Neurogenesis is restricted to specific areas

Compared with high-rate turnover tissues such as the intestine for example, the brain presents limited regenerative capacities, in terms of intensity as well as in terms of location. The stem cells in the brain comply with the requirements defining stem cells: they self-renew, and are multipotent, i.e. they can differentiate into functional neurons, astrocytes and oligodendrocytes. The NSC are localised in three areas: the subventricular zone (SVZ) lining the lateral ventricle, the subgranular zone of the dentate gyrus (DG) in the hippocampus and a recently acknowledged area of the hypothalamus, the median eminence (ME). In the parenchyma, some scattered dividing cells have also been described, but these dividing cells are progenitors and not stem cells, in the sense that they do not self-renew, and that their in vivo fate is limited to the glial lineage. They will not be considered here.

The NSC of the SVZ give rise to neuroblasts migrating to the olfactory bulb, where they differentiate into GABAergic olfactory bulb interneurons and integrate the existing neuronal networks⁽ Reference Lazarini and Lledo ^{1 )}. They take part in the maintenance of the network, the incorporation of new olfactory stimuli, or the olfactory memory. The fate of the NSC has been precisely documented in the SVZ: NSC (type B cells) derived from the embryonic radial glial cells divide slowly and give rise to transient-amplifying cells (type C cells), precursors of the neuroblasts (type A cells) which migrate into the rostral migratory stream to reach the olfactory bulb⁽ Reference Lazarini and Lledo ^{1 –} Reference Doetsch ^{3 )}. NSC have also the possibility to differentiate into astrocytes and oligodendrocytes⁽ Reference Jessberger, Toni and Clemenson ^{4 ,} Reference Menn, Garcia-Verdugo and Yaschine ^{5 )}.

In the DG of the hippocampus, the NSC are designated by different names, but the successive events are very similar: the NSC are named radial and horizontal type 1 NSC. They divide slowly into type 2 neuronal precursors, which cycle more intensely. The precursors differentiate into neuroblasts and are finally integrated into the network of the granular zone after 4 weeks of maturation as glutamatergic granule cells. The difference in this structure resides in the fact that the neuroblasts migrate shorter distances, and differentiate in the layer of subgranular cells. They establish connections with the pyramidal layer of cornu Ammonis (CA) 3 zone of the hippocampus⁽ Reference van Praag, Schinder and Christie ^{6 ,} Reference Toni, Laplagne and Zhao ^{7 )}. Correlative studies by ablation and deletion have established that the new neurons are involved in the maintenance of spatial memory, learning capacities, and the retention of new memories⁽ Reference Zhao, Deng and Gage ^{8 )}. There seems to be a correlation between the survival of newborn cells and spatial-memory performances⁽ Reference Borcel, Perez-Alvarez and Herrero ^{9 ,} Reference Trouche, Bontempi and Roullet ^{10 )}. DG neurogenesis has also been linked with anxiety and depression or depression susceptibility in humans. This aspect has been under intense focus after the groundbreaking studies showing that the anti-depressant fluoxetine could increase cell division in the DG, and that this effect could correspond to the time needed for the treatment to have an impact in the patients⁽ Reference Encinas, Vaahtokari and Enikolopov ^{11 ,} Reference Lee, Kim and Yim ^{12 )}.

If the existence of neurogenesis is now firmly established in rodents’ brains, the relevance of these observations in humans has been debated. In humans, SVZ neurogenesis is very active until 2 years of age, with large numbers of immature neuroblasts migrating in the rostral migratory stream, but is nearly extinct in adulthood⁽ Reference Sanai, Nguyen and Ihrie ^{13 )}. A similar observation has been made in non-human primates⁽ Reference Wang, Liu and Liu ^{14 )}. By contrast, human DG neurogenesis remains very active in adults: it has been estimated that 700 neurons/d are added in the DG⁽ Reference Spalding, Bergmann and Alkass ^{15 )}.

