Does Adult Neurogenesis Persist in the Human Hippocampus?

Whether new neurons are added to the human brain in childhood or adulthood is of widespread interest. Our recent observations suggest that newborn neurons in the adult human hippocampus (HP) are absent or very rare (Sorrells et al., 2018). A subsequent study proposes that large numbers of new neurons continue to be produced in the adult human HP (Boldrini et al., 2018). This has stimulated discussion and re-appraisal of this topic. Human studies have intractable caveats, making it important to interpret both positive and negative observations, including our own work, with a critical lens. A recent minireview (Kempermann et al., 2018) attempts to offer a perspective on the field and discusses the challenges of working with human tissue. Yet it fails to critically evaluate the evidence for adult human hippocampal neurogenesis and how, for example, halogenated thymidine analogs or carbon 14 (¹⁴C) birthdating are susceptible to false positives or sample processing effects (Extended Discussion in Sorrells etal., 2018).

Although Kempermann et al. (2018) compare our study with Boldrini et al. (2018), the identification of the cells and the type of staining obtained are quite different. We used DCX (doublecortin) co-expression with PSA-NCAM (polysialylated neural cell adhesion molecule), Tuj1, or NeuroD to identify hippocampal young neurons. In children, DCX⁺ cells, co-stained with the above markers, were small (5–10 μm) and elongated with a leading process (Figures S1A–S1C; Extended Data Figures 5B, 5G, and 6C in Sorrells et al., 2018). In adults, we observed DCX staining in small, round cells not only in the granule cell layer, but also in the hilus of the dentate gyrus (DG) and in other brain regions, including white matter (Figure S1D). Many appear to be non-neuronal glial cells. Some ramified DCX-labeled cells with morphology of differentiated neurons were also observed in the HP and other brain regions (Extended Data Figure 7 in Sorrells et al., 2018; seen in Liu et al., 2018). Given the possibility that our staining could be affected by the postmortem interval (PMI), we studied samples from surgical resections and autopsies with different clinical histories and from different sources. We also analyzed two adult cases with a PMI < 5 hr to fixation that were perfused with 4% paraformaldehyde before brain removal. We found a consistent pattern: young neurons and progenitor cells are present in the anterior and posterior human hippocampus in children, but not adults. In the same adult postmortem samples where we did not see DCX⁺PSA-NCAM⁺ cells in the HP, we detected young neurons in the ventricular wall (Extended Data Figure 6B in Sorrells et al., 2018) and entorhinal cortex (Figure S1E), in line with previous reports. The absence of young neurons in the adult DG, therefore, cannot be attributed to poor antigen preservation.

Consistently, other studies show young neurons with a typical elongated morphology before the age of 5 years (Cipriani et al., 2018; Knoth et al., 2010; Dennis et al., 2016). In these reports, the adult samples had very few DCX⁺ cells (1–7 cells/mm²), and the examples shown were either the small, round cells or large, mature-looking neurons (frequently highly ramified). In the companion letter (Tartt et al., 2018), additional examples of these complex cells are shown with neurofilament co-expression, consistent with mature neuronal identity. We used electron microscopy to show that DCX⁺young neurons in DG of children are different from the DCX⁺ cells in adults (Extended Data Figures 6A, 7C, and 7E in Sorrells et al., 2018). We also studied the PSA-NCAM⁺ cells in the adult DG; these cells were DCX^– and had large, round nuclei. Their broad distribution in the hippocampus, complex morphology, and NeuN expression suggested they were mature neurons.

Kempermann et al. highlight the value of stereological measurements in Boldrini et al. (2018). However, stereology is only a useful technique if what is being counted is correctly identified. Boldrini et al. interpret small, round cells or large, ramified DCX⁺ cells as young neurons (Figure 3 in Boldrini et al., 2018). The examples of DCX⁺PSA-NCAM⁺ cells shown in this study lack elongated nuclei and leading processes: key features of young neurons in other species, in the DG in children (Figures S1A–S1C; Dennis et al., 2016; Knoth et al., 2010) or in the adult ventricular-sub-ventricular zone (V-SVZ) (Wang et al., 2011). PSA-NCAM⁺ cells with a prominent process were also considered new neurons, but these cells were more than twice the size (20~30 μm) of typical immature neurons (Figures 2 and 3). Ki-67 is used to infer neural progenitor proliferation, but precursors of oligodendrocytes, microglia, and parenchymal astrocytes continue to divide in adults. Their Ki-67 staining in many of the images is non-nuclear (e.g., co-expressed with nestin in the cytoplasm or processes) or is in the nuclei of large GCN (e.g., Figures 1F, 1I, [ 1K, 1L, and 4C). Moreover, SOX2 or nestin expression is used to identify progenitor cells, but these proteins are frequently expressed by differentiated glial cells i (Komitova and Eriksson, 2004), and neither was quantified with Ki67.

