Discovery of a Cyclic Choline Analog That Inhibits AnaerobicCholine Metabolism by Human Gut Bacteria
The anaerobic conversion of choline to trimethylamine (TMA) by the human gut microbiota has been linked to multiple human diseases. The potential impact of this microbial metabolic activity on host health has inspired multiple efforts to identify small molecule inhibitors. Here, they use information about the structure and mechanism of the bacterial enzyme choline TMAlyase (CutC) to develop a cyclic choline analog that inhibits the conversion of choline to TMA in bacterial whole cells and in a complex gut microbial community. In vitro biochemical assays and a crystal structure suggest that this analog is a competitive, mechanism-based inhibitor. This work demonstrates the utility of structure-based design to access inhibitors of radical enzymes from the human gut microbiota.
The human gastrointestinal (GI) tract is colonized by trillions of microorganisms that have coevolved with the host. These organisms play an important role in maintaining host health by performing critical biological functions, including metabolizing inaccessible dietary components, synthesizing essential vitamins and nutrients, providing protection from invading pathogens, and priming the immune system. Gut microbial metabolic activities have also been linked to human disease. Indeed, metabolomics experiments have showed strong correlations between levels of human serum metabolites made or modified by gut microbes and various disease states. Anaerobic gut microbial choline metabolism is a disease-associated activity that has received much attention as a potential target for therapeutic intervention. While choline is an essential nutrient for the host, contributing to cell membrane function, methyl transfer events, and neurotransmission, gut anaerobes use choline as a source of carbon and energy, cleaving its C−N bond to generate acetaldehyde, which is further processed by alcohol dehydrogenase CutO and aldehyde oxidoreductase CutF to give ethanol and acetyl coenzyme A, and trimethylamine (TMA), a uniquely microbial metabolite. While gut microbes generate TMA from multiple sources, including the nutrients carnitine and glycine betaine, choline is considered to be the major precursor.
Anaerobic choline metabolism is a disease-linked gut microbial activity and a promising target for inhibition. (A) Generation of TMA from choline by gut microbes and its subsequent processing in the human body. (B) Crystal structure of CutC (PDB = 5FAU) and (C) proposed mechanism for choline cleavage.
In the human body, TMA is further oxidized to trimethylamine-N-oxide (TMAO) by the liver enzyme flavin-dependent monooxygenase 3 (FMO3). Elevated levels of TMAO in human plasma have been linked to multiple diseases such as nonalcoholic fatty liver disease (NAFLD), cardiovascular disease, chronic kidney disease, type 2 diabetes,and atherosclerosis. If mutations in the gene encoding FMO3 result in a deficiency of this enzyme, TMA builds up in the body and its excessive excretion leads to a distinct odor. This inherited metabolic disorder is known as trimethylaminuria or “fish odor syndrome”. The correlation between TMA production by the gut microbiota and diseases, as well as its prevalence in the human gut microbiota, makes this metabolic activity a prominent target for manipulation and further study. They have discovered and biochemically characterized the enzymes responsible for the key C−N bond cleavage reaction that converts choline to TMA: the glycyl radical enzyme (GRE) choline TMA-lyase (CutC) and its corresponding radical S-adenosylmethionine (SAM) activating protein (CutD).
Previously reported small molecule inhibitors of anaerobic bacterial choline metabolism.
GREs are a family of oxygen-sensitive enzymes that use protein-based radical intermediates to catalyze chemically challenging reactions. Well characterized family members include pyruvate formate-lyase, class III (anaerobic) ribonucleotide reductase, benzylsuccinate synthase, and 4- hydroxyphenylacetate decarboxylase.The glycyl radical is installed by a dedicated partner enzyme that is a member of the radical SAM enzyme superfamily. A crystal structure of unactivated CutC with choline bound in the active site, together with site-directed mutagenesis, supports a mechanism in which the glycyl radical of CutC abstracts a hydrogen atom from an adjacent cysteine (Cys489), generating a thiyl radical intermediate that initiates the reaction with choline by abstracting the pro-S hydrogen atom from C1. Deprotonation of the C1-OH by Glu491, followed by C−N bond cleavage via a “spin-center shift”, would then eliminate TMA.