Another very important role of neurogenesis in both areas is that the generation of new neurons is involved in neuronal repair after ischaemic stroke. Neuroblasts from the SVZ migrate to the damaged areas such as the cortex and the striatum and serve as a replacement of the dead neurons⁽ Reference Jin, Sun and Xie ^{16 ,} Reference Liu, Solway and Messing ^{17 )} and ischaemic stroke enhances also neurogenesis in the DG⁽ Reference Tan, Preston and Wojtowicz ^{18 )}. NSC transplants can survive and migrate into the ischaemic area⁽ Reference Kelly, Bliss and Shah ^{19 )} and neuroblast transplantation helps the restoration of behaviour and motor skills in rodents⁽ Reference Chintawar, Hourez and Ravella ^{20 –} Reference Nori, Okada and Yasuda ^{22 )}.

There is a great similarity in the markers of the NSC and progenitors from both areas: nestin and Sox2 are the classical markers for the NSC, and are not expressed anymore along the line of differentiation, polysialylated neuronal cell adhesion molecule (PSA-NCAM) and doublecortin are the most commonly used markers for neuroblasts, which once fully differentiated into neurons, can be labelled with NeuN or HuC/D. Using these markers enables to trace the fate of the newly generated cells in vivo.

The hypothalamus has been less studied. There have been reports showing the existence of dividing cells for long⁽ Reference Xu, Tamamaki and Noda ^{23 )}; but this structure has recently gained further interest because studies have pointed out that the neurogenesis in this area is modulated by high-fat diets, a point that will be discussed below. Tanycytes are glial cells that line the third ventricle. Four different types have been characterised, α 1 and 2, and β 1 and 2, the function and properties of which differ with their location along the ventricle⁽ Reference Rodriguez, Blazquez and Pastor ^{24 )}. The β2 tanycytes, at the bottom of the third ventricle in the ME region and the α2 subtype, bordering the lateral walls of the third ventricle in the paraventricular and arcuate nuclei, are now considered as NSC, because they proliferate and display markers such as nestin and Sox2⁽ Reference Lee and Blackshaw ^{25 )}. They migrate under intermediate differentiation into the arcuate nucleus (ARC) of the hypothalamus, and differentiate fully into pro-opiomelanocortin (POMC) neurons. POMC neurons are anorexigenic neurons; they decrease food intake through αMSH release, a proteolytic product of POMC. A fraction of these new neurons differentiate in a second wave into NPY and Agouti-related peptide (AgRP) neurons, which are orexigenic, and increase food intake⁽ Reference McNay, Briancon and Kokoeva ^{26 ,} Reference Padilla, Carmody and Zeltser ^{27 )}. The origin of the feeding circuitry lies in the embryonic period, and the remodelling is very intense postnatally, and up until adolescence in the rodent⁽ Reference Tan, Preston and Wojtowicz ^{18 )} but persists into adulthood⁽ Reference Li, Tang and Cai ^{28 )}. Disruption of the neurogenic process leads to modifications of food intake: an acute degeneration of AgRP neurons leads to a severe anorexia in mice, while a slow and progressive deletion does not have a significant effect on food intake, because neurogenesis can compensate and replace the degenerated neurons⁽ Reference Pierce and Xu ^{29 )}. Therefore, the loss of POMC neurons in ARC could favour the development of obesity.

The neurogeneses in the three areas present similarities on several aspects: (1) the stem cells display markers specific to the glial lineage⁽ Reference Barry, Pakan and McDermott ^{30 –} Reference Pinto and Gotz ^{32 )}; (2) the succession of events involved, i.e. proliferation followed by migration and differentiation; and (3) the nature of the markers. Yet, because each area is also a particular niche, their regulations can present some differences.

Regulation and effectors

As mentioned above, NSC are not scattered in the parenchyma, but are concentrated in particular areas named ‘niches’ which provides the environment necessary to protect and preserve the cell pool. The niches are localised in proximity to ventricles (SVZ and ME) and/or capillaries (DG). Therefore, they are in close contact with effectors and the NSC pool can react to signalling modifications, either by dividing or staying quiescent. The signals can be endogenous; they include hormones, growth factors or neurotrophins. They can also be external, and that is how diet and nutrition can influence the process.

