# Two Places the Adult Brain Still Makes Neurons

> A neuron counts as new only when the same cells carry two facts at once: a birth record, a label showing that their DNA was copied at a known time, so they were most likely born then or later, and a neuron identity. This lesson teaches that rule first, with birth labels such as tritiated thymidine and BrdU, and then uses it to read every experiment that follows. Adult rodents make neurons in two places, the subventricular zone, whose young neurons migrate to the olfactory bulb, and the dentate gyrus of the hippocampus, where the rate falls steeply with age. Astrocyte-like stem cells make them, many newborn cells die early, niche signals and experience can act on division or on survival, and the survivors wire into working circuits, where they help with tasks such as telling similar places apart. In people, the two lines of evidence that carry a birth record, BrdU in cancer patients and bomb-test carbon-14 in neuronal DNA, point to some adult neurogenesis in the hippocampus. Stains for young neurons carry only a state, they disagree, and the rate and role in people are still unsettled.

Course: Developmental Neuroscience (Rebuilding, evolving, and going wrong). Level: advanced university and graduate. About 32 minutes.
Lesson page (link this one when citing): https://stemhop.com/lessons/adult-neurogenesis/

Builds on: The Ventricular Zone: Radial Glia as Stem Cells · Tangential Migration: Interneurons and the Long Way Round · Too Many Neurons: Naturally Occurring Cell Death

## What you'll learn

- Explain how a birth label such as tritiated thymidine or BrdU, combined with a neuron marker in the same cell, shows that a neuron was born in adulthood, and why a young-neuron marker alone does not.
- Describe the two well-established neurogenic niches of adult rodents, the subventricular zone with its rostral migratory stream to the olfactory bulb and the subgranular zone of the dentate gyrus, and the astrocyte-like stem cells and B to C to A cell lineage that make their neurons.
- Describe how adult-born neurons mature and join existing circuits, and what they are thought to do in rodents, including pattern separation in the dentate gyrus and odour learning in the olfactory bulb.
- Explain how niche signals such as Notch and Wnt and experience such as running change the output of new neurons, and tell an effect on division from an effect on survival.
- Summarise the human evidence, from bomb-test carbon-14 dating of neurons to stains for young neurons, and explain why the rate and role of adult neurogenesis in people remain unsettled.

**Inject an adult rat with a label that marks newly made DNA. A month later, a labelled cell sits in its hippocampus. Is it a new neuron?**

For most of the twentieth century the working rule was simple: the neurons you are born with are the neurons you keep, and those you lose are not replaced. Adult rodents turned out to be an exception: they keep making neurons in two places. This lesson starts with how anyone can tell that a neuron is new, because every later claim rests on it, then asks which cell makes them, how many are kept and what they do, and ends with people, where the question is still argued over.

## How you know a neuron is new

To find a newborn cell you need a mark that only newborn cells carry. A cell copies its DNA once per division, in S phase, and a neuron that has stopped dividing normally does not copy its genome again. Give an animal a labelled form of thymidine, the DNA building block used only in DNA, and every cell in S phase while the label is present builds it into its chromosomes.

Read the label with care: it dates the DNA, not the neuron. A labelled precursor can divide again after the label has cleared, and each daughter then inherits about half as much label, a quarter after two divisions. So a strongly labelled neuron found weeks later most likely descends from a cell that was copying its DNA while the label was present; its own last division came that day or later. Label can also enter a cell that never divided: through DNA repair, or when a dying neuron starts to copy its DNA and fails (abortive DNA synthesis). So studies need controls for both.

*Figure (AI-generated illustration).* A labelled DNA building block (BrdU) is incorporated into DNA by cells that are copying it. A labelled neuron found weeks later most likely descends from a cell that was copying its DNA while the label was present, so it was born that day or later, although DNA repair can also label a cell, so controls are needed. Antibodies let the same cell also be stained as a neuron. Image: a three-step strip. A label is given to a dividing cell but not to a non-dividing neuron; the label is built into the daughter cells' DNA; weeks later a neuron has a labelled nucleus while an old neuron does not, and antibody icons mark the label and a neuron-only protein. Labelling suggests, but does not prove, birth after the dose; DNA repair can also incorporate label. Chromosomes are drawn schematically.

Joseph Altman did this in rats in the 1960s with [tritiated thymidine](https://stemhop.com/methods/thymidine-analog-birthdating/), thymidine carrying radioactive hydrogen. He found labelled cells in the dentate gyrus of the hippocampus and in cells moving from the wall of the lateral ventricle to the olfactory bulb.

