Connectome (18 page)

Perhaps the early connectome is like a rough draft. I said above that the initial wiring and the creation of connections are guided by genes but also subject to randomness. And earlier I mentioned the theory that synapse elimination in the adult brain is driven by weakening, which in turn is driven by experience. By the same arguments, experience is likely to be the main driver of synapse elimination in the developing brain. And perhaps the elimination of many synapses from a branch leads to its pruning. These destructive processes refine the rough draft to produce the adult connectome.

This scenario is slightly misleading, however, because it suggests that creation and destruction occur in two distinct phases. The writing analogy clarifies why this is implausible. While working on a rough draft, I both add and remove words. There is
net
word creation because additions outnumber deletions. It's the other way around in the later phase of refinement, when the total number of words is decreasing. So it would be a mistake to think that before age two it's only synapse creation that occurs, and thereafter it's only synapse elimination. Net creation occurs early and net elimination occurs later, but both processes happen throughout life. Even in adulthood, when the total number of synapses remains roughly constant, both creation and elimination are taking place.

If synapse creation is mostly random while synapse elimination is driven by experience, shouldn't enriched cages cause synapse number to
decrease
in rats? Recall the finding of William Greenough and other researchers (mentioned in Chapter 5)—that synapses increase in number. We can only speculate, but here's one plausible scenario. Let's suppose that synapse elimination does happen at a greater rate in the brain of an enriched-cage rat, because it is learning more, but then, to replace the eliminated synapses, the brain steps up the creation of new ones. If creation more than compensates for elimination, the result is a net increase in synapse number. In this speculation, the increase in synapse number is the
effect
of learning rather than its cause.

The oxymoron
creative destruction
was central to the Austrian economist Joseph Schumpeter's theory of economic growth and progress. Its first word referred to the creation of new companies by entrepreneurs, and its second to the destruction of inefficient companies by bankruptcy. Brain development, writing an article, and economic growth all involve an intricate interplay between creation and destruction. Both processes are required for complex
patterns
of organization to evolve. When seen in this light, it verges on futile to measure progress by counting the total number of synapses in a brain, words in an article, or companies in an economy. It's the organization of the brain that matters, not the number of synapses.

 

By now you should have some appreciation for the intricacies of brain development. There are plenty of ways for such a complex process to go wrong. Disruption of the earliest steps of development, the creation and migration of neurons, is expected to cause abnormalities that are easy to see, such as microcephaly and lissencephaly. But disruption of the later steps of development could lead to
connectopathies,
disorders of neural connectivity. The total number of neurons and synapses would be normal, but they would be connected in a less than ideal way.

Remember the story of the Cray-1 supercomputer, which contained hundreds of thousands of wires totaling 67 miles in length? Remarkably, the first time it was powered up, it worked properly. The workers who built it had succeeded in connecting every single wire correctly. Your brain is far more complex, containing millions of miles of “wire.” It's a wonder that any brain can ever develop correctly at all.

As I mentioned earlier, the corpus callosum fails to develop in rare individuals. This connectopathy is visible in an MRI scan because the callosum is ordinarily so large. But given our inability to see brain connectivity clearly, it's likely that the vast majority of connectopathies remain undiscovered. These will be revealed as our technologies for finding connectomes advance.

Earlier I zeroed in on the most puzzling aspect of autism and schizophrenia—the lack of a clear and consistent neuropathology. Studies of twins convinced researchers years ago that autism and schizophrenia have some basis in faulty genes. But exactly which of the tens of thousands of genes are faulty? Most researchers now suspect that many of the culprits are somehow involved in brain development. Autism and schizophrenia are said to be
neurodevelopmental
disorders, in which the brain fails to grow normally. They are fundamentally different from neurodegenerative disorders like Alzheimer's disease, in which an originally normal brain starts to fall apart.

What is the evidence behind this suspicion? The case is more clear-cut for autism, as its symptoms are detected in early childhood. Whatever the neuropathology may be, it must have emerged during gestation and infancy, when the brain was growing most rapidly. Earlier I mentioned that autistic children have larger brains on average. Looking at brain growth over time reveals a more complex picture. The autistic brain is slightly smaller
than average at birth, larger than average from age two to age five, and average again by adulthood. In other words, the
rate
of brain growth is abnormal in autistic children. This suggests a developmental abnormality, but conclusive proof would require identifying a clear and consistent neuropathology that emerges in the womb or during infancy.

In the first half of the twentieth century, researchers did not believe that schizophrenia was neurodevelopmental. They hypothesized that the schizophrenic brain was normal during childhood and that it started to degenerate in late adolescence or young adulthood, triggering the first episode of psychosis. But they failed to find neuropathologies that should accompany a degenerating brain, so the hypothesis had to be abandoned.

Today many researchers speculate that schizophrenia, like autism,
is a neurodevelopmental disorder. It turns out that many schizophrenics experienced slight delays in learning to talk, move, and socialize, so perhaps their brains were already slightly abnormal in childhood. Their brain development might even have veered off course in the womb: Statistical studies suggest that pregnant mothers exposed to famine or viral infection are more likely to give birth to children who later develop schizophrenia.

So here's what researchers believe: Autism and schizophrenia are caused by some neuropathology, which is caused by abnormal brain development, which is caused by some combination of abnormal genetic and environmental influences. Neuroscientists are just beginning to find the genes, which could help them close in on the relevant developmental processes. This sounds encouraging, but I'm embarrassed to admit that the most important question has still not been answered: What is the neuropathology? Without data, theories have abounded. Since these are far too numerous to review exhaustively, I'll focus on the one that makes the most sense to me—the theory that autism and schizophrenia are connectopathies.

Recall that the autistic brain grows faster than normal in early childhood. The overgrowth is somewhat greater in the frontal cortex than in other lobes, perhaps because too many connections are created between neurons there. In addition, researchers speculate that too
few
connections
are created between the frontal cortex and other regions of the brain.

It's distressing to realize that this theory of autism is based on phrenological evidence and couched in phrenological terms. As I've mentioned, the enlargement of the autistic brain is only statistical, governing only averages. Diagnosing autism in an individual child based on the size of the brain or its regions would be grossly inaccurate. Statements about “too many” or “too few” connections are just as crudely phrenological as “too large” or “too small.” If autism is caused by a connectopathy, the difference will probably be found in the organization of connections, rather than in their overall number. The connectopathy would be invisible to our current technologies; hence the failure to find a clear neuropathology for autism.

Could schizophrenia, too, be caused
by a connectopathy? Here the most tantalizing evidence comes from studies of synapse elimination. Earlier I mentioned that adults have fewer synapses than babies, but I did not describe exactly when the reduction occurs. Researchers have found that synapse number declines rapidly after the peak in infancy, stays roughly constant during childhood, and drops rapidly again in adolescence.
Perhaps something goes wrong in the schizophrenic brain during this second reduction. The defect is probably not as simple as too few or too many synapses, as that kind of neuropathology would have been detected by now. Maybe the wrong synapses are eliminated, and this pushes the brain over the edge to psychosis.

Finding a clear and consistent neuropathology should be a central goal of research on autism and schizophrenia. We will need to go beyond phrenological methods if these disorders are connectopathies; we will need the technologies of connectomics. In fact, I believe that studying autism and schizophrenia without connectomics is like studying infectious diseases without the microscope. Seeing the microbes that cause disease is not by itself a cure, but it accelerates research toward one. Similarly, finding a neuropathology that is truly distinctive of a mental disorder is not a cure by itself, but it's a step in the right direction.

For the sake of argument, however, let's consider the opposing view. Maybe searching for neuropathology is a waste of time. A genomics enthusiast might say that autism is caused by defective genes, so we should focus on finding them and not waste time with connectomes.

Indeed, the rapid progress of genomics is stunning. When genomic technologies were slow and expensive, researchers focused on a few rare families with a history of many afflicted members. Now it's possible to rapidly screen the genomes of large populations to find abnormalities. Researchers have discovered abnormalities in many different genes associated with autism and schizophrenia. This is exciting progress, but there are also limitations.

Genomics can predict with high confidence that a child who is born with certain genetic defects will develop autism or schizophrenia. But it cannot predict the vast majority of cases, because no single known defect can account for more than 1 or 2 percent of all cases, and most account for far fewer. In this sense, genomics is currently ineffective for predicting autism or schizophrenia
in individuals, much as neo-phrenology cannot predict the IQ of individuals.

Genetic testing is much more successful at predicting Huntington's disease (HD), a neurodegenerative disorder that typically strikes in middle age. HD begins with random involuntary, jerky movements and eventually progresses to cognitive decline and dementia. Because only one gene is involved, HD is much simpler to predict than autism. An abnormal version of the gene can be detected by a highly accurate DNA test. A positive result means that the individual will develop HD,
and a negative result means that the individual will not.

Understanding the genetics of autism and schizophrenia
is much trickier, given that so many genes are involved. One way forward is to say that autism is actually composed of a large number of autisms, each one caused by a different defective gene. We could study each autism independently and develop a different treatment for each one. This strategy is being pursued by many researchers now, and I expect it will be the most successful one in the short run. But in the long run a complementary strategy will also be fruitful. It may be the case that diverse genetic defects all produce the same neuropathology. I believe we should focus on identifying that neuropathology and treating it.

A genomics enthusiast might argue that treating the neuropathology is not the right approach, because it doesn't strike at the cause. If defective genes cause mental disorders, we should use gene therapy to replace the bad copy of the gene with a good copy. Researchers have experimented with this strategy by engineering animals with genetic defects that lead to brain disorders. In some cases, they have had remarkable success in treating adult animals by correcting the genetic defect.
Such research could eventually lead to therapies for human patients. But this strategy may not always work, or may be only partially successful. If the genetic defect primarily disrupts brain function in the present, then correcting it should solve the problem. But if the defect did most of its damage in the past by altering brain development, correcting it now may not be as helpful.

An analogy may clarify the issue. Imagine that you're suffering from depression because your marriage is breaking up. You go to an old-fashioned psychoanalyst for help, and you're told that your problems spring from the bad relationship you had with your mother when you were growing up. That may be true, but does this insight really help you fix the problem? Now that you're all grown up, replacing your mother with an adoptive mother would have little effect.

Saying that mental disorders are caused by defective genes is the modern way of blaming one's parents: It's not obvious how to use this historical explanation as a basis for treatment. Gene therapy on an adult with a brain that failed to develop normally might be as ineffective as replacing an adult's mother.

Now suppose that a mental disorder is caused by a connectopathy. A true cure requires correcting the abnormal connectivity. So the obvious question is: How much can we change our connectomes, and what is the best way to do it?

In the game of life, you are dealt genes. You can't change your genome; it's the hand you must play. The genomic worldview is pessimistic, constrained on all sides. In contrast, your connectome changes throughout life, and you have some control over that process. The connectome bears an optimistic message of possibility and potential. Or does it? How much can we really change ourselves?