Liquid Biopsy Caught EGFR C797S in a Patient Progressing on Zipalertinib, a First Outside the Laboratory

Five years, four plasma draws

The man presented with dyspnea and a right-sided pleural effusion after a 45 pack-year smoking history, and cytology confirmed adenocarcinoma. Brain MRI found a 5-mm intracranial lesion, treated with radiotherapy. First-line pemetrexed and carboplatin, then pemetrexed with bevacizumab, produced a partial response of the pulmonary lesion after eight cycles. Poziotinib went in at 16 mg once daily nine months after diagnosis and was cut to alternating 8 mg and 4 mg because of cutaneous and mucocutaneous toxicity. Mobocertinib followed at month 39, escalated from 80 mg to 160 mg once daily, and imaging showed bone and intracranial progression anyway. Amivantamab at 1050 mg every two weeks lasted two months before new hepatic lesions appeared. Zipalertinib, started at month 48, produced clinical and radiologic improvement that the authors describe as most evident intracranially.

The tissue and the plasma named different insertions

Comprehensive genomic profiling on tissue reported D770_N771insSVD. Every Guardant360 report in the paper, including the figure legend that carries the four collection dates, names S768_D770dup instead. The report does not address the discrepancy, and the case narrative treats the plasma variant as the driver whose allele frequency fell from 35.1% to 0.5% under zipalertinib. Anyone reading a resistance trajectory built from two platforms has to reconcile a tissue call and a cell-free DNA call that do not use the same variant nomenclature.

No tissue at progression, so bypass activation stays untested

The authors label a 0.05% allele frequency early subclonal emergence, note that no tissue was obtained at progression, and list what that leaves untested: bypass activation, including MET amplification, which was not detected on cell-free DNA, and histologic transformation. Microsatellite instability was not detected. Response was judged by the treating oncologist on serial PET-CT and brain MRI, with RECIST 1.1 not applied prospectively. The preclinical anchor is work by Kagawa and colleagues in JTO Clinical and Research Reports in 2023, which showed C797S conferring resistance to CLN081 in Ba/F3 cells engineered with exon 20 insertion mutations.

Cys797 is a shared dependency across the oral class

Zipalertinib forms an irreversible bond with the thiol group of Cys797 through Michael addition. Swapping that cysteine for serine removes the anchor the molecule needs, and the authors write that covalent EGFR inhibitors, listing zipalertinib, mobocertinib, poziotinib and sunvozertinib, all theoretically carry the same vulnerability, while amivantamab works through receptor internalization and immune effector function and does not. One passage in the introduction has aged badly: it describes amivantamab as the only FDA-approved agent for this population. The FDA granted sunvozertinib accelerated approval on July 2, 2025 for locally advanced or metastatic NSCLC with EGFR exon 20 insertions after platinum-based chemotherapy, and cleared the Oncomine Dx Express Test as its companion diagnostic the same day. The manuscript was accepted nine months later.

Why This Matters to the APO|APE Reader

When C797S appeared, the team added gefitinib to zipalertinib, on preclinical grounds that C797S without T790M may retain sensitivity to a first-generation reversible inhibitor, and the report names BLU-451, BBT-176 and JIN-A02 as fourth-generation non-covalent candidates along with EGFR-directed PROTACs. A laboratory that runs the assay has to decide where its reporting threshold sits. A 0.05% call arrived on the same report as a driver that had collapsed to 0.5%, and it was the smaller number that changed the treatment plan. Serial plasma at fixed intervals is what surfaced it, since the tissue block predated four lines of therapy and no repeat biopsy was taken.