Like many biologists working then did, Michael White spent much of the 1990s toying with firefly luciferase and green fluorescent protein. And, like many others, White was surprised to find that gene expression varied extensively within and among single cells.
"Luciferase has got a relatively short half-life, so it began to show, quite early on, the fact that transcription varied between cells both over time and at the same time," says White, now a professor of systems biology at the University of Manchester in the UK, of his early work using luciferase as a reporter to monitor gene expression in single cells. "It gave the indication that not only was transcription more dynamic than people suspected at the time, but also that there was more heterogeneity within a cell population than people thought."
Soon, more biologists began to examine the phenomenon of phenotypic heterogeneity within clonal cell populations, loosely attributing it to the newly discovered noisiness of gene expression. They wondered about the cause of the randomness they had observed. Was it intrinsic, controlled by regimented biochemical processes? Or was it the result of extrinsic factors, perhaps a product of the cell's environment?
And then came the physicists.
"Starting with the work of Michael Elowitz [at the California Institute of Technology] and others — a lot of it in prokaryotes — began the idea that transcription was stochastic and very interesting," White says. "And then you began to get [Harvard University's] Sunney Xie and people like that beginning to come in from a physics background doing really quantitative work to understand these processes in simple, defined systems."
Everything is illuminated
Having begun his academic studies in theoretical physics before moving on to mathematics, Arjun Raj found himself in search of an application during graduate school. While working on his mathematics PhD, "I got really interested in this idea of asking quantitative questions and using quantitative tools in biology to actually measure things and actually test ideas in living cells," he says. Raj's foray into mathematical biology was largely driven by a desire to "ask biological questions for which we can get quantitative answers," he adds.
As part of his doctoral research, Raj adapted a fluorescence in situ hybridization and digital imaging microscopy-based RNA visualization technique — developed by Albert Einstein College of Medicine's Robert Singer and his colleagues in 1998 — to examine stochastic gene expression.
Now a commercial assay marketed by Biosearch Technologies, Raj's single-molecule FISH enables the detection, localization, and quantification of individual RNA molecules. "That allows us to get a really quantitative picture of gene expression and transcription," he says. Now leading a systems biology lab at the University of Pennsylvania, Raj and his group are using this single-molecule RNA FISH tool to address spatial questions. "One of the things we've been focusing on lately is: Where are those molecules within the cell?" he says. "Are they in the nucleus or cytoplasm?"
Manchester's White is also tracking the position of single molecules within single cells. Analyzing stored time-lapse images, White and his colleagues have observed the transcription factor NF-κB on the move. "It showed us unexpectedly — from what people had seen before — that in fact it [NF-κB] was oscillating into and out of the nucleus," he says. "You couldn't see that at the population level because the cells were different to each other, [but] at the single-cell level it was very clear."
Elsewhere in Europe, Olivier Gandrillon at the Université Claude Bernard Lyon in France is also using fluorescent reporters, and he is investigating the role of chromatin dynamics in gene expression stochasticity. Because of this work, "we have a vision of how a gene is open or closed that is very different from what I would have expected before," Gandrillon says. "Before, I would have said it's just always mostly on, and [that] sometimes transcription stops, but that's not the case." Rather, reporter experiments have shown that some genes are only "open for a very short period of time, and during that time, there is a very intensive transcription scenario that goes on, and then it closes very fast. You have to wait for a long time before it opens again," Gandrillon says. "And that's really something unexpected. I wouldn't have pictured a gene behaving like that before doing those series of experiments."
Still, questions of how and why single cells and single molecules behave the way that they do, as well as how those actions affect biological systems, linger.
"Currently, [when it comes to] the things people are really interested in, there are two broad camps: One is sort of 'What are the sources of this variability? What actually causes it?'" says Penn's Raj. "There's also another camp of people trying to figure out what the consequences of this variability are, [and whether] the cell ever uses variability to its advantage."
Indeed, for Gandrillon, the main push for researchers in the field is to "focus on the mechanisms that regulate stochasticity in gene expression, [and] how the cell augments or represses variability in gene expression for its own needs," he says.
Spotlight on cause
At present, it is unclear the extent to which noisy intrinsic and extrinsic factors contribute to the stochastic nature of gene expression. "We are far from understanding all the molecular mechanisms at stake," Gandrillon says. But there have been hints.
Several groups have shown that the rate and dynamics of transcription — which is now known to occur in pulsatile bursts — play a leading role. Raj says that low copy numbers of molecules within single cells contribute greatly to gene expression variability. "It's almost like a thought experiment: If you just have a few copies of a gene per cell, then it's inherently going to be highly variable, because either the polymerase is bound or it's not bound," Raj says. "If you have small numbers of molecules, you're going to have a lot more fluctuations about the mean."
