Метагеномное секвенирование выявило редкие вирусы, вызывающие вспышки заболеваний

In 2019, Nichol Cano, an otherwise healthy mother of three in Fresno, California, began experiencing worsening headaches. She had no history of headaches; they had seemed to start after a liposuction surgery in Tijuana, Mexico. She saw her primary-care doctor and then a neurologist, who diagnosed her as having migraines. She was given a standard migraine treatment—an injection and a pill—but her pain only intensified. “It felt like my skull was being crushed and my eyes were going to pop out of my head,” she told me. She went to a local E.R., which again treated her for migraines and sent her home. A few weeks later, Cano’s husband, Javier, was at work, supervising a manufacturing warehouse, when he received an incoherent voice mail from Cano. She was alone at home, disoriented. Eventually, an ambulance came and rushed her to the hospital. This time, E.R. doctors performed a spinal tap: a needle was poked into her back to collect a sample of cerebrospinal fluid. She recalls the doctors telling her that her opening pressure, which reflects the force that the fluid exerts on the brain, was the highest that they had ever seen. The result strongly suggested meningitis—an inflammation around the brain and spinal cord often caused by bacteria, viruses, or other pathogens—yet a battery of tests failed to detect any underlying infection. The doctors speculated that a fungus might be responsible, but they didn’t want to give her potentially toxic antifungal drugs without first confirming their hunch. “They were saying there was nothing more they could do,” Cano said. A doctor on Cano’s team had heard about a new and powerful type of genetic testing called metagenomics, which can search for thousands of different microorganisms in a single sweep. The doctor reached out to Michael Wilson, a leader in the field of metagenomics who works as a neuroinfectious-disease specialist at the University of California, San Francisco. Within a week, Cano was in Wilson’s office, describing her ordeal. He arranged for her to be admitted to U.C.S.F. “We’re going to figure this out, and you’re going to be fine,” he told her. The usual way of diagnosing an uncommon infection could be compared to finding a loose page in a library and trying to determine which book it came from. The page will surely contain clues—maybe the subject matter points to a particular shelf, or the author uses telltale turns of phrase—but sometimes a match can’t be found. There are diagnostic tests for many conditions, but a doctor’s list of potential diagnoses may be long, and the culprit might be too rare to come readily to mind. Metagenomics, in contrast, works more like a search of the entire Google Books database. First, scientists sequence the full range of genetic material in a patient’s sample. Then they compare what they’ve found to a vast data set of known pathogens. Matches can reveal infections that doctors either could not culture in a laboratory or never thought to look for. At U.C.S.F., Cano underwent another spinal tap as Wilson tried to find the cause of her headaches. His colleagues sequenced all of the genetic material that they could find in her cerebrospinal fluid and ran it through their genetic library of pathogens. At last, metagenomics produced a match: Candida parapsilosis, a yeast that typically lives on our skin without causing harm. “She had no business having yeast in her cerebrospinal fluid,” Wilson told me. One item in Cano’s medical history stood out to him: before her liposuction procedure in Tijuana, she’d received an injection of epidural anesthesia near her spinal cord. He now suspects that it was contaminated. When Cano finally learned what was making her sick, she broke down in tears. Treatment for the fungus was gruelling: she was put on two potent antifungals, one of which caused vivid hallucinations, and she needed a shunt in her head to relieve pressure on her brain. Seven years later, the shunt remains in place, and she only recently stopped taking antifungal medications. Still, she told me, “My body feels stronger, and I know I’m going to be O.K.” Infectious microorganisms have beleaguered humanity for far longer than people have been able to see and understand them. In the late seventeenth century, Antoni van Leeuwenhoek, a self-taught Dutch scientist, used a handmade microscope to observe what he called “animacules”—tiny living organisms—in rainwater. Two centuries later, Louis Pasteur and Robert Koch helped establish germ theory, which holds that many diseases are caused by infectious agents that multiply in the body. Koch devised a method for growing bacteria in a shallow glass dish; Julius Petri improved on it by adding a snug lid. It was now possible for doctors