Introduction

Despite earlier evidence pointing to the existence of dividing cells in the adult brain, it is only recently that a consensus has been reached that new neurons are added to the adult mammalian brain. In the 1960s, a series of studies by Joseph Altman using [3-H]-thymidine autoradiography reported evidence of newborn neurons in adult rat and cat brains (Altman 1962, 1963; Altman and Das 1965, 1966). However, these findings were ignored for three decades due to technical concerns and lack of acceptance of this concept by the field.

After three decades of debate (Eckenhoff and Rakic 1988; Kaplan 1985; Rakic 1985), several new technologies, including the use of synthetic thymidine analogue 5-bromo-2′-deoxyuridine as well as multi-color immunolabeling for cell type-specific markers and confocal microscopy with stereological calculation of immunolabeled cells, were developed that helped to confirm the existence of adult neurogenesis. These technological developments allowed researchers to determine the existence of adult-born neurons; eventually, adult neurogenesis was confirmed in the dentate gyrus (DG) of the hippocampus of rodents by our group and others (Kempermann et al. 1997b; Kuhn et al. 1996; Seki and Arai 1999). Subsequently, two important findings were reported. First, using several different in vivo and ex vivo techniques, researchers found adult neurogenesis not only in the DG of mammalian model animals but also in the DG of the human hippocampus (Eriksson et al. 1998; Knoth et al. 2010; Roy et al. 2000; Spalding et al. 2013). Because the hippocampus is strongly associated with learning and memory, these findings prompted researchers to hypothesize that adult neurogenesis could play functional roles in human cognition. Second, several studies revealed that the process (proliferation, migration, differentiation, maturation, etc.) of adult hippocampal neurogenesis is highly regulated by experience as well as by environmental and biological factors (i.e., aging) (Cameron and McKay 1999; Gould et al. 1999; Kempermann et al. 1997b, 1998, 2002; Mirescu et al. 2004, Tanapat et al. 1999; van Praag et al. 1999). Although the proportion of adult-born neurons is a minor population compared with the total number of DG neurons, the continuous addition of new neurons over a lifetime (approximately 700 cells per day in human) implies that adult neurogenesis could add substantial structural and functional plasticity into the tri-synaptic circuits of the hippocampus (Spalding et al. 2013). Since the rate of adult neurogenesis is regulated, the degree of hippocampal plasticity (meta-plasticity) could be regulated through the modulated of adult neurogenesis. These studies have prompted the field to hypothesize and investigate the roles of hippocampal adult neurogenesis in cognitive function and brain plasticity.

In this review, we focus on the contribution of adult neurogenesis to structural and functional plasticity in the DG of the hippocampus in the rodent model.

Cell biological features of adult neurogenesis

Soon after the discovery of adult-born neurons in the DG of rodent and human brains, the morphological and molecular features of cells related to adult neurogenesis were extensively investigated (Fig. 1). Adult neural stem cells (type 1 radial glia-like cells; RGLs) are found in the narrow area between the granule cell layer and the hilus, the so-called sub-granular zone (SGZ). The SGZ provides an essential environmental niche for RGLs, and the microenvironment in the SGZ allows RGLs to proliferate and maintain the neural stem cell pool. One prominent feature of adult hippocampal neurogenesis is that the rate of neurogenesis is regulated by an individual’s experience (described below). Since the tri-synaptic circuit in the hippocampus is strongly associated with learning and memory, the addition of adult newborn neurons in the DG could provide another layer of structural and functional plasticity in those processes in addition to the plasticity of existing circuits.

Fig. 1
figure 1

Neural circuits in the hippocampus (upper panel). Adult neural stem cells in the hippocampus (radial glia-like cells, type 1 cells) and their differentiation in the SGZ of the DG (lower panel) with cell type-specific transcription factors

Full size image

Two essential characteristics of neural stem cells are self-renewal and multipotency to generate specific neural cell types. Clonal analysis showed that RGLs in the DG have these two features (Bonaguidi et al. 2011, 2012). However, an alternative “disposable stem cell” model was proposed following the observation of serial labeling of dividing cells, in which activated RGLs differentiated into astrocytes after several rounds of cell division (Encinas et al. 2011). These two models are not mutually exclusive, and recent evidence suggests that RGLs are a heterogeneous population, differentially responding to the stimuli depending on their subtype (Jhaveri et al. 2015). Further investigation of RGLs using single-cell RNA sequencing methods currently under development should help to reveal the nature of the heterogeneity of RGLs (Lacar et al. 2016; Shin et al. 2015). RGLs generate transient amplifying progenitors (type 2a and 2b cells), type 2 cells give rise to neuroblasts (type 3) that are still able to proliferate, and neuroblasts differentiate into DG neurons (Fig. 1) (Bonaguidi et al.; Lugert et al. 2010; Suh et al. 2007). In addition to DG neurons, adult neural stem cells in the SGZ give rise to a small population of glial cells, such as astrocytes (Encinas et al. 2011; van Praag et al. 1999).

