Science oft resembles the federal tax code: the rules are rigid, but they also keep changing. So it has been with the study of neurogenesis, or the creation of neurons in the human brain. Not long ago a hard-and-fast rule held that neurons could neither divide nor emerge from elsewhere. The neurons you were born with, in short, were the ones you took to your grave. That dogma began to change in the 1980s, however, when Fernando Nottebohm of the Rockefeller University discovered that neurons were dividing in the forebrains of canaries. It appeared that neurons divided after all.

Neuroscientists resisted this idea at first. It clashed with all established facts and belief. But slowly the accumulation of data moved minds. Now newborn brain cells are popping up in studies like mushrooms after rain. Scientists have even found that neurogenesis increases after physical exercise--a great relief to baby boomers who fried too many brain cells in college. The latest research shows that new neurons are hatching in the most hallowed of all brain regions, the hippocampus, the seat of declarative and spatial memory. These discoveries raise the prospect that we might learn to manipulate neurogenesis to relieve ailments such as stroke and cognitive decline.

Skeptical scientists, however, have met all this news with an important question: What use are new neurons if they do not somehow wire themselves into the existing circuitry of the brain--and how are these inexperienced neurons going to do that? The difficulty of incorporating new cells into the intricate, tightly woven fabric of neural connections in the grown-up brain was always one of the stronger arguments against the existence of new neurons in the first place. What good are these cellular neophytes if they merely become passive bystanders?

In the past few years, however, findings from several labs have shown that new neurons do connect with existing circuitry. A recent paper by neuroscientists Nohjin Kee, Cátia M. Teixeira and their colleagues at the University of Toronto contributes a key piece of evidence. It shows that new neurons indeed integrate themselves into functional networks in the hippocampus and that these new recruits actually boost memory--or at least participate in making new memories. It is perhaps the clearest demonstration yet that new neurons join existing teams and do real work. How did researchers crack that nut?

The discovery required a microscope, a tank of water and a mutant mouse.

The Swim Test
First, some memory basics. Memories are not held inside neurons, the brain's cells. Rather they are set in the connections between neurons, called synapses--tiny gaps across which the signal-emitting finger of one neuron (an axon) sends a message to the signal-receiving finger of another neuron (a dendrite). Memories are created when nerve cells in a circuit increase the strength of their connections.

Kee and his fellow researchers first trained mice in a standard memory test device called a Morris water maze. This test consists of plunking a mouse in a vat of milky water in which a small platform is hidden somewhere just below the surface. The mouse swims in a highly motivated manner until it finds the submerged life preserver, where it can stand comfortably with its head above water.

If placed in the tank again, the mouse remembers the platform's location and swims to it pronto. With subsequent trials the mouse gets really good at finding the platform quickly. The mouse thus demonstrates learning--specifically, spatial learning, given that the rodent cannot see the platform but must remember its location by using landmarks placed on the tank walls or suspended overhead. (Incidentally, this test works well because mice are good swimmers and take naturally to the challenge.) Synapses are strengthened when the animal is subjected to the stressful or cognitively challenging experience of finding its way to the platform.

To make memories stick, neurons must turn on genes to manufacture proteins that will cement more strongly the synapses shared among them. The molecules that establish current flow around synapses, as with all proteins in the body, degenerate and are replaced constantly over a period of hours or days. Scientists have known since the 1960s that turning on genes was somehow involved in making memories permanent, because genes tell cells to produce proteins, and new proteins must be synthesized in the neural networks within minutes of an experience for it to be coded in memory.

After training mice in this water task, Kee and his colleagues looked to see if the task switched on learning-associated genes in new hippocampal neurons. They knew they were looking at new neurons because they had injected the mice with a marker for dividing cells--one of the DNA bases that has been chemically modified so that it shows up in tissue slides under a microscope. Because cells must make DNA to divide, newly minted cells carry this fluorescent marker (bromodeoxyuridine) in their nuclei. When Kee and his team looked through the microscope at the hippocampus of mice that had learned to find the platform, they indeed found that certain "memory" genes known as c-fos and arc were turned on in many of the new neurons. Nevertheless, as Kee knew, it is possible for these genes to get turned on by forms of mental stimulation other than learning. How would they know these new neurons were actually storing memories of where the submerged platform was hidden?

The Control
Enter the mutant mouse. As a control, Kee and his co-workers also put through the water maze a mouse that had a disabling mutation in an enzyme (CaMKII) that is essential for making memories. This mouse could find the platform, but it could not carry the memory to the next test; it had to find the platform afresh every time. As it turned out, its hippocampus had as many newborn neurons as that of the more learned mice, but its memory gene, c-fos, was not switched on any more than it was in control animals that had not been trained. The memory gene's activation seemed to be what made the learning possible.

The researchers also found that by training the mice at different time intervals after the bromodeoxyuridine injection they could determine just how old the new neurons had to be before they joined the adult circuits of memory storage. They found that new neurons are not involved in memory until they reach the age of about four to six weeks. Yet, like impressionable children, these new neurons were actually more likely to participate in learning than were the adult neurons already established in networks.

The study might have been stronger if the team had used an additional mutant of a different type or various drugs to interfere with memory [see "Erasing Memories," by R. Douglas Fields; SCIENTIFIC AMERICAN MIND, December 2005] instead of using only the CaMKII mutant mouse. This is because this enzyme is involved in turning on the c-fos gene for reasons unrelated to memory formation. Also, the researchers did not report whether or not the gene more closely associated with memory, arc, was switched on in the mutant mice.

Yet the experiment carries substantial weight. It shows that although most neurons in the adult brain do not divide, about 1 to 2 percent of the population at any given time is new in parts of the brain, including the hippocampus. It now seems clear that these new neurons are preferentially recruited into brain circuits that record new spatial memories. But the reasons for this recruitment--and how plugging new neurons into existing circuits might affect old memories--remain about as opaque as the instructions for the federal 1040 tax form.