Lovastatin lactone may improve irritable bowel syndrome with constipation (IBS-C) by inhibiting enzymes in the archaeal methanogenesis pathway

Methane produced by the methanoarchaeon Methanobrevibacter smithii ( M. smithii) has been linked to constipation, irritable bowel syndrome with constipation (IBS-C), and obesity. Lovastatin, which demonstrates a cholesterol-lowering effect by the inhibition of HMG-CoA reductase, may also have an anti-methanogenesis effect through direct inhibition of enzymes in the archaeal methanogenesis pathway. We conducted protein-ligand docking experiments to evaluate this possibility. Results are consistent with recent clinical findings. METHODS: F420-dependent methylenetetrahydromethanopterin dehydrogenase ( mtd), a key methanogenesis enzyme was modeled for two different methanogenic archaea: M. smithii and Methanopyrus kandleri. Once protein models were developed, ligand-binding sites were identified. Multiple ligands and their respective protonation, isomeric and tautomeric representations were docked into each site, including F420-coenzyme (natural ligand), lactone and β-hydroxyacid forms of lovastatin and simvastatin, and other co-complexed ligands found in related crystal structures. RESULTS: 1) Generally, for each modeled site the lactone form of the statins had more favorable site interactions compared to F420; 2) The statin lactone forms generally had the most favorable docking scores, even relative to the native template PDB ligands; and 3) The statin β-hydroxyacid forms had less favorable docking scores, typically scoring in the middle with some of the F420 tautomeric forms. Consistent with these computational results were those from a recent phase II clinical trial ( NCT02495623) with a proprietary, modified-release lovastatin-lactone (SYN-010) in patients with IBS-C, which showed a reduction in symptoms and breath methane levels, compared to placebo. CONCLUSION: The lactone form of lovastatin exhibits preferential binding over the native-F420 coenzyme ligand in silico and thus could inhibit the activity of the key M. smithii methanogenesis enzyme mtd in vivo. Statin lactones may thus exert a methane-reducing effect that is distinct from cholesterol lowering activity, which requires HMGR inhibition by statin β-hydroxyacid forms.

METHODS: F420-dependent methylenetetrahydromethanopterin dehydrogenase ( ), a key methanogenesis enzyme was modeled for two mtd different methanogenic archaea: and . Once M. smithii Methanopyrus kandleri protein models were developed, ligand-binding sites were identified. Multiple ligands and their respective protonation, isomeric and tautomeric representations were docked into each site, including F420-coenzyme (natural ligand), lactone and β-hydroxyacid forms of lovastatin and simvastatin, and other co-complexed ligands found in related crystal structures. RESULTS: 1) Generally, for each modeled site the lactone form of the statins had more favorable site interactions compared to F420; 2) The statin lactone forms generally had the most favorable docking scores, even relative to the native template PDB ligands; and 3) The statin β-hydroxyacid forms had less favorable docking scores, typically scoring in the middle with some of the F420 tautomeric forms. Consistent with these computational results were those from a recent phase II clinical trial ( ) with a proprietary, NCT02495623 modified-release lovastatin-lactone (SYN-010) in patients with IBS-C, which showed a reduction in symptoms and breath methane levels, compared to placebo.
CONCLUSION: The lactone form of lovastatin exhibits preferential binding over the native-F420 coenzyme ligand and thus could inhibit the activity of in silico

