Can you extrapolate the analytical results from the study of a cell population to what is going on in a single cell? As technology and analytical methods improve, scientists are starting to ask those questions and coming up with some surprising results.
At Fluidigm, Kenneth Livak, Ph.D., senior scientific fellow, has focused on meeting the challenges of analyzing single-cell gene-expression data. Practical steps for analyzing single-cell qPCR data had to be developed because the stochastic nature of eukaryotic transcription at the single-cell level meant that conventional methods for analyzing qPCR data often do not apply.
The company has developed protocols and developed an instrument, the BioMark™ HD, that enables qPCR analysis of transcripts from single cells. In population studies the data yields an average value rather than a real measure of individual cells.
The data can be misleading because this average is dominated by a small number of cells that have a high number of transcripts. Consequently, average measures of gene expression don’t give the real picture of what’s going on in the single cell as most cells in a population, even a homogeneous population, have a substantially lower number of transcripts that the average value.
“Single-cell gene expression is intrinsically noisy,” noted Dr. Livak. “This is based on the observation that eukaryotic transcription occurs in pulses with bursts of transcription interspersed with inactivity when transcripts decay. Consequently, there is significant variation from cell-to-cell that can range as much as 10- to 1,000-fold for every gene we look at.
“Normalization to housekeeping genes doesn’t help, as these housekeeping genes can vary as much as the genes we’re studying. We find that it is best to use ‘unnormalized’ data that actually is normalized on a per cell basis.”
More specifically, Dr. Livak reported that the solution is to look at enough genes in enough cells to apply a multivariant analysis to get a robust signature as to what is going on at the level of single-cell gene expression. A multivariant analysis, either hierarchal clustering or principal component analysis, allows you to look at the pattern of all genes in the study to get an understanding of what is happening at the individual cell level.
At Fluidigm, the approach is to take FACs-sorted tissue culture cells and plate them individually directly into lysis buffer in wells in a standard 96-well PCR plate. Reagents are added and the plate is put in a thermal cycler for the multiplexed reverse-transcriptase step. Ideally, in this multiplex preamplification step, sufficient cycles are run so that single cDNA molecules will generate at least five molecules per qPCR reaction chamber in the next step.
The cDNA pools are transferred by eight-channel pipetter to one side of the microfluidic chip. On the other side, singleplex PCR primers for the 96 genes to be analyzed are added. The chip is then put within the Biomark HD instrument where all pairwise combinations are mixed. The PCR reactions are run and analyzed in real time within the instrument, yielding 9,216 data points. The data output is presented as a histogram that shows the number of cells with each expression level bin for all the genes in the study.