Skip to main content

Bosutinib: Don’t Believe the Label!

Now here’s a worrisome thought, if you’re doing kinase research. A tyrosine kinase inhibitor in the clinic against Bcr-Abl, bosutinib (SKI-606), is also being used as a research tool in a number of academic groups. But they’re probably not using what they think they’re using.
This article has the details. The compound has a dichloromethoxy aryl group hanging off of it, and apparently someone has been making (or made one good-sized batch of) the wrong isomer. Instead of 2, 4-dichloro-5-methoxy, many commercial samples appear to be 3,5-dichloro-4-methoxy. This got noticed at first by inspection of an X-ray structure deposited in the Protein Data Bank, 3ZZ2, from a group at Oxford. A postdoc at Stanford saw that something was off, and at the same time, he was having trouble matching his own X-ray data with the known structure of the compound.
The explanation wasn’t what anyone wanted to hear, for sure. The two groups had purchased their material years apart, from different vendors. The count of vendors selling the wrong material is now up to thirteen and climbing. That link also suggests a possible earlier source of the problem: some of the commercial supplies of 2,4-dichloro-5-methoxy aniline are not the right material. Whoever made this bosutinib may well have thought that they were right on target.
Odds are, some batch of the wrong stuff has been resold through the supplier community since at least 2006 – this sort of thing goes on all the time. But the tricky part here is that LC/MS wouldn’t have told you that there was a problem, unless you had an authentic sample to check the retention times (which would have been pretty darn close, anyway, I’d guess). The mass is, of course, the same. And the NMRs of the authentic and mis-labeled stuff would be different, but not on casual inspection, for sure (same number of aryl protons). No, I would have let this stuff through, I’ve no doubt about that. Makes a person wonder what else on the shelf is the wrong material, doesn’t it?

34 comments on “Bosutinib: Don’t Believe the Label!”

  1. Virgil says:

    Golden rule in our group is to ask for both proton and carbon NMR on anything we contract out synthesis of. When it comes to Sigma et al. however, you just kind of assume it’s the right stuff, which makes this case very annoying.

  2. RTW says:

    I guess no one takes Melting Points any more, or actually compares NMR with literature to see that they have what they think they have while working through a synthesis? I blame the suppliers for not doing their do-diligence, particularly the supliers of the anilines. The 2,4 DiCl cmpd has a reported MP of 51 deg C. Unfortunately a quick google search doesn’t afford that for the 3,5 cmpd. I would however expect it to be at least a few degrees different.

  3. Anonymous says:

    I’d expect quite a difference in 1H-NMR. The symmetry for that ring is different, the two protons would be equivalent in the 3,5-dichloro-compund, thus no coupling.

  4. anana-mouse says:

    I worked for a major pharma house for over 30 years before being layed off and I can confidently say that, in last five or so years in my career, I was forced to check my starting materials from outside contract manufacturers (predominantly India, Russia and China)to make sure that what I got was indeed what was labled. There were an unfortunate number of wrong compounds which slipped through our in-house QC. I would be willing to bet that that most of what was contracted as synthetic intermeditates for our use was never properly vetted by use prior to being registered in our in-house data base.

  5. JC says:

    Hydrogenate off those chlorines & then take the NMR.

  6. Stiv says:

    @JC – you don’t need to do hydrogenation – and I disagree with Derek here – a cursory look at the proton NMR would easily differentiate the two compounds. These are not complicated molecules.
    Bosutinib would have 5 distinct signals in the aromatic region, the isomer only 4 due to the symmetry in the chlorine-containing ring. The singlet from the protons in that ring – integrating to two protons – should have set off the alarms. It’s hard to believe that no one – from all of the suppliers down to the end users – even examined the NMR data with a critical eye.

  7. LittleGreenPills says:

    @5 and 8
    The spectra are available through the link provided by 2. The spectra are indeed different, but both produce only 4 aromatic proton signals. For the authentic compound two signals are coincidentally overlapping.
    We typically collect a full NMR data set (proton, COSY, HSCQ, HMBC, and ROESY) on all compounds.

  8. stiv says:

    @9 – bad luck that, the overlapping signals.
    A simple DEPT would have easily distinguished – 4 vs 5 lines in the aromatic region – or a C13 – 14 vs 16 lines – but I don’t know that I would have gone to the trouble…… I guarantee that Levinson and Boxer will always do so in the future though.

  9. Rick Wobbe says:

    Proton NMR, COSY, ROESY, HSCQ, HMBC, DEPT… Sounds like the chemist’s equivalent of a lot of Western blots. 😉

  10. milkshake says:

    these custom synthesis companies who caused huge waste of federal grant money (by supplying a wrong compound) have been selling unauthorized samples of the drug for research purposes and they were breaking the law in the process – it perhaps happens frequently but it is actually illegal to sell someone’s else compound.
    Anybody is free to re-synthesize a patented compound for his own internal research purposes but they are not allowed to sell it for profit to third party without explicit approval of the compound owner. I presume Pfizer has composition of matter on bosutinib. So any screwup is responsibility of the custom synthesis companies and excuses about insufficient NMR documentation in the original Pfizer patents/papers will not get them off the hook.

  11. weirdo says:

    Gotta side with milkshake on this one. Tocris sells this molecule under license from Pfizer, so there is no excuse for infringement (if there ever is). You buy it from anyone else, well, you get what you pay for.

  12. TsOH says:

    According to the PKC Pharma link, the original Wyeth patents contain two different sets of NMR, both claiming to be bosunitib. If these are in fact from two different compounds, wouldn’t this be a patent problem? At the very least it would be nice to know which spectrum corresponds to the one currently in human clinical trials at Pfizer.

  13. luysii says:

    An analogous problem has occurred with cell lines. According to Wikipedia “because of their adaptation to growth in tissue culture plates, HeLa cells are sometimes difficult to control. They have proven to be a persistent laboratory “weed” that contaminates other cell cultures in the same laboratory, interfering with biological research and forcing researchers to declare many results invalid. The degree of HeLa cell contamination among other cell types is unknown because few researchers test the identity or purity of already-established cell lines. It has been demonstrated that a substantial fraction of in vitro cell lines — estimates range from 10% to 20% — are contaminated with HeLa cells. Stanley Gartler in 1967 and Walter Nelson-Rees in 1975 were the first to publish on the contamination of various cell lines by HeLa.[23]”
    So if you think you’re testing your antineoplastic against colon cancer, lung cancer etc. you may really be testing it against the humble HeLa cell.

  14. Nuclear Option says:

    “…a group at Oxford” is the Knapp Laboratory. A fountain of brilliant science, in this case like many others placed into the public domain in advance of publication. Nicely done, Stef.

  15. Secondaire says:

    This happened to a graduate school labmate. Let’s see if I remember the story…she was making some kind of heterocycle via a condensation of some hydrazide with some carbonyl compound. She had to resynthesize the lead compound for re-assay and ordered the hydrazide from a custom synthesis company. Was supposed to be the p-methyl. N months later, the numbers didn’t make sense. She looked back at the NMRs, and something seemed a little off. Anyway, cutting past the intermediary steps that ended up involving a crystallography subcontract or something…
    …She had been sent the o-methyl compound.

  16. Hap says:

    The people working with the compound might be blamed a little for not buying from Pfizer or one of its approved companies, but it seems not trivial to pick up the error at bosutinib. The people who made the compound, particularly the contract manufacturers, seem more responsible to me – if your name is on it, you ought to be pretty careful with it. Blaming Pfizer for minimal data in a patent (chortle) seems unreasonable when you couldn’t be bothered to check your intermediates – while the end product might be hard to check, the amines should be pretty distinct, both by 1H and by 13C. At that stage, the symmetry should be a giveaway that something is wrong.
    I am confused, though, as to how someone managed to make the isomeric aniline in bulk and not notice. The correct compound and the isomer can’t be derived from the same starting material – the correct aniline is (probably) derived from m-anisidine, while the isomer is derived (probably) from the p-anisidine. The isomers don’t differ much in price from Aldrich as of 2009-10 (and while bulk supplies should be cheaper, I am assuming that the price order or difference should be similar), so there doesn’t seem to be a good economic reason to use the wrong anisidine. I couldn’t find a commercial supplier (other than random contract people) for the 4-acetanisidide, while the 3-acetanisidide is available but dear. The 4-anisidide might be available by nitration of anisole and reduction and acetylation, while the 3-isomer would require a more indirect route, I think (though you might be able to get it from resorcinol by reduction, formation of the enamide with acetamide, and reoxidation). Maybe there is an economic reason to use the 4- and not the 3-, but I’m not sure.
    I can see having done this – I don’t remember ever checking my starting materials. I could have had C11, C13, or C14 diamine as a starting material, for example, and it would have been hard to tell (though greasy and symmetrical CH2s have that problem, and eventually the mass would have been off). If it’s really important (like doing expensive protein crystal structures), you probably need to check beforehand.
    I wonder if this makes the FDA even more squeamish about the proliferation of subcontracted work going into to drugs.

