Friday, November 16, 2012

On the trail of bosutinib isomers


“Huh, that looks a bit funny.”

The publication of a paper by Nicholas Levinson and Steven Boxer in PLoS ONE in April 2012 hit the biochemistry world like a bombshell.  Its deceptively innocuous title “Structuraland spectroscopic analysis of the kinase inhibitor bosutinib and an isomer of bosutinibbinding to the Abl tyrosine kinase domain” hid a very worrisome discovery: two distinct chemical compounds were being sold by chemical suppliers as “bosutinib.”  In one fell swoop, dozens of research papers and experimental results were cast into doubt.

Bosutinib (developed by Pfizer) is a selective kinase inhibitor, currently in clinical trials as a chemotherapeutic agent.  Levinson, from Stanford University, was working on a crystal structure of bosutinib bound to a tyrosine kinase called Abl.  He noticed a problem with the electron density around the area of the aniline ring – the expected chlorine at the 2 position seemed to be missing from his electron density map, and instead seemed to present at the 3 position.  Somewhat worried about this, Levinson checked the crystal structure of botsutinib bound to serine threonine kinase 10, recently deposited in the Protein DataBank by Stefan Knapp and coworkers from England’s Oxford University.  Upon closer inspection, their electron density data showed that the 2-chloro atom on the aniline ring was missing, and a chlorine atom was instead located in the meta position.  The authors had noted in their title that the compound was “radiation damaged”, but Levinson was now convinced they were afflicted by the same problem he was seeing.

“What is bosutinib?”

Levinson and his supervisor Boxer immediately subjected their “bosutinib” sample to a battery of tests.  Multi-dimensional NMR experiments quickly revealed that not only was the chlorine atom at the 3 position instead of the 2 position, but the other chloro and methoxy group seemed to be switched as well.  What they had been working on was not bosutinib, but in fact a bosutinib isomer.  The difference is subtle – the isomer has the same mass and would give the same elemental analysis results.  It also has some kinase inhibitory activity, so even biological activity assays could be fooled.  The key to distinguishing the isomers is either X-ray, or detailed NMR analysis (preferably 13C NMR) which would reveal the symmetry present in the aniline ring in the bosutinib isomer, but neither analysis is routinely done on reagents purchased commercially.

Levinson and Boxer notified the company who supplied the wrong isomer – LC Laboratories, a subsidiary of PKC Pharmaceuticals – who immediately launched a comprehensive investigation.  What they uncovered was a wide-spread – indeed, world-wide – problem that probably went back as far as 2006.  PKC Pharmaceuticals gathered unequivocal physical, HPLC, TLC, and spectroscopic evidence that at least two different compounds have been, and possibly still are, being offered for sale under the name bosutinib by at least 18 different biochemical suppliers.  What is worse, PKC found some other spectroscopic discrepancies that may indicate the existence of yet a third isomer.  “What is bosutinib?” is now a real and pressing question, as it impacts on many researchers whose results from studies based on the wrong isomer may need repeating.  One thing that does seem certain is that the compound being tested in clinical trials IS the real thing – Pfizer makes it and tests it in house, and they insist that no isomeric material has ever been administered to humans.

“How deep does the rabbit hole go?”

The only reasonable explanation for the production of the isomeric material seems to be if the wrong aniline precursor was used in the synthesis.  This could be due to choosing an incorrect synthesis of the anilinic isomer, purchasing the wrong isomer, or purchasing the right isomer but receiving a wrong compound.  The latter case is probably most worrisome, and indeed, PKC has found that at least one incorrect compound is being sold as the required aniline.  They are currently testing samples of “2,4-dichloro-5-aniline” obtained from 28 worldwide vendors in order to locate a possible company that may be selling the wrong aniline to bosutinib producers.

Overall, instances of incorrect isomers being sold in the marketplace are very rare – bosutinib may be only the second example. Yet the implication that the problem goes back to an incorrect precursor is troubling, not just because many more bosutinib analogues are being generated by the medicinal chemistry community and may be propagating further structural errors.  Other groups may have purchased the incorrect aniline for their own syntheses, leading to structural errors in molecules of a completely different class.  And unfortunately there is no simple mechanism to alert the wider scientific community of this problem.  In the end, the bosutinib saga serves as a warning to researchers never to take the identity of purchased reagents for granted.

This essay was shortlisted for the Royal Society of Chemistry Science Communication Competition, 2012.