|Home | About | Journals | Submit | Contact Us | Français|
Hippocampal function varies in a subregion-specific fashion: spatial processing is thought to rely on the dorsal hippocampus, while anxiety-related behavior relies more on the ventral hippocampus. During development, neurogenesis in the dentate gyrus proceeds along ventral to dorsal as well as suprapyramidal to infrapyramidal gradients, but it is unclear whether regional differences in neurogenesis are maintained in adulthood. Moreover, it is unknown whether young neurons in the adult exhibit subregion-specific patterns of activation. We therefore examined the magnitude of neurogenesis and the activation of young and mature granule cells in dentate gyrus subregions in adult rats that learned a spatial water maze task, swam with no platform, or were left untouched. We found that both adult neurogenesis and granule cell activation, as defined by c-fos expression in the granule cell population as a whole, were higher in the dorsal than the ventral dentate gyrus. In contrast, c-fos expression in adult-born granule cells, identified by PSA-NCAM or location in the subgranular zone, occurred at a higher rate in the opposite subregion, the ventral dentate gyrus. Interestingly, c-fos expression in the entire granule cell population was equivalent in water maze-trained rats and swim control rats, but was increased in the young granule cells only in the learning condition. These results provide new evidence that hippocampally-relevant experience activates young and mature neurons in different dentate gyrus subregions and with different experiential specificity, and suggest that adult-born neurons may play a specific role in anxiety-related behavior or other non-spatial aspects of hippocampal function.
A large number of new neurons are added to the dentate gyrus (DG) of adult mammals (Cameron and McKay, 2001). There is considerable evidence that these adult-born granule neurons are physiologically functional (Esposito et al., 2005; Ge et al., 2006; Snyder et al., 2001; van Praag et al., 2002; Wang et al., 2000) and contribute to hippocampus-dependent behaviors (Santarelli et al., 2003; Saxe et al., 2006; Saxe et al., 2007; Shors et al., 2001; Shors et al., 2002; Snyder et al., 2005; Winocur et al., 2006). Despite the wide range of behaviors that appear to require new neuron function, little is known about the specific contribution of new neurons to hippocampal function.
Functional gradients have been shown to exist within the hippocampus. Based on partial lesion and regional inactivation studies, the predominant view is that the dorsal hippocampus is particularly critical for spatial learning, whereas the ventral hippocampus is involved in regulating fear and anxiety (Bannerman et al., 2004; Kjelstrup et al., 2002; Moser et al., 1995; Pentkowski et al., 2006; Pothuizen et al., 2004). Other studies have suggested that the entire hippocampus is involved in processing spatial information but that the role of the ventral region is different from that of the dorsal region (de Hoz et al., 2003; Jung et al., 1994; McDonald et al., 2006). Thus, although their exact roles are not entirely agreed upon, it does seem clear that the dorsal and ventral regions contribute in different ways to hippocampus-dependent behaviors. The DG can also be divided into suprapyramidal and infrapyramidal “blades,” which lie dorsolaterally and ventromedially, respectively. Blade-dependent differences in excitability, GABAergic inhibition and exploration-induced Arc expression (Chawla et al., 2005; Ramirez-Amaya et al., 2005; Scharfman et al., 2002) indicate that cells in the two blades also contribute to different aspects of hippocampal function.
Comparisons across anatomically-defined regions of the hippocampus may provide an approach for understanding new neuron function (Sahay and Hen, 2007). Most studies of adult neurogenesis have either examined a small portion of the DG or have used stereological methods to quantify the total number of new neurons throughout the entire DG. Therefore, unlike developmental neurogenesis (Schlessinger et al., 1975), little is known about anatomical gradients of adult neurogenesis. In the present study we examine anatomical gradients of adult neurogenesis and use immunohistochemistry for the immediate-early gene c-fos (Fos) to compare activation rates in different subregions, in young granule cells as well as in the overall granule cell population. Fos is expressed in an activity-dependent manner and has been used as a marker for activated neurons in several studies of granule cells in the DG (Countryman et al., 2005; Vann et al., 2000; Worley et al., 1993), including adult-born granule cells (Jessberger and Kempermann, 2003; Kee et al., 2007; Tashiro et al., 2007). Here we find that young and mature granule cells show different anatomical gradients of activation and different specificity for the experiences that increase Fos expression.
