Anyway, the latest nomenclature problem I came across concerns methylated histamine metabolites. I was under the impression that the major inactivation product was N-tau-methylhistamine, generated by the action of HNMT. However, some papers refer to this as 1-methylhistamine, others as 3-methylhistamine and there's even a mention of N-gamma-methylhistamine (thankfully the alpha amino group on the sidechain doesn't seem to cause any further problems). I turned to NCBI's PubChem, believing this to be fairly definitive but there I found that both ring N-methylated derivatives had "tau" as a synonym. My notion that N-tau-methylhistamine was the only methylated metabolite was destroyed when I saw that histamine was also a substrate for INMT and yielded the alpha-N-methyl and "non-tau"-N-methyl products. Mercifully I came across an IUPAC definition page [PDF] describing the labeling situation. It would appear that "tau" is actually tele ("far") and refers to the nitrogen furthest from the sidechain. The other ("pi") is actually pros ("near"). These are N-3 and N-1, respectively, in the formal nomenclature. I can certainly see why authors preferred to talk about N-tau-methylhistamine than N3-methylhistamine as there is less scope for mistakes.

I'm going to continue to hang on to my over-inclusive list of small compound (but not necessarily xenobiotic) methyltransferases;

• O-Methyltransferases
• COMT (soluble and membrane-bound forms)
• HIOMT
• N-Methyltransferases
• INMT
• NNMT
• PNMT
• HNMT
• S-methyltransferases
• TPMT
• TMT

I believe that I have already mentioned that my mathematical skills are fairly limited. It was thus a surprise to realise that you cannot calculate a unique fraction unbound when provided with Ks, [L] and [P]. Instead you have to make the approximation that the concentration of empty binding sites is not appreciably affected by the presence of ligand, then;

fu = Ks/([P] + Ks)

I came up with an alternative approximation which is less error-prone than this;

fu = Ks/( ([P]-[L]) + Ks )

This is version is always better for any ligand to binding site ratio of greater than about 1 : 1.2. I have never seen this approximation used in the literature. This is probably because in most studies binding sites are usually present in great excess (although, since protein binding is not my specialty I could be very wrong).

Related to this, I have adjusted my model to collapse the hysteresis loop under the baseline conditions and can now see the loop opening again when I perturb the system. Are there accepted terms for this phenomenon? Hopefully they don't involve "hysteresis hysteresis" or "retro-hysteresis" or somesuch.

This barrier to switching between disciplines, even those that are closely related, is widely recognised and is best crossed with a guide familiar with the terrain on both sides. I have seen colleagues in the same quandary as me when they come across new terms in drug metabolism, such as "phase III". This concept arose when the role of transport in the excretion of metabolites was appreciated and reinforced the notion that true detoxication hadn't taken place until the xenobiotic (either as parent drug OR metabolite) had left the cell, organ and body. So there was phase I (functionalisation), phase II (conjugation) and phase III (metabolite transport). My argument has been that transport should be termed "phase 0", as xenobiotics will be generally exposed to transporters BEFORE they are exposed to phase I (or phase II) enzymes. For some reason my idea doesn't seem to have caught on...

Trying correlations between metabolic stability and physicochemical properties (measured or calculated) is common and easy to perform but the mileage varies considerably. I find it to be most useful to steer within a series as more global comparisons often blur the trends together. I have seen better correlations between properties such as PSA and ClogP and pharmacokinetic parameters, especially as the compounds become metabolically more stable.

To best steer chemists towards metabolic stability you need to identify the structural changes that make the most difference. Ideally this means comparisons between pairs or larger groups of similar compounds (groups that can ideally be represented by Markush structures). Picking those groups autimatically is the crux of the problem. I am now beginning to really appreciate how much work is entailed in clustering compounds by structure. The most common methods create bit strings ("fingerprints") or key lists from moiety assignments and then align them, in apparently much the same way that protein or nucleic acid sequences are aligned. The choice then depends on the keys (e.g. MACCS, Daylight), the definitions of similarity (e.g. Tanimoto, Euclidean) and the clustering techniques (e.g. unweighted, weighted, Ward's). I have been trying all of the ones available to me and have been surprised at the differences in the results. I was also surprised (and a little disappointed) that some pairs of very similar compounds I had spotted with the naked eye failed to be clustered by any of the search combinations. I am talking to our computational chemists about alternatives, such as shape-based similarity searching (e.g. Tversky). I will be very interested to see how things work out. In the meantime I am putting together some Perl scripts to help automate finding clusters or spikes of stability and instability. I need to be able to build for the long term as these challenges can only get more difficult as more compounds accumulate (and I am charged with interpreting data for more and more projects). At this stage it is still fun. I hope that it stays that way.

