Each cell, even within the same tissue, is unique. It inhabits its own specific microenvironment within a 3-D space and is therefore acted upon by a unique set of stimuli, which often leads to the expression of certain sets of genes. While some of this gene expression variability is indeed random, other variability leads to critical differences, such as a neuron's discernment of the scent of a rose from that of formal-dehyde, the response of one tumor cell to chemotherapy versus another's resistance, or one cell's propensity to become cancerous over another one nearby. Analyzing an individual cell's transcriptome—that is, the set of transcribed genes that serve as the source of each specific cell's individuality—could give insights into how transcriptional variability is generated. Until now, such single-cell experiments have proven technically challenging, since isolating and quantifying such minute amounts of mRNA is difficult. But several intrepid researchers are adapting different techniques to make it easier.
FLAGing TasteWhile individual olfactory neurons are renowned for expressing different odorant receptors, individuality extends to other neurons as well. Defining the transcriptional profile of single neurons would greatly illuminate the molecular mechanisms responsible for achieving such functional diversity.
At the University of Tokyo's Graduate School of Science, Yuichi Iino and his group in the Department of Biophysics and Biochemistry are exploring ASE gustatory neurons in Caenorhabditis elegans (1,2). In C. elegans, the ASE neuron class responds to gustatory cues like salts, amino acids, and small metabolites. This cell type consists of a pair of functionally different subtypes: ASEL (left), which senses Na+; and ASER (right), which senses Cl-. These two neurons are bilaterally symmetrical but express a different set of chemoreceptors. Though it's known that their asymmetry is developmentally programmed, the precise mechanisms rendering ASEL/R cells different from other types of neurons and from each other is still being elucidated.
To examine these cells more closely, Iino's group FLAG-tagged poly A–binding protein (PABP) and put the construct under the control of che-2, a promoter that is active only in these cells. FLAG is a polypeptide tag often linked to proteins with recombinant DNA technology, and facilitates later affinity purification. By immunoprecipitating the FLAG tag, bound and purified mRNA can then be amplified by PCR and used for microarray analysis. Instead of cDNA microarrays, they used genome wide-microarrays because the genes expressed in a small set of neurons tend to be under-represented in cDNA libraries.
The results showed 8 of the 13 genes known to be expressed only in this cell type, validating the method. The team also identified nine new genes—four of which with unknown functions. All were found to be under the regulatory feedback loop that controls ASEL/R specification, and examination of their promoter regions revealed that their expression is dependent on the cell specific zinc finger transcription factor CHE-1. These results corroborate a previous study using comparative serial analysis of gene expression (SAGE) to delineate the expression profile of ASE cells (3). Like most other studies using SAGE, this one used pools of cells. Studies such as Iino's illustrate that neurons, even within a single class, can have very different functional roles from one another, and that only by examining small numbers of cells can the causative gene expression differences be identified.
Stochastic ProfilingWhile delineating a single cell's transcriptome at a particular time point can provide key insight into development, cancer is another area where transcriptional profiling could play a significant role. Joan Brugge, chair of the Department of Cell Biology at Harvard Medical School, is interested in understanding the cellular processes and pathways involved in the initiation and progression of breast cancer. While working as a postdoctoral fellow in her lab, Kevin Janes developed a method for identifying variations in molecular programs like protein synthesis, oxidative stress response, or signaling through specific pathways in single cells using stochastic profiling (4). Janes, now an assistant professor at the University of Virginia, was awarded a 2009 National Institutes of Health New Innovator Award based on this work.
Janes notes that not only is there nonuniformity in the expression of certain genes among individual cells, but this nonuniformity can be mechanistically important—for instance, in determining why only afraction of tumor cells expressing a given receptor respond to treatment with a drug targeting that receptor. His goal is to “map which transcriptional heterogeneities correlate to phenotypic heterogeneities,” even if those phenotypes do not manifest until weeks or even years later. “This is a starting point for general questions of cell-to-cell heterogeneity; to examine specific instances of divergences in a cell population,” he says.
This stochastic sampling is really a small-sample, rather than a single-cell, approach to transcriptional profiling. In it, mRNA from 10 microdissected cells is amplified by poly A PCR to yield approximately 10ng unlabeled cDNA. As this is an insufficient amount for microarray analysis, a fraction of the sample is reamplified before labeling. Labeled cDNA is then hybridized to microarrays; 7000–8000 transcripts were routinely detected in this way. As a proof-of-principle, Janes and his colleagues examined RNA from MCF7 breast cancer cells seeded in a reconstituted basement membrane to promote the formation of a cluster of cells called an acinus. This is an established 3-D culture model for mammary epithelial acinar morphogenesis. Each acinus is clonal, and therefore isogenic. Variable expression of some genes has been noted between cells, but the significance of this cell-to-cell variability remains unknown. 547 genes—12% of all transcripts—were found to be heterogeneously expressed among cells within the acinus. Many of these were not previously suspected to be variably regulated during morphogenesis.
Until now, finding heterogeneously expressed genes was done in a directed fashion: researchers either looked at the transcription of a specific gene or the transcription of genes that disrupt a known expression nonuniformity, such as that seen in patterning during development. Occasionally, researchers would find nonuniform gene expression simply by chance.
Because it samples ten cells instead of one, Janes says that his stochastic profiling method is both accurate and reproducible. “It is ten times easier to get quantitative measurements, but we can still extract single-cell information.” Also, he says, sampling ten pools of ten cells is also more cost-effective than running one hundred single-cell assays to achieve a reliable sample size.
Which Match Will Ignite?Assistant Professor of Biology Hui Zong, from the Institute of Molecular Biology at the University of Oregon, is interested in figuring out how cancer starts. Zong has been using mouse genetics to generate and label only a handful of cells that have the potential to form tumors, rather than all the cells of a particular lineage (5). “When I keep monitoring them, it turns out that only one focal tumor forms in a given mouse, suggesting that some additional problems within a single cell pushed it over the edge while other cells couldn't do it,” says Zong. He describes the scenario as similar to testing which match could light a fire.
“Before tumors form, we don't know which cell has the potential or which match will ignite. After tumors form, it already went over the edge…the match is already burned up, so we can't determine what caused it to be the one to ignite,” explains Zong.
Zong hopes to use Janes' stochastic profiling technique to figure out what makes a specific precancerous cell malignant. Since only a few precancerous mutant cells actually go on to become tumors, Zong cannot get the answers he needs from studying whole-cell populations; he must look at each cell individually to find the unique transcriptional signature that initiated tumorgenesis.
Still, some of the heterogeneity in gene expression between individual cells can be random. “If you measure a single cell, you may never see that cell again,” warns Janes. A good sampling of the population is therefore required, as fluctuations that are seen repeatedly are probably real. But techniques like the one he and others have developed (6) may help us understand why only certain cells play out their genetic destinies or respond predictably to an inducing agent, while others do not.
