Functions and Dysfunctions of Adult Hippocampal Neurogenesis

Synaptic Inputs

Studies using oncoretroviruses for birth-dating and labeling have shown that newborn granule cells in young adult mice develop a single primary dendrite with multiple branches that reaches the molecular layer within 7 days and exhibits rapid growth between 7 and 17 days, followed by modest growth for at least two months (Sun et al. 2013). Functional electrophysiological characterization of labeled newborn neurons in acute slices has revealed a stereotypic integration process in which GABAergic synapses precede glutamatergic synapse formation (Espósito et al. 2005, Ge et al. 2006, Overstreet-Wadiche et al. 2006b). Although radial glia-like precursors exhibit functional GABA_A receptors and tonic responses to ambient GABA (Song et al. 2012), the first functional synaptic inputs appear to form onto proliferating neuroblasts within four days of birth (Song et al. 2013, Tozuka et al. 2005). Postmitotic newborn neurons continue to exhibit tonic GABA responses while their GABAergic synaptic responses mature (Espósito et al. 2005, Ge et al. 2006). Several recent studies, using pair recording (Markwardt et al. 2011), optogenetics (Song et al. 2013), and rabies virus–based retrograde transsynaptic tracing (Deshpande et al. 2013, Li et al. 2013, Vivar et al. 2012), have identified multiple interneuron subtypes that innervate newborn neurons within weeks of birth (Figure 1b), including parvalbumin-expressing basket cells, somatostatin-expressing HIPP (hilar perforant path-associated) cells, HICAP cells (hilar interneuron with commissural-associational pathway-associated axon terminals), and MOPP (molecular layer perforant pathway) cells, such as neurogliaform cells/Ivy cells. The sequence of GABAergic synapse formation by various interneuron subtypes remains unknown. Notably, newborn neurons exhibit initial depolarizing responses to GABA, which gradually shift to hyperpolarizing responses within two to three weeks of birth (Ge et al. 2006, Overstreet Wadiche et al. 2005). Depolarizing GABAergic signaling promotes survival, maturation, and synapse formation and activation in newborn neurons (Chancey et al. 2013, Ge et al. 2006, Pontes et al. 2013, Song et al. 2013, Tozuka et al. 2005). Initial GABAergic synaptic inputs are not sufficient to elicit action potentials and are therefore unlikely to be directly involved in information processing. Instead, new neuron outputs are controlled by glutamatergic synaptic inputs. Electrophysiological analyses have shown that the first detectable glutamatergic synaptic responses emerge in 11–14-day-old newborn neurons, and these responses mature over the next several weeks, accompanied by increased density of dendritic spines (Chancey et al. 2014, Espósito et al. 2005, Ge et al. 2006). Recent optogenetic and rabies virus–based retrograde transsynaptic tracing suggested that glutamatergic synaptic inputs onto newborn neurons originating from mossy cells form ahead of those by perforant pathway fibers from the entorhinal cortex (Figure 1b) (Chancey et al. 2014, Deshpande et al. 2013, Kumamoto et al. 2012). In addition, studies revealed inputs from cholinergic septal neurons at early stages (Deshpande et al. 2013, Vivar et al. 2012). Upon maturation, adult-born granule cells appear to exhibit general properties that are indistinguishable from developmentally born granule cells (Ge et al. 2007; Laplagne et al. 2006, 2007), although differences in some specific characteristics cannot be ruled out.

Synaptic Outputs

Newborn neurons extend a single axon from the base of the cell body that follows a stereotypic pathway through the hilus to reach CA3 within 7 days and establishes mature primary projection patterns within 21 days (Sun et al. 2013). Electron microscopic analyses of retrovirally labeled newborn neurons have shown synaptic structures associated with cells in both hilus and CA3 within 14 days and mossy fiber en passant boutons reaching morphological maturation within 8 weeks (Faulkner et al. 2008, Toni et al. 2008). Optogenetic activation of newborn neurons confirmed functional glutamatergic synaptic outputs onto hilar mossy cells and interneurons and CA3 neurons (Gu et al. 2012, Toni et al. 2007). Two- to four-week-old adult-born neurons synthesize and corelease GABA, in addition to glutamate (Cabezas et al. 2012, 2013). However, this GABA release appears to modulate presynaptic mossy fiber excitability only by activating GABA_B autoreceptors and GABAergic postsynaptic responses have not been detected (Cabezas et al. 2012). In general, we know much less about properties of synaptic outputs of newborn neurons compared with their inputs, information critically needed to better understand adult neurogenesis functions. Future studies using new tools, such as optogenetics, anterograde transsynaptic tracing, and whole-mount imaging, are needed to provide a more complete picture about different targets of adult-born neurons and temporal dynamics of functional synapse formation.