Adult neurogenesis in the forebrain is evolutionarily conserved across mammals, birds, reptiles, amphibians and fish. However, compared to other species, neurogenic regions in the mammalian adult brain are restricted to specialized brain areas such as the SGZ of DG and the subventricular zone (SVZ) of the lateral ventricle. The DG of the mammalian hippocampus is highly convoluted and much larger than a homologous structure in other species, and adult hippocampal neurogenesis in the SGZ niche of the DG is highly conserved in most mammals (Patzke et al. 2015). Interestingly, however, in cetaceans, adult neurogenesis in the DG is less evident (Patzke et al. 2015). It is not quite clear yet whether cetaceans lost the ability to maintain adult neural stem cells in the SGZ niche of the adult brain during evolution. It is possible that adult hippocampal neurogenesis is an evolutionarily novel system that developed to increase the structural and functional plasticity in the hippocampus. It may also be possible that, due to the higher metabolic requirement to maintain neural stem cells in the brain of cetaceans, which live mostly underwater, adult neural stem cells are depleted during ontogeny. Of note, the existence of postnatal neurogenesis in the human olfactory bulb, which had previously been postulated, has been questioned despite the fact that high proliferative activity in the SVZ and migrating cells in the rostral migratory stream were observed (Bergmann et al. 2012; Curtis et al. 2007; Sanai et al. 2011). This apparent loss of neurogenesis the adult olfactory bulb in humans may be because humans rely more on the visual systems than rodents, and, in parallel, humans have lost a number of olfactory receptor genes. In this light, the retention of neurogenesis in the adult hippocampal neurogenesis in humans implies that it plays some significant role in behavior. Intriguingly, the neuroblasts and/or newborn neurons that are most likely derived from the SVZ in humans migrate into several other brain regions, including the frontal cortex and the cingulate cortex in the infant brain and the striatum in the adult brain (Ernst et al. 2014; Paredes et al. 2016; Sanai et al. 2011). Because birth is one of the largest environment changes that occurs during our life, and because birth regulates brain development (Toda et al. 2013), neurogenesis and migration into the human frontal cortex occurring during perinatal periods may provide additional plasticity to adapt to environmental changes after birth. These observations also imply that the processes of adult hippocampal neurogenesis in the human brain could also differ from those in other mammalian brains, although the differences have not yet been extensively characterized. It would be intriguing to figure out how adult hippocampal neurogenesis and perinatal neurogenesis in the human brain have evolved to functionally contribute to cognition through comparing the neurogenesis-dependent functions and genetic/environmental factors involved in adult/perinatal neurogenesis. Recent technical progresses in generating human hippocampal and cerebral organoids may enable us to eventually address some of these evolutionary questions (Bredenoord et al. 2017; Sakaguchi et al. 2015).

Extrinsic factors regulating adult neurogenesis

Morphogens, cytokines and growth factors

A number of extrinsic and intrinsic factors have been found to regulate adult neurogenesis at various developmental stages, including morphogens/growth factors and their receptors, hormones, neurotransmitters, cell adhesion molecules, cytoplasmic factors, transcriptional factors, and epigenetic modifiers. Most molecular factors are common to both embryonic and adult neurogenesis, but some have only been reported during the processes of adult neurogenesis. It is likely, and becoming apparent in some cases, that these unique molecular factors may help explain how adult neural stem cells react to complex environmental changes in different contexts that underlie the increase in the degree of plasticity observed in adult neurogenesis.

To identify the molecular factors regulating adult neurogenesis, a useful and robust in vitro system has been developed to isolate and culture adult neural stem cells. Our group has established protocols to isolate and propagate FGF-2-responsive neural progenitor cells from several areas of rat brains (Gage et al. 1995; Palmer et al. 1995) and mouse brains (Ray and Gage 2006). Furthermore, these neural progenitors possess multipotency, the ability to differentiate into several neural cell types (Palmer et al. 1997, 1999). Isolation of hippocampal adult neural progenitors allowed us to test their cellular plasticity, and the transplantation of adult rat hippocampal neural progenitors into several brain areas revealed that they could proliferate and mature in the neurogenic brain regions but not in non-neurogenic regions (Gage et al. 1995; Suhonen et al. 1996). In addition, transplanted neural progenitors can differentiate into non-hippocampal neurons (Suhonen et al. 1996), indicating that environmental cues play crucial roles in specifying the cell fate of adult-born neurons. In vitro neural stem cell culture technology has allowed researchers to prepare relatively pure material for biochemical and cell biological assays and has facilitated the identification of a number of essential factors, both cell-intrinsic and cell-extrinsic environmental, in addition to in vivo transgenic approaches.

Morphogens and cytokines have been found to be essential for the formation and maintenance of RGLs in the SGZ. For example, sonic hedgehog (Shh) signaling through the primary cilia is essential for the formation and maintenance of RGLs (Ahn and Joyner 2005; Breunig et al. 2008; Han et al. 2008; Li et al. 2013). A recent report showed that some hippocampal RGLs derive from the ventral hippocampus and that distinct sources of Shh signaling contribute to the formation and maintenance of RGLs, depending on developmental timing (Li et al. 2013). It is not yet clear whether all RGLs in the SGZ of the adult brain derive from the ventral hippocampus, since RGLs could be generated locally (Li and Pleasure 2005), but it will be intriguing to determine whether the heterogeneity of RGLs is derived from their developmental origins.

Bone morphogenetic protein (BMP) signaling is also implicated in the regulation of adult neurogenesis. BMPs are a subgroup of the transforming growth factor-beta of cytokines, and BMPs are supplied by the local microenvironment. BMP signaling regulates the balance between proliferation and quiescence of RGLs through BMPR-IA and Smad4, as well as the amplification and maturation of intermediate neural progenitor pools through BMPRII (Bonaguidi et al. 2008; Bond et al. 2014; Mira et al. 2010). Thus, BMP signaling modulates several aspects of the process of adult neurogenesis using different receptors, suggesting that the specific receptor-signaling cascades may contribute to organizing the strength of downstream BMP signaling in the context of adult neurogenesis. Recently, it was reported that the levels of Bmp4 and Bmp6 are increased in endothelial cells and microglia of the hippocampus, and the attenuation of BMP signaling could increase the proliferation of neural progenitors in the aged hippocampus (Yousef et al. 2015). These findings suggest that the increase in BMP secretion as a result of aging of the environmental niche could be one reason that adult neurogenesis is reduced during aging, implying that the reduction in adult neurogenesis during aging could be the consequence of systemic changes in the brain. Additionally, it appears that an increase in Noggin, which is a BMP signaling antagonist, mediates the effect of antidepressant treatment through an increase in adult neurogenesis (Brooker et al. 2017). Thus, the modulation of BMP signaling could be a therapeutic target for treating adult neurogenesis-related diseases. Interestingly, certain levels of BMP4 in combination with FGF-2 reversibly induce the quiescent state of neural stem cells in vitro. Epigenetic profiling of these cells has identified nuclear factor one X as a mediator to induce the quiescent state (Martynoga et al. 2013). Since BMP signaling also promotes astrogenesis (Bonaguidi et al. 2005), the tight control of BMP signaling seems to be critical in the balance between the maintenance and differentiation of adult neural stem cells. Although it is still not clear how much BMP4/FGF-2-induced quiescent neural stem cells in vitro resemble RGLs in vivo, further intensive characterization, such as transcriptional and epigenetic comparisons, may be able to establish a powerful in vitro model to investigate bona fide quiescent neural stem cells.

