At A Glance
Name: Sanjiv Sam Gambhir
Position: Director, Crump Institute for Molecular Imaging, UCLA School of Medicine
Associate Professor, Department of Molecular & Medical Pharmacology and Department of Biomathematics, UCLA
Background: M.D., Medicine, UCLA
Ph.D., Biomathematics, UCLA
B.S., Physics, Arizona State University
You published a paper online in PNAS last week describing a system for imaging protein-protein interactions in living mice. What is special about this method?
The key advance in this work is that it’s designed to measure protein-protein interactions in living subjects — instead of what researchers normally do, which is measure protein-protein interactions in intact cells under a microscope, or after lysing the cells. The idea is then that you can test drugs in living animals, or you can study developmental biology in living animals.
We recently developed a method that was also published in PNAS, about half a year back [Mar 5;99(5):3105-10], which adapts the yeast two-hybrid system for living animals. In that approach, when two proteins interact, you turn on a reporter gene. We proved that we could detect, for the first time, interactions of two proteins in living mice using optical cameras. The limitation with the yeast two-hybrid approach is that it can only measure protein-protein interactions in the nucleus of the cell; the two protein partners have to be interacting in the nucleus.
What did you do to overcome this limitation?
We started to look for ways to study two proteins interacting anywhere in the cell. The two genes we introduce into the cell each encode a split reporter, which is fused to the protein partner of interest.
We chose firefly luciferase as a reporter, because we have been using it extensively in mice. We used luciferase that we split at key places, so that when the two halves of the enzyme are brought together [and you add substrate], you produce light, which is detectable in the mouse.
We use both an approach where the two split halves are brought together through complementation and [one where this happens] through a process known as intein-mediated reconstitution.
The proteins are expressed at levels similar to what they would be found under physiological conditions. We induce them with a promoter that is inducible by a drug, in this case TNF alpha.
How do complementation and reconstitution differ?
Complementation just brings the two split halves in close proximity, and when the two interacting proteins stop interacting, the two halves also stop interacting. It’s real-time, on-off, on-off, on-off. In reconstitution, as the two proteins of interest interact, it leads to the two split reporters coming together, and then they are spliced into a mature protein, so that even when the two protein partners stop interacting, you still have reporter signal present. They both ended up having equal sensitivity in living animals.
How can you detect a signal from inside a mouse?
The optical reporter can only produce light, bioluminesce, if you provide substrate, and the substrate has to get into the cells. We give substrate to the mice via the intraperitoneal or the IV route. The substrate goes everywhere through the mouse, but when it comes into contact with the complementing reporters or the reconstituted reporter, you produce light. Most of the light never makes it out of the mouse, but we can detect the light from deep within the animal, because the mouse is placed in a black box, and then a cool CCD camera that’s very sensitive detects the light, anywhere within the mouse.
It will also work in rats, but it wouldn’t work in humans, because they are too deep. In humans, we use other techniques like positron emission tomography (PET) or micro-PET. In that case, we use reporters that are enzymes that are capable of trapping radioactive probes, which can give signals that can be detected at much deeper depths than optical light can.
In this work you only looked at two proteins, MyoD and Id, known to interact very strongly. Have you since tested the technique with other proteins?
We are now testing other protein partners in addition to the ones that we tested in that paper. So far, with other — weaker — partners, we also get significantly detectable signal.
What are the main limitations of the method?
The main limitation of this technique is that it’s not a high-throughput screen, like many proteomics cell techniques are. Because the method is designed for animal use, it’s not designed to screen which two proteins are interacting. Its primary use is when you know two proteins are interacting, and you are trying to develop drugs to block that interaction, or you want to know, in a time course of animal development, when two proteins are starting to interact.
The other limitation is, it does require this very sensitive optical CCD camera, which is not as routinely available in any lab. It’s a special piece of equipment that costs as much as $150,000.
Also, you can only measure the effects of a drug on a particular pair of proteins, right?
Yes. All you can assess is whether that drug is doing its job in inhibiting the two proteins that you are interested in. If it has side effects, if it’s also blocking two other proteins, or doing other things, that you can only indirectly see by using other measures.
So in proteomics, would this method be used more downstream?
It might be used to study how drugs modulate the protein-protein interaction. Once you know two proteins are interacting, let’s say, in a particular signal transduction pathway, and you are developing drugs to inhibit that interaction, this gives you a direct way to test those drugs in the animals prior to use in humans.
Are you thinking about making this commercially available?
The next goal is to partner with drug companies that want to test specific protein-protein interactions in living animals using these types of technologies, which they would license from UCLA. The provisional patent for this [technique] has been filed.
What are you going to do next?
A lot of different things. We are starting to test the technique with protein partners that have weaker interactions, and with protein partners where there is known drug development, because what we want to do is test whether those drugs, then, can be evaluated in the mice through the protein-protein interaction signal. We are also working on split reporters that are more sensitive, that produce more light in the living animals, and on split reporters that would be applicable to humans by using PET technology. All those kinds of studies are now beginning.
Why is it important to test how a drug interferes with a protein-protein interaction in vivo?
You can do all kinds of testing in vitro or in cell culture, and it’s not predictive of what happens in vivo at all. In vivo, the cells are in their normal environment, and you have got to deal with all the cells that they are in contact with. The bigger issue is that when you put a drug in cell culture, it’s not what the drug is going to do in vivo. You have got issues of delivery of the drug to the target site, clearance from the kidney, the hepatobiliary system. Many drugs work great when you drop them into cell lines, but that’s because there is no physiological mechanism that normally occurs in vivo. That’s why we are big believers in that most of biology will shift to more and more in vivo techniques, because they are more predictive of what likely is to happen in a human.
I think more and more people in the proteomics world need to be aware that techniques are now emerging that let them test things in living animals. People have to start saying: ‘How do we go to the next level of systems biology and study systems in the context of intact organisms?’