Introduction
Irritable bowel syndrome (IBS) affects as many as 45 million people in the United States, and up to 23% of the worldwide population 1 .
Depending on the region, as many as 43.3% of these patients will have irritable bowel syndrome with constipation (IBS-C) 2 . The illness affects both men and women; however, two-thirds of diagnosed sufferers are women. Studies have linked methane production to the pathogenesis of constipation and IBS, as well as obesity 3 . Methanogens -i.e. anaerobes that respire hydrogen to produce methane -are found in many habitats supporting anaerobic biodegradation of organic compounds, including human and animal intestinal tracts 4,5 . Archaea are the only confirmed, naturally occurring biological sources of methane. Methanobrevibacter smithii (M. smithii) is the predominant methanogen in the human intestine accounting for 94% of the methanogen population 3 .
The isoprenoid biosynthesis for the main cell membrane components in archaea (archaeol) relies on the same enzyme that catalyzes the biosynthesis of the isoprenoid cholesterol in humans -HMG-CoA reductase (mevalonate pathway) 6 . It has been previously suggested that statins, i.e. known HMG-CoA reductase inhibitors, can also interfere with the biosynthesis of the archaeal cell membrane and thus inhibit archaeal growth 7 . Statins, specifically lovastatin, have been shown to lower methanogenesis in human stool samples 8 and can inhibit archaeal cell membrane biosynthesis without affecting bacterial numbers as demonstrated in livestock and humans. Lovastatin is a secondary metabolite produced during fungal growth and is found in oyster mushrooms 9 , red yeast rice 10 , and Pu-erh 11 .
Humans and archaea utilize the HMGR-I isoform for isoprenoid biosynthesis 12 . Mevastatin and lovastatin were both shown to inhibit growth of several rumen Methanobrevibacter isolates in the ~10 nmol/ml range 3 . While it is believed that statins inhibit methane production via their effect on cell membrane biosynthesis mediated by inhibition of HMG-CoA reductase, there is accumulating evidence for an alternative or additional mechanism of action where statins inhibit methanogenesis directly 13 . In one case, in silico molecular docking of the methanogenic enzyme F420-dependent NADP oxidoreductase (fno) showed that both lovastatin and mevastatin had higher affinities for the F420 binding site on fno than did F420 itself. It has been suggested that lovastatin may act as an inhibitor of fno 14 . Once models for A5UMI1 and Q02394 were developed with StructFast, ligand binding sites were identified by inference from the respective PDB templates used in modeling and from the Eidogen SiteSeeker algorithm 23 . SiteSeeker looks for concave, surface features sufficiently exposed to enable ligand binding while also considering evolutionary conservation of sequence. In addition to sites identified by SiteSeeker, other sites were manually inferred within PyMOL v1.8 after aligning models and templates containing their respective co-complexed ligands. Residues on model structures with a 7Å cutoff of co-complexed ligands within the templates were exported and also processed as sites.
Ligands were carefully prepared considering different protonation states, isomers, and tautomers. We standardized charges, added missing hydrogens, enumerated ionization states, ionized functional groups, generated tautomers and isomers, and generated starting-point 3D coordinates for each ligand using BIOVIA's

Amendments from Version 2
• Added text and hyperlink related to AutoDock Vina at the end of the third paragraph of methods: "AutoDock Vina has been tested against the Directory of Useful Decoys, performing frequently better than many commercially available docking programs." • Added "dehydrogenase" in first paragraph of the conclusion.
• Added in the conclusion: "While in vitro ligand binding experiments were not conducted, the docking simulations suggest that these dehydrogenases in the main methanogenesis pathway may be possible targets." • Added hyperlinks to facilitate description of PDB targets used to model Q02394 In each case, ligand-binding sites were readily inferred from the ligand binding sites found in the respective template structures.
The modeling of sequence A5UMI1 was straightforward given its high 52% sequence homology to 3IQZ. Other templates (e.g. 1U6I, 1U6J, 1U6K, and 1QV9) were possibilities, but each had slightly lower resolutions, earlier deposit dates, and/or were in apo form. Each 3IQZ chain (A-F) was considered, given the possibility that one template-chain might offer additional or different insight into possible ligand binding locations. The Eidogen SiteSeeker algorithm identified only one site when template chains A, C, D were used, while two sites were identified in models leveraging template chains B, E, F. Unfortunately, the H4M site from the 3IQZ template was not easily inferred into any of the A5UMI1/3IQZ-based models, because 3IQZ has multiple chains involved in H4M binding.
Modeling sequences from PDB templates is done with individual chains. Quaternary modeling using models of individual chains can be very challenging. We manually modeled the H4M site as described in Figure 1. Since our aim was to dock all ligands across all possible ligand binding sites, we included the sites identified by inference (i.e. where ligands were present in templates), by the SiteSeeker algorithm run across single chain models, and by manually modeled sites as described by Figure 1. A total of 10 ligand-binding sites (Table 1) were identified across all the Q02394 and A5UMI1 models.