Neurotrophins and growth factors are positive regulators of neurogenesis: brain-derived neurotrophic factor (BDNF), insulin-like growth factor or vascular endothelial growth factor all increase cell proliferation, and/or promote cell survival or cell differentiation in the DG⁽ Reference Bergami, Rimondini and Santi ^{33 ,} Reference Sairanen, Lucas and Ernfors ^{34 )}. A broad range of hormones can also act on neurogenesis: oestrogen, progesterone and dehydroepiandrosterone promote DG neurogenesis⁽ Reference Li and Shen ^{35 ,} Reference Sicard, Ehrhart-Bornstein and Corbeil ^{36 )}, while corticosteroids, the last messengers of the hypothalamic–pituitary–adrenal axis, diminish cell proliferation and differentiation⁽ Reference Denis, Potier and Vancassel ^{37 )}. Peptide hormones can also modify neurogenesis: leptin promotes cell proliferation in the DG⁽ Reference Garza, Guo and Zhang ^{38 )} and insulin promotes neuronal differentiation⁽ Reference Han, Wang and Xiao ^{39 )}. Gut peptides can modify cell renewal: ghrelin, peptide YY and cholecystokinin can increase cell proliferation in the DG and ghrelin acts positively on all steps of neurogenesis in the SVZ⁽ Reference Li, Kim and Kim ^{40 ,} Reference Stanic, Paratcha and Ledda ^{41 )}. Chronic injections of glucagon-like peptide-1 increased neurogenesis in a model of Alzheimer’s disease⁽ Reference Parthsarathy and Holscher ^{42 )}; the effect of satiety peptides have been described on DG neurogenesis so far, and their effect on the proliferation and differentiation of β2 tanycytes is not known. It would be interesting to examine this point. Indeed, although the mechanisms involved in the three areas unroll very similarly, they are not necessarily modified by the same factors: significant differences in the length and importance of the successive steps may be responsible for the differences in sensitivity towards effectors. For instance, prolactin enhances SVZ neurogenesis and thereby increases the olfactory potential of the mother in preparation for maternal behaviour but DG neurogenesis is not modified by prolactin⁽ Reference Larsen and Grattan ^{43 )}. Conversely, BDNF increases cell proliferation in the DG, but does not affect cell proliferation in the SVZ⁽ Reference Galvao, Garcia-Verdugo and Alvarez-Buylla ^{44 )}.

A common feature of neurogenesis in the three areas is the vulnerability to the ageing process. Following embryonic neurogenesis, proliferation is very active in the postnatal period and decreases significantly until mid-age, then keeps decreasing at a slower pace during senescence in the hippocampus⁽ Reference Bondolfi, Ermini and Long ^{45 ,} Reference Kuhn, Dickinson-Anson and Gage ^{46 )}. Recently, a study by Zhang et al. ⁽ Reference Zhang, Li and Purkayastha ^{47 )} showed that inflammation in the hypothalamus increases with age, as measured by NF-κB activation in the microglia; this ageing process negatively regulated gonadotropin-releasing hormone (GnRH) synthesis, and GnRH delivery into the third ventricle could reverse age-induced neurogenesis decline.

Dietary influences

Environmental factors are efficient actors in the regulation of neurogenesis and diet or dietary intakes take an important part in these interactions. Several aspects can be considered: the quantity of ingested food or its composition, as developed below, and even its texture: rat hippocampus proliferation was decreased in animals fed soft food, and after being fed a powdered diet for 1 year, mice had lower counts of pyramidal neurons in the CA1 and CA3 regions of the hippocampus than mice fed a solid diet⁽ Reference Aoki, Kimoto and Hori ^{48 ,} Reference Nose-Ishibashi, Watahiki and Yamada ^{49 )}.

Mechanisms involved

From the examples, a general outline can be drawn, showing that molecules or events that increase inflammation tend to lower neurogenesis, while anti-inflammatory molecules or pathways restore it. Therefore, the mediators of the inflammatory pathway represent the central targets and the balance upon which the different effectors can impinge. We will develop here some examples of these targets, such as the cAMP response element binding protein (CREB), BDNF, NF-κB, sirtuin 1 (Sirt1), and the energy-sensing kinase AMP-activated protein kinase (AMPK), and we will show how they all inter-relate.

