Here Is a Human Being (34 page)

That powerful susceptibility genes couldn’t be found also explained why so many of the risk numbers reported by the consumer genomics companies needed to be taken with mammoth grains of salt. 23andMe, for example, typed customers for just two SNPs conferring susceptibility to multiple sclerosis.
¹⁰
SNPedia, meanwhile, listed twenty-eight of them, all with modest effects.
^*
11
This was the legacy of GWAS: for complex diseases like MS, which are determined by many genes and the environment, any single genetic risk factor was likely to be extraordinarily weak.
¹²
And even if one had multiple genetic risk factors, as I did for MS, those factors interacted in ways that we were still a long way from understanding well enough to make predictions, let alone help guide treatment.

David is athletic but neither tall nor imposing. He is thin and sometimes socially awkward, almost shy, often looking down with hands in the pockets of his jeans as he makes his way down the hall to a lab meeting, to get a coffee, or to catch a flight to Johannesburg or Taipei. He wears wire-rimmed glasses beneath a mop of thick, unruly hair. His lab is remarkably productive but fairly small given that it oversees an eight-figure annual budget. They are a close-knit bunch, meeting regularly for beer, jogging, meals, trips to the beach, etc. David would not have it any other way. He is free with his money. He rides a motorcycle and smokes cigars. He appreciates a funny joke, a well-told story, a nice bottle of wine, a good song. He enjoys life … as hopeless and futile as it may be.

David and a few other like-minded people had come to believe that there had to be a better way find the elusive risk factors for inherited traits that GWAS had failed to uncover.
¹³
,
¹⁴
If, as they suspected, these factors were rare, then they would not be found on the “SNP chips” used in GWAS. So how to find them?

Sequence.
¹⁵

As they continued to chase the genetic basis of infectious disease, David and a group of collaborators amassed samples from a cohort of fifty hemophilia patients who had been exposed to HIV-infected blood but remarkably had not gotten infected. Why not? David assumed there must be one or more genetic variants that made them resistant to the virus. Those variants, he reasoned, would be found by sequencing. Kevin’s lab would sequence these fifty people; David’s people would analyze the data (when this book went to press, the hemophilia project was ongoing). In the process, Kevin, an Illumina partisan since buying his lab’s first Genome Analyzer in 2007, would put all of the company’s new toys through their paces.
¹⁶

I saw an opportunity.

“I think you should sequence me,” I said to Kevin. “I am already going to be sequenced and phenotyped out the wazoo.” (I certainly hoped that that was true.) “My information will be completely public, so there will be zero IRB issues. You won’t need to establish a cell line because I’m right down the hall—if you need more blood you know where I live. I am the ultimate control sample … Well, unless of course you’re studying anxiety, metabolic syndrome, or nearsightedness—then I’m a case.”

“Let me talk to David,” Kevin said. “It’s a question of money.”

I pressed the “ultimate control” argument and David began to weaken. “Maybe if you were up for having more of your tissues sampled,” he said. “I’ll give some thought to it. If I were you I would try to talk to me about it some time when I’m not sober.”
¹⁷

“Done and done,” I said.

It began with a five-minute blood draw in Clinical Research Coordinator Kristen Linney’s office followed by a routine chemical extraction of DNA from my white blood cells. But after that my sample sat for a while. Why? Because the Institutional Review Board at Duke (upon which I happen to serve) didn’t quite know what to do with me: most IRBs have not made provisions for returning results to research subjects, even lunatics like me who’ve
already
seen much of that same data and put it on the Internet for the whole world to examine. It’s a situation that IRBs have not ever had to think about—until now. Over the next week the request to have me sequenced and my data returned to me and the world would travel from Kristen’s office to the senior chair of the Duke IRB, to the associate dean of research support services, to the director of Duke’s Center for Bioethics and Humanities, and then back to the associate dean of research support services, then on to the dean of the medical school and the vice dean for research, then back to the IRB chair, back to Kristen, and finally, back to David, Kevin, and me.
¹⁸