Fourteen adult (10 weeks old at the start of experiments) male Long Evans rats (Charles River, Quebec) were used in the following experiments. All animals were individually housed, and all treatments conformed to animal health and welfare guidelines of the University of Toronto. To label adult-born DG granule cells, all rats were given 2 injections of 5-bromo-2’-deoxyuridine (BrdU; Sigma, 200 mg/kg/injection, dissolved at 20 mg/ml in saline, 0.007N NaOH) spaced 10 hours apart [Cameron and McKay, 2001; Snyder, 2005]. Beginning three weeks after BrdU injections, all rats were handled 5 minutes per day for 5 days to minimize stress associated with behavioral procedures. Four weeks after BrdU injections, rats were divided into three groups that were either trained in the Morris water maze (see below), put in the water maze with no platform (swim controls), or left untouched (cage controls). All rats were perfused exactly 2 hours after their first water maze trial (~90 min after their last trial) or at the same time of day (cage controls) to assess activity-dependent Fos expression.
On the final day of the experiment, 8 rats were trained in the Morris water maze, a hippocampus-dependent task (Morris et al., 1982) known to induce Fos expression in adult-born granule neurons (Jessberger and Kempermann, 2003; Kee et al., 2007; Tashiro et al., 2007). The testing facility was as previously described (Snyder et al., 2005). Briefly, the pool was 180 cm in diameter and filled with water. Water was kept at 25°C and non-toxic white paint was added to hide a 10 cm wide platform present in the SE quadrant. Distal cues were present on the walls of the room to allow rats to develop a spatial strategy for escaping the water by mounting the platform. Rats were trained in pairs for 16 trials. Each trial lasted 60 sec, and rats were allowed to remain on the platform for 10 sec after finding it. If a rat failed to locate the platform by the end of the trial, it was guided there by the experimenter. Trials were 1–2 min apart, allowing initially-naïve rats to quickly develop a spatial memory for the platform location. Latency to find the platform and mean proximity to the platform were calculated for each trial using HVS Image software (Buckingham, UK). For the proximity measure (Gallagher et al., 1993), the distance of the rat from the platform location was calculated at 0.1 sec intervals. The mean of these values was then calculated for each trial. A rat that searches in the correct area of the pool will have a low mean distance value, providing a measure of spatial bias for the platform location that should decrease across trials if animals learn the location of the platform. In addition, if the search pattern is spatially selective, the difference between the mean distance from the platform location and from a hypothetical “platform” location at the opposite side of the pool should increase across blocks. Following training, rats were returned to their home cages until being perfused exactly 2 hours after their first training trial. Swim control rats (n=3) were placed in the pool in the absence of a platform and allowed to swim for 60 sec, 45 sec, 30 sec and 15 sec (4 trials of each) to approximate the swim times of the trained rats. They were perfused 2 hours after the first swim trial. Cage control rats (n=3) were perfused directly from their home cages with no behavioral manipulation.
Animals were perfused with phosphate-buffered saline followed by 8% paraformaldehyde. Brains were fixed in paraformaldehyde for an additional 24 hours. The right hippocampus was extracted and sectioned perpendicular to its long axis to enable comparable analyses along the entire axis (Gaarskjaer, 1978; Rapp and Amaral, 1988). This axis is most precisely described as septo-temporal, but it will be referred to here as the dorso-ventral axis, as this terminology is more commonly used in studies of hippocampal function. Sections were cut at 40 µm using a vibratome for a total of ~200 sections.