to culture many of the pathogens that make us sick. Microbial cultures were a major advance, but they had limits. The bacterium that causes tuberculosis, for example, grows very slowly, so culturing a sample to diagnose an infection can take weeks. Many other fungi, parasites, and viruses are difficult or impossible to culture. The advent of polymerase chain reaction, or PCR, testing, in the nineteen-eighties, enabled scientists to make numerous copies of the DNA or RNA in a sample, and that dramatically increased the sensitivity of diagnostic tests. But this method still required doctors to know what they were looking for, and to test for each pathogen one at a time. A patient with respiratory symptoms could be PCR-tested for influenza, rhinovirus, and RSV, but, if all of them came back negative, then the underlying cause remained unknown. Then, in 2001, a biochemistry professor at U.C.S.F. named Joe DeRisi worked with several colleagues to develop what they called the ViroChip. Building on research that he had conducted as a graduate student at Stanford, DeRisi programmed a robot to print DNA snippets from every virus that had been fully sequenced onto glass slides. When scientists collected samples from a patient’s viral infection, tagged them with a fluorescent dye, and placed them on the ViroChip, a particular spot on the slide would glow, revealing what kind of virus it was. The chip could even help identify never-before-seen infections. In 2003, DeRisi used the ViroChip to help determine the cause of a mysterious infectious outbreak in East Asia: a novel coronavirus later named SARS-CoV. (It is now called SARS-CoV-1, and the virus that causes COVID-19 is SARS-CoV-2.) He became convinced that metagenomics could transform the science of diagnosis. Wilson joined DeRisi’s lab at U.C.S.F. in 2013. One day, a pediatric-medicine professor at the University of Wisconsin offhandedly told them about an immunocompromised fourteen-year-old boy who had been repeatedly hospitalized with severe meningoencephalitis. Aggressive treatments weren’t working, the colleague said, and conventional testing—including a biopsy of the boy’s brain—had not revealed what was making him sick. Metagenomic testing on his cerebrospinal fluid ultimately found Leptospira, the bacteria that cause leptospirosis. He’d probably acquired it while swimming in Puerto Rico, where he was on a missionary trip. The boy was ultimately cured with intravenous penicillin. After that, the team was inundated with requests from doctors seeking diagnostic help. DeRisi,Wilson, a collaborator named Charles Chiu, and several others co-founded a company called Delve Bio, which specializes in metagenomic testing of cerebrospinal fluid. In May, I visited a glass-and-steel building on the outskirts of Boston that serves as Delve Bio’s East Coast headquarters. A tiny plastic tube of cerebrospinal fluid had just been shipped to the company—from a patient with an unknown infection or inflammatory disorder in their central nervous system. I watched two technicians in gowns and surgical masks place the tube in a centrifuge, a circular machine that shook it violently. (Brad Murray, Delve’s C.E.O. and one of the company’s co-founders, told me that extracting genetic material from any fungi that might be in the sample was “like trying to get the meat out of a walnut.”) When the shaking stopped, the sample was removed and then readied for sequencing. In a sequencer that resembled a large fridge, millions of short nucleic-acid fragments would be identified in the span of ten hours. The data were then uploaded to a cloud-based analysis platform to determine whether any of the snippets came from fully sequenced pathogens. By the following day, metagenomic testing on the sample would reveal a bacterium that can be treated with a new, targeted antibiotic, and which the patient’s doctors hadn’t tried yet. As Murray told me about cases that Delve has helped solve, I was reminded just how diverse infectious diseases can be. A ten-year-old with the same type of shunt as Cano turned out to be infected with a novel species of Moraxella, a bacterium that normally lives in the upper airways. A sixtysomething woman, who was confused and struggling to walk, was found to have neurocysticercosis, a disease in which microscopic tapeworm larvae colonize the brain. I had memorized lists of these pathogens in medical school, but as a physician I rarely encountered them in my clinical practice. They probably wouldn’t have made it onto my list of potential diagnoses. The applications of metagenomics go far beyond medicine. It’s possible to sequence all of the DNA in a spoonful of pond water, for example, to get a sense of which species live there. If you detect invasive algae, you might launch a campaign to clean it up; if