Brain-derived neurotrophic factor and cAMP response element binding protein

BDNF is a well-recognised effector of neurogenesis. This neurotrophin is involved in brain development and neural plasticity. It mediates its effect through binding to the membrane receptor tropomyosin receptor kinase B (TrkB). The variations of BDNF in the hippocampus induced by environmental modifications and dietary interventions are directly correlated with the variations noted in neurogenesis: dietary restriction, which increases neurogenesis, and increases BDNF mRNA expression and protein⁽ Reference Andrade, Mesquita and Assuncao ^{83 )}. On the other hand, overnutrition decreases BDNF levels. It is here interesting to underline that BDNF, apart from its role on neurogenesis, acts also on food intake, since knock-out mice deficient in bdnf display hyperphagia and develop obesity⁽ Reference Gray, Yeo and Cox ^{84 )}. Very similarly, the effects of flavonoids on neurogenesis have been connected with parallel modifications of BDNF in whole brain or hippocampus: ingestion of naringin, oroxylin A or blueberry extracts leads to a significant increase in BDNF levels⁽ Reference Jeon, Rhee and Seo ^{85 –} Reference Williams, El Mohsen and Vauzour ^{87 )}. Resveratrol promotes BDNF synthesis from astroglia⁽ Reference Zhang, Lu and Wu ^{88 )} and in the hippocampus⁽ Reference Rahvar, Nikseresht and Shafiee ^{89 )}, and by promoting BDNF synthesis reverses hippocampal atrophy in chronic fatigue mice and reverses the effect of mild stress on cognition⁽ Reference Liu, Xie and Yang ^{90 ,} Reference Wang, Gu and Wang ^{91 )}.

The effects of n-3 PUFA supplementation are similar: a large body of studies has unanimously shown that, in experimental models, the dietary supplementation of DHA or a mixture of n-3 PUFA could increase BDNF levels, or normalise them after mild brain injury in rats⁽ Reference Wu, Ying and Gomez-Pinilla ^{92 )}. DHA supplementation at 1·25 % could increase the effects of voluntary exercise in mice, with a parallel increase in BDNF in the brain. Quite interestingly, the benefits of n-3 PUFA were also transmitted by the maternal diet since prenatal n-3 PUFA supplementation to a micronutrient-imbalanced diet protected brain neurotrophins in both the cortex and hippocampus in the adult rat offspring⁽ Reference Sable, Kale and Joshi ^{93 )}.

In most cases the similarity of actions on BDNF synthesis could be related to the effects on synthesis of CREB and its phosphorylation. CREB is an ubiquitous transcription factor, highly expressed in the brain in which it is involved in many signalling pathways serving many functions, ranging from neuronal survival to memory formation⁽ Reference Finkbeiner ^{94 )}. Its activation through phosphorylation is the result of the action of the mitogen-activated protein kinases, or protein kinase C kinase. Once phosphorylated, CREB enters the nucleus where it binds to the CREB response element. One of its target gene is Bdnf. Experimental data showed that n-3 PUFA supplementation after traumatic brain injury enhanced CREB protein contents⁽ Reference Wu, Ying and Gomez-Pinilla ^{95 )} and reduced oxidative damage in injured rats. Curcumin enhanced CREB expression in rat brain⁽ Reference Kumar, Antony and Gireesh ^{96 ,} Reference Xu, Ku and Tie ^{97 )}, and resveratrol increased the phosphorylated state of CREB after ischaemia⁽ Reference Lin, Chen and Zhang ^{98 )}.

Inflammatory NF-κB v. anti-inflammatory Sirt1 and AMP kinase

Another factor affecting neurogenesis is the balance between pro- and anti-inflammatory pathways.

Ageing, depression and obesity have all been related to low-grade neuroinflammation and its oxidative stress companion⁽ Reference Tang, Purkayastha and Cai ^{99 ,} Reference Chaudhari, Talwar and Parimisetty ^{100 )}. NF-κB is the major and central mediator of inflammation. This transcription factor, once activated, can induce the transcription of the inflammatory cytokines such as TNFα, monocyte chemoattractant protein-1 and IL-1β or IL-6⁽ Reference Baker, Hayden and Ghosh ^{101 )}. NF-κB activation has also been cited in the different situations leading to an increase or decrease of neurogenesis: all the signals, either endogenous such as age and stress, or external such as overnutrition, lead to an activation of NF-κB in the brain, either in the hippocampus or the hypothalamus, and decrease neurogenesis in these areas.