  17. Ben Zene says:

    A few comments
    1. Milkshake is correct, the only companies who are legitimately allowed to sell Bosutinib are those who have licensed it from Pfizer – namely Sigma, Tocris and maybe TRC. In this case I believe Pfizer has actually supplied these vendors with Bosutinib for resale purposes so the quality should be pretty much guaranteed.
    2. Many of the posts come from chemists thinking like chemists. These products are generally sold to pharmacologists and other more biologically oriented researchers who at best might think in terms of running an LC-MS but would be pretty unfamiliar with techniques such as NMR. So they are unlikely to spot structural discrepancies. And often they may only buy 1-10mg of expensive material which does not leave much for a full chemical analysis. Probably customers should ask the supplier to provide the batch specific analytical data they have on file (not just a CoA) and then ask for some help interpreting the data if necessary.
    3. Another related problem is that companies selling these kinds of products often identify new products to add to their range by reading pharmacological literature. Have you seen the standard of structure reporting in these journals? It’s often shockingly bad. Stereochemistry in particular is viewed as optional (what, carbon atoms form chiral centers?) and leaving off or transposing the odd methyl group or heteroatom here or there would be perfectly acceptable. How this kind of thing gets past review is beyond me (but that’s a whole other series of recent posts!).
    4. Congratulations to the scientists in this case who identified the problem, it was a nice piece of detective work on their part.
    Caveat emptor!

  18. petros says:

    A problem I encountered years ago when we were screening for subtype selective GPCR antagonists was that the supplier of the receptor subtypes had mislabelled two of them!. With no good pharmacological tool available then, there was no easy way to distinguish them.

  19. exGlaxoid says:

    I have seen the wrong compound put into a commercial chemical bottle several times. Once was even Aldrich, where the chemical in the bottle was reasonably pure, but not the material listed on the label. Fortunately, the MW was off, so it was easy to determine. It has become more common to receive materials that are less pure than labelled, but not often the entirely wrong compound, in my experience to date.
    I have also seen internally made compounds at big pharma with the wrong structure. One project spent months remaking a screening hit compound, only to discover the material was dead, and the LC retention time was slightly off, despite correct MW by MS. We only had a few mg in DMSO to test, and NMR was only able to show that it was not the structure previously assigned to it.
    After a fairly large amount of work, it was discovered that someone had used an isoquinoline SM rather than a quinoline SM in the original work years back, and no one had ever noticed. The original bottle was even found, amazingly, and was labelled correctly, the chemist just made a simple mistake. But it had passed QC just fine. Even the NMR was not off by much, but was clearly different from desired material. So NMR is useful, but most useful when you can compare it to a known from the same frequency NMR and same solvent.
    So it is useful to check SMs, especially once you have a molecule of interest. And it is even more important to check final product or any compound for pharma uses for complete QC, not just an LC-MS, like so many contract companies.

  20. Hap says:

    The isomeric anisidine should also give a mess of chlorination products – the o,o-dichloroanisidine should be resistant to react to form the diarylamine, and so the m,m-dichloroanisidine should react preferentially to give the observed isomer, but there could be an o,m-dichloroanisidine, and other chlorinated amines – which might also give an isomeric byproduct. The desired anisidine, on the other hand, should undergo regioselective dichlorination to give the desired dichloro compound as the major (if not sole) product.

  21. Vinylog says:

    Not sure if anyone else picked up on this, but the ligand in PDB 3ZZ2 is the *3,4*-dichloro-5-methoxy.
    This is different than the ‘bosutinib isomer’ structure at the top of the C&EN article, nor discussion of the 3,5-dichloro-4-methoxyaniline later in the article.
    It would certainly be trickier to catch the 3,4-dichloro isomer by NMR since the aniline moiety, like bosutinib proper, is asymmetrical.
    I have no idea what that actually means in the scheme of things, but it would appear that Vollmar et al. used yet a DIFFERENT isomer mislabeled as bosutinib.
    Isomers, isomers everywhere…

  22. Hap says:

    How would you even make the 3,4-dichloro-5-methoxyaniline if you were trying?
    At least it could have come from the correct anisidine.

  23. daen says:

    It is scary how much you’re dependent on your suppliers. I used to work for a company doing fragment-based screening, many which were purchased from suppliers. I now wonder if some anomalous results might not have been because of a similar problem. We’d already come across issues with custom oligo suppliers, and indeed one of my first jobs in the company was to automate QC checking of those — we’d usually get a batch of 20mers, 100 or so at a time — doing MALDI/TOF on those by hand was NOT fun. And it was surprising how variable quality was even from the same supplier …

  24. barry says:

    in this era of electronic notebooks and automated analysis, we Med.Chemists should be routinely recording 1HNMR of starting materials and linking those data to the products we submit. Those reference NMRs are a great resource for identifying contaminants in final products. And years later, you might learn that you didn’t use what you thought you used.
    Even more so for production/scale/process labs.

  25. petros says:

    Re RTW
    3,5-Dichloro-4-methoxyaniline, m.p. 77°-79°,(US4139621)
    big difference

  26. petros says:

    Re RTW
    3,5-Dichloro-4-methoxyaniline, m.p. 77°-79°,(US4139621)
    big difference

  27. @23
    I didn’t have space to address this in my article, but von Delft tells me at that resolution it’s hard to tell chloro and methyl groups apart, so they aren’t sure if they have 3,4-dichloro-5-methoxy aniline or the 3,5-dichloro-4-methoxy aniline. Also, they don’t have any more of that bosutinib left to analyze, so they’ll likely never know.

  28. simpl says:

    Active ingredients are to some extent public domain on approval. USP et al have identity tests, admittedly somewhat later, when generics become relevant. The test chosen (often IR) should be specific for likely mix-ups.
    Similarly, purchasing of samples from any source should be possible, since the patent must describe the chemical and show an advantage. Protection is coupled to sale of the medical effect; other uses, including research, are not covered per se. The cover may be more through a gentleman’s agreement, as the original supplier should be a good and cheap bulk source, and gets easier support for external researchers. Direct distribution of small samples takes real administrative effort. We have always shipped analytical and pharmacological quantities (mg) as a service, but restrictively, as shipping has become more tricky and the number of requests has risen.

  29. Vinylog says:

    Ah, thanks for clearing that up.
    Now that @24 mentions in, it would be rather trickier to make the 3,4-dichloro compound, so in all likelihood, it’s the 3,5-dichloro-4-MeO that was used.

  30. darwinsdog says:

    I suppose this is the karma of the universe for those occasions when ‘we’ (in the western big pharma sense) have contracted a synthesis to an overseas lab and drawn the structure incorrectly. I actually know of an example of this (OK only one time vs. say ten times having found the wrong reagent in a commercial bottle) from a company I was formerly at and the poor Indian CRO eventually completed the synthesis months overdue all because the person that drew the structure put a “C” where a “N” ought to have been and nobody was double checking it.

  31. Steve Hayward says:

    At ACD/Labs, we thought it would be interesting to see if an Automated Structure Verification (ASV) system such as ours would be able to flag the original bosutinib structure for review and/or pass the isomer when compared with experimental NMR data. We are working with Phil Keyes of Lexicon Pharmaceuticals, and you can see our initial results here:

  32. A follow-up has been posted on our blog, detailing how a Computer-Assisted Structure Elucidation program would deal with the NMR data for the bosutinib isomer.

Comments are closed.