Ligand processing
The key ligands for this effort included lovastatin (lactone and hydroxyacid forms), F420, and simvastatin (lactone and hydroxyacid forms) and processed ligands that were found in the PDB templates used to model each sequence: 803, F42, H4M, I2C, FE9, SIM, 116, HMG, and 882. The latter four (SIM, 116, HMG, 882) were ligands found in the positive control templates for completeness.
The PDB often contains problematic ligand structures, so we processed both PDB ligands and ligands extracted from PubChem 29 for lovastatin (lactone and hydroxyacid forms), F420, and simvastatin (lactone and hydroxyacid forms). It should be noted, the PDB considers 803 as lovastatin (lactone form), F42 as coenzyme-F420, and SIM as simvastatin (hydroxyacid form), though their actual structural forms may vary depending on the PDB entry. This is why we also use PubChem structural representations for lovastatin, F420, and simvastatin.
It is well established that the β-hydroxyacid form and not the closed-ring lactone form of lovastatin is the active HMGR-binding form of the molecule 30 . Simvastatin and lovastatin are commercially available in the lactone form; they behave as prodrugs which inhibit HMGR only after the opening of the lactone ring into the hydroxyacid form 31,32 . The degree of hydrophobicity of imidazole derivatives correlates with improved activity against human methanogenic archaea 33 .
Each ligand was computationally processed in the same way prior to docking. BIOVIA's (Accelrys') Pipeline Pilot was used for this ligand preparation. First, stereochemistry and charges were standardized, then ionized at pH 7.4, then tautomers (if present) were enumerated, and finally initial 3D models were determined. AutoDock Vina explores ligand 3D conformation, so the initial 3D models were simple starting points. Additionally, ligands were processed without the above standardization, ionization, and tautomer exploration. Each ligand representation was considered in the docking runs. Ligands processed with the standardization sequenced contained the prefix "STD_," and ligands without standardization contained the prefix "RAW_." Together, these expanded ligand representations can help gauge the docking algorithm's sensitivity to the ligand's structural representation.
Docking multi-ligand variations/multi-receptor sites A total of 88 ligand variations were systematically docked into the 10 identified binding sites across all the A5UMI1 and Q02394 models for a total of 880 docking simulations. Even though AutoDock Vina achieves two orders of magnitude speed-up and significantly improves the accuracy of the binding mode predictions compared to AutoDock 4, 880 docking simulations could have taken several weeks to complete. To accelerate the effort, we requisitioned a compute cluster in the Amazon EC2 34 cloud environment for approximately three days at a cost under $60.
The docking process scores ligand conformations based on ligand conformation and ligand-to-receptor interactions within a grid box. After the 880 docking simulations were complete, we rescored all docked ligand variations against their respective full model structures. This enabled a more realistic rank ordering given possible overlap with a docked ligand and other portions of a model not represented in the rectangular box. This also served as an internal control, since rescoring was completed independently of the docking simulations.
Since it is unknown which (if any site) might actually engage the ligands of interest, we calculated the average, minimum, and maximum affinity of each ligand/variation for each of the 10 sites. The top-two sites (highlighted in bold in Table 2) were used to then rank order each ligand. Table 3 shows the rank ordered ligands using the AutoDock Vina overall score, which considers steric interactions (Gauss 1, Gauss 2, and steric), dispersion/repulsion, hydrophobic interaction between hydrophobic atoms, and, where applicable, hydrogen bonding.
Given the rank ordering in Table 3, several observations became evident: 1) Consistent with Sharma et al. 14 , the lactone form statins docked into each site with favorable site interactions (i.e. lower docking scores) as compared to F420 for the same sequence/site grouping.
2) The statin lactone forms generally had more favorable docking scores, even relative to the native template PDB ligands.
3) The statin hydroxyacid forms had less favorable docking scores and typically scored in the middle with some of the F420 forms.
4) The F420 scores were generally the lowest for each sequence/site models of A5UM1 and Q02394.  Figure 2 depicts the best scoring lovastatinlactone and -hydroxyacid poses in the A5UMI1 modeled site (top) and the Q02394 modeled site (bottom). The A5UMI1 modeled site is more spherically form fitting while the Q02394 modeled site is more elongated. The A5UMI1 site also contains a greater concentration of hydrophilic resides (depicted in cyan in Figure 2). In each modeled site, the best scoring lactone and hydroxyacid form were docked roughly in the same position with similar interactions, however the lactone form contained more favorable intermolecular feature.      and perceive additional detail depicted). Lovastatin-lactone had better AutoDock scores and more favorable calculated affinitiesdespite having fewer hydrogen bond interactions. Both ligands appear to be interacting with ARG-255, ARG-150, and GLN-153, though F420 seems to also interact with ARG-244. F420's fit is also considerably more constrained, which explains why its AutoDock Vina score is 4.4× worse than lovastatin-lactone's score.

Conclusions
Given the large number of ligand-to-site docking scenarios, we were able to observe several key trends that together suggest that statin binding is likely for the two key dehydrogenase targets in question A5UMI1 and Q02394. In most cases, the lactone form appears to have preferential binding over the hydroxyacid form and F420. And in many cases, lovastatin/lactone and simvastatin/lactone appear to have preferential binding to even the native ligands found in the PDB templates used to model Q02394 and A5UMI1. While in vitro ligand binding experiments were not conducted, the docking simulations suggest that these dehydrogenases in the main methanogenesis pathway may be possible targets. These results are also consistent with those from a recent phase II clinical trial (NCT02495623 35 ) with a proprietary, modified-release lovastatinlactone (SYN-010) in patients with constipation-predominant, irritable bowel syndrome, which showed a reduction in symptoms and breath methane levels compared to placebo. Given that the lactone form seems to preferentially bind, the next stage of the project is to identify molecules with similar features to lovastatin-lactone that also show similar or better receptor-site interaction potential.

