July 1, 2014 (Vol. 34, No. 13)

Richard A. A. Stein M.D., Ph.D.

Historically, the vast majority of experimental studies that have scrutinized biological phenomena have done so by surveying populations of cells. While it is expedient, this approach has a significant shortcoming—it yields measurements that reflect population averages. Such measurements say little or nothing about individual cells.

Nevertheless, it is possible to visualize variability at the single-cell level, even in genetically identical cells grown under identical conditions. This kind of variability becomes relevant in many contexts including cancer evolution, resistance to antimicrobial or chemotherapeutic agents, cell differentiation, and virus-host interactions.

Increasingly, scientists are exploiting advances in biotechnology to assess cell-to-cell variability. Indeed, they are already using single-cell approaches to enrich their understanding of biological processes. Single-cell approaches are even driving a conceptual shift, a reassessment of the notions that have accumulated about cellular and molecular processes. Because these processes shape development and disease, a deeper understanding of them promises to powerfully impact diagnostic and therapeutic strategies.

“We do not have a full understanding of the complexity of various responses in biological processes, and one of the reasons is that there are many different cell types, many different cell states, and thousands of millions of interactions that determine an overall population response,” explains Rahul Satija, Ph.D., postdoctoral fellow at the Broad Institute of MIT and Harvard University. “We hope that single-cell analyses will help us increase the resolution to understand these biological processes.”

In a recent analysis, Dr. Satija and colleagues compared multiple technologies for their ability to amplify low-input and low-quality RNA samples. “We had been thinking about how to expand the use of transcriptome analysis beyond the more traditional applications in expression profiling,” says Dr. Satija.

Extending this work to profile single cells in a pilot project, Dr. Satija and colleagues made a somewhat unanticipated observation. After sequencing 18 single cells, they noticed huge differences from one cell to another in the way they were responding to a stimulus.

“Even though we have long appreciated that cells are different, we have been unable, due to technical challenges, to examine cells individually and deeply,” adds Alex K. Shalek, Ph.D., postdoctoral fellow in the Departments of chemistry and chemical biology at Harvard University. “[But now] there is a wonderful opportunity for genome-scale single-cell analyses to give us new insights into biology.”

In experiments using dendritic cells derived from mouse bone marrow, Drs. Shalek and Satija, together with collaborators, found a bimodal variation in the mRNA abundance in different cells responding to the same stimulus. The investigators found that using dendritic cells for single-cell analyses offered several advantages. These cells are relatively thoroughly characterized. They have a robust response to lipopolysaccharide activation—a response that has been intensely studied at the population level. They are clinical relevant. And they are post-mitotic.

“The first thing we had to explore was whether the differences in the single-cell transcriptomes we measured were just artifacts of the amplification, because when a single cell’s genetic material is amplified to generate a library, there are biases that can be introduced at every step,” recalls Dr. Shalek.

Single-cell RNA sequencing revealed that, while some artifacts existed, the differences that were visualized most likely reflected biologically relevant changes, and RNA-fluorescence in situ hybridization, an imaging approach that does not use sequencing or amplification, conferred additional confidence that the observations are biologically meaningful and not technical artifacts.

“Once we became convinced that these heterogeneities are real, we wanted to understand why the cells were behaving in this way, and what they can teach us in terms of their biology,” states Dr. Satija.

An analysis of the single-cell expression profiles helped identify two distinct subpopulations of cells based on their inflammatory cytokine expression levels, and incorporating an analysis of their maturation markers revealed that the two groups of cells represent two distinct maturity states. This indicated that RNA-Seq is powered to distinguish between closely related but developmentally different cellular states of the same cell type.

Calculations of the correlation in expression profiles between every pair of induced genes in all the single cells helped identify a cluster of over 130 genes that were changing in a correlated way. Many were relevant for the antiviral response, including two master antiviral response regulators, Irf7 and Stat2.

In cells isolated from an Irf7 knockout mouse, the expression of most bimodally expressed transcripts decreased, but that of the remaining ones was not significantly affected, and the Stat2 gene was not affected in this genetic background. These findings suggest that Stat2 may be acting in either a parallel network with (or upstream of) Irf7. They also illustrate the strength of this approach in dissecting cellular regulatory circuits.


Researchers are exploiting advances in biotechnology to assess cell-to-cell variability. Many scientists are already using single-cell approaches to enrich their understanding of biological processes. [Eskemar/iStock Photos]

SMART Technology

“High-throughput sequencing created a paradigm shift because, compared to a couple of decades ago when sequencing was performed to understand the sequence, currently it has become more like a tool to understand biology,” says Andrew Farmer, Ph.D., vice president, research and development, Clontech Laboratories. “This technology is ideal for performing RNA sequencing on very small amounts of RNA, such as the RNA isolated from single cells.”

Dr. Farmer refers to SMART technology, which stands for Switching Mechanism at 5′ end of RNA Template. It takes advantage of two characteristics of the Moloney murine leukemia virus: reverse transcription and template switching.

