March 1, 2007 (Vol. 27, No. 5)

With the introduction of fluorogenic probes and fluorometric thermal cyclers that enable the monitoring of amplification reactions in real time, the sensitivity and precision of gene expression analyses has improved dramatically. Utilizing these assays, often called quantitative or real-time RT-PCR, populations of mRNAs can now be analyzed from samples as small as a single cell. The sensitivity of these assays was significantly improved by the development of techniques such as laser-capture dissection for the isolation of individual cells from tissue slices.

However, improved sensitivity exposes problems inherent to extremely sensitive detection. In real-time PCR, the threshold cycle (the number of cycles of amplification required before fluorescence signals first become clearly visible) is linearly proportional to the logarithm of the number of template molecules initially present in the sample. However, when the sample possesses only a few target molecules, that relationship is subject to statistical uncertainties during the early stages of PCR. A sample containing only eight template molecules may behave like a sample possessing twenty-five molecules or like a sample containing only two molecules. Sometimes researchers can overcome this statistical limitation by analyzing the sample many times and then calculating the mean threshold cycle as a measure of target-copy number.

Since the number of mRNAs molecules in a cell varies from zero to a few thousand, with most species of mRNA being present in less than a hundred copies, when RNA species are amplified from a sample containing all of the RNA from a single cell, there is substantial uncertainty in the measured copy number. In order to overcome this problem, researchers developed more reliable ways to perform single-cell RT-PCR. One of these methods is based on digital PCR. In digital RT-PCR, a sample containing RNA from a single cell is divided into a large number of reaction wells, so that each well is likely to receive only a single template molecule. Fluorescent probes (such as molecular beacons or TaqMan probes) are utilized to light up the wells that contain amplified template molecules. The number of illuminated wells provides a direct measure of the number of target molecules in the sample.

Originally, digital PCR was performed in 96- or 384-well format and therefore had a dynamic range of 0–100 target molecules. However, the dynamic range of these assays has been improved by the introduction of assay platforms possessing thousands of individual wells (Figure 1). In one format, there are 3,000 wells etched into the surface of a glass slide, each with a volume of a few nanoliters. The sample is distributed among the wells, and real-time PCR is performed. In a related format, 1,200 chambers are created in a microfluidic device by the intersection of rows and columns of channels and valves.

In a third format, termed BEAMing (Beads, Emulsions, Amplification, and Magnetics), the sample is distributed into hundreds of thousands of microdroplets in a thermostable water-oil emulsion. Each microdroplet also contains a paramagnetic bead that binds to the amplified DNA and becomes labeled by fluorescent reporters associated with the amplified DNA. At the end of the amplification, the beads are separated from the emulsion and counted by a fluorescence-activated cell sorter. In the fourth format, the sample is diluted in a gel, and PCR is performed within the gel. Since the gel limits the diffusion of the amplicons during PCR, molecular colonies form at the positions of the original target molecules, and the number colonies indicates the number of targets that were in the original sample. Even more interesting formats are likely to follow.

Believing that the results obtained from exponential amplification of a few molecules will always suffer from statistical vagaries despite these innovations, other investigators have tried to do away with PCR altogether. A particularly attractive alternative is to perform in situ hybridization with several oligonucleotide probes against the mRNA target. Each of these probes is labeled with multiple fluorophores so that when they all bind to the same mRNA molecule at the same time, the target molecule appears as a fine fluorescent spot under a fluorescence microscope. All of the spots present in the cell can simply be counted, providing an accurate and integral value for the number of target mRNA molecules expressed in the cell.

Figure 1

Stochastic Forces

However, armed with these precise methods, as researchers examine the expression of mRNAs in individual cells, they are finding that gene expression in cells is itself subject to stochastic forces. Warren et al., utilizing digital RT-PCR, analyzed phenotypically identical hematopoietic stem cells for the expression of transcription factor gene PU.1 and expression of the housekeeping gene GAPDH. They found that each cell expressed very different numbers of mRNAs for each of these genes. Using real-time RT-PCR, Bengtsson et al. found that even though insulin producing cells in mouse pancreatic islets are identical in other respects, the level of their insulin mRNA expression varies as much as ten-fold. Similarly, in situ hybridization analyses of cultured cells shows that the number of mRNA molecules varies extensively from cell-to-cell (Figure 2).

Recognizing that large cell-to-cell variations in gene expression are the norm, rather than the exception, many investigators have begun to explore the origins and consequences of these variations. Utilizing their in situ hybridization approach, Raj and his colleagues found that large-scale variations in gene expression occur between isogenic population of cells because mRNAs are not synthesized at a steady rate, but are synthesized in bursts beginning and ending in a random manner. Cells that exhibit a large number of mRNAs are those that at the moment of observation are in the middle of a burst of RNA synthesis. Cells that are observed to possess just a few mRNAs have either not experienced a burst yet, or had produced a burst of synthesis so long before observation that most of the mRNA molecules had been degraded. They also found that bursts of mRNA synthesis in different genes occur independently of each other.

The origin of stochastic mRNA synthesis may lie with unique mechanisms that open up the chromatin in which the gene is embedded, rending it conducive for mRNA synthesis, and then close down the chromatin, shutting off synthesis.

How do cells achieve their characteristic homogenous phenotypes, given that the number of molecules of a given type of mRNA in each cell is so variable? Part of the answer is that proteins generally stay around in cells longer then mRNAs do. Pre-existing pools of proteins receive periodic supplements as a consequence of transient bursts of mRNA synthesis. Since the size of the protein pools is relatively large, it is buffered against variations in mRNA level. Thus, levels of proteins in cells vary less then levels of mRNAs in cells. However, the “lifetimes” of different proteins are different from one another, and thus variations in the level of short-lived proteins are more subject to variations in the level of their respective mRNAs. To cope with such variations, organisms may have developed other, yet unidentified, mechanisms. In some situations, these variations will be beneficial, serving as an extragenetic substrate for adaptation to transient variations in the environment.

The ultimate aim of gene expression analyses is to determine the levels of different proteins that are present in a cell; some investigators like to extend the profiling to the levels of cellular metabolites. Assuming that levels of proteins are proportional to the levels of mRNAs, biologists have used mRNA analyses as surrogates for more difficult to obtain protein profiles. However, the divergence of mRNA levels and protein levels from cell to cell highlights the limitations of this paradigm.

Conventional RT-PCR and microarray-based analyses are typically performed by homogenizing the RNA contents of a large number of cells, and the results are then presented as being representative of the behavior of the entire cell population. Single-cell RT-PCR and in situ hybridization procedures are revealing significant and intriguing alterations to that simplistic assumption.

Figure 2

The author is supported by National Institutes of Health Grant GM-070357.

Sanjay Tyagi, Ph.D., is at the Public Health Research Institute, New Jersey Medical School, University of Medicine and Dentistry of New Jersey. E-mail: [email protected].

Previous articlePeptisyntha Expands Production Capacity in Brussels.
Next articleArgenta Adds New Target to Genentech Collaboration