A common feature of pre- and postsynaptic integration of newborn neurons is the apparent competition with mature granule cells for innervation from afferent axons and efferent connections to invade and replace preexisting synapses (Toni et al. 2007, 2008). Therefore, adult neurogenesis not only continuously adds new individual units to the dentate gyrus, but also induces structural plasticity of mature neurons, including mature granule cells and hilar mossy cells and interneurons, presynaptic terminals of entorhinal inputs, and postsynaptic sites on CA3 neurons.

Basic characterization of the adult hippocampal neurogenesis process has provided critical information on when and how newborn neurons could contribute to brain functions. For example, 14-day-old adult-born neurons already exhibit functional glutamatergic synaptic inputs and outputs and can therefore participate in neural processing during immature stages. Developmental and synaptic integration patterns of adult-born neurons are largely consistent with those described for dentate granule cells generated during development (Liu et al. 1996, Overstreet-Wadiche et al. 2006a, Zhao et al. 2006); therefore, it seems that adult-born neurons may participate in the same neuronal circuits as do preexisting ones. A fundamental question follows: How can a small population of adult-born neurons make meaningful contributions to brain functions in the presence of millions of mature neurons of the same type?

Distinct Cellular Properties

Electrophysiological analyses showed that, compared with mature neurons, immature adult-born neurons are highly excitable (Dieni et al. 2013, Mongiat et al. 2009). As a result, they are very efficient in generating action potentials, even with weak glutamatergic inputs (Marín-Burgin et al. 2012). Immature newborn neurons also exhibit a lower induction threshold and larger amplitude of associative long-term potentiation (LTP) of perforant path synaptic inputs compared with mature granule cells in acute slices under identical conditions (Ge et al. 2007, Schmidt-Hieber et al. 2004). This enhanced synaptic plasticity is partially due to a lack of strong GABAergic inhibition in immature neurons (Ge et al. 2008). Adult-born neurons exhibit such properties only during a critical period between approximately three and six weeks after birth and depend on developmentally regulated synaptic expression of NR2B-containing N-methyl-D-aspartate (NMDA) receptors (Ge et al. 2007). Similarly, in vivo field recordings showed that four-week-old newborn neurons exhibit enhanced LTP at mossy fiber synaptic outputs onto CA3 neurons (Gu et al. 2012). Therefore, adult-born neurons have distinct cellular properties compared with mature neurons, and the transient nature of such properties may provide a fundamental mechanism allowing adult-born neurons within this critical period to serve as major mediators of experience-induced plasticity.

Distinct Circuitry Properties

One major difference between immature and mature granule neurons is in GABAergic inputs and, in particular, a lack of strong perisomatic inhibition of immature neurons (Ge et al. 2008, Li et al. 2012, Marín-Burgin et al. 2012). As a result, four-week-old newborn neurons exhibit a lower firing threshold owing to an enhanced excitation/inhibition balance involving feedforward inhibitory circuitry. Combined with higher intrinsic excitability, weak afferent activity recruits a substantial proportion of immature neurons while activating few mature granule cells, as shown by calcium imaging in acute slices (Marín-Burgin et al. 2012). These observations suggest a model in which immature neurons with a low activation threshold and input specificity comprise a population of integrators that are broadly tuned during a finite developmental period and may encode most features of the incoming afferent information. However, mature granule cells generated during both development and adult neurogenesis, owing to their high activation thresholds and input specificity, serve as pattern separators. In this model, activity patterns entering the dentate gyrus could undergo differential encoding through immature neuronal cohorts that are highly responsive and integrative and, in parallel, through a large population of mature granule cells with sparse activity and high input specificity.

Long-distance modulatory inputs may also differentially affect immature and mature granule cells and impact information processing. Dentate granule cells are known to receive dopaminergic inputs from the ventral tegmental area (Gasbarri et al. 1997). Dopamine causes a long-lasting attenuation of medial perforant path inputs to newborn neurons through D1-like receptors and decreases their capacity to express LTP, whereas dopamine activation via D2-like receptors suppresses synaptic inputs onto mature granule cells but does not influence their LTP expression (Mu et al. 2011). Whether other long-projection modulatory inputs differentially regulate newborn and mature granule cells remains to be determined.

These studies, mostly in vitro characterizations in acute slice preparations, have demonstrated differential properties of adult-born neurons during immature stages and provided the framework for how adult-born neurons could make unique contributions to specific brain functions. Indeed, many computational modeling and animal behavior studies support the immature neuron model of how adult neurogenesis may contribute to hippocampal functions (Deng et al. 2010). This model does not rule out potential contributions of mature adult-born neurons because these neurons are also plastic in response to neuronal activity and could be involved in different aspects of learning and memory (Lemaire et al. 2012, Ramirez-Amaya et al. 2006). One critical parameter of this model is the rate of new-neuron maturation or the duration of immature states. The entire neurodevelopmental process takes an estimated eight weeks in young adult mice (Ge et al. 2007, Zhao et al. 2006). This maturation rate is affected by numerous environmental, pathological, and pharmacological factors (Piatti et al. 2011, Zhao et al. 2006) and exhibits significant differences among species (Brus et al. 2013). Notably, neuronal maturation in the dentate gyrus of adult macaque monkeys appears to be longer than six months (Kohler et al. 2011). It is tempting to speculate that the lengthened time course for adult-born neuron maturation in nonhuman primates, and possibly in humans, may help to maintain immature neuronal properties over a longer life span.

Much remains to be learned about basic properties of adult-born neurons. One major roadblock is the lack of effective approaches to directly examine physiological properties of newborn neurons in awake behaving animals. Our current methodology is limited to immediate early gene expression as an indirect readout of neuronal activation, which has produced conflicting results on whether adult-born neurons are preferentially recruited into active networks over preexisting neurons upon specific behavioral stimulation (Ramirez-Amaya et al. 2006, Stone et al. 2011). Future studies of newborn neurons in vivo, such as targeted recording of firing properties (Leutgeb et al. 2007; Neunuebel & Knierim 2012, 2014) or calcium imaging using miniature microscopes (Ziv et al. 2013), will provide essential new information and significantly advance the field.

Cognitive Functions

A central dogma regarding information processing through the trisynaptic circuit of the hippocampus has been that the dentate gyrus mediates pattern separation, the ability to distinguish similar stimuli and contexts, whereas CA3 mediates pattern completion, the reinstatement of activity patterns correlated with complete contexts and associations using only partial or degraded information. Building on Marr’s description of the information-processing capacity of the hippocampus based on its structure and intrinsic connectivity, an emergent model proposed that the densely packed dentate gyrus can support the orthogonalization of inputs arriving from the entorhinal cortex (pattern separation) (Marr 1971, Treves & Rolls 1994). The dentate gyrus, in turn, directly projects to CA3, a hippocampal subregion with extensive recurrent connections and thus the putative capacity to reactivate stored patterns using partial inputs (pattern completion). The idea that the dentate gyrus could support pattern separation received empirical support following the observation that cells in the dentate gyrus region are sparsely activated and have very low firing rates. Many behavioral assays of the dentate gyrus and CA3 function have been designed to test this hypothesis, and indeed results suggest that each of these regions may play a role in pattern separation and completion, respectively (Hunsaker & Kesner 2013). Recent evidence also supports an essential role for newborn granule neurons in mediating pattern separation, as inferred from behavioral deficits in mnemonic discrimination when adult neurogenesis is impaired (Aimone et al. 2011, Sahay et al. 2011b). For example, ablation of adult neurogenesis via irradiation impairs the ability of the mice to discriminate stimuli with little spatial separation, but not stimuli widely separated in space, in two spatial memory tasks (Clelland et al. 2009). Conversely, increasing the number of adult-born neurons by deleting the proapoptotic gene Bax from adult neural precursors and their progeny enhances their ability to differentiate between overlapping contextual representations (Sahay et al. 2011a). In one striking example, mice engineered to block synaptic release via tetanus toxin light-chain expression in mature dentate granule cells, but not in most immature neurons younger than four weeks, exhibit improved discrimination of very similar contexts in a fear-conditioning test, and blockade of adult neurogenesis via irradiation impairs context discrimination, suggesting a predominant role of immature neurons in mediating pattern separation (Nakashiba et al. 2012). However, these mice exhibit defects in the water-maze task and fear-conditioning tasks with partial cues presented, which suggests a critical role for mature granule cells in rapid pattern completion. One must recognize that the computational definition of pattern separation and completion implies a strict input–output relationship at the neural level that is not necessarily congruent with the use of this concept to describe behavior (Santoro 2013). It is still not clear, for example, that an animal that appears to pattern complete by behaving indistinguishably in a full-cue versus partial-cue context is doing so by using attractor dynamics in the CA3 region or by engaging mechanisms that compensate for the activation of only a subset of synapses to recruit the full complement of synapses involved in encoding (Knierim & Zhang 2012). Likewise, it is not clear that similar contexts are encoded by minimally overlapping ensembles of dentate granule cells.

