Hold Up On Some of Those Antibiotic Combinations

BY DEREK LOWE

Here's a rather unnerving article on using combinations of antibiotics to treat disease. There have of course been a lot of studies in this area, but they have tended to look at easier-to-measure outcomes like effects on bacterial growth. Not that that doesn't make sense! You'd think that the combinations that have the strongest effects there would also be the ones that have the best effects overall, but that's what this new work is challenging. If you keep your eye on things and wait to see what the effects are on total bacterial clearance, the authors say, things change.

And to be fair, some of this shows up in the growth experiments, too. You can see additive effects with some combinations, positive synergy (once in a while), where the combination is better than you'd have expected versus the two ingredients, and (unfortunately) negative synergy as well, where the combination is actually less effective than you would have predicted. Now, if you listen to nothing but management seminars (not a lifestyle I'm recommending), you'd never believe that there is such a thing as negative synergy out in the real world, but yep, things really can end up as less than the sum of their parts. As the literature review at the beginning of this paper notes, the usual situation when this happens is that the activity of one of the combination drugs ends up being suppressed, while the other one goes along as usual - you don't usually see "reciprocal suppression". Here's one such earlier study and here's another, although there are many others.

This team has come up with a high-throughput cell viability assay system that can be taken out further than usual, with good enough signal/noise to get a reading on bacterial persistance at longer time points. They're using the same strain of Staphylococcus aureus engineered with two different fluorescent markers and co-cultured. This generated a gargantuan pile of microplate images as they tested 14 different antibiotic combinations - the starting measurements were the "early killing" effects of each drug individually at a 90 minute time point (which is certainly enough to show effects on fast-dividing critters like S. aureus), and then they looked at the combinations at that time point as well. They also looked at growth inhibition effects at that same time point.

These experiments showed some complex patterns. There were both synergistic and antagonistic combinations, but these didn't always match up between the growth-inhibition measurement and the "outright killing" ones. Many of these interactions are already well-known in the antibiotic and clinical literature - there's a general antagonism, for example, between "bactericidal" agents and "bacteriostatic" ones (the latter slow growth but don't kill outright). It's been terribly clear since the 1950s, for example, that combining penicillin-type antibiotics ("cidal" drugs that mess up the bacterial cell wall) with tetracycline-type antibiotics (bacteriostatic drugs that inhibit protein synthesis) is a really bad idea and leads to much higher levels of treatment failure and deaths due to infection. The problem is that the former only really kill multiplying bacteria, while the latter prevent bacteria from multiplying in the first place, wiping out the benefit of the penicillin component completely. Overall in the 90 minute experiments here, there was more of that non-reciprocal suppression mentioned above in the killing data - quite a bit of it, in fact, but no actual reciprocal suppression where each drug hurt the other one's activity. 

But that reciprocal suppression behavior was "pronounced and widespread" in the long-term clearance data at 8 hour time points, which is the surprise here. These turned out not to be related at all to the growth-inhibition data, but did show a pretty strong correlation to the early-killing data, which suggests that you might be able to use that (much easier) experiment as a proxy or at least as an early warning system. But you'll have to be careful if you try that, because (for example) some of the strong nonreciprocal suppression examples seen in that early-killing data basically disappeared by the 8 hour measurements. Examples of this are the interaction between tobramycin and the protein synthesis inhibitors, or between trimethoprim and ciprofloxacin. You'd have marked those down via the early-killing data, but the 8-hour clearance data showed that the effect wasn't there any more. That said, some of the classic interactions turn out to be even worse when you look at the longer time points, such as one mentioned earlier between the beta-lactams and the tetracyclines, which seems to slide over from suppression of the beta-lactams to suppression of both partners. The team tried various growth conditions, antibiotic concentrations, and various long time points, but these reciprocal interactions persisted under all sorts of conditions.

The authors went on to try 63 different multidrug combinations of clindamycin, tetracycline, fusidic acid, meropenem, ciprofloxacin, and oxacillin, and found that the situation doesn't get any better. It's known from past experiments that adding in more drugs almost always make the growth-inhibition effect stronger in bacterial assays (as opposed to early-killing measurements, which don't change that much). But in these new clearance measurements, the multidrug combinations were worse across the board. To be sure about that, they repeated the measurements with combinations of five more diverse antibiotics (cefoxitin, linezolid, cefazolin, minocycline, and pristinamycin) and saw the same effects. These experiments are covering a lot of antibiotic space, as folks who have worked with bacteria will appreciate from those lists, and it certainly looks like this clearance effect is (unfortunately) robust. The worst offenders were (as before) combinations of "cidal" and "static" agents, though, so we at least already should know to avoid those.

But there's some good news as well: adding in bacteriostatic drugs that don't depend as much on the rate of cell metabolism (daptomycin or mitomycin) actually potentiated things very strongly with all of the above mixtures (!) These drugs are presumably targeting the metabolically inactive bacteria that the other combinations are (disastrously) missing. And this paper uncovered another interaction that clinicians should be aware of: it turns out that adding a β-lactamase inhibitor can actually lower the clearance efficacy of various drug combinations against β-lactamase-resistant strains, which is not what you would have predicted from first principles. The authors note that combinations of macrolides or doxycycline with the commonly used amoxicillin/clavulanate pair (sold as Augmenin) are prescribed for the treatment of community-acquired pneumonia, and that their work indicates that this has a good chance of making the overall antibiotic efficacy worse, not better. We've clearly got a lot more to learn about these combinations, but this looks like it could be a really useful addition, and should set off a lot more research in this direction.

https://www.science.org/content/blog-post/hold-some-those-antibiotic-combinations