At A Glance
Name: James Watters
Position: Instructor, Department of Medicine, Washington University School of Medicine
Background: Postdoc, Washington University School of Medicine — 2001-2004; PhD, genetics, Harvard University School of Medicine — 2001
James Watters of the Washington University School of Medicine in St. Louis just finished his postdoctoral fellowship, and is now an instructor of medicine at the school. He brings a strong interest and background in genetics to the position, and was the lead author on a recently published PNAS paper describing the use of genome-wide phenotypic screens for assessing genetic influence on the cytotoxicity of chemotherapeutic agents. Watters recently discussed this research with Inside Bioassays.
Tell me about the assay technology that you used in your research?
It’s a colorimetric assay using Alamar blue vital dye indicator, which is a redox sensitive dye that turns color based on how viable a cell population is. We [seed cells] on 96-well plates ourselves, and then we incubate them with various concentrations of drugs, and then we add the Alamar blue to it. Depending on how viable the cells are, the dye will change color because the more viable cells that are present, the more reduced the cellular environment is. As the dye becomes more reduced, its absorbance properties change, and so we can [put] the plate [in a spectrophotometer], and based on its absorbance properties, we can determine how viable the cell population is compared to a non-treated control population.
Does this dye provide a range of absorbance, or is it just one color or the other?
It changes. The dye really absorbs at two wavelengths — 570 and 600 nm. And depending on how reduced the dye is, it absorbs more at 570 and less at 600, and vice-versa. So it’s a continuous shift, and that’s how in the end we can get a continuous variable reading, because as the cells change from more viable to less viable, you get anywhere in the middle in terms of relative absorbance between 600 and 570
So this tells you more than whether a cell is dead or alive — whether it’s dying, as well?
I think what’s really happening is that when we increase the dose — we do a whole range of concentrations — a larger percent of the population will be dead. So it’s not really accurate to say that there are degrees of dying, per se, but more like a larger or smaller percentage of the overall population is actually dead.
Has this vital dye been around for a while?
Yes, it’s been around for a while. When we were first setting up this experiment, there was the standard MTT assay, which is almost the same thing, but these cells are not attached cells, they’re sort of floating around. So it’s hard to aspirate the media, and add the MTT reagent. But Alamar blue is easier for us because we’re doing a pretty high-throughput experiment, and we can just add it directly to the media, it changes colors, it’s not toxic to the cells, and fewer manipulations of the plate are required. We can just incubate the cells for the predetermined amount of time, and simply add the dye to each well, come back a few hours later, and see the color change, and read it directly on the [spectrophotometer].
You mentioned you’re doing this in relatively high throughput. How are you enabling that?
We have a 96-well plate reader and a robotic loader. When it comes to actually setting up the plates — culturing the cells, and counting the cells, and setting up the plates — we still do that by hand. But when it comes to reading the plates on the [spectrophotometer] — which can be a time-consuming thing — rather than sitting there and reading each plate by hand, we have a robotic plate loader that we can stack plates into, and it automatically feeds the plates into the stack. So a human being doesn’t have to sit there and read all the plates. The machine loads the plates into the spec reader one-by-one, which reads it and gives us the viability output.
Is this a commercially available system or did you design it yourselves?
It’s a robotic plate loader from Tecan. We also have a 96-well plate reader from Tecan. We’ve had for about a year or so. It has two towers for loading plates, and a robot arm for the plate reader. It measures each plate’s absorbance at the two different wavelengths, and then based on the relative absorbance, it does a calculation. And so we calculate viability at each dose, including one dose which is an untreated control. From that, we calculate a percent viability at each dose relative to the untreated control, and that’s the actual phenotype we use to do gene discovery.
Tell me a little about the greater implications of comparing cell viability against different therapeutic compounds.
The two compounds we are using are cytotoxic chemotherapy compounds that are used to treat various solid tumors — for example, colorectal cancer, non-small cell lung cancer, and breast cancer. These kill cells by two distinct mechanisms. One is a DNA anti-metabolite, so it interferes with the process of DNA replication and creating new nucleotides for DNA replication — that’s 5-fluorouracil. The other compound we’re looking at is called docetaxel — the trade name for that is Taxotere, made by Aventis. That compound acts by binding to microtubules, which are structures inside the cell that are normally very dynamic — they are extending and shortening during cell division and many other cellular processes. Docetaxel acts by binding to microtubules and stabilizing them, preventing them from shortening and that prevents the cell from being able to undergo cell division and growth.
We wanted to ask the question: Using a model system, what proportion of differences in drug sensitivity is due to genetics, or inherited factors in a family? Because we have this system where we can measure quantitatively, what proportion of phenotypic differences — in other words, what proportion of drug sensitivity in different cell lines — is due to the genes inherited from the patient’s parents?
Because these cell lines are derived from large families that are studied all over the world, we know a lot about their DNA sequences and variations among cell lines, and because of that, we’re able to correlate changes in DNA sequence with changes in drug sensitivity, and identify regions of the genome that seem important for drug sensitivity. The reason that’s important is that right now, in pharmacogenetics, many studies are limited to looking at one, or a very small number of candidate genes, and you have to have some pre-existing hypothesis or reason for looking at those genes. And we wanted to develop a system that would let us scan the whole genome without any pre-existing assumptions — really have an unbiased search of the genome — and let the biology point us in the right direction of areas of the genome that are important. Rather than having candidate genes going in to the experiment, this helps us identify new candidate genes that we wouldn’t have had a reason before to look at. So the idea was to develop an unbiased discovery system that would let us look at candidate genes with the highest chance of being important in clinical association studies.
When you say ‘scanning the entire genome’…
We look at genetic markers or DNA sequence variations all over the genome — on every chromosome. We look at genetic markers — meaning microsatellite markers or single nucleotide polymorphisms — on every chromosome, all over the genome.
So this approach is not that common in pharmacogenetics?
That’s true. There are other ways of doing a genome-wide assessment, and I guess it depends on your definition of genome wide. Microarray analysis is whole-genome expression profiling, so you’re looking at phenotypic differences, be it drug sensitivity or disease susceptibility, and correlating that with global changes in RNA expression. When it comes to associating DNA changes with drug sensitivity, it’s usually done by candidate gene association, rather than a whole-genome-wide scan.
So this could be applied to any type of high-throughput screening for drug action, right?
Absolutely. It doesn’t have to just be cytotoxic chemotherapy. We’re just measuring cell death — that’s our endpoint. But really any phenotype you can generate in a relatively high-throughput fashion in cells, you can do this exact same approach. So if you wanted to look at molecules that modulate transporter molecules, or say, tyrosine kinase inhibitors, and you want to look at the degree of inhibition of phosphorylation, you can look at those things. Really, for any phenotype that you can measure relatively rapidly, accurately, and quantitatively, you can apply this paradigm.
What’s next for your group?
First of all, we’ll be following up on the actual genes that reside in the areas we found, so we’ll be looking at biologically interesting candidates and validating those in a few ways — maybe altering their function with gene expression knockdown — and then looking for associations with variants in these genes and clinical outcomes in patient populations that have been treated with these drugs. The next step for the [PNAS paper] will be going from identified regions to actually finding the genes and showing that they’re clinically relevant. The other thing is that we’re applying this approach to other drugs. We looked at two drugs here, and we’re expanding our panel to a wide range of chemotherapy drugs.