1 and Fig. 2). Metaphase analysis demonstrated that almost all EGFR-amplified parent cells had four chromosome 7 s. Three of them contained a single copy of EGFR and the other contained multiple copies of EGFR (EGFR-ampch7) ( Fig. 3A). By G-banded karyotype analysis of chromosome 7, we found that the EGFR-amplified parent cells had four different type of chromosome 7 s (n, a, b and c) and

clone 4D8 had three different type of chromosome 7 s (n, b, EPZ5676 in vivo and c) ( Fig. 3B). Since the chromosome 7 s (n, b and c) other than EGFR-ampch7 (a) were shared with both parent cells and clone 4D8, it can be considered that clone 4D8 was emerged by loss of an EGFR-ampch7 in EGFR-amplified parent cells. Next, we determined whether the EGFR-unamplified cells were originally present in the parent cell population and evenly proliferated as EGFR-amplified cells, or whether these emerged constantly as part of the parent cell population under normal cell culture conditions. For this purpose, we isolated and expanded two Trichostatin A cell line EGFR-amplified clones, 3B4 and 4F7, from the parent cells, and found that these clones contain 2.5% and 1.0% of EGFR-unamplified cells, respectively ( Fig. 3C and Supplementary

Table 2). Furthermore, we isolated two EGFR-amplified clones from each of 3B4 and 4F7. These four clones again had 0.6–2.4% of EGFR-unamplified cells (Supplementary Table 2). These findings indicate that a small population of EGFR-unamplified cells emerges constantly in parent cells under normal cell culture

conditions (without erlotinib) by means of the loss of an EGFR-ampch7 in EGFR-amplified cells. The IC50 values of resistant cells B10 and D11 to erlotinib (0.68 and 2.0 μM, respectively) were approximately the same as that of clone 4D8 (0.76 μM). The Ribonucleotide reductase level of expression and phosphorylation of EGFR in B10 cells were markedly decreased, but the phosphorylation of AKT and ERK were not completely inhibited by 1 μM of erlotinib (Fig. 4A) as with clone 4D8. Both of these resistant cells had three copies of EGFR, and >99.99% of their populations were classified as EGFR-unamplified because no EGFR-amplified cells were detected in more than 10,000 cells ( Fig. 4B, C and Supplementary Fig. 2A and B). By direct sequencing analysis, the parent cells were shown to have only the E746-A750 deletion in exon 19, as described previously [15], whereas clone 4D8 and B10 and D11 resistant cells contained both the wild-type and the E746-A750 deletional sequences ( Fig. 4D). However, by melting curve analysis, we found that approximately 2% of the parent cell population had the wild-type allele and 98% had the E746-A750 deletion allele, whereas in clone 4D8 and B10, D11 resistant cells, approximately 60% of the population had the wild-type allele and 40% had the E746-A750 deletion allele ( Fig. 4E).