The scientific evidence against Bruce Ivins, the 62-year-old Army scientist who killed himself while about to be indicted for the anthrax murders, is finally emerging. Last week, the Federal Bureau of Investigation (FBI) laid some of its cards on the table. One key document, scientists say, now enables a reconstruction of the trail that led the FBI from the deadly letters back to Ivins's lab at the U.S. Army Medical Research Institute of Infectious Diseases (USAMRIID) in Fort Detrick, Maryland.
The investigation relied heavily on outside labs such as The Institute for Genomic Research (TIGR) in Rockville, Maryland, which sequenced a large number of anthrax samples; it also required the development of new genetic tests. Although none of the steps was revolutionary or particularly inventive, researchers say, combining them to solve a criminal case was. Surprisingly, many past speculations on the forensic science were wrong on one point: Sophisticated fingerprinting techniques for Bacillus anthracis developed at Northern Arizona University (NAU) in Flagstaff, widely rumored to be crucial, didn't play a significant role.
Scientists say they need many more details to decide the merits of the case against Ivins. But despite the bureau's widely ridiculed mistakes--including an early focus on Ivins's former colleague Steven Hatfill--"the scientific evidence is probably really strong," says Steven Salzberg, a former TIGR researcher now at the University of Maryland (UMD), College Park. "They've got some very good people," Salzberg says. "The impression that they're not good may just come from their style. They never tell you anything."
The main document unsealed last week is an October 2007 affidavit by Thomas Dellafera, a postal inspector. Filed in support of a warrant to search Ivins's home, cars, and a safety box, the 25 pages of text didn't spell out the details of the evidence. But a close reading of the four paragraphs about the FBI's genetic analysis helps clarify how the bureau approached the problem, says microbiologist Jeffrey Miller of the University of California, Los Angeles.
The key to understanding the investigation is that the anthrax used in the attacks didn't have a single, uniform genetic makeup, a source close to the investigation says. Each of the envelopes likely contained many billions of spores; within such a population, there are always subpopulations of cells bearing mutations that set them apart from the majority. The same minorities would presumably have been present in the "mother stock” of anthrax from which the spores were prepared.
However, standard sequencing--which would require the DNA from thousands of spores--would have resulted in a "consensus sequence" for the spores, in which such rare mutations were simply drowned out. To find them, researchers used a different technique: They grew spores from the envelopes on petri dishes, generating hundreds or even thousands of colonies per dish, each the progeny of a single spore. They then searched for colonies that looked different from the majority; the affidavit mentions variations in "shape, color, texture." (Those colonies might have been rough instead of smooth, or much smaller than most, Miller says.) Next, they set out to find the mutations that made those colonies different.
To do that, the FBI used a brute-force approach: It had the entire genomes of the bacteria in the minority sequenced. TIGR--which merged into the J. Craig Venter Institute in 2006--sequenced "probably somewhere between 10 and 20" such genomes in the years after the attacks, Salzberg says. TIGR could not handle live anthrax cells itself; the FBI gave the lab purified DNA produced by Paul Keim's lab at NAU, Salzberg says. Claire Fraser-Liggett, who led TIGR at the time and is now also at UMD, declines to discuss details of the investigation. But two other sources confirm TIGR's role.
Comparing the sequence of the variant colonies to an original B. anthracis strain called Ames, widely used in research, identified a number of mutations, says Salzberg; they included single-nucleotide polymorphisms, a change of a single base pair, and tandem repeats, in which a short piece of DNA is repeated a variable number of times.
The FBI then had scientists at other labs develop tests that allowed them to screen any anthrax sample for four of these mutations. Such assays "are very easy to design," for instance, using a polymerase chain reaction-based strategy, says evolutionary biologist Richard Lenski of Michigan State University in East Lansing; molecular biology labs do it all the time.
Armed with the four tests, the FBI examined more than 1000 anthrax isolates, collected from 16 labs that had the Ames strain in the United States and several more in Canada, Sweden, and the United Kingdom. In only eight of those samples, they found all four mutations seen in the envelope samples; and each of these eight, the affidavit says, was "directly related" to a "large flask" of spores, identified as RMR-1029, which Ivins had created in 1997 and of which he was the "sole custodian."
That still leaves many questions open, researchers say. One thing that needs to be explained, says Miller, is whether the eight isolates that were "directly related" to RMR-1029 were all found at USAMRIID, or whether some came from other laboratories. In the latter case, it's unclear why the FBI ruled out those labs as the potential origin. (One clue that the affidavit offers is that USAMRIID is the only lab in Maryland or Virginia, the states where the particular envelopes used in the attacks were sold.)
It's also unclear how many of the 1000 samples had fewer than four, but more than zero, of the mutations. "If a whole bunch of them had two or three," that would increase the odds that the perfect match at USAMRIID was just a false positive, Lenski says. Another key question, he adds: Where in the anthrax genome did the four mutations occur? If they were in hypervariable regions, that would also probably make the case against Ivins weaker.
Whether the analysis would hold up in court seemed to be front and center in the FBI's thinking, says Salzberg. For instance, when researchers from TIGR and NAU published a comparison of two anthrax strains in Science in 2002 (14 June 2002, p. 2028), a top FBI researcher named Bruce Budowle encouraged them to include a statistical analysis to estimate the data's accuracy, Salzberg says. "Budowle felt it would be useful to have it all go through peer review, in case it went to court," he says.
The FBI has invested heavily in microbial forensic expertise since 2001, and Budowle has co-authored many papers on the topic. But the bureau farmed out much of the scientific bench work, in part because the Marine Corps doesn't allow bioweapons agents at its base in Quantico, Virginia, where the FBI Laboratory is located. The work was "highly compartmentalized," says a source close to the investigation: Most labs didn't know exactly what the others were doing.
The affidavit is very unclear about whether the spore preparations might have undergone physical or chemical treatments to make them disperse more easily--still a point of major confusion, says Barbara Hatch Rosenberg, a bioweapons specialist at Purchase College in New York. Scientists at the Armed Forces Institute of Pathology reported in October 2001 that the spores sent to U.S. Senator Tom Daschle's office had been mixed with silica to make them more easily dispersible. However, in congressional briefings and in a paper published in the August 2006 issue of Applied and Environmental Microbiology, FBI officials described the powder as a simple spore preparation without additives.
The affidavit reports that there was "an elemental signature of Silicon within the spores" in all four letters that were recovered. This silicon signature is later cited as part of the evidence linking the mailed anthrax to the flask of spores that Ivins had access to. But what the silicon was for, or whether other samples were tested for the signature, remains unclear.
Science aside, the affidavit relies heavily on circumstantial evidence. For instance, it notes unexplained spikes in Ivins's nighttime lab activity right before the two waves of letters were sent. It also claims that he tried to mislead investigators to hide his involvement. In April 2002, he submitted samples from his lab that tested negative for the four mutations, according to the affidavit; but on 7 April 2004, an FBI agent seized the RMR-1029 flask, which tested positive for all four. Ivins insisted he had given agents RMR-1029 the first time around, however.
One of the weak points in the affidavit is Ivins's motive, says Gregory Koblentz, a biodefense specialist at George Mason University in Fairfax, Virginia. The FBI suggests that Ivins was afraid of losing his job if the government ended a project he was working on that was trying to solve regulatory issues around the so-called AVA anthrax vaccine. It "seems a bit of a stretch" that Ivins would have thought his job hinged on that project, says Koblentz. His group "would have had plenty of other anthrax vaccine-related work to keep them busy." A glaring omission, meanwhile, is any evidence placing Ivins in Princeton, New Jersey, on any of the days the envelopes could have been mailed from there.
A spokesperson for the FBI's laboratory declined a request to interview Budowle and referred scientific questions to the FBI's Washington, D.C., field office. "In the near future the FBI will determine the best way to address the science involved in the anthrax case," she e-mailed Science. Many suspect that with so many burning questions, a full account of the evidence--including the scientific details--is now just a matter of time.
With reporting by Yudhijit Bhattacharjee.