Diet and environment are two of the biggest influences on gut microbiota. So maybe it’s only appropriate that gut microbiota researcher Patrick Schloss often uses food and gardening analogies to describe his research problem.
“Typically, when a patient goes into the hospital, he or she gets a whole smorgasbord of antibiotics,” he says. And since exposure to antibiotics is one of the biggest risk factors for the hospital-acquired infection, Clostridium difficile, “that puts a huge pool of people at risk.” The nasty gut invader causes diarrheal infections that cost 15,000 lives and $4.8 billion in added healthcare costs each year.
“What we really wanted to know is whether we can predict who’s going to get C. difficile, how severe the infection will be, and whether we can do something to reduce the risk,” says Schloss, a microbiologist at University of Michigan in Ann Arbor who oversaw the study published this week [http://mbio.asm.org/content/6/4/e00974-15.full] in mBio.
Schloss and his former graduate student, Alyxandria Schubert, first began playing around with different antibiotic treatment scenarios in mice to see what their resulting, altered microbiota communities would look like. Schubert gave mice 16 different drug regimens, using various titrated doses and recovery times after dosing.
“At that point, we thought it would be cool to take all these different microbiota communities we had generated and see if we could make any predictions about susceptibility to C. difficile,” says Schloss. Those communities varied widely and a correlation analysis showed that relationships were not black and white, with different treatments yielding differing abundances of certain gut bacterial species. The bottom line was that multiple bacterial species worked together to provide resistance to C. difficile colonization, and who was on each protective team was very context dependent. (Image: Multiple bacteria species (red) keep Clostridium difficile (green) from colonizing a mouse’s gut (purple). Credit: Vincent Young and Christine Bassis)
“As we started playing with the data, we wondered if we could model this more mathematically?” They turned to a machine-learning algorithm that uses a forest of decision trees to shift through the complicated data set. “You can think of it like the spam filter on your email. If it sees certain words like ‘prince’, ‘Nigerian’, and ‘millions,’ then the filter decides an email is spam."
In this case, the team built a mathematical model to ask whether the make-up of a mouse’s particular gut microbial community could predict susceptibility to C. difficile infection. The answer was yes—for each mouse ‘patient’s’ microbiome, the model predicted susceptibility correctly within an error of less than 10%.
The analysis also revealed the complicated nature of the relationships at play to make a gut resistant or sensitive to C. difficile. Resistance was associated with the presence of members of the Porphyromonadaceae, Lachnospiraceae, Lactobacillus, Alistipes, and Turicibacter. Susceptibility was associated with losses of these populations and concurrent gains in members of Escherichia or Streptococcus.
But, Schloss notes, susceptibility is not simply an all-or-nothing scenario—it’s all about the context of the total microbiome players. “I think about it as a buffet—you have to mix and match different ingredients to get resistance or sensitivity to C. difficile.”
That means it won’t be easy to nail down the mechanism of how the microbiota team players provide resistance. “I don’t think it’s going to be one simple mechanism,” says Schubert, now a postdoctoral fellow with Schloss. “It could be nutrient competition, stimulating the immune system in different ways, or bacteria secreting molecules that prevent C. difficile from germinating.”
However, the study’s findings could lead to better ways of categorizing hospital patients as being at low or high risk of infection. Schubert notes that patients’ gut microbiota could be sequenced and assessed for which key species are missing and those bacteria could be added back in a probiotic supplement.
Schloss sums up with a gardening analogy: “C. difficile is like a weed. If you clear out some of the gut microbiota using antibiotics, then you expose space for the weed to grow. You need a complex community to fill all the niche holes and keep C. difficile out.”