This article was originally published on the Scientific American website on March 10, 2016.
Alexia had been in-and-out of intensive psychiatric therapy for nearly two decades by the time we met. She suffered from bipolar disorder, which meant that she cycled between explosions of boundless energy and black holes of suicidal despair. Despair brought her to our unit.
Her long chart chronicled how previous psychiatrists had emptied the armory: antidepressants, antipsychotics, anticonvulsants, mood stabilizers, group and intensive inpatient therapy, psychotherapy, dialectic and cognitive behavioral therapy. Nothing had a lasting effect.
What struck me was the shotgun approach: try everything. Her medications spanned the molecular gamut: some stopped the disposal of the neurotransmitter serotonin, allowing more to be present in the brain; some focused on norepinephrine; others blocked the action of dopamine; yet others had an unknown target but had proven helpful to some patients. The imprecise approach to treating this most sophisticated of organs, the brain, seemed odd.
I was a third year medical student at the time and was asked to speak with her, design a treatment plan, and present it to the team later in the day.
We met in a small room with Ikea furniture. Alexia wore a loose black t-shirt and grey sweat pants with faded pink hospital socks—the ones with traction padding. I was stiff with naiveté. She seemed entirely limp. She moved and spoke slowly, as if each whisper betrayed some tragic secret.
She shared a story consistent with what I’d read in her chart. She hadn’t been affected by the profound forgetfulness characteristic of severe depression. To my surprise, beneath her cloud of depression, she was sharp and lucid and witty. An intelligent woman who enjoyed literature. She painted.
It was clear she suffered deeply. Her pain, torment incarnate. But beyond literary descriptors, what was going on in her brain and how could we treat it was unclear.
After we spoke, I presented her case to my team and asked about her litany of medications. They explained that trial-and-error approaches to medication—based on guidelines from large epidemiological studies—were standard care for patients with bipolar disorder and depression.
Alexia had reported the “signs and symptoms” of depression to her clinician and so had been put on medications that had been successful in patients with similar symptoms. Diagnosis and treatment based on observable signs and symptoms—as opposed to an underlying cause—is known as a descriptive classification.
Descriptive classifications have long dominated psychiatry—in large part because they were developed before we understood the underlying causes, or etiology, of those signs and symptoms.
“Depression” reveals as much about the brain as “jaundice” does about the liver: something’s wrong. Neither term is actually a disease, but rather a sign or symptom of an underlying disease process. Like jaundice, depression has many causes which are not captured by the symptom “depression.”
Scottish psychiatrist Thomas Clouston (1840-1915) classified symptoms into three main groups: melancholia, mania, and paranoia. This classification became widely popular and used by the community. Emil Kraepelin (1856-1926), a German psychiatrist, extended this classification to include clinical course and patient outcome, yet he didn’t resolve the question of etiology: why does mental illness occur?
Freud, like many others of his time, attempted to explain mental illness in terms of the brain’s pathophysiology. But, because so little was known about the brain, he was forced to develop what he called “metaphorical expressions” like the id and ego to explain the causes of mental illness. Although his system proved helpful, Freud lamented:
The deficiencies in our description would probably vanish if we were already in a position to replace the psychological terms with physiological or chemical ones…We may expect [physiology and chemistry] to give the most surprising information and we cannot guess what answers it will return in a few dozen years of questions we have put to it. They may be of a kind that will blow away the whole of our artificial structure of hypotheses. (Freud, Beyond the pleasure principle, 1920)
But why bother coming up with an “artificial structure” at all? Patients needed help.
Imagine it’s 1920 and Alexia presents to your clinic with suicidal thoughts. What do you do? There are no chemical or molecular therapies—neurotransmitters wouldn’t be discovered until 1921—or useful knowledge of brain pathways. So you treat her using a linguistic tool: metaphor.
Pharmacologic therapies developed in the post-War era presented new tools to treat patients with mental illness. By targeting the function of a particular molecule, they also informed new etiologic theories: for example, increasing the brain’s available serotonin with a Selective Serotonin Reuptake Inhibitor (SSRI) helps resolve the symptom of depression and suggests that depression is caused by inadequate serotonin. However, patients who weren’t helped by an SSRI could sometimes helped by a Selective Norepinephrine Reuptake Inhibitor (SNRI), suggesting more than one cause for depression.
To understand the underlying causes of depression, neuroscientists turned to genetics and brain imaging. The goal is to find a common marker—whether a gene or measure of brain structure or function—that indicates the presence of a mental illness.
One of the great difficulties of this endeavor has been looking for common markers within diagnoses based in descriptive classifications.
Joel Gelernter, Director of Yale University’s Division of Human Genetics in Psychiatry, recalls that, “about 20-25 years ago, people would do studies of psychiatric traits and genetics with 50-100 samples. Around this time, people were looking for the gene for schizophrenia, or the gene for bipolar disorder, which is viewed nowadays as a pretty ridiculous idea. Of course, we didn’t know it then because we hadn’t done the experiments.”
The problem was that one laboratory would discover gene X as a marker for bipolar disorder, another laboratory would discover gene Y, and neither could reproduce the work of the other. Eventually, these failed experiments shifted the search from a single, blatantly abnormal gene to multiple, slightly abnormal genes.
