Date: Fri May 12 2006 - 12:54:32 EDT
Hi everybody! Thanks to all of you who sent suggestions directly to me. Many comments were made to me regarding adding or pre-coating with BSA to all plastics that come in contact with cells, optimizing volumes and using smallest possible polypropylene tubes and centrifugation times and g forces. I mainly received discouraging comments about the beads idea, although 2 of you seamingly had some improvements with beads. While these are the kind of comments I was after, after receiving Dr Shapiro's comments and discussing with my staff, I realized that effectively the separation of our positive versus negative populations is usually pretty poor. Here is more details: We want to characterize a cell population of rabbit endothelial progenitor cells grown in culture... Most of you will know that finding antibodies that work with rabbits is not trivial, but even for our human cell projects, the separation on markers such as CD34 and KDR is rarely sharp. So we actually use anti-human Ab that cross-react with rabbit Ag. At the moment we use a directly labeled CD34 but undirectly labeled KDR (we have plan to switch to a directly-labeled KDR Ab). We are looking for endothelial markers like VE-cadherin or PECAM1 that work in rabbits, but have not been successful so far. At the very best, the positive peak overlaps on the negative peek by approximately 50% of the surface under peak. As far as CV, I would be pretty confortable for these measurements with a pretty high CV as the results we are after only need to be kind of qualitative, ie, be able to say that about 15 to 25% of the cells express CD34 and about 65 to 85% express PECAM, for example. The other step where we lose cell counts is the increased mortality after the labelings/washings. While we start with more than 80% live cells, we end up with about 40% live cells... I don't know if it is possible now with these added information to comment on the minimal number of cells we would need... Maybe you will have even more questions to do this evaluation. I realize it started as a bit naive question on my side... but I really thank you all for your comments. Eric De : Howard Shapiro <hms@shapirolab.com> Pour : Eric.Rheaume@icm-mhi.org, Cytometry Mailing List <cytometry@flowcyt.cyto.purdue.edu> cc : Objet : Re: Smallest possible cell number for (preferably flow) cytometry 05/11/2006 09:47 PM Éric Rhéaume wrote: >I have a basic yet critical question for us, about just how small the cell >number could be scaled down to characterize (2-3 color analysis) a >sub-population of cells representing approximately 5-20% of the total cell >population. Our cell samples are very modests in size but we need to do a >minimal cell characterization. Beside the maths, from talking to people who >do flow cytometry experiments, it appears that one of the limiting factors >is the increased cell loss with smaller number of cells during the washing >steps for stainings. > >Is there any way (beside decreasing washing volumes and using BSA) to >increase the yield on less than 50 to 100 k cells samples? Are there some >additives that could be used? Could adding cell counting beads to increase >the pellet volume, be a good idea? Any experience with that or a similar >approach and which beads should be used? Our goal would be to do the >labellings on say... about 10 000 cells... Suggestions for newcomers to >microscopy analysis are also welcome (critical factors, software, >protocols)... It is really not possible to answer the question without knowing how you plan to "characterize" the cells. If you have a 5% to 20% cell population, and you analyze a total of 10,000 cells, you will collect data from 500 or so cells at the low end of the range and from 2,000 cells at the high end. With respect to the cell counts themselves, the precision on a 500 cell count can be no better than 4.4% (Poisson statistics); that on a 2000 cell count can be no better than 2.2%, so, if you are looking for substantial changes in the fraction of your overall population representing your cells of interest, you could probably get away with analyzing 10,000 cells. Bear in mind, however, that selective losses during washing steps can also affect your measurement precision, always for the worse. If, on the other hand, you wanted to measure DNA content and use mathematical modeling to compute the fraction of your cells of interest in each phase of the cell cycle, you'd probably need at least 10,000 of those cells, meaning you'd need to analyze 50,000 to 200,000 cells overall. Most mathematical modelers would prefer 50,000 cells of interest, and that raises the ante. If you are characterizing the population of interest simply by phenotype, the breadths of the distributions of fluorescence associated with each of the markers measured, and the distances between "positives" and "negatives," make a difference; if you cannot completely discriminate between your cells of interest and the rest of the population, it becomes difficult to make any statement about the cells of interest. Adding beads to increase the pellet volume does not sound like a good idea; you'd need at least as many beads as cells to even double the pellet volume, meaning you'd have to analyze at least twice as many events to collect data from your cells of interest, and the beads might also stick to the cells or otherwise interfere with characterization. Rephrasing the question with more specifics will probably get you a more specific answer. -Howard (Embedded image moved to file: pic14794.jpg)