Once we had institutional approval, I assumed we would be off and running. But around that time, the Shianna lab’s sequencing runs began to fail at an alarming rate: nearly one in two weren’t working. The fluorescent signals emitted by each sequenced base were decaying too rapidly; thus the early cycles could be used from each run but the later ones could not; they were too faint to allow the computer to distinguish one base from another. Within a few days it became clear that it wasn’t the machines, but rather the chemicals. And it was not just the Shianna lab. Genome Analyzers all over the world were failing. Illumina, which was forced to eat millions of dollars, told customers it was no longer shipping reagents until it could isolate the problem.
¹⁹

In the meantime, I went back and scrutinized my SNPedia data. Was there anything interesting there that I had missed? I didn’t think so. I was seven times more likely to go bald than most men … duh. I needed only to look in the mirror to confirm that. I had plenty of risk factors for coronary artery disease, type 2 diabetes, stroke, obesity … again, duh. I had had a grandparent on each side die before age sixty from a heart attack. My dad had had a quadruple bypass when he was sixty.

I opened the SNPedia “Medicines” menu with a bit more optimism: as David and others had shown, pharmacogenomics was one of the most promising places for personal genomics to make a difference. Esther Dyson told me that perhaps the most useful thing she had learned from 23andMe was her family’s sensitivity to the blood thinner warfarin.
²⁰
As I witnessed firsthand when my nephew Noam developed deep-vein thrombosis (see chapter 10), warfarin is a tricky customer: not enough of it and you might suffer a debilitating blood clot. Too much and you might bleed to death.
²¹
But unlike warfarin, most of the known drug response markers we knew about did not usually alter one’s sensitivity to a clinically meaningful extent.

An exception was a series of markers in the ABCB1 gene. ABCB1 is expressed most strongly in tissues that serve either as barriers (the blood-brain barrier, the placenta) or are involved in eliminating waste from the body (kidney, liver, intestines). One can imagine how this gene might impede drug response: by raising a molecular “gate” or by causing the drug to be flushed from the system more rapidly.
²²
But as a team of German researchers found, some versions of ABCB1 genotypes do just the opposite: they
enable
drug response.
²³
Perhaps the gene could also lower a gate or keep a drug in the system for an extended period. In any event, most Caucasians, including me, were significantly less likely to respond to certain antidepressants because of their common ABCB1 alleles. I had been on ten milligrams of Lexapro (escitalopram) since 2008 and, as my wife could tell you, it had drastically changed my life for the better, mostly by reducing my anxiety level and keeping my frequent companion, a deep and abiding sense of impending doom, at bay. Yet my ABCB1 alleles would betray me, no?

I went back to SNPedia and looked at the four other non-ABCB1 SNPs that had been associated with depression and for which I had been typed. They were a mixed bag. One was associated with a poorer response to clomipramine,
²⁴
another antidepressant but of a different chemical class than Lexapro, that is, probably irrelevant to me. Another was associated with major depression in Mexican Americans.
²⁵
But 97 percent of Europeans, including me, did not carry the risk allele … okay, next. Another was associated with suicidal thoughts in people taking Lexapro’s predecessor, citalopram (Celexa).
²⁶
Despite having the “normal” allele, I admit that in my darkest moments I had had such thoughts.

The last “depression SNP,” found in a gene that encoded brain-derived neurotrophic factor, was arguably the most interesting. BDNF is an important gene: it’s expressed like gangbusters in our brains and is crucial to the way our neurons grow and develop. Knock out both copies in mice and they’ll have problems with coordination, balance, hearing, taste, and breathing … and they’ll die soon after birth.
²⁷
Knock out just one copy and they’ll have problems learning and remembering.
²⁸
Like 28 percent of Europeans, I carry one copy of a variant in BDNF that changes a valine to a methionine at position 66 of the protein. Val66Met, as it’s known in mutation parlance, has been well studied; however, being well studied doesn’t always equate to real understanding. It appears by some accounts to be modestly associated with bipolar disorder,
²⁹
though BDNF has been touted as a “depression gene” in humans for many years despite inconsistent evidence and frequent failures to replicate. Some psychiatric geneticists think it’s time to move on.
³⁰
But we know that Val66Met has functional consequences. People with the valine allele exhibit higher BDNF activity; they tend to perform better on memory tasks. And the anatomy of their brains is different than people who are Met/Met.
³¹
As I wrote this, pharmacogenomic data were scanty, but some studies suggested that people with lower levels of BDNF responded better to antidepressants.
³²

