3% donkey serum. We here used the following primary antibodies: rat anti-BrdU (Harlan Seralab, Leicestershire, UK) 1:500, mouse anti-NeuN (Chemicon, Temecula, CA) 1:100, goat anti-doublecortin (Santa Cruz Biotech, Santa Cruz, CA) 1:200, goat anti-calretinin (Chemicon) 1:200 and rabbit anti-GFP (Abcam, Cambridge, UK) 1:200. Secondary antibodies, raised in donkey and conjugated with either FITC, Rhodamine X, CY-5 or Biotin were purchased from Jackson Laboratories (distributor: Dianova) and diluted 1:250 for immunohistochemistry.

Immunohistochemistry and microscopy

For phenotyping of BrdU-immunoreactive cells, every twelfth section of the coronal section series was subjected to triple-fluorescence labelling. After pretreatment for BrdU detection (see above) and a blocking step with TBS-plus (containing 3% donkey serum and 0.1% Triton-X-100), sections were incubated in the respective primary antibody at 4 °C for 48 h. After washing sections in TBS and TBS-plus, we exposed them to the respective fluorochrome-conjugated secondary antibodies for 4 h at room temperature in the dark. Labelled sections were mounted in polyvinyl alcohol with diazabicyclo-octane as antifading substance.

For quantification of BrdU- and calretinin-immunoreactive cells, every sixth section of coronal section series was subjected to immunohistochemical staining. Here, primary antibodies were recognized with biotinylated secondary antibodies and visualized using the peroxidase method (ABC system, Vectastain, Vector Laboratories, Peterborough, UK) and nickel-intensified diaminobenzidine (DAB) as chromogen. BrdU- and calretinin-immunoreactive cells in the subgranular zone and granule cell layer of the dentate gyrus were counted throughout the entire rostro-occipital extend of the hippocampus using a light microscope (Leica). The optical dissector method was modified in that cells appearing in the uppermost focal plane were excluded from analysis as described previously (Kempermann et al. 1997).

BrdU-positive cells were phenotypically characterized in triple-fluorescent labelled sections using confocal laser scanning microscopy (Leica). All confocal analyses were conducted in sequential scanning mode to avoid cross-bleeding between channels and series of images along the z-axis (one optical section/1 µm) of BrdU-positive cells were taken to demonstrate colocalization of immunoreactivity against BrdU and the respective antigen in the same cell. Fifty BrdU-positive cells randomly chosen from the subgranular zone and granule cell layer of the dentate gyrus were examined for immunoreactivity of the respective antigen (Nestin-GFP, Dcx, calretinin and NeuN). Percentages of immunohistochemical phenotypes were multiplied with the total number of BrdU-immunoreactive cells to achieve the absolute number of the respective phenotype.

Statistical analyses

All statistical analyses of morphological and behavioural data were performed with statview 4.5.1 for Macintosh. Factorial analyses of variance (anova) were used for multiple comparisons, followed by Fisher post hoc test, where appropriate. Comparisons with a P-value of <0.05 were regarded as statistically significant.

Results

Morris water maze training (hidden and cued version) decreased measures associated with adult hippocampal neurogenesis

We found that 9 days after BrdU, 1–4 days after the learning stimuli, there was a significant difference regarding the number of BrdU-positive cells in the GZ and SGZ of the dentate gyrus between the different experimental groups (one-way anovaF2,17 = 7.832; P < 0.01; Fig. 2a). Post hoc analyses showed that animals trained on the cued version of the Morris water maze (VIS) had significantly fewer hippocampal BrdU-positive cells than controls (CTR) that had been housed under standard laboratory conditions during the entire experiment (P < 0.01). The comparison of BrdU-positive cells in animals trained on the hidden version of the Morris water maze (HID) and CTR showed a similar decrease that, however, did not reach conventional statistical significance (P = 0.085). Interestingly, HID animals had significantly more BrdU-positive cells than VIS (P < 0.05): the training-induced reduction was less in animals trained on the hidden version. We noted that the variance of BrdU cell counts seemed to be higher in CTR than in HID and VIS and were concerned that this might be due to outliers, which might distort the results. We thus analysed our data in scattergrams and found evenly scattered plots in all groups (shown for the BrdU data in Fig. 2b).