Canonical Wnt/β-catenin signaling and non-canonical Wnt signaling also play fundamental roles in the regulation of adult neurogenesis (also reviewed in Goncalves et al. 2016b; Inestrosa and Arenas 2010). Wnt3, presumably secreted from astrocytes, induces the nuclear translocation of β-catenin and subsequently activates the expression of NeuroD1 and Prox1, which facilitates proliferation and neuronal cell fate specification (Karalay et al. 2011; Kuwabara et al. 2009; Lie et al. 2005; Song et al. 2002). Ephrin-B signaling also regulates the proliferation and cell fate specification by activating the Wnt signaling pathway (Ashton et al. 2012). Recently, our group also found that non-canonical Wnt/Planer cell polarity signaling regulates the neurogenesis as well as morphogenesis of DG neurons (Schafer et al. 2015). Thus, similar to the BMP signaling pathway, the same morphogens could work in the different stages of adult neurogenesis through distinct pathways.

In addition to locally secreted factors, recent studies have shown that systemic milieux play prominent roles in the regulation of adult neurogenesis. Using homo- or heterochronic parabiosis between young and old animals, one study has shown that several chemokines, such as CCL11, increase with age, and the increase in these chemokines reduces adult neurogenesis and impairs hippocampus-dependent learning and memory (Villeda et al. 2011). Subsequent studies showed that β2-microglobulin, a component of the major histocompatibility complex class 1 molecule that is known to be involved in synaptic plasticity (Corriveau et al. 1998; Huh et al. 2000), inhibits adult neurogenesis. β2-microglobulin is somehow secreted, circulates in the blood, and, interestingly, the levels of β2-microglobulin increase with age (Smith et al. 2015). A recent report also found a positive systemic factor for cognitive functions from human umbilical cord plasma (Castellano et al. 2017). Therefore, although the source of systemic changes due to aging and the underlying molecular mechanisms through which those changes regulate adult neurogenesis remain elusive, identifying the positive and negative systemic factors for adult neurogenesis/plasticity should enable researchers to develop therapeutic tools for age-dependent cognitive decline.

Cell–cell contact-mediated signaling such as Notch signaling is also essential to maintain RGLs. Notch signaling regulates proliferation and cell fate specification of RGLs and intermediate progenitor cells, and it plays fundamental roles in many aspects of developmental processes as well as in the maintenance of adult brains (Ables et al. 2011; Imayoshi and Kageyama 2014; Lui et al. 2011). Depletion of the downstream mediator of the Notch pathway, RBPj, induced the transient increase in adult neurogenesis, which subsequently caused the depletion of adult neural progenitors and reduced the adult neural pools (Ehm et al. 2010). Similarly, conditional knockout of Notch1 in RGLs resulted in a reduction in the numbers of adult-born neurons as well as RGLs and intermediate neural progenitors (Ables et al. 2010). Interestingly, physical activity rescued the total number of newborn neurons in the conditional Notch-1 KO mice without resulting in an increase in RGLs and intermediate progenitors, in contrast to the normal condition (Suh et al. 2007). Thus, Notch signaling is critical for the proliferation and activation of RGLs and intermediate progenitors but is not necessary for the proliferation of neuroblasts. Physical voluntary activity or the modulation of downstream signaling of physical activity may provide a therapy for neurogenesis-related disorders without increasing the RGL and intermediate progenitor pools.

Synaptic and non-synaptic regulation of newborn neurons

One of the unique features of adult neurogenesis is that a number of neurotransmitters are already in place to regulate the process. Compared to the embryonic brain, the adult brain has built up highly organized neural circuits and the DG receives intensive glutamatergic innervations from the entorhinal cortex (the perforant pathway) as well as local inputs from several types of GABA (γ-aminobutyric acid)-ergic interneurons (Fig. 2) (Amaral et al. 2007). In addition, the DG receives a variety of modulatory inputs, including cholinergic and GABAergic inputs from the septal nuclei, glutamatergic inputs from the supramammillary area, serotonergic inputs from the raphe nuclei, noradrenergic inputs from the nucleus locus coeruleus and dopaminergic inputs from the ventral tagmental area (Fig. 2) (Amaral et al. 2007). Thus, the anatomical features of the neural circuits innervating the DG imply that experience-dependent neural activity, as well as the emotional state of animals, has the potential to directly and/or indirectly regulate adult neurogenesis on demand. In fact, glutamatergic inputs through NMDA receptors were found to be critical for the survival of immature neurons in a time-dependent manner (Tashiro et al. 2006). There are several sources of glutamatergic input onto adult-born neurons. Combining electron microscopy and retrovirus-based newborn neuron tracing, our group found that some of the inputs likely originated from the entorhinal cortex, and newborn neurons preferentially contacted preexisting axonal boutons (Toni et al. 2007). Two- to four-week-old adult-born neurons compete with existing mature DG neurons to receive glutamatergic synaptic inputs from the entorhinal cortex to be integrated into the circuit (Toni et al. 2007). Interestingly, decreasing the spine density of mature DG neurons by a genetically inducible system increased the integration and survival of adult-born neurons, and the increased integration of adult-born neurons in aged animals improved contextual memory (McAvoy et al. 2016), suggesting that increasing adult hippocampal neurogenesis may be implemented to improve memory precision in aged humans.

Fig. 2
figure 2

Synaptic and non-synaptic inputs during adult hippocampal neurogenesis

Full size image

More recently, it has been shown that vesicular release from astrocytes is also critical for the NMDA receptor-dependent integration of newborn DG neurons (Sultan et al. 2015). Furthermore, by taking advantage of a rabies virus-based monosypantic retrograde tracing technique (Wickersham et al. 2007), synaptic inputs onto adult-born neurons were traced and a variety of afferents have been located, including mossy cells, interneurons, CA3 neurons, and mature DG neurons, in addition to cholinergic septal cells in the lateral entorhinal and perirhinal cortex (Deshpande et al. 2013; Vivar et al. 2012). With neuronal maturation, the fraction of inputs from hilar cells and the entorhinal cortex increased, whereas the fractions of inputs from mature DG neurons decreased, suggesting that individual inputs probably contribute to distinct steps in the survival, integration, and maturation of adult-born neurons. This connectivity between adult-born neurons and CA3/mature DG neurons was surprising, and it would be interesting to determine how these feedback/feed-forward inputs onto adult-born neurons modulate the integration of adult-born neurons.

