In the case of sample shortage, usually biopsy, you have to use whole sample for RNA extraction but in the single cell transcriptome analysis you use just pick up one cell and make cDNA. Then finally microarray or sequencing is performed.
I couldn't find any paper worked on disease by this technique. I wanna study on COPD/ asthma. We want to isolate single fibroblast/epithelial cells and study their expression state.
There are several concerns:
1- Is it possible to amplify the single cell cDNA without bias? (mentioned in the papers that it is possible but I'm not sure)
2- Can a single cell in the complex tissue represent its real situation in the tissue. For e.g. If we analyze fibroblast single cells, can we judge about what happens in the tissue.
3- Limited single cell separation techniques. Most of the works is performed on stem cells and neurons.
4- Biopsy microarray or single cell microarray. which one is more informative.(For this case)
5- SOLiD system, its expenditure and availability is the next question (not hard to find this answer)
I am still just a Masters Candidate, however I have worked in several Genomics Cores.
1) There are RNA pre-AMP kits which claim to Amplify up as little as 1 ng of total RNA while maintaining the relative expression profile. I have never done this personally. My Boss at the time expressed her doubts if it would work or not.
First Google Hit. for "RNA preamp"
http://www.sabiosciences.com/NanoPreAMP.php
2) Logically, I do not believe single Cells would represent the whole tissue. Any individual cell could by anywhere in the cell division cycle and that would obviously have a large effect on any individual cell's transcriptome.
The great thing about your study is that you are actually using single cells. this gives you a lot of power rather than just having averages of a bunch of cells at once. Cell populations are very heterogeneous and we often miss very importnat information that is "hidden" in variablility of expression. Variance of gene expression identifies altered network constraints in neurological disease. see the paper below.
Mar JC, Matigian NA, Mackay-Sim A, Mellick GD, Sue CM, Silburn PA, McGrath JJ, Quackenbush J, Wells CA.
PLoS Genet. 2011 Aug;7(8):e1002207. Epub 2011 Aug 11.
Single cell technology is the way of the future! You are actually measuring single cells of a tissue but you will have to do a lot of replicates to get decent data. We have been doing some work on single cells recently. Yes, you will need to amplification of cDNA or RNA at some stage. That is just the way it is as the technologies availbale for interrogating the samples is just not sensitive enoughto do it.... yet. We are using a BioMark from Fluidigm at the moment for measuring 96 single cells with 96 sets of qPCR primers and we also do 48x48. As you would expect, it can be a bit vairable in its detection ability but we have compared it to arrays results (with pooled samples) with pretty decent correlations. Our pre-amp is a Life Tech "Cell to CT" kit and we do a first round PCR of 16 cycles using gene specific PCRs, then dilute 1:5 and use that for our qPCR.
3) Single cell prep techniques will be best performed with a FACS machine such as BD FACS Aria. You could also dissociate your sample of choice and do serial dilutions and observe single cells in a 96 well plate, but that is very much more random. It is very possible though.
4) This question relies on what you want to answer. You will need to do a LOT of single cell arrays to get any kind of significance. This of course =$$$$$$$$. If your question is "transcriptional changes in disease tissue vs controls" I'd do biopsies. You can get more patient samples and more answers per buck.
5) SOLiD is VERY expensive and a large beast to tangle with in terms of bioinformatics. If you want to measure splice variants than it is certainly the way to go, but means you have to sequence deeper and cost you more money.
Great answer, Anthony! I wonder - do you indeed see large variability in gene expression profile between 48 or 96 different cells which have "theoretically" the same background? And another thing - do you use kind of normalization gene or evaluate only based on Ct values? Do you use C1 unit for pre-processing or do pre-amp etc manually?