Re-evaluating the link between neuropsychiatric disorders and dysregulated adult neurogenesis

Abstract

People diagnosed with neuropsychiatric disorders such as depression, anxiety, addiction or schizophrenia often have dysregulated memory, mood, pattern separation and/or reward processing. These symptoms are indicative of a disrupted function of the dentate gyrus (DG) subregion of the brain, and they improve with treatment and remission. The dysfunction of the DG is accompanied by structural maladaptations, including dysregulation of adult-generated neurons. An increasing number of studies using modern inducible approaches to manipulate new neurons show that the behavioral symptoms in animal models of neuropsychiatric disorders can be produced or exacerbated by the inhibition of DG neurogenesis. Thus, here we posit that the connection between neuropsychiatric disorders and dysregulated DG neurogenesis is beyond correlation or epiphenomenon, and that the regulation of adult-generated DG neurogenesis merits continued and focused attention in the ongoing effort to develop novel treatments for neuropsychiatric disorders.

By 2020, neuropsychiatric disorders are predicted to be the second highest cause of global disease burden¹. Current treatment for disorders such as major depressive disorder (MDD), bipolar disorder, schizophrenia, post-traumatic stress disorder (PTSD) and substance-related or addictive disorders includes pharmacological intervention, which provides relief for many people. However, these therapies are ineffective for as many as 30% of individuals, and they are often accompanied by substantial side effects^2–4. Equally concerning is that a high percentage of treated individuals relapse^5–8. These facts call for aggressive expansion of the current neuropsychiatric-disorder treatment toolkit.

Clues to treatment and understanding of neuropsychiatric disorders come from the fact that many symptoms of these disorders are reminiscent of abnormal DG function. Similarly to the hippocampus that surrounds it, the DG has a role in memory and mood regulation, and it is also instrumental in the processing of contextual similarities and differences known as ‘pattern separation,’ as well as the processing of intrinsically positive or rewarding stimuli (Box 1)^9–16. The DG is also highly sensitive to stressful experiences, and baseline DG function can be enhanced or inhibited by stress (Box 1)^17–19. Illustrative of aberrant DG function, many people diagnosed with MDD, bipolar disorder, schizophrenia, PTSD or substance-related or addictive disorders have memory dysfunction and mood fluctuations or abnormalities^20–22. Humans diagnosed with MDD, PTSD or schizophrenia show aberrant pattern separation^22–24, whereas those diagnosed with MDD, bipolar disorder, schizophrenia or substance-related and addictive disorders have abnormal processing of rewarding stimuli²⁵.

Box 1. The circuits and functional output of the hippocampal DG.

The healthy hippocampal DG receives input (black lines) from the limbic system (Entorhinal cortex (Ent) via perforant path (PP), perirhinal cortex (PRH), medial septum and (not shown) contralateral DG and recurrent collaterals from hippocampal CA3) and midbrain and hindbrain modulatory regions (VTA/SN, LC, DR)¹⁶. Many other non-DG connections that might affect the DG indirectly are also indicated (gray lines). The DG and hippocampus are highly sensitive to stress (for example, physiological, environmental) owing to direct inputs from the hypothalamic–pituitary–adrenal axis (HPA, red lines, red shading) and putative, indirect effects from the HPA to Ent (red dotted line)^17–19. In turn, the HPA axis is influenced by the DG in part through hippocampal, subicular and Ent projections to the HPA axis. Working in concert, these and other brain circuits contribute to functional DG output, which are shown in schematic form to represent memory, mood, pattern separation and reward. As discussed in the main text, neuropsychiatric disorders are marked by dysregulated memory, mood, pattern separation and reward processing, symptoms that are correlated with impaired DG function and/or impaired generation or function of new DG neurons which may be affected by stress (Table 1). CA1, cornu ammon 1; CA3, cornu ammon 3; DG, dentate gyrus; DR, dorsal raphe; Ent, entorhinal cortex; HPA, hypothalamic–pituitary–adrenal; LC, locus coeruleus; NAc, nucleus accumbens septi; PFC, prefrontal cortex; PP, perforant path; PRH, perirhinal cortex; SN, substantia nigra; VTA, ventral tegmental area.

Furthermore, in humans and animal models, stressful experiences can enhance the susceptibility to and severity of these disorders and their symptoms^18,22,26,27, which also suggests a role for the DG in these disorders (Table 1). Finally, successful treatment of these disorders in humans and the application of similar therapies in animal models normalize or enhance DG function (Fig. 1). For example, brain-stimulation therapy is commonly applied to brain cortical regions to ‘jump start’ or recalibrate brain circuitry in MDD²⁸. However, stimulation of the entorhinal cortex (Ent), a brain region functionally upstream from the DG (Box 1), improves memory in both humans and mice^29,30.

