DLL3 expression on circulating tumor cells predicts response to bispecific antibody tarlatamab

Small cell lung cancer (SCLC) is a neuroendocrine malignancy characterized by high metastatic potential and poor clinical outcomes. Recently, tarlatamab, a DLL3-directed Bispecific T-cell Engager Therapy (BiTE), has demonstrated effectiveness in SCLC, but clinical outcomes vary, and there is no available biomarker to stratify patients. Efforts to identify predictive biomarkers for tarlatamab have largely centered on immunohistochemistry staining of tumor tissue, but the near-universal expression of DLL3 in archival biopsies does not reflect the clinical response rates of ~40%. Here, we aimed to determine whether single-cell analysis of DLL3 expression in circulating tumor cells (CTCs) immediately prior to initiation of tarlatamab therapy could identify patients most likely to benefit. We quantified DLL3 protein expression on CTCs isolated with the CTC-iChip, an automated microfluidic platform that enriches intact tumor cells through negative depletion of red blood cells and platelets using inertial microfluidics and removal of leukocytes using immunomagnetic sorting. Enriched cells were analyzed by single-cell multispectral fluorescence imaging. In parallel, single-cell RNA sequencing of SCLC tumor biopsies was performed to assess heterogeneity in DLL3 expression. In our study, we performed pre-treatment CTC profiling of a prospective cohort of 20 patients with advanced SCLC, distinguishing patients as DLL3 High (≥25% DLL3-positive CTCs) or DLL3 Low . We found that DLL3 High patients consistently derived clinical benefit (SD/PR), whereas most DLL3 Low patients progressed on therapy (85% sensitivity, 100% specificity). In addition, 3 of 5 patients with clinical grade 2 cytokine release syndrome were found to have evidence of tumor cell lysis (circulating tumor fragments) in the days following the first tarlatamab infusion. Subsequent ongoing longitudinal analysis of CTCs at the time of tarlatamab acquired resistance revealed two distinct patterns: loss of DLL3 expression on CTCs or persistence of epitope expression, accompanied by systemic T cell dysfunction. Importantly, in the case of a reduction in DLL3 expression, other neuroendocrine or SCLC-enriched epitopes (SEZ6 or B7-H3) remained detectable, indicating preserved lineage identity. Together, these findings demonstrate that CTC-based quantitation of DLL3 using unbiased microfluidic enrichment provides a real-time, non-invasive biomarker for stratifying patients most likely to benefit from the bispecific antibody tarlatamab. For immune-based cancer therapies that are uniquely dependent upon epitope expression by cancer cells, CTC-based measurements may provide a robust biomarker for guiding therapeutic interventions. Beyond SCLC, future directions include assessing generalizability to extra-pulmonary neuroendocrine cancers known to express DLL3, including medullary thyroid cancer, GI, and GU cancers.