The Tangled Tree

by David Quammen

  All those researchers from Harvard and elsewhere wanted to understand why. The best clues lay in aspects of bdelloid life history that I’ve already mentioned: their tolerance for desiccation, their reproduction without sex. Desiccation can be damaging to membranes and molecules, even when a creature survives the drought, and biologists suspect that such drying-and-rehydrating stresses cause bdelloid DNA to fracture and leave cell membranes leaky. Given that they’re surrounded in their environments by living bacteria and fungi, plus naked DNA remnants from dead microbes, the porous membranes and fracturing could make it easy for alien DNA to enter even the nuclei of bdelloid cells and to get incorporated into bdelloid genomes as they repair themselves. Let me say that again: broken DNA, as a cell fixes it, using ambient materials, may include bits that weren’t part of the original. If that mended DNA happens to be in cells of the germ line, the changes will be heritable. Baby rotifers will get them and, when the babies mature, pass the changes along to their own daughters. Thus a bacterial or fungal gene can become part of the genome of a lineage of animals.

  Furthermore, the absence of sexual reproduction in bdelloids, the absence of recombination, might leave them especially needy of just such new genetic possibilities. Variation is the raw material of adaptation, as you know, and no lineage survives throughout time and vicissitudes without it. Mutation provides only tiny changes slowly, at the scale of one base in the DNA molecule replaced by another. Sexual recombination, by contrast, makes big rearrangements of what’s already there. The tiny changes alone may not be enough. Omit sex, and you streamline reproduction but sacrifice adaptability. Parthenogenetic populations can thrive in the short term, but in the long term, they tend to go extinct. All this is relevant. Maybe bdelloid rotifers, reproducing asexually for millions of years—with no remixing of gene combinations, only mutations to supply incremental change—have gotten much of their freshening innovation from HGT.

  If so, it’s an aspect of evolution that was unimagined by Charles Darwin. And it goes far beyond the bdelloids.

  58

  It started showing up among insects. Again, this was supposed to be impossible. There were fervent doubters. Alien genes cannot move from one species to another, they insisted. The germ line of animals, meaning the eggs and the sperm and the reproductive cells that give rise to them, is held separate from such influences. It’s sequestered behind what biologists call the Weismann barrier, named for August Weismann, the German biologist of the nineteenth century who defined the concept. (Germline, as now commonly composited, is a good word because it suggests the linearity of that lineage of cells.) Those cells and their DNA are isolated within the ovaries and testes, sequestered from genetic changes that may occur in the rest of the body. Bacteria cannot cross that barrier, Weismann’s barrier—so said the skeptical view—to insert bits of their own DNA into animal genomes. Impossible. And again it turned out to be possible.

  One of the big revelations came in 2007 from a team that included a young postdoc researcher named Julie Dunning Hotopp, then at the Institute for Genomic Research (TIGR), a private entity founded by the brilliant and audacious J. Craig Venter and located in Rockville, Maryland. Venter is the wildcat geneticist who competed against a huge and publicly funded international research initiative, the Human Genome Project, to assemble the first complete (or nearly so) sequence of the human genome. Dunning Hotopp joined TIGR after that ruckus, and her work is notable in its own right. She had come from Michigan State University with a newly minted PhD in microbiology but also an aptitude for what’s called computational biology, meaning the analysis of huge amounts of biological data using computer and mathematical skills. It’s essentially the same as what I’ve earlier mentioned as bioinformatics. Dunning Hotopp teamed with another postdoc, a fellow named Michael Clark at New York’s University of Rochester (where she had gotten her own undergrad degree), and with their two mentors, for a study to see whether bacterial genes might be sneaking into the genomes of insects and other invertebrate animals, such as head lice, crustaceans, and nematode worms. The answer, for eight of the genomes they checked, was a strong yes.

