Bacteria as Biosensors of Environmental Damage
Source: Eric Dubinsky and Dan Hawkes and LBNL TABL
ESD ecologist Eric Dubinsky was part of a scientific team whose work, recently reported on in mBio (the online open-access journal of the American Society for Microbiology), has shown the power of microbial communities to predict sources of contamination in the environment. The team found that the rapid sequencing of microbiomes in place at environmental sites can be used to monitor damage caused by human activity.
The team hypothesized that because bacteria are already continuously monitoring and responding to changing environmental conditions, then perhaps they could act as an onsite environmental surveillance network. Tracking the make-up of those bacterial communities could be digitized through DNA sequencing.
To find out if bacterial community structure was predictive of contamination, they employed a method called supervised machine learning, using the Random forest algorithm that relies on an ensemble of thousands of decision trees. The team sampled 93 different monitoring wells near Oak Ridge, some of which are known to be contaminated with uranium and nitrate from the early development of nuclear weapons at the site. From each sample, the microbial 16S rRNA gene was extracted and sequenced, yielding a total of 26,943 unique species of bacteria. After weeding out the species in very low abundance or narrowly distributed across the well sites, the team was left with 2,972 species to use as “features” in a machine-learning model.
Using DNA data, this model could accurately predict uranium contamination of the groundwater 88% of the time and nitrate contamination 73% of the time. Then, testing the model within a vastly different environment and contaminant—on 60 samples from the Deepwater Horizon spill in the Gulf of Mexico previously collected both before and during the 2010 oil spill—the model accurately predicted oil contamination 98% of the time.
Even more impressive, the model could predict the presence of previous contamination in samples where the chemical signature of oil hydrocarbons was absent. In other words, even after the oil had been degraded by bacteria, the presence of a particular bacterial community structure could still identify that the contamination event had taken place.
Dubinsky contributed the data showing that microbial community composition is a near-perfect predictor of locations with oil contamination in the Gulf of Mexico—even areas impacted by previous contamination where chemical signatures of oil were no longer present. This shows that microbial communities can be more sensitive than chemical measurements in tracking the fate of oil spills and potentially other contaminants.
This work was recently highlighted in a press release by the American Society of Microbiology: http://www.asm.org/index.php/newsroom/92-news-room/press-releases/93516-bacterial-communities-can-act-as-precise-biosensors-of-environmental-damage
Smith, M.B., A.M. Rocha, C.S. Smillie, S.W. Olesen, C. Paradis, L. Wu, J.H. Campbell, J.L. Fortney, T.L. Mehlhorn, K.A. Lowe, J.E. Earles, J. Phillips, S.M Techtmann, D.C. Joyner, D.A. Elias, K.L. Bailey, R.A. Hurt, Jr., S.P. Preheim, M.C. Sanders, J. Yang, M.A. Muelleer, S. Brooks, D.B. Watson, P. Zhang, Z. He, E.A. Dubinsky, P.D. Adams, A.P. Arkin, M.W. Fields, J. Zhou, E.J. Alm, and T.C. Hazen (2015), Natural bacterial communities serve as quantitative geochemical biosensors. mBio, 6 (3), e00326.15; DOI: 10.1128/mBio.00326-15.
To read an earlier paper which presented microbial community data from the spill, go to: http://pubs.acs.org/doi/abs/10.1021/es401676y