I'll leave discussions on interpreting in vitro covalent binding data for another time.

Anyway, I am a fan of the columnist, John Dvorak, and regularly read his blog and PC Magazine column. I was somewhat surprised to read his recent post about big pharma supposedly ignoring a heroic little Australian company who supposedly has a miracle cure for the flu. Unfortunately Mr. Dvorak appears to have forgotten to check for signs of bogus science (especially signs 2 and 6). A little research would find that zanamivir is one of a class of compounds (neuraminidase inhibitors) that has been around for some time and were once thought to have the potential to be a real flu killer. Oseltamivir (Tamiflu) is another drug in the same class. There are also a couple of compounds (amantadine, rimantidine) targeting the virus' M2 ion channel. Unfortunately flu virus seems to find it remarkably easy to develop resistance to all of these agents and, as one of the comments to the blog post notes, the window of opportunity for treatment is also very narrow. There's also no reason to invoke conspiracy theories as the only current competitor to a potential anti-flu agent is a vaccine and, as clearly illustrated this year, it is not as if the market is saturated with flu vaccine manufacturers. Considering the size of the market and the profit potential, if Relenza-Biota Holdings had a blockbuster in the wings they would have been bought by one of the pharma giants. As with any such broad target, there's plenty of need to go around.

If anyone else is reading this then best wishes for 2005 (or 4703).

Anyway, as a comparative latecomer (1992) to the wonderful world of transporters I am glad to see that things are becoming a little more settled. As genome projects reach fruition and we can answer the question of "just how many transporters are there in this family?" the divisions into superfamilies and families can become a little more confident. A prime example of this is the OATP superfamily as this used to occupy SLC21, one of the >40 families in the SLC solute carrier superfamily. Unfortunately a gene family wasn't big enough to contain all of the members so now they have their own SLCO superfamily. The downside is that scientific presentations are now frequently accompanied by sidebar explanations intended to remind the audience that "this is SLCO1B1, which used to be SLC21A6, and it encodes OATP1B1, which most of you know as OATP-C, but a few of you call OATP2 or LST-1". My guess is that protein will remain OATP-C in most people's minds for some time to come. The chances of MDR1 becoming referred to as ABCB1 seem really slim and will have to wait for the pioneers in the transporter field to retire. Until that time I will have to carry my drug transporter nomenclature cheat sheet with me to all meetings.

• Cytochromes P450: You could almost argue that CYP1A2 is an "amine oxidase" because of the major role in aromatic amine (especially mutagenic aromatic amine) N-hydroxylation. There are also other P450 enzymes that are not bad at this kind of reaction.


• Flavin-containing monooxygenases: These are great at creating N-oxides (as well as S- P- and Se- oxidation) so they deserve the "amine oxidase" label as well.


• Monoamine oxidases: As the name suggests both MAO A and MAO B are professional amine oxidases. The substrate specificity used to seem to be a little narrow to be truly interesting but the finding that MAO B can metabolise some amide derivatives of valproic acid suggests that it cannot be so easily summed up.


• Polyamine oxidase: This other FAD-containing enzyme is another that has largely been ignored in xenobiotic metabolism but the report that it can metabolise milacemide again suggests that it may occasionally cross over from working with endogenous substrates.


• Semicarbazide-sensitive amine oxidases (SSAO): These are the copper-containing enzymes with the weird topaquinone group. AOC1 is diamine oxidase or histaminase and is mainly expressed in the kidney. AOC2 is the form found in the retina. AOC3 is also known as VAP-1 (vascular adhesion protein) and appears to be the source of plasma SSAO activity. The method for release from endothelial cells has yet to be fully elucidated but it might involve a metalloproteinase. Interest in this group of enzymes is moving away from roles in metabolism and more towards functions in pathological consitions such as diabetes and heart failure.

I think I had better leave non-P450 enzymes for another commentary.