Figure 2.

Morphological data. Experiment 1: Morris water maze training (hidden and cued version) decreased the number of BrdU-positive cells in the dentate gyrus as compared to standard laboratory conditions (a). Consistent with this finding (and independent of BrdU), the number of calretinin-positive cells, early postmitotic cells in adult hippocampal neurogenesis, was decreased by Morris water maze training as well (f). Phenotyping of BrdU-positive cells revealed that the visible version of the Morris water maze significantly decreased the number of neuronal progenitors at different developmental steps (b–e) as compared to controls (CTR). Training in the hidden version (HID) of the task decreased the number of late neuronal progenitors as compared to CTR (d, f) and significantly increased the number of early progenitors as compared to visible-trained animals (e). Experiment 2 (g): animals trained in the hidden version of the Morris water maze task did not differ from time-yoked controls regarding hippocampal BrdU-positive cells, supporting the notion that differences between HID and VIS in experiment 1 might have been due to physical activity. Experiment 3 (h): Pre-exposure to the water maze prior to BrdU injections and Morris water maze training balanced the effect of Morris water maze exposure on hippocampal BrdU-positive cells described above. This supports the hypothesis that decreased hippocampal neurogenesis induced by Morris water maze training (experiment 1) was due to stress associated with the procedure. All cell numbers were counted in the subgranular zone and granule cell layer of the right dentate gyrus. Bars reflect mean ± SEM. *P < 0.05, **P < 0.01. HID, hidden version of the Morris water maze; VIS, visible version of the Morris water; HID yoked, time-yoked control to HID; VIS yoked, time-yoked control to VIS; CTR, standard-housed control group.

To further substantiate the effect of Morris water maze training on adult hippocampal neurogenesis, we analysed the number of calretinin-positive cells in the dentate gyrus, which allows to assess the aspects of adult hippocampal neurogenesis independently of BrdU. Calretinin is transiently and specifically expressed in early postmitotic adult-born hippocampal neurons and disappears with the subsequent development to mature granule cells (Brandt et al. 2003). The calretinin data mirrored the effects on BrdU-positive cells in the dentate gyrus; mice trained in the cued or the hidden version of the Morris water maze had significantly less calretinin-positive cells in the SGZ and GZ of the dentate gyrus than control animals (one-way anova: F2,17 = 6.888; P < 0.01; Fisher's post hoc analysis VIS versus CTR: P < 0.01; HID versus CTR: P < 0.05; HID versus VIS: P = 0.38; Figs 2e and 3).

Figure 3.

Calretinin-expressing cells. Calretinin is expressed transiently during the early postmitotic period during adult hippocampal neurogenesis. To some degree, the number of calretinin-positive cells can serve as a surrogate marker to assess net neurogenesis (Brandt et al. 2003). Here, the number of calretinin-positive cells in the subgranular zone (arrow) reflects the decrease in adult neurogenesis seen in VIS (c) as compared to CTR (a) and HID (b) as depicted in Fig. 2(f). Scale bar (in ‘a’ for all panels), 250 µm.

Phenotyping of BrdU-positive cells demonstrated that the decrease of BrdU-positive cells by Morris water maze training translated in a decrease of BrdU-positive cells expressing the marker NeuN (one-way anova: F2,17 = 6.42; P < 0.01; Fisher's post hoc analysis CTR versus HID: P < 0.05; CTR versus VIS: P < 0.01; HID versus VIS: P = 0.13; Fig. 2b), showing that net neurogenesis was decreased in all animals exposed to the Morris water maze.

During distinct developmental steps of adult hippocampal neurogenesis, the intermediate filament Nestin and the microtubule-associated protein Doublecortin are sequentially expressed, with an overlap on one descriptively distinct cell type, type 2b cells (Kempermann et al. 2004a; Kronenberg et al. 2003). Further phenotypic analysis of hippocampal BrdU-positive cells in Nestin-GFP-reporter gene mice to visualize Nestin-expressing neural cells focused on the coexpression of Nestin and Doublecortin (type 2 cells) and the expression of Doublecortin, but not Nestin, by BrdU-positive cells (type 3 cells). We observed that type 3 cells were significantly decreased in both groups exposed to Morris water maze training as compared to controls (one-way anova: F2,17 = 5.173; P < 0.05; Fisher's post hoc analysis CTR versus HID: P < 0.05; CTR versus VIS: P < 0.01; HID versus VIS: P = 0.23; Fig. 2d), but the more immature type 2 cells were only decreased in VIS but not HID as compared to CTR (one-way anova: F2,17 = 4.8; P < 0.05; Fisher's post hoc analysis CTR versus HID: P = 0.58; CTR versus VIS: P < 0.05; HID versus VIS: P < 0.05; Fig. 2e).