Sequential fluorescent double-labeling was performed for Fos, followed by BrdU or PSA-NCAM, on free floating sections. For BrdU/Fos double-immunolabeling, sections were first incubated with rabbit anti-Fos antibody (1:10,000; Calbiochem; 3 days at 4°C) followed by Alexa568 goat anti-rabbit secondary antibody (1:200; Molecular Probes; 2 hours at room temperature). Sections were then treated with 1N HCl at 45°C for 40min to denature DNA and expose BrdU. Sections were then incubated with rat anti-BrdU antibody (1:200; Accurate; 1 day at 4°C) followed by Alexa488 goat anti-rat secondary antibody (1:200; Molecular Probes; 2 hours at room temperature). PSA-NCAM/Fos staining was done similarly, using rabbit anti-Fos (1:10,000; Calbiochem; 3 days at 4°C) then Alexa568 goat anti-rabbit followed by mouse anti-PSA-NCAM (1:200, Chemicon; 3 days at 4°C) then Alexa488 goat anti-mouse. All antibodies were diluted in phosphate-buffered saline containing 0.03% Triton X-100. One rat in the water maze-trained group had poor immunolabeling and was excluded from histological analysis.
For regional comparison, sections were assigned to one of four equal bins along the dorso-ventral length of the DG (D1=dorsal, D2=mid-dorsal, V3=mid-ventral, V4=ventral; see Fig. 2 insets). Within each section, attention was paid to whether cells were located within the infrapyramidal or suprapyramidal blade of the DG.
Three different methods of identifying adult-born granule neurons were used in this study, in order to optimize the population size for each analysis and provide independent confirmation of findings with different detection methods (summarized in Table 1).
Quantification of Fos+ and BrdU+ cells was performed on every 10th section throughout the dorso-ventral extent of the DG (19–22 sections per rat). Nuclei were counted as positive for Fos expression if the Fos signal stood out clearly compared to the surrounding tissue and if the staining evenly filled the entire nucleus. BrdU and Fos were quantified in the same sections. Analysis of Fos staining in PSA-NCAM+ cells was performed on every 5th section in the D1 and V4 quartiles. Every BrdU+ and PSA-NCAM+ cell was examined for Fos expression with an epifluorescence microscope (Olympus BX51) and a 60× oil immersion lens (N.A. 1.25). Potential double-labeled cells were then further examined using confocal microscopy (Olympus FV300) to confirm or reject colocalization of the two labels. To enable the objective characterization of a PSA-NCAM+ cell as Fos-positive or Fos-negative, the brightness of the Fos signal was measured as follows. At the middle focal plane of the cell, where Fos staining was brightest, the intensity of the nuclear Fos signal was measured and compared to the background intensity, measured in a random larger region of the hilus in the same optical field that appeared devoid of Fos staining. Cells with Fos staining brighter than 1.2× background were considered positive. This cutoff was chosen to approximate the staining intensity judged positive by eye. PSA-NCAM+ cell density (irrespective of Fos staining), which is labor-intensive due to the large number of cells with intertwined processes, was measured in two sections per animal – one from the middle of D1 (dorsal pole) and one from the middle of V4 (ventral pole). Total granule cell density was calculated using the optical fractionator method (West, 1993) in three Hoechst33258-counterstained sections (spaced 240 µm apart) from the middle of D1 and of V4 . Total granule cell number per section was estimated by counting cells at 60× with a 15 µm × 15 µm counting frame and 175 µm × 80 µm sampling grid, using Stereoinvestigator software (Microbrightfield). For all analyses, cross-sectional area of the granule cell layer was measured at 4× and multiplied by the section thickness (40 µm) to yield region volumes. Cell densities were then determined by dividing the number of cells in each region of the granule cell layer by its respective volume.
