An assessment of the existence of adult neurogenesis in humans and value of its rodent models for neuropsychiatric diseases

Abstract

In sub-mammalian vertebrates like fishes, amphibians, and reptiles, new neurons are produced during the entire lifespan. This capacity diminishes considerably in birds and even more in mammals where it persists only in the olfactory system and hippocampal dentate gyrus. Adult neurogenesis declines even more drastically in nonhuman primates and recent evidence shows that this is basically extinct in humans. Why should such seemingly useful capacity diminish during primate evolution? It has been proposed that this occurs because of the need to retain acquired complex knowledge in stable populations of neurons and their synaptic connections during many decades of human life. In this review, we will assess critically the claim of significant adult neurogenesis in humans and show how current evidence strongly indicates that humans lack this trait. In addition, we will discuss the allegation of many rodent studies that adult neurogenesis is involved in psychiatric diseases and that it is a potential mechanism for human neuron replacement and regeneration. We argue that these reports, which usually neglect significant structural and functional species-specific differences, mislead the general population into believing that there might be a cure for a variety of neuropsychiatric diseases as well as stroke and brain trauma by genesis of new neurons and their incorporation into existing synaptic circuitry.

INTRODUCTION

These days there is a popular and fateful trust that the continuous addition of new neurons is essential for our thinking and the formation and preservation of our memories. It is also believed that diminished neurogenesis in adult human is involved in the pathogenesis of a variety of neuropsychiatric diseases that range from autism and schizophrenia to depression and Alzheimer’s and that therefore new neurons are necessary for their prevention and useful in their treatment. This belief implies that extant stem cells in the adult human brain can generate new neurons that migrate to assume their appropriate positions and then mature and integrate into already established healthy, or deteriorated neuronal networks, where they would perform normal physiological functions. This creates a belief in the possibility that the addition of new neurons can enhance brain function in normal aging and that it can repair diseased or injured brains in conditions such as trauma and stroke as well as major neuropsychiatric disorders, including memory loss, depression, and dementia. Given those powers, it is not surprising that the idea of getting new neurons is very attractive and ignites remarkable public interest.

The field of adult neurogenesis took off in the mid-90s, when several reports confirmed the presence of newly formed neurons in the dentate gyrus of the hippocampus and in the olfactory bulb of rodents [1, 2]. These initial studies were soon replicated in the macaque [3–6]. This excitement quickly led to overreach and soon the daily birth of thousands of new neurons that would supposedly populate the neocortex of the adult macaque brain was reported [7]. This claim, labeled as the discovery of the century, ignited the imagination of scientists and general public alike. The possibility of regenerating the neocortex was born. Unfortunately, despite receiving thousands of citations, this report could not be repeated even by the same authors and therefore it has been discredited [8–12].

In 1998, Eriksson and co-workers [13] published a highly influential paper reporting that postmortem brain tissue of cancer patients that were injected with bromodeoxyuridine (BrdU) during treatment contained cells co-labeled with both BrdU and neuronal markers (see below for critical technical analysis). Since then, there has been a number of human studies suggesting the presence of adult neurogenesis in humans, although most of them did not provide conclusive evidence, as they only showed the presence of dividing cells, without determining whether they differentiate into neurons, glia, or other cell types [e.g., [14]. Another influential study based on analysis of the incorporation of ¹⁴C in neurons described the birth of 700 new neurons daily in the human hippocampus. There are also a number of studies on the postmortem human brain that claim variable rates of neurogenesis and the presence of dividing progenitors and cells labeled with doublecortin (DCX), a protein typically expressed in newly born and migrating immature neurons [13, 15–21].

In the present invited Expert Review, because of the relevance to the readers of Molecular Psychiatry, we will focus first on technical problems and then on contradictions and logical discrepancies that this phenomenon would pose in the adult human brain and caution that the valuable results of many animal models do not translate directly to human (see Fig. 1).

Fig. 1. Schematic representation of the general relation between the amount of adult neurogenesis and cognitive abilities in different vertebrates derived from a common ancestor (⋆).

It is paradoxical that adult neurogenesis, a capacity claimed to be very important for very many human qualities and pathologies, in the course of evolution, has decreased to negligible levels in the human while still remaining substantial in vertebrates, such as fishes and amphibians.

