Fig 2: PD-L1–vInt4-overexpressed mice displayed resistance in αPD-L1 treatment.(A–D) Mice were injected with 5 × 105 MC38 cells with overexpression of WT PD-L1 or PD-L1–vInt4 and treated with 75 μg of anti–PD-L1 or control IgG 3 times a week by i.p. injection. n = 10–12 mice/group. Similar experiments were performed 3 times. Each line represents the tumor volumes for each mouse. Each figure represents a mice group of: (A) MC38_mPD-L1 WT, treated with control IgG; (B) MC38_mPD-L1 WT, treated with αPD-L1; (C) MC38_mPD-L1–vInt4, treated with control IgG; or (D) MC38_mPD-L1–vInt4, treated with αPD-L1. The regimen of treatment is indicated by black arrows for control IgG and red arrows for αPD-L1. (E) Kaplan-Meier curves of treated mouse groups in A–D. **P < 0.01 by Gehan-Breslow-Wilcoxon test. (F) Serum PD-L1 levels of mice in experiments A–D were sequentially measured by ELISA. Sera were collected just before the endpoint of the experiment or at sacrificing. Data represent mean ± SEM. **P < 0.01 by 2-sided Mann-Whitney U test. Similar experiments are conducted twice for A–F.

Fig 3: Overexpression of soluble PD-L1 variants did not help the tumor growth alone.(A) Mice were injected with 1 × 106 MC38 cells transduced with empty vector (n = 6). (B) MC38_PD-L1–KO cells transduced with empty vector (n = 6). (C) MC38_PD-L1–KO cells transduced with murine WT PD-L1 (n = 6). (D) MC38_PD-L1–KO cells transduced with mh #1 (n = 6). (E) MC38_PD-L1–KO cells transduced with mh #2 (n = 6). (F) MC38_PD-L1–KO cells transduced with mm (n = 6). Each line represents the tumor volumes for each mouse. mh #1 and #2 are different clones. Each group was compared with 2-sided Mann-Whitney U test.

Fig 4: PD-L1–vInt4 demonstrated its function as a decoy in vitro.(A) Evaluation of NFAT activity by measuring luminescence when PD-1 effector cells preincubated with αPD-L1 and secreting PD-L1 variants were cocultured with K1_PD-L1 WT cells. PD-L1–v242 and αPD-L1 were added in 3:1 molar ratio, whereas the ratio was 6:1 with PD-L1–vInt4 and αPD-L1. n = 3. **P < 0.01 by paired 2-tail Student’s t test. (B) Comparison of NFAT activity by measuring luminescence — this time using PD-1 effector cells preincubated with nivolumab and secreting PD-L1 variants. Secreted variants do not significantly suppress the signaling; paired 2-tail Student’s t test. Each condition was compared with the third bar counted from the left end in A and B. (C) Evaluation of NFAT activity by measuring luminescence. PD-1 effector cells were cocultured with K1 or K1_PD-L1 WT cells. PD-L1–vInt4 was added at a physiologically plausible concentration and at a 10-fold concentration. Secreted variants do not significantly suppress the signaling; paired 2-tail Student’s t test. Each condition was compared with the bar in the left end in C. Data represent mean ± SEM in A–C. (D) Jurkat–PD-1 cells incubated with or without Fc-tagged PD-L1–vInt4 were evaluated with FACS. As for nivolumab, it was added 30 minutes before PD-L1–vInt4.

Fig 5: PD-L1–vInt4 expression was found in various types of cancer.(A) The percentage of cases expressing PD-L1–vInt4–specific sequence in intron 4 with FPKM ≥ 1, ≥ 2, ≥ 5, and ≥ 10 in each TCGA cancer type. (B) Structural changes associated with PD-L1–vInt4 expression in a case of HNSCC (TCGA-CV-5443-01) and STAD (TCGA-BR-8361-01). Aberrant transcripts are colored in red. (C) PD-L1–vInt4 was detected in some of the clinical samples. Data were obtained from CCLE.