|Setting the Record Straight|
|Two studies find no evidence that humans have acquired genes directly from bacteria|
Edward R. Winstead
June 25, 2001
A new study effectively dismisses the notion that humans have acquired genes directly from bacteria during evolution. The study is the second in five weeks to refute the claim that 113 human genes came to vertebrates through direct, or horizontal, transfer from bacteria. The original claim was published in February in the Nature paper on the human genome sequence.
In the current issue of Nature, researchers at GlaxoSmithKline, in Collegeville, Pennsylvania, say there is no evidence that bacterial genes have ever migrated to vertebrate genomes. The researchers conducted evolutionary, or phylogenetic, analyses for 28 of the 113 proposed human genes. The 28 human genes have counterparts in nonvertebrate species or are more closely related to species other than bacteria, according to the paper.
"We find absolutely no evidence of bacteria-to-vertebrate gene transfer for the set of 28 genes we analyzed, as well as for the other genes we're now working on," says Michael J. Stanhope, head of Evolutionary Biology, Bioinformatic Sciences, at GlaxoSmithKline. A research team led by Stanhope and James R. Brown constructed evolutionary trees for the 28 genes and is analyzing others in the original set of 113.
Last month, another research team published a study showing that the original claim was an error. Researchers at The Institute for Genomic Research (TIGR), in Rockville, Maryland, reported in Science Express that most of the proposed genes have counterparts in invertebrate databases. They trimmed the original list of candidates to about 40 human genes and said that none of these is likely to have been transferred from bacteria.
Steven L. Salzberg and colleagues at TIGR tested the claim by repeating the original analysis using similar methods but additional data. They simply did not believe that 113 human genes are likely to have "entered the vertebrate (or prevertebrate) lineage by horizontal transfer from bacteria," as the International Human Genome Sequencing Consortium proposed in Nature.
The claim implies that humans can pick up bacterial genes through an infection by a pathogen. It also implies that there are risks in consuming genetically modified food organisms because of the possible transfer of genes from the food organism to the host.
"We researchers think that such events are enormously unlikely, and the public needs to know that," says Michael Stanhope at GlaxoSmithKline. Citing concerns, particularly in Europe, about the safety of genetically modified foods, he adds: "The original claim does nothing but fuel paranoid thoughts about gene transfer."
The two follow-up studies were undertaken almost immediately in an effort to rectify a mistake in a paper that will be widely read and cited in the years to come. "We are trying to set the record straight," says Stanhope. "The Nature human genome paper and its companion in Science are arguably the two most important biological publications in the history of publishing science."
If the bacteria-to-human finding takes hold, experts have warned recently, then time and money will be wasted on unnecessary studies. The original claim has already been repeated as fact in the scientific literature. Trends in Genetics published an article in May stating:
Correcting the scientific record may be easier than changing the public's perceptions about gene transfer. The claim made headlines in February, when interest in the human genome was high. In materials for the press, the public sequencing consortium ranked 11 important discoveries about the human genome. Bacteria-to-vertebrate gene transfer was fourth on the list.
Horizontal gene transfer is a real phenomenon in nature. Bacteria can pick up genes from other bacteria; and mitochondria, organelles in human cells that once were free-living bacteria, have transferred genes to humans. But unlike bacterial genomes, vertebrate genomes are relatively sheltered.
It would be remarkable for even one or two bacterial genes to have entered vertebrate genomes at some point in time, says Stanhope. The biology is complex: For a bacterial gene to be passed on, it would have to penetrate a vertebrate 'germ' cell, which gives rise to eggs or sperm. The transferred gene would have to be maintained. And for its long-term survival, the gene presumably would have to serve a purpose in its new home.
The statement that so extraordinary an evolutionary event has occurred 113 times struck Stanhope as "incredible" and highly improbable. "We read the claim and thought, 'How could this possibly happen?'" recalls James Brown, who co-led the GlaxoSmithKline study. The two re-analyses cite the flawed methodology of the original.
Drawing conclusions about gene transfer based on computational analyses was the fundamental flaw of the original, the critics agree. The authors of the Nature human genome paper used programs such as BLAST to search databases for matches to human gene sequences. The BLAST searches generated lists of 'hits'genes with similar DNA sequences.
"You cannot infer evolutionary relationships from a list of BLAST hits," says Brown, who is head of Portfolio Support, Microbial Bioinformatics, at GlaxoSmithKline. "An evolutionary biologist would look at the hit list and do a phylogenetic analysis to verify the relationships of those genes."
The strategy of the new study was to find all homologous genes for each of the 28 sequences and diagram the relationships using evolutionary trees. By contrast, the original analysis excluded data on a range of invertebrates. Based on BLAST searches that did not screen all available sequence data, the authors identified 113 genes that appeared to exist only in humans and bacteria.
But most of these genes clearly exist in other species. The GlaxoSmithKline researchers easily found matches in nonvertebrate databases, as did the TIGR group. Some invertebrate hits ranked below bacterial genes on the list of results, indicating less sequence similarity between matches than for higher hits. But genes evolve at different rates, which is one reason BLAST comparisons cannot be used to determine evolutionary relationships.
"For some of the 28 sequences, the top ten or so BLAST hits might be for bacterial genes," says Brown. "But further down on the list would be a matching sequence from yeast or worm."
Eleven of the 28 candidates had invertebrate sequence counterparts from the "EST others database," a publicly available online database of partial gene sequences, or Expressed Sequence Tags. Subsequent phylogenetic analysis showed that human and matching sequences from nonvertebrates were close relatives.
"Say you do the BLAST search and the top nonvertebrate hit is a bacterium, but several sequences down in the report you find C. elegans," says Stanhope. "When you put those sequences in an alignment and do the phylogenetic analysis, it becomes clear that C. elegans is the 'sister' group to vertebratesand that bacteria are nowhere near vertebrates on the evolutionary tree."
Salzberg's group at TIGR argues that the loss of genes in invertebrate species is likely to explain cases where a gene appears to exist only in humans and bacteria. Or, they say, a gene may be present in invertebrates but has mutated beyond recognition.
The new analysis revealed that one of the 28 human genes is a possible case of reverse gene transfer: a vertebrate gene picked up by a bacterium. "Reverse gene transfer occurs much more easily, because you're taking DNA and putting it in a unicellular organism that divides asexually," says Stanhope.
The circumstances of a possible vertebrate-to-bacteria transfer are so mysterious that researchers admit ignorance. "The accurate determination of which host was involved in any transfer event requires a great deal of sampling of as many genomes as possible," says Stanhope.
The original claim, he argues, is symptomatic of a larger problem in biological science: "My philosophical opinion is that the field of biology has recently been preoccupied with data management issues. And in building bioinformatic tools to handle all those data, we have sometimes been missing the biological focus."
"Bringing order to genomic sequences is a biological problem that needs a biological perspective," Stanhope continues. "The field is tilted a little toward software engineering, and you could make an argument that this mistake is symptomatic of where we are in this regard."
The GlaxoSmithKline researchers submitted their paper to Nature because that's where the original claim appeared. "We felt that it was Nature's responsibility to want to set the record straight," says Stanhope, adding: "The editors treated it like any other paper."
See related GNN article
. . .