DISREGARD OF SPECIES-SPECIFIC DIFFERENCES

Alexander Pope’s aphorism that “the proper study of mankind is man” is very true but not very practical when the best approaches require invasive experimental interventions in vivo. Humans benefit enormously from animal experimentation and without it a large portion of medical and pharmaceutical advances would not be possible. So, animal experimentation is both necessary and useful for the advancement of science. However, it is imperative to take into consideration that there are limitations inherent to the different species, which may dictate different biological behaviors under the same conditions. One extreme example of this is to cut the tail of a salamander and observe that it will completely regenerate, including nerves [e.g., [22]]. This, of course, is not the case of a mouse tail or a human spinal cord.

Most studies on adult neurogenesis are carried out in rodents, mostly mice, in which numerous chemicals and nutrients have been tested and shown to affect it significantly in both directions. This renders the output unpredictable, which also indicates that this trait is not a tightly regulated process. For example, blueberries were found to increase neurogenesis, while a high-fat diet decreases it in males but has no effect on females [23, 24]. A soft diet decreases it in the hippocampus but not in the olfactory bulb, a finding explained based on the rate of mastication [25]. Another study showed that mice having caffeine at the equivalent of up to 30 expresso coffees a day have decreased neurogenesis, but the effect can be reversed by doubling the dose [26]! Commonly, activities such as learning, running, exposure to an enriched environment, or caloric restriction are reported to increase adult neurogenesis while alcohol and/or stress to decrease it. Regarding pharmacology, it has been reported that hormones, morphogens, growth factors, cytokines, glucocorticoids, antidepressants, opioids, seizures, excitatory neurotransmission and many other all affect it [23, 27, 28]. It can be argued that some of the described factors have a significant but small effect while others, like running or an enriched environment, have a strong effect. This might be true, but there are also caveats. While voluntary running in laboratory mice increases proliferation up to fourfold, it does not occur equally in all mouse lines. Running B6 mice will produce about 1000 new neurons/km, while D2 animals will produce only about 400 new neurons/km [29]. Importantly, when wild rodents caught outside were tested for running in an impoverished environment (a cage), they did not show changes in neurogenesis [30, 31]. In addition, many studies that implicitly support adult human neurogenesis are actually done in relatively young animals, most of them strains of laboratory mice that have been inbred for hundreds of generations, producing results that do not even apply to the wild long-tailed wood mouse [32], also see [33] for a thorough review]. These findings strongly question the value of inbred strains of mice as models for adult neurogenesis in the wild mice, not to mention for humans. Therefore, it is perplexing that so many researchers, as well as people in the public, believe that those results apply to human.

In addition, it was reported that epilepsy in mice increases the number of new neurons, and if this applies to humans, then epileptic patients should have larger hippocampi and have better memory and problem-solving skills than normal controls. In contrast, it was reported that a decreased rate of neurogenesis in adult mouse does not cause depression, and therefore neurogenesis itself is not a primary factor leading to depression [34]. Finally, severe reduction of adult neurogenesis by knock-out of specific genes like cyclin D2 has little behavioral implications regarding learning and memory, the action of antidepressants, or epilepsy [35]. Adult neurogenesis was initially linked to antidepressants because both processes take a period of 6–8 weeks to produce an outcome. However, more recently introduced drugs, such as ketamine, have their effects within a few hours of administration [36] and therefore their effect cannot be related to neurogenesis. Paradoxically, among the most frequent claims of the value of adult neurogenesis is that it is essential for forming new and recalling old memories or exactly the opposite, to regulate forgetting [e.g., [37]. Thus, after >10,000 papers published on adult neurogenesis in rodents the data of its time course, rate, or functional characteristics are contradictory [38].

In relation with the issues described above, it should be noted that some of the treatments and processes that regulate adult neurogenesis in rodents also are expected to have an effect on their physiology or cognition. For example, running is well known to increase the number of capillaries, improve cardiovascular function, and release trophic factors and endorphins. Those changes are naturally expected to produce an improvement in motor and cognitive output both in rodents and in humans, irrespective of neurogenesis. Therefore, research might be better aimed to study those shared features of hippocampal biology.

PROBLEMS WITH METHODS OF DETECTION OF NEW NEURONS

In this section, we caution that most methods and markers used for detecting the generation and introduction of new neurons into the synaptic circuitry of the adult human brain are not specific and/or adequate without additional evidence.