Table 1.

Overview of studies correlating neurogenesis in the DG and neuropsychiatric disorders. Given that more than 2,000 publications have contributed to the correlative findings summarized here, review publications are primarily cited in Table 1 to direct the reader to comprehensive and referenced tables in the literature

Correlated with neuropsychiatric disorders	Stage of neurogenesis (human postmortem)			Hippocampal		Normalize/improve with treatment? (human/rodents)
	Proliferation	Differentiation	Survival	Volume (human)	Network activity (human/rodents)
Major depressive disorder	Decreased/nc133–135	–	Decreased*104	Decreased136	Decreased59,61	✓/nc42,97,104,133,134
Bipolar disorder	nc135	Increased125	Decreased (trend)125	Decreased/nc137,138	Decreased/Increased/nc139	✓99,140,141
Schizophrenia	Decreased132,135	Decreased142	–	Decreased143	Increased/Decreased92,93,144	✓101
Post-traumatic stress disorder	–	–	–	Decreased145	Decreased94	✓100
Substance-related and addictive disorder	Decreased146	Decreased#146	–	Decreased/nc147,148	Increasedˆ95	✓41

Asterisk (*), fewer DG granule neurons in MDD versus MDD with medication and controls;

^#decreased dendritic arborization of immature neurons in heroin addicts;

carat (ˆ), increased connectivity between hippocampus and reward-related brain regions. nc, not changed;

dash (–), not studied.

Figure 1.

Neurogenesis in the DG and its sensitivity to neurotransmitter systems with relevance to common therapies. (a) In the adult mouse DG (a′ hippocampus in gray, dorsal DG in tan, main schematic in a is expanded view of red bar in the inset), new neurons are generated over time, depicted as green cells maturing through developmental ‘stages’ of proliferation, differentiation and survival, a process that takes ~4 weeks. Mature DG granule neurons (right, green cells) receive diverse input from other DG cells (Mossy Cells, blue; interneurons, red and pink) and other limbic (lateral and medial entorhinal cortex (LEnt, MEnt) in dark blue; medial septum in purple) and midbrain and hindbrain regions (ventral tegmental area/substantia nigra (VTA/SN), dorsal raphe (DR), locus coeruleus (LC) in orange). The main somatic input to new DG granule neurons is from inhibitory interneurons (red flathead lines; for example, DG basket cells). Late-stage differentiation and surviving cells are darker green, whereas proliferating cells are lighter green. (b) As new DG neurons develop, they are regulated by an increasingly diverse set of neurotransmitters. The first input is nonsynaptic (ambient) GABA (red bar) from DG interneurons (red input, a), which eventually transitions to include synaptic GABA as well (white gradient in red bar). The last being synaptic glutamate (bottom dark blue bar, b) from Ent neurons (dark blue arrows (a)). Treatments for neuropsychiatric disorders act on many of the neurotransmitter systems that regulate new neurons (b). Ach, acetylcholine; DA, dopamine; DG, dentate gyrus; DR, dorsal raphe; Ent, entorhinal cortex (LEnt, lateral Ent; MEnt, medial Ent); GABA, gamma-aminobutyric acid; Glu, glutamate; LC, locus coeruleus; NE, norepinephrine/noradrenergic; SN, substantia nigra; VTA, ventral tegmental area.

One aspect of DG physiology that has offered hope with regard to the normalization of DG function in neuropsychiatric disorders is its ability to give rise to new neurons throughout life (Fig. 1). These new neurons are found in the DG of mammals ranging from rodents to humans³¹, and work in mice shows that new neurons are able to wire and fire correctly within the DG (Fig. 1a)^27,32–35.

New DG neurons are dynamically regulated by environmental and physiological factors^32,36–40. Notably, the number of new neurons is increased by neuropsychiatric medications such as antidepressants and altered by drugs of abuse^41,42, and this correlative link between new DG neurons and neuropsychiatric disorders is a well-established concept (Table 1)^{18,26,43–50}.

However, in the past decade, new technologies have emerged that enable inducible manipulation of new DG neuron number, morphology or activity in mice and rats. As highlighted in Table 2 (full details in Supplementary Table 1), studies that employ these modern approaches indicate a role for new DG neurons in memory, mood, pattern separation and reward^29,51–68. In addition, studies using animal models of depression coupled with these new technologies show that behavioral symptoms can be produced or ameliorated by select manipulation of new DG neurons. On the basis of this recent work in mice and rats, we posit that the connection between neuropsychiatric disorders and dysregulated DG neurogenesis is beyond correlation or epiphenomenon, and that the regulation of new DG neurons merits continued and focused attention in the ongoing effort to develop novel treatments for neuropsychiatric disorders.

