Acquired Resistance of Lung Adenocarcinomas to Gefitinib or Erlotinib Is Associated with a Second Mutation in the EGFR Kinase Domain

Picture the clinic room on a good day. A patient with an epidermal growth factor receptor mutant lung adenocarcinoma starts gefitinib or erlotinib, and within weeks, the cough eases, the scans brighten, and the tumor shrinks. It feels like a switch flipped. Then, months later, another switch: growth returns. The cancer learned something. That whiplash—deep response, then relapse—is the puzzle that Pao and colleagues set out to solve. There was a hint from history. When imatinib revolutionized chronic myeloid leukemia, resistance eventually traced back to mutations right in the kinase domain of BCR-ABL, often in a gatekeeper residue that controlled drug access. Similar stories popped up in KIT and PDGFR-alpha. So the team asked a focused question in lung cancer: do EGFR mutant tumors that initially respond to gefitinib or erlotinib acquire a second change in the EGFR kinase domain that blocks the drugs? They went straight to the business end of the protein. Using tumor material from patients who had responded and then progressed, they sequenced EGFR exons 18 through 24—the catalytic domain—and, in parallel, KRAS exon 2. Why KRAS? Because KRAS mutations are classic drivers of primary resistance in lung cancer; if they appeared after therapy, that would point to a different path to escape. The sampling was careful and serial: original lung tumors where available, new biopsies from lesions that grew on treatment—spine, lung, pleural effusions—and, for some cases, more than one site in the same patient. On the lab side, they added something crafty: not just asking if there was a mutation, but if the new mutation was on the same allele as the old sensitizing one. That’s critical. If the same copy of EGFR carries both, then the resistant clone is building directly on the drug-sensitive template. Here’s the headline. In six patients who either developed acquired resistance or recurred quickly after adjuvant gefitinib, three tumors picked up a second EGFR mutation in exon 20: T790M. The threonine at position 790 was swapped for methionine. It wasn’t there before therapy, but in the progressing disease, it was. That’s the pattern you’d expect for a resistance allele under drug pressure. Just as important is what they didn’t see: pretreatment tumors almost never carried this change. In a set of 155 untreated tumors that had already been sequenced across EGFR exons 18 to 21, not one had T790M. And in a much larger tally—nearly 1,300 non-small-cell lung cancers profiled for those exons—just one case showed it. So T790M wasn’t quietly lurking as a common baseline feature. It was linked to what happened after the drug went in. Numbers on a page are one thing. Cells in a dish can tell you how hard that mutation pushes. The group turned to three non-small-cell lung cancer lines that cover the key genotypes. H3255 carries the classic sensitizing L858R mutation in EGFR. H1975 carries both L858R and the T790M gatekeeper change. H2030 has wild-type EGFR but a KRAS mutation—useful for teasing apart EGFR dependent and EGFR independent resistance. When they exposed the cells to gefitinib for 48 hours and measured viability, the split was stark. H3255 was exquisitely sensitive—the drug concentration that cut growth in half hovered around 0.01 micromolar. H1975 sat about two orders of magnitude to the right, with a half-maximal inhibition near 1 micromolar. H2030 behaved much like H1975, not because of EGFR itself but because KRAS drove the show. Erlotinib produced the same ranking. That’s the functional footprint of T790M: it shifts the dose-response curve so far that you’re essentially out of the therapeutic range. If you zoom in from cell growth to the phosphorylation that sends signals downstream, you see the mechanism start to come into focus. In transiently transfected 293T cells, they expressed different forms of EGFR: wild-type, L858R, an exon 19 deletion common in responders, and the same two but with T790M layered on top. Then they pulsed the cells with gefitinib or erlotinib and asked a simple biochemical question: does tyrosine phosphorylation, including the phospho-EGFR to total EGFR ratio, drop? In constructs without T790M, yes—tyrosine phosphorylation fell with increasing drug. In constructs with T790M, no—the bands barely changed. That’s the molecular echo of resistance. The gatekeeper mutation sits in the ATP-binding pocket, and the threonine to methionine swap bulks up the pocket in a way that clashes with gefitinib and erlotinib while leaving ATP accommodation largely intact. It’s the same logic as the T315I mutation in BCR-ABL, translated to EGFR. Back to the patients for a second, because the allele story matters. Using a mix of sequencing strategies, including cloning complementary DNA that spans exons 19 to 21, Pao’s team showed that, in the H1975 line and in patient material, the T790M call often sat on the very same allele as the original L858R mutation or exon 19 deletion. In plain terms, the drug-sensitive EGFR allele had sprouted a shield. They didn’t stop at Sanger peaks. They built a targeted assay that capitalized on a quirk of the T790M change: that 2369 C-to-T substitution creates