elife Episode 2: Plants keep Thyme, cancer drug resistance and clear corneas

Cancer researchers and doctors all over the world are getting increasingly excited about the new era of molecular therapies designed to target specific gene faults in tumours. But while these drugs tend to show impressive results, their effectiveness soon wanes and the cancer comes back after a number of months or years, as it evolves resistance to the therapy.

Kat Arney spoke to Martin Nowak, Professor of Mathematics and Biology at Harvard University, who has been using mathematical models of tumour evolution to try and understand why this happens. His results could change the way that these new treatments are used.

Martin - The initial effect of the drug is often amazing and doctors see a decline in advanced cancers to an extent that they have never witnessed before in any situation. So, this is extremely promising, but many of these therapies exactly as you say, they are short-lived and the reason is because the cancer cells develop resistance.

Kat - They kind of evolve don't they and change to escape the drug.

Martin - Yes, exactly. So, we are witnessing here an evolutionary process on a pretty fast timescale and here's also where the mathematics comes in because the mathematics of biology is very developed to understand evolution. So, evolutionary dynamics can be described with exact mathematical equations. We can describe how populations get a resistance, get mutations and then how mutations respond to selection pressure. This is what we're doing for cancer here.

Kat - How do you go about starting to model what seems like such a complex biological process?

Martin - The first question would be an estimate of how big the population of cancer cells in the body is. So, if you look at the particular lesion of a certain size then very often we're confronted with approximately 1 billion cancer cells. And then we give the treatment and the observations suggest that there are single point mutations that can actually confer resistance to the treatment. So the first question that would arise is; at the time to start treatment, are the resistant mutations already present in the patient, or do the resistant mutations emerge once therapy has started.

Kat - So, that would be single changes to a gene that would make the cells resistant to a treatment or not, to find out whether they're there at the start or whether they kind of come on later.

Martin - Yes, that's exactly right. So now, the question is; supposed, there are a number of point mutations in the oncogene that can do this between say, 1 and 10 or so. The point mutational rate is fairly well known approximately 10 to the minus 9 per base per cell division. So, given this point mutation rate, given the population size, we can fairly accurately estimate the probability that the resistant cells are already there before you started treatment.

Kat - So, by doing some maths, you can kind of figure out the rate at which these cells are mutating and evolving, and kind of track back to work out whether they were there at the start.

Martin - Yes, exactly. So, if you are telling me basically there are 10 possible point mutations, the population start is 10 to the 9, the point mutation is 10 to the minus 9 then there's some very high probability that we can conclude that the resistant mutations are already there by the time the doctor starts the treatment. So maybe 1 in a million cells has the resistant mutation and this is below clinical detection. So, if you examine the cancer, you would conclude that there's no resistant cell present because the frequency is so small. But then you start treatment and within approximately 20 to 30 weeks, you observe that the treatment fails. If you now look and analyse the genetic composition of the oncogene, you will find that the resistant mutations have emerged.

Kat - This seems to me to be incredibly significant for treatment because at the moment, with these targeted treatments, we tend to test one at a time. So, you do an analysis of a patient's tumour. You say, "Okay, you need this drug" and then you give it to them for a bit, it works and then it fails, and then you try another one. What conclusions can you draw from your research about how we should use these drugs now?

Martin - So, what we have done in this paper is really to ask the question, supposed that you are in a situation where you have two targeted therapies against the same cancer. If you now lose two therapies at the same time then what is the probability that the cancer is now cured. This really depends on whether or not there exist single point mutations in the genome that give simultaneous resistance to both drugs. So, if you are in a situation where there is even only a single point mutation in the whole genome that confers resistance to both drugs at the same time, the situation again is very gloomy and the chances for curing patients with this type of combination therapy are fairly small. So, what we really need to design are drugs that work against the same cancer and moreover, there's not a single point mutation in the genome that keeps rise to resistance to both drugs at the same time. Then the cancer has to respond by developing at least two-point mutations. One that gives resistance for drug A, one for drug B, and that is not so easy for the cancer.

Kat - Cancer cells only have a limited range of ways they can get around these drugs. They're not completely immortal so you would think that eventually, if we had enough targeted drugs, we would be able to have enough ways to hit cancer so it couldn't get round the treatment.

Martin - Yes, this would be the idea, but always keep in mind, it doesn't matter how many drugs you have, if the cancer cell has a mechanism to give resistance to all those drugs simultaneously. So, you really must develop drugs that do not have resistance mutations in common and here the surprising observation of our current study is that a single point mutation of the genome that would give you resistance to both drugs would defeat the treatment. The very important aspect is that whatever combination therapy you use, do not use it sequentially because to use first one drug and then wait until it fails, and then another drug is a certain recipie for complete treatment failure. To use two drugs simultaneously as opposed to sequentially is a much better idea and this is rare in the clinical practice of cancer treatment at the moment.

