NAME: Christopher Love
TITLE: Assistant professor, chemical engineering, MIT; associate member, Broad Institute; associate faculty, Ragon Institute of Massachusetts General Hospital, MIT, and Harvard
BACKGROUND: PhD, physical chemistry, Harvard University, 2004; researcher in immunology at Harvard Medical School, 2004-2005; researcher at the Immune Disease Institute, 2005-2007
Researchers from the Massachusetts Institute of Technology and Broad Institute have developed a reverse transcription PCR-based method to detect gene expression in single cells in a massively parallel fashion.
The technique involves partitioning single cells in picoliter-scale PCR reaction volumes in elastomeric microwell arrays, then performing one-step, single-cell RT-PCR to directly detect copies of mRNA transcripts.
According to the researchers, the method permits the detection of mRNA transcripts of interest for more than 6,000 single cells in parallel per assay with high sensitivity and specificity; and has a variety of applications, including the detection of replicating intracellular pathogens such as retroviruses.
In a paper published online this month in Lab on a Chip, the researchers demonstrate how they combined the method with image-based cytometry to examine the relationships between gene expression and cellular secretion of antibodies in a clonal population.
Christopher Love, the corresponding author on the paper, took a few moments this week to discuss the method, its development, and applications with PCR Insider.
What led you to develop this method?
One of the challenges that we've been trying to address in the lab is understanding how to make multiple different kinds of measurements on phenotypic and genotypic qualities of individual cells. We've spent a lot of time in the last couple of years thinking about how to extract information about secreted proteins from cells; and how to look at what kinds of cells are present based on surface markers that would distinguish lineages and differentiation states.
One of the things we found was that you can use that information to go back to individual cells and extract that manually, one at a time, to look for particular genes — for example, antibodies against particular antigens related to pathogens … or for use in biochemical reagents, where we could capture those genes. But it's really been one cell at a time. Meanwhile we've been doing these massively parallel processes on thousands of cells at a time for these other attributes. That led us to start thinking about how we could also then be able to identify expression of particular genes within those cells that we could couple directly to a number of these other traits — the types of cells that are secreting.
On first glance, your method seems similar to digital PCR. Are there any similarities?
Digital PCR is a method for counting transcripts based on individual reactions. You isolate individual transcripts into small volumes to detect the presence or absence of it.
In this case, we're actually measuring the expression of the genes from one cell. And an individual cell may have one copy or multiple copies. Presently the tool allows us to detect the presence or absence of particular transcripts, which we think will be useful in being able to identify, for example, if a cell is infected with retroviruses like HIV – there are certainly applications there that we are considering.
The challenge for us now is how we go from essentially end point measurements to now quantifying the numbers of transcripts present in each cell. So a traditional dPCR reaction is looking at one or a very small number of cells; whereas with the [single-cell] approach the opportunity is to be able to look at thousands or tens of thousands of cells in parallel.
I think the technique that is most closely related to what we're doing in these little containers is traditional qPCR or RT-qPCR. These little reaction volumes would be equivalent to those that you would run on a 96- or 384-well plate. Here we're just doing it in 100 picoliters. So the opportunity now is to understand the ways in which we can monitor the evolution of signal over time so we can quantify it, much in the way we do with qPCR methods.
So this is currently a "1" and "0" scoring method like dPCR, and not quantitative?
Yes, at the moment it allows us to detect the presence or absence of expression, so the quantification is really in the number of events that occur. So we can say that there are "X" number of cells that express a particular gene. And in the one example in the paper we are trying to understand the relationship between mRNA expression that codes antibodies with the actual ability of that cell to secrete those antibodies; and to look at the subset of cells that actually are in the process of secreting antibodies.
How are you considering modifying this so you can quantify gene expression?
We have started to explore a number of ideas that might allow quantification of transcripts. Some of the early results are looking very interesting, but we still have a ways to go.
Can you share any of these?
Not at this time.
Can you provide a little more detail about how you combined the single-cell PCR analysis with phenotypic analysis to match gene expression and phenotype?
One of the questions we've really been interested in is how to make multiple measurements from the same cell. We've chosen to establish a common platform — in this case it's an array of sub-nanoliter wells that can isolate individual cells. That platform we can use as a stamp, or an engraving plate, to print microarrays of the supernatants of those cells. And that allows us to detect what's being secreted. We can also use it simply as an array for imaging to track what kinds of cells are present based on live-cell stains or surface-expressed proteins. And so each of those data are spatially registered, so now we can go back and ask for a given cell what surface markers did it have, and what proteins was it secreting? And as a terminal measurement, we can measure whether or not they express a particular gene.
The main result of this particular paper was that only a small subset of human B cell hybridomas that transcribed a gene for their antibodies actually secreted the antibody. Was that a surprising finding?
I don't think it's vastly surprising given that people who work within this space of looking at heterogeneity among cells. There is an understanding that transcript expression occurs, and it is modulated at a much different frequency and is well-removed from other downstream functional activities, which are modulated by a number of activities including transcript expression.
I think that what has been missing in the analytical technologies is the opportunity to measure directly the relationship between the functional activity and transcript expression. Now, we can integrate those measurements directly, and I think that's really where the value of the technology lies, in those integrative measurements.
The paper mentions other applications, such as detecting and functionally phenotyping cells infected with pathogens, and amplifying specific genes from many cells in parallel for downstream genetic analysis by sequencing. You also mentioned HIV applications earlier. Are any of these areas particularly promising?
One of the challenges in HIV is understanding where certain cells that are infected [are located] within the immune system. We understand that T cells are infected, and that certain macrophages can be infected, but when [a patient] goes on antiretroviral therapy, often times certain reservoirs of cells still harbor the virus, which basically limits you from being fully cured. Understanding where those cells are and what kinds of cells those are would be very important. But they're very rare. So we think that the opportunity now is to apply these types of technologies to look at those questions, and look for very rare cells that are infected with retroviruses like HIV; and to be able to determine what the phenotypic characteristics of those cells are in relation to the infection.
Is there commercial potential for this? Is it patentable?
We have filed a provisional patent around the methodology. There are efforts now to look at how best to commercialize it, including a potential spinout company.
The challenge for single-cell analysis currently is integration of measurements around a given cell; particularly in examples where you have very limited cells to work with. The challenge is to be able to do that on a statistically relevant number of cells.
And that's where the massively parallel aspect of this comes in.
Exactly.