Tributyrin (TB) degrading clones exhibited butyrate producing capability.
Since TB is an ester derived from a glycerol and 3 molecules of butyrate, we postulated that enzymes that can catabolize TB may produce butyrate. Hence, we sought to establish an efficient screening method for high-throughput identification of bacterial clones that express such TB-degrading enzymes. We took advantage of the fact that agar plates containing dissolved TB (Tributyrin Agar, TBA) appear turbid, and that upon TB degradation, the agar turns transparent. Indeed, we observed that bacterial clones producing TB-degrading enzymes, such as PAO1 strain of Pseudomonas aeruginosa, form colonies with halos of a cleared zone on TBA (data not shown). Using this simple yet effective screening scheme (Fig. 1A), we screened a metagenomic library consisting of 5,760 bacterial artificial chromosome (BAC) clones, where each DNA insert was purified from the gut microbiome of BALB/c mice . An E. coli DH10B that contains the pIndigoBAC-5 vector with no insert was used as negative control. Two clones (33E2 and 54E5) exhibited prominent halos around the colony on TBA following 48h growth (Fig. 1B). These two TB-degrading clones were subsequently isolated and evaluated for butyrate production. Quantitative measurement of butyrate using HPLC implicated that 33E2 and 54E5 produced > 8 times and > 3 times more butyrate, respectively, than the control strain in the presence of 20 mM of the substrate TB. This quantitative butyrate measurement is consistent with what was observed on the TB plate (Fig. 1B). In addition, we also confirmed that overnight cultures of 33E2 and 54E5 produced the characteristic smell which signalled butyrate production.
Sequence analysis of 33E2 and 54E5 reveals the genes responsible for TB degradation.
In order to better understand the genes related to TB degradation, the vectors of 33E2 and 54E5 were sequenced and analysed. Consequently, the sequence of 33E2, which contains a 10.4 kb insert, was fully assembled as a complete circular contig. In contrast, the inserted sequence of the 54E5 clone was assembled as three contigs. Based on the lengths of these three contigs, the insert size of the 54E5 clone is estimated to be larger than 64.7 kb. Full-length sequences of these inserts are currently undergoing the deposition to NCBI database.
When BLASTn searches were performed using the sequences of the 33E2 and 54E5 inserts as query, no significant alignment was retrieved, suggesting that the inserts of these two clones are derived from microbes with unknown genome sequences (Table 1 and Table 2). However, the 33E2 insert is most similar to a region of Blautia producta genome with 74.3% identity (Table 1). Third, fourth and fifth highest-ranking hits of the BLASTn search were identified as genomic sequences from Clostridium lentocellum, Herbinix sp. and Cellulosilyticum sp. bacteria, respectively, all of which belong to the Lachnospiraceae family [24, 25]. Contig1 and contig3 of the 54E5 insert share regions of high homology with genomic regions of Duncaniella sp. (Table 2). No sequence with significant homology to contig2 was detected by BLASTn. Disregarding contig2, most of the hits in Table 4 are associated with members of the Muribaculaceae family . These results suggest that the DNA inserts of 33E2 and 54E5 have originated from bacteria belonging to Lachnospiraceae and Muribaculaceae, respectively. Of note, the predicted origin information of the 33E2 and 54E5 DNA fragments are consistent with the information derived from the BLASTp search of amino acid sequences as query.
The sequence of the 33E2 insert contains 9 ORFs including the first ORF that encodes a putative esterase (Fig. 2A). Although the species origin of the DNA fragment cannot be ascertained, all of the ORFs within the 33E2 insert encode proteins highly homologous to those of Lachnospiraceae family, whose members are well-known producers of short-chain fatty acids . The list of proteins encoded by 33E2 insert genes is provided in Supplementary Information (Table S1). The protein encoded by ORF1 (795 bp, 240 aa) is most homologous to the alpha/beta hydrolase fold domain-containing protein of Cellulosilyticum lentocellum with 60.46% amino acid sequence identity. Second-ranked is a protein with the same description and produced by another species of the same genus, Cellulosilyticum (Fig. 2A). Other proteins in the list are acetylesterases, either from the genus of Clostridium or Herbinix (Fig. 2A). These results strongly suggest that TB-degrading capability of the 33E2 clone can be attributed to the enzymatic activity of this gene product.