Adult hippocampal neurogenesis has also been implicated in other aspects of contextual and spatial memory (Deng et al. 2010). By varying the timing between adult neurogenesis ablation and behavioral tests, studies have also pinpointed essential roles of adult-born immature neurons at various stages for these functions, especially when the task is difficult (Deng et al. 2009, Denny et al. 2012). These results corroborate findings of special cellular properties associated with immature stages and support the immature neuron model of the contribution of adult neurogenesis to brain functions. Adult neurogenesis has also been implicated in memory consolidation and the reorganization of memory traces to extrahippocampal structures, such as the prefrontal cortex (Kitamura et al. 2009). One study has shown that decreased adult hippocampal neurogenesis is accompanied by a prolonged period of hippocampus-dependent associative fear memory, whereas increased adult neurogenesis is associated with accelerated reorganization of memory traces that rely less on the hippocampus. In another interesting study, longitudinal activity data were collected on a large number of inbred mice, which shared one large enriched environment, to explore the relationship among cognitive challenges, adult hippocampal neurogenesis, and the development of individual behavioral traits. The size of the roaming area explored by an individual mouse was positively correlated with the amount of hippocampal neurogenesis (Freund et al. 2013), suggesting that one function of adult neurogenesis may be to shape the neuronal circuitry according to individual needs and improve adaptability over the life course of the individual.

There are many inconsistencies in the current literature regarding effects of various manipulations of adult neurogenesis levels on behavioral test outcomes, which have been summarized in previous reviews (Deng et al. 2010, W.R. Kim et al. 2012). Many possible factors contribute to these contradictory findings, such as differences in the genetic background, experimental manipulation, and behavioral paradigms. One major limitation of traditional approaches is the chronic nature of manipulations that affect adult neurogenesis throughout multiple phases of the learning process; therefore, it is not always clear whether adult-born neurons contribute to encoding, consolidation, storage, and/or retrieval processes. In addition, compensatory network changes may occur following ablation of adult neurogenesis (Singer et al. 2011). Newly available tools now allow for specific manipulation of the activity of adult-born neurons at distinct stages of maturation and during specific stages of learning and recall (W.R. Kim et al. 2012). One recent study of retrovirally targeted newborn neurons in the adult mouse dentate gyrus showed that optogenetic suppression of four-week-old, but not two- or eight-week-old, newborn neurons during recall trials impairs contextual fear memory and spatial memory retrieval (Gu et al. 2012). Given the increasing availability of sophisticated genetic models to target specific populations of neural progenitor subtypes or newborn neurons at specific maturation stages, combined with optogenetic and pharmacogenetic tools to manipulate neuronal activity with spatial and temporal precision, future studies will be able to directly address questions of how and when newborn neurons contribute to neural function.

Mood Regulation

It is now well-established that stress negatively regulates progenitor proliferation and new-neuron survival (Gould et al. 1992), whereas clinical antidepressant treatments, including electroconvulsive therapy and chemical antidepressants (Malberg et al. 2000), promote proliferation of neural progenitors and maturation of newborn neurons during adult hippocampal neurogenesis (Sahay & Hen 2007, Warner-Schmidt & Duman 2006). Such effects are evolutionarily conserved from rodents to nonhuman primates (Perera et al. 2007) and potentially to humans (Boldrini et al. 2009). Ablation of adult neurogenesis does not appear to alter affective phenotypes at basal levels but abolishes some antidepressant-induced behaviors in rodents (Santarelli et al. 2003) and nonhuman primates (Perera et al. 2011). Emerging evidence suggests a critical role of adult hippocampal neurogenesis in the stress response by suppressing the hypothalamic-pituitary-adrenal (HPA) axis. In mice with adult neurogenesis ablated, mild stress leads to increased levels of stress hormones and greater stress responses in behavioral tests (Schloesser et al. 2009, Snyder et al. 2011). Furthermore, blockade of adult neurogenesis abolishes the antidepressant effect of hippocampal regulation of the HPA axis after chronic stress (Surget et al. 2011). The mechanism by which adult-born neurons regulate the HPA axis under basal conditions and upon antidepressant treatment remains to be determined. Anatomical studies have revealed that whereas the dorsal hippocampus projects primarily to cortical areas that mediate cognitive processes such as learning and memory and navigation and exploration, the ventral hippocampus projects to the limbic system, including the amygdala, the nucleus accumbens, and the hypothalamus (Fanselow & Dong 2010). Future studies are needed to address how adult-born neurons, especially those in ventral regions, influence neural pathways involved in emotional experience and affective states. It will also be interesting to test the immature neuron model of adult neurogenesis function in the context of mood regulation. Development of optogenetic and pharmacogenetic approaches to target newborn neurons at specific maturation stages in the dorsal or ventral dentate gyrus will facilitate these efforts (Kheirbek et al. 2013).