Of his team's single-molecule RNA FISH approach, Raj says that "the nice thing [about] counting molecules in single cells is that you can get a sense of the variability between single cells. So if we look from one cell to the next we can say ... 'This cell has five mRNAs, this one has a hundred. Why is that, what's going on?' And that would point to some sort of transcriptional variability." The spatial organization of transcription within cells may also play an important functional role, he adds.
Researchers are increasingly looking toward chromatin remodeling as a potential source of transcriptional variability. "We've got a real challenge in terms of trying to understand how the different chromatin marks on the histones are controlled, how the chromatin conformation is controlled, how transcription links to translation or RNA stability, as well," Manchester's White says.
To Gandrillon's mind, though, transcription is far from the only factor to consider.
"There's also the fact that division, repartition in the two sister cells is unequal, which is a source of variability. Translation is probably an important step, [but] is absolutely overlooked as of now. Cell-to-cell interactions are also probably very important, and they're mainly overlooked as of now," he says. And those are but a few examples.
Trouble is, "everything has to be done in order to understand what are the various levels at which this is regulated. This is the most open challenge I think — just to make sure that any and every source of noise is really well understood," Gandrillon adds. "In the end, what matters is noise ... and this might be different from cell to cell, from physiological setting to another physiological setting."
Without integration, though, researchers risk not seeing the forest for the trees. "I think the biggest challenge is: How do we expand our single-cell toolkit to interrogate more than just RNA and a few proteins?" Raj says.
Spotlight on consequence
Of course, there must be some reason why biologists had overlooked heterogeneity for so long. Perhaps they simply viewed biological systems as a sum of their parts, and justifiably so.
"When you think about it, you're really dealing at the level of stochastic processes, which have to somehow combine to give rise to — shall we say, at least in some sense — an ordered process," White says. The data seem to say that they do, at least in the case of his NF-κB signaling experiments.
"The NF-κB system does actually have organized heterogeneity that is actually engineered by the system," he adds. And for good reason. "Production of something like TNF-α is really important to have that under control, because variation in TNF-α is very dangerous because of the positive feedback on inflammation. So it makes sense at a tissue level you want to control that very, very tightly, but maybe instead of controlling that in every cell, you control it across the group of cells," White says. "And so it does suggest the idea that controlling heterogeneity is important, and therefore inevitably that heterogeneity becomes, perhaps in the future, a new type of drug target that you might control to manipulate processes."
For Penn's Raj and others, their research focus is on whether variability is of some benefit to cells. "There have been a few pieces of work that show that is indeed the case — that you can use variability, that it has some usefulness," he says. However, Raj adds, there is also evidence to suggest cells work to minimize variability, though that is still under investigation.
Considering the potential consequences of inter- and intracellular heterogeneity could alter biologists' long-held understanding of the systems they study. For most biologists, Raj says, variability does not yet "matter to them until we show them there's some real consequences to it, some real meat to the whole thing." Until then, he adds, "the onus is still on us to show that, hey, these bursts are important because of X, Y, Z biological phenomenon that are dependent on this. And it's right to wait until then, [to say] that that's something to consider when they do their analyses."
A biologist himself, Lyon's Gandrillon says he became interested in the stochastic nature of gene expression because he couldn't help but notice variability in the systems he studied. But he quickly found experimental data was not enough to explain what he had observed.
"That's the difficulty, but also the beauty of this subject. You have to have interdisciplinary research because you have to have a model of what goes on in your cells. So, when you couple both model and experiments, then you can really go deep into the molecular mechanisms," Gandrillon says. Because of this, when it comes to understanding biological noise, "it's still physicists leading the game."
Mathematical, computational, and other models are needed to put experimental data into context. "You have to have a model, otherwise you just measure things but cannot make sense of it," he adds. "I am sure the conclusions will get into the biology core community at some point, and people will have to tend to think about their systems in a different way. … But it's going to take a long time before every biologist integrated the fact that variability is a key feature of every, and any, biological system."
"It's an exciting field because it [really] impacts on the basics of biology and biological systems," Gandrillon says.
Manchester's White says progress in the field to date has been technology- and team-driven. "Certainly it requires a combination of theory, maths, and biological measurements to push it forward further," White says. Overall, "wherever we look in cells, the surprise is that we see more dynamics than you might previously have expected — things are more dynamic. And I think that has been, for me, the abiding message for everything we've done," he adds.
For Gandrillon, the most important outcome of research in this area will be how it impacts scientists' visions of biology. "Most biologists were trained to see biological systems as being deterministic, where everything goes from A to B and B to C and everything is ordered, kind of a program, and cells just follow the program. And I think that stochasticity just makes that explode," he says. "You have to rethink the way that biological systems behave. They are not computers, they don't follow a program. They are always using randomness. Randomness is at the heart of biological systems from the start. It's kind of obvious now, [but] it really forces us to rethink the way biologists think about biological systems."