Thus, HID had significantly more earlier neuronal progenitor cells (type 2, still proliferating) than VIS, whereas the comparison of the number of later neuronal progenitors (type 3, still proliferating) did not exhibit a statistically significant difference between HID and VIS. Furthermore, HID and VIS did not differ with regard to the number of calretinin-positive cells which represent an early postmitotic population (following the type 3 stage) in the context of adult hippocampal neurogenesis (one-way anova: F2,17 = 6.888; P < 0.01; Fisher's post hoc analysis VIS versus CTR: P < 0.01; HID versus CTR: P < 0.05; HID versus VIS: P = 0.38; Fig. 2f).

Comparison of latencies to reach the platform and the totally moved distances at different trials of acquisition training indicated an increased performance with subsequent trials, demonstrating that learning had occurred in animals trained on the hidden or cued version of the Morris water maze (Fig. 4). The probe trials given at day 4 of the water maze procedure showed that animals trained in the hidden version of the task spent significantly more time in the target quadrant than in the remaining quadrants (one-way anova: F3,48 = 6.293; P < 0.01; Fig. 4).

Figure 4.

Behavioural data. Decreasing latencies to reach the platform (in seconds), decreasing totally moved distances (in pixels, as measured by the automated tracking system) during trials and the time spent in the target quadrant on the probe trial (in percentage of total trial duration) demonstrate that animals trained on the hidden version of the Morris water maze task (full circles) in experiments 1, 2 and 3 had learned the spatial location of the platform. Latencies and swim paths to reach the platform for animals trained on the cued version of the Morris water maze are shown (open circles). For clarity, latencies and distances of two consecutive trials are shown grouped. All bars reflect mean ± SEM.

Did physical activity account for effects of the Morris water maze training on adult hippocampal neurogenesis?

Taken together, the above data indicated that despite the overall loss of precursor cell proliferation in the SGZ, animals trained on the hidden version of the Morris water maze retained more early neuronal progenitors (type 2), but did not differ with regard to later developmental steps of adult hippocampal neurogenesis from cued trained animals (type 3 and CR stage).

Paralleling these findings, HID animals were found to travel longer average distances per trial to reach the platform than VIS [distance travelled on one trial (mean ± SEM) in centimetres: HID: 716.29 ± 45.67 cm; VIS: 497.27 ±  41.47 cm; df = 333; t-value = 3.551; P < 0.01], raising the possibility that a different degree of physical activity might account for the differential effects on adult hippocampal neurogenesis. This explanation would be in line with earlier reports that physical activity preferentially affected the proliferation of early type 2, but not later neuronal progenitor cells (type 3) (Kronenberg et al. 2003).

We therefore performed an additional experiment (experiment 2) and compared mice learning the Morris water maze to a control group, swimming a matched distance per day but in the absence of any platform to be found. We observed no significant difference regarding the number of BrdU-positive cells between HID animals and the time-yoked control group with their similar amount of physical activity on water maze training (t-test: df = 12; t-value = 0.944; P = 0.364; Fig. 2g).

Did stress associated with Morris water maze training explain the decrease of adult hippocampal neurogenesis observed in experiment 1?

We hypothesized that stress related to Morris water maze training may be an important confounder when trying to study the effects of spatial learning in the Morris water maze on adult hippocampal neurogenesis in mice and might explain the decrease of measures associated with adult hippocampal neurogenesis induced by Morris water maze training in experiment 1.

To test this hypothesis, we designed an experiment analogous to the former, but included a pretraining phase during which we habituated the animals to the task settings in order to reduce stress levels at the actual training. Prior to training, the mice were repeatedly exposed to the pool (three 2-min trials for 3 days; experiment 3).