a new site for a restriction enzyme called NlaIII. After amplifying exon 20 with fluorescently labeled primers, they digested the product and analyzed it. Wild-type gives a 106 base pair fragment; T790M introduces a 97 base pair fragment. In H1975, you saw both. In H2030, only the wild-type band. And in patient samples collected after resistance, the 97 base pair band appeared even when the pretreatment sample had shown only the 106 base pair band. The sensitivity of that approach was good enough to spot mutant alleles that made up roughly three percent of the DNA in a background of wild-type, a bit sharper than direct sequencing, which needed about six percent. That matters when resistant clones are just starting to dominate. You can even feel the clonal dynamics across body sites. In one patient, the T790M signal grew stronger in pleural fluid collected at progression than in the earlier lung biopsy, signaling selection. In another, every tumor sample carried a particular exon 19 deletion from the start, but T790M only appeared in the metastases that grew on therapy. Using a nearby single-nucleotide polymorphism as a mile marker, the team could tell the balance of alleles shifted as treatment proceeded—again, a sign that the resistant copy was expanding. A third patient showed the same arc: pretreatment material with only the sensitizing mutation, then a post-treatment pleural effusion now positive for T790M. Across patients, different sites agreed with each other, suggesting that once a resistant clone breaks through, it can seed multiple lesions. One tempting alternative explanation would be that the tumors simply turned on KRAS to bypass EGFR. Not here. In the progressing samples they analyzed, KRAS exon 2 was clean, matching the cell line pattern where KRAS mutations mostly mark tumors that were never EGFR addicted to begin with. That’s useful clinically because it separates the concept of primary resistance—where KRAS is a usual suspect—from acquired resistance, where T790M emerged as the recurring culprit in this cohort. Now, a pause to keep us honest. T790M wasn’t the whole story. Three of the six patients who progressed did not carry the mutation in the material tested. Some resistant tumors likely use other mechanisms—amplifying EGFR itself, engaging alternate pathways, or changing lineage programs. Even among T790M positive cases, those mutations can be present at different fractions in different sites. Resistance is messy because evolution is messy. Still, the gatekeeper lesson travels. As Pao and colleagues pointed out, the T790M substitution does for EGFR what T315I does for BCR-ABL and analogous positions do in KIT and PDGFR-alpha: it protects the kinase in a way that standard inhibitors struggle to overcome. Structural comparisons even hinted that drugs binding differently—lapatinib was one example discussed in that era—might bypass the steric roadblock. The biochemical data backed that intuition: gefitinib and erlotinib knocked down phosphorylation in wild-type and L858R EGFR but barely touched constructs carrying T790M. If you’re treating patients, the immediate implication is pragmatic. The genotype you started with is not the genotype you’ll have at progression. That argues for re-biopsy when a tumor on gefitinib or erlotinib begins to grow again. In this series, T790M emerged after therapy in half of the cases they could analyze, and detecting it changed the conversation. It meant the tumor was still EGFR driven but had altered the pocket, pointing toward the need for differently designed inhibitors rather than a total pivot to non-EGFR strategies. And if you’re designing drugs, this paper felt like a gauntlet thrown. The methionine at 790 blocks the old path into the pocket; so you need to find a new path. That notion—combined with the clear, allele level evidence that resistant clones carry both the original "on" mutation and the new "shield"—helped crystallize the case for second-generation inhibitors that either bind irreversibly or engage conformations that T790M doesn’t defend as well. It also made the case for more sensitive assays to catch emerging resistant alleles when they’re still a minority, because a three percent signal in a pleural effusion today can be the dominant clone on next month’s scan. Step back to where we started: the emotional arc of response and relapse. What Pao and colleagues gave us is a molecular arc to match—a before and after anchored by a single residue in the kinase domain. They tied patient timelines to exon level changes, linked those to live cell behavior, and connected all of it to a structural rationale that made the resistance intelligible rather than mysterious. It didn’t explain every case, but it explained enough—and with enough specificity—that a path forward became visible. So the next time you see a dramatic response to an EGFR inhibitor, hold two truths at once. First, you are watching a dependence being exploited; the tumor runs on that mutant kinase, and the drug pulls the plug. Second, evolution is patient. The clone that can stay plugged in under pressure—the one with T790M on the same allele as L858R or an exon 19 deletion—will take over if you let it. Knowing that, in detail, is why we can meet the second act with something smarter than surprise.