The first, and the most cited claim of adult neurogenesis in humans, using BrdU, is a paper by Eriksson et al. [13] in which 5 patients with cancer were given 250 mg of intravenous BrdU to assess the proliferative activity of the tumor cells. One single patient with a similar cancer, but not injected with BrdU, was used as control together with rats and mice injected with BrdU. Five to seven sections through the hippocampus were examined postmortem from each of the injected patients, who died 16–781 days post BrdU injection. In all of them, BrdU-labeled cells were detected in the subgranular zone (SGZ), the granule cell layer, and hilus of the dentate gyrus, while labeling was not detected in the patient not injected. BrdU labeling was interpreted as evidence of the presence of newly generated cells without taking into consideration critical problems with the methodology. First, the sample, and the number of brain sections per sample, was very small and all the humans under study were terminally ill, e.g., the study did not include a single healthy human control. Second, the study ignored the intrinsic limitations of BrdU, that, although it labels dividing cells, it is not a specific marker of cell division, but an indicator of DNA synthesis, a process that may occur independent of mitosis such as when damaged cells try to repair DNA or are undergoing apoptotic processes [e.g., [39–42], reviewed by Duque and Rakic [42]]. Postmitotic (possibly dying) neurons can re-enter the cell cycle and therefore could get labeled with BrdU [43–46]. In fact, as reported by Eriksson et al., the number of BrdU⁺ cells declined with increasing interval between BrdU injection and time of death, which would indicate that the BrdU⁺ cells were dying; a process more congruent with being ill than being newly born.

Another more recent, commonly cited paper for neurogenesis in adult humans used ¹⁴C as a method of detection [16]. This method seems appropriate when the results are negative since no new cells can be created without incorporating carbon. However, positive results do not necessarily imply neurogenesis, since incorporation of ¹⁴C into the DNA, as discussed earlier for BrdU, does not occur only at the time of cell division but also in postmitotic neurons re-entering the cell cycle for any reason or undergoing DNA repair or methylation, processes that occur at high rates in the hippocampus [47, 48] and seem to have been greatly underestimated in the Spalding report [33]. DNA replication in degenerating and dying cells has been well documented [e.g., [43, 45, 49]]. Also, it was shown that hypoxia–ischemia induces DNA synthesis without cell proliferation in dying neurons in adult rodent brain [46] and there is an increase in neurogenesis markers after hypoxic–ischemic encephalopathy in humans [50]. Hence, older and dying neurons that re-enter the cell cycle can incorporate markers used to label new neurons. Other technical inadequacies of the methodology are thoroughly discussed by Duque and Spector [33], including contamination artifacts and large corrections and assumptions. Finally, as pointed out by Sorrells et al. [51], the results differ from histology studies [e.g., [15, 52, 53]] and rely on isolation of nuclei using NeuN, which could represent subpopulations of oligos and microglia [54, 55]. Finally, this technology has produced similar results in other brain areas [e.g., caudate nucleus [56]], which have been challenged [57] and could not be corroborated by any other study.

Another set of evidence supporting adult neurogenesis comes from studies using immunohistochemistry against specific markers of progenitors or immature neurons. For obvious reasons, those studies on human are done on postmortem or surgical biopsy brain tissue that has its own limitations, mostly due to a postmortem delay and a suboptimal fixation.

However, accepting the limitations, performing proper controls, and exercising caution on the interpretation of results can go a long way. For example, the problems with recent studies by Boldrini et al. [17], Moreno-Jimenez et al. [18], Tobin et al. [19], and Flor-Garcia et al. [20], which claim adult neurogenesis in human, are commented at length in the dual perspective article by Sorrells et al. [51]. While technically sophisticated, these studies have not considered a host of confounding factors that compromise the validity of the results and are conducive to the incorrect interpretation of such results. On the other hand, the work of Knoth et al. [15], Doorn et al. [58], Dennis et al. [52], Mathews et al. [59], Sorrells et al. [53], Cipriani et al. [60], and Seki et al. [61] also covered at length in the dual perspective articles reported the absence of neurogenesis in humans using the same methodologies, which is in our opinion the correct conclusion.

Among the studies on human tissue, by far the most comprehensive analysis was performed by Sorrells et al. [53]. This study concluded that the phenomenon fell to undetectable levels in human adults. The study was performed on n = 59 well-preserved postmortem and postoperative human tissue samples from fetal to adult ages and included ultrastructural analysis of the dentate gyrus, revealing that contrary to other species, humans lack the SGZ, the specific niche where hippocampal neurogenesis takes place. Simultaneously, Cipriani et al. [60] provided solid evidence for the absence of neurogenesis in adult and aged human hippocampus, and both studies included infant and child controls where the authors could detect in fact the presence of markers of neurogenesis that decreased precipitously after adolescence [14]. However, concurrent and subsequent studies [17–19] claimed the existence of granule cell neurogenesis based mostly on the presence of DCX, and in the case of Boldrini et al. [17], they also used nestin to identify progenitors although those nestin-expressing cells looked exactly like regular astrocytes.

Expression of markers such as DCX, Calbindin, Calretinin, NSE, polysialylated neuronal cell adhesion molecule (PSA-NCAM), SOX2, and proliferative cell nuclear antigen (PCNA), associated with neurogenesis, are not unique to newly born neurons or other cell types. For instance, Calbindin and Calretinin are markers of subpopulations of GABAergic interneurons that are abundant in the cerebral cortex, and Calbindin is expressed in mature granule cells [62–64]. DCX is expressed not only in some adult neurons but also in oligodendrocytes and microglia [e.g., [54, 55, 65], and DCX⁺ cells may represent a pool of immature cells, not newly generated cells [38], that could undergo protracted maturation in adulthood. De-maturation of adult hippocampal neurons is also possible and has been demonstrated with, for instance, fluoxetine treatment [66]. Some reports specifically warn of the dangers of ignoring confounding factors and show no significant relationship between the number of DCX⁺ cells and working memory [67]. It is also shown quantitatively that few granule cells are added to the adult hippocampus [68] and that PSA-NCAM is widely expressed in many mature inhibitory interneurons [69]. SOX2, which is heavily used to label neural progenitors, is also expressed in some mature astrocytes. Finally, PCNA is expressed after DNA repair [70] and in adult postmitotic ependymal cells [71], to mention only some examples of the limitations of the available markers.

Given the technical difficulties, it is not surprising that the results of studies on human tissue are very mixed and confusing, even when the tissue is obtained from patients with the same illness. For instance, in epilepsy, some report an increase [72–75] and some report a decrease [76, 77] of neurogenesis markers. Likewise, in depression there are reports of reduction (i.e., MCM2) [78] or an increase in markers (i.e., DCX) [79]. In normal aging, the results are also mixed, contradictory, and highly controversial [16–18, 52, 53, 59, 80, 81]. In the recent dual perspective articles, Moreno-Jimenez et al. [21] defend the existence of large amounts of human adult neurogenesis, while Sorrells et al. [51] using more comprehensive analysis argue, as we do, that few, if any, new neurons are born and incorporated in the adult human hippocampus.

A smaller hippocampal volume has been established as a biomarker for major depressive disorder [82–84] and therefore increase of hippocampal volume in response to treatment is emerging as a useful neuroimaging technique to assess outcomes [e.g., [85]]. However, the changes in hippocampal volume are likely related to vascularization of folded areas more vulnerable to anoxia due to induced vasoconstriction [86] than to the emergence of new neurons. The relationship between markers whose expression is usually interpreted as evidence of new neurons in the hippocampus and compromised vascularization conducive to hypoxia, with and without an inflammatory response, should be further investigated in humans. Increased blood flow specific to the dentate gyrus and not the whole hippocampus has been shown to occur in the mouse [87] and about 37% of all dividing cells are immunoreactive for endothelial markers [88], which subsequently disappear.

Most recently, an alternative approach based on transcriptome analysis has become available to study neurogenesis in the dentate gyrus. More specifically, messenger RNA is extracted, amplified, and sequenced to obtain detailed information about gene expression at a single-cell level. Analysis of the results allows cells to be classified in subtypes based on the combined expression of markers. Transcriptome analysis in developmental and young adult mouse has proven very efficient to reveal the presence of all the components of the neurogenic process, such as neuronal progenitors, newborn neurons (neuroblasts), and immature neurons [89]. Therefore, the power of this technique stands on its capacity to reveal the whole picture of a process, instead of identification of cells based on a single marker like the case of DCX for immature neurons or Nestin or Sox2 for progenitors. Based on this potential, transcriptomics has been proposed by key figures in the field as a method to finally settle the controversy regarding adult neurogenesis in humans [90–93].

A pioneer study on the adult human brain by Habib et al. [94] rendered a cluster of cells in the hippocampus that was classified as neural stem cells (NSCs) based on the expression of some putative NSC markers. Posterior re-analysis of these data by the Alvarez-Buylla laboratory showed that this cluster was specifically and highly significantly enriched in markers of motile cilia and ependymal cells, suggesting that they were ependymal cells and not NSCs [51]. Recently, a comprehensive transcriptomic study on the dentate gyrus of adult human [95] showed lack of a neurogenic trajectory in the hippocampus. This study could not find a defined population of neuronal progenitors or newly formed neurons expressing the typical markers of neuroblasts, including DCX. These transcriptomic studies on the adult human hippocampus have shown that DCX is mostly expressed in inhibitory interneurons and only a few cells with a mature granule cell phenotype could be found expressing DCX mRNA [51, 94, 95]. Thus, the results reveal that there are not neural progenitors or newly born neurons in the adult human dentate gyrus and that DCX might be present in mature granule cells and interneurons.

In summary, a thorough review of the literature shows that there is no scientific convincing evidence of the generation and incorporation of new neurons into the circuitry of the adult human brain, including the dentate gyrus of the hippocampus.

PROBLEMS AND CONTRADICTIONS OF THE BASIC CONCEPT

Apart from the lack of scientifically proven evidence that new neurons are generated and incorporated into circuits of the adult human brain, even if a few were added, there are many logical problems with the interpretation of its possible significance. These include suggestions for their healing capacity to treat multiple and diverse neuropsychiatric disorders. As listed below, some of the problems with those overoptimistic suggestions can be refuted without any laboratory research:

1. If there is continuous neurogenesis over many decades of human life, why is it that structures involved do not continue to grow? For example, if neurogenesis is abundant and continuous in the adult human hippocampus, why do older people not have bigger hippocampi than younger ones? If the size is maintained by a corresponding constant and equally large amount of neuronal cell death, why has neuronal death not been demonstrated to occur in the same place, at the same time and same rate, as new neurons are generated?

2. Independent of whether new neurons die soon after being produced, it seems logical that an abundant number of new cells in the adult hippocampus, or anywhere adult neurogenesis is supposed to occur, would require an abundant number of progenitors caught during mitotic cell division. Why are there are no abundant mitotic or cycling cells visible? The presence of ph3, Ki67, or MCM2+ cells is reported in some studies, but these cells are few and whether they are neurons or glia is usually not established.

3. Why would it be evolutionarily advantageous for any structure in the adult human brain, with already connected and functional neurons, to make room for new neurons to develop and mature dendrites and axons and form new synaptic connections? In addition, would these processes not be extremely disrupting to the structure’s ongoing functional operations and for the retaining of acquired knowledge? And, if the cells have this capacity why are damaged structures not repaired automatically after trauma or disease?

4. If adult neural progenitors divide actively, as often has been reported, at some point tumors of neurons should exist. It appears that adult neuroglial stem cells can give rise only to glia-based tumors (gliomas).

5. If adult neurogenesis is essential for human memory and cognition, which are evolutionarily new functions, why do humans not have higher rates of neurogenesis than lower vertebrates, such as fishes and frogs? And why does adult neurogenesis, even in rodents, occur only in the phylogenetically older olfactory bulb and hippocampus (archicortex) and not in the neocortex?

CONCLUSION

We have written this critical review by invitation from Molecular Psychiatry, whose many readers are physicians, to caution them not to be misled by the panacea of data on adult neurogenesis in rodents to believe that all results are directly relevant and applicable to understand the pathogenesis of disease or to treat human neuropsychiatric disorders. There are many major methodological and conceptual problems with the field of adult human neurogenesis. Although it is true that good-quality postmortem adult human brain tissue continues to be difficult to obtain and that the regulations for obtaining it and using it in research are significant burdens for those interested, obtaining and preserving good samples is not impossible. Studying adult neurogenesis in alternative species affords the advancement of knowledge but authors should be careful not to over-interpret their findings by neglecting technical problems and species-specific differences. As pointed out in this review, there is not definitive evidence that there are new neurons that are incorporated into the circuitry of the adult human brain. So why do so many believe in it? As Victor Hamburger, co-discoverer of the nerve growth factor, said at an informal meeting: “A single report of an incorrect finding that many people like, takes more than hundreds of papers with negative findings to make an acceptable correction”. He was talking about a different subject, but it certainly applies to the popular claims on the significance of neurogenesis in the adult human brain. The promise of curing a variety of neuropsychiatric disorders with different pathogenesis via fostering of adult neurogenesis has deviated substantial private and public resources from proven preventive approaches to preserve the good, already connected, working old neurons that endow us with some of our most human characteristics and faculties.