An endogenous inhibitor of NF-κB is Sirt1. Sirt1 is a deacetylase enzyme regulating energy metabolism and cell survival. Among other targets, the RelA/p65 component of the NF-κB complex is deacetylated by Sirt1 on Lys130, which inhibits the transcription ability of NF-κB⁽ Reference Yeung, Hoberg and Ramsey ^{102 )}. By this mechanism, Sirt1 can prevent the deleterious effects of a high-fat diet and protect from hepatic steatosis⁽ Reference Pfluger, Herranz and Velasco-Miguel ^{103 )}. Therefore, Sirt1 activators could also be valuable in preventing the inflammatory pathway. Dietary restriction is a strong inducer of Sirt1 expression and activity. The first compound detected among a screening of small molecules to be an inhibitor of Sirt1 was resveratrol⁽ Reference Howitz, Bitterman and Cohen ^{104 )}, and it was later demonstrated that resveratrol could mimic the effects of energy restriction and delay the ageing process through its up-regulation of Sirt1⁽ Reference Wood, Rogina and Lavu ^{105 )}. It is now well admitted that nutrients can modulate inflammation through Sirt1 involvement. Other flavonoids, such as quercetin or rutin for instance, can also up-regulate Sirt1 expression⁽ Reference Davis, Murphy and Carmichael ^{106 ,} Reference Su, Yu and Chen ^{107 )}. Because of its broad enzymic activity, Sirt1 can take part in many pathways. It has been demonstrated that Sirt1 activity could modify the differentiation fate of NSC, favouring astrogliosis to the expense of neurogenesis under oxidative stress⁽ Reference Prozorovski, Schulze-Topphoff and Glumm ^{108 )}. Sirt1 is also essential for cognition and neuronal plasticity⁽ Reference Michán, Li and Chou ^{109 )}. It is interesting to mention that the overexpression of Sirt1 stimulates BDNF expression⁽ Reference Zocchi and Sassone-Corsi ^{110 )}, and is neuroprotective in a model of Huntington’s disease⁽ Reference Jiang, Wang and Fu ^{111 )}. By its deacylating activity, Sirt1 is involved in a large range of effects. Among the numerous Sirt1 targets, we will focus on AMPK, because its involvement in NSC physiology and its regulation by dietary factors attract increasing interest.

AMPK is a fuel-sensing kinase that drives the cellular metabolism from catabolic to anabolic depending on the energy level of the cell. Its activity is up-regulated when the ATP levels of the cell are low. AMPK and Sirt1 share many cellular targets, are activated by the same stimuli, and also regulate each other⁽ Reference Ruderman, Xu and Nelson ^{112 )}. Experimental evidence from transgenic mice lacking the AMPKβ1 subunit demonstrates that this kinase is involved in NSC proliferation, since the mice display cerebral atrophy. AMPK could also improve neuronal survival under low energy conditions⁽ Reference Dasgupta and Milbrandt ^{113 )}. Several studies have shown that bioflavonoids can induce, in vitro and in vivo, an activation of the AMPK pathway. Naringenin, quercetin and quercetin 3-O-glycosides extracted from plants in muscle cells, or in the brain of old mice, up-regulate AMPK activity. Apigenin induces the activation in human keratinocytes and its activity is also up-regulated by resveratrol in mouse primary neurons⁽ Reference Eid, Martineau and Saleem ^{114 –} Reference Vingtdeux, Giliberto and Zhao ^{118 )}. Similarly, n-3 PUFA exert their protective and anti-inflammatory effects through the up-regulation of AMPK⁽ Reference Wu, Ying and Gomez-Pinilla ^{95 )}. The implication of Sirt1 has not been extensively investigated in these studies, but has been demonstrated in others⁽ Reference Jing, Yao and Liu ^{119 ,} Reference Xue, Yang and Wang ^{120 )}.

For clarity we have described the two pathways under two distinct paragraphs, but they actually are also intertwined and it is interesting here to notice that the mediators susceptible to act on neurogenesis are also found at the intersection of the fields of inflammation, energy homeostasis and neuronal plasticity, as recapitulated in Fig. 1.

