Identification of food and nutrient components as predictors of Lactobacillus colonization

Abstract

A previous double-blind, randomized clinical trial of 42 healthy individuals conducted with Lactobacillus johnsonii N6.2 found that the probiotic's mechanistic tryptophan pathway was significantly modified when the data was stratified based on the individuals' lactic acid bacteria (LAB) stool content. These results suggest that confounding factors such as dietary intake which impact stool LAB content may affect the response to the probiotic treatment. Using dietary intake, serum metabolite, and stool LAB colony forming unit (CFU) data from a previous clinical trial, the relationships between diet, metabolic response, and fecal LAB were assessed. The diets of subject groups with high vs. low CFUs of LAB/g of wet stool differed in their intakes of monounsaturated fatty acids, vegetables, proteins, and dairy. Individuals with high LAB consumed greater amounts of cheese, fermented meats, soy, nuts and seeds, alcoholic beverages, and oils whereas individuals with low LAB consumed higher amounts of tomatoes, starchy vegetables, and poultry. Several dietary variables correlated with LAB counts; positive correlations were determined for nuts and seeds, fish high in N-3 fatty acids, soy, and processed meats, and negative correlations to consumption of vegetables including tomatoes. Using machine learning, predictors of LAB count included cheese, nuts and seeds, fish high in N-3 fatty acids, and erucic acid. Erucic acid alone accurately predicted LAB categorization, and was shown to be utilized as a sole fatty acid source by several Lactobacillus species regardless of their mode of fermentation. Several metabolites were significantly upregulated in each group based on LAB titers, notably polypropylene glycol, caproic acid, pyrazine, and chondroitin sulfate; however, none were correlated with the dietary intake variables. These findings suggest that dietary variables may drive the presence of LAB in the human gastrointestinal tract and potentially impact response to probiotic interventions.

Keywords: Lactobacillus johnsonii N6.2; dietary intake; lactic acid bacteria; machine learning; microbiome; multivariate analysis; probiotic.

Figure 1

Principal component analysis (PCA) of H-LAB and L-LAB groups by habitual intake of (A) meats (oz), (B) vegetables (oz), (C) dairy (oz), and (D) monounsaturated fats (mg). The variables for each of the two dimensions and their % contribution to the variance are given in the eigenvector charts for each PCA plot.

Figure 2

Identification of biomarkers for LAB abundancy by most frequently predictive and greatest importance dietary intake variables of random forest models trained with (A) meats, (B) monounsaturated fatty acids erucic acid (“M221,” C22:1) and oleic acid (“M181,” C18:1).

Figure 3

Lactobacillus species showed a visible ability to utilize erucic and oleic acid as a sole source of fatty acids. Growth curves of (A) Lactobacillus johnsonii N6.2, (B) L. johnsonii ATCC 33200, (C) Lactobacillus amylovorus ATCC 33620, (D) Ligilactobacillus murinus 35,020, (E) Limosilactobacillus reuteri TD1 (F) Ligilactobacillus fermentum 14,391 (G) Lactipantibacillus plantarum 25,302, and (H) Lacticaseibacillus casei ATCC 334, were grown in MRS-NT supplemented with 0.1% oleic acid (MRS-O), 0.1% erucic acid (MRS-E), Tween 80 (MRS-T), without Tween 80 (MRS-NT), and with Tween 80 and 0.1% DMSO (MRS-TD).