Table 2.

Overview of studies linking DG neurogenesis and features of psychiatric disorders in a causative manner

Approach to manipulate new neurons or DG
Disrupt or inhibit	Enhance or stimulate	Animal model	Behavioral assessment of DG function	Reference	Influence of manipulation on DG function
MEMORY
Disrupt	–	–	Acquisition	51	Impaired
Disrupt	–	–	Retrieval	66	Impaired
–	Enhance	–	Acquisition/retrieval	118	nc
Disrupt	–	–	Acquisition/retrieval+	29	Impaired
Inhibit	–	–	Acquisition/retrieval	52	Impaired
Disrupt	–	–	Acquisition	53	Impaired
Inhibit	Stimulate	–	Acquisition/retrieval+	54	Impaired (inhib or stim)
Disrupt	Enhance	–	Retrieval*	67	Enhanced or impaired
Inhibit, Disrupt	–	–	Retrieval	55	Impaired
Disrupt	–	–	Acquisition/retrievalˆ	65	Impaired
Inhibit	Stimulate	–	Acquisition/retrieval	56	Impaired (inhib or stim)
–	Enhance	–	Acquisition/retrieval+	57	Enhanced
MOOD
Disrupt	–	Stress (chronic)	Anxiety, stress-induced despair	58	Increased
Disrupt	–	–	Despair	59	Increased
Disrupt	–	Stress (chronic)	Stress-induced despair	60	Increased
Disrupt	–	–	Anxiety, stress-induced despair	66	nc
–	Enhance	–	Anxiety, stress-induced despair	118	Decreased or nc
Disrupt	–	Stress (acute)	Anxiety, despair	62	Increased
Disrupt	–	Stress (chronic)	Anxiety	61	Increased
Disrupt	–	–	Anxiety, despair	53	nc
Inhibit	Stimulate	–	Anxiety	54	Decreased (stim)
Disrupt	–	Stress (chronic)	Stress-induced anxiety, despair#	63	Increased
–	Enhance	Stress hormone	Anxiety, stress-induced despair	68	Decreased
Disrupt	–	–	Anxiety§	65	Increased
Disrupt	–	Stress (chronic)	Stress-induced anxiety	108	Increased
Inhibit	Stimulate	–	Anxiety	56	nc
–	Enhance	–	Anxiety, stress-induced despair	57	nc
PATTERN SEPARATION
–	Enhance	–	Pattern separation	118	Enhanced
Disrupt	–	–	Pattern separation	53	Impaired
Inhibit	–	–	Pattern separation	56	Impaired
–	Stimulate	–	Pattern separation	57	Enhanced
REWARD
Disrupt	–	Cocaine self-admin	Rat model of addiction	64	Increased
Disrupt	–	Stress (acute)	Natural reward	62	Decreased

References cited target and manipulate new DG neurons or DG activity and assess a DG function (memory, mood, pattern separation, reward) relevant to neuropsychiatric disorders. Influence on DG function for memory and pattern-separation outcomes are given as enhanced, impaired or nc. Influence on DG function for mood- and reward-related are given as increased, decreased or nc. For study inclusion criteria, detailed list of results and specific figure panels that led to our conclusions in Table 2, see Supplementary Table 1.

Asterisk (*), altered neurogenesis post-learning;

carat (ˆ), dependent on type of conditioning (trace or delay);

cross (+), cognitive flexibility;

pound sign (#), only after adrenalectomy;

dollar sign (§), only after, not before, memory testing;

nc, not changed; self-admin, self-administration; —, not examined.

Adult DG neurogenesis

Adult DG neurogenesis in laboratory animals and humans shares common features^69–71. For example, new DG neurons occur throughout the hippocampal DG, from the dorsal (or septal) to the ventral (or temporal) DG (mouse DG shown in Fig. 1a), albeit with distinct functional contributions along the DG longitudinal axis^72–75. Also, as with the process of embryonic neurogenesis, adult DG neurogenesis in humans and mice and rats is a process, not a timepoint, with maturing cells proceeding through “stages” (Fig. 1a)^27,71. In general, the proliferation, differentiation and eventual survival to maturity of new DG neurons takes 2–4 weeks in mice and rats (Fig. 1a), but weeks longer in primates and humans, as revealed by postmortem studies^31,76. New DG neurons in both rodents (mice, rats) and humans do not replace existing embryonic-generated DG neurons, but rather contribute to the ongoing turnover (addition and subtraction) of newborn neurons in a restricted part of the DG (Fig. 1a)^9,71,77–80.