Kat - How quickly do you think you can get this message out to people who are treating cancer patients and for example, running clinical trials of these new targeted drugs? It seems like an incredibly important thing for doctors to know about.

Martin - Yes, I think it can happen quickly. I was involved in a similar study in 1995 for HIV virus and there, we actually were able to show that single drug therapy of the HIV virus gives rise to complete resistance in approximately 4 weeks depending on the drug because there's a very rapid turnover of the virus. This moved the field very quickly, I mean within in a year to combination therapy. I mean, for me, it's fascinating to see and to hope that mathematics can really help to inform experimental scientists and doctors in the clinic. And so, my hope would be that in a few decades from now, our understanding of cancer will be engineer-like and then we can really use an engineering-like approach to actually design and use the optimum therapies and prevent many, many people from dying from cancer.

What do you do if you want to see how an immune T-cell responds to meeting the antigen molecules that it's programmed to detect? 

Geoff -   My name is Geoff O'donoghue.  I'm a post doc at UC Berkeley in the lab of Jay Groves.  The general question we were interested by was how the immune system is able to detect form pathogens, things like virus and bacteria, and how this occurs at the molecular level.  So, we were looking at T-cells, CD4 or helper T-cells and we were looking at the T-cell receptor and the T-cell receptor binds a pathogenic molecule on the surface of another cell.  It's just like a lock and a key basically at the receptor on the T-cell surface is a receptor and it has a ligand that it's looking for.  Once it finds that, it binds and initiate an immune response specifically only for that one pathogen.

Chris -   So, you look at these T-cells, CD4 cells specifically.  How do you study that interaction?

Geoff -   We study the interaction using optical imaging.  Ideally, we would look at the interaction at the interphase between a living T-cell and what's called an antigen presenting cell.  But that's a very difficult thing to image because you have two curved surfaces trying to image the interphase between those two curved surfaces with high temporal and spatial precision is very difficult.  The trick that we use is to replace the antigen presenting cell with, basically, a biomimetic surface that is flat.  We put on the surface the support lipid bilayer.  So, this is a very good mimic of a cell.  Attached to this support lipid bilayer, we attached the pathogenic molecules.  In those molecules are free to diffuse laterally in two dimensions and then interact the T-cells once we flowed those on top of this surface.

Chris -   And you can see where these pathogenic molecules are going, can you?

Geoff -   Yes.  What we did was label each molecule on the surface with a tracer molecule where we can relatively easily observe.  We basically observed a field of these individual molecules, diffusing around in two dimensions.  Once we flowed the T-cells in, the T-cells bind some of those pathogenic molecules on a surface.  Once that occurs, we observed that instead of diffusing around, the tracer molecules on the surface flow way down.  And then we can relatively easily observe.  The way that technique that we use was basically to use a very long camera exposure time, the images are similar to the long time lapse images you see of a city where the cars on a road are blurred out because they're moving so quickly, but the buildings in the background are very stationary and very clearly resolved.  We took the same approach to distinguish between single molecules that are bound to their ligands in the living T-cell and once they were unbound, instantly diffusing on the surface.

Chris -   And so, this enables you to tell when a T-cell has interacted with one of those ligand molecules on the biomimetic surface.  Are you then able to tell how the T-cell responds to that interaction?

Geoff -   Yes.  So, the first thing we do is measure the number of seconds that we observe this interaction happening.  That tells us something about the time component that the T-cell is detecting.  And then we're looking at a molecule inside the cell which is an enzyme and induces a cascade of chemical reactions inside the T-cell.  And so, what we did was two-colour single molecule imaging.  We looked at the pathogen on the surface and the catalyst inside the cell, and we measured the number of molecules that are recruited to the reaction site involved.

Chris -   So, you can very neatly here, identify when the T-cell docks or binds onto the thing on the cell surface, or the pretend cell surface.  You can see exactly when the T-cell begins to respond and how vigorously a T-cell responds.  That's amazing!  So, what did you find?

Geoff -   What emerged was that interactions between the pathogenic material and its receptor were surprisingly long lived.  It wasn't clear from the previous literature whether or not these would be short much, much less than a second or seconds to minutes long.  We discovered that these interactions occurred for a surprisingly long time.  what we also observed is that one molecule could induce a chemical cascade inside the T-cell.  The T-cell is very sensitive and can detect single molecules or a handful of molecules that will induce a chemical reaction inside the cell.  The next question is to look at not just one pathogen but to see how T-cells respond to a panel of these pathogens and then kind of get at that question, is a large number of short interactions equivalent to a small number of very long lived interactions.  That's I think the next step and that would go some way towards kind of really getting at the question of how T-cells distinguish between self and non-self.