Predictive Effect of Spatial Protein Signature in Advanced Non-small Cell Lung Cancer Patients Treated with Bispecific Antibody (KN046) Immunotherapy

Background: Immunotherapy, targeting programmed death ligand-1 (PD-L1) and cytotoxic T lymphocyte-associated antigen-4, has shown significant antitumor activity and favourable safety profile in patients with advanced non-small cell lung cancer (NSCLC), while predictive biomarkers remain largely unknown. Here, we investigated the predictive effect of spatial protein expression signature of KN046, a bispecific antibody (bsAb), as a treatment in advanced NSCLC patients. Methods Digital spatial profiling was used to investigate protein expression in both tumor and stroma areas of formalin-fixed paraffin-embedded sections at baseline. Regions of interests (ROIs) were recorded after tricolor fluorescence labeling. The geometric means of different protein expression of different response groups were combined to construct the tumor signature, the stroma signature and the tumor + stroma signature. The accurate score was determined by receiver operating characteristic curve and the area under the curve (AUC). Results Eight patients with partial response (PR) and nine with progressive disease (PD) were enrolled in the study. Among 133 ROIs, 14 proteins expressed differently between PR and PD groups (both P  < 0.05). Among tumor areas, tumor necrosis factor superfamily member 4, CD276, B-cell lymphoma-2 and CD45RO were highly expressed in PR group, while transmembrane protein 173 was highly expressed in PD group (both P  < 0.05). Among stroma areas, PR group exhibited a greater expression of T cell immunoglobulin and mucin domain 3 (Tim-3), PD-L1, complement component 3 receptor 4 subunit (CD11c) and Beta-2-microglobulin (B2M) than those in PD group ( P  < 0.05). The stroma signature showed a superior predictive performance (AUC = 0.840, P  < 0.05) to the tumor + stroma signature (AUC = 0.732, P  < 0.05) and the tumor signature (AUC = 0.670, P  < 0.05). The predictive effect of the stroma signature was much better than that of PD-L1 expression or tumor mutation burden. Conclusions In NSCLC, we firstly found that there are differences in the expression of immune-related proteins in different spatial regions. And we successfully developed a stroma signature with Tim-3/PD-L1/CD11c/B2M which could better predict the response to KN046. This signature might potentially complement the limitations of PD-L1 and TMB measurements and be useful in clinical practice about bsAbs.