Metabolite stereochemistry and salt forms: L-carnitine, peptides and statins #1114

Three groups of metabolites had structures that did not match the compound the model means. L-carnitine carried the D-carnitine structure and ids. Most peptides had D or undefined residues. The statins were written as their calcium or sodium salts, some of them as two-molecule dimers. This issue lists what was changed and why, and the points that were deliberately left open.

1. L-carnitine is (R)-carnitine

MAM02348 "L-carnitine" (BiGG crn) had the (S)/D-carnitine structure and ids in every field except HMDB. All human carnitine enzymes in the model (CPT1/2, CRAT, BBOX1) use L-carnitine, which is (R)-carnitine. For example, Rhea writes the CRAT reaction with (R)-carnitine (RHEA:21136). The model has no second carnitine that could be the D form.

Formula and charge (C7H15NO3, 0) are unchanged.

Carnitine-shuttle reaction ids. Reactions that use carnitine linked to MetaNetX reactions containing (S)-carnitine. Each id was replaced by the MetaNetX reaction whose equation is the same except that it has (R)-carnitine (MNXM1105736 -> MNXM1105737).

• 94 ids that were already in v2.0.1 are corrected here. Seven of those had no (R)-carnitine counterpart in MetaNetX, so their id was removed: MAR02605, MAR02607, MAR02759, MAR02761, MAR02790, MAR02897, MAR04634.
• The ModelSEED ids that pointed to the D-carnitine form were also replaced: rxn15328 -> rxn01726 on MAR02591, MAR02593, MAR02780, MAR03030 (carnitine acetyltransfer), and rxn15330 -> rxn01727 on MAR06982 (BBOX1).

Examples:

Left open: MAR06598 and MAR06600 ((11Z)-octadecenoylcarnitine, CPT1/CPT2) keep MNXR206165 because the 18:1 acyl-CoA and acylcarnitine reactions are handled in the open fatty-acid issues.

2. Peptides are all-L

Rule. 221 peptide metabolites (di- and tripeptides, gamma-glutamyl peptides, glutathione conjugates and the THF polyglutamates) had SMILES with D or undefined alpha-carbons. The structures came from Recon3D/VMH and were copied into MetaNetX. The names give L-amino-acid residues, and the model hydrolyses these peptides to L-amino acids. Each structure was rebuilt with L residues, keeping the model formula and charge. PubChem and MetaNetX ids were then set to the entries whose InChIKey matches the all-L structure (skeleton and stereo layers). MetaNetX ids of the D-configured entries were removed.

Examples:

Renamed peptides. For three peptides the name gave an amide residue (Asn/Gln), but the formula, BiGG id and hydrolysis reaction all give the acid (Asp/Glu). The name had been expanded wrongly from the BiGG abbreviation. The metabolites and their exchange reactions are renamed:

Left unchanged on purpose:

• D-alanyl-D-alanine (MAM03164) stays D,D.
• The peptides changed by fix: curate metabolite structure, formula and identity inconsistencies (#1085) #1104 were skipped.
• Ids with undefined stereo (not D) were kept.
• No PubChem entry exists for the all-L form of MAM02105, MAM03458, MAM03470, MAM03871, MAM03894 and MAM04014, so their PubChem id is left empty.
• The generic MAM00097 "(5-L-glutamyl)-L-amino acid" still carries the gamma-Glu-Ala structure and id.
• Spelling slips in peptide names ("Isolecyl", "Metheonine", "Cystyl", "Glutamatsyl") are cosmetic and were not changed.

3. Statins: free molecules at physiological charge

Salts and dimers. The rosuvastatin and pitavastatin metabolites were hemicalcium salts: two statin molecules plus Ca2+ in one formula (for example, rosuvastatin C44H54CaF2N6O12S2). The pravastatin, fluvastatin and cerivastatin metabolites were sodium salts. In solution the salt dissociates, and all other carboxylic acids in the model, including the atorvastatin, simvastatin and lovastatin acids in the same block, are written as the free anion. The dimers also forced 2:1 coefficients into their reactions. 45 statin metabolites were therefore set to a single molecule: open acids as the carboxylate (-1), lactones neutral, and CoA esters at -4 like every other acyl-CoA.