The sequence of the 54E5 insert contains 32 ORFs and 90.6% (29 ORFs) of the encoded proteins are highly homologous to proteins from the Bacteroidales order (Fig. 2B). The very first gene in contig3 (1194 bp, 397aa) was determined to encode a protein, whose amino acid sequence is highly homologous to a Muribaculaceae bacterium esterase with 86.9% identity. Proteins retrieved from the BLASTp search using this protein as query similarly include esterases or alpha/beta hydrolases from species of the Bacteroides genus (Fig. 2B). Hereafter, we named ORF1 of 33E2 and ORF1 of the 54E5 contig1 as tbe1 and tbe2, respectively, with tbe standing for “tributyrin esterase”.
Characterization of Tbe1 and Tbe2.
In an effort to shed light on the molecular nature of Tbe1 and Tbe2, the respective amino acid sequences were examined using InterPro and SignalP5.0. Analysis of Tbe1 revealed the lack of a signal peptide in the amino acid sequence implying that Tbe1 is probably an intracellular protein. In contrast, Tbe2 amino acid sequence includes a signal peptide sequence, from residue 1 to 19 (Fig. 3A, red underlined), which suggests that Tbe2 may be an extracellular protein. Furthermore, amino acid sequences of Tbe1 and Tbe2 were compared with those of the known TB esterases of Lactobacillus lactis (LL_Tbe) and Streptococcus pneumoniae (SP_Est) [27, 28]. All of the proteins contained an alpha-beta hydrolase conserved domain (InterPro entry: IPR029058), which is common to hydrolytic enzymes (Fig. 3A). In addition, we also detected the presence of a serine-histidine-aspartate (SHD) catalytic motif, included within the active sites of most alpha-beta hydrolase fold super family proteins  (Fig. 3A). Furthermore, all SHD motifs found in our analysis contained 3 consensus sequences (LLHG, GLSMGG, and DFL) (Fig. 3A, highlighted in black).
We then asked whether Tbe1 and Tbe2 proteins are structurally similar to a known esterase. Structures of Tbe1 and Tbe2 were simulated based on the 3D structure of S. pneumoniae EstA  from the RSCB Protein Data Bank (PDB entry: 2UZ0) as template. Tbe1 shares the overall structural similarity with the template to a greater extent than Tbe2 does (Fig. 3B). However, the SHD triad, found in both Tbe1 and Tbe2 conserved within the catalytic core region, is nicely superimposed with that of the S. pneumoniae esterase. These results strongly suggest that Tbe1 and Tbe2 are responsible for TB degradation and butyrate biosynthesis in the 33E2 and 54E5 clones.
Relative activity of Tbe1 or Tbe2 with the known TB esterases.
Next, we compared the enzyme activity of Tbe1 and Tbe2 with previously reported tributyrin esterases [27, 30]. To this end, we further cloned genes encoding the esterase of Lactococcus lactis (LL_Tbe) and Streptococcus pneumoniae (SP_Est). In order to evaluate the enzyme activities under the same condition, all four genes were cloned into the same location in the pBAD24 plasmid, which is downstream of the promoter sequence of tbe1. The bacterial strains were then incubated with 4-nitrophenyl butyrate (p-NPB), a substrate of the enzyme. Hydrolysis of p-NPB by the esterase enzyme releases 4-nitrophenolate, which can be detected by absorbance at 400nm. Based on Fig. 4, the esterase of S. pneumoniae was the most potent among the four tested enzymes. Of note, Tbe2 is more active than Tbe1, and Tbe1 demonstrated almost identical activity to the L. lactis enzyme. While using purified enzymes is more preferable for a precise quantification of the enzyme’s kinetic parameters, these results clearly suggest that Tbe1 and Tbe2, which we identified in the current study, are either as equally as or more active than the esterase of L. lactis, a probiotic strain.
Tbe1 and Tbe2 degrade TB and produce butyrate.
Our bioinformatic analyses clearly suggest that Tbe1 and Tbe2 are the most probable proteins contributing to conversion of TB into butyrate. In order to conclusively determine the roles of Tbe1 and Tbe2, we cloned tbe1 and tbe2 genes into another E. coli host and monitored TB-induced butyrate production. The host strain we used for Tbe expression is the tEc strain that we isolated from the mouse intestines . tbe1 gene was amplified to include ~100 bp sequence upstream of the tbe1 ORF and cloned into pBAD24 plasmid (pBAD24::tbe1). Inclusion of the upstream sequence was intended to enable transcription of tbe1 using its endogenous promoter. tbe2 gene was inserted right at the junction of the 54E5 plasmid (Fig. 2B), so we were unable to clone tbe2 with its endogenous promoter. Therefore, tbe2 gene was cloned in place of the tbe1 ORF in pBAD24::tbe1, and the resultant plasmid was named pBAD24::tbe2. When tEc strains harbouring either plasmids were grown in the presence of 5 mM TB, prominent increases in butyrate production level were observed (Fig. 4). Butyrate production levels in these two cultures were >16 times higher than the control culture. The characteristic scent of butyrate was again noticed in these two cultures.
Of interest, the level of butyrate produced by tbe2 gene was almost identical to that by tbe1 gene. In Fig. 1C, butyrate production by 54E5 was substantially less than that by 33E2. We speculate that this discrepancy was caused by the absence of tbe2 gene’s own endogenous promoter in the original 54E5 clone. In 54E5, tbe2 gene expression was probably induced by the upstream sequence part of the pIndigoBAC-5 plasmid. When tbe2 was transcribed in the presence of the tbe1 promoter in pBAD24::tbe2, greater butyrate production was achieved.
In vivo on-site butyrate production protects mice against DSS-induced colitis.
Our results so far demonstrate that TB together with Tbe-expressing E. coli cells can efficiently produce butyrate, a beneficial microbiome-derived metabolite. To examine whether or not butyrate produced by this system can alleviate intestinal inflammation, we used DSS-induced mouse model of acute colitis. Mice were divided into 4 groups that received different treatments, as illustrated in Fig. 5A. Mice administered 2.5 % DSS developed acute colitis characterized by weight loss, bloody diarrhoea, and watery stool, and these outcomes were collectively reflected in the increase of the DAI (Disease Activity Index) score  (Fig. 5B). In TB-only group, the DAI scores remained persistently high even after DSS administration was discontinued, suggesting that TB itself did not exert any positive effects toward restoration of a healthy gut condition in mice. It was of particular interest that E. coli cells expressing tbe2 provided significant beneficial effects in mice suffering from severe colitis (Fig. 5B). E. coli cells expressing tbe1 did not seem to confer any protective effects in comparison to the empty vector control group (Fig. 5B). It is not clear why only tbe2-expressing E. coli cells led to the amelioration of DAI scores.
Another phenotype of the DSS-induced acute colitis is the shortened intestinal length as gut inflammation manifests . On day 12, the last day of the experiment, mice from each group were sacrificed and the lengths of their colons were measured. The colon lengths of mice that received Tbe1- or Tbe2-expressing E. coli cells were significantly longer than those of control mice (Fig. 5C), indicating the positive effects of Tbe against DSS-induced colitis. Moreover, the intestinal tissues looked significantly improved by Tbe1 or Tbe2. Hyper-inflammatory responses evidenced by neutrophil infiltration (black arrow) and fluid accumulation (white arrows) were observed in DSS-pretreated mice that received only TB (Fig. 5D). Still higher degrees of neutrophil staining were observed in mice that received control E. coli cells (Fig. 5E, black arrow). In contrast, intestinal tissues looked considerably improved when DSS-pretreated mice were treated with TB and E. coli cells that express Tbe1 (Fig. 5F) or Tbe2 (Fig. 5G). These results clearly suggest that on-site butyrate production by coupling TB and microbiome-derived novel esterases can exert beneficial effects in the inflamed mouse intestines.