Under these conditions, comparison of the number of BrdU-positive cells in the GZ and SGZ of the dentate gyrus did not reveal a statistically significant difference between the groups (one-way anova: F4,34 = 1.307; P = 0.29; Fig. 2h), which suggests that the decrease of measures associated with adult hippocampal neurogenesis in experiment 1 could have been largely due to stress related to the procedures. The behavioural measures (latencies to reach the platform, totally moved distance) showed that training was accompanied by subsequently improved performance in all experiments (Fig. 4). In addition, mice trained on the hidden version of the water maze task spent significantly more time in the target quadrant than in the other quadrants during probe trial (one-way anova: F3,58 = 14.234; P < 0.01; Fig. 4).

Discussion

The detection of learning-related regulation of adult hippocampal neurogenesis might be confounded by physical activity and stress

In the present study, we intended to examine the effect of learning the Morris water maze on adult hippocampal neurogenesis in mice. Specifically, we examined hippocampal cell populations that had gone through S-phase of the cell cycle 5–6 days prior to the first training session.

This is the period where most of the regulation aiming at cell survival occurs. The time-point of morphometric analysis was chosen because BrdU-labelled hippocampal cell populations reach a quantitatively stable level quickly after cell divisions, and therefore, we assume that regulation of survival that occurs during this vulnerable period should be readily detectable. It is likely that regulation occurs also at later stages of neuronal development than addressed in the present study. However, the calretinin stage likely comprises the phase of highest synaptic plasticity in the course of neuronal development in the adult dentate gyrus (Brandt et al. 2003; Kempermann et al. 2004a). Given the comparatively lower ranges of regulation measurable on the level of cell numbers and marker expression at the late stages of development, these late effects become increasingly difficult to detect. The finding of increased synaptic plasticity and increasing maturation in the first few weeks (Jessberger & Kempermann 2003; van Praag et al. 2002; Snyder et al. 2001; Wang et al. 2000), however, suggests that other parameters might be successfully addressable during this period.

In the present study, we found a robust decrease of measures related to adult hippocampal neurogenesis (BrdU-positive precursor cells and calretinin-positive immature neurons in the dentate gyrus) by both training in the cued and hidden version of the Morris water maze as compared to animals that had been housed under standard laboratory conditions during the entire experiment (Fig. 2a,f). In addition, we report a differential regulation of early (type 2 cells) hippocampal progenitors by HID and VIS, whereas such a differential effect was not detectable on the level of later hippocampal progenitors (type 3 cells, calretinin-positive cells; Fig. 2d,f). However, given the relatively small differences in terms of absolute cell numbers, the differential effect on differentiation phenotypes, although statistically significant, should not be over-interpreted.

Several aspects of Morris water maze training other than hippocampus-dependent spatial learning do or might have effects on adult hippocampal neurogenesis and could therefore confound the assessment of spatial learning-related processes on hippocampal neurogenesis.

First, water maze training and pretraining procedures included a variety of potential stressors; the mice were handled and before training received several intraperitoneal injections of BrdU. Water maze training itself, particularly during the initial period, constitutes a major source of stress. The stress response to water maze training as assessed behaviourally (vocalizations, struggling) and physiologically (serum glucocorticoids) is rather acute, levels out quickly and is more pronounce in females than in males (Beiko et al. 2004). Pre-exposure to water maze stressors prior to training markedly reduced the stress response during training (Beiko et al. 2004). Stress has been reported to be a powerful negative regulator of adult hippocampal neurogenesis, largely mediated by glucocorticoid release from the adrenal gland upon activation of the hypothalamic-pituitary-adrenal axis, but the relationship is not linear (Fuchs et al. 2001; Gould et al. 1997; Heine et al. 2004). In order to adapt our mice to the water maze and reduce stress levels associated with training, we pre-exposed the animals to the water maze setting prior to BrdU injections and water maze training. Pre-exposure reversed the negative regulatory effect of Morris water maze training on adult hippocampal neurogenesis observed in experiment 1, where no pre-exposure had been conducted (Fig. 2g). Thus, we suggest that stress associated with water maze training might be an important confounder, when studying the effect of spatial learning on hippocampal neurogenesis in mice.