In the SMART approach, a pool of messenger RNA (mRNA) from a cell lysate is incubated with an oligo(dT) primer that binds to the poly-A tail of the mRNA. The reverse transcriptase copies the message starting from the 3′ end.

“The key trick of SMART is that when the reverse transcriptase reaches the 5′ end of the message, its terminal transferase activity adds several additional nucleotides that are not encoded in the template, and they complex with a second oligonucleotide that is included in the reaction,” explains Dr. Farmer. “This process promotes template switching.”

An advantage of the SMART approach is that it does not require any adaptor ligation or purification steps in order to generate the double-stranded cDNA for downstream applications. “Through this method we get incredible sensitivity that allows us to get down to single-cell levels,” asserts Dr. Farmer.

Another advantage is that, because the switching occurs at the 5′ end, and because reverse transcription starts from the 3′ end, the products are strongly enriched for complete messages. “There are many techniques that can generate sequencing libraries from RNA, but these generally suffer from the shortcoming that they are biased to the 3′ end of the message and do not adequately capture sequence from the 5′ end,” continues Dr. Farmer. “The SMART technology, however, strongly favors complete full-length messages. This is important in many circumstances.”

Sensitivity remains a challenge. “Although we can do this with single cells, we are still not picking up all of the messages,” cautions Dr. Farmer. Improving the sensitivity will help profile smaller and smaller amounts of RNA.

An additional challenge cited by Dr. Farmer is that the method was developed to target mRNA. The current oligo(dT)-primed approach used for single cells only captures RNA molecules that have a poly-A tail, but there are many other RNA species in the cell, such as regulatory RNA molecules, that do not have a poly-A tail.

“These would be missed simply because the technology primes with an oligo(dT) primer,” notes Dr. Farmer. “Ideally, we would like to prime with a random primer, but avoiding the ribosomal RNA, which represents 80–90% of the cellular RNA, becomes then the bigger challenge.”

While investigators at Clontech have developed an approach in which ribosomal RNA is removed from the sample, currently the lower limit of its sensitivity is 10 ng. “This is still in the thousands of cells range,” observes Dr. Farmer. “Our goal is to push these methods to the single-cell level.”

Splicing Variants

“We have been using plasmon nanorulers to detect and quantitate mRNA in live cells at single-cell resolution, and we were able to extend this approach to splice variants,” says Joseph Irudayaraj, Ph.D., professor of biological engineering at Purdue University. Over 90% of the human genes undergo alternative splicing, and perturbations in this process have been implicated in a large number of human diseases.

However, quantitating splicing variants has been challenging experimentally. In the approach that Dr. Irudayaraj and colleagues developed, 40 nm gold nanoparticles, functionalized with oligonucleotides, can specifically hybridize with mRNA molecules to form nanoparticle dimers that produce a strong and a very specific spectral peak shift as a result of plasmon coupling. This shift can subsequently be detected using dark-field hyperspectral images, allowing the individual mRNA molecules to be tracked and quantitated in live cells.

In a proof-of-concept experiment, Dr. Irudayaraj’s team characterized the spatial and temporal distribution, in vitro and in vivo, of three BRCA1 splice variants at the single-cell resolution. “This technology allows us to capture the complexity of the cellular heterogeneity,” asserts Dr. Irudayaraj.

Parasite Genomes

“Our main motivation for pursing single-cell analyses was that in malaria patients, particularly in Africa, we find more than one parasite strain in almost every single infected individual,” says Ian H. Cheeseman, Ph.D., staff scientist at the Texas Biomedical Research Institute. The presence of multiple parasite genotypes in the same individual complicates genetic analysis because deep sequencing data is not informative about which mutations are from the same genetic background.

“There are a number of major unanswered questions about malaria infections, such as how many different strains are present in a single infection, how they are related to one another, and ultimately how this influences the disease severity and pathology,” insists Dr. Cheeseman.

Previous methods used blood from infected individuals to extract parasite DNA. “There is a lot that this approach does not tell us,” notes Dr. Cheeseman. “And the things that it does not tell us are exactly the ones that we really want to find out.”

To capture the parasite genome at the single-cell level, Dr. Cheeseman and colleagues exploited the fact that red blood cells do not have nuclei. By taking advantage of a DNA dye while performing flow cytometry, the investigators were able to selectively identify and isolate red blood cells that are infected with the parasite. “This helped us efficiently isolate single cells from infected individuals, and subsequently we were able to use them to analyze the DNA,” explains Dr. Cheeseman.

This strategy revealed the possibility of generating parasite genomic sequences directly from the blood of infected individuals. It also unveiled information that has so far been resistant to scrutiny using classical experimental approaches.

Previous articlebluebird bio Acquires Pregenen for Up to $139.9M
Next articleWhat to Look for in a Next-Generation Sequencer