Intriguingly, the connectivity of adult newborn neurons also depends on environmental effects. When animals were reared in an enriched environment during the critical period for the development of adult-born neurons, adult-born DG neurons received a considerably higher amount of inputs from local neurons (hilar interneurons, mature DG neurons, mossy cells) as well as distal neurons (cholinergic septal neurons and the entorhinal cortex) (Bergami et al. 2015). More surprisingly, innervations from interneurons in the CA1 and CA3 of the hippocampus and from the mammillary bodies were almost exclusively observed in mice reared in an enriched environment, suggesting that an experience-dependent reorganization of the presynaptic connectivity of adult-born neurons may underlie the effects of environmental enrichment on adult neurogenesis (Kempermann et al. 1997b), and that reorganized neural circuits, including feedback inhibitory inputs from CA1 and CA3 onto adult-born neurons, could modulate synaptic integration and information processing in the DG. Interestingly, physical voluntary exercise shows slightly different effects on the connectivity compared to enriched environment, suggesting that the reorganization of connectivity onto the adult-born neurons is really context dependent. This finding may also explain the different consequences between the relationship of exercise and enrichment on adult neurogenesis and behavior. Thus, the integration of newborn neurons depends on activity-dependent competitive processes (Tashiro et al. 2006, 2007) and neuronal network activity as well as local release from astrocytes to regulate these processes.

Newborn DG neurons also receive tonic GABA inputs, presumably from ambient GABA, which are thought to be the first synaptic inputs onto adult-born neurons (Esposito et al. 2005); these inputs activate immature neurons due to a higher concentration of intracellular chloride ions in immature neurons (Ge et al. 2006). This GABA-induced excitation of immature neurons regulates the dendritic maturation as well as synaptic integration of immature neurons, but it has been controversial whether GABA inputs actually elicit action potentials in young adult-born neurons (Ge et al. 2006; Mongiat et al. 2009; Overstreet Wadiche et al. 2005). In an elegant recent report, more systematic electrophysiological analyses in GABAergic synaptic inputs onto the young adult-born neurons have been conducted and revealed that GABAergic inputs from the molecular layer actually depolarized doublecortin-positive immature neurons but did not evoke action potentials (Heigele et al. 2016). However, when GABA inputs from both the molecular layer and the granule cell layer were temporally summed in the range of gamma frequency, this combination efficiently evoked action potentials in young adult-born neurons. However, GABA inputs that were too strong inversely interfered with the initiation of action potentials, suggesting that young adult-born neurons responded to GABA inputs in a biphasic manner (Heigele et al. 2016). Furthermore, the study revealed the spatial and temporal integration dynamics of GABAergic and glutamatergic inputs to elicit action potentials in young adult-born DG neurons. This finding is quite striking, since gamma waves have been implicated in seeking behavior as well as memory association (Igarashi 2015). Thus, these findings suggest that the physiological state of the hippocampus could regulate the recruitment of young DG neurons into neuronal networks as well as their integration into the hippocampal circuits through GABAergic signaling. It would be interesting to compare the neuronal activity of young DG neurons and the power of gamma waves in a different environment such as an enriched environment or physical voluntary running.

In agreement with this line of research, the power of hippocampal gamma waves was found to be greater in animals housed in an enriched environment (Shinohara et al. 2013); more recently, researchers using optogenetics and chemogenetics have shown that GABAergic inputs from parvalbumin-positive interneurons, which are thought to be responsible for generating gamma waves, were found to be essential for enriched environments to enhance the integration and maturation of young DG neurons (Alvarez et al. 2016). GABA released from parvalbumin-positive interneurons also regulated the maintenance of quiescent RGLs through γ2-subunit-containing GABAA receptors (Song et al. 2012), as well as supporting the survival of newborn neurons (Song et al. 2013). A recent report has shown that diazepam binding inhibitor, an endogenous negative modulator of GABAA receptor, was enriched in RGLs and regulated the proliferation of RGLs and intermediate neural progenitors (Dumitru et al. 2017), suggesting that GABA signaling is tightly tuned to new neurons, responding to local neuronal activity and experience. Importantly, in addition to the higher excitability of adult-born DG neurons compared to mature DG neurons, immature neurons exhibited NR2B receptor-dependent higher synaptic plasticity between 4 and 6 weeks of cellular age, a so-called critical period for the development of adult-born neurons (Ge et al. 2007).

These characteristic features of adult-born DG neurons allow them to respond to a broad range of synaptic inputs and to compete with mature neurons to be integrated into the existing neural circuits. Local network activity in the hippocampus regulates the proliferation and integration rates of neural stem cells/immature DG neurons through both glutamatergic and GABAergic signals. The newly integrated DG neurons project feed-forward excitation into the CA3 pyramidal neurons, mossy cells and hilar interneurons (Gu et al. 2012; Toni et al. 2008; Zhao et al. 2006), which is critical for memory retrieval, as well as providing feedback inhibition into the mature DG neurons to control the spare coding in the DG (Temprana et al. 2015), which plays prominent roles in the structural plasticity of hippocampal circuits and the entire hippocampal network activity. Recent studies reported that DG neurons also directly project to the CA2 pyramidal neurons making synaptic contacts, in contrast to previous observations (Kohara et al. 2014; Llorens-Martin et al. 2015). The CA2 pyramidal neurons innervate the deep layer of CA1, suggesting the existence of an alternate to the tri-synaptic circuit in the hippocampus (Kohara et al. 2014) (Fig. 1). Furthermore, adult-born DG neurons project to the CA2 region (Llorens-Martin et al. 2015), and the CA2 region plays key roles in social memory and contextual discrimination (Hitti and Siegelbaum 2014; Wintzer et al. 2014). It would be intriguing to investigate the functional difference and convergence between the classical tri-synaptic circuit and the novel CA2-mediated synaptic circuit.

Intrinsic factors regulating adult neurogenesis

Extrinsic signals activate stage-specific transcription factors (Fig. 1) that are fundamental to regulating cell type-specific gene expression and eventually proliferation, differentiation, cell fate specification, and functional maturation.

