Thursday, April 16, 2015

Spot-Checking Antibacterials in PLoS One

Update (April 17) - Commenters here and at Derek's place have set off alarm bells to potentially doctored spectra. Stay tuned...

Ever looked at a molecule and thought something was

A little while ago, I browsed through the latest edition of PLoS One, the open science journal. I don't often go there for chemistry publications, but curiosity struck, so I typed, "antibacterial NMR" (I think) into the search box. Up came this paper, from 2013, announcing:
  1. A novel alkaloid, xinghaiamine A, isolated from a marine bacterium, that...
  2. Had novel structural features, such as a sulfoxide and acenaphthylene ring, rarely seen, and...
  3. That showed decent lead activity against MRSA strains (2.7-5.5 uM MIC).
First thought: "Wow!" Second thought: "If this is so great, why is it buried in PLoS One?"

Then I happened to look at the proposed structure for this potential panacea:

Jiao, Zhang, Zhao, Hu, Suh, PLoS One, 2013

[Rubs eyes in disbelief].

Sulfonylated "ladderane" substructure.
A (4,7) disubstituted acenaphthylene bridge.
A previously-unknown-to-science (and virtually unknown to SciFinder) 4-4-5-6 ring system.
Can that really be correct?

Naturally, I made a model of one "half" of the presumed dimeric molecule:

Xinghaiamine A, pseudo-axial conformation,
assumed diastereomer of the (4-4-5-6) ring system.
The red circles are supposed to be bonded together...
I've taken apart and reassembled this molecule a few times, and can't seem to access any diastereomer that 1) looks energetically minimized, and 2) can actually have the acenaphthylene bridge the way the authors specify without "strain energy" popping out the plastic bonds.

Perusing the analyticals in the Supporting Information, I'm puzzled by a few more bits:

Jiao, Zhang, Zhao, Hu, Suh, PLoS One2013

The IR, briefly addressed in the text as suggestive of "[the] presence of a sulfoxide functional group" due to the 1082 cm-1 band, but contains peaks suggestive of:
  • alcohols, primary amines: 3383 cm-1 band
  • heteroaromatics: 743, 846 cm-1 band
The published 1H NMR has no peak labeling or integration (tsk, tsk).

That the COSY shows only one set of strong H/H correlations out of the 6 posited protons on the conjugated ring, which should not fit with the proposed structure. Nor should the seeming high symmetry of the ring fit with the proposed substitution.

Readers, what do you make of this? What kept this out of a more mainstream chemistry journal? 
Do you agree with the assigned structure? (Am I just shouting in the wind here?)

April 17 - Uh-oh...check out these anonymous commenters' finds on Derek's blog:
Potential editing of the H/H COSY

...and the HMBC


  1. I am guessing that if they did try for a better journal, the reviewers would have expressed sincere doubts and requested further evidence which the authors presumably could not provide. Hence they went down the tier until they found a journal which wouldn't question it. There are many isolation papers that get it wrong, but most aren't as obvious as this.

  2. So, some quick MM structure optimizations give us a truly wonky looking beast (feel free to post the pictures if you'd like, SAO) which would release about 720 kJ/mol upon breaking that C8-C25 bond. For those following at home, this molecule would definitely be what the experts refer to as "thermally unstable." Probably a pretty good hand warmer, though.

  3. With the IR analysis....not sure where the C=O peak at 1725.28 comes from (2dps really?). Also no 1510 peak for aromatic C-C or aromatic C-H @ just above 3000. Plus the 1080 peak looks much more like a C-O.

    If I were to guess this is oxidized R1 rather than anything exotic. But what wold I know I'm just a lowly polymer chemist.

    Red Jim

    1. That absorption at 1725 looks way too small for a carbonyl though.

  4. Any structure is possible with microsoft paint!

    HMBC peaks @ 8.2

    Proton NMR peak at 1.45

  5. Interesting find! If you take a look at the metadata from the images it appears that the NMRs were "created" using Adobe Photoshop CS2

  6. In response to the above, the use of Adobe Photoshop CS2 could have been simply to scale the image size down, or crop out superfluous data eg. aquisition parameters. I wouldn't say that is evidence to indicate foul play.

  7. If the image were in the main paper, that might make sense, but I think the key images are in the SI, and there doesn't seem to be a good reason to need to reduce size or cut parameters out of the image there. (If they were generated on paper, perhaps, although I would assume the scanned images would fit or could be rotated to fit standard page sizes.) I would probably want the NMR parameters to be present in the SI image, in case someone wanted to reproduce the work. It might not be evidence of foul play, but with all the other funny image pieces, it does not improve the conclusions.