Second, Morris water maze training is associated with locomotion, and physical activity robustly increases hippocampal neurogenesis (van Praag et al. 1999a, 1999b). Our findings may suggest that physical activity is a relevant confounder: First, the finding that HID had more BrdU-positive cells than VIS (Fig. 2a) was paralleled by the observation that HID were physically more active during Morris water maze training than VIS. Furthermore, HID increased the number of early neuronal progenitors (type 2 cells), but not late neuronal progenitors (type 3 cells, calretinin-positive cells), as compared to VIS (Fig. 2d,f; but note caveat stated above), which corresponded to the pattern in which physical activity has been reported to affect adult hippocampal neurogenesis (Kronenberg et al. 2003). How the pronounced effect of physical exercise on type 2 cells is achieved mechanistically is unknown at present. Given the important regulatory roles of VEGF (Fabel et al. 2003) and IGF-1 (Trejo et al. 2001) in exercise-induced increases of hippocampal neurogenesis, however, specificity might be implemented at sites of action of these growth factors. Finally, animals trained on the hidden version did not differ from a time-yoked control group (that had a similar degree of physical activity on water maze exposure) with regard to BrdU-positive cells in the dentate gyrus (Fig. 2g). We assume that physical activity during swimming has a similar upregulating effect than voluntary wheel running, the paradigm that has been used to address activity-dependent regulation of adult neurogenesis (van Praag et al. 1999b). However, voluntary swimming paradigms cannot be implemented for mice, and our present data indicate that the normally used protocols for time-yoked controls in the water maze task seem to be confounded in this regard.

Third, besides spatial learning, learning the Morris water maze involves non-spatial learning strategies. It is unknown if and to what degree such strategies might impact on hippocampal neurogenesis, and thus, they have to be taken into account as potential confounders as well.

Based on our present data, we hypothesize that the observed effects of Morris water maze training on adult hippocampal neurogenesis in mice (experiment 1) might be largely explainable by physical activity and stress related to the procedure and may thus confound the recognition of potential effects of spatial learning on adult hippocampal neurogenesis under these conditions. Our data do not rule out that a direct effect of learning the water maze on adult hippocampal neurogenesis in mice occurs. However, our data demonstrate that if such effect exists, it may be masked by confounding parameters such as physical activity and stress. Because these two factors with their well-documented strong influence on adult hippocampal neurogenesis act in opposite directions, the potentially small direct survival-promoting effect of learning on the survival of new neurons will often be lost in other less-specific influences of the learning situation.

The Morris water maze might thus not be an optimal tool to investigate possible learning-dependent effects on adult hippocampal neurogenesis in mice. Previous studies on the effects of Morris water maze training on adult hippocampal neurogenesis had yielded heterogeneous results. Gould et al. (1999) found that adult hippocampal neurogenesis was specifically increased by the hidden version of the water maze, thus suggesting a specific regulation of neurogenesis by learning-related processes. There are differences in the study design, which might contribute to the discrepant results; Gould et al. used male Sprague Dawley rats, whereas we studied female C57BL/6 mice. Second, in terms of absolute time, in the present study, the animals were trained at a shorter absolute interval after the BrdU application than in the Gould study. However, the interval in our study was chosen exactly based on the rationale underlying the choice of interval length in the Gould study and was meant to reflect the species differences in the duration of neuronal maturation. Our previous data have shown that the downward sloping survival curve of the newly generated cells is the steepest in the interval we chose here (Kempermann et al. 2004a). We assume that the early postmitotic phase, during which the new neurons make first mature functional contacts (Hastings & Gould 1999; Jessberger & Kempermann 2003; Markakis & Gage 1999) exactly reflects the vulnerable period of hippocampal neurogenesis (Brandt et al. 2003; Kempermann et al. 2004a). Third, a few technical differences in the training protocols existed. For example, on day 4 of the water maze procedure, Gould et al. continued to train the animals, whereas in the present study (following the Lipp–Wolfer protocol; Wolfer et al. 1998), probe trials were given with the rationale to obtain an additional measure to judge spatial learning.