you find an endangered frog, you might halt development in the area to save it. DeRisi told me how, a few years ago, there was a large die-off of leopard sharks in the San Francisco Bay. Autopsies on the sharks showed extensive bleeding in their brains—a sign that they’d died from infection. Using metagenomics, DeRisi’s team, in collaboration with the California Fish and Wildlife agency, identified the culprit as Miamiensis avidus, a parasite not previously known to infect leopard sharks. It probably hitched a ride into the bay on runoff from heavy rains. “The beauty of a hypothesis-free approach,” DeRisi told me, “is that, when you’re confronted with the data, you can’t be weighed down by your cognitive biases of what you think is wrong.” Of course, a technique that involves millions of genetic fragments might also create millions of opportunities to be misled. Shangxin Yang, a clinical microbiologist at U.C.L.A. who has studied metagenomics for fifteen years, told me about an immunocompromised patient who was treated with antifungals after fungal DNA turned up in metagenomic blood tests. Only later did doctors realize that the genetic material had probably come from a mushroom that the patient had consumed, and not from a pathogen. “We are starting to see a lot of things that we didn’t expect,” Yang told me. The scale of the sequencing, and of the reference data sets, makes it difficult to detect a reliable signal in the noise. Our bodies are colonized by trillions of cells from hundreds of microbial species, and they may show up in metagenomic tests even if they don’t cause any harm. One large study, published in 2019, found that nearly twenty-three per cent of healthy people had microbial DNA in their blood. Patricia Simner, who directs the clinical-microbiology molecular laboratory at the Mayo Clinic, told me that some genetic libraries of pathogens are painstakingly curated for accuracy, but others are peppered with inaccuracies. In one study of five thousand healthy individuals, for example, researchers thought that they’d found fifty types of bacteria that were significantly more common in males than in females. The finding turned out to be an illusion: the reference sequences for the bacteria erroneously included fragments of the human Y chromosome. “These microbial reference libraries are enormous, and they’re constantly changing,” Simner told me. The sensitivity of metagenomic testing also depends heavily on the type of sample being tested. Cerebrospinal fluid bathes the brain and spinal cord, so it is particularly useful for detecting infections in those tissues. Metagenomic testing on blood samples, which is also becoming more common, is not always conclusive; localized infections don’t always reach the bloodstream. Indeed, when Yang and his colleagues studied the outcomes of a thousand metagenomic blood tests, they found that the results meaningfully improved patient care in only sixteen per cent of cases. Ishminder Kaur, an associate clinical professor of pediatrics at U.C.L.A. and a lead author of the study, told me that metagenomic testing shouldn’t be ordered automatically. “In the right patient and the right situation, it’s a good add-on test,” she told me. “It’s not a replacement for getting tissue from the source of infection.” Yang worries that some clinicians are approaching metagenomics much as they are approaching artificial intelligence: they are rushing to adopt the technology without adequately studying its limitations. “We believe that technology can solve all our problems,” he told me. “This is very different from years ago, when we used to learn it before we used it.” The risk is that doctors treat the results of metagenomics tests as an ultimate truth, rather than what it is: another data point requiring interpretation. None of these concerns has dampened Yang’s enthusiasm for metagenomics, however. The technology’s promise was on display this spring, when metagenomics helped identify the cause of a sickness aboard the M.V. Hondius cruise ship: the Andes hantavirus. Around the same time, hospitals in Ituri province, in the Democratic Republic of the Congo, began to fill up with patients suffering from fevers, diarrhea, and vomiting. Tests for malaria and typhoid fever came back negative, as did PCR tests for the common Zaire strain of Ebola. Only weeks later, in mid-May, did clinicians send patient samples to Kinshasa, the capital of the D.R.C., for metagenomic analysis. Within sixteen hours, sequencing had identified the virus that was making people sick: the Bundibugyo variant of Ebola. The problem wasn’t that the PCR tests had been wrong; after all, the patients did not have the strain of Ebola for which they were tested. The problem, it seemed, was that the patients had received the wrong kind of test. ♦