Where the structure could be resolved from the name and literature, a SMILES and PubChem id were added. The glucuronides of rosuvastatin (MAM03919, PubChem 71752114), pitavastatin (MAM03912, PubChem 131881387) and fluvastatin (MAM03581) are written as 1-O-acyl glucuronides, because in the model they lactonise and only the acyl glucuronide can do that.

Cerivastatin was one CH2 short. Cerivastatin is C26H34FNO5 (PubChem 446156), and the model's cerivastatin family was one CH2 short throughout. M-1 (O-demethylation), M-23 (hydroxylation of one methyl of the isopropyl group) and M-24 (both) are defined relative to cerivastatin (PMID:10976657, review of human pharmacokinetics). Every member gets +CH2: cerivastatin MAM03514 C26H33FNO5 -1, M-1 MAM03515 C25H31FNO5 -1 (PubChem 29977428), M-23 MAM03517 C26H33FNO6 -1, M-24 MAM03518 C25H31FNO6 -1.

Cerivastatin M-22 and M-31 are lactones. M-22 (MAM03516) and M-31 (MAM03519) are made in the model by "spontaneous conversion" of the M-1 and M-23 glucuronides and are opened again by hydrolysis. They are the neutral lactones of M-1 and M-23: M-22 C25H30FNO4 0, M-31 C26H32FNO5 0. The M-1 and M-23 glucuronides (MAM03523, MAM03524) are acyl glucuronides (C31H39FNO11 -1, C32H41FNO12 -1).

Fluvastatin 3R,5S. The fluvastatin acids had the 3S,5R SMILES. Fluvastatin and its lactone are 3R,5S, so fluvastatin MAM03579 (PubChem 37888681), 5-hydroxy-fluvastatin MAM03285 (PubChem 71777519), 6-hydroxy-fluvastatin MAM03305 and N-desisopropyl-fluvastatin MAM03547 get the 3R,5S structure. MAM03305 and MAM03547 have no PubChem id, because the PubChem entries found for them are the other enantiomer.

Reactions rewritten

All statin reactions balance with the new formulas. Exchange and transport reactions (1:1 in the same metabolite) needed no change.

GPR on rosuvastatin glucuronidation. MAR12829 had no GPR and now has UGT1A1 or UGT1A3 or UGT2B7, with reference PMID:26412035. Only these three UGTs contribute significantly to lactonisation, via the acyl glucuronide, of six statins including rosuvastatin (PMID:26412035, recombinant human UGT subtypes). The pitavastatin reaction MAR12812 already carries these UGTs (PMID:12519692).

Open points (not changed)

• Drug reactions that still use SAM/SAH or H2O2 to balance. These come from the Recon3D drug module and are balanced, but they are not P450 chemistry: MAR12535 and MAR12539 (cerivastatin O-demethylation/hydroxylation, SAH -> SAM and H2O2), MAR12564 (fluvastatin N-desisopropylation, 3 SAH -> 3 SAM), MAR12569 (pravastatin "desacyl dehydro", 5 SAH -> 5 SAM), MAR12869 (pravastatin triol, H2O2), and MAR12808/MAR12809 (fluvastatin propionic-acid branch).
• Metabolites whose identity is not verified. Pitavastatin M-3 (MAM03915) and M-13 (MAM03914) are not defined in the abstracts of PMID:12519692 or PMID:10077432 (rat, rabbit and dog bile), so they have formulas only. The fluvastatin des-isopropyl propionic-acid branch (MAM03892, MAM03893) does not match a C17 structure: the formula of MAM03892 was only made consistent with MAR12809. For the fluvastatin threo isomer (MAM03985) the enantiomer is unknown. Many pravastatin, fluvastatin and cerivastatin metabolites have no SMILES because the position or stereochemistry cannot be resolved from the name.
• Fluvastatin was not among the statins tested in PMID:26412035, so the evidence for its acyl glucuronide (MAM03581, lactonised in MAR12857) is indirect.
• MAM03478 "atorvastatin-ether-glucuronide -G1" has the formula of a di-glucuronide.

Implemented in #1107. The carnitine reaction ids added since v2.0.1 are corrected in #1105 (see the comment on #967).