Data availability
F1000Research: Dataset 1. Raw data for 'Lovastatin lactone may improve irritable bowel syndrome with constipation (IBS-C) by  . From modelling perspective, the paper is well-structured and it provides enough Methanopyrus kandleri information to reproduce the results, highlighting common problems involving modelling techniques (e.g. different ligand structures in databases). Considerable computational effort was put into identifying the most probable binding site as well as the active form (lactone versus hydroxyacid) of the ligands. Maybe the major drawback is the lack of results confirming the results, what currently seems difficult to in vitro achieve on the basis of the previous authors' reply. Results from the clinical trial do not fully ensure that the mechanism goes via these targets, so I would suggest putting this correlation/clinical trial into context. Apart from that, I understand that validation of all modelling results is not always possible and the in vitro authors use a proper tone to expose their conclusions.
As a minor point, some comments (e.g. target description) on the PDB entries used to model Q02394 (page 4) would be useful. The authors modeled the structure of F420-dependent methylenetetrahydromethanopterin dehydrogenase (mtd) and systematically predicted possible binding sites in order to test docking of lactone form of statins vs. β-hydroxyacid form of statins, as well as the native ligand F420-coenzyme and other co-complexed ligands found in related structures. They used AutoDock Vina for docking and scoring and their results suggest that: for each modeled site the lactone form of the statins had more favorable site interactions compared to F420; the statin lactone forms generally had the most favorable docking scores, even relative to the native template PDB ligands; and the statin β-hydroxyacid forms had less favorable docking scores. Can the authors explain why would the lactone form of lovastatin be a privileged ligand for the predicted binding sites when compared to the other tested ligands? Did the authors try to use another docking program to reproduce these findings?
Authors also suggest that the lactone form of lovastatin could inhibit the activity of the key M. smithii methanogenesis enzyme mtd in vivo. This is in agreement with the phase II clinical trial (NCT0249562335) the authors refer to. The clinical trial showed a reduction in symptoms and breath methane levels in patients treated with lovastatin-lactone (SYN-010) when compared to placebo. However, it is not clear from the clinical trial reference if SYN-010 inhibits mtd. It would be beneficial for this manuscript to add a reference that indicates the target of SYN-010.
No competing interests were disclosed.

Competing Interests:
I have read this submission. I believe that I have an appropriate level of expertise to confirm that it is of an acceptable scientific standard, however I have significant reservations, as outlined above.

Steven Muskal
Thank you for your comments and questions.
The key dehydrogenases targets considered in this work [ ] both interact with A5UMI1 and Q02394 F420. However, statin lactone v. F420 binding experiments for these targets were not in vitro conducted. In the absence of this data, our modeling and subsequent docking of the lactone statins generally show preferential docking as compared to F420 -which suggests these targets may interact with the lactone form of the statins.
We selected AutoDock Vina not only because it is open-source and readily available to other researchers, but also because it has has been well tested and found to be "a strong competitor against other programs, and at the top of the pack in many cases" (see ). Researchers with access to commercial docking packages are http://vina.scripps.edu/index.html welcome to compare the AutoDock Vina results and data provided with this paper.
Indeed, Synthetic Biologics' SYN-010 lowered breath methane and improved stool frequency in a Phase 2 Four Week Study in Patients with Irritable Bowel Syndrome with Constipation (IBS-C). This Phase 2 study did not identify the specific ligand-target interaction. However, work by Marsh -Lovastatin Lactone Inhibits Methane Production in Human Stool Homogenates ( et al. http://content.stockpr.com/syntheticbiologics/db/220/608/file/Lovastatin+in+Human+Stool-FINAL+Poster+%28ACG ) does shed some light. While a messy culture, was the predominant methanogen in the M. smithii stool. In this work, lovastatin lactone was identified as the only effective methane inhibitor. Considering other mechanisms by which statins may inhibit methanogenesis, we suggested the dehydrogenases in the main methanogenesis pathway as possibilities.
No competing interests were disclosed. The manuscript describes modeling studies suggesting that Lovastin lactone, a statin, inhibits growth of methane-forming archaeon Methanobrevibacter smithii by inhibiting an F420 dependent-enzyme involved in CO2 reduction to methane, namely F420-dependent methylene-tetrahydromethanopterin dehydrogenase (Mtd). Modeling studies had previously indicated that another F420-dependent enzyme, F420H2:NADP oxidoreductase in methanogens could be a site of inhibition (reference 14). The results are interesting, however, before indexing the following information has to be added: