In a surprising new study, researchers using image-analysis methods similar
to those employed in facial-recognition software have made a discovery that
rules out the two main theories scientists had created to explain how bacteria
self-organize into multicellular aggregate mounds. The study by researchers
from Rice Univ.
and the Univ. of Georgia appears online in the Proceedings of the National Academy of
Sciences.
The find is important for the study of biofilms. Federal health officials
have estimated that as many as 80% of all microbial infections arise from
biofilms, and scientists know that the same bacteria can be up to 1,000 times
more resistant to antibiotics if they’re living inside a biofilm rather than
living on their own. To better fight biofilms, scientists have been scrambling
to understand the biochemical and biophysical mechanisms that allow bacteria to
form aggregates, reorganize, and interact.
“The results of our analysis were really surprising,” said study
co-author Oleg Igoshin, assistant professor in bioengineering at Rice.
“Our results didn’t support either of the major competing theories people
have come up with. Those theories were each predicated on the idea that as the
bacterial mounds were forming and reorganizing, the individual bacterium were
drawn toward one or another of them by some sort of chemical signal.
“That doesn’t appear to be the case at all,” Igoshin said.
“We didn’t find any neighbor-related factors between the groups at all.
Instead, there seems to be a signaling mechanism within the group itself that
trumps everything else.”
The study involved the bacterium Myxococcus
xanthus, a
common soil bacteria that’s often studied for its ability to self-organize into
various patterns. In the wild, M. xanthus are content
to collectively hunt other bacteria. But when food is scarce, they stream
together into aggregates containing up to 100,000 cells and form spores. The
resulting aggregate mounds are large enough to be carried away to better
environs by the wind or passing insects.
To study this behavior in the lab, Igoshin and Rice co-authors postdoctoral
fellow Chunyan Xie and graduate student Haiyang Zhang created a computer
program that could analyze thousands of still frames from microcinematic movies
of M. xanthus. The movies were created
in the laboratory of Univ.
of Georgia collaborator
and co-author Lawrence Shimkets. The movies showed how M. xanthus
streamed together to form aggregates. One hallmark of the M. xanthus streaming process is that less than half of the
aggregates that initially form will survive through the end of the process. The
factors that control this ripening are not understood.
In designing their image-analysis application, Igoshin’s team had the
computer scrutinize every aggregate—frame-by-frame—throughout the streaming
process. The computer cataloged 33 properties for each aggregate, including
things like area, perimeter size, distance to, and size of the nearest
neighbor. After all the data were collected, the team ran a statistical
analysis to find out if any feature or combination of features could be used to
predict which aggregates would eventually win out over their neighbors.
“We found that size mattered most,” Igoshin said. “Not size
in relation to neighbors, which is something people had previously thought
might matter, but size of the aggregate itself. We found that if we answered
one question—is the size of an aggregate beyond a certain threshold—then we
could accurately predict whether the aggregate would survive with 90%
accuracy.”
Igoshin said some of the image analysis methodologies that the team applied
to study M. xanthus are similar to
ones that Chunyan Xies used for facial recognition analyses in her previous
work. He said scientists have only recently begun to apply these sorts of image
analysis techniques to fundamental biological questions like bacterial
self-organization.
“One of the most exciting aspects of this study is the fact that we can
apply these methods much more broadly to study self-organization in other
bacteria and unicellular organisms,” Igoshin said. “In fact, this
kind of analysis is sorely needed, because most of the existing methods to
study these phenomena are qualitative rather than quantitative. As a
discipline, we need quantitative methods if we want to conduct side-by-side comparisons
between real-world and computer-generated results.”
