BilR is a gut microbial enzyme that reduces bilirubin to urobilinogen

Abstract

Metabolism of haem by-products such as bilirubin by humans and their gut microbiota is essential to human health, as excess serum bilirubin can cause jaundice and even neurological damage. The bacterial enzymes that reduce bilirubin to urobilinogen, a key step in this pathway, have remained unidentified. Here we used biochemical analyses and comparative genomics to identify BilR as a gut-microbiota-derived bilirubin reductase that reduces bilirubin to urobilinogen. We delineated the BilR sequences from similar reductases through the identification of key residues critical for bilirubin reduction and found that BilR is predominantly encoded by Firmicutes species. Analysis of human gut metagenomes revealed that BilR is nearly ubiquitous in healthy adults, but prevalence is decreased in neonates and individuals with inflammatory bowel disease. This discovery sheds light on the role of the gut microbiome in bilirubin metabolism and highlights the significance of the gut-liver axis in maintaining bilirubin homeostasis.

Fig. 1. Identification of bilirubin-reducing bacterial strains.

a, Illustrated representation of the haem degradation pathway. Key human enzymes are labelled with grey text. b, Diagram of the structures of bilirubin and urobilinogen. The bonds reduced during bilirubin reduction are shown in red. c, Results of fluorescence assay screening of bacterial strains. Measurements from n = 3 independent biological replicates are shown as black points. Bars show the ratios of the samples’ fluorescence to a corresponding abiotic media sample with bilirubin added. Error bars indicate 1 s.e. above and below the mean values. The grey line marks a ratio of 5, above which the sample was considered to be positive for bilirubin reduction. Clostridium sp. M62/1 and Clostridium citroniae WAL-17108 are represented by single data points.

Fig. 2. Putative bilirubin reductase operons.

The phylogenetic tree shows the relationship between five bilirubin reducers and five non-reducers. Genes are represented as arrows. Genes are coloured to show predicted domains.

Fig. 3. Confirmation of bilR(S) bilirubin reductase activity.

a, Schematic of the Clostridium symbiosum and Clostridioides difficile bilRS construct. P_tac is a tac-promoter, P_T7 is a T7 promoter, and LacO is a lac operator. b, Schematic of the Ruminococcus gnavus bilR construct. c, Fluorescence assay comparing bilirubin reduction activities of E. coli 10-beta transformed with the empty vector, Clostridium symbiosum construct and Clostridioides difficile construct (two-sample one-sided t-test). d, Metabolomics confirmation of urobilin production (two-sample one-sided t-test). e, Fluorescence confirmation of the bilirubin reduction activity of E. coli 10-beta transformed with the Ruminococcus gnavus bilR construct (two-sample t-test). The bar heights show the mean of n = 4 independent biological replicates used to generate the plots and statistics in c and d. A total of n = 3 independent biological replicates were used to generate the plots and statistics in e. Error bars indicate 1 s.e. above and below the mean. P values of a two-sample one-sided t-test to determine whether the mean of the samples is greater than the mean of the vector controls are provided above the brackets. Individual data points are indicated by black points on each plot.

Fig. 4. Identification of a bilR clade.

a, Gene tree constructed from putative bilR sequences and related reductases from the Old Yellow Enzyme family. Experimentally confirmed bilirubin reducers are labelled in the tree. Clade 1 indicates the overall bilR clade, and clade 2 indicates the short bilR clade. The same tree with all bootstrap values and species labels is included in Extended Data Fig. 8. b, AlphaFold-predicted structure of the Ruminococcus gnavus BilR with sequences coloured to show their degree of conservation within clade 1 based on a ConSurf conservation analysis. A docked bilirubin (green) and FMN (orange) molecule are shown on the structure, and the insets show the potential site of interaction between the R167 residue and the putative enzyme substrates. The HGDR residues are coloured purple in the insets. c, Diagram showing the conservation of positions within the clade 1 BilR sequences. The conserved HGDR motif positions are coloured. d, Fluorescence assay results comparing the bilirubin reductase activities in E. coli 10-beta transformed with the vector control, Ruminococcus gnavus bilR and Ruminococcus gnavus bilR with the Aspartic Acid, Arginine (DR) residues at position 166–167 mutated to Alanine, Alanine (AA). The bar heights are based on the mean of n = 6 independent biological replicates used to generate the plots and statistics in d. Error bars indicate 1 s.e. above and below the mean values. P values of a two-sample one-sided t-test to determine whether the mean of the native Ruminococcus gnavus bilR was higher than the mean of the vector control or mutant are provided. Individual data points are indicated by black points on each plot.

Fig. 5. Taxonomic distribution of bilirubin reductase.

The cladogram shows the relationships between different taxa with detected bilirubin reductase genes. The outer rings show the presence of the short bilR gene (red) and long bilR gene (orange).