In contrast to these shared traits, most of what is known about the specific development and afferent input of new DG neurons is understood in mice and rats. Once mature, new DG neurons become glutamatergic granule neurons that receive glutamatergic input from the upstream Ent and inhibitory GABAergic feedback from DG interneurons⁸¹ and project to the downstream hippocampal CA3 region (Box 1)¹⁶. In this sense, fully mature adult-generated DG neurons are indistinguishable from embyronic-generated DG granule neurons. However, as they mature, new DG neurons receive specialized inputs and outputs and dynamically express key intrinsic ion channels and neurotransmitter receptors (Fig. 1a,b)^32,82,83. For example, younger new DG granule neurons receive direct innervation from mature DG granule neurons and downstream hippocampal cornus ammonis region 3 (CA3) pyramidal neurons⁸⁴, and these inputs disappear as the new neuron matures. In addition, younger new DG granule neurons neither receive nor send feedback inhibition⁸¹, whereas older new DG neurons do. The location of inputs to new DG neurons is crucial for determining their influence on new neuron function^33,35,83. For example, most glutamatergic inputs are on the dendritic tree of new neurons, whereas inhibitory inputs are on or near the soma (Fig. 1a)^33,35. These inputs form gradually over the >4 weeks that it takes for new neurons to mature (Fig. 1b), with nonsynaptic, ambient GABA as the first input and synaptic glutamate from Ent as the last^33,35,83. Taken together with the fact that new neurons have a lower threshold for firing relative to embryonic-generated neurons and that they are sensitive to activity-dependent regulation^36,85–87, the emerging view is that immature adult-generated DG neurons are distinct from mature adult-generated and all embryonic-generated DG neurons in that they are “young and excitable”⁸⁸, function independently of inhibitory GABA circuits⁸⁹ and receive direct innervation from their glutamatergic “elders”⁸⁴. In fact, some work suggests that adult-generated neurons remain functionally distinct from their embryonic-generated counterparts⁹⁰.

Neuropsychiatric disorders and adult DG neurogenesis

It has long been known that the development of new neurons is regulated by antidepressants and drugs of abuse^41,42. In fact, treatments for neuropsychiatric disorders (including antidepressants, electroconvulsive treatment, antipsychotics and mood stabilizers, all of which can improve DG functional output in humans diagnosed with or animal models of neuropsychiatric disorders) act on many of the very neurotransmitter systems that regulate new neurons (Fig. 1b)^26,91. Such studies raise the idea that neuropsychiatric disorders are linked with fewer new DG neurons, and that treatment or remission is linked to more or a normalized number of new DG neurons.

Although this correlative relationship between new DG neurons and neuropsychiatric disorders relies on only a few human postmortem publications, there is much evidence in animal models that these disorders are linked to impaired generation or function of new DG neurons^26,47. New DG neurons contribute to hippocampal volume and appropriate DG function and activity, and in humans, mice and rats, these disorders are also marked by impaired hippocampal volume and/or DG network activity (Table 1). For example, the relative activity between the DG and the downstream hippocampal CA1 region (Box 1) is decreased in stressed rats as compared to control rats⁵⁹, which is notable, given that stress is a major precipitating factor in the exacerbation of neuropsychiatric symptoms. In animal models of bipolar disorder and schizophrenia, the stress- or novelty-induced activity of the DG is diminished^92,93. Additionally, humans with PTSD have decreased hippocampus and amygdala activity relative to undiagnosed, trauma-exposed controls⁹⁴. Finally, long-term cocaine users have decreased hippocampal gray matter, and abstinent heroin-dependent users have enhanced connectivity among hippocampus and reward-related brain regions relative to individuals who do not use drugs^95,96.

Additional support for a correlation between the presence of fewer new DG neurons and the presence or exacerbation of neuropsychiatric symptoms comes from studies in both humans and rodents in which a decreased number of new neurons or lower DG volume is normalized or improved after treatment or during remission (Table 1). For example, postmortem studies show that chronic antidepressant treatment restores proliferation deficits in humans with MDD⁹⁷. Chronic treatment with the mood stabilizer lithium increases new-neuron proliferation and differentiation in animal models, and the bipolar disorder medication valproate rescues aberrant DG activity and manic-like behavior in animal models^98,99. Furthermore, anti-psychotic drugs, such as haloperidol, risperidone or clozapine, restore neurogenesis in animal models of schizophrenia, and antipsychotics can also ameliorate anxiety and restore new-neuron proliferation in an animal model of PTSD^100,101.

