Personalized B cell response to the Lactobacillus rhamnosus GG probiotic in healthy human subjects: a randomized trial

Abstract

The specific effects of administering live probiotics in the human gut are not well characterized. To this end, we investigated the immediate effect of Lactobacillus rhamnosus GG (LGG) in the jejunum of 27 healthy volunteers 2 h after ingestion using a combination of global RNA sequencing of human biopsies and bacterial DNA sequencing in a multi-visit, randomized, cross-over design (ClinicalTrials.gov number NCT03140878). While LGG was detectable in jejunum after 2 h in treated subjects, the gene expression response vs. placebo was subtle if assessed across all subjects. However, clustering analysis revealed that one-third of subjects exhibited a strong and consistent LGG response involving hundreds of genes, where genes related to B cell activation were upregulated, consistent with prior results in mice. Immunohistochemistry and single cell-based deconvolution analyses showed that this B cell signature likely is due to activation and proliferation of existing B cells rather than B cell immigration to the tissue. Our results indicate that the LGG strain has an immediate effect in the human gut in a subpopulation of individuals. In extension, our data strongly suggest that studies on in vivo probiotic effects in humans require large cohorts and must take individual variation into account.

Keywords: Lactobacillus rhamnosus GG; B cell activation; human transcriptomics; immediate in vivo effect; probiotics.

Figures

Figure 1.

Experimental design. A total of 29 healthy volunteers were enrolled in a cross-over design. At visit 1, volunteers were assessed for eligibility, given a questionnaire and informed consent was acquired. At visit 2, 15 subjects (blue) ingested LGG solution and remaining 15 subjects ingested placebo solution (red). Two hours after ingestion an upper GI endoscopy was performed: biopsies and intestinal fluid from jejunum were sampled for RNA-seq and bacterial DNA sequencing, respectively. At visit 3, the same procedure as in visit 2 was repeated, but subjects given LGG in visit two received placebo and vice versa (see color highlights). Two subjects developed conditions that required medication between visits 2 and 3 and were therefore excluded from the study, so the final number of subjects was 27 as indicated

Figure 1.

Experimental design. A total of 29 healthy volunteers were enrolled in a cross-over design. At visit 1, volunteers were assessed for eligibility, given a questionnaire and informed consent was acquired. At visit 2, 15 subjects (blue) ingested LGG solution and remaining 15 subjects ingested placebo solution (red). Two hours after ingestion an upper GI endoscopy was performed: biopsies and intestinal fluid from jejunum were sampled for RNA-seq and bacterial DNA sequencing, respectively. At visit 3, the same procedure as in visit 2 was repeated, but subjects given LGG in visit two received placebo and vice versa (see color highlights). Two subjects developed conditions that required medication between visits 2 and 3 and were therefore excluded from the study, so the final number of subjects was 27 as indicated

Figure 2.

Fraction of LGG reads from bacterial DNA sequencing. Y-axis shows the fraction of reads originating from LGG vs. all bacteria in luminal fluid from jejunum taken 2 h after LGG or placebo digestion as boxplots. Triangles show individual samples (N = 23; two runs failed). X-axis shows treatment

Figure 3.

Differential expression analysis of LGG vs. placebo response. (a) Expression analysis using all subjects as a single group. Y-axis shows the number of differentially expressed genes (LGG vs. placebo), as a function of cutoffs based only on significance, or significance and effect size (X-axis). Opaque colors show up-regulation, pale colors show down-regulation. Numbers on bars show the number of up/down-regulated genes. (b) Hierarchical clustering of normalized RNA-seq libraries. Y-axis shows Euclidean distance in the tree, created by complete linkage. Each subject is represented by two leaves, colored by treatment. Note that libraries always cluster by subject, not treatment (indicated by subject number below leaf pairs). (c) Multidimensional scaling (MDS) plot based on log2 LGG/control RNA-seq expression values. Axes represent dimensions 1 and 2. Dots represent subjects, colored by sex. Two major groups (defined as LGG-responders and non-LGG-responders, based on subsequent analysis, see main text and panel D) were observed, as indicated by color. Three subjects were outside these groups and were considered as outliers (excluded in subsequent analyses). The subset of genes (N = 1389) that showed LGG responsiveness were used for MDS analysis (see Methods). (d) Expression analysis using groups defined in panel C. Y-axis shows the number of differentially expressed genes (LGG vs. placebo) within groups defined in panel C (X-axis; blue bars show results for LGG-responders, red bars show results for non-LGG-responders, where solid color indicate up-regulated genes and and pale colors down-regulated genes). Upper panel shows the results of analysis using FDR<0.05 as cutoff, lower panel shows the results of analysis using FDR<0.05 and absolute log2 fold change >0.5. Numbers on bars show the number of up/down-regulated genes. (e) Relation between LGG vs. placebo expression change in LGG-responders and non-LGG-responders. Y-axis shows the average RNA-seq log2 fold change (LGG vs. placebo) in non-LGG-responder subjects. X-axis shows the average RNA-seq log2 fold change (LGG vs. placebo) in LGG-responder subjects. Only genes that were differentially expressed in LGG-responders are shown (FDR<0.05). Dots are colored by the number of genes falling into respective plot area. Dotted lines show log2 fold change of 0 (no change) on both axes. Dashed line shows the diagonal (Y = X)

Figure 3.