To characterize activity in a broader population of adult-born cells, we calculated the proportion of Fos+ cells that were located in the subgranular zone (SGZ), defined for this purpose as the deepest row of granule cells in the granule cell layer, bordering the hilus. The granule cell layer follows an outside-in gradient of formation that does not produce strict layering of granule cells by age but nonetheless results in the deepest portion of the granule cell layer containing a high proportion of granule cells that are young (Crespo et al., 1986; Dayer et al., 2003; Kempermann et al., 2003; Seri et al., 2004) and possess immature electrophysiological properties (Wang et al., 2000). The number of Fos+ cells in the subgranular zone and the total number of Fos+ cells in the granule cell layer (subgranular zone plus other layers) were counted in two sections per animal, one from the middle of D1 and one from the middle of V4 and the percentage of Fos+ cells in the subgranular zone was calculated (100 × FosSGZ / Fostotal). The percentage of Fos+ cells in the subgranular zone was compared across subregions (D1 vs. V4) and was also compared, within each subregion, to the chance value expected if Fos+ was distributed equally in all layers (=100/# of layers).
Statistical analyses were performed using Systat software (http://www.systat.com). Comparisons were made using 2-way or 3-way ANOVA and Tukey’s HSD for post-hoc analyses, with significance set at p<0.05.
Trained rats showed a decreased latency to find the hidden platform over 16 trials in the water maze (4 blocks of 4 trials) (Fig. 1a; One-way repeated measures ANOVA, main effect of block, F3,31=22, p<0.0001). There were significant differences between all blocks in the latency to find the platform (post-hoc, p<0.05) except for blocks 3 and 4. To verify that decreases in latency reflected spatial search strategies, the mean distance to the platform location, a measure of spatial bias (Gallagher et al., 1993), was also compared across trials. Two-way repeated measures ANOVA, with block and platform location as factors, showed significant main effects of block (F3,42=16, p<0.0001), platform location (F1,42=26, p=0.0002) and a block × platform location interaction (F3,42=4.5, p<0.01; Fig. 1b). Rats had a significant spatial bias towards the correct, target platform location on blocks 3 and 4: the mean distance to the platform location was significantly less than the distance to the equivalent location in the opposite quadrant of the pool (p<0.001; Fig. 1b).
The density of 4-week-old neurons, labeled with BrdU, was compared across dorso-ventral position (see Methods) and blade. BrdU+ cell counts were done only in water maze-trained rats rather than comparing across treatment groups, because BrdU+ cells were 4 weeks old, an age at which young granule cells are no longer susceptible to cell death (Dayer et al., 2003; Kempermann et al., 2003). In addition, rats were treated identically until 2 hours before perfusion, and this interval is almost certainly too short to see a differential survival effect (Olariu et al., 2005). Two-way ANOVA showed a significant main effect of dorso-ventral position on the density of BrdU+ cells (F3,40=6.4, p=0.001; Fig. 2,,3).3). Post-hoc analysis showed that the dorsal DG had significantly more neurogenesis than the ventral DG: both dorsal quartiles (D1&D2) had ~70% higher BrdU+ cell density than the ventral-most quartile (V4; p<0.01 in both cases). There was also a significant main effect of blade, with BrdU+ cell density significantly higher in the infrapyramidal blade than in the suprapyramidal blade (F1,40=8.1, p<0.01; Fig. 3). There was no blade × dorso-ventral interaction (F3,40=0.7, p=0.6).
Since differences in BrdU+ cell density were most pronounced between the dorsal-most and ventral-most quartiles, we focused on these two regions for our analysis of the endogenous marker of young neurons, PSA-NCAM. As for BrdU+ cell analysis, only water maze-trained rats were examined, because the post-treatment survival interval was too short to expect treatment effects on the number of young neurons. A two-way ANOVA on PSA-NCAM+ young neuron density in regions D1 and V4 found no significant effect of blade but found a significant effect of dorso-ventral position on PSA-NCAM+ cell density; PSA-NCAM+ cell density was higher in the dorsal DG than in the ventral DG (F1,24=6.5, p=0.02; Fig. 3b). Thus, the dorso-ventral gradient of neurogenesis is similar whether measured in four-week-old BrdU+ cells or in immature PSA-NCAM+ cells.
Greater BrdU+ and PSA-NCAM+ cell density in the dorsal DG could reflect a higher ratio of young:mature granule cells or it could be caused by a more general increase in the density of granule cells of all ages in the dorsal compared to ventral DG. To distinguish between these possibilities a two-way ANOVA on total granule cell density in regions D1 and V4 was performed. We found no significant effect of blade (F1,12=0.4, p=0.5) or dorso-ventral position on total granule cell density (F1,12=1, p=0.3; dorsal 433,913±28,911 cells/mm3, ventral 396,650±16,677 cells/mm3, mean±se; Fig. 3C). Thus, since overall granule cell density is constant, we conclude that regional differences in BrdU+ and PSA-NCAM+ (and Fos+, below) cell density reflect differences in the proportion of granule cells expressing these markers.
Immediate-early genes contribute to signaling cascades required for induction of long-term synaptic plasticity and consolidation of long-term memory (Kubik et al., 2007), and their expression patterns parallel those seen physiologically in “place cells” (Guzowski et al., 1999; Leutgeb et al., 2005). Together, these features make immediate-early genes such as Fos suitable for identifying neurons acutely involved in hippocampal functioning. A three-way ANOVA comparing the effects of treatment, dorso-ventral position and blade on Fos expression in adult rats revealed several significant effects. There was a significant main effect of treatment on the density of Fos+ cells (F2,80=21, p<0.001; Fig. 2, ,4).4). Maze-trained and swim control rats had approximately twice as many Fos+ cells as cage controls (post-hoc tests, p<0.001), confirming the previously observed activity-dependence of Fos staining in the dentate gyrus (Countryman et al., 2005; Kee et al., 2007; Vann et al., 2000). Interestingly, Fos+ cell density was not different between water maze-trained rats and swim controls (p=0.99). There was a significant main effect of dorso-ventral position (F3,80=22, p<0.001), with the dorsal-most quartile (D1) having greater Fos density than all other quartiles (post-hoc tests, p<0.001), consistent with place cell data showing that, during exploration, the proportion of neurons displaying place fields is greater in dorsal CA1 than in ventral CA1 (Jung et al., 1994). A significant main effect of blade was also found, with the suprapyramidal blade having a higher Fos+ cell density than the infrapyramidal blade (F1,80=68, p<0.001). Additionally, there was a significant treatment × blade interaction (F2,80=17, p<0.001), with post-hoc tests showing that both water maze training and swimming increased Fos+ cell density only in the suprapyramidal blade (p<0.001 versus suprapyramidal blade of cage control animals and versus infrapyramidal blades in all groups; Fig. 2, ,4).4). This restriction of the treatment effect to the suprapyramidal blade is consistent with the expression pattern previously seen for the immediate-early gene Arc (Chawla et al., 2005). Finally, there was no significant interaction between treatment and dorso-ventral position on Fos+ cell density (F6,80=2, p=0.08), indicating that granule cells in all dorso-ventral subregions were equally recruited.
Low numbers of BrdU+/Fos+ cells, as expected based on the small population of BrdU+ cells and low frequency of Fos expression throughout the dentate gyrus, prevented statistical comparisons using BrdU to identify young neurons. PSA-NCAM is a reliable marker of young neurons expressing characteristic enhancement of long-term potentiation (Schmidt-Hieber et al., 2004). Because it is expressed in new neurons for 3–4 weeks (Seki, 2002), it identifies a larger population of young neurons than two BrdU injections (4.4× larger in the current study), while following the same dorso-ventral gradient (Fig. 3).