Neuropsychiatric symptoms and neurogenesis—a causative relationship?

Recently, techniques have emerged that enable more focused exploration of the correlative–causative relationship of neurogenesis and neuropsychiatric disorders (Table 2; see also Supplementary Table 1). These include studies in mice and rats that ablate new neurons at a specific ‘stage’ of their development, manipulate the number, structure or activity of new DG neurons, or couple ‘classic’ new-neuron deletion techniques, such as irradiation, with behavioral investigations that reflect modern understanding of new neuron function^29,51–65. For example, by using retroviral-mediated gene transfer to express light-sensitive channels in proliferating precursors, researchers are able to control new neuron firing remotely⁵². Recently developed transgenic mouse lines also enable the targeting of new neurons⁵⁶ or of mature DG granule neurons⁵⁴ with these light-sensitive channels, and they enable inducible gene expression or deletion selectively in new neurons^{53,56,62,63,65} or mature DG granule neurons^54,55,57. The temporal control that researchers have over the firing of the targeted cell population in these optogenetic studies is ideal for behavioral investigations. As discussed below, most data are available for memory and mood, although studies have also confirmed a role for new DG neurons in pattern separation and reward^29,51–65.

Memory

Neuropsychiatric disorders are accompanied by memory dysfunction^20–22. Given that recent studies support the idea that manipulation of new neurons can alter memory, it is possible that the activation of new neurons might be useful for normalizing memory dysfunction seen in neuropsychiatric disorders. For example, studies in mice have found that the ablation or inducible silencing of new or mature DG neurons impairs aspects of memory function^9,26,102, with some finding impairments in memory acquisition or the learning of a new memory^52,56, and others finding impaired retrieval, or expression of the memory after it has been formed^52,55. These studies, all of which manipulate adult or embryonic-generated neurons during or before learning, suggest that enhancing neurogenesis might improve memory.

Notably, understanding the timing of neurogenesis disruption is crucial to understanding the role of new neurons in memory. For example, disruption of neurogenesis after learning has taken place impairs memory retrieval, which suggests that new neurons are needed as well for ‘forgetting’^66,67. Given that certain neuropsychiatric disorders (PTSD, substance-related and other addictive disorders) can be considered as being marked by aberrantly strong memories, it is feasible that postlearning enhancement of neurogenesis might help to drive the forgetting of such memories.

Also relevant to neuropsychiatric disorders is recent work showing that the ‘age’ or stage of new DG neurons determines whether the cells are involved in functional DG output^52,56. For example, the stimulation of 4-week-old new DG neurons—but not of 2-week-old immature neurons or 8-week-old fully mature neurons—is required for memory retrieval⁵². This timing fits with that of new-neuron integration into DG-CA3 circuitry (Fig. 1a)³², and with the concept that there is a ‘critical period’ during which new neurons are needed for successful memory processing¹⁰³.

Mood disorders

With regard to mood disorders, MDD, anxiety and PTSD are strongly linked to abnormal hippocampal structure and function, and impaired mood and greater anxiety are hallmarks of these human disorders^18,26. Humans diagnosed with MDD often have an exacerbated or prolonged stress response and lack the typical hippocampal inhibition of the stress axis relative to individuals without depression, which again indicates that this mood disorder is linked to a dysfunctional hippocampus¹⁸. Notably, many correlative studies support the idea that rodents (mice and rats) exposed to chronic stress and humans diagnosed with MDD have decreased or dysfunctional neurogenesis, and that antidepressant treatment drives neurogenesis^47,104. Techniques used to ablate new neurons led to a depressive-like phenotype in mice and rats in some studies, but not others^47,105. These mixed messages as to the role that new neurons have in depression raises concerns that neurogenesis is an epiphenomenon, and thus not a reasonable target for treatment¹⁰⁶.

However, a relatively consistent conclusion from the recent studies in rodents is that new DG neurons are required for antidepressant efficacy^58–61(Table 2; Supplementary Table 1). One clue to a potential underlying mechanism is that new neurons buffer the body’s stress axis responsiveness and associated behaviors^62,107. For example, a loss of new neurons results in a hyperactive stress response in mice⁶². In fact, it is now hypothesized that having a decreased number of DG neurons contributes to the development of depressive-like behaviors particularly under stressful conditions, such as the psychosocial stress experienced during social defeat stress⁶³. Indeed, increased DG neurogenesis in mice is sufficient to attenuate anxiety and depressive-like behavior under stress-like conditions in which the rodents are given the stress hormone corticosterone⁶⁸. In further support of the role of new DG neurons in buffering the stress response, the positive effect on stress-induced anxiety of a neurogenic-compound-induced increase in neurogenesis is blunted by the ablation of neurogenesis¹⁰⁸.

