To circumvent the difficulty of isolating specific cell populations by MACS from dissociated complex animal tissue, when their proportions reached levels similar to that of the background, we developed the "Three-step MACS" strategy. Cells of interest are defined by their expression of a particular gene(s) of interest rather by than their natural cell surface markers or size. A two-component transgenic cell surface protein, for two sequential rounds of MACS, is expressed under the promoter control of the endogenous gene of interest by means of gene targeting and the generation of transgenic tissue. An initial step to remove dead cells is also used. Here, we describe proof-of-concept experiments, using the biotin acceptor peptide (BAP)-low-affinity nerve growth factor receptor as the two-component protein. The first component, the BAP, can be biotinylated in specific subsets of cells expressing a particular gene by expressing the biotinylating enzyme, hBirA ϭ humanized BirA (hBirA), under the promoter control of another gene defining the specific subpopulation. We showed that a rare population of cells (1.1% of the 13.5 days postcoital mouse embryo) could be enriched to a sufficiently high purity (84.4%). From another sample with 0.1% of our cells of interest, we achieved a 40.3% pure sample. The low cost, speed, and technical ease of the Three-step MACS also make it scalable and hence, an ideal method for preparing sufficient quantities of biological samples for sensitive, high-throughput assays.