One of the prominent transcription factors is a sex-determining region Y box2 (Sox2), which is highly expressed in type1 and type 2a cells and acts as a mediator of Notch signaling and regulator of Shh expression (Ehm et al. 2010; Favaro et al. 2009). Sox2 also regulates the expression of Notch signaling components (Bani-Yaghoub et al. 2006), so these factors constitute a feedback loop to maintain the state of neural stem cells. Sox2 also works as a hub to recruit co-transcriptional regulators, and transcriptional factors such as Prx1, Brn2 (Lodato et al. 2013; Shimozaki et al. 2013) as well as epigenetic regulators HDAC1, MeCP2, Chd7 and RMST have been found to cooperate with Sox2 in neural stem cells (Engelen et al. 2011; Kuwabara et al. 2009; Ng et al. 2013; Szulwach et al. 2010). Sox2 regulates other critical transcription factors such as the nuclear orphan receptor Tlx (Shimozaki et al. 2012), which is critical for the regulation of adult neurogenesis as well as hippocampus-dependent learning and memory (Shi et al. 2004; Zhang et al. 2008). We recently found that one of the nuclear pore complex proteins, Nup153, interacts and cooperates with Sox2 to maintain the cellular state of neural stem cells, supporting the idea that a Sox2 partnership plays a central role in maintaining the identity of neural stem cells (Toda et al. 2017).

The other prominent transcription factor is the repressor element 1-silencing transcription (REST) and its co-factors. These factors have been reported to regulate multipotency and play indispensable roles in adult neurogenesis (Ballas et al. 2005; Gao et al. 2011; Mukherjee et al. 2016). REST is expressed from type 1 to mature DG neurons; it controls quiescent RGLs to retain neural stem cell pools and directly regulates several downstream pathways, including neurogenic genes as well as ribosome biogenesis and cell cycle. REST also non-cell-autonomously regulates neuronal differentiation through the secretion of Secretogranin II (Kim et al. 2015).

Other transcription factors that are intensively expressed in adult neural stem cells to regulate quiescence and proliferation are basic helix–loop–helix (bHLH) transcription factors such as Hes1, Hes5, Ascl1 (Mash1), Neurog1, Neurog2, Ids, Olig1, and Olig2. These factors are downstream mediators of Notch signaling, and research in embryonic neural stem cells has revealed that the oscillatory expression of bHLH factors is vital to maintaining a proliferative state and that sustained expression of single bHLH factors results in differentiation into neurons, astrocytes or oligodendrocytes (Imayoshi et al. 2013; Imayoshi and Kageyama 2014; Kageyama et al. 2015). Therefore, it will be intriguing to examine whether bHLH factors also oscillate in the quiescent adult neural stem cells. In the context of adult neurogenesis, Ascl1 has been found to have multiple functions. Originally, it was shown that Ascl1 promoted neurogenesis (Tomita et al. 1996). Subsequently, it was reported that Ascl1 was expressed in type 2a intermediate progenitors, which in turn mostly give rise to DG neurons. However, overexpression of Ascl1 in the dividing neural progenitor in the DG gives rise to oligodendrocytes (Jessberger et al. 2008), showing that adult hippocampal progenitors possess substantial cellular plasticity in vivo. In addition, a recent report found that a small subset of type 1 RGLs also expressed moderate levels of Ascl1, and about 30% of proliferating RGLs were Ascl1-positive (Andersen et al. 2014). Furthermore, neurogenic stimuli, such as kainite acid administration or conditional knockout of RBPJk, induced the expression of Ascl1. Additionally, the study found that the induction of Ascl1 in RGLs was essential to activate quiescent RGLs, and conditional knockout of Ascl1 in RGLs strongly blocked neurogenesis due to the failure to exit from a quiescent state. The study also found that Ascl1 directly bound to several cell cycle regulator genes such as Ccnd2 and Rm2 to regulate their expression, providing direct evidence for the regulation of cell cycle by the expression of Ascl1 (Andersen et al. 2014). The same group also found that the destabilization of Ascl1 by E3-ubiquitin ligase Huwe1 in proliferating neural stem cells was necessary for the cells to return to a quiescent state, and conditional knockout of Huwe1 resulted in the increase in Ascl1 and depletion of adult neural stem cell pools (Urban et al. 2016). Taken together, these studies indicate that not only the regulation at the transcriptional level but also the proteostatic regulation of bHLH factors are critical to maintaining adequate levels of adult neurogenesis throughout life.

The other transcription factors regulating adult neurogenesis are FoxOs, Tbr2, and CREB (Hodge et al. 2012; Jagasia et al. 2009; Nakagawa et al. 2002; Renault et al. 2009). At the stage of fate specification, NeuroD1 and Prox1 play prominent roles in the specification of DG neurons. Activation of canonical Wnt signaling up-regulates the expression of NeuroD1 through the activation of the β-catenin/TCF/LEF complex (Gao et al. 2009; Kuwabara et al. 2009), which subsequently activates the expression of Prox1 (Karalay et al. 2011). Prox1 is an essential factor for acquiring the characteristic features of dentate granule neurons, and loss of Prox1 in post-mitotic immature DG neurons misdirects them to differentiate into CA3-like neurons (Iwano et al. 2012). Importantly, the expression of Prox1 is maintained even after the maturation of DG neurons, implying that the sustained expression of Prox1 is also essential to maintain the identity and function of DG neurons. After fate determination, Krüppel-like factor 9 (Klf-9) is a transcriptional factor regulating the functional maturation of DG neurons (Scobie et al. 2009). The expression of Klf-9 is induced by neuronal activity, and loss of Klf-9 affected the survival of adult-born neurons as well as neurogenesis-dependent synaptic plasticity (Scobie et al. 2009), suggesting that Klf-9 works as a downstream mediator of synaptic inputs onto maturating adult-born DG neurons. Taken together, these well-organized transcriptional cascades are tightly regulated from the activation of RGLs to the maturation of DG neurons through a combination of environmental and synaptic inputs.

