Introduction
Fecal microbiota transplantation (FMT) is an emerging treatment for infectious and autoimmune diseases, whereby donor feces are implanted in a patient’s intestinal tract. This treatment cures recurrent Clostridium difficile infection (rCDI) in 85% of cases (van Nood et al., 2013) and there is some evidence it may be effective for other diseases, including inflammatory bowel disease (Moayyedi et al., 2015, Suskind et al., 2015), metabolic syndrome (Ridaura et al., 2013, Vrieze et al., 2012), and autism (Hsiao et al., 2013). Putative mechanisms of FMT efficacy focus on the trillions of bacteria that inhabit the gastrointestinal tract, the gut microbiota. FMT is thought to restore these bacteria (Shahinas et al., 2012, Youngster et al., 2014), which may then alter host metabolism (Floch, 2015, Trompette et al., 2014), inhibit pathogens (Britton and Young, 2014), and effect changes in host immunity (Furusawa et al., 2013, Ivanov et al., 2009, Round and Mazmanian, 2010).
Precision engineering of the gut microbiota with bacterial isolates in pure culture offers the therapeutic potential of FMT without the risks associated with the use of raw fecal matter (Petrof and Khoruts, 2014). Whether this next generation of microbiome-based therapeutics will effectively replace FMT will depend on (1) whether the “active ingredients” of FMT that carry out a desired mechanism can be identified, (2) whether these strains engraft in a patient’s gut, and (3) whether they are sufficiently abundant to produce a clinical response.
While the mechanism may be studied using in vitro or animal models of disease (as for traditional small-molecule drugs), engraftment and abundance in humans are less well understood. If engraftment is governed by simple rules, such as the law of mass action, it may be highly deterministic and easy to predict. Alternatively, if engraftment is governed by contextual factors, such as genetics, diet, antibiotics, and the immune system, it may vary considerably among patients and be difficult to predict. A quantitative model of bacterial engraftment would accelerate drug discovery efforts by pinpointing the bacteria that engraft at high abundance in a given host. However, no such model exists, and despite significant advances in our understanding of FMT (Li et al., 2016), surprisingly few principles of bacterial engraftment in a human host are known.
Direct tests of bacterial engraftment are difficult, as animal models do not capture important aspects of human biology, and experiments in humans present regulatory and ethical challenges. However, there is already an ongoing large-scale experiment of engraftment in humans: the use of FMT to treat recurrent C. difficile infection.
C. difficile is a Gram-positive pathogen that causes severe diarrhea and is responsible for 500,000 infections, resulting in 30,000 deaths per year (Lessa et al., 2015). It is often carried asymptomatically in the gut, where it is normally inhibited by gut commensals. Disruptions to this inhibition, often via broad-spectrum antibiotics, allow C. difficile to proliferate. First-line treatment with antibiotics can cure this infection, but in 20% of cases C. difficile spores persist and reinitiate the cycle of infection. FMT is thought to break this cycle by restoring the protective gut microbiota that inhibit the growth of C. difficile and prevent recurrence of the infection (Aroniadis and Brandt, 2013).
Bacterial engraftment is therefore thought to be responsible for the efficacy of FMT, yet few studies have examined the factors that promote the engraftment of individual strains. Here, we develop a method for strain inference, Strain Finder, and combine it with techniques from machine learning to quantitatively model bacterial engraftment in diverse human hosts. We uncover the principal factors that govern bacterial engraftment after FMT and show that these rules appear to generalize to the treatment of another disease, metabolic syndrome.
