Background: Immune checkpoint inhibitors (ICIs) is effective in treating tumors expressing PD-L1, however, the reported response rate was low. Inter-tumoral heterogeneity might be one of the reasons for the insensitive of ICIs. In patients with multiple tumors (defined as co-occurrence of more than one malignant lesion in a single or multiple sites), the situation is more complicated. Methods: A total of 67 patients who provided tumor tissues from at least two anatomical sites for assessment of PD-L1 expression by immunohistochemistry (22C3 and/or 28-8 antibodies) was enrolled in this study. Cells with ≥1% staining was defined as a positive result. The sequencing data of 61 patients was also retrospectively reviewed. Targeted sequencing was performed with panels of up to 450 cancer-related genes (at least 47 genes in common). Matched peripheral blood samples were also tested as control to exclude germline mutations. Results: At sample level, the PD-L1 positivity by cancer type was as follow: cervical cancer/melanoma/adrenal gland carcinoma (100% [1/1]), thymic cancer (100% [5/5]), colorectal cancer (80% [8/10]), esophageal squamous cell carcinoma (62% [8/13]), ovarian cancer (57% [4/7]), biliary tract cancer (50% [3/6]), lymph node metastasis (50% [1/2]), pancreatic cancer (50% [2/4]), lung squamous cell carcinoma (43% [3/7]), endometrial cancer (40% [4/10]), breast cancer (29% [2/7]), gastric cancer (25% [2/8]), lung adenocarcinoma (15% [4/26]) and liver cancer (14% [1/7]). All the provided tissues were positive or negative for PD-L1 in 42 patients and were partly positive and partly negative in 25 patients. At patient level, differentially mutated genes between the two above-mentioned groups of patients included CCND1 (10.5% vs 39.1%, P = 0.0116), FGF19 (7.9% vs 30.4%, P = 0.0324), FGF3/4 (7.9% vs 34.8%, P = 0.0143) and TERT (21.1% vs 0%, P = 0.0199). Tumor mutation burden and microsatellite instability status were comparable between the two groups. Consistent with previously reported (not specifically recruiting patients with multiple tumors), EGFR exon 19 deletion was preferentially mutated in PD-L1 negative tumors (16.7% [4/24] vs 0% [0/7], P = 0.5497) in non-small-cell lung cancer, and PD-L1 expression was correlated with EGFR wild-type (wild-type: 75% [3/4] vs mutated: 45% [9/20], P = 0.5901) and ALK mutation (mutated: 25% [1/4] vs wild-type: 5% [1/20], P = 0.3116) in lung adenocarcinoma, although statistical significance was not reached. Conclusions: For patients with multiple tumors, inter-tumoral heterogeneity of PD-L1 expression is common. Several actionable gene mutations were found to be more commonly occurred in PD-L1-heterogenous tumors. Physicians should carefully consider PD-L1 and genomic assessments by case in order to select appropriate patients for ICI treatments.