Adult hippocampal neurogenesis reduces memory interference in humans: opposing effects of aerobic exercise and depression

Abstract

Since the remarkable discovery of adult neurogenesis in the mammalian hippocampus, considerable effort has been devoted to unraveling the functional significance of these new neurons. Our group has proposed that a continual turnover of neurons in the DG could contribute to the development of event-unique memory traces that act to reduce interference between highly similar inputs. To test this theory, we implemented a recognition task containing some objects that were repeated across trials as well as some objects that were highly similar, but not identical, to ones previously observed. The similar objects, termed lures, overlap substantially with previously viewed stimuli, and thus, may require hippocampal neurogenesis in order to avoid catastrophic interference. Lifestyle factors such as aerobic exercise and stress have been shown to impact the local neurogenic microenvironment, leading to enhanced and reduced levels of DG neurogenesis, respectively. Accordingly, we hypothesized that healthy young adults who take part in a long-term aerobic exercise regime would demonstrate enhanced performance on the visual pattern separation task, specifically at correctly categorizing lures as "similar." Indeed, those who experienced a proportionally large change in fitness demonstrated a significantly greater improvement in their ability to correctly identify lure stimuli as "similar." Conversely, we expected that those who score high on depression scales, an indicator of chronic stress, would exhibit selective deficits at appropriately categorizing lures. As expected, those who scored high on the Beck Depression Inventory (BDI) were significantly worse than those with relatively lower BDI scores at correctly identifying lures as "similar," while performance on novel and repeated stimuli was identical. Taken together, our results support the hypothesis that adult-born neurons in the DG contribute to the orthogonalization of incoming information.

Keywords: depression; exercise; hippocampus; interference; neurogenesis; pattern separation.

Figures

Figure 1

Correct classification of similar lures pre- vs. post-exercise in Experiment 1 (A), and in those with below median and above median BDI scores in Experiment 2 (B,C). (A) Median change in % correct identification of lures as “similar” (bias corrected ranked difference scores) following exercise for responders vs. non-responders. (B) Mean % correct identification of lures as “similar” (bias corrected, ± standard error) when the original target appeared in the same block. (C) Mean % correct identification of lures as “similar” (bias corrected, ± standard error) when the target appeared in a previous block. *p ≤ 0.05.

Figure 2

Percent change in fitness scores (VO2peak) vs. change in mean % correct identification of lures as “similar” following exercise (r = 0.74, p < 0.01).**p ≤ 0.01.

Figure 3

The behavioral pattern separation task used in Experiment 2. In each block, everyday objects are displayed 1 at a time in a sequence of presentation trials followed by a sequence of recognition trials. In recognition trials, the object can either be old (an exact repetition of a previously presented object), similar, but not identical, to a previous target, or a new, completely unrelated foil. Each block has a distinct visual context (background image). In recognition trials, for repetitions and similar items, the associated target may have been seen within the same block (same temporal and visual context) or within a previous block (different temporal and visual context).

Figure 4

Mean bias toward pattern separation vs. pattern completion (mean probability of correctly identifying lures as “similar” minus probability of misidentifying lures as “old,” ± standard error) vs. BDI score in Experiment 2 (A) and vs. age in Toner et al. (B). (A) Mean bias scores when related target appeared in the same block as lure, for those with below median vs. above median BDI scores. (B) Mean bias scores for younger vs. older adults. Adapted from Figure 2 in Toner et al. (2009) and reproduced with permission from Cold Spring Harbor Laboratory Press. *p ≤ 0.05.

Figure 5

Mean bias toward pattern separation vs. pattern completion (mean probability of correctly identifying lures as “similar” minus probability of misidentifying lures as “old,” ± standard error) in Experiment 2 for those with below median vs. above median BDI scores, (A) for “more similar” lures, when target was within the same block, (B) for “less similar” lures, when target was within the same block, (C) for “more similar” lures, when target appeared in a different block, and (D) for “less similar” lures, when target appeared in a different block.**p ≤ 0.01.

Figure 6

Mean % correct identification of similar lures (bias corrected) as a function of BDI scores, (A) when they appear in the same block as the original target (r = −0.272, p = 0.05), (B) when they are presented in a different block from the original target (r = −0.297, p = 0.03).*p ≤ 0.05.