In addition to transcription factors, accumulating evidence has shown that a number of epigenetic factors, such as histone modifiers, DNA methylase/demethylase, and microRNA, are involved in the regulation of adult neurogenesis. Histone modifications are often associated with the transcriptional status of genes, and the N-terminal tails of histones receive several post-translational modifications, such as acetylation, methylation, and phosphorylation. Histone deacetylases (HDACs) have pleiotropic roles in the gene regulation of neurogenesis. Selective knockout of HDAC2 in adult neural stem cells resulted in mis-expression of Sox2 in immature neurons and impaired the survival of adult-born neurons (Jawerka et al. 2010). HDAC1 interacts with Sox2 and presumably represses the expression of NeuroD1 as well as LINE-1 retrotransposons (Kuwabara et al. 2009).

DNA methylation has been intensively studied and plays significant roles in the regulation of temporal genes during development and genomic stability, as well as in the context of adult neurogenesis. Methyl-CpG binding proteins are major readers of DNA methylation, and they bind to methylated DNA to recruit other factors to modulate gene expression. Methyl-CpG binding protein 1 (MBD1)-knockout mice show impairments in neurogenesis and the maintenance of RGLs in the adult brain (Jobe et al. 2017; Zhao et al. 2003). Mechanistically, MBD1 directly binds the regulatory genome regions of FGF-2, miR-184, and miR-195, and represses their expression to regulate the proliferation and differentiation of adult neural stem cells (Li et al. 2008; Liu et al. 2013, 2010). Methyl-CpG binding protein (MeCP2) is another well-studied factor, since mutations in MeCP2 have been found to be associated with Rett syndrome (Amir et al. 1999; Guy et al. 2001). In MeCP2-knockout mice, the maturation of adult-born neurons such as spinogenesis is perturbed (Smrt et al. 2007), and the phosphorylation of MeCP2 also modulates adult neurogenesis (Li et al. 2014). MeCP2 has multiple functions in gene regulation and interacts with several other factors, including the key transcription factor Sox2, the HDAC complex, the NCoR repressors, and the microprocessor Drosha complex (Lyst et al. 2013; Nan et al. 1998; Szulwach et al. 2010; Tsujimura et al. 2015), and they work together as a hub for epigenetic regulation. MeCP2 also maintains genome stability by inhibiting L1 retrotransposition (Muotri et al. 2010). Interestingly, recent findings indicate that DNA methylation involves many more dynamic processes than previously thought (Guo et al. 2011; Zeng et al. 2016), and that neuronal activation of DG neurons by electroconvulsive stimulation induces Gadd45b in mature neurons, which mediates active DNA demethylation, and induces Gadd45b to demethylate the promoter of FGF1 and BDNF genes. This demethylation eventually increases the expression of target genes, and this activity-induced gene expression presumably promotes the proliferation and morphogenesis in a non-cell-autonomous manner (Ma et al. 2009). A recent study that has systemically examined the histone modification and DNA methylation among excitatory neurons, and two types of interneurons revealed that different neural cell types possess distinct DNA methylation profiles in addition to gene expression (Mo et al. 2015). Surprisingly, the study found that non-CpG methylation within intragenic regions associated well with gene expression compared to CpG methylation, suggesting that another layer regulating DNA methylation is in place to regulate cell-type-specific gene regulation. It would be intriguing to investigate the role of non-CpG methylation in the context of adult neurogenesis.

In addition to these canonical epigenetic factors, our group has recently discovered that one of the nuclear pore complex proteins, Nup153, interacts with Sox2 and regulates the maintenance of neural progenitors through direct gene regulation (Toda et al. 2017). Although a classical role of nuclear pore complex proteins is the regulation of nuclear-cytoplasmic transport, accumulating evidence suggests that nuclear pore proteins also play prominent roles in gene regulation. Thus, our finding highlights another layer of epigenetic gene regulation by the nuclear pore proteins, which could be associated with non-random genome distribution in the nucleus or nuclear structure, the so-called nuclear architecture.

Environmental effects on adult neurogenesis

One of the intriguing and fundamental features of adult neurogenesis is that the processes of neurogenesis are regulated by an individual’s behavior, experience, and emotional/biological status. This on-demand neurogenesis in response to physiological and environmental signals provides an additional layer of plasticity in the hippocampal circuits. An enriched environment has been shown to induce structural and functional changes in the brain (Bartoletti et al. 2004; Fiala et al. 1978; Rosenzweig et al. 1962). Our group tested whether environmental enrichment could lead to a higher number of neurons in the DG. Strikingly, exposure to an enriched environment significantly increased the number of newborn cells and the volume of the granule cell layer and improved the speed of spatial learning (Kempermann et al. 1997b). Subsequently, studies revealed that voluntary running selectively increased the number of newborn cells, whereas environmental enrichment promoted the survival of newborn cells (van Praag et al. 1999). Still, it remains unclear how environmental enrichment and voluntary running can increase proliferation and survival; one possibility is, as described above, that environmental signals could change the neural network activity in the hippocampus that modulates synaptic inputs on immature neurons as well as epigenetic regulation.

In contrast, stress and aging decrease the number of adult-born neurons in the DG through corticosteroid signaling (Cameron and McKay 1999; Gould et al. 1997; Kuhn et al. 1996). These studies highlight the fact that the rate of adult neurogenesis and/or that of the integration of newborn neurons could be regulated by environmental and biological factors in both a positive and a negative direction, and its consequences could affect brain function. These studies serve as a basis for follow-up studies to investigate how environmental stimuli and pathological conditions regulate and affect adult neurogenesis, how adult-born neurons contribute to brain function, and how we can implement the accumulating knowledge to develop medicines to treat people.

Functional impacts of adult-born neurons

Since the discovery of adult neurogenesis, the central question has been how adult-born neurons influence cognition and behavior. To address this question, several methodologies have been applied to ablate adult neurogenesis or silence adult-born neurons. Accumulating evidence has suggested that adult-born neurons play significant roles in cognition, learning and memory, although the results of studies have also shown inconsistencies depending on the methodology used to ablate/reduce adult-born neurons, the species and strains used, the experimental design for the behavioral assessment, and the timing of tests after the manipulation of adult neurogenesis; all these factors have made it difficult to obtain a consensus regarding the basic principles of the functional roles of adult neurogenesis. In this section, we summarize the experimental observations based on different strategies and discuss future directions for research.