Dysfunction via Loss-of-Function

Fragile X syndrome, the most common form of inherited intellectual disability, is caused by the functional loss of fragile X mental retardation protein (FMRP). Fmrp null mice exhibit deficits in some forms of hippocampus-dependent learning, accompanied by reduced adult hippocampal neurogenesis due to impaired neuronal differentiation and survival (Luo et al. 2010). Deletion of Fmrp specifically in adult neural progenitor cells using nestin-CreER^T2 mice recapitulates defects in both adult neurogenesis and hippocampus-dependent learning, and furthermore, restoration of Fmrp expression in adult neural progenitors alone is sufficient to rescue learning deficits in Fmrp null mice (Guo et al. 2011). These striking results suggest a causal role of adult neurogenesis dysfunction in learning impairments associated with fragile X syndrome, at least in animal models. Whether this is generalizable to other disorders remains to be seen.

One major regulator of adult hippocampal neurogenesis is aging. Over the subject’s lifetime, the rate of adult neurogenesis decreases dramatically, from rodents to primates, which may contribute to the dysfunction of hippocampus (Ming & Song 2011). The rate of decline of adult neurogenesis during aging is much more robust in rodents (about tenfold) than in humans (about fourfold) (Spalding et al. 2013).

Dysfunction via Gain-of-Function

Dentate granule cells may play a central role in the pathogenesis of temporal lobe epilepsy, one of the most common human seizure-related disorders (Houser 1992). In animal models of epilepsy, pilocarpine-induced status epilepticus leads to a dramatic and prolonged increase in dentate neural progenitor proliferation (Parent et al. 1997). However, many of these newborn neurons integrate aberrantly, displaying hilar basal dendrites with spines, ectopic hilar localization of the cell body, and mossy fiber sprouting (Jessberger et al. 2007, Kron et al. 2010), similar to what has been observed in postmortem dentate gyri of patients with temporal lobe epilepsy (Houser 1992). Eliminating cohorts of newborn neurons decreases status epilepticus–induced mossy fiber sprouting and ectopic granule cells (Kron et al. 2010) and attenuates spontaneous recurrent seizures in mice (Jung et al. 2004). Separately, deletion of PTEN (phosphatase and tensin homolog deleted on chromosome ten) in a small percentage of dentate granule cells born postnatally is sufficient to cause spontaneous seizure within four weeks, accompanied with aberrant granule cell morphology seen in epilepsy (Pun et al. 2012). Collectively, these studies provide strong evidence that dysfunction of adult hippocampal neurogenesis plays a causal role in epileptogenesis.

In another gain-of-function example, retrovirus-mediated knockdown of Disrupted in Schizophrenia I (DISC1), a risk gene for major mental illness (Thomson et al. 2013), leads to aberrant integration of newborn dentate granule neurons in the adult mouse hippocampus, including ectopic location of the cell body to the outer granule cell layer and molecular layer, aberrant axonal targeting beyond CA3, hyperexcitability, and aberrant formation of synaptic inputs and outputs, due in part to hyperactivation of the mTOR pathway in newborn neurons (Duan et al. 2007; Faulkner et al. 2008; J.Y. Kim et al. 2009, 2012). Dysregulated adult hippocampal neurogenesis following DISC1 knockdown in one cohort of newborn neurons is sufficient to cause several behavioral phenotypes, including pronounced learning and memory deficits (in the object-place recognition task and the spatial version of the Morris water maze), as well as clear anxiety and depression-like phenotypes (in the forced-swim test and elevated plus maze) (Zhou et al. 2013). Inactivation of these aberrant neurons reverses specific behavioral phenotypes, indicating a causal role of adult neurogenesis dysfunction in behavioral impairments.

Notably, the impact of adult neurogenesis dysfunction due to gain-of-function on animal behavior is generally more pronounced than that seen in loss-of-function conditions. This finding may not be surprising because the complete absence or removal of a system may trigger the recruitment of alternative pathways to compensate (Singer et al. 2011), whereas miswiring of newborn neurons can be more detrimental, especially given their high excitability and unique properties. These findings have significant implications for future cell-replacement therapy in which correct wiring of transplanted neurons could be essential for functional benefits and for avoiding potential side effects.