Coronal sections (40 μm) were cut through the entire DG of one hemisphere (randomly chosen) of the brain with an oscillating tissue slicer. For BrdU peroxidase staining, a 1:12 series of sections were mounted onto glass slides and pretreated by heating in 0.1 M citric acid (pH 6.0). Tissue was then incubated in trypsin, followed by 2N HCl and overnight in primary mouse anti-BrdU (1:200) and 0.5% Tween 20. The next day, tissue was incubated for 1 hr in biotinylated anti-mouse antibody (1:200), then in avidin-biotin-horseradish peroxidase (1:100), and lastly in diaminobenzidine. After rinsing in phosphate buffer, slides were counterstained with cresyl violet and cover-slipped with Permount. For quantitative analysis, estimates of total numbers of BrdU-labeled cells were determined using a modified unbiased stereology protocol (Gould, et al., 1999b
; West, Slomianka, & Gundersen, 1991
). BrdU-labeled cells in the subgranular zone (SGZ), granule cell layer (GCL) and hilus on every twelfth unilateral section throughout the entire rostrocaudal extent of the DG were counted blindly at 1,000X on a Nikon Eclipse E400 light microscope, avoiding cells in the outermost focal plane. The number of cells was multiplied by 24 to obtain an estimate of the total number of BrdU-labeled cells in the hippocampus.
Conditioning was evaluated using repeated measures ANOVA. Blocks of training trials were used as the repeated measures (blocks of 20 trials for the first 100 and blocks of 100 for the remaining 700 trials) and the type of training procedure (Trace versus CTC) as the independent measure. The type of training procedure did not alter the % of CRs that were emitted [F(1, 22)=0.01; p>0.05], nor was there an interaction between blocks of training trials and type of training (p>0.05). Thus, responding during training with trace and CTC was similar (). As expected, there was a main effect of trials [F(11, 242)=13.39; p<0.001], as the percentage of CRs increased across blocks.
To evaluate the potential effect of overall performance during training on cell survival, animals were categorized into those that reached a criterion of 60% CRs during training (good learners) or those that did not (poor learners), as used in previous studies (Leuner, et al., 2004
). Of the good learners, 7 had been trained with the standard trace procedure and 6 had been trained with CTC. Of the poor learners, 5 had been trained with trace and 6 with CTC. As expected, those classified as good learners emitted a greater % of CRs across blocks [F(11,66)=11.65; p<0.001; F(11,55)=9.49; p<0.001 separate repeated measures ANOVA for good learners trained for 800 trails on Trace or CTC, respectively], whereas the poor learners did not (p>0.05) (). Within-subjects comparisons indicated that the %CRs in animals that learned well (reached 60% CRs) did not further increase during the last 200 trials of training (p>0.05), indicating that they had reached asymptotic performance. There was no difference in spontaneous eyeblink rates between good learners and poor learners before any training occurred [F(1,22)=0.10; p>0.05] (data not shown).
Overall, learning rather than training increased the number of cells that remained in the hippocampus one day after training had ceased [F(1,34) = 0.76; p>0.05] (). The animals that reached a criterion of 60% CRs in either task (Trace or CTC) possessed more labeled cells than did naïve animals that were kept in their home cages during the training procedure [Trace: F(1,18)= 9.53; p<0.01; CTC: F(1,17)= 4.78; p<0.05] (). This effect of learning was evident in the combined counts from subgranular zone and granule cell layer [Trace: F(1,18)= 5.75; p<0.05; CTC: F(1,17)= 5.24; p<0.05] () and did not occur in the hilus (p>0.05; data not shown). Since the majority of cells in the subgranular zone and granule cell layer mature into neurons (Christie & Cameron, 2006
), these data suggest that learning rescues cells that will become neurons. Moreover, the animals that learned well possessed more cells after training than did the animals that learned poorly [Trace: F(1,10)= 15.9; p<0.005; CTC: F(1,10)= 5.16; p<0.05] ( and ). Again, the effect of learning was evident in the subgranular zone and granular cell layer [Trace: F(1,10)= 5.08; p<0.05; CTC: F(1,10)= 6.72; p<0.05] (), but not in the hilus (p>0.05; data not shown). The number of BrdU labeled cells in animals that reached criterion (good learners) did not differ between animals that were trained on trace or CTC (p>0.05).
Learning during trace conditioning or CTC enhanced the survival of new born cells in the dentate gyrus
BrdU labeled cells (shown with black arrows) in the dentate gyrus of hippocampus from similar sections of an animal that learned well during training (reached 60% CRs) and an animal that learned poorly. Images were magnified 1000x.
The number of BrdU labeled cells in subgranular and granule cell layer of individual animals correlated with the % of CRs during training on the third session (trials 400–600) [r = 0.54; p< 0.01] () and the last session (trials 600–800) [r = 0.49; p< 0.05] (). The number of cells also correlated with the total number of CRs that were emitted across all 800 trials of training [r = 0.42; p< 0.05]. The number of BrdU labeled cells (SGZ and GCL) did not correlate with the %CRs during training on the first (trials 1–200) or second (trials 200–400) session. The number of cells in the hilus did not correlate with the %CRs during any session of training (p>0.05).
