You can fluorescently label cells (genetic or membrane) and sort them out. You can even study properties of each cell population individually, as well as assess protective effect of one population on another. See my paper attached.
Article Interaction and self-organization of human mesenchymal stem ...
maybe treatment of the feeder layer with mitomycin C to inhibit cell growth in long-term culture assays would be an option for you - in that case cells from the feeder layer would be simply diluted out in consecutive cancer cell culture rounds.
It really depends as to how pure you want your cancer cell population to be- which would depend on your application.
1. In general, you may try pre-plating your cells. This simply means (as you might know), after you split your cells, transfer them to a gelatin coated dish. incubate the cells at standard conditions (37C, 5%CO2) for 30-45 mins. -- observe under the microscope to be sure. In this duration, your MEFs would attach to the gelatin coating while your cancer cells would still be in suspension.
carefully, tilt the dish, and aspirate your cancer cells and re-plate them ontoa fresh dish, without MEFs, of course. on subsequent passaging, you would observe almost no MEFs in your cultures.
2. Another fool proof way is by Magnetic sorting, using magnetic beads. (MACS)- a bit expensive technique, but gold standard.
3. You could also stain your cancer/ or MEFs and use conventional FACS for sorting your cells.
I use pre-plating for sex-determination of my mESC lines, and it works fine for me. Of course, this might and would, change the karyotype and other properties of your cell lines, as a response /shock of losing MEFs would change their morphology etc.