Differential expression analysis of LGG vs. placebo response. (a) Expression analysis using all subjects as a single group. Y-axis shows the number of differentially expressed genes (LGG vs. placebo), as a function of cutoffs based only on significance, or significance and effect size (X-axis). Opaque colors show up-regulation, pale colors show down-regulation. Numbers on bars show the number of up/down-regulated genes. (b) Hierarchical clustering of normalized RNA-seq libraries. Y-axis shows Euclidean distance in the tree, created by complete linkage. Each subject is represented by two leaves, colored by treatment. Note that libraries always cluster by subject, not treatment (indicated by subject number below leaf pairs). (c) Multidimensional scaling (MDS) plot based on log2 LGG/control RNA-seq expression values. Axes represent dimensions 1 and 2. Dots represent subjects, colored by sex. Two major groups (defined as LGG-responders and non-LGG-responders, based on subsequent analysis, see main text and panel D) were observed, as indicated by color. Three subjects were outside these groups and were considered as outliers (excluded in subsequent analyses). The subset of genes (N = 1389) that showed LGG responsiveness were used for MDS analysis (see Methods). (d) Expression analysis using groups defined in panel C. Y-axis shows the number of differentially expressed genes (LGG vs. placebo) within groups defined in panel C (X-axis; blue bars show results for LGG-responders, red bars show results for non-LGG-responders, where solid color indicate up-regulated genes and and pale colors down-regulated genes). Upper panel shows the results of analysis using FDR<0.05 as cutoff, lower panel shows the results of analysis using FDR<0.05 and absolute log2 fold change >0.5. Numbers on bars show the number of up/down-regulated genes. (e) Relation between LGG vs. placebo expression change in LGG-responders and non-LGG-responders. Y-axis shows the average RNA-seq log2 fold change (LGG vs. placebo) in non-LGG-responder subjects. X-axis shows the average RNA-seq log2 fold change (LGG vs. placebo) in LGG-responder subjects. Only genes that were differentially expressed in LGG-responders are shown (FDR<0.05). Dots are colored by the number of genes falling into respective plot area. Dotted lines show log2 fold change of 0 (no change) on both axes. Dashed line shows the diagonal (Y = X)

Figure 4.

Gene ontology (GO) and pathway analysis of genes upregulated after LGG treatment in the LGG-responder group. (a) Over-representation of GO terms. Y-axis shows top 10 Biological Process GO terms (FDR<0.05 and enrichment score >3), sorted after enrichment score (X-axis). Bars are colored by over-representation FDR on -log10 scale. (b) Over-representation of KEGG pathways. Plot is organized as in A, but shows over-represented KEGG pathways (FDR<0.05 and enrichment score >3). (c) Over-representation of REACTOME pathways. Plot is organized as in A, but shows top 15 over-represented REACTOME pathways (FDR<0.05 and enrichment score >3). (d) LGG response of gene in the B cell response pathway. Pathway schematic is based on the KEGG database. Boxes indicate genes or complexes, colored by their LGG response (average log2 fold change LGG vs placebo); note that colors are capped at log2 fold change +-2

Figure 4.

Gene ontology (GO) and pathway analysis of genes upregulated after LGG treatment in the LGG-responder group. (a) Over-representation of GO terms. Y-axis shows top 10 Biological Process GO terms (FDR<0.05 and enrichment score >3), sorted after enrichment score (X-axis). Bars are colored by over-representation FDR on -log10 scale. (b) Over-representation of KEGG pathways. Plot is organized as in A, but shows over-represented KEGG pathways (FDR<0.05 and enrichment score >3). (c) Over-representation of REACTOME pathways. Plot is organized as in A, but shows top 15 over-represented REACTOME pathways (FDR<0.05 and enrichment score >3). (d) LGG response of gene in the B cell response pathway. Pathway schematic is based on the KEGG database. Boxes indicate genes or complexes, colored by their LGG response (average log2 fold change LGG vs placebo); note that colors are capped at log2 fold change +-2

Figure 5.

STRING interaction analysis plot of LGG-upregulated genes. Circles indicate genes upregulated after LGG treatment in the LGG -responsive group (FDR<0.05, log2 fold change >0.5; gene names within boxes). Lines between genes represent evidence of interaction or co-expression (from the STRING database), where thicker lines indicate stronger evidence. Gene circle color indicates whether the genes were also upregulated following LGG treatment in mouse. Genes with no connections in the STRING database are not shown. Two major interaction clusters are evident, dominated by immune response genes (including B cell response), and proliferation/cell cycle genes (blue and beige background, respectively)

Figure 5.

STRING interaction analysis plot of LGG-upregulated genes. Circles indicate genes upregulated after LGG treatment in the LGG -responsive group (FDR<0.05, log2 fold change >0.5; gene names within boxes). Lines between genes represent evidence of interaction or co-expression (from the STRING database), where thicker lines indicate stronger evidence. Gene circle color indicates whether the genes were also upregulated following LGG treatment in mouse. Genes with no connections in the STRING database are not shown. Two major interaction clusters are evident, dominated by immune response genes (including B cell response), and proliferation/cell cycle genes (blue and beige background, respectively)