Current studies have manipulated new DG neurons or DG activity in the absence of substantial stress to address whether mood regulation is mediated by them. Notably, direct stimulation or inhibition of new DG neurons does not change anxiety behavior under basal, nonstressed conditions⁵⁶. However, direct stimulation of mature DG granule neurons in either the dorsal or ventral DG has an innate anti-anxiety effect⁵⁴. This fits with prior work showing that the dorsal DG is more linked to spatial or contextual functions, whereas the ventral DG is more linked to mood and related emotional functions^72,74,75, and suggests that subregional targeting of DG activity is a potential intervention for mood disorders. This functional neuroanatomical gradient is also evident in studies in which neurogenesis is manipulated. For example, whereas reversible silencing or stimulation of new neurons in the dorsal—but not ventral—DG impairs memory acquisition and retrieval, direct stimulation of ventral DG granule cells—but not of new neurons—decreases anxiety in mice, as compared to controls^54,56. Thus, acute stimulation of ventral DG granule neurons could be a very useful intervention for relieving anxiety.

Pattern separation

Pattern separation is a computational term often applied to the DG to explain how cortical inputs representing spatial and contextual information converge onto the DG yet diverge into distinct outputs in the downstream CA3, a hippocampal region involved with the complementary function of pattern completion¹⁰⁹. However, pattern separation is also a very relevant concept in neuropsychiatric disorders because overgeneralization (impaired pattern separation) is a barrier to symptom relief and is seen in many individuals with neuropsychiatric disorders²². For example, humans diagnosed with PTSD have diminished pattern separation; they generalize cues associated with a traumatic memory (crime victimization or war, for instance) to nonthreatening contexts, which thus drives fear and anxiety in these nonthreatening contexts²². Notably, this deficit is apparent in tests of pattern separation unrelated to the particular trauma: humans diagnosed with PTSD or related anxiety disorders perform poorly on neuropsychological tests of pattern separation²². By contrast, some humans diagnosed with autism spectrum disorder have enhanced pattern separation and note even slight differences in contexts or routines¹¹⁰. In PTSD, the severity of symptoms is correlated with a smaller and less active DG, as compared to healthy controls. These connections raise the possibility that the normalization of pattern separation will be accompanied by normalization of DG activity and diminished symptom severity.

Indeed, targeting new neurons in rodents can mimic these findings in humans: inhibition or ablation of new neurons impairs pattern separation, whereas stimulation of mature DG neurons or inducible enhancement of new neurons enhances pattern separation^{53,56,57,68,111,112} (Table 2 and Supplementary Table 1). In fact, when computational models of DG pattern separation account for the addition of new DG neurons, there is reduced overlap of activated GCs by previous memories, and thus improved pattern separation as compared to controls²⁷. As with learning and memory, the ‘stage’ of the new neurons matters in regard to pattern separation. For example, pattern separation is reliant on young new neurons (just integrated into DG circuitry) but not on older mature new neurons or embyronic-generated DG neurons¹¹³. These aspects will be useful for guiding and fine-tuning potential future new-neuron-based therapies for pattern separation.

Reward

With regard to reward, the ablation of new DG neurons in rats increases cocaine drug-taking and drug-seeking, as compared to controls, which suggests that decreased adult neurogenesis is a vulnerability factor in a rat model of cocaine addiction⁶⁴ (Table 2 and see also Supplementary Table 1). Given work showing that environmental-enrichment- or running-induced increase in adult neurogenesis correlates with a reduction of addictive-like behaviors in rats and mice^114,115, it would be interesting to selectively and inducibly increase the number of new DG neurons or stimulate existing new DG neurons to test whether this would diminish drug-taking or drug-seeking in animal models of addiction. However, the only other study that has inducibly manipulated new neurons and examined reward-based hedonic behavior found the opposite result: the deletion of new DG neurons decreases a mouse’s preference for a rewarding substance, in this case, sucrose⁶². Given that it is challenging to compare an animal model of cocaine addiction with oral sucrose intake, more studies are warranted to see how the manipulation of new neurons influences reward processing.