**Predict.** Under a light microscope, black grains sit over one nucleus in the dentate gyrus, with a glial cell pressed against it. Is the labelled cell a neuron? What second fact, about the same cell, would you need to be sure?

> **Check your prediction.** The light microscope could not tell. The grains might belong to a glial cell pressed against a neuron, so they gave a birth record but not an identity. That is a large part of why Altman’s claims were set aside for years.
>
> The proof came in steps. In rats, Michael Kaplan and James Hinds used electron microscopy in 1977 to show that thymidine-labelled cells in the adult dentate gyrus and olfactory bulb were neurons. Songbirds gave a second, clear case. In adult canaries, label given to dividing cells appeared first in the ventricular zone and weeks later in neurons of the song nucleus HVC, and electron microscopy confirmed that they were neurons. So the new neurons were born at the ventricle and migrated in (Goldman and Nottebohm, 1983).
>
> Then the label changed. Bromodeoxyuridine (BrdU) is a thymidine look-alike that an antibody can detect, so the same cell can be stained for BrdU and for a neuron-only protein such as NeuN. That made the identity check routine.

> **The one idea to carry through this lesson.** A neuron counts as new only when the same cells carry two facts at once: a **birth record**, a label showing that their DNA was copied at a known time, so, with the controls above, they were born then or later, and a **neuron identity**. A young-neuron marker such as doublecortin, a protein made by immature neurons, is a third kind of fact. It tells you what state a cell is in, not when the cell was born.

**What the rule covers.**

- **Which evidence.** In single cells, a label given to the animal supplies the birth record and a separate check the identity: tritiated thymidine with [electron microscopy](https://stemhop.com/methods/electron-microscopy/), or BrdU with a neuron protein. A retrovirus that integrates only in dividing cells, or another inherited label, passes to every descendant, so it records ancestry as well as a division after infection.
- **What it does not cover.** The rule tests whether neurons were born. It does not test whether they are wired in or what they do, which needs other experiments, such as recording, blocking their birth and testing behaviour. Nor does it give a rate per day, which needs a model on top of the birth record.

## Two places adult rodents keep making neurons

**Film transcript: Who Makes Them, and How Many Are Kept**

### Film transcript: Who Makes Them, and How Many Are Kept

Adult rodents make new neurons. Which cell makes them, and how many are kept? They appear in two places. One is the subventricular zone, in the wall of the lateral ventricles. Its young neurons migrate forward in chains to the olfactory bulb. The other is the dentate gyrus of the hippocampus. There, the rate falls steeply with age.

Four kinds of cell sit in that wall. Ependymal cells line it, with astrocyte-like B cells beneath. Both divide rarely. Fast-dividing C cells multiply, and young neurons, the A cells, migrate away. In 1999, Fiona Doetsch and colleagues infused adult mice with a drug that kills cells copying DNA, then stopped. Pause and predict. Which cells survive? And which one rebuilds the zone?

The fast dividers and most neuroblasts died. The rare dividers survived. Then the astrocyte-like cells divided. Precursors came back, then neuroblasts, in the normal order. That makes them the leading candidate, not yet proof. Proof needs an inherited label showing the neurons descend from an astrocyte-like cell, in the same cells as neuron identity. This retrovirus writes its genes into the chromosome of a dividing cell, so every descendant inherits them. This virus enters only through one receptor. The mice were engineered to make it only in cells making GFAP. So in untreated mice, only astrocyte-like cells were infected. Later, the marker turned up in new olfactory bulb neurons.

The dentate gyrus follows the same plan, with radial, astrocyte-like stem cells. How many are kept? In mice, many newborn cells die early, mostly within the first days. Later, while they mature, survival also depends partly on their own activity. Young neurons lacking NMDA receptors survive less well than their neighbours. Other survival signals may help too. More new neurons can mean more division, or more young cells kept. Two time points help tell them apart. After one dose of BrdU, a one-day count is an index of division. At four weeks, label plus a neuron protein counts labelled cells that are neurons by then. Pause and predict. If running acted only on division, what would you see at each time point?

More labelled cells at one day, and more new neurons at four weeks, in the same proportion. In those mouse studies, running mainly raised division. Enriched housing mainly raised survival. Takeaway: the stem cell looks glial, and born is not the same as kept.

From the film

- Adult rodents make neurons in two places, the subventricular zone and the dentate gyrus, where the rate falls steeply with age.
- Astrocyte-like stem cells make the neurons; after antimitotic treatment in mice, surviving B cells regenerated C cells and then neuroblasts.
- More new neurons can mean more division or more survival; in the classic mouse studies, running acted mainly on division and enriched housing mainly on survival.