Ambrogini et al. (2000) reported an increase of hippocampal neurogenesis in rats, specifically in the spatial learner group as well. However, in that study only, cell densities and ratios were compared and no absolute numbers of BrdU-positive cells taken into consideration, so that the small observed difference might be spurious. Also, factors that may have differentially affected the volume of tissue constituents could confound this result. In a more recent report, Ambrogini et al. (2004) reported a decrease in the density of BrdU-positive cells in the rat dentate gyrus, specifically in the group trained in the hidden version of the Morris water maze as compared to the cued trained and control group. The second study by Ambrogini et al. more closely resembled the one by Gould et al. with regard to timing of BrdU administration and training than the first one. If this observation holds, it might not be species differences alone that explain the differences between our present and past data (van Praag et al. 1999b) and the results by Gould et al. (1999) and the initial claim by Ambrogini et al. (2000).

Döbrössy et al. described a differential effect of spatial learning in the Morris water maze on proliferating cells in the rat dentate gyrus. The cells that were generated during the early phase of learning appeared to be more susceptible to elimination, whereas those that were born during the later phase of learning were induced to proliferate more (Döbrössy et al. 2003). Besides species differences, the different temporal relationship of BrdU-injections and Morris water maze training in this study and the one by Döbrössy et al. provides a possible first explanation why we might not have seen similar differential effects of spatial learning on adult hippocampal neurogenesis, although our approach included the distinction of early (type 2) and late (type 3) precursor cells and early postmitotic neurons (stage of calretinin expression).

All mentioned studies have examined the effect of Morris water maze training on adult hippocampal neurogenesis in rats, while in the present study and our previous study (van Praag et al. 1999b), C57BL/6 mice were used. The rat studies, however, used time-yoked control groups as well, so that they controlled for some aspects of Morris water maze training not directly related to spatial learning such as physical activity and stress (Gould et al. 1999; van Praag et al. 1999b). Consequently, the discrepant results of this and previous studies might reflect species differences with regard to regulation of adult hippocampal neurogenesis by different aspects of Morris water maze training. In addition, a species-specific differential stress response associated with the exposure to a water pool in rats and mice appears particularly likely, given that rats swim in their natural habitat, whereas mice do not. Rats outperform mice in several maze tasks in water, but not on dry land (Whishaw & Tomie 1996). Furthermore, the fact that investigators of previous studies used male animals, while in the present study, females were used, may be particularly relevant with regard to stress responses induced by the experimental procedure, since females display more pronounced acute stress responses to water maze training (Beiko et al. 2004).

Other behavioural models involving hippocampal function such as trace eyeblink conditioning have been linked to direct effects on adult neurogenesis (Gould et al. 1999; Leuner et al. 2004), but it remains to be shown that this applies to complex cognitive tasks as well. At least, environmental paradigms that are ethologically more relevant than the task used in the present study have found an upregulation of hippocampal neurogenesis in mice (Kempermann et al. 1997), rats (Nilsson et al. 1999) and birds (Barnea & Nottebohm 1994, 1996). Although the possibility that these effects are caused by factors other than learning can not be excluded, the cognitive demands required in the enriched environment are different from those in the water maze. In any case, ethologically more relevant tasks (Gerlai & Clayton 1999) will have to be developed in order to assess the immediate functional contributions of adult neurogenesis.

Aspects of hippocampal neurogenesis not dealt with in the present study may of course be highly relevant for learning-related processes in the hippocampus. Physiological regulation with functional implications might potentially occur beyond the level of cell proliferation and the determination of cell survival. For example, adult-generated neurons are particularly susceptible to activity-dependent synaptic plasticity (Schmidt-Hieber et al. 2004; Snyder et al. 2001; Wang et al. 2000). As a consequence, the addition of new hippocampal neurons may contribute to learning-related processes, without being quantitatively regulated by learning itself.

Conclusion

In the present study, we could not find evidence of a specific and measurable effect of learning the water maze task on the survival of newly generated neurons in the adult hippocampus. However, we found further support for the idea that physical activity is an important stimulus regulating adult hippocampal neurogenesis and that stress can counteract physiological upregulation.

Acknowledgments

This work was supported by Volkswagenstiftung and Deutsche Forschungsgemeinschaft. The authors would like to thank Irene Thun, Silke Kurths and Ruth Zarmstorff for technical support, and Robert A.M. Brown and Brian Wiltgen for helpful comments and discussion.

Ancillary