  These transferred genes came from bacteria in the genus Wolbachia, a group of aggressive intracellular parasites that infect at least 20 percent of all insect species on Earth. Wolbachia bacteria target the germline cells of the animals they enter, especially ovaries and testes, and, once established, a Wolbachia infection is passed from mother to offspring within her infected eggs. It does not usually pass within infected sperm. Wolbachia work around that constraint, the absence of sperm-to-offspring transmission, and proliferate themselves by manipulating the reproductive outcomes of their hosts. They do it in four different ways: killing male offspring before they can hatch, turning males into females, triggering parthenogenesis (virgin females delivering more females), and sabotaging the viability of uninfected eggs when fertilized with Wolbachia-infected sperm. The net result of their interference is to change the male–female ratio, shifting whole populations of insects quickly toward more Wolbachia-infected females producing more Wolbachia-infected offspring. These are evolutionary wins for Wolbachia. Given the broad range of insects infected (plus many other arthropods and nematode worms), and the sizes of those populations, Wolbachia is an extraordinarily successful group of parasites. “Arguably,” according to one expert, “the spread of Wolbachia represents one of the great pandemics of life on this planet.”

  Being intracellular parasites means, of course, that Wolbachia bacteria take up residence not just inside the host but also inside cells of the host, closely adjacent in each cell to the nucleus with its DNA. Since they invade not just any cells but primarily the germline cells, that puts Wolbachia close to the very molecules of DNA that will be duplicated in the production of egg cells and passed to offspring. Such proximity seems to offer special opportunity for getting Wolbachia DNA spliced into the insect’s DNA. Julie Dunning Hotopp and her colleagues discovered, by scrutinizing genome sequences from twenty-six different critters, that four insects and four nematode worms had taken aboard Wolbachia genes by horizontal transfer. The most dramatic case was one species of fruit fly, which had accepted almost the entire genome of Wolbachia (more than a million letters of code) into its own nuclear genome.

  The fruit fly in question, Drosophila ananassae, is a favorite laboratory animal, and its genome had already been sequenced by other researchers. The sequence was publicly available. That published version omitted the Wolbachia genome, probably not because it hadn’t appeared during sequencing but because the sequencing team assumed it reflected a bacterial contamination, an error, in their lab work. Researchers at the time were so reluctant to believe that bacterial genes could be transferred into animal genomes that, before publishing a new genome sequence, they routinely edited out the bacterial stretches. Dunning Hotopp and her coworkers took a different approach. Up in Rochester, Clark raised the flies in a laboratory and cured them of their Wolbachia infections, using antibiotics. Under microscopic inspection, their ovaries were clean; the only Wolbachia genes left behind, therefore, would be those actually embedded in the fly’s own genome. Clark then mailed the fly DNA to Dunning Hotopp, in Rockville, where she handled most of the sequencing and the computational analysis.

  “Anything having to do with the animal, he did,” she told me, when I visited her lab. “Anything having to do with the computer, I did. And then some of the stuff in between, we both did.” What they found amid the fly genome, to their own surprise, was almost the entire Wolbachia genome. Their methods were sound, and Science published their paper. It got attention in popular media, including the New York Times and the Washington Post, and was generally well received among the genome biologists.

  Generally, but not universally. Their dramatic and well-supported findings were dismissed outright in certain quarters, Dunning Hotopp told me. She was in Baltimore now, not Rockville, when I called on her, at the Institute for Genome Sciences of the University of Maryland, and contin
uing her work on horizontal gene transfer in animals. Her career was thriving, she had received a prestigious grant of support from the National Institutes of Health, and she was pushing the study of HGT in new directions. Her group had recently discovered evidence, for instance, of bacterial DNA transferred horizontally into the genomes of human tumors. What that dizzying revelation means is still unclear, but there’s at least some chance that such insertions might play a role in causing cancer.

  For the cancer-related work, she and her colleagues used bioinformatics to scan a vast number of human genome sequences, from several sources, looking for stretches that resembled bacterial—not human—DNA. One of their sources was a publicly available database called the Cancer Genome Atlas, containing genome sequences from the tumors of thousands of patients. The genomes of tumors are often different, in small but important ways, from the genomes of patients suffering the cancer, because tumor cells mutate as they replicate. Hotopp’s team did find bacterial DNA lurking within some normal human genomes, an interesting result. More peculiar and disquieting, though, was that they found it 210 times more common in tumor cells than in healthy cells.