Harnessing the DG to treat neuropsychiatric disorders

Given the above studies, it is intriguing to consider targeting new neurons to recalibrate dysfunctional DG activity to treat neuropsychiatric disorders. For example, in contrast to the healthy DG wherein Ent projections to the DG and its new neurons drive memory processes, mood regulation, pattern separation and processing of rewarding stimuli (Fig. 2a)^9–16, this functional output is disrupted in neuropsychiatric disorders, which perhaps leads to or exacerbates symptoms (Fig. 2b). Regardless of whether the dysfunctional DG is due to changes in Ent input, the number of new DG neurons, regulation of DG activity, or some other parallel process, we propose two putative approaches for targeting new DG neurons as a way to normalize DG activity and functional output (Fig. 2c).

Figure 2.

Proven and proposed approaches to targeting new DG neurons to recalibrate DG functional output in neuropsychiatric disorders. (a) In the healthy DG, input from the Ent into the DG drives new neurons and ultimately results in functional DG output, such as memory, mood, pattern separation and reward (green shaded triangle). When DG activity or function is aberrant, as is the case in many neuropsychiatric disorders (b, gray shaded triangle), new neurons are often decreased in number or function, perhaps in part owing to a dysfunctional upstream input from the Ent (indicated by a question mark (?)). (c) We propose that inducible increase in new neuron number, activity or DG activity can improve neuropsychiatric symptoms (blue box, arrow, green shaded triangle), whereas inhibition of new neuron number, activity or DG activity can block such improvement (gray box, flat head line, gray shaded triangle). Thus, targeting new neuron number, activity or DG activity might be a novel treatment for neuropsychiatric disorders. Recalibration of DG functional output is feasible via the stimulation of upstream DG regions (e.g., Ent, d) or increased number of new neurons or increased activity of new neurons or DG (e). For (d), it has already been shown in mice that Ent stimulation (blue box, bold font) enhances neurogenesis and improves learning and memory in both humans and mice (green shaded triangle, d). It remains to be tested whether this upstream stimulation and enhanced new neuron number or activity will affect mood, pattern separation or reward (indicated by question mark (?)). For (e), it has been shown in mice that direct stimulation or induction of adult neurogenesis (blue box) can improve memory, mood and pattern separation, although most of the work in this regard has shown the opposite (e.g., direct inhibition or suppression of adult neurogenesis impairs memory, mood and pattern separation). Although it has not been tested whether direct stimulation or induction of adult neurogenesis can improve reward, the converse has been shown: the suppression of adult neurogenesis enhances vulnerability in animal models of aberrant reward (e.g., addiction). DG, dentate gyrus; Ent, entorhinal cortex.

The first approach—increasing Ent input to DG and thus increasing the activity of DG neurons—might be useful for disorders that have common traits of impaired learning and memory induced by hippocampal or DG dysfunction and Ent afferent dysfunction, such as MDD, PTSD and schizophrenia. In such cases, recalibration of DG functional output is feasible via the stimulation of upstream DG regions, such as deep-brain stimulation of Ent (Fig. 2d). Indeed, it has already been shown that Ent stimulation enhances neurogenesis in mice and improves memory in both humans and mice^29,116. More recently, Ent stimulation is also able to activate a DG memory in a mouse model of Alzheimer’s disease¹¹⁷, although the role of new neurons is not assessed in that study. It remains to be tested, however, whether this upstream stimulation and the subsequent putative enhanced new neuron number or activity will have beneficial effects on mood, pattern separation or reward (Fig. 2d). However, given that lesioning of Ent impairs pattern separation in mice⁸⁴, it is reasonable to hypothesize that Ent stimulation would improve pattern separation, and in turn, improve PTSD symptoms. Notably, whereas Ent-stimulation-induced memory improvement in mice acts via increased neurogenesis²⁹, it remains unclear whether Ent-stimulation-induced improvement in DG function is always reliant on intact neurogenesis.

The second approach—increasing new DG neuron number or activity—builds off the unique physiology and positioning of new DG neurons and the many neuropsychiatric disorders that are linked with dysfunctional DG activity (Fig. 2e). Many studies support the idea that the ablation or silencing of new neurons impairs aspects of memory, mood, pattern separation and the response to rewarding stimuli (Table 2 and Supplementary Table 1). By contrast, inducible increase in the number of new DG neurons formed, or inducible modulation of their spines, improves pattern separation and cognitive flexibility, respectively, in mice^57,118, and stimulation or induction of new or mature DG neurons decreases anxiety in mice^54,68. These studies suggest that a therapy based on the induction of new neurons has the potential to boost pattern separation and normalize learning and memory deficits (Fig. 2e). However, the stimulation of new neurons or mature DG neurons can also inhibit DG function^54,56. Although the discrepancies between studies that stimulate new neurons might be due to technical issues, this urges caution when developing and employing such new-neuron-based therapies.