Our latest molecule of interest had a thioether moiety. CYP3A seemed to love that and initially metabolised it to the sulphoxide. This lost the therapeutic activity but had no effect on an annoying off-target activity. The pharmacokinetics of the metabolite were very similar to parent and the plasma concentrations tended to reach a level equivalent to, or higher than, parent. CYP3A had another go at the sulphoxide and converted it to the sulphone. This actually increased the off-target activity. Plasma concentrations were still pretty substantial, reaching about 50% of those of parent. Finally CYP3A got it right and hydroxylated a group at the other end of the molecule and it rapidly disappeared from sight. Thankfully when they were shown these data the chemists agreed to remove the sulphur for their follow-on series. I keep telling them to get rid of nitrogen and oxygen as well but for some reason they claim they can't get by without them. Perhaps this is just as well or we'd have another "octafluoro-chickenwire" series to deal with.

Anyway, we saw an interesting one last week in samples from mouse hepatocytes. The mass was +162 and the parent compound contained a nice juicy N-methyl group. Our initial guess was demethylation + glucuronidation. Unfortunately the supplementary evidence didn't support this as there was no cleavage back to demethylated parent and the characteristic fragmentation of the glucuronic acid moiety (e.g. loss of water and carbon dioxide) didn't occur. Having learned a lesson we checked higher masses and there was absolutely no sign of a mass consistent with a glutathione conjugate (with or without demethylation). Since the sample was from hepatocytes we wondered if there could be quantitative cleavage by GGT to the cysteinyl glycine conjugate of the demethylated metabolite. This is unlikely but in fact Jösch and co-workers (1998) report that the liver can take a glutathione conjugate all the way through to a mercapturic acid derivative (who needs kidneys!). So this was unlikely but possible. We were resigning ourselves to purifying enough for NMR when my associate found a literature precedent for metabolism of the subgroup we suspected was important in our molecule. He ordered some new cofactors and tried an in vitro conjugation reaction. Bingo! The metabolite turns out to be the direct N-glucoside. This route is comparatively rare in humans, rats and dogs etc. but mice seem to favour it a lot. Now that my eyes have been opened, the more I look in the literature the more I see that mice are associated with glucosidation of some series of compounds. This is not a fact of which I was previously aware. I am familiar with the other R.T. Williams era rules ("cats don't make glucuronides", "pigs don't make sulphates", "dogs are poor acetylators" etc.) but there's always one more rule to learn

I find it fascinating that the same UGT enzymes that catalyse glucuronidation also catalyse glucosidation. The odd thing is that the aglycone appears to determine/bias the nature of the conjugate, so for some compounds the mice are making glucuronides instead even though the cofactors are presumably still present at the same ratio. One fact I haven't been able to glean from the literature is the relative concentrations of UDP-glucose and UDP-glucuronic acid in the endoplasmic reticulum lumen. It seems that two separate transporters are responsible for their presence there but I don't know enough about the kinetics to model what is happening.

Human FMO6: This gene passes the simple functional test in that it can be transcribed and doesn't contain premature stop codons or any other nastiness. However there is almost certainly no active FMO6 protein generated due to horribly inefficient RNA processing generating all sorts of odd transcripts. This was originally reported in 2002 by Hines and co-workers and, more recently, Hernandez et al. (2004) have suggested that FMO6 be reclassified as a pseudogene. Unless bioinformatics programs become very good at the quantitative assessment of splicing then gene/pseudogene distinctions will have to be made at the functional level rather than simply relying on predicting inactivating truncations.

CYP2D7: This was originally classified as a pseudogene as the genomic sequences available indicated that no active protein should be produced. This changed earlier this year when Pai et al. found that a frameshift mutation could result in a functional transcript being generated by tissue selective splicing in the brains of half of the individuals they examined. This may eventually lead to CYP2D7 being reclassified as the 58th human cytochrome P450 enzyme. This expression of CYP2D7 in the brain could also be functionally important as CYP2D7 is more efficient than CYP2D6 at converting codeine to morphine.

We'll have to wait and see if FMO6 is "downgraded" to a pseudogene and/or CYP2D7P loses its 'P'.

The reference in question is Burk et al. (2002) and is a fascinating read. The mRNA levels are still much lower than those for CYP3A4 but considering the interesting substrate specificity of CYP3A7 I think the relevance will be toxicologic rather than pharmacokinetic.

Mother nature is always fascinating.