Antibiotics and Grants - 5

For the next installment in our series on grant writing for the discovery, optimization and development of anti-bacterials (much of this holds for antivirals and antifungals as well), I would like to focus on the preclinical development of a lead compound(s).  Hang on to your hats as this is where things start to cost money.

There are two key interrelated parts to this, efficacy and toxicology.  I prefer to conceptualize both of these in terms of drug exposure – although it is possible to do everything on a mg/kg basis.  Other important aspects revolve around metabolism, drug interaction and mutagenic potential. These will have to be discussed at another time.

Hopefully, during your optimization program, you have already established an in vivo proof of concept (see no. 4 in this series).   You should have also tested your compounds for activity against the hERG channel (correlates with cardiotox). Your efficacy goal in preclinical development is to establish an estimated efficacious dose and to define the PK driver for efficacy.  These goals are best achieved in the neutropenic mouse models of thigh infection and lung infection (pneumonia). The readout is colony forming units per gram of tissue.  In setting up the model, be sure your contractor is taking a sample at 1-2 hours post infection to establish the baseline colony count in the tissue.  Some make the mistake of comparing everything at the 24 hour time point which, in the untreated animals, represents growth at the site.  Your efficacy endpoint should be the exposure required to achieve either bacteriostasis (for bacteriostatic drugs) or some level of bacterial killing (for bactericidal drugs) (usually one or two logs). 

Before you get too deeply involved in these efficacy models, you should have some idea of drug toxicity or at least tolerability.  I recommend doing a maximum tolerated dose (MTD) test (non-GLP) early.  I also recommend using the same route of administration as you plan to use in your efficacy models (IV, Sub-cut or oral).  If you choose IV – be warned that IV bolus administration sometimes produces intolerability that disappears with a slow bolus, with infusion or with sub-cut administration. Obviously, an overlap between the MTD and the efficacious dose is not so good.  Be aware of solubility issues when looking at IV dosing.

IF you had a positive signal in the hERG screen (not mentioned in the last blog on screening - sorry) - confirm with a formal patch-clamp assay.  If still positive at some relevant concentration (less than 30 uM for e,.g.) - run an in vivo screening model - like in the guinea pig.  IF you have in vivo tox - rethink your choice of lead or find a way to be sure that the toxic dose is not achieved during therapy (not always possible). 

Another critical piece of information you will need is protein binding – both in the mouse and in human sera. For this stage of preclinical development, I recommend equilibrium dialysis where you can correct for non-specific binding to plastic.  If non-specific binding is a problem – try doing everything in the presence of a small amount of tween-80. Test several drug concentrations as some antibiotics (tigecycline for eg.) are subject to concentration dependent protein binding.  Since it’s the free drug that is required for antibacterial activity in vivo, understanding protein binding is key.

The exposure measure is that which is most closely related to efficacy. To establish this, animals are dosed either once per day or multiple times per day usually up to six times.  This allows us to distinguish the importance of Cmax (once daily dosing favored), time (multiple doses favored) and AUC (everything looks roughly the same). 

This sounds simple right?  Not really.  There are a number of exceptions here.  The biggest one involves drugs with a long half-life in the mouse.  In this circumstance, it is not always easy to achieve a distinction between the importance of Cmax/MIC, time > MIC and AUC/MIC in a 24 hour model.  Some experts advocate going as long as three days here.  Some antibacterials, especially those that are primarily bacteriostatic, do not work well in the neutropenic model.  For them, an immunocompetent mouse is preferred. Finally – these models don’t work for every antibiotic – but they do for over 90% in my experience.

Which models and how many strains? At a minimum, I would do the thigh model to establish systemic efficacy parameters.  If you do not plan to develop the antibiotic for pneumonia – you probably don’t need to bother with the lung.  The other issue with the lung model is that mice are not always people.  Success and predictability here depend on transport of the drug into the epithelial lining fluid (ELF) of the lung. Mice may differ from people quite considerably.  For each model, I recommend studying several strains – at least three Gram positives (usually S. aureus) and three Gram negatives (all E. coli, all Klebs, all Pseudomonas).  To add different genera, study at least two strains of each.

Once you’ve gotten this far and you know that the drug is tolerated at some multiple of the efficacious exposure, you should proceed to a non-GLP dose ranging study where you will examine the effect of repeated doses (I recommend 5-7 days) where the maximum daily dose is just under the MTD and the minimum is just below the efficacious dose. You can follow clinical signs and clinical laboratory values or, under some circumstances (where you already know something about organ tox) you could carry out focused histopath.  Another option is to save the tissues for histopath in case you decide you need it later based on the clin path data.

The dose ranging tox and the efficacy testing all set you up to do your GLP repeat dose tox study.  Hopefully the doses to be used will be obvious from the prior studies. I recommend at two week study (NOT LESS) (which must be done in two species – usually rat and dog).  You should count on this requiring about four months from start to receipt of a QA’d draft report.  Foir your regulatory submissions (IND, IMPD) you do not need final reports early on – but you should have drafts for which the data have been QA’d. You don’t want to go changing data after you have submitted your regulatory documents.