Strategy for rigidifying choline analogs leading to N-methyl-3-pyrrolidinol 1 or N-methyl-3-piperidinol 2.
Hydrogen atom abstraction from Cys489 by the resulting acetaldehyde-centered radical would regenerate the thiyl radical and produce acetaldehyde. Proton transfer would then prepare the CutC active site for the next round of catalysis. Targeting microbial choline metabolism to reduce TMA generation (leading to lowered TMAO levels) has been proposed as a potential therapeutic approach for the treatment of cardiometabolic diseases. As CutC is the enzyme responsible for TMA generation, it may be an ideal therapeutic target. Hazen and co-workers reported the first CutC inhibitor, 3,3-dimethyl-1-butanol (DMB), replacing the nitrogen atom of choline by a carbon . They demonstrated that orally administered DMB reduced plasma TMAO levels in mice. They also observed that DMB treatment reduced foam cell formation induced by a high choline diet in an atherosclerosis mouse model. However, in our hands DMB poorly inhibits CutC activity in vitro and in bacterial cell culture, suggesting these effects may arise from interactions with a different target. Subsequently, Hazen et al. reported greatly improved rationally designed CutC inhibitors, including iodomethylcholine (IMC)28 and various N, N-dimethylprop-2-yn derivatives. Bhandari and co-workers recently described a pyridine derivative, trigonelline, that reduces TMA production from choline by inhibiting growth of choline metabolizing bacterial strains, like Citrobacter f reundii, and also directly inhibits FMO3.
Binding of (S)-6 to CutC reveals key interactions important for recognition. (A) Crystal structure of unactivated CutC soaked with (S)-6 reveals interactions between bound inhibitor and CutC active site residues. Black dashes represent H-bond interactions, purple dashes represent CHO interactions, gray dashes represent a potential π–π interaction, and orange dashes represent the path of radical transfer. Chemical structure of (S)-6 with numbered atoms is shown next to the structure. (B) Simulated annealing composite omit map (2Fo – Fc, blue mesh) contoured at 1.5σ showing the location of the inhibitor. (C) Previously described crystal structure of unactivated CutC bound to choline (PDB 5FAU). Interactions between CutC and choline are color coded as described above with the exception of a cation−π interaction which is shown in gray dashes. (D) Comparison of choline-bound (protein residues are colored in light green) and (S)-6-bound (colored in green) CutC active sites. Choline and (S)-6 were omitted from the structure for clarity.
More recently, Boehringer-Ingelheim used a matrix-assisted laser desorption/ionization time-of-flight (MALDITOF)-based high-throughput screen with whole cells of Clostridium hathewayi to identify new inhibitors of choline metabolism. This study revealed the ability of an inhibitor of choline transport to prevent TMA production. Concurrently, they reported an additional CutC inhibitor, betaine aldehyde (BA) (in vitro IC50 = 26 μM), which was identified by screening a focused library of choline analogs for activity toward the cut gene cluster-containing human gut isolate Escherichia coli MS 200-1. A cocrystal structure of BA bound to unactivated CutC showed the formation of a thiohemiacetal with Cys489 of CutC, suggesting this compound could to be a covalent, reversible inhibitor. They also found that BA inhibited TMA formation from choline in whole cell assays with multiple choline-degrading human gut bacteria (EC50 = 0.4−2.5 mM range). Building on this proof of concept work, they sought to use structural information to design more potent CutC inhibitors.
