Quantifying Antibiotic Resistance Evolution in Clinically-Relevant Microbes

Drug resistance is often observed when we treat infected patients with drugs that were discovered,
or designed, usually at great cost, with the express purpose of curing people of their infectious disease.
This happens, for example, to HIV patients, malaria sufferers or when a pathogenic microbe, like E. coli, finds its way into someone's bloodstream. Cancers can soon become resistant to the chemotherapeutic agents we throw
at them too, & all because of evolution.

The evolutionary march towards drug resistance can take time. It can be years, or
decades, after the introduction of a new drug before we see confirmation of clinical resistance to it and
a ten-year timescale is thought typical of many antibiotics. Unfortunately, this stops pharmaceutical companies from
seeking new antibiotic molecules. After all, why should they spend 10 years, at great cost, seeking to cure
a disease with a pill that is profitable in the marketplace for only 10 more years?

Intriguingly, drug resistance in tumours is seen in patients on a much shorter timescale,
sometimes within months of the start of chemotherapy, depending on the drug used, the tumour
type, and on the individual patient. So why should we not observe a similar phenomenon for antibiotics?
In fact, we do, & we are now seeing the emergence of datasets showing that bacterial pathogens
can evolve resistance within individual patients because of changes to the DNA of that bacterium
in a matter of mere weeks, even days; & it can be lethal.

This proposal cites a 2015 study (Blair et al, PNAS) whereby resistance to antibiotic treatment in a
blood-borne Salmonella infection was traced, week-by-week, over a 20-week period, whereupon the patient died.
That whole-genome sequencing study, using a range of computer and physical modelling techniques
designed to track evolution in real time, showed very precisely how the resistance profile of the infection quickly
changed by altering expression levels and structures of a variety of proteins within the Salmonella population.

Within a week the population had doubled the amount of efflux protein it was making, moreover, it was now making even better efflux proteins than the original, infecting Salmonella. The efflux proteins are used to pump the antibiotics from inside Salmonella cells to prevent the antibiotic from hitting its target, so they stop working, but this was just one of a variety of mechanisms identified that were shown to correlate with the changes in drug resistance that took place during treatment.

It is important to mention 'plasmids', loops of DNA that are disseminated across the planet by different
microbial species that provide resistance to a range of antibiotics, given these, it seems our future ability to deal with microbial infection sits in a terribly parlous state if something is not done to mitigate such rapid evolution. But what can be done?

Importantly, the 2015 study hints at possibilities. It shows that bacteria become susceptible to some
antibiotics as they increase resistance to others; in other words there are cross- or collateral-sensitivities that emerge
during treatment. So, sometimes, one could use one, and then another antibiotic. This is not outlandish, it is an
idea that has been trialled in the clinic for Helicobacter pylori infections, but little else, so we now need to find
novel cross sensitivities. We also need new ways of combining antibiotics into novel cocktails, & some of those are
proposed here too.

I claim that by bringing to bear modern tools of mathematical modelling and data analysis on microbes that
are subjected to antibiotics in the laboratory, by observing how they respond, we can find weak spots
in their defences that will help clinicians design new therapies & give pharma companies new
methodologies to use within their analysis pipelines. Indeed, this is happening now & I am seeking funding to continue the efforts of my group in this task.

The purpose of this proposal is to broaden the impact of a research programme I began 7 years ago whose goal it was to formalise concepts about antibiotic treatment, going from theory, to the lab & now to the clinic.

Then, I reasoned that mathematical models could build on our increasing dexterity at manipulating microbial genetics & the increasing ease with which we gather evolutionary data to make wholly new contributions to how evolutionary biologists saw the path towards antibiotic resistance. I thought we might generate new
ideas for treatment along the way because the process of mathematical modelling provides very different ways of thinking to those medical & clinical microbiologists are used to.

While those differences provided opportunities, they also presented difficulties both in communication and in finding clinical relevance of the theory, but the first stage of the fellowship was spent overcoming these difficulties and in finding common ground.

That approach was highly worthwhile and the subsequent outcomes were important. The impact of that process on our understanding of just how rapid drug-resistance evolution is has been a feature of the papers we publish in world-leading journals. For example, I was able to show that synergistic antibiotic interaction patterns could morph in antagonistic patterns because of drug-resistance evolution; this is new.

The impact I first sought was realistically narrow, targeting a better understanding
of the "evolutionary pharmacology" of antibiotics. But I am now in a position to build on this & I am broadening the scope of those prior studies to actively, & realistically, seek impact in the clinic.

Although there is natural chasm between mathematical modelling studies and clinical treatments, I have identified two routes to clinical impact to push over the next two years.

First, I will publish analyses showing that mathematical models are consistent with the idea that antibiotic "mixing" and "cycling" are not likely to show appreciable differences in clinical trial datasets on the degree of antibiotic resistance they each select for. Core to the field, this is a controversial topic so this will be an important
contribution written to be accessible to a wide community of non-mathematical researchers.

Second, Project 1 describes a study that will show how different patients, with their different genomes, metabolise antibiotics differently. This means those patients are likely to need different treatment regimens, but those treatments can only be explored once we have some understanding of how the human genome
impinges on the antibiotic dosage in the body. This proposal will commence that study, initially using an in vitro system we recently designed for this purpose.

As my research programme has developed, it has become increasingly clear that the treatment of cancers, bacterial infection, plant fungal species, malaria, even HIV, share common features despite their many differences. One of my goals is to share ideas across communities where drug resistance is a core problem in order to form a consensus on "best practise".

It is through this common ground that I began to talk with Astazeneca over a year ago as we share a large number of questions that we both tackle from a common mathematical modelling perspective. We are now engaged in seeking solutions in each other's domain where our main overlap now is optimal combination design. We independently hit on the idea that one protein could be targeted with two drugs to keep resistance at bay & we are now testing this together, in both cancers and microbes.

Finally, if I am able to develop the culture device described in project 3 then this has the potential to benefit a wide variety of microbiological labs working in universities that are interested in studying drug resistance in a spatially-extended context. There are many of these but that requisite device is currently absent from the market.