Blockade of PD-1/PD-L1 signaling pathway by monoclonal antibodies (MAbs) to release the anti-tumor activity of preexisting tumor specific T cell immunity has initiated a new era for tumor immunotherapy. Administration of anti-PD-1 MAbs (nivolumab and pembrolizumab) in either monotherapy or in combination with anti-CTLA-4 MAbs or traditional chemotherapy has achieved a tumor regression rate of 30%–50% in dealing with melanoma, non-small cell lung cancer, etc.(Larkin et al., 2015). The approval of anti-PD-L1 atezolizumab and avelumab by US Food and Drug Administration (FDA) since 2016 has provided additional choices in dealing with multiple tumors aside from anti-PD-1 and anti-CTLA-4 MAbs as immunotherapeutic medication. The structures of the two therapeutic anti-PD-1 MAbs, nivolumab and pembrolizumab, complexed with PD-1 have been reported which elucidated the molecular basis of MAb-based anti-PD-1 immunotherapy (Tan et al., 2016a, b; Na et al., 2017; Tan et al., 2017). Complex structures of avelumab and BMS-936559 with PD-L1 were also reported which contributes a better understanding of the molecular basis of MAb-based anti-PD-L1 checkpoint blockade therapy (Lee et al., 2016; Liu et al., 2017). In addition, two additional anti-PD-L1 MAbs are in clinics or phase III trials, atezolizumab and durvalumab. Durvalumab (MEDI4736) is a fully human IgG1 MAb targeting PD-L1 that was developed by AstraZeneca, and has been approved by US FDA very recently. Multiple Phase III clinical trials are still ongoing in non-small cell lung cancer, head and neck cancer, urothelial cancer, etc.(NCT02542293, NCT02369874, NCT02516241, etc.). A Phase Ib report demonstrated that durvalumab is well tolerated and showed promising anti-tumor efficacy in nonsmall cell lung cancer patients (Antonia et al., 2016). However, the molecular basis of durvalumab-based anti-PD-L1 reactivity and binding characteristics compared to the other three MAbs used in clinics has not yet been elucidated. In the present study, we expressed the two-Ig-domain PD-L1 and single chain Fv fragment (scFv) of durvalumab as inclusion bodies in Escherichia coli cells. Soluble proteins were obtained by in vitro refolding, and the two refolded proteins survived well in gel filtration (Fig. S1). Subsequently, crystal screen was performed with the durvalumab-scFv/PD-L1 complex proteins, and well-diffractable crystals grew in 3.5 mol/L sodium formate, pH 7.0 (See more details in supplementary information). The complex structure of durvalumab-scFv/PD-L1 was determined by molecular replacement at a resolution of 2.3 Å (Table S1). The binding of durvalumab to PD-L1 involves both of its heavy chain (VH) and light chain (VL)(Fig. 1 A). All of the three complementarity-determining regions (CDRs) of VH and CDR1 and CDR3 of VL contribute to interactions with PD-L1, leaving LCDR2 without any contacts. Previous reports on the anti-PD-1 MAbs revealed that the binding of these MAb is mainly located on the loops of PD-1, ie, the N-terminal loop of PD-1 for nivolumab interaction and the C’D loop for pembrolizumab. However, the binding of avelumab and BMS-936559 is mainly located on the strands of the front-β-sheet face of PD-L1. Here, the binding of durvalumab on PD-L1 was also mainly located on the front β-sheet face which is constituted by A, G, F, C, and C’strands of the IgV domain of PD-L1. A detailed analysis of the interactions between durvalumab and PD-L1 shows an unbiased contribution from VH and VL of durvalumab in binding to PD-L1. The A, G, and F strands of PD-L1 provide major hydrogen bond interactions with durvalumab (Fig. 1 B). D26 of …