By using GEPIA2, differential expression between LUAD or LUSC and normal controls were performed with the option of matching TCGA normal and GTEx data (Tomczak et al., 2015; Tang et al., 2017). GSE32683 and GSE116959 datasets were used to evaluate the difference of CD73 expression between LUAD tumor tissue and non-tumor tissue (Selamat et al., 2012; Moreno Leon et al., 2019). GSE32683 dataset contains gene expression profiling of 60 LUAD and matched adjacent non-tumor lung tissue. GSE116959 dataset contains profiling of 57 LUAD samples and 11 peritumoral normal lung tissues. GSE37745 dataset were used to analyzed the difference of CD73 expression between LUAD and LUSC (Botling et al., 2013; Lohr et al., 2015; Jabs et al., 2017). GSE31552 and GSE103512 datasets were used to analyzed the difference of CD73 expression between LUSC and normal controls (Lin et al., 2014; Brouwer-Visser et al., 2018).

GEPIA2 and KM plotter (Menyhárt et al., 2018) were used to evaluate the prognostic value of CD73 in LUAD and LUSC. For GEPIA2, the source data was based on TCGA database, we choose a median expression as cut-off for splitting the high-expression and low-expression cohorts. For KM plotter, the source data was based on GEO datasets (GSE19188, GSE29013, GSE30219, GSE31210, GSE3141, GSE37745, GSE50081), all possible cut-off values between the lower and upper quartiles are computed, and the best performing threshold is used as a cut-off.

LUAD cell line-A549 and LUSC cell line-NCI-H520 cells were cultured with DMEM plus 10% FBS (SiJiQing, Hangzhou, China) in 5% CO₂ incubator at 37°C. CD73 overexpression recombinant lentivirus (Genechem, Shanghai, China) were transfected into A549 and NCI-H520 cells according to manufacturer’s instructions. Stable cell models were screened with 2.5 μg/ml puromycin, named as A549-CD73 and NCI-H520-CD73. Reverse transcription-quantitative PCR and Western-blot were used to evaluate the CD73 expression level in cell models. 100μM APCP (Sigma-Aldrich) treatment were used to inhibit the CD73 enzymatic activity in A549-CD73 cells.

The cells were seeded in 96 well plate at a density of 2 × 10³ cells per well. Cell proliferation were detected by CCK8 assay according to manufacturer’s instructions (KeyGENