Connectome

by Sebastian Seung

  The bullet metaphor illustrates two important principles that apply to all medical treatments, not only drugs. First, there should be a specific target, and second, the ideal intervention should selectively affect only that target—that is, avoid “side effects.” These principles aren’t upheld by our remedies for brain disorders, which remain distressingly primitive. The surgeon’s knife seems hopelessly crude for altering the brain’s intricate structure, yet sometimes there is no other way. You’ve heard that neurosurgeons treat severe cases of epilepsy by removing the part of the brain where the seizures originate. But overzealous surgery can lead to catastrophe, as you saw in the case of H.M. To minimize side effects, it’s important to target as small a region as possible.

  Epilepsy surgery simply removes neurons from a connectome. Other procedures are intended to break the wires of neurons without killing them. In the first half of the twentieth century, surgeons attempted to treat psychosis by destroying the white matter connecting the frontal lobe to other parts of the brain. The infamous “frontal lobotomy” was eventually discredited and replaced by antipsychotic drugs. Yet psychosurgery is still practiced today as a last-ditch measure when other therapies fail.

  Before considering other types of interventions, I’d like to step back to imagine the ideal one. I’ve said that certain mental disorders might be caused by connectopathies. If that’s the case, true cures would require establishing normal patterns of connectivity. You might regard this prospect as hopeless if you’re a connectome determinist. But even if you’re more optimistic, you can’t deny that the complexity of the brain’s structure is daunting. Merely seeing connectomes is difficult enough, and repairing them seems even harder. It’s unclear how any of our technologies could be up to the challenge.

  But the brain is naturally endowed with mechanisms for connectome change—reweighting, reconnection, rewiring, and regeneration—that are exquisitely controlled. Since genes and other molecules guide the four R’s, they could serve as targets for drugs. I doubt you’re surprised by the idea of the connectome as the target for medications, given that you’ve been reading this book. But you might wonder whether the idea is consistent with what you know from other sources.

  According to well-known theories dating back to the 1960s, certain mental disorders are caused by surplus or deficiency of neurotransmitter, which explains why they are relieved by drugs that alter neurotransmitter levels. Depression, for example, has been attributed to a dearth of serotonin, which is thought to be corrected by antidepressant medications such as fluoxetine, more commonly known as Prozac. (The drugs are supposed to increase serotonin levels by preventing neurons from sucking the molecule back up after secreting it. Recall that a number of such housekeeping mechanisms exist for keeping neurotransmitters from lingering in the synaptic cleft.)

  But there is a problem with this theory. Fluoxetine affects serotonin levels immediately, yet it lifts mood only after several weeks. What could account for this long delay? According to one speculation, the serotonin boost causes other changes in the brain over the longer term. Perhaps it’s these changes that relieve depression, but what exactly could they be? Neuroscientists have looked for effects of fluoxetine on the four R’s, and found that it increases the creation of new synapses, branches, and neurons in the hippocampus. Moreover, as I mentioned in the discussion of rewiring, fluoxetine restores ocular dominance plasticity in adults, possibly by stimulating cortical rewiring. This doesn’t prove that the drug’s antidepressant effects are caused by connectome change, but it has certainly opened the minds of neuroscientists to the idea.

  In this chapter I will focus on the prospect of finding new drugs that specifically target connectomes for the treatment of mental disorders. Let me emphasize, though, that other types of treatment are also important. Drugs may only increase the potential for change. To actually bring about positive changes, drugs could be supplemented by training regimens that correct behaviors and thinking. This combination could direct the four R’s to reshape connectomes for the better. In my opinion, the best way to change the brain is to help it change itself.

  ***

  There’s no doubt that drugs have greatly advanced the treatment of mental disorders. Antipsychotics treat the most dramatic symptoms of schizophrenia, the delusions and hallucinations. Antidepressants can enable the suicidal to lead normal lives. But current drugs have limitations. Can we find new ones that are even more effective?

  Our most successful drugs are for infectious diseases. An antibiotic like penicillin cures infections, killing bacteria by punching holes in their outer membranes. A vaccine consists of molecules that make the immune system more vigilant against a bacteria or virus. In short, an antibiotic corrects infection, while a vaccine prevents it.

  These two strategies also apply to brain disorders. Let’s consider prevention first. During a stroke, most neurons remain alive but damaged, and only later do they degenerate and die. Neuroscientists are working to find “neuroprotective” drugs that would minimize damage to neurons right after a stroke and thereby prevent death later on. The same strategy extends to diseases that destroy neurons for no apparent reason. For example, no one knows for sure why dopamine-secreting neurons degenerate and die in Parkinson’s disease. Researchers hypothesize that the neurons are under some sort of stress, and would like to develop drugs that reduce it.

  Some cases of Parkinson’s disease are caused by defects in a gene that encodes a protein called parkin. An obvious therapy would be to replace the faulty gene. Researchers are attempting to do that by packaging a correct version inside a virus and injecting it into the brain, where they hope the virus will infect the dopamine-secreting neurons and protect them from degeneration. This “gene therapy” for Parkinson’s has been tried in rats and monkeys so far, but not yet in humans.

  Death is just the last step in the degeneration of a neuron, which is generally a long drawn-out process. You might compare it to the slow decline of a person who starts out weak and is then hit by a cascading progression of ailments, each worse than the last. To find clues, researchers look carefully at the various stages of degeneration in neurons, much as physicians observe the progression of symptoms in diseased patients.

  Such observations are helpful because they narrow the search for molecular causes, the potential targets for neuroprotective drugs. In addition, they pinpoint the very first steps of degeneration. The timing is critical; intervening at the outset is likely to be more effective at preventing cell death later on. Early intervention is also important for treating cognitive impairments, which often emerge long before significant neuron death. These symptoms may occur because connections are lost well before neurons actually die.

  In general, it’s important to see degeneration more clearly, and to see it at its earliest stage. The images acquired by the tools of connectomics will help us do that. Serial electron microscopy will reveal exactly how a neuron deteriorates. We will also obtain more precise information about which neuron types are affected and when. All this is bound to be helpful in the search for ways to prevent neurodegeneration.

  Can we also find ways to prevent neurodevelopmental disorders? To do this, we must diagnose them as early as possible, before development has veered too far off course. Even while the fetus is still in the womb, genetic tests can be performed to predict whether disorders such as autism and schizophrenia are likely to emerge later on. But accurate predictions may require combining genetic testing with examination of the brain.

  I argued earlier that microscopy of dead brains, with its high spatial resolution, will be necessary for determining whether a brain disorder is caused by a connectopathy. That method might yield good science, but by itself it will be useless for medical diagnosis. That being said, once a connectopathy has been fully characterized by microscopy of dead brains, it should become easier to use diffusion MRI to diagnose it in living brains. In general, it’s easier to detect something if you know exactly what you are looking for.
<
br />   Behavioral signs will also be informative for some disorders. Some schizophrenics exhibit mild behavioral symptoms when they are children, before the first onset of true psychosis. Perhaps careful detection of such early symptoms, combined with genetic testing and brain imaging, could accurately predict schizophrenia.

  Early diagnosis of neurodevelopmental disorders will pave the way for prevention. Connectomics will help us identify exactly which processes of brain development are involved, making it easier for us to develop drugs or gene therapies that prevent connectopathies or other abnormalities from developing.

  The goal of prevention seems ambitious enough; it’s even more challenging to repair the brain when the damage has already been done. After injury or degeneration has caused neuron death, is there any recourse? A pessimistic answer comes from regeneration denial, one flavor of connectome determinism. Since it’s generally true that no new neurons are added in adulthood, the brain has limited power to heal itself after injury. Is there any way to overcome this?

  Other animal species, such as lizards, are able to regenerate large parts of their nervous systems after injury. And human children regenerate better than adults do. In the 1970s, when physicians realized that children’s fingertips regenerate like lizards’ tails, they stopped attempting to reattach severed fingertips through surgery; now, they simply let the fingertips grow back. Hidden powers of regeneration might lie dormant in adults, and the new field of regenerative medicine seeks to awaken them.

  Injury naturally activates regenerative processes in the adult brain. A main site of neuron creation is known as the subventricular zone. Immature neurons, known as neuroblasts, normally migrate from there to the olfactory bulb, a brain structure dedicated to smell. Stroke increases the creation of neuroblasts and can divert them from the bulb to the injured brain region. Since this natural process might contribute to recovery after stroke, some researchers are trying to develop artificial means to promote it.

  Another route to regeneration is to transplant new neurons directly into the damaged region. This might work better than trying to promote migration from a distant location like the subventricular zone. Parkinson’s disease, as I’ve mentioned, involves the death of dopamine-secreting neurons. Researchers have attempted to replace them by transplanting healthy neurons from fetuses. Amazingly, some neurons were shown to survive in recipients’ brains for over a decade, although it’s unclear whether the transplants actually did much to alleviate the symptoms. The experiments, conducted with cells isolated from aborted fetuses, raised thorny ethical issues. A further complication of transplantation was that patients’ immune systems could reject the new cells as foreign.

  We can now avoid both of these problems, thanks to a recent advance that allows the culturing of new neurons customized to a particular patient. A skin cell can be “deprogrammed” to become a “stem cell,” one that has effectively “forgotten” its former life as a skin cell. Owing to its newly ambiguous identity, this stem cell can now be “reprogrammed” to divide and produce neurons in vitro. (The Latin term in vitro, which means “in glass,” refers to the artificial environment used for culturing molecules, cells, or tissues isolated from an organism. At first that environment was typically a glass container, but plastic ones are more common now.) Researchers have used this method to create dopamine-secreting neurons from the skin cells of Parkinson’s patients. They are planning to transplant the neurons back into the patients’ brains to treat them.

  Whether created naturally or added by transplantation, most new neurons die. Without “taking root,” new neurons presumably cannot survive. Regenerative treatments will thus require enhancing the integration of new neurons into the connectome, a process that depends on promoting the other three R’s—rewiring, reconnection, and reweighting.

  The adult brain may hold untapped potential for making these changes. Earlier I referred to the fact that most recovery happens during the three-month period just after stroke. According to one speculation, this is a critical period, analogous to the one during brain development, with production of similar molecules that promote plasticity. Once this window closes, plasticity plummets and the rate of recovery slows. Perhaps stroke therapies should aim at keeping the window open, extending the natural processes of recovery.

  As we’ve seen, rewiring may be difficult in the adult brain. After injury, though, neurons appear to grow new axonal branches more easily. If researchers can identify the molecular reasons why, it may be possible to promote rewiring of the adult brain by artificial means, which would help integrate new neurons into the brain as well as allow existing neurons to change their functions. Similarly, since creation of new synapses happens at a greater rate in the injured brain, there may be natural molecular processes that could be manipulated to promote reconnection.

  Could we also correct neurodevelopmental disorders, fixing the brain after it has wired up improperly? If you’re a connectome determinist, you’d probably regard correction as futile and instead focus all your efforts on prevention. But it’s not clear whether completely accurate and early diagnosis of neurodevelopmental disorders will be possible, so we have no choice but to think about correction too. This will require the most extensive connectome changes of all, and therefore the most advanced control of the four R’s.

  I’ve stressed the treatment of malfunctioning brains, since these are the connectomes most in need of change, but people also want drugs for enhancing normal brain function. Many university students drink coffee while studying. While caffeine may help them stay awake, it has little effect on learning and memory. Nicotine improves the mental abilities of smokers, but that’s only relative to their substandard performance when deprived of cigarettes. Can we find more effective drugs than these? For example, we’d really like a drug that promotes the connectome changes necessary for learning or remembering new information or skills. Also useful would be drugs to help us forget. Perhaps these could promote the elimination of cell assemblies or synaptic chains formed after traumatic events, or those implicated in bad habits or addictions.

  We have a long wish list for drugs, both for preventing brain disorders and for correcting them. Unfortunately, the pace of discovery is slow. New drugs appear on the market every year, often with great fanfare, but many are not really new; they’re just variants of old drugs, and unlikely to be significantly more effective. Most antipsychotics and antidepressants are variants of drugs discovered by accident over half a century ago. Few drugs are truly new; few draw on recent advances in neuroscience.

  The challenges of drug development are not unique to mental disorders, of course. Creating new pharmaceuticals is a hugely risky business. It can take many years to develop candidate drugs. Only those deemed most likely to succeed are tested in human patients, yet nine out of ten fail in this last stage, turning out to be toxic or ineffective. This is a huge waste of money, given that clinical trials incur a significant fraction of the investment required to bring a new drug to the marketplace. (Total cost estimates range from one hundred million to a billion dollars. ) Everyone desperately wants better drugs—those who suffer from diseases, those who treat them, and those who invest huge sums of money trying to develop therapies. How can drug discovery be accelerated?

  Historically, most drugs have been discovered by chance. The first antipsychotic was chlorpromazine, known in the United States by the trade name Thorazine. This belongs to the phenothiazine class of molecules, the earliest of which were originally synthesized in the nineteenth century by chemists attempting to create dyes for the textile industry. In 1891 Paul Ehrlich discovered that one of them could be used to treat malaria. During World War II, the French pharmaceutical company Rhône-Poulenc (a forerunner of today’s Sanofi-Aventis) tested many phenothiazines looking for more malaria drugs; when they failed to find any effective ones, they started looking for antihistamines. (You may have taken medications of this type for allergies.) Then a physician discovered that phenothiazines could enhance
the actions of surgical anesthetics. Rhône-Poulenc researchers switched to testing for this new application, and discovered that chlorpromazine was effective. After giving the drug to psychiatric patients as a sedative, doctors realized that it specifically reduced symptoms of psychosis. By the end of the 1950s, chlorpromazine had swept through the psychiatric hospitals of the world.

  The first antidepressant medications, iproniazid and imipramine, were discovered around the same time, in stories with similar twists and turns. Iproniazid was originally developed for tuberculosis, but had the unexpected side effect of making patients unreasonably happy. Psychiatrists eventually realized that it could be used to treat victims of depression. Meanwhile, the Swiss company J. R. Geigy (an ancestor of Novartis), having heard about Rhône-Poulenc’s success with chlorpromazine, decided to play catch-up by looking for an antipsychotic of their own. They tried testing imipramine, which chemists had synthesized by modifying a phenothiazine. It was a failure for treating psychosis, but fortunately it turned out to relieve depression.

  So researchers did not intend to develop the first antipsychotics and antidepressants. They were just lucky and alert enough to stumble upon them during this golden age of the 1950s. More recently, there has been growing excitement about “rational” methods of drug discovery built upon our modern understanding of biology and neuroscience. How do these methods work?

  Recall that cells are composed of a huge variety of biological molecules, which are involved in many kinds of life’s processes. (Earlier I talked about the important class of biomolecules known as proteins, which are synthesized based on blueprints encoded in genes.) A drug is an artificial molecule that interacts with the natural ones in cells. Ideally, according to the magic-bullet principle, the drug should interact with a specific type of biomolecule but not with other types.