Prediction of Response to Systemic Corticosteroids in Active UC by Microbial Composition-A Prospective Multicenter Study

Abstract

Background: Corticosteroids are used for induction of remission in patients with moderately to severely active ulcerative colitis. However, up to one-third of patients fail to this therapy. We investigated if fecal microbial composition or its metabolic capacity are associated with response to systemic corticosteroids.

Methods: In this prospective, multicenter study, patients with active ulcerative colitis (Lichtiger score ≥4) receiving systemic corticosteroids were eligible. Data were assessed and fecal samples collected before and after 4 weeks of treatment. Patients were divided into responders (decrease of Lichtiger Score ≥50%) and nonresponders. The fecal microbiome was assessed by the 16S rRNA gene marker and analyzed with QIIME 2. Microbial metabolic pathways were predicted using parsimonious flux balance analysis.

Results: Among 93 included patients, 69 (74%) patients responded to corticosteroids after 4 weeks. At baseline, responders could not be distinguished from nonresponders by microbial diversity and composition, except for a subgroup of biologic-naïve patients. Within 4 weeks of treatment, responders experienced changes in beta diversity with enrichment of ascribed beneficial taxa, including Blautia, Anaerostipes, and Bifidobacterium, as well as an increase in predicted butyrate synthesis. Nonresponders had only minor longitudinal taxonomic changes with a significant increase of Streptococcus salivarius and a microbial composition shifting away from responders.

Conclusion: Baseline microbial diversity and composition seem to be of limited use to predict response to systemic corticosteroids in active ulcerative colitis. Response is longitudinally associated with restoration of microbial composition and its metabolic capacity.

Keywords: butyrate; corticosteroids; microbiome; ulcerative colitis.

Figure 1.

Serum and fecal biomarkers in responders (n = 69) compared with nonresponders (n = 24) at baseline (V1) and after 4 weeks of corticosteroid therapy (V2) displayed with boxplots. Significances are indicated with bars and stars within the diagrams. *P ≤ 0.05; **P ≤ .01; ***P ≤ .001. Green: responders, purple: nonresponders. LCN-2 = Lipocalin-2.

Figure 2.

Comparison of alpha and beta diversity in responders vs. nonresponders to corticosteroids at baseline (V1). A, Alpha diversity showed no difference in responders compared with nonresponders at V1 determined by species richness, Shannon entropy and phylogenetic diversity (Kruskal Wallis: P > .05; response, n = 69; nonresponse, n = 24). B, Unweighted UniFrac showed no qualitative phylogenetic difference in the microbial composition between responders and nonresponders at V1 (ANOSIM: P > .05; response, n = 69; nonresponse, n = 24). Green: responders, purple: nonresponders. Abbreviation: NS, not significant.

Figure 3.

Changes in alpha and beta diversity and taxonomy in responders during the 4 weeks of corticosteroid therapy. A, Alpha diversity showed no significant changes in samples of responders at V1 compared with V2 determined by species richness, Shannon entropy and phylogenetic diversity (Kruskal Wallis: P > .05; V1 n = 69; V2 n = 63). B, Beta diversity showed a significant qualitative (Unweighted UniFrac) phylogenetic difference in the microbial composition of responders at V1 compared with V2 (PERMANOVA: P < .05; V1 n = 69; V2 n = 63). C, D, LEfSe analysis showed 23 significantly discriminative features in responders comparing microbial taxonomy in fecal samples at V1 and V2 with relative abundance (Logarithmic LDA score ≥4.0; P < .05; V1 n = 69; V2 n = 63). Dark green: baseline (V1), light green: after 4 weeks of corticosteroid therapy (V2). Abbreviation: NS, not significant. *P ≤ .05; **P ≤ .01.

Figure 4.

Changes in alpha and beta diversity and taxonomy in nonresponders during the 4 weeks of corticosteroid therapy. A, Alpha diversity showed no significant changes in samples of nonresponders at V1 compared with V2 determined by species richness, Shannon entropy and phylogenetic diversity (Kruskal Wallis: P > .05; V1 n = 24; V2 n = 22). B, Beta diversity showed no significant differences in the microbial composition of nonresponders at V1 compared with V2 as shown for unweighted UniFrac (PERMANOVA: P > .05; V1 n = 24; V2 n = 22). C, LEfSe analysis showed 8 significantly discriminative features in nonresponders comparing microbial taxonomy in fecal samples at V1 and V2 (Logarithmic LDA score ≥4.0; P < .05; V1 n = 24; V2 n = 22). Dark purple: baseline (V1), light purple: after 4 weeks of corticosteroid therapy (V2). Abbreviation: NS, not significant. *P ≤ .05; **P ≤ .01.

Figure 5.

Comparison of alpha and beta diversity and taxonomy in responders vs nonresponders to corticosteroids at week 4 (V2). A, Alpha diversity was significantly reduced in responders compared with nonresponders after 4 weeks of corticosteroid treatment determined by species richness, Shannon entropy and phylogenetic diversity (Kruskal Wallis: P < .05; response n = 63; nonresponse n = 22). B, Unweighted UniFrac showed a significant qualitative phylogenetic difference in the microbial composition between responders and nonresponders at V2 (ANOSIM: P < .05; response n = 63; nonresponse n = 22). C, LEfSe analysis showed four significantly discriminative features at V2 that were associated with response (Logarithmic LDA score ≥4.0; P < .05; V1 n = 63; V2 n = 22). Green: responders, purple: nonresponders. *P ≤ .05; **P ≤ .01.

Figure 6.

Predicted butyrate production by the gut microbial communities at baseline (V1) and after 4 weeks of corticosteroid therapy (V2) compared between responders and nonresponders and between time points. A, Comparison of butyrate production rate between responders and nonresponders at V1 and V2. P values are based on Wilcoxon rank sum exact test. B, Comparison between V1 and V2 stratified by therapy response to corticosteroids. Connected dots refer to the same patient. P values are calculated from the paired Wilcoxon signed rank exact test. V1 n = 93, V2 n = 85.