Fig. 6. Presence of bilR in the human gut during development and disease.

a, Percentage of infant gut metagenomes missing bilR during their first year of life. The period of highest jaundice susceptibility is indicated by a shaded blue area on the plot. b, Comparison of bilR absence in samples from healthy adults and infants in their first month of life. c, Comparison of the percentage of samples with no bilR detected from healthy adults and adults with IBD (Crohn’s disease (CD) or ulcerative colitis (UC)). The number of metagenomic samples included in each dataset is indicated above each bar. The P values for each comparison show the results of a test of equal proportions to determine whether the fraction of samples with no bilR detected was different between groups, without adjusting for multiple testing.

Extended Data Fig. 1. Diagram of the heme degradation pathway in humans.

Chemical structures are shown for each intermediate metabolite along with the names of the enzymes involved and if they are host or bacterially encoded.

Extended Data Fig. 2. Negative controls for reduction screening.

Plots show corresponding fluorescence assays for experiments shown in Fig. 1C, without bilirubin added to the culture. Individual data points from n = 3 independent biological replicates are shown as black points and the bar heights indicate the mean values of the replicates. Error bars indicate one standard error above and below the mean values. The gray line marks a ratio of 5 for the fluorescence to the abiotic media sample. Data was not obtained for C. citroniae WAL-17109.

Extended Data Fig. 3. AlphaFold predicted BilR and BilS structures.

The C. symbiosum BilR and BilS were aligned to the R. gnavus BilR structure and are shown in the same orientation to show the similar fold structure. The structures are colored based on the predicted protein domains.

Extended Data Fig. 4. Structural alignment of the predicted BilR and 1PS9 structures.

a) Alignment of the full structures of 2,4-Dienoyl CoA reductase (1PS9) structure (blue) and the AlphaFold predicted structure of the R. gnavus BilR protein (red). b) Visualization of the active site on the aligned proteins using the same color scheme. The Arginine 167 residues from the BilR is highlighted in light red and the Tyrosine 166 residues from the 1PS9 protein is highlighted in light blue. The metabolites and residues of interest are labeled.

Extended Data Fig. 5. Metabolomics confirmation of bilirubin reduction.

Extraction ion chromatograms for the detection of Mesobilirubin (m/z 589.3 →m/z 301.2) Urobilin (m/z 591.4→ m/z 343.2) and Stercobilin (m/z 595.4→m/z 345.2). a) Mesobilirubin standard. b) Urobilin standard. c) Stercobilin standard. d-g) pCW-dif-bilRS biological replicates. h-k) pCW-sym-bilRS biological replicates. l-o) Vector control samples. p-s) BHI media control samples. Panels a and a have different Y and X axis scales compared to the other plots to show the corresponding peaks.

Extended Data Fig. 6. Confirmation of reduction of bilirubin, mesobilirubin to urobilin.

a) Fluorescence assay comparing mesobilirubin reduction activity of C. difficile CD3, C. symbiosum WAL-14163, and R. gnavus to an abiotic media controls with mesobilirubin added. b) Fluorescence assay comparing mesobilirubin reduction activity of heterologously expressed C. symbiosum bilRS in E. coli to the vector control (two-sample one-sided t-test). Bar heights in plots indicate the mean values of n = 3 independent biological replicates. The gray line marks a ratio of 1, equivalent to the mean control fluorescence. Individual data points are indicated by black points on each plot. Error bars indicate one standard error above and below the mean values. P-values of a two-sample one-sided t-test testing if the mean of the samples are greater than the means of the controls are shown on panels a and b.

Extended Data Fig. 7. Metabolomics confirmation of mesobilirubin reduction.

Extraction ion chromatograms for the detection of Mesobilirubin (m/z 589.3 →m/z 301.2) Urobilin (m/z 591.4→ m/z 343.2) and Stercobilin (m/z 595.4→m/z 345.2). a) Mesobilirubin standard. b) Urobilin standard. c) Stercobilin standard. d-f) E. coli transformed with pCW_gna_bilRS incubated with bilirubin (d), mesobilirubin (e) or nothing (f). g-h) C. difficile incubated with mesobilirubin (g) or nothing (h). i-j) C. symbiosum incubated with mesobilirubin (i) or nothing (J). k-l) E. coli transformed with pCW_sym_bilRS incubated with mesobilirubin (k) or nothing (l). m-n) C. symbiosum incubated with bilirubin (m) and incubated with bilirubin and co-injected with the urobilin standard (n). o-p) E. coli strains transformed with empty pCW vector incubated with bilirubin (o) and mesobilirubin (p).

Extended Data Fig. 8. Gene tree of bilR clade.

This tree is the same as the tree represented in Fig. 4a. Bootstrap values are included based on 1000 bootstrap iterations. Species with experimentally confirmed BilR genes are shown with white backgrounds. Clade 1 of the bilR genes is highlighted in blue, and Clade 2 is highlighted in orange. Clades not containing bilR genes are collapsed.

Extended Data Fig. 9. Circular dichromism confirmation of mutant bilR structural similarity.

Circular dichroism spectra for the wild type BilR protein (blue) and DR166AA mutant BilR protein (red).

Extended Data Fig. 10. Frequent absence of bilR during the first three months of life.

Plot showing the percent of samples within each five-day age bin with no bilirubin reductase detected. Points are sized based on the number of samples included in each bin.