Applied in adult rodents, the rule finds new neurons all their lives in the two places Altman had pointed to. In the dentate gyrus the rate falls steeply with age: in rats, neurogenesis there has dropped sharply by middle age (Kuhn, Dickinson-Anson and Gage, 1996).

The first place is the **subventricular zone** (SVZ), a thin layer in the wall of the lateral ventricles. Neuroblasts, young neurons that can still migrate, line up in chains and travel forward along the **rostral migratory stream** to the olfactory bulb, the first relay for smell, where they become interneurons, local neurons that connect nearby cells. Carlos Lois and Arturo Alvarez-Buylla showed this long-distance migration in adult mice in 1994; the stream returns as a case of tangential migration in [that lesson](https://stemhop.com/lessons/tangential-migration/).

The second is the **subgranular zone** (SGZ) at the inner edge of the dentate gyrus, the input stage of the hippocampus, a sheet of small, densely packed granule neurons.

## Which cell is the stem cell?

A **stem cell** here means a cell that can divide to make more of itself and also make cells that become neurons. Finding it in the subventricular zone meant telling apart four kinds of cell that sit there. The letters are the field’s labels; the roles are what matter.

- **Ependymal cells** line the ventricle itself: a single sheet of cells with beating cilia that rarely divide.
- **Astrocyte-like cells**, called B cells, lie just beneath them. They look and stain like astrocytes, cells long thought only to support neurons, and they divide rarely.
- **Fast-dividing precursors**, called C cells or transit-amplifying cells, divide quickly a few times, multiplying the number of cells on the way to neurons.
- **Neuroblasts**, called A cells, are young neurons that migrate in chains toward the olfactory bulb.

In 1999 Fiona Doetsch, Alvarez-Buylla and colleagues used an **antimitotic** drug, one that kills cells while they copy their DNA. They infused it into the brains of adult mice for several days and then stopped.

**Predict.** Which of the four cell types would you expect to survive the drug? If the zone later rebuilt itself, what would the responsible cell have to do first, and what would it make, in what order?

> **Check your prediction.** The fast-dividing precursors died, because at any moment many of them are copying DNA. Most neuroblasts, made from those precursors, were gone too. The cells that divide rarely, the ependymal cells and the astrocyte-like B cells, survived.
>
> After the drug was stopped, the astrocyte-like cells began dividing. Fast-dividing precursors reappeared, and then neuroblasts, in the same order as the normal sequence. The ependymal cells did not rebuild the zone.

*Figure (AI-generated illustration).* In the adult mouse subventricular zone, astrocyte-like B stem cells generate transit-amplifying C cells, then A-cell neuroblasts. Neuroblasts migrate along the rostral migratory stream to the olfactory bulb and become interneurons. After antimitotic treatment removed C cells and most A cells, surviving B cells divided and regenerated C cells followed by neuroblasts. Image: an adult mouse SVZ schematic shows ependymal cells lining the lateral ventricle and the sequential B stem cell to C precursor to A neuroblast pathway. A cells migrate along the rostral migratory stream to an interneuron in a separate olfactory bulb region. Two smaller panels show elimination of C cells and most A cells by an antimitotic drug, followed by surviving B cells dividing and restoring C cells and then A cells.

Surviving and rebuilding make the astrocyte-like cells the leading candidate. That is not yet proof that they make neurons in a normal brain: a cell that survives a poison and divides afterwards might be repairing the zone, not running it. Proof needs the rule’s two facts in the same cells, with a label that also records ancestry: one showing that the neurons descend from an astrocyte-like cell.

So the group used a virus as an inherited label. This kind of retrovirus writes its genes into the chromosomes of a dividing cell it infects, so every descendant of that cell inherits them, however many times it divides. The virus they used can enter a cell only through one receptor, and the mice were engineered to make that receptor only in cells that make the astrocyte protein glial fibrillary acidic protein (GFAP), the astrocyte-like cells. In untreated mice the virus entered only those cells, and later the marker it carried turned up in new neurons of the olfactory bulb. The neurons had descended from astrocyte-like cells, and were born after the infection; the ancestry was the new information. So the lineage runs:

> **B cell (astrocyte-like stem cell) → C cell (fast-dividing, transit-amplifying precursor) → A cell (neuroblast) → olfactory bulb interneuron**

The adult stem cell looks like a glial cell, which fits development: the embryonic progenitors of the brain, the radial glia, are glial in character too ([that lesson](https://stemhop.com/lessons/radial-glia-progenitors/)). The dentate gyrus follows the same plan. Its stem cells are radial, astrocyte-like cells whose bodies sit at the inner edge of the granule layer; they make fast-dividing precursors, and those make young granule neurons.

## Born is not the same as kept

More new neurons can come from two different steps: more division of precursors, or more of their young offspring kept as neurons. The rule, used at two time points, helps tell them apart. One day after a dose of [BrdU](https://stemhop.com/methods/thymidine-analog-birthdating/), the labelled cells are mostly precursors that were copying their DNA, plus some of their first daughters, so that count is an index of division. Weeks later, cells carrying both BrdU and NeuN are the labelled cells that became neurons and were kept. Dividing the late count by the early count gives a neuron yield per early labelled cell. It is not quite a survival rate, because labelled precursors can divide again after the first day and some daughters become glia; studies that need each step pinned down measure those too.

**Predict.** Mice live with or without a running wheel and get one dose of BrdU. Some are examined after one day, others after four weeks. If running acted only on division, how would runners differ from controls at each time point? What if it acted only on survival?

> **Check your prediction.** | If running acts on | BrdU cells at 1 day | BrdU-and-NeuN cells at 4 weeks | Late count ÷ early count (neuron yield) |
> | --- | --- | --- | --- |
> | division only | more | more, in proportion | the same |
> | survival only | the same | more | higher |
>
> The table is the simple version: it assumes that after the first day the labelled cells divide again, choose a fate and die at the same rates in both groups.
>
> In the classic mouse studies, the two kinds of experience acted mainly at different steps. Running acted mainly on division: mice with a running wheel had more dividing precursors and more new dentate gyrus neurons than mice in standard cages (van Praag, Kempermann and Gage, 1999). Enriched housing acted mainly on survival: mice housed in larger cages with toys and companions had more surviving new neurons and more dentate granule cells (Kempermann, Kuhn and Gage, 1997). In rodents, learning tasks that depend on the hippocampus also mostly change how many new neurons survive, and the effect depends on when the cells were born (reviewed by Ming and Song, 2011). The MCAT question below gives made-up counts to try this with numbers.

So a study that reports more new neurons has to say which step changed.

*Figure (AI-generated illustration).* More new neurons can come from two different steps. Counting labelled cells one day after the dose indexes division; counting cells that carry both the label and a neuron protein weeks later counts those kept as neurons. Only the pair of counts tells the two apart, so a study reporting more new neurons has to say which step changed. Each row is a hypothetical outcome; in the classic studies, running acted mainly on division and enriched housing mainly on survival. Image: four panels in two rows. In the more-division row, runners have twice as many labelled cells as controls one day after the label and twice as many labelled neurons four weeks later. In the more-survival row, runners and controls have the same number of labelled cells at one day, but runners have more labelled neurons four weeks later. Cell counts are schematic.

Survival is not a small step. In mice, many newborn cells die early, and most of that loss comes within days of birth (Sierra and colleagues, 2010). A later step, while the cells mature, depends partly on the young neuron’s own activity: young dentate neurons lacking N-methyl-D-aspartate (NMDA) receptors, a type of glutamate receptor, survived less well than their neighbours (Tashiro and colleagues, 2006). The shape is familiar from [the lesson on the neurotrophic hypothesis](https://stemhop.com/lessons/neurotrophic-hypothesis/): make too many, then keep only some. The evidence here does not show a target-derived factor like nerve growth factor (NGF) doing this job, though trophic signals may still contribute.

Local signals in the niche also set the pace. Notch signalling helps keep stem cells at rest, in reserve ([that lesson](https://stemhop.com/lessons/notch-signaling/)), although some Notch signals also sustain stem cells once they are dividing. Wnt pushes precursors on toward neurons: in adult rats extra Wnt3 increased neuron production, while blocking Wnt signalling nearly abolished it (Lie and colleagues, 2005) ([that lesson](https://stemhop.com/lessons/wnt-signaling/)).

## New neurons join working circuits

A new neuron is useful only if it is wired in. In the dentate gyrus, new granule neurons grow dendrites and send axons into the hippocampal circuit over several weeks. For a time while they mature, young granule neurons fire more readily than older ones in response to weak input (reviewed by Miller and Sahay, 2019).

That excitability may be what makes them useful. A leading hypothesis is that a small, excitable young population helps the dentate gyrus give two similar inputs different outputs. That job is **pattern separation**, telling apart similar places or events. Mice in which adult neurogenesis was blocked did worse at distinguishing nearby locations (Clelland and colleagues, 2009). In the rodent olfactory bulb, the steady supply of new interneurons has been linked to odour learning.

*Figure (AI-generated illustration).* A leading hypothesis, not a settled result: a small, excitable young population helps the dentate gyrus give two similar inputs different outputs, the job called pattern separation. Mice in which adult neurogenesis was blocked did worse at distinguishing nearby locations. Image: two schematic circuits. On the left, two similar inputs reach a sheet of dentate neurons that includes three gold young, more excitable neurons, and the two outputs stay separate. On the right, with neurogenesis blocked, the same two inputs reach a sheet with no young neurons and the two outputs overlap heavily. The circuit is schematic, not anatomy.

The first sign that new adult neurons join working circuits came from songbirds: in 1984 John Paton and Fernando Nottebohm recorded from new neurons in the canary song nucleus HVC and found that some of them responded to sound.

### Design problem

You suspect that running helps mice tell apart two similar places because it adds new dentate granule neurons. Design experiments to test the three links in that claim: that running adds new neurons, that the better discrimination needs them, and that the effect is specific.

> **One way to do it.** First, show that running adds neurons, not just dividing cells. Give BrdU to runners and controls, and several weeks later count cells positive for both BrdU and NeuN. Count BrdU cells one day after the dose too, so you can see whether running changed the early count, the later neuron yield, or both.
>
> Second, test the behaviour. Train runners and controls to tell apart two nearby locations, and also two far-apart ones. If new neurons help with pattern separation, runners should do better mainly when the locations are close.
>
> Third, test whether the new neurons are needed. Block adult neurogenesis in runners, for example with X-irradiation of the hippocampus or a genetic method that kills dividing precursors, and ask whether their advantage on nearby locations goes away. Check that the block does not stop the mice running, harm other cells, or impair the easy far-apart task; otherwise a loss could have another cause.

## New neurons in the adult human brain

**Film transcript: Do You Make Them Too?**

### Film transcript: Do You Make Them Too?

Adult rodents make new neurons. Do adult humans make them too? BrdU cannot be given to healthy people for research. But in the 1990s, some cancer patients received it for clinical reasons. After death, their dentate gyrus held cells with both BrdU and a neuron protein. Both facts, in single human cells. Bomb tests in the 1950s and early 1960s sharply raised the carbon-14 in the air. That carbon ends up in the DNA of dividing cells. A neuron's DNA keeps the level of its last division. Neurons at the back of the cortex match the person's own birth year. They were born around then. So were olfactory bulb neurons. Pause and predict. Hippocampal neurons do not match the birth year.

What can you conclude, and what not? An offset beyond measurement error points to neurons added after birth, unless repair or stray nuclei explain it. But a sample averages many nuclei, so it cannot say how many, or when.

One model estimate is about 700 new neurons per hippocampus per day. Stains for young neurons show a state, not a birth date. In people, it is debated whether they mark only new neurons. And the stains disagree. One study found none in adults. Others found thousands. Pause and predict. Suppose 700 new neurons arrive each day, and each stays young and marked for weeks or months. How many marked cells should one hippocampus hold?

Rate times time. Even three weeks gives about 15,000 across one hippocampus. These are toy numbers. So a count of none needs an explanation. A far lower rate in those donors. A young-neuron protein lost when tissue waited, or was fixed for long. Or a young state that is short, or unmarked, in people. Or sections and stains that missed cells that were there.

The human evidence that carries a birth record points to some new neurons. How many new neurons people make, and what they do, is still open. Takeaway: ask whether the same cells carry a birth record and a neuron identity.

From the film

- In people, the evidence that carries a birth record, such as bomb-test carbon-14 dating, points to some hippocampal neurogenesis.
- Stains for young neurons show a state, not a birth date, and studies using them disagree.
- How many new neurons people make, and what they do, is still open.

Do adult humans do this too? Recent reviews read the evidence as increasingly favouring some adult neurogenesis in the human hippocampus, from several independent methods, but its rate, how much it varies between people and what it does are still unsettled. The rule explains what each human method can and cannot show, because the methods differ in which facts they supply.

[BrdU](https://stemhop.com/methods/thymidine-analog-birthdating/) cannot be given to healthy people for research, because it is built into the DNA of every dividing cell. But in the 1990s some patients with cancer received it for clinical reasons, to measure how fast their tumours were dividing. Peter Eriksson, Fred Gage and colleagues examined the brains of some of these patients after death and found cells in the dentate gyrus that carried both BrdU and neuron markers such as NeuN (Eriksson and colleagues, 1998). Those cells most likely descended from cells that copied their DNA while the patients were adults, and were neurons when they died.

| Human method | Birth record? | Neuron identity? | In the same cells? |
| --- | --- | --- | --- |
| BrdU in cancer patients | yes, BrdU in DNA | yes, NeuN | yes, single cells |
| Bomb-test carbon-14 in sorted nuclei | yes, the carbon-14 of the air when the DNA was made | yes, nuclei sorted with a NeuN antibody | only as an average over many nuclei |
| Staining for doublecortin (DCX) or PSA-NCAM | no, a state, not a date | a young-neuron state, if the marker means what it is taken to mean | not applicable |

The two methods that carry a birth record, BrdU in patients and carbon-14, say more about when cells were born than a marker can, though each has limits of its own.

### Carbon-14: a birth record everyone carries

Jonas Frisén’s group at the Karolinska Institute found a label already in the atmosphere. Nuclear weapons tests in the 1950s and early 1960s sharply raised the carbon-14 in the air, and the level fell after the 1963 treaty that banned tests in the atmosphere. Carbon from the air ends up in the DNA of every cell that divides, so the DNA of a cell that never divides again keeps the carbon-14 of its birth year. A neuron’s birth here means its last division, not the person’s birth.

> **carbon-14 in the DNA of a cell = carbon-14 in the air in the year that DNA was made**

Sorting nuclei from post-mortem brain into neurons and other cells, the group found that the DNA of neurons at the back of the cerebral cortex matches the air of the person’s own birth year, so those neurons are as old as the person (Spalding and colleagues, 2005). So are olfactory bulb neurons (Bergmann and colleagues, 2012), unlike in rodents; in human infants the stream of young neurons toward the front of the brain thins out during the first 18 months of life and is nearly gone by adulthood (Sanai and colleagues, 2011).

**Predict.** In the same donors, hippocampal neurons carry carbon-14 that does not match the birth year. What can you conclude, and what can you not?

> **Check your prediction.** You can conclude that some hippocampal DNA was made after birth, beyond what measurement error and normal early-postnatal development explain, so some hippocampal neurons were added later (Spalding and colleagues, 2013), unless DNA repair, cells with extra copies of their DNA or stray non-neuronal nuclei in the sorted sample add carbon. Each is a reason for a control, not a reason to drop the rule.
>
> You cannot say how many neurons were added, or when. A sample averages many nuclei:
>
> > **measured carbon-14 = (fraction born at birth × level at birth) + (fraction born later × average level when they were born)**
>
> On the falling side of the curve, a donor born in 1972 whose hippocampal neurons read like the air of 1976 could have had all of that DNA made in 1976. Or most neurons were born around birth and some at scattered later times, which pull the average down. The reading alone cannot tell which, and because the sample averages carbon-14, not years, 1976 is not simply the mean birth year either.

Turning the reading into a rate needs a model of how neurons are added. Spalding and colleagues fitted one and estimated about 700 new neurons per hippocampus per day (its assumptions are under Going deeper). A later study reported new interneurons in the adult striatum, beside the ventricle wall (Ernst and colleagues, 2014), a site that is less established.

### Markers: a state, not a birth date

Staining tissue for proteins of young neurons has given opposite answers. The proteins are doublecortin (DCX) and PSA-NCAM, an adhesion protein coated with a long sugar chain, carried by young, migrating neurons. In 2018 Sorrells, Alvarez-Buylla and colleagues found that young neurons in the human dentate gyrus fell sharply in childhood and were not detected in adults. The next month Boldrini and colleagues reported thousands of immature neurons in healthy people from adolescence into old age, and in 2019 Moreno-Jiménez, Llorens-Martín and colleagues, working with tissue fixed under tightly controlled conditions, reported thousands of DCX-positive neurons into old age.

**Predict.** A birth rate and a marker count can be compared only through time: if new neurons arrive at a steady rate and each stays alive and marker-positive for a set time, the number marked at any moment is the rate times that time. Take the carbon-14 estimate of about 700 new neurons per hippocampus per day as that rate, and suppose each stays DCX-positive for a few weeks to several months; the human duration is not known. How many DCX-positive cells should one hippocampus hold at any moment? Can a count of none be squared with that?

> **Check your prediction.** | If each cell stays DCX-positive for | DCX-positive cells per hippocampus at any moment |
> | --- | --- |
> | 3 weeks (21 days) | 700 × 21 = 14,700, about 15,000 |
> | 3 months (90 days) | 700 × 90 = 63,000 |
> | 6 months (about 180 days) | 700 × 180 = 126,000 |
>
> Even the shortest phase predicts more than ten thousand across one hippocampus. These are toy numbers: 700 a day is a model average of neurons that persisted, not a measured entry into the DCX state. Still, a finding of none in adults is hard to square with them unless the rate in those donors was far lower than the average estimate, the antigen was lost in handling, the young state in humans is short or does not carry the marker, or the sections sampled and the stain missed cells that were there. The arithmetic does not pick among these, but it turns a contradiction into a precise question.

Why would careful laboratories disagree? Two limits matter most. DCX staining weakens with time after death and with long fixation, so a negative result may mean the antigen was lost. And in humans it is debated whether DCX and PSA-NCAM mark only newborn neurons. If the adult human dentate gyrus holds cells that are immature in their markers but were born long ago, a marker count and a carbon-14 estimate could both be correct and still describe different things. The donor collections also differ in age, cause of death, epilepsy and medication, and [single-nucleus](https://stemhop.com/methods/single-cell-rna-seq/) RNA sequencing has pointed both ways.

None of the rodent functions above transfers to people automatically. Whatever adult neurons humans make, the carbon-14 data point to the hippocampus and striatum, not the olfactory bulb where rodents put most of theirs.

## Putting it together

| Experiment | Animal | Result | What it supports | What it does not establish |
| --- | --- | --- | --- | --- |
| Tritiated thymidine (Altman, 1960s) | rat | labelled cells in the dentate gyrus and in cells moving to the olfactory bulb | cells copied their DNA in these regions after the dose, usually because they divided | that the labelled cells are neurons |
| Electron microscopy of labelled cells (Kaplan and Hinds, 1977) | rat | labelled cells are neurons | birth record and identity in the same cells | rate or function |
| Thymidine in canaries (Goldman and Nottebohm, 1983) | adult canary | label first in the ventricular zone, weeks later in HVC neurons | new neurons are born at the ventricle and migrate in | what they do |
| Antimitotic drug, then recovery (Doetsch, 1999) | adult mouse | astrocyte-like cells survive and rebuild the zone in order | astrocyte-like cells are the stem-cell candidates | that they make neurons in an untreated brain |
| Retrovirus entering only GFAP-expressing cells (Doetsch, 1999) | adult mouse | the inherited marker appears in new olfactory bulb neurons | astrocyte-like cells make neurons | how many olfactory bulb neurons come from them, or that every GFAP-expressing cell is a stem cell |
| BrdU at one day and at four weeks, with experience | mouse | running mainly raises division; enriched housing mainly raises survival | experience acts at different steps | what the extra neurons do |
| Neurogenesis blocked (Clelland, 2009) | mouse | worse at telling apart nearby locations | a role in pattern separation | that people use new neurons this way |
| BrdU in cancer patients (Eriksson, 1998) | human | BrdU and NeuN in the same dentate gyrus cells | adult-born human neurons exist | the rate, or that healthy people make them at the same rate |
| Bomb-test carbon-14 (Spalding, 2005 and 2013; Bergmann, 2012) | human | cortex and olfactory bulb neurons as old as the person; hippocampal neurons younger | some hippocampal neurons are added after birth | a rate without a model, or which single cells |
| DCX and PSA-NCAM staining (2018, 2019) | human | none in adults in one study; thousands in others | a young-neuron state in some tissue | when those cells were born |

In rodents the picture is well established: two niches, glial-like stem cells, many newborn cells lost early, and survivors that join circuits. In people, the evidence that carries a birth record points to some hippocampal neurogenesis. How much, how it varies, and what it does are still open.

## Going deeper

### The discovery, step by step

Altman’s method was [autoradiography](https://stemhop.com/methods/isotope-tracers/): a photographic emulsion laid over a thin brain section shows tritiated-thymidine-labelled nuclei as black silver grains. His dentate gyrus report, with Gopal Das, came in 1965 and his olfactory bulb report in 1969. Even after Kaplan and Hinds had answered the identity objection in rats, there was contrary evidence from primates: in 1985 Pasko Rakic reported that thymidine given to rhesus monkeys after puberty, examined from 3 days to 6 years later, labelled glial cells but no neurons in the regions he examined.

The canary experiment started from a hormone effect. Adult female canaries given testosterone start to sing, and HVC grows. In 1983 Steven Goldman and Fernando Nottebohm injected [tritiated thymidine](https://stemhop.com/methods/thymidine-analog-birthdating/) into adult female canaries and found that 1.8 to 4.0 percent of HVC neurons were labelled, with or without the hormone, so the hormone was not needed for the new neurons to appear.