Learning, memory and cognitive flexibility

One common strategy used to study the function of adult neurogenesis is to ablate adult-born neurons and examine the cognitive performance of animals. To assess the functional roles of hippocampal adult neurogenesis, many studies have investigated hippocampus-dependent cognitive abilities such as contextual and spatial memories after the ablation of adult-born neurons. Initially, antimitotic reagents such as methylazoxymethanol acetate (MAM) or temozolomide (TMZ) were used to reduce adult neurogenesis in the adult rat and mouse (Garthe et al. 2009; Shors et al. 2001). Treatment by MAM reduced adult neurogenesis in the hippocampus and resulted in deficits in hippocampus-dependent trace eyeblink conditioning but not in hippocampus-independent delay eyeblink conditioning tasks (Shors et al. 2001). After withdrawal of MAM, as adult neurogenesis recovered, the trace eyeblink conditioning recovered (Shors et al. 2001), suggesting that adult neurogenesis plays a role in hippocampus-dependent contextual memory. Similarly, MAM treatment canceled out the beneficial effect of environmental enrichment on long-term memory during a novel-object recognition task in the adult rat but made no changes in contextual fear memory (Bruel-Jungerman et al. 2005; Shors et al. 2002). Interestingly, the ablation of adult neurogenesis by TMZ in mice affected not only the speed of learning but also the flexibility in the Morris water maze test (Garthe et al. 2009). Thus, adult-born neurons may function not only in learning processes but also in the plasticity of learning strategies.

Since the chemical ablation of adult neural stem cells could involve other side effects, several groups subsequently developed novel strategies to selectively ablate adult neurogenesis, such as focal X-ray irradiation and genetic ablation based on transgenic animals using neural progenitor-specific promoters (Clelland et al. 2009; Deng et al. 2009; Imayoshi et al. 2008; Meshi et al. 2006; Saxe et al. 2006, 2007). As opposed to the results using MAM in rats (Shors et al. 2002), focally irradiated mice, depleted of adult neurogenesis, showed impairments in contextual fear conditioning and a formation of long-term potentiation (LTP) in the DG but not in cued fear conditioning and spatial memory 3 months after irradiation (Saxe et al. 2006). Similarly, although MAM blocked the effect of enriched environment on novel-object recognition test in rats (Bruel-Jungerman et al. 2005), the ablation of adult-born neurons by focal X-ray irradiation onto the mouse hippocampus did not alter the effects of an enriched environment on spatial learning with the Morris Water maze (Meshi et al. 2006). In another study, 11 days after X-ray irradiation of the whole brain, the magnitude of LTP was not altered but the retention of LTP was enhanced in the DG of rats (Kitamura et al. 2009), suggesting that learning-induced LTP in the DG might be reversed by adult-born DG neurons, which might be the basis of the decay of the hippocampus-dependent memory as well as hippocampus-dependent cognitive flexibility (Garthe et al. 2009). Since the feedback inhibition from young adult-born DG neurons suppresses the activity of mature DG neurons (Temprana et al. 2015), the ablation of newborn neurons may temporarily increase the excitability of mature DG neurons. In line with this notion, a recent study reported the role of adult neurogenesis in the clearance of hippocampus-dependent memory and infantile amnesia (Akers et al. 2014; Epp et al. 2016). However, another study reported that an episodic memory during the infantile amnesia period was not completely cleared but stored as a latent memory for long periods of time. (Travaglia et al. 2016). The study also showed that, although recent fear contextual memory was slightly impaired in the irradiated mice 5 weeks after irradiation, remote fear contextual memory was not perturbed in the irradiated mice (Kitamura et al. 2009). Interestingly, the study found that adult neurogenesis modulated the speed of memory transfer from the hippocampus to elsewhere, most likely into the neocortex, suggesting that adult neurogenesis could also modulate the relocation and long-term storage of memory in the brain.

As an independent approach to specifically ablate neural stem cells, or temporally control the rate of adult neurogenesis, transgenic animal models were developed: Nestin-CreERT2, in which a tamoxifen-inducible Cre recombinase is expressed under the control of a neural stem cell-specific nestin promoter (Imayoshi et al. 2008), or Nestin-tk and GFAP-tk mice, in which a thymidine kinase is expressed under the control of neural stem cell/glial cell-specific promoters (Deng et al. 2009; Saxe et al. 2007). Ablation of adult-born neurons by the combination of Nestin-CreERT2 and NSE-DTA (diphtheria toxin A) mice impaired spatial learning and memory with the Barnes maze test (Imayoshi et al. 2008). Treatment of Nestin-tk or GFAP-tk mice by ganciclovir (GCV) allowed researchers to temporally reduce adult neurogenesis. After a 6-week GCV treatment, GFAP-tk mice exhibited a deficit in contextual fear conditioning but not in cued conditioning (Saxe et al. 2006). On the other hand, the reduction of adult neurogenesis by a 2-week treatment of Nestin-tk mice with GCV impaired spatial long-term memory in the Morris water maze but did not affect learning processes and memory extinction (Deng et al. 2009). In the same mouse model with the contextual fear memory test, extinction of short-term memory was impaired but not learning itself, and the effect reverted after the recovery of neurogenesis (Deng et al. 2009), suggesting a specific role for adult neurogenesis in the process of learning and memory.

Overall, a number of studies support the idea that adult hippocampus neurogenesis plays significant roles in hippocampus-dependent cognitive functions, including learning, memory, and cognitive flexibility, although the results and interpretations vary depending on the method used to ablate adult-born neurons, the species used in the studies, and the behavioral assays; therefore, further investigations are required. It has been noted that the genetic background and ages of animals strongly influence the rate of adult neurogenesis (Kempermann et al. 1997a; Kuhn et al. 1996), and thus the impacts of adult neurogenesis on hippocampal function may vary among different species and strains; we need to carefully take this point into account when comparing different studies.