  Human cells are continually exposed to bacteria—the ones that live routinely in our guts and on our skin; the ones that infect us sometimes. That intimate juxtaposition has consequences. One consequence, unsuspected before but suggested by this Hotopp study in 2013, is that bits of naked bacterial DNA, possibly from broken-open bacterial cells, may often get integrated into cells (not necessarily germline cells) of a person’s body. Into cells of the stomach lining, for instance. Or blood cells. By “integrated,” what I mean is, not just absorbed or injected into the human cell but patched into its DNA. The good news about any such horizontal transfer, bacterial DNA into nongermline human cells, is that the change isn’t heritable. It won’t be passed to future generations. The bad news is that it might trigger cancer.

  How? By disrupting the cell genome in a way that allows runaway cell replication.

  Hotopp and her colleagues looked especially at two kinds of human cancer, acute myeloid leukemia and stomach adenocarcinoma. In the leukemia cell genomes, they found stretches resembling the DNA of Acinetobacter bacteria, a group that includes infectious forms often picked up in hospitals. In the stomach tumor genomes, they found pieces suggesting Pseudomonas, the genus including Pseudomonas aeruginosa, a nasty bug that also inhabits hospitals and medical equipment, and is especially feared for its resistance to multiple antibiotics. Bacteria have been linked to human cancer in the past—for instance, Helicobacter pylori, an intestinal bug associated with gastric ulcers—and the simplest hypothesis was that by causing inflammation, the bacteria damage DNA and sometimes lead to cancerous mutations. The alternate hypothesis offered by Hotopp’s team, supported by genome data and now crying for further investigation, is that horizontal transfer of bacterial DNA may discombobulate one human cell, in the stomach, in the blood, wherever, and turn it cancerous. Putting horizontal gene transfer on the list of suspected human carcinogens brings it out of the realm of microbial arcana.

  Even before that provocative suggestion, while she was still looking at insects, Dunning Hotopp had faced adamant resistance among a few influential biologists, including some Nobel Prize winners, to her and her colleagues’ discoveries of HGT in the animal kingdom. “No, that’s all artifacts, there’s no way that’s true,” was the tenor of these responses. An artifact, in scientific parlance, is an illusion produced by a methodological mistake. “I have biologists who come into my office,” she said, “and it’s just, like, ‘No, it’s got to be an artifact. You have to be able to explain it some other way.’ ” Animals don’t experience horizontal gene transfer, period. Humans, certainly not.

  “Do you ever say to them, ‘Is that a faith-based statement?’ ” I asked. What I meant was: it seemed almost as though the Weismann barrier had become a theological dogma.

  She mused about that for a moment and allowed that some scientists did appear to be more religious about science than about religion. A touch of faith-based genomics? “I think it is,” she said.

  59

  And yet in the background of this situation, fresh in memory both to Dunning Hotopp and to those who doubted her findings, was an episode that helps illuminate why such critics were skeptical. It involved an embarrassing overreach of scientific claims—an overreach that had put large-scale HGT into human germlines, not just into tumors, and not just into fruit flies and other insects. This intersects with Dunning Hotopp’s story for several reasons, one being that the correctors of the overreach included four of her former colleagues from Craig Venter’s TIGR.

  The race to sequence the human genome, a bitter contest between Venter’s private team and the publicly funded effort, had ended in a negotiated, face-saving tie. The public effort had combined a huge group of government- and university-supported collaborators, known as the International Human Genome Sequencing Consortium. Lots of money had been spent, and no one wanted to admit that duplication of activities had caused waste. Public access to the data, versus private proprietorship, was also at issue. President Bill Clinton announced the brokered finish at a White House ceremony, a fancy press event, on June 26, 2000. Prime Minister Tony Blair was patched in by video, because British scientists and resources had played a sizable role, after which Venter and his counterpart from the public effort, Francis Collins, made polite remarks. What the two groups had to offer at that point, though no one stressed this qualification, was just a rough draft of the genome, consisting of two more-or-less overlapping versions.