In 1999 two groups using BrdU reported new dentate gyrus neurons in adult monkeys (Gould and colleagues, and Kornack and Rakic, both in macaques), with Kornack and Rakic estimating a rate about ten times lower than in rodents.

### A rival candidate for the stem cell

In the same year as Doetsch’s experiments, Johansson and colleagues, from Frisén’s group, claimed that the ependymal cells lining the ventricle are the stem cells of the adult zone. Doetsch’s drug experiment pointed instead to the astrocyte-like cells, since the ependymal cells survived the drug but did not rebuild the zone. This lesson follows Doetsch’s evidence, but the ependymal claim was a serious rival at the time.

### Reading the bomb curve

Carbon-14 is made at a nearly steady rate in the upper atmosphere. Above-ground tests, at their height between 1955 and 1963, nearly doubled the level in northern hemisphere air; afterwards it fell year by year as the extra carbon-14 moved into the oceans and the biosphere. Plants fix that carbon dioxide, and people eat the plants or animals that ate them. Carbon-14 decays slowly (half-life about 5,730 years), so over a lifetime the decay is negligible; the clock is the changing level in the air.

In 2005 Kirsty Spalding, Frisén and colleagues first showed that DNA carbon-14 tracks the atmospheric curve. They sorted the nuclei with an antibody to NeuN and measured carbon-14 by accelerator mass spectrometry; in the occipital cortex, non-neuronal cells were younger than the person.

### Two donors on the curve

Try this before opening the answer.

Two people donate hippocampus and cortex: one born in 1950, before the bomb peak, and one born in 1972, after it. In the person born in 1950, the carbon-14 of cortical neuron DNA matches the air of 1950, but the hippocampal neuron DNA holds more. What does each result mean, and why is the person born before the peak the more decisive test?

> **One way to do it.** The cortical value matches the birth year, so essentially all cortical neurons were born around birth. The hippocampal value is higher, and in 1950 the air held the low pre-bomb level. Every year from about 1955 through the decades these donors lived had more carbon-14 than 1950, so, unless DNA repair, extra DNA copies or stray non-neuronal nuclei add bomb carbon, the value can rise only through DNA made after 1955, when this person was at least five. Most likely some hippocampal neurons were born after early childhood.
>
> In the person born in 1972, neurons born later carry less carbon-14 than birth-year neurons, so they lower the value. How much depends on how many were added, when, and how fast the curve was falling, and the shortfall has to be larger than measurement error. Near 1972 the curve is steep, so a few months’ error in when the birth-year neurons were made, or in how long carbon takes to reach the body through food, shifts the expected value by an amount that can rival a small signal. Before the bomb the curve is nearly flat, and those errors matter much less. So in the person born in 1950 any rise is easy to interpret, though its size still depends on how many neurons were added and when.

### The hippocampal turnover estimate

Spalding and colleagues (2013) fitted the carbon-14 of hippocampal neurons from donors born across the twentieth century to turnover models. They estimated that about one third of hippocampal neurons are subject to exchange, with about 700 new neurons added per hippocampus per day, a turnover of 1.75 percent a year within the renewing fraction. The estimate rests on the model chosen, on pure neuronal sorting, and on DNA repair adding little carbon.

### The human studies in detail

Sorrells and colleagues (March 2018) also stained for dividing progenitors. Young neurons fell most steeply in the first year of life, a few remained at 7 and 13 years, and none were detected at ages 18 to 77, in post-mortem controls or in tissue removed during epilepsy surgery. A similar decline in juvenile macaques led them to conclude that adult human hippocampal neurogenesis does not continue or is extremely rare.

Boldrini and colleagues counted in whole post-mortem hippocampi from healthy people aged 14 to 79, and found similar numbers of intermediate progenitors at every age, with a smaller pool of resting stem cells in the front of the dentate gyrus in older people. Moreno-Jiménez and colleagues found DCX-positive neurons into the ninth decade, with fewer as Alzheimer’s disease advanced.

[Single-nucleus](https://stemhop.com/methods/single-cell-rna-seq/) RNA sequencing has not ended the argument. Franjic and colleagues (2022) found signs of neurogenesis in adult macaques, pigs and mice but not in adult humans, and found doublecortin (DCX) in many human neurons that were not immature. Dumitru, Frisén and colleagues (2025) combined single-nucleus RNA sequencing, the proliferation marker Ki67 and machine learning, and reported dividing neural progenitors in the adult human dentate gyrus.

---

From STEMhop (https://stemhop.com), free lessons by Christopher Deppmann. Lesson text © the authors. Practice questions: https://stemhop.com/lessons/adult-neurogenesis/
