Chemicals vs. drugs (Part 1): The end of bacitracin / the buzz around halicin

This newsletter is the first of a series — here are the links to Part 2, Part 3, Part 4, and Part 5. 

Dear All,

Today’s theme is a 3-part journey that could be subtitled “Many are called, but few are chosen,” or “It’s easy to kill bacteria — steam, fire, and bleach are consistently effective — but those aren’t drugs!”

Our first stop is with a very old compound, bacitracin. Approved in 1948 because of its activity vs. S. aureus and thus long before the 1962 FDA amendment that standardized our approach to drug evaluation (link), it fell from regular usage due to its substantial toxicity. But, an IM (intramuscular) formulation of bacitracin remained on the list of FDA-approved drugs until it was withdrawn on 31 Jan 2020 (link). To learn more about the saga of IM bacitracin, I have previously written a 3-part series of newsletters discussing the FDA Advisory Committee held last year before finally retiring the IM formulation. Go here to find the first newsletter and follow the links to read parts 2 and 3 (or, click here and here).

So, is this a drug or just another toxic chemical? Well, it depends on your point of view. I think there was a time in the 1950s when the paucity of other choices and the rise of penicillin-resistant S. aureus would have made IM bacitracin better than nothing and hence, in context, it was a useful drug. But, today we have many other choices for this organism and hence bacitracin has been appropriately deleted from the systemic formulary. Its human use is now limited to topical application and it also has important veterinary uses. 

Colistin (IV) is obviously a similar story. Despite its clear toxicities, we brought it back in to clinical use because other agents were failing. But, and as I’ve recently argued (link), it’s now time to put it back into the category of “not a drug.” It may still have inhaled uses, but IV use should come to a screeching halt!

Our second stop is with an article that just came out in Cell by Stokes et al. (link). Entitled A Deep Learning Approach to Antibiotic Discovery, the authors sought to use machine learning to identify antibacterial activity in compounds where such activity would not have otherwise been anticipated. You should read the paper yourself, but here’s the short version:

• As a metric for possible antibacterial activity, the authors measured an E. coli MIC in 1,760 approved FDA drugs + several hundred additional natural products. 
• These data were used to build a predictive model that was then applied to 6,111 molecules in various stages of development as human drugs.
• Hits were obtained and one of them, c-Jun N-terminal kinase inhibitor SU3327, was identified as having strong antibacterial activity.
• Further work with SU3227 (renamed halicin by the authors, by inference an homage to HAL from 2001: A Space Odyssey, link) showed that SU3227 was active against a wide range of bacteria (Acinetobacter, Pseudomonas, and others) as well as against M. tuberculosis.
• The authors showed that SU3227 had activity when applied topically in a murine wound infection model.
• And finally, the authors report that their algorithm has found additional candidate antibiotics in a large compound library.

So, is this a drug or just another toxic chemical? Well, and with the caveat that I’d be happy to be proven wrong, my guess would be that SU3227/halicin is not a drug. At a very simple level, compounds that kill nearly everything often turn out to kill indiscriminately and hence are too toxic to use as systemic drugs. Consistent with this, c-Jun N-terminal kinases are involved in many critical biological pathways and work on inhibitors appears limited by the need to find highly selective variants (link). Perhaps SU3227/halicin could be developed for topical use, but that’s a space with very little unmet need. 

Putting it all together, I think it is premature to call SU3227/halicin a drug or to think that this approach to discovery will dramatically shorten the overall development cycle for new antibiotics. The paper by Stokes et al. is truly fun to see and their insights might open up new chemical space for an antibiotic that will eventually get approved, but the work to be done from the point of showing in vitro activity is enormous: toxicology, manufacturing, formulation, in vivo pharmacology, and a lot more. It would be really cool if finding a new molecule could be reduced to something done in silico via a gadget from Star Trek but that is something for the distant future.

As related notes:

• For those who want to dig more deeply into machine learning, I suggest the excellent blog by Derek Lowe in Science Translational Medicine (link).
• The CO-ADD project (https://www.co-add.org/) offers a free in vitro screening service for compound collections and could be thought of as the bench-based version of the search conducted in the halicin paper. Compound libraries are screened at no cost against a key panel of susceptible and drug-resistant bacterial, and fungal pathogens. Check it out! 22 Jan 2020 addendum: CO-ADD has recently published a paper (Frei et al., link) in which they observe that metal-containing compounds have an unusually high validated hit rate.

In summary, antibacterial discovery is a tricky area and it makes you realize that the antibiotics we do have are true marvels! And while you’re thinking about drugs vs. chemicals, let me remind you of two prior newsletters that approach this topic:

• The shorthand concept of Lipinski’s Rule of 5 (Ro5) is an interesting tool for making guesses about whether a molecule is likely to play well with biological systems and hence possibly become a drug (link).
• Lynn Silver, one of the true masters of antibacterial discovery, has over time written a series of must-read articles on the perils and pitfalls of antibiotic discovery. Go here to get educated! 24 Feb 2020 update: And, go here to read her commentary on the paper by Stokes et al.

All best wishes, –jr

PS: This newsletter has a follow-up here and here.

John H. Rex, MD | Chief Medical Officer, F2G Ltd. | Operating Partner, Advent Life Sciences. Follow me on Twitter: @JohnRex_NewAbx. See past newsletters and subscribe for the future: https://amr.solutions/blog/