KRAS Mutations and Primary Resistance of Lung Adenocarcinomas to Gefitinib or Erlotinib

Some lung cancer patients on gefitinib or erlotinib responded dramatically — tumors shrinking, scans clearing, and lives extended. Others on the exact same drug and the exact same dose got nothing. They had the same diagnosis and the same treatment but opposite outcomes. Hold that paradox for a moment. A team led by William Pao figured out that those two groups of patients weren't the same at all. Their tumors carried different mutations, and one of those mutations was essentially a molecular stop sign hiding in plain sight. The story starts with the epidermal growth factor receptor, or EGFR — a protein that sits on the cell surface and relays growth signals from outside into programs that tell the cell to divide. In a subset of lung adenocarcinomas, somatic mutations in EGFR's kinase domain alter that signaling in ways that make tumors dependent on it. Gefitinib and erlotinib were designed to block that kinase activity directly, cutting off the growth signal at the source. And when EGFR mutations are present, these drugs can work strikingly well. Pao and colleagues had already established that EGFR mutations predict sensitivity to these drugs. In their cohort of sixty lung adenocarcinomas with known outcomes, seventeen of twenty-two drug-sensitive tumors — seventy-seven percent — carried EGFR mutations. Not one of the thirty-eight drug-resistant tumors had an EGFR mutation. Every single tumor that harbored an EGFR mutation responded to treatment. That's a hundred percent observed response rate, with a confidence interval running from eighty-two to one hundred percent. The p-value on that association was approximately six point eight times ten to the negative eleven — a number so small it's essentially not a number, it's a certainty. But that still left a problem. EGFR mutations explain who responds. They don't fully explain who doesn't. Many patients with lung adenocarcinoma lack those sensitizing EGFR changes. Some of those patients get the drug anyway, and it doesn't work. What's driving their resistance? That's the question this paper set out to answer. The answer was KRAS. KRAS encodes a GTPase — think of it as a molecular relay switch — that operates downstream of EGFR in the same signaling cascade. Normally, EGFR activates KRAS transiently, KRAS carries the signal forward, then turns itself off. But activating mutations in KRAS, most commonly at codons twelve and thirteen in exon 2, lock that switch in the permanent "on" position. When that happens, it doesn't matter what you do upstream. Block EGFR all you want — KRAS keeps firing. The growth signal bypasses the drug entirely. Pao and colleagues reasoned that if KRAS is stuck on downstream of EGFR, then gefitinib and erlotinib should be useless in those tumors. To test this, they screened all sixty adenocarcinomas — the same tumors they'd already analyzed for EGFR — for KRAS exon 2 mutations. Genomic DNA was polymerase chain reaction amplified and sequenced in both directions, with all mutations confirmed in an independent DNA isolate. Sensitive and refractory tumors were determined by Response Evaluation Criteria in Solid Tumors criteria, reviewed by a single pathologist and radiologist blinded to outcomes. The results were clean. Nine of thirty-eight drug-resistant tumors — twenty-four percent — carried KRAS exon 2 mutations. Zero of twenty-one drug-sensitive tumors had KRAS mutations. None. The p-value on that separation was zero point zero two. When you look at it the other way: all seventeen EGFR-mutant tumors responded to treatment, and all nine KRAS-mutant tumors did not. That combined comparison reached a p-value of three point two times ten to the negative seven. The two groups don't overlap at all. The specific KRAS mutations found in those refractory tumors were classic exon 2 substitutions: G12C, G13C, G12D, G12S, and G12V — changes at codons twelve and thirteen that have been catalogued in non-small-cell lung cancer before. Pao and colleagues also note that KRAS mutations tend to occur in patients with heavier smoking histories, while EGFR mutations are more common in patients who have smoked fewer than one hundred cigarettes in their lifetimes. That epidemiological fingerprint adds another layer to the molecular picture. There's one more finding worth dwelling on, because it's biologically elegant: EGFR and KRAS mutations were mutually exclusive in this dataset. No tumor had both. This isn't a coincidence. These two alterations drive the same downstream signaling program — having one is apparently sufficient; having both is redundant. The mutual exclusivity means the mutations partition lung adenocarcinomas into biologically distinct subgroups, and that partitioning maps almost perfectly onto drug response. Pao and colleagues ran drug-specific analyses as well. For gefitinib, EGFR mutations appeared in nine of twelve sensitive tumors and zero of twelve refractory ones, while KRAS mutations appeared in zero of twelve sensitive and five of twelve refractory. For erlotinib, the EGFR signal was even sharper — eight of ten sensitive versus zero of twenty-six refractory — though the KRAS association in the erlotinib group alone didn't reach statistical significance, likely because those patients were enriched for bronchioloalveolar carcinoma cases, which have a low baseline rate of RAS mutations. The clinical implication the paper draws from all of this is direct: test both genes before prescribing either drug. A patient whose tumor carries an EGFR mutation has a very high likelihood of responding. A patient whose tumor carries a KRAS exon 2 mutation has essentially no likelihood of responding — at least to these agents, in this data. Pao and colleagues use careful language: patients with KRAS mutations "will not experience significant tumor regression" with gefitinib or erlotinib. That phrasing is deliberate. It's a clinical recommendation embedded in a results section. The authors are careful about the limits of what they're claiming. Tumors that are EGFR-mutation negative should not automatically be denied treatment — there are drug-sensitive tumors in this cohort that lack both mutations, meaning other mechanisms of sensitivity exist and aren't yet characterized. They explicitly call for validation in large prospective trials using standardized mutation-detection methods. They even document the sensitivity thresholds of their own assays: the fluorescence-based capillary electrophoresis for EGFR exon 19 deletions detected mutant alleles comprising as little as six percent of total DNA, compared to twelve percent for direct sequencing. The Sau96I restriction assay for the L858R point mutation detected mutant alleles down to three percent. That level of methodological transparency is itself an argument for standardization — you can only compare results across labs if everyone is using tools of comparable sensitivity. What this paper represents, stepping back, is an early proof-of-concept for something that would reshape oncology: the idea that a small number of specific genetic alterations can tell you, before treatment begins, not just who will benefit from a drug but who definitively won't. The positive prediction — EGFR mutations pointing toward treatment — had already been established. What Pao and colleagues added is the negative prediction: KRAS mutations as a stop sign. That second signal is just as clinically valuable as the first. Sparing a patient from a drug that won't work — with its costs, its side effects, and its weeks of misplaced hope — is a medical achievement. The broader shift the paper gestures toward is from anatomy-first to biology-first oncology. A lung adenocarcinoma isn't just a tumor in a lung. It's a tumor with a specific mutational identity, and that identity determines how it will behave when you throw a targeted drug at it. Pao and colleagues were working with sixty samples and two genes. The principle they demonstrated — that tumor genotype can guide treatment selection in both directions — is the foundation on which modern precision oncology is built. This lecture was created by ennepō. Go to https://ennepo.ai to Discover, Create and Follow the latest research in your field. Read when you can. Listen when you want to.