Methods:

For rational donor selection, the microbiome profile of HE patients and healthy controls were assessed. HE patients demonstrated a relative reduction of Lachnospiraceae and Ruminococcaceae compared to controls. Leveraging the cross-sectional microbiome data, a donor was selected from the donor database using Random Forrest analysis, to complement the relative deficiencies of the patient microbiota 3. The same donor was used for all FMT-assigned participants. These participants underwent five days of pre-FMT antibiotics after which 90ml (27 grams of stool) of the donor material containing approximately 2.7 E12 CFU was administered via enema. SOC participants did not undergo any interventions or antibiotic therapy. All participants were followed in keeping with the visits reported in Figure 1A. There were no FMT-related serious adverse events in the five months post-study as judged by the data safety monitoring board. For the long-term study, all participants were followed prospectively at least every 2 months through chart review, phone calls or in-person with specific focus on hospitalizations and HE episodes. Eligibility, cognitive testing, further donor selection and methods are in the supplement. At the long-term visit, we re-administered cognitive tests (Psychometric hepatic encephalopathy score, PHES and EncephalApp Stroop) and obtained stool (Figure 1A) 3. Data were compared to pre-FMT, post-antibiotics and early post-FMT values. Microbial composition was studied using 16S rRNA sequencing. Principal component (PCO) and Linear discriminant analysis effect size (LEFSe) analysis were performed.