Without knowing my BDNF levels, I knew this was all speculation. But as far as BDNF being a common “depression gene,” David Goldstein was having none of it. “Until fairly recently, psychiatric geneticists would start out believing in a gene because a role for it made such good sense, and then they would go out and find themselves a polymorphism in the gene that showed an association with
something:
a cognitive trait, an imaging trait, or with some disease of interest. That the association statistics were weak was largely judged acceptable because it all made such ‘good sense.’ Of the many famous genes implicated in this era, none survived into the genome-wide phase when researchers cleaned up their acts and insisted upon consistent standards of evidence. Even career-creating ‘discoveries’ like polymorphisms in BDNF associated with cognitive performance have failed to find support in the much better powered and better controlled genome-wide studies using large samples.” Both he and our colleague Anna Need pointed out, however, that truly
rare
variants in BDNF might yet turn out to be responsible for some subset of psychiatric illness.
³³

This never-ending exploration of my own cellular White Pages was giving me a headache. I turned my attention back to sequencing. Eventually a millionth of a gram of my DNA would make its way into Kevin’s lab and into the pipeline, unstable reagents be damned. Kevin warned me that because of the reagent problems, there was a fair chance my sequencing runs would fail; we would know in the next day or so.

Inside the large windowless room that housed the sequencers, the hum was constant. Internal fans kept the machines cool. Every piece of equipment seemed to have its own computer. The first step was for my DNA to be fractionated by a vibrating machine called a nebulizer into super tiny bits some 200–400 base pairs long—the DNA needed to be small before it could be amplified en masse in a later step. If the pieces were too big, they would take up too much room on the flow cell, the small glass slide where the business of DNA sequencing would get done. The ends of the DNA were “polished” enzymatically to make them uniform, and two unique adapters (short stretches of synthetic DNA) were attached to the ends with another enzyme (ligase) to the fragments. The adapters would allow the enzyme to recognize every fragment equally and sequence it. The fragments were then amplified by PCR (see chapters 5 and 9) to make sure there were sufficient ligated fragments for the actual sequencing reaction still to come. Once this step was done, the sample preparation steps were over: a “library” of my genome—millions of fragments with adapters attached—was ready. “Everything to this point is pretty straightforward,” Kevin said. “Illumina produces a kit for these steps and it’s basically idiotproof.”
³⁴

The surface of the flow cell was a tiny grid the size of a microscope slide: eight lanes and one hundred microscopic tiles (think ceiling tiles) per lane. The goal was to get about two hundred thousand amplified “clusters” of DNA per tile. But we’re getting ahead of ourselves. The next step was to coat the surface of the flow cell with a dense lawn of primers (again, short stretches of single-stranded DNA) that were “flapping in the breeze.” These were complementary to the adapters attached to the fragments of my DNA: they would recognize each other and stick together, or hybridize. They would do exactly what their name suggested: prime the reaction by giving the enzyme a known sequence to start with. But first we separated the strands of my DNA with sodium hydroxide; making the clusters single-stranded meant they would bind to the primers on the flow cell. Once the single-stranded, adapter-ligated fragments were bound to the primers on the flow cell, they could be amplified again to scale up the reaction. This step was called bridge amplification because each end of a ligated fragment would form the middle of a bridge with one of the primers on the surface of the flow cell. At the end of this process, which was done on a separate bench away from the sequencing machines at a so-called cluster station, each fragment was amplified about a thousandfold. The DNA was also double-stranded for the moment; DNA is more stable in this state and can be kept in the fridge for a couple of months.
³⁵