We therefore examined Fos expression in this larger population of PSA-NCAM+ young neurons in the dorsal (D1) and ventral (V4) poles of the DG. A two-way ANOVA (treatment × dorso-ventral position) showed that treatment affected PSA-NCAM+/Fos+ cell density in a subregion-specific manner (Fig. 5). There was a significant main effect of treatment (F2,20=9.1, p<0.01) with water maze-trained rats having greater PSA-NCAM+/Fos+ cell density than both swim control and cage control rats (both post-hocs p<0.01; Fig. 5d–f). There was no difference between swim control and cage control animals (post-hoc p=0.98). We found no main effect of dorso-ventral position on PSA-NCAM+/Fos+ cell density but there was a significant interaction between treatment and dorso-ventral position (F2,20=3.9, p<0.05). This was attributable to a significant elevation in PSA-NCAM+/Fos+ cell density in the ventral DG of water maze-trained rats relative to the dorsal DG in the same rats and relative to dorsal and ventral regions of swim control and cage control rats (all post-hocs p<0.01). There were no overall or subregion-specific differences in PSA-NCAM+/Fos+ cell density between cage controls and swim controls (all post-hocs p>0.6). The elevated ventral PSA-NCAM+/Fos+ cell density in water maze-trained rats appeared to be greater in the suprapyramidal blade than the infrapyramidal blade (Fig. 5f), but data were collapsed across blade for the statistical analysis in order to achieve a normal distribution with small numbers of cells. Taken together, these findings indicate that young neurons are activated by platform location training in a water maze but are not activated by swimming in the same spatial arena in the absence of a platform. In addition, these data show that greater activation of young neurons occurs in the ventral region of the dentate gyrus, even though this region has fewer young neurons and less overall Fos activation.
To obtain a complementary measure of regional activity in a broader population of adult-born neurons, we also analyzed the proportion of Fos+ cells that were located in the SGZ in water maze-trained rats (Fig. 5b,c,g). The granule cell layer follows an outside-in gradient of formation that continues through adulthood, resulting in a high proportion of young neurons in the subgranular zone (Crespo et al., 1986; Dayer et al., 2003; Kempermann et al., 2003). By two-way ANOVA we found no difference between the suprapyramidal and infrapyramidal blades in the proportion of Fos+ cells in the subgranular zone but observed a significant effect of dorso-ventral location (F1,24=47, p<0.001; Fig. 5g). The proportion of Fos+ cells found in the SGZ was more than three times higher in the ventral DG than in the dorsal DG. Moreover, in the dorsal DG the proportion of Fos+ cells found in the SGZ was below chance levels based on the number of rows in the granule cell layer (8.1%; chance = 14.6% based on 6.8 layers; T6=4.2, p<0.01), while in the ventral DG the proportion was above significantly higher than chance levels (28.8%; chance = 18.2% based on 5.5 layers; T6=3.9, p<0.01). These findings are consistent with the increased activation of PSA-NCAM+ cells seen in the ventral DG. Additionally, they suggest that in the ventral region younger neurons are more likely to be activated than older granule cells, while in the dorsal region the opposite is true.
Anatomical subregions are apparent during development of the DG, when the ventral DG forms prior to the dorsal DG and the suprapyramidal blade forms before the infrapyramidal blade (Schlessinger et al., 1975). The higher density of new neurons in the infrapyramidal blade and dorsal DG observed in the current study matches gradients of neurogenesis observed late in development (Schlessinger et al., 1975), suggesting that regional differences in adult neurogenesis are remnants of developmental gradients.
A handful of studies have previously looked for dorso-ventral gradients of adult neurogenesis, with conflicting results showing either no significant difference (Banasr et al., 2006; Olariu et al., 2007), more new neurons in the dorsal DG (Dawirs et al., 1998; Ferland et al., 2002), or more neurogenesis at both poles with less in the middle (Silva et al., 2006). The current study differed from all previous studies because sectioning was done perpendicular to the dorso-ventral axis which divides the dentate gyrus into subregions that map closely onto the known anatomical segregation of inputs to the DG (Dolorfo and Amaral, 1998; Pitkanen et al., 2000). Sectioning along this axis also produces sections with a similar shape and thickness throughout the dorso-ventral extent, minimizing the changes that can complicate comparative analyses within the DG (Gaarskjaer, 1978; Rapp and Amaral, 1988). By normalizing cell counts to the volume of tissue analyzed (i.e. density), we eliminated the inevitable problem of dividing the dentate gyrus, which has no visible dorso-ventral boundaries, into pieces that are not exactly the same size. Since overall granule cell density was equivalent across subregions, gradients in densities of cells expressing the different markers reflect similar gradients in the proportions of total granule cells expressing these markers.