Fig. 1

Adult neurogenesis is a flexible process, which can be modified by positive or negative influences. Pro-inflammatory molecules or mechanisms activate NF-κB, which induces the transcription of pro-inflammatory cytokines, while anti-inflammatory molecules by their action on sirtuin 1 (Sirt1) activate cAMP response element binding protein (CREB), brain-derived neurotrophic factor (BDNF) and AMP-activated protein kinase (AMPK) and restore cell renewal in the brain. The influence of gut microbiota is here hypothetical. For a colour figure, see the online version.

Rodent experimentation has amply demonstrated the disadvantages/benefits of diets/molecules on neurogenesis and their functional consequences such as the effects on learning, mood or energy homeostasis. The relevance of these studies to the human situation is hampered by the difficulty in measuring human neurogenesis so this point is mainly based on correlations: for instance, a recent study tested the supplementation of resveratrol in elderly and proved it efficient in protecting memory and glucose metabolism⁽ Reference Witte, Kerti and Margulies ^{121 )}. Similarly, components of the Mediterranean diet improved cognition in elderly⁽ Reference Huhn, Kharabian Masouleh and Stumvoll ^{122 )}. So data generated from animal experimentation have opened the way to many more studies on the usage of diets or natural compounds as therapeutic agents.

Gut microbiota: a possible effector?

Gut microbiota is now considered as a major factor in the modulation of the host’s physiology, including neural functions through the gut–brain axis. Seminal data obtained from the comparison of germ-free (GF) mice with their conventional counterparts have shown that the absence of intestinal microbiota was associated with a decreased anxiety but a stronger response to stress⁽ Reference Diaz Heijtz, Wang and Anuar ^{123 )}, but GF male F344 rats exhibited increased anxiety compared with SPF counterparts⁽ Reference Crumeyrolle-Arias, Jaglin and Bruneau ^{124 )}. The genetic background of the animals has been proposed as an important factor⁽ Reference Crumeyrolle-Arias, Jaglin and Bruneau ^{124 )}. GF mice display a lower abundance of BDNF in the hippocampus, and an increased monoamine turnover in the striatum⁽ Reference Diaz Heijtz, Wang and Anuar ^{123 )}. Another group has shown higher hippocampal concentrations of 5-hydroxytryptamine in male GF mice⁽ Reference Clarke, Grenham and Scully ^{125 )}.

A robust body of evidence has shown that gut microbiota composition differs between fat and lean animals, with a higher proportion of Firmicutes and a lower presence of Bacteroidetes in obese animals⁽ Reference Tremaroli and Backhed ^{126 )}. The modifications pointed out in animals are consistent with the modifications of the gut microbiota detected in obese patients⁽ Reference Turnbaugh, Hamady and Yatsunenko ^{127 )}. Moreover, microbiota alone seems able to modify the host’s metabolism, as exemplified by the recent clinical observation, resulting from a faecal transfer⁽ Reference Alang and Kelly ^{128 )}. The mechanisms involved are still under clarification, but hypotheses suggest that microbiota composition could either promote a more efficient energy harvest from the nutrients, or act through the lowering of the systemic inflammation level associated with obesity. Indeed, the inflammation develops with different kinetics in the organs, and the brain is the first one to be harmed, since a low level of inflammation is perceived after only 7 d of a high-fat diet⁽ Reference Tan, Preston and Wojtowicz ^{18 )}. Furthermore, the installation of obesity is accompanied with a deregulation of numerous hormones levels, such as insulin, leptin, adiponectin⁽ Reference Song, Kang and Kim ^{129 )} and the gut satiety peptides such as glucagon-like peptide-1. We have seen in the paragraph above that these factors can positively modify neurogenesis, and, therefore, their deregulation could contribute to the development of a downward spiral aggravated by an impairment of neuronal repair. Recent data from obese and non-obese patients show that gut microbiota and brain microstructure are associated⁽ Reference Fernandez-Real, Serino and Blasco ^{130 )}. Whether ‘obese’ or ‘non-obese’ microbiota can modulate neuroinflammation or act directly through bacterial metabolites would be interesting questions to study, particularly in the hypothalamus, but also in the hippocampus, with regard to depression or decline of the cognitive abilities linked with obesity. Moreover, since the characteristics of the diet (for example, animal v. plant proteins) themselves can modify the composition of gut microbiota⁽ Reference Zhu, Lin and Zhao ^{131 )}, dietary recommendations should be made to drive the intestinal microbes’ mass to beneficial and optimal proportions.