In addition to the postnatal decline in young hippocampal neurons we and others observe, a subgranular zone (SGZ) niche of proliferative stem cells is not evident in the human hippocampus. Instead, Ki-67⁺ cells in the adult human hippocampus are dispersed across the DG (Figure 2 in Sorrells et al., 2018), with no evidence of increased cellularity in a putative human SGZ. Furthermore, the vasculature in the rodent SGZ forms a dense tangential plexus, which is considered important for neurogenesis (Palmer et al., 2000). This plexus is not present in the human DG in Boldrini et al. (Figure 4 in Boldrini et al., 2018) or in our samples. If granule cell neurons (GCNs) were born further away from the GCL in a “dispersed SGZ,” we would expect to find examples of elongated, migrating DCX cells between the hilus and the GCL, but such cells have not been observed in our study or that of others.

Studies supporting the presence of adult human hippocampal neurogenesis are not consistent with each other: some report a sharp decline and small, negligible contribution in adults (Knoth et al., 2010; Dennis et al., 2016). Others support continuous high levels of neurogenesis in old age (Spalding et al., 2013; Boldrini et al., 2018), but show extremely high variability. In Spalding et al. (2013), many of the samples have ¹⁴C levels consistent with no addition of new neurons. Whether this reflects true variability between individuals is unclear. Kempermann et al. suggest a decoupling hypothesis to explain the paucity of DCX expression in the human DG, proposing a “latent” period when putative young neurons do not express DCX. Against this view, our study shows that in children, DCX expression is in young neurons at different stages of maturation, including cells that have already developed typical dendrites and axons of GCNs (Figure S1C; Extended Data Figure 5C in Sorrells et al., 2018). This suggests that neurons continue to express DCX as they differentiate. An important remaining question, therefore, is how long do human GCNs require to fully mature? This process could take several months and possibly years. An extended period of maturation of a subpopulation of neurons could contribute to protracted plasticity.

If there was a large pool of newborn GCNs at different stages of maturation in the adult human DG, they should be readily detectable. Instead, the observations from our adult human samples and those of others (Cipriani et al., 2018; Knoth et al., 2010; Dennis et al., 2016) indicate that these cells are not detected, or are extremely rare. How do we move forward? We all concur that postnatal hippocampal neuronal recruitment is a fascinating phenomenon and an important process to understand as it is revealing new mechanisms of plasticity and possibly repair. We are also in agreement that new neurons continue to be recruited in the hippocampus of children, and perturbations to this process could have life-long consequences. But there is a need to unify criteria for cell identification and further study the identity of round or ramified cells that appear to express DCX in adults. Single-cell transcriptomic data may provide additional markers and patterns of gene expression that define the different stages of differentiation of GCN in the human DG. Each new approach must be viewed objectively and critically. We hope this discussion highlights the importance of well-preserved human samples for research and will stimulate more work to further improve tissue preservation and the exchange of materials and protocols.

Supplementary Material

1

Supplementary Figure S1: Young neurons in the human hippocampus decline to undetectable levels in the adult. A, DCX⁺ cells in the 3 week old human DG. At this age these small, elongated DCX⁺ cells express PSA-NCAM or (right) NeuroD or TUJ1. B, Individual DCX+ neurons at 7 years of age in the human GCL with small, elongated cell bodies and few processes. C, DCX⁺ neuron in the 13 year old human DG with prominent dendrites and an axon; 13 years is the oldest age where these cells were detected. D, (Left, Middle) DCX⁺ cells in the 19 and 36 year old hippocampus: hilus, GCL, and nearby white matter (wm). (Right) DCX⁺ OLIG2⁺ cells in the 13 year old hippocampus. E, 35 year old temporal lobe section with (Left) DCX⁺ neurons and (Right) DCX⁺ PSA-NCAM⁺ neurons in the entorhinal cortex (EnCx) are absent in the granule cell layer (GCL) of the same samples.