Pattern separation

In addition to contextual memory, hippocampal adult neurogenesis has been postulated to function in the process of pattern separation, a computational process that is employed to discriminate inputs from entorhinal cortex to separate information and to encode that information as distinct memories, based on the firing patterns of groups of neurons. Due to its anatomical and physiological features (Aimone et al. 2014; Deng et al. 2010), the DG has been considered to be a pattern separator, and this notion has been experimentally tested (Leutgeb et al. 2007; McHugh et al. 2007). The addition of highly excitable newborn DG neurons could impact the process of pattern separation and may shape spatial and temporal pattern separation. The details of pattern separation combined with adult neurogenesis are reviewed in Aimone et al. (2014) and Deng et al. (2010). To experimentally test the pattern separation hypothesis, focally irradiated mice exhibited impaired spatial pattern separation. Therefore, the DG was suggested to function as a pattern separator for spatial or episodic memory (Clelland et al. 2009). Similarly, attenuation of adult neurogenesis by treatment with TMZ or GCV in Nestin-tk mice impaired spatial pattern separation and CA3 population coding (Niibori et al. 2012). Complementary approaches that increased adult neurogenesis with voluntary running or inhibited programmed cell death by ablating Bax improved pattern separation (Creer et al. 2010; Sahay et al. 2011) and also supported the idea that adult-born neurons were in place to regulate spatial pattern separation. However, in contrast, another study showed that ablation of neurogenesis by X-ray irradiation improved working memory in a low-memory load/high interference radial maze-task (Saxe et al. 2007). Thus, depending on the temporal and spatial information, immature neurons may have distinct functions in pattern separation.

Recent advanced transgenic approaches also allowed researchers to compare the functions of mature and immature DG neurons using cell-type-specific expression of tetanus toxin (Nakashiba et al. 2012). When the synaptic transmission from mature DG neurons was inhibited, the ability to discriminate different contexts in the contextual fear memory test was enhanced, but this enhancement was blocked by ablating adult-born neurons with irradiation, supporting the idea that immature DG neurons are responsible for spatial pattern separation. Another study investigated the functional impact of adult-born DG neurons from a different angle by increasing the integration of adult-born neurons. The study found that, as the integration of adult-born neurons increased, local connectivities involved in adult-born neurons were reorganized and the precision of contextual fear memory was improved in both young and old mice, suggesting that a fraction of integrated adult-born DG neurons is crucial for pattern separation (McAvoy et al. 2016).

Overall, adult-born neurons in the DG seem to be functionally important in many different aspects of hippocampus-dependent function such as short-term and long-term memory formation, the flexibility of learning strategies, and pattern separation.

Emotional control by adult neurogenesis

In addition to learning and memory, accumulating evidence suggests that adult neurogenesis in the DG regulates emotional status, such as anxiety and depression, in conjunction with cognitive flexibility. It has been shown that hippocampal adult neurogenesis is very sensitive to the effects of stress from the perinatal period to adulthood (Gould et al. 1997; Lehmann et al. 2013; Lemaire et al. 2000; Mirescu et al. 2004), and that adult neurogenesis is required for some of the beneficial effects of antidepressants through 5HT1A receptors (Santarelli et al. 2003). Interestingly, most studies have shown that the ablation of adult-born neurons does not affect the baseline levels of anxiety (David et al. 2009; Kitamura et al. 2009; Saxe et al. 2006), suggesting that adult neurogenesis is important for the stress response but not in every case (Bessa et al. 2009). It is not yet clear how adult hippocampal neurogenesis regulates anxiety levels, but some evidence indicates that adult-born neurons buffer the stress responses, and that the depletion of adult-born neurons augments stress responses and increases anxiety and depression-like behavior (Snyder et al. 2011). Since the ventral subiculum in the hippocampus inhibits the activity of the hypothalamo-pituitary-adrenocortical axis, a critical adaptive system against physical and psychological challenges (Jankord and Herman 2008), adult-born neurons may somehow balance the activity levels of the ventral subiculum. Recent optogenetic approaches have revealed that the ventral hippocampus is associated with social memory and anxiety, and that the activation of the ventral DG neurons can reduce anxiety levels (Kheirbek et al. 2013; Okuyama et al. 2016), suggesting that differential circuits along the dorsoventral axis of the hippocampus may process distinct information. However, it is possible that the dorsal hippocampus could also regulate anxiety-/depression-related behavior through the amygdala (Ramirez et al. 2015). Further investigation will be required to figure out how adult-born neurons buffer stress responses at the circuit levels.

Future directions

As summarized above, since adult neurogenesis was “rediscovered”, the molecular mechanisms underlying adult hippocampal neurogenesis have been intensively studied. A number of intrinsic and extrinsic factors regulate the maintenance, proliferation, and differentiation of adult neural stem cells/progenitors. In addition, spatially and temporally organized synaptic/non-synaptic inputs onto adult-born neurons regulate the integration of adult-born neurons into the existing neural circuits, which provides substantial structural plasticity into the tri-synaptic hippocampal circuits. Since the number of adult-born neurons and the integrated numbers of neurons into the existing circuits are regulated by several intrinsic and environmental factors, as summarized above, these observations suggest that the degree of hippocampal plasticity, which is adult hippocampal neurogenesis-dependent meta-plasticity, can also be regulatable. It would be intriguing to investigate to what extent the adult-hippocampal neurogenesis-dependent meta-plasticity can be regulated in physiological and pathological conditions. Importantly, it is not yet clear how adult-born neurons modify the network activity and contribute behaviorally to hippocampal functional plasticity in vivo. Most of our knowledge about the function of adult neurogenesis comes from behavioral studies using the ablation of adult-born neurons or from electrophysiological analysis in vitro. Therefore, we do not yet know how adult-born neurons contribute to hippocampal functions in the natural context. Several groups, including ours, have been developing novel live-imaging methodology to measure neuronal activity in the DG in awake, behaving mice using a multi-photon microscopy or microendoscopy (Danielson et al. 2016, 2017; Goncalves et al. 2016a; Kirschen et al. 2017; Pilz et al. 2016). In combination with cell type-specific expression of calcium indicators, these methods allow us to track neural activity in young adult-born DG neurons, mature DG neurons, mossy cells, and interneurons in the hilus over time, and enable us to measure the impact of adult-born neurons on hippocampal plasticity. Although current methodologies require the removal of the neocortex and/or the CA1 regions for imaging the DG, in vivo imaging will likely uncover more precisely the functions of adult-born DG neurons depending on the timing of maturation, the state of the neural network, and the history of neuronal activity. Further technical developments, such as 3-photon microscopy, may one day allow intact imaging of the DG (Ouzounov et al. 2017).

Looking back on the relatively short history since the rediscovery of adult neurogenesis, remarkable progress in our understanding of the structure and function of this most plastic process in the adult nervous system has been made, and new technologies are promising a deeper understanding.