They also forced the community to re-evaluate how mental illnesses are classified and diagnosed.
Studies have shown that depression involves the amygdala, a brain region that processes emotional information. The amygdala uses the neurotransmitter serotonin to coordinate emotion processing with other brain regions that regulate attention, cue memory, form judgments, and perceive sensory information like happy or sad faces.
Consider four patients, each with amygdala dysfunction, each with a different cause: the first doesn’t produce enough serotonin; the second doesn’t have enough serotonin receptors; the third lacks neural connections from the amygdala to other brain areas; the fourth has a completely separate brain region outside the amygdala (perhaps involving norepinephrine?) that inappropriately inhibits amygdala function. Yet they are all “depressed.”
Patrick Sullivan, who heads the Psychiatric Genomics Consortium (PGC, more on this below) wrote, “the fact that disease definition in psychiatry is descriptive poses particular problems. The diagnostic process relies heavily on signs and symptoms without recourse to biological means of distinguishing affected from unaffected individuals.”
Geneticists have begun to target “types/flavors” of depression that run in one family and are, therefore, more likely to result from a similar cause. Because family members are more genetically similar than the general population you can more accurately compare how and where the genes of depressed and not-depressed family members differ. This is known a “linkage analysis.” Linkage analyses provide candidate genes that can be evaluated in larger samples using a genome-wide association study.
To perform a genome-wide association study Gelernter explains you “gather DNA from cases and controls, sift through the entire genome, and identify the differences. When all other sources of differences between the samples are accounted for, the distinctions that are left must account for the genetic difference between samples.”
But genome-wide association studies presented an unexpected statistical problem: while a single, blatantly abnormal gene could be detected with 50-100 samples, multiple, slightly abnormal genes required about 1,000 times more samples. The problem existed because small individual differences in genetic structure occur naturally. Detecting abnormalities related specifically to bipolar disorder requires you to filter out these individual differences, which requires many more D.N.A. samples.
John Krystal, Chair of Yale University’s Department of Psychiatry and editor of the preeminent journal Biological Psychiatry, told me that, “by the late 1990’s, the statistical geneticists had predicted that the required sample sizes to detect gene variants…approached 100,000 patients and 100,000 controls.”
Such a group of samples was too large for one scientist or one laboratory to feasibly gather—there simply wasn’t time or money. To succeed, statisticians essentially told the field that they needed to restructure their entire enterprise and work together to pool their resources. Not something the field wanted to hear.
While the National Institute of Mental Health (NIMH) offered funding incentives that no-doubt catalyzed the shift to team science, Krystal explained “the lack of replicability forced the field to conform to the predictions of the statistical geneticists.”
In 2014, the Psychiatric Genomics Consortium (PGC), one of the most productive ventures in team science, published a report in the leading journal Nature that identified 108 genes associated with schizophrenia risk. Krystal explains, “the sample size necessary to do this was over 37,000 patients and almost 120,000 controls…[a] magnitude that would never have been accomplished with one site working by itself. It was only possible through these very large scale collaborations.”
Yet 108 genes, each related to a similar set of symptoms? What if these different genetic abnormalities caused similar downstream abnormalities in genetic expression? To answer this question, the PGC developed a statistical method for mapping the pathways from genes to genetic expression called “pathway analysis.”
Their pathway analysis, published in Nature Neuroscience, analyzed D.N.A. from over 60,000 participants and focused on schizophrenia, major depressive disorder, bipolar disorder, autism-spectrum disorder and attention deficit-hyperactivity disorder. What they found was surprisingly simple: certain biological “themes” spanned certain disorders. Bipolar disorder was more associated with genetic pathways that control histone methylation, the primary mechanism whereby D.N.A. is turned on and off.
Theoretically, genetic pathways could provide a larger, better drug target than a specific gene. Abnormal genetic pathways could also be more consistent across patients than one of, say, 108 abnormal genes, which makes for a more useful marker.
In addition to helping create new therapies, markers could guide clinicians towards existing therapies that are most likely to help a specific patient.
Gelernter wrote in a recent Biological Psychiatry article, “Drug response to phenotypes are likely to be much less complex than psychiatric diagnosis phenotypes—in some cases, a great deal of clinical variation hinges on a single genetic variant, which could be the point of interaction between the drug and a particular biological target.”
For example, one study showed that in patients with bipolar disorder, variation in the gene GADL1 had a strong effect on an individual’s response to lithium, one of the drugs Alexia had been trialed on. While we didn’t test her for this gene, I wondered whether a genetic test for a panel of medications would have shortened Alexia’s nearly two decades of trial and error.
After four weeks on our service, Alexia improved. Her mood stabilized with a combination of lamotrigine and cognitive behavioral therapy, a mix of mechanism and metaphor. Her thoughts of suicide cleared and her literary inclinations switched from Sylvia Plath to Zen poetry.
We had succeeded in breaking her bipolar cycle but we wondered how long this would last and when her next crisis would be. Our team agonized over how best to follow-up her treatment: try different therapies and therapists? Daily or weekly? New drugs?
The scope of these questions underlined how little we knew about what was going on in Alexia’s brain. Had we known more, we could have helped more.