Numerous behavioral studies have suggested that the dorsal and ventral hippocampus have dissociable roles in spatial memory and anxious/fearful behavior: the dorsal hippocampus contributes to spatial water maze memory, but not fear-related behavior in an elevated plus maze, whereas the ventral hippocampus contributes to fear-related behavior but not spatial learning (Bannerman et al., 2004). These distinct behavioral roles are supported by separation of anatomical inputs to the hippocampus. The ventral hippocampus receives strong direct projections from the hypothalamus and amygdala, structures involved in responding to anxiogenic, fearful and stressful stimuli (Petrovich et al., 2001; Pitkanen et al., 2000; Risold and Swanson, 1996; Swanson and Cowan, 1977), while the dorsal hippocampus receives input from the dorsolateral portion of the entorhinal cortex (Dolorfo and Amaral, 1998), which is specifically required for spatial learning (Steffenach et al., 2005).
In the current study we observed higher Fos+ cell density in the dorsal DG and suprapyramidal blade, consistent with previous studies showing dorso-ventral and suprapyramidal-infrapyramidal gradients of activity during spatial navigation (Chawla et al., 2005; Jung et al., 1994; Ramirez-Amaya et al., 2005). Additionally, we found that the dorsal DG had higher Fos+ activity, regardless of treatment, suggesting that the dorsal DG is generally more active than the ventral DG and that experience recruited granule cells from all dorso-ventral DG subregions equally. Interestingly, we found that activity levels were equivalent in rats learning a platform location and those swimming in the absence of reinforcement, suggesting that the granule cell population responds similarly during active learning of a spatial location and undirected exploration. This is perhaps not surprising since rats can learn about their environment regardless of whether or not learning is being guided and measured by a researcher. This activation in swim control animals is also consistent with place cell activation seen during free exploration (Leutgeb et al., 2007) and the proposed role for the hippocampus in “automatic recording of attended experience” (Morris and Frey, 1997).
We found two major distinctions between the activation of young granule cells and older granule cells that suggest that young granule cells in the ventral dentate gyrus play a different role in water maze learning than mature granule cells. The first difference was that Fos expression in PSA-NCAM+ young granule cells showed a ventral over dorsal gradient of Fos expression, opposite to the gradient seen in the overall granule cell population. Importantly, the larger number of activated young granule cells in the ventral dentate gyrus does not simply reflect a larger number of new neurons in this region, because neurogenesis showed the opposite gradient. This gradient difference was replicated in an analysis of Fos expression in adult-born neurons in the subgranular zone. This subgranular zone analysis also allowed us to directly compare Fos expression in younger and older neurons across dorso-ventral subregions. In the ventral region, the younger granule cells of the subgranular zone were more likely to be activated by water maze training than granule cells in the more superficial layers, while in the dorsal area, the older granule cells were more likely to be activated than the younger subgranular neurons. No previous studies have compared immediate-early gene expression in young neurons in different DG subregions. However, a few studies have examined activation of adult-born neurons and found either enhanced activation (Kee et al., 2007; Ramirez-Amaya et al., 2006) or no difference in activation (Jessberger and Kempermann, 2003; Tashiro et al., 2007) of new cells relative to older cells. These studies are difficult to directly compare because of differences in stimuli used for activation (water maze versus unreinforced exploration) and prior experience of the animals (previous water maze training, enriched housing conditions, etc.).
