Self-renewing stem-like cells in the adult hippocampus have captured the imagination of neuroscientists and clinicians for decades. Unfortunately, there have been relatively few studies investigating the functional role of AHN in humans. Throughout this thesis, I have described a number of studies in which we undertook the challenge of identifying indirect correlates of AHN in humans as well as elucidating the functional role of adult-born granule cells in everyday memory. We accomplished this by assessing various lifestyle- and blood-based factors known to influence neurogenesis from the animal literature and comparing these factors to behavioural performance on tasks which tested the proposed roles for AHN in learning and memory.
There has been much speculation regading the functional role of neurogenesis in humans. Computational modellers and theorists have proposed several distinct roles for AHN in cognition based on what is known of their location in the brain, functional connectivity with surrounding regions and physiological poperties. Some have proposed that the constant turnover of newborn cells in the hippocampus would allow memory storage for novel events, while avoiding interference with older memories, a computational process termed pattern separation (Becker et al., 2005; Chambers and Conroy, 2007; Appleby and Wiskott, 2009; Becker et al., 2009; Weisz and Argibay, 2009; Aimone and Gage, 2011). However, events occurring close together in time may be subject to enhanced interference because the same population of cells would be firing in response to each event encountered. This process of increasing interference between events occurring close together in time is referred to as pattern integration (Aimone et al., 2006). Indeed, some studies have shown paradoxical improvements in working memory tasks following ablation of neurogenesis (Saxe et al., 2007). However, working memory circuits outside of the hippocampus may be responsible for such improvements. Across short timescales, the majority of behavioural evidence from rodents has actually demonstrated that the role for neurogenesis in cognition, although seemingly widespread, converges on one function in particular. The formation of context-shock associations is impaired in animals lacking neurogenesis (Saxe et al., 2006; Winocur et al., 2006; Imayoshi et al., 2008; Warner-Schmidt et al., 2008; Wojtowicz et al., 2008; Hernandez-Rabaza et al., 2009; Ko et al., 2009; Guo et al., 2011; Nakashiba et al., 2012; Pan et al., 2012b), especially when the shock is relatively weak or training paradigm relatively short (Drew et al., 2010; Pan et al., 2012a, 2013). Animals lacking neurogenesis are also impaired at discriminating between overlapping odor pairs (Luu et al., 2012) or between nearby, but not far apart spatial locations (Clelland et al., 2009). In contrast, upregualting neurogenesis via aerobic exercise or genetic manipulation has been shown to increase AHN and leads to enhanced behavioural pattern separation or CFC performance (Creer et al., 2010; Sahay et al., 2011; Kohman et al., 2012). While it seems like a wide-variety of tasks require adult-born granule cells, many, if not all of these tasks require overcoming interference. All of these tasks require the animal to form separate representations of similar stimuli, regardless of whether the stimuli are different contexts, objects, spatial locations or odours. This is why I say that AHN is required for a wide-variety, yet specific set of memory tasks. The behavioural requirements of tasks shown to depend on neurogenesis have differed substantially, but the psychological construct shown to rely on AHN has been fairly consistent.
Neurogenesis may further help separate similar events occurring over longer time periods (Becker, 2005; Aimone et al., 2006; Becker and Wojtowicz, 2007). A distinct pool of newborn neurons would help to add a degree of contextual novelty to similar events that are separated by a sufficient amount of time. Without new cells being added to the hippocampal network, the same populations of cells would end up reperesenting multiple different memories, leading to catastrophic interference (Wiskott et al., 2006). This account of the role for neurogenesis in learning and memory has generally been reffered to as the memory retention hypothesis throughout this thesis. In contrast, others have proposed that the addition of newborn cells to the hippocampus would result in existing connections being altered in such a way that information is lost (Feng et al., 2001; Deisseroth et al., 2004; Frankland et al., 2013). This account of the role for neurogenesis in learning and memory has generally been referred to as the memory clearance hypothesis throughout this thesis. Behavioural evidence from non-human animal studies has supported the memory retention hypothesis, especially for spatial or context-rich memories. For instance, rodents with ablated neurogenesis display marked deficits in remembering the platform locaton following MWM training across long, but not short timescales (Snyder et al., 2005; Deng et al., 2009; Jessberger et al., 2009; Kitamura et al., 2009; Inokuchi, 2011; Pan et al., 2012a, 2012b, 2013). In contrast, some studies have shown impaired long-term retention of fear memories in younger mice with relatively high rates of neurogenesis compared to their older counterparts (Akers et al., 2012). When older mice had wheel-running- or antidepressant drug-induced enhancement of neurogenesis, they were impaired on tests of remote memory compared to control mice (Akers et al., 2014). Therefore, it would seem that behavioural evidence from rodents has also supported the memory clearance hypothesis. The persistence or clearance of memories as a result of ongoing neural turnover in the DG may depend on the type of memory. There is evidence to suggest that spatial memories are always dependent on the hippocampus (Snyder et al., 2005; Deng et al., 2009; Jessberger et al., 2009). For these memories that are permanently hippocampal-dependent, AHN may help keep overlapping events disctinct from one another, thereby promoting long-term retention of the original memory. On the other hand, fear memories may be supported by regions outside of the hippocampus (Kitamura et al., 2009). Thus, for those memories that can be supported by extrahippocampal structures, AHN may accelerate the process of systems consolidation (Kitamura et al., 2009), shifting the dependence of the memory from the hippocampus to neocortical regions (McClelland et al., 1995; Squire and Alvarez, 1995; Maviel et al., 2004; Squire and Bayley, 2007). While animal studies have provided valuable clues as to the importance of AHN in learning and memory, it has nonetheless become apparent that studying the functional role of neurogenesis directly in humans is the critical next step that must be taken in order to alleviate some of the confusion generated in non-human animal studies.