  Eight months later, on February 15, 2001, the Consortium published a provisional analysis of their human genome sequence in the journal Nature. (Venter and his group published their own analysis, almost simultaneously, in the journal Science. The full sequence itself was too long to print in any journal, since it ran to about 3.2 billion bases and would have filled many book-length volumes.) Listed first among more than two hundred coauthors on the Consortium’s paper was Eric S. Lander, then of the Whitehead Institute for Biomedical Research, in Cambridge, Massachusetts. Lander’s priority reflected the fact that the Whitehead’s Center for Genome Research, led by him, had contributed more letters of code to the final assemblage than any other participating group. His authorial priority also entitled him to a good share of the discomfort when, soon afterward, one of the paper’s major conclusions was shown to be credulous and overstated, if not outright wrong.

  “Hundreds of human genes appear likely to have resulted from horizontal transfer from bacteria,” the Consortium authors wrote. This had occurred not perhaps in the recent past, they added, but sometime during vertebrate evolution. Hundreds of human genes? More precisely, they put the number at 223. What Lander and his coauthors were saying was that these bacterial genes had come to our vertebrate ancestors not through parentage but on a shortcut, by infective heredity. What was the evidence? The 223 suspects matched bacterial genes very closely but weren’t present in certain eukaryotic creatures outside the vertebrate lineage—not in a yeast, not in a worm, not in a fly, not in mustard weed. So the 223 genes hadn’t come down to us vertically, throughout roughly a half billion years of evolution. They must have come across more recently by horizontal transfer. Yes?

  No. Not necessarily, said a paper by Steven L. Salzberg and three coauthors that appeared soon after the Consortium’s analysis. Salzberg and his coauthors all worked within Venter’s TIGR, giving a certain nuance of rivalry to their response; but it stood on its own and was published by Science. Salzberg himself was TIGR’s director of bioinformatics, so he knew a thing or two about crunching big biological data. (One of his coauthors, Jonathan Eisen, would later be a professor at the University of California, Davis, and write an influential blog titled “The Tree of Life.”) The team of Consortium authors had made two simple mistakes, Salzberg’s group argued. They had failed to look at enough other eukaryotic genomes outside the vertebrate lineage for the possible presence of the
supposedly leaping genes, and they had failed to take seriously the chance that those ancient genes had merely been lost from the four genomes they did look at: the yeast, the worm, the fly, and the mustard weed. Salzberg and his colleagues examined additional data and found an interesting trend: the more nonvertebrate eukaryotic genomes they scrutinized, the fewer genes seemed uniquely shared by bacteria and humans. By the time they finished, the original 223 had been reduced to 41, with a steady downward trend suggesting that further genome sequences, if available, might drop the number to zero. HGT among humans began to look like an illusion.

  Other scientists, even some deeply engaged with the subject of horizontal gene transfer, found the Salzberg critique persuasive. Ford Doolittle and two colleagues wrote a comment in Science, calling the original claim about 223 transferred genes “the most exciting news” from the Human Genome Project so far, but concluded that it was “probably overenthusiastic.” William F. Martin, a tall American biologist at Heinrich Heine University in Düsseldorf, known for his ferocious intelligence, his important ideas, and his bluntness, called the Consortium’s claim “at least an overstatement, very probably a gross exaggeration and possibly altogether erroneous.” The New York Times took note of this “fresh skirmish in the genome wars,” under a headline about HGT into humans being “Hotly Debated by Rival Scientific Camps.” To the Times reporter, Steven Salzberg said that he’d been surprised by the Consortium’s report of 223 alien genes, but that as he read further, “I was immediately struck by the fact it was likely to be an error, because the method was simply wrong.” Eric Lander, also reached by the Times, didn’t admit being wrong but declined to insist he was right.

  This is how science proceeds: by fits and starts, by claims and critiques, by new answers in the light of better data. The fuss over those 223 genes was no disaster for Lander and the Consortium, no scandal, and arguably not even a major embarrassment. It was a correction—a call for greater caution and broader thinking, the oxymoronic combination that makes for genuine scientific advance. It reminded everyone that the prospect of horizontal gene transfer from bacteria (or other microbes) into the human genome is a boggling thing, a trespass on our sense of identity, and an improbability against which there should ever be a high standard of proof. But the other point worth noting about the 223 episode is that it wasn’t the end of that discussion. It was the beginning.