In this experiment, cells that were born one week before trace conditioning were more likely to survive provided that learning occurred. The type of training task was inconsequential: that is, learning during training with the standard trace procedure in which the stimuli are discontiguous was effective, as was training with a trace procedure in which contiguity is established by simultaneous presentation of the CS and the US together after the trace interval. These results confirm previous findings showing that learning a trace conditioning procedure enhances the survival of newly generated cells in the adult hippocampus (Gould, et al., 1999b
) and that discontiguity between the CS and the US is not a necessary feature for this effect to occur (Leuner, et al., 2006a
). In a recent study, we found that animals with hippocampal lesions could associate the CS with a US across a trace interval, provided that the CS was presented again in combination with the US (Bangasser, et al., 2006
). Here we find that learning under these training conditions increased the number of new cells that survived, indicating that the newly generated cells are not responding exclusively to tasks that depend on the hippocampus for learning. There is one potential caveat to this conclusion. In the lesion study (Bangasser, et al., 2006
), conditioning was assessed with fear conditioning (freezing) rather than an eyeblink response, as used here. It seems unlikely that the choice of behavioral response would matter and therefore, we tentatively conclude that the increase in cell survival is not limited exclusively to learning that depends on the hippocampus. Additionally, just because the CTC task does not require the hippocampus for learning, it does not mean that the hippocampus is not used when it is present.
Irrespective of the training regimen, animals that learned better by the end of training retained more new cells in their hippocampus than those that did not learn as well. The cells were located in the subgranular zone and granular cell layer, where post-mitotic daughter cells reside as they differentiate into neurons (Christie & Cameron, 2006
). There was no effect of learning on the number of cells that remained in the hilus, where fewer new neurons reside. In previous studies, the vast majority of the cells (~80%) that remained in the hippocampus after learning possessed neuron-specific markers (Leuner, et al., 2004
; Leuner, et al., 2006a
; Gould, et al., 1999b
). It is therefore assumed that the cells here would become neurons, if they were not already. The increase in BrdU cell number after learning was significant, although proportionately less here than in some previous studies (Leuner, et al., 2004
; Gould, et al., 1999b
). The reasons for the differences probably reflect, at least in part, the fact that more cells were present in the naïve controls. This could be due to the age of the animals when they were injected with BrdU. Even in adulthood, the number of new cells decreases significantly between about 2 and 9 months of age (McDonald & Wojtowicz, 2005
). In the initial study (Gould, et al., 1999b
), we used adult animals, but did not confine our measurements to young adults (~65 days of age) used here and more recently (Leuner, et al., 2006a
). Also, the overall level of conditioning achieved after 800 trials even in those that reached criterion (i.e. the good learners) was not as high as in other studies, which regulates the number of cells that survive.
The results from these studies indicate that learning and not training increases the survival of new cells in the dentate gyrus. The effects are therefore not attributable to “enriched environment” or movements associated with the training procedure. This supports previous results indicating no effect of unpaired stimuli on cell survival (Leuner, et al., 2004
; Gould, et al., 1999b
). Exactly what determines whether a given task will increase cell survival is unclear at this time, but most likely involves differences in the electrophysiological responses during training. Trace conditioning is known to enhance cell excitability in the hippocampus (Moyer, Thompson, & Disterhoft, 1996
), but so does delay conditioning (Berger, Rinaldi, Weisz, & Thompson, 1983
), which does not rescue the new cells from death (Leuner, et al., 2006a
;Gould, et al., 1999b
). More subtle differences in how hippocampal neurons respond to trace conditioning (Gilmartin & McEchron, 2005
) likely account for the differential effects of the various training procedures on neurogenesis.
Here we report several correlations between the amount of learning and the number of cells that remained in the dentate gyrus. Other investigators have also reported correlations (Drapeau, Mayo, Aurousseau, Le Moal, Piazza and Abrous, 2003
, Kempermann, & Gage, 2002
), although typically between cells generated and performance on hippocampal-dependent tasks. For example, the number of cells born in the hippocampus of aged animals, weeks after training, correlated with their performance on a spatial maze task (Drapeau, et al., 2003
). The effect reported here is different in that the correlation emerges as a function of learning itself and thus reflects the fate of cells that were already present at the time of the learning experience. In a previous study, we did find that the degree of responding early in training (200 trials) correlated with the number of new cells that survived (Leuner, et al., 2004
). However, the animals were not trained to asymptote, and therefore we do not know which ones would have learned, given the opportunity. Here we show that animals that learned after training for 800 trials retained more cells than animals that did not learn, but were trained for as many trials. Therefore, it can be concluded that the effect of trace conditioning and perhaps other training tasks on neurogenesis is related to learning and not simply to training. Moreover, the correlation between the number of learned responses and cell number that occurs early in training is maintained until the end of training when most animals have reached asymptote. These data suggest that acquisition rescues the cells from death and the number of cells at the end of training relates to the level of performance that was achieved.