Potential Mechanism of CutC Inhibition by (S)-6
Examining the conformation of choline within the CutC active site, they noticed that the trimethylammonium group is oriented gauche to the hydroxyl, with an average dihedral angle of 61°. The aromatic ring of Phe395 may stabilize this conformation of choline by forming a face-on cation−π-like interaction with C2 (C2 to ring center distance: 3.8 Å), which would be expected to have a partial positive charge due to the adjacent quaternary ammonium. Overlaying choline- and BA-bound CutC crystal structures showed that BA binds within the active site in a similar conformation to choline, maintaining contacts with key active site residues. Together, these structures suggested the possibility of building this conformation into an inhibitor scaffold by forming a ring between C1 and an N-methyl group, leading either to N-methyl-3-pyrrolidinol 1 or N-methyl-3-piperidinol 2. These modifications could potentially improve inhibitor potency by constraining its conformation and thereby reducing the entropic penalty of binding. These compounds (synthesis detailed in the SI) were tested for inhibition of both anaerobic choline metabolism by E. coli MS 200-1 and CutC activity in vitro.
Both 1 and 2 inhibited TMA formation from choline in vitro and in bacterial cell culture. They decided to prioritize compound 2 for further optimization as it was more active in whole cells. They found that moving the hydroxyl group to the 4-position (compound 3) led to a 3- and 2-fold loss in vitro and in whole cell activities, respectively. Expanding the size of the heterocyclic ring (compound 4) did not improve activity relative to 2. Oxidation of the alcohol to a ketone (5) led to a compound that was 20-fold less active toward whole cells, indicating that installing an electrophile would not improve activity. This loss of potency could also arise from the absence of a hydrogen atom on C3 if its abstraction is critical for activity. They next introduced an alkene into the piperidine ring (compound 6), hypothesizing that this modification could affect potency by activating the C3 hydrogen atom for abstraction and/or altering the piperidine ring conformation. This modification improved inhibitory activity 100-fold both in whole cells and in vitro. The addition of a methyl group to this derivative at position 5 (compound 7) decreased activity. An acetylated analog (compound 8) was 100-fold less active than 6 in whole cells.
They then turned their attention toward the N-methyl group, replacing it with both a carbon atom (compound 9) and an oxygen atom (compound 10). The methylene containing derivative 9 was 25- and 230-fold less active than 6 in vitro and in whole cells, respectively. While 10 possessed activity similar to that of 6 in the whole-cell assay, it was 1,000- fold less active toward CutC in vitro, suggesting that it may interact with an additional, unknown target. These results highlight the importance of the amine for activity toward CutC and suggest this part of the molecule may be mimicking the trimethylammonium group of choline. Next, they varied the substituent on the piperidine nitrogen atom. Introduction of bulkier substituents like n-propyl (compound 11), cyclobutyl (compound 12), and benzyl (compound 13) all reduced activity. They also tested the dimethylated quaternary ammonium analog 14, and they found this compound possessed similar activity to 6 in whole cells but was slightly less potent against CutC in vitro.
Finally, they synthesized and tested both enantiomers of compound 6. While (S)-6 possessed similar activity to the racemic mixture, (R)-6 was at least 3-fold less active both in whole cells and in the CutC assay. Overall, this optimization effort revealed (S)-6 to be the most promising inhibitor, with a high ligand efficiency (0.98) and fit quality score (1.38). To gain further structural insights into the interactions between (S)-6 and CutC, they obtained a cocrystal structure of an unactivated enzyme−inhibitor complex.
Following incubation of wild-type CutC with 10 mM (S)-6, CutC was crystallized using similar conditions to those reported previously.24 The resulting structure determined to 2.3-Å resolution revealed nonproteinogenic electron density within the CutC active site. Based on the observed electron density they modeled compound (S)-6, which fits well in the composite omit map density. Overall, inhibitor (S)-6 bound in a similar fashion as choline, exploiting some of the same key binding interactions observed within the CutCcholine complex. Of interest, the alcohol moiety of (S)-6 is within hydrogen bond distance (2.7 Å) of Glu491, placing the C3 hydrogen within 3.8 Å of Cys489. Similar to the CH−O bond interactions observed in the CutC−choline bound structure that are involved in securing the trimethylammonium moiety of choline in place, comparable CH−O interactions are observed between the N-methyl group, C2, and C6, which are adjacent to the tertiary amine of (S)-6, and residues Asp216, Tyr506, and Tyr208, respectively. Disruption of these interactions could provide a rationale for the decrease in activity observed for compounds 9−14 where the N-methyl functionality has been substituted. Perhaps unexpectedly, the two carbons that they anticipated would mimic the choline backbone (C2 and C3) are not located in the same region of the active site; instead, (S)-6 binds in a conformation amenable for π−π interaction between C4−C5 of (S)-6 and F395. This interaction may contribute to the boost in potency observed upon introducing this functional group. The similar binding orientation of (S)-6 compared to choline is also reflected in the observation that active site residues do not rearrange upon binding the different molecules.
Overall, the structures are very similar, with low RMSD values (0.28 Å over 792 residues out of 846). Together, our inhibitor optimization studies and structural data allow us to propose that cyclic compound 6 may be a mechanism-based inhibitor. The similarity of its binding mode to that of choline, coupled with the proximity of the C3 hydrogen atom of (S)-6 to Cys489, suggests that this compound could undergo C−H abstraction to form an inhibitor-centered ketyl radical. Notably, such a radical would be stabilized by the adjacent alkene, which could derail catalysis and lead to inhibition. This proposal is consistent with the dramatic improvement in activity observed upon introducing the alkene into the cyclic scaffold. It is also consistent with improved potency for the (S) versus (R) enantiomer, as the (S) enantiomer should position the C3 hydrogen optimally for abstraction. To test this hypothesis, they synthesized an analog of 6 that is deuterated at C3 (15) and assayed it both in whole cells and in vitro. Racemic 15 displays a 3- and 2-fold loss of activity in whole cells and in vitro against CutC in comparison to 6.
This preliminary result provides support for our mechanistic hypothesis, as the loss in activity could potentially arise from a reduced rate of deuterium abstraction due to a kinetic isotope effect. An alternative mode of inhibition that they cannot yet rule out would be the oxidation of 6 to an enone followed by conjugate addition of C489 at C5. The cut genes are widely distributed among human gut bacteria and likely encode a major pathway for choline utilization in the GI tract. They, therefore, tested the potency of our best lead compound (S)-6 and our reference inhibitor BA for inhibition of TMA production from choline by additional Gram-negative (Proteus mirabilis, Escherichia coli MS69-1, Klebsiella sp. MS 92-3) and Gram-positive (Clostridium sporogenes) choline-metabolizing human gut isolates. Compound (S)-6 was effective against all strains analyzed, with IC50 values ranging from 5 to 97 μM, which is at least 20-fold more active than betaine aldehyde.
Finally, (S)-6 inhibited the conversion of choline to TMA by a human fecal suspension ex vivo with an EC50 of 60 μM, confirming its efficacy in a complex community more reminiscent of the gut microbiota. Although (S)-6 possesses a somewhat modest biochemical IC50, the small shift between our in vitro and ex vivo IC50 values demonstrates its potential utility as a tool for in vivo studies. In summary, they have discovered a cyclic choline analog that inhibits anaerobic choline metabolism by both Gram-negative and Gram-positive human gut isolates as well as a more complex gut community. Structural information and preliminary mechanistic studies suggest 6 may be a mechanism-based inhibitor that is processed by CutC to generate a stabilized radical intermediate. This finding adds another molecular scaffold to the growing arsenal of small molecules that inhibit gut microbial choline metabolism.
Discovery of a Cyclic Choline Analog That Inhibits Anaerobic Choline Metabolism by Human Gut Bacteria Maud Bollenbach, Manuel Ortega, Marina Orman, Catherine L. Drennan, and Emily P. Balskus ACS Medicinal Chemistry Letters 2020 11 (10), 1980-1985 DOI: 10.1021/acsmedchemlett.0c00005