Because new neurons are highly sensitive to activity-dependent regulation, a related approach would be to stimulate DG activity to subsequently stimulate new neurons. Intriguingly, ‘neurogenic’ compounds exist that drive neurogenesis and/or improve some aspect of DG functional output in an activity-dependent manner^108,119. However, so far, no neurogenic compounds act directly and solely on new DG neurons. A major challenge is identifying specific receptors or proteins on new DG neurons that could be exploited for treatment purposes. This is a particularly problematic obstacle in that new DG neurons move through so many developmental stages (Fig. 1a,b). However, the fact that new neurons develop much more slowly in primates than in rodents⁷⁶, yet a larger proportion of hippocampal neurons turn over in humans than in rodents^31,120, raises the possibility that identifying a new-neuron target might be more feasible in humans.

One final way to target new neurons would be to regulate their microcircuitry to normalize dysfunctional DG activity. For example, the activation of certain DG interneurons in rodents drives sparse encoding by feedback inhibition onto mature GCs⁸¹, which might improve pattern separation, as has been studied indirectly¹²¹. In addition, the stimulation of other DG interneurons in mice promotes progenitor survival and thus adult neurogenesis¹²², which might also enhance sparse encoding and pattern separation. Indeed, the regulation of any local mechanism that drives sparse encoding and the accompanying improvement in temporal precision of activation might be a candidate for this approach¹²³. Such studies are not currently feasible in humans because of the limited knowledge of new-neuron microcircuitry in humans and the invasive nature of such approaches.

Conclusions and future directions

A dysfunctional DG is considered to be an endophenotype or biological correlate of most neuropsychiatric disorders^{22,24,59,61,124,125}. The normalization or recalibration of aberrant DG function holds potential for treatment of these disorders. New DG neurons have unique connections and physiological properties that enable them to both sense and influence DG activity. Thus, we propose that targeting new DG neurons is one way of normalizing or recalibrating a dysfunctional DG (Fig. 2e). Another way of recalibrating a dysfunctional DG is upstream stimulation of the Ent (Fig. 2d). We think that both of these approaches merit consideration for therapeutic normalization of DG functional output, and we predict that—in animal models at least—this normalization occurs in part via the stimulation of new DG neurons.

There are numerous knowledge gaps that remain. First, specific DG output circuits in mice and rats need to be thoroughly defined, as has been done in regard to the ventral hippocampus–nucleus accumbens circuits for mood regulation¹²⁶, because incorrect neuroanatomical or functional targeting of a circuit might impair an already disrupted network¹²⁷. This is feasible, given the next generation of functional neuroanatomical techniques, which employ genetic promoter enhancers with cell-specific genetic- and viral-mediated, gene-transfer-induced labeling and lineage-specific transgenic lines^128,129. Second, a broader scope of behaviors should be tested in mice and rats, particularly as new functions of the DG are recognized¹³⁰. It is notable how few studies cross the ‘memory–mood’ divide, or even consider other DG functions, such as reward. Here we have highlighted papers that examine both memory and mood^53,54,65, mood and reward⁶², or memory, mood and pattern separation^56,57. Given that neuropsychiatric disorders are marked by a dysfunctional DG, and that the DG has a range of functional output, we encourage researchers to more frequently and pointedly cross the memory–mood divide and to explore more work with animal models of neuropsychiatric disorders. Third, additional human and postmortem work is warranted, particularly with regard to development, regulation, and most ambitiously, in vivo imaging^31,131. Finally, although neurogenesis does exist in the postnatal primate and specifically human DG^31,131, the process is presumed to take much longer than in the mouse or rat, and it is more challenging to detect cells in select stages of neurogenesis, given the limitations of current markers and the modes of assessment of new neurons in the adult human. This is relevant because postmortem human neurogenesis studies most commonly measure proliferating cells, not immature or specifically new neurons, because of technical obstacles. Should any future therapies emerge that, for example, normalize the proliferating cells that are reduced in humans diagnosed with either schizophrenia or MDD^97,132, a post-treatment waiting period would be warranted to assess whether indeed the therapy has any influence on DG function. Therefore, although the common model of neurogenesis and its timeline (Fig. 1a) is a simplified representation of our current knowledge, it should continue to be rigorously tested and refined, particularly in regard to human neurogenesis.

We hope that this current view of new DG neurons and neuropsychiatric disorders will drive interest in testing the postulates put forth here. These recent studies underscore that the regulation of adult-generated DG neurogenesis merits continued and focused attention in the ongoing effort to develop novel treatments and to enable expansion of the toolkit available to treat neuropsychiatric disorders.