The second difference between young and old granule cells was related to the specific experiences that activated the young and older granule cells. Mature granule cell activation, as seen through Fos expression throughout the granule cell layer, occurred to the same degree in water maze trained rats as in swim control animals. However, young granule cells did not show this expected pattern. Unlike the overall granule cell population, PSA-NCAM+ young granule cells showed a large, 4-fold, increase in activation in the water maze learning condition but no change in activation in the swim control condition relative to the home cage condition. Interestingly, Gould et al. (1999) found that survival of young granule neurons increases after spatial training in the water maze but not in the swim control condition, further supporting the idea that hippocampus-dependent learning experiences have special significance for young granule cells.
The fact that water maze training induces Fos expression in young granule cells suggests that these young neurons may play a role in spatial learning and memory. Although spatial tasks such as the hidden-platform water maze used in this study are often considered to be more dependent on the dorsal hippocampus, studies have demonstrated a role for the ventral hippocampus in spatial memory as well (de Hoz et al., 2003; Ferbinteanu et al., 2003; McDonald et al., 2006). Studies looking for impairment after loss of adult neurogenesis have not found evidence that young granule neurons are needed for acquisition in the water maze (Madsen et al., 2003; Saxe et al., 2006; Shors et al., 2002; Snyder et al., 2005). However, inhibition of adult neurogenesis does impair long-term retention of spatial platform location (Snyder et al., 2005), suggesting that the Fos+ young neurons in the ventral DG in the current study may be involved in long-term spatial memory formation. Consistent with a role in long-term memory, the ventral hippocampus requires more time than the dorsal hippocampus to consolidate water maze memories (de Hoz et al., 2003), and it is primarily the ventral DG that is activated by the retrieval of remote water maze memory (Gusev et al., 2005).
Alternatively, it is possible that the young neurons activated in the ventral DG are responding to non-spatial aspects of experience in the water maze. Water maze learning has strong anxiogenic and stressful, in addition to spatial, components (Beiko et al., 2004), suggesting that the new neurons in the ventral DG may contribute to the expression of the fear-or anxiety-related behaviors that are believed to be mediated by the ventral hippocampus (Bannerman et al., 2004). This role for young neurons would be consistent with the idea that young neurons are involved in depression and anxiety-related behavior and/or the effects of antidepressants on those behaviors (Drew and Hen, 2007; Duman, 2004; Santarelli et al., 2003; Sapolsky, 2004). However, the lack of increase in Fos expression in young neurons in the very stressful swim control condition, argues that new neurons are not simply activated by all fearful or stressful experiences. It may be the case, though, that young granule neurons in the ventral dentate gyrus are important for learning that occurs in a fearful context. The impairment in contextual fear conditioning observed after loss of new neurons is consistent with this possibility (Saxe et al., 2006; Winocur et al., 2006; Wojtowicz et al., 2008), as are findings that the ventral, as well as dorsal, hippocampus is important for consolidation of contextual fear memory (Rudy and Matus-Amat, 2005; Sutherland et al., 2008). More broadly, it may be that the non-spatial functions proposed for the hippocampus, and in some cases for the ventral hippocampus in particular, activate young neurons during learning. These functions include resolving conflicts between two possible responses (McNaughton and Wickens, 2003), inhibiting locomotor activity (Bast and Feldon, 2003), or inhibiting affectively positive associations or memories (Davidson and Jarrard, 2004). It will be important in the future to determine which types of experience, spatial or non-stressful or otherwise, activate young neurons in the dorsal DG and to discover whether mnemonic and anxiety-related roles are dissociable or are in fact part of a common function for these new neurons.
The authors would like to thank Sarah Rabbett and Preethi Ramchand for assistance with histological analyses. This research was supported by an Ontario Graduate Scholarship (JSS), CIHR grant (JMW) and by the Intramural Program of the National Institutes of Health, National Institute of Mental Health, Z01-MH002784 (HAC).