In the set of studies that comprise this thesis, we have demonstrated that change in aerobic capacity following chronic physical activity correlates with change in performance on a putatively neurogenesis-dependent visual pattern separation task. On the other hand, stress and depression scores had opposing effects on behavioural pattern separation performance. Importantly, neither exercise response nor depression scores predicted performance on other trial types within the BPS-O, repeated or novel items, nor the visuo-spatial CANTAB® PAL task. We have also shown that lower stress and depression scores are associated with improved visual object recognition on repeated items following a two-week delay from the study phase. Further, on two-week delayed retention tests, participants scored near chance at identifying lures as “similar”, regardless of stress and depression levels. Interestingly, they more often misclassified these items as “new”, as opposed to “old”. Our results provide indirect evidence from human participants that AHN is important for pattern separation across shorter delays, while contributing to the persistence of memories for repeated items across extended time intervals. Future studies could explicitly test the memory clearance hypothesis in humans by measuring recognition memory across longer timescales with a pro-neurogenic intervention, such as long-term exercise, in between study and test.
Pattern separation and memory retention (or clearance) may be coexisting phenomena. The addition of immature neurons to the hippocampus may at first bias the network towards pattern separation, as opposed to pattern completion, thereby reducing interference between events (Yassa and Reagh, 2013). In turn, the amount of pattern separation may decide what information will be subject to reconsolidation and what information will be cleared (Yassa and Reagh, 2013). Reconsolidation is the process whereby an existing memory becomes susceptible to change. If an event is considered similar, but not the same as, a previously stored event then the original memory may be modified to accommodate the discrepent information. Thus, the constant addition of adult-born neurons to the hippocampus may serve as a means of adding contextual information to existing memories. However, the original memory may be altered so drastically during reconsolidation that it is no longer accessible (essentially cleared), depending on the amount of interference between the original memory and the novel event. Indeed, a number of computational models predict that the addition of newborn neurons to an existing circuit would hinder retrieval of previously stored memories (Deisseroth et al., 2004; Weisz and Argibay, 2009, 2012). On the other hand, if two events are considered one in the same, then information may be strengthened, although more generalized in nature. As these adult-born neurons that once contributed to pattern integration or pattern separation continue to mature and establish new synaptic connections with the pre-existing circuitry, they may destabilize previously established memories in the hippocampus, leading to the loss of previously stored information (Josselyn and Frankland, 2012; Frankland et al., 2013; Yassa and Reagh, 2013). In turn, the clearance of older memories would make room for new ones and the newborn neurons would become part of the physical storage site for new memories (Josselyn and Frankland, 2012). Thus, both processes may be beneficial in their own way. Whether or not memory clearance is a benefit or detriment to memory performance really depends on the relative importance of information is being cleared.
Correlates of Neurogenesis in Humans and Animals
Neurogenesis is down-regulated in a variety of neuropsychiatric disorders, so being able to characterize AHN in vivo is critical for better disease prevention and/or treatment. Unfortunately, there is no way to non-invasively quantify newborn cells in the living human hippocampus. Therefore, it has been difficult to improve our understanding of how neurogenesis influences the onset or recovery from certain disorders associated with downregulated neurogenesis, such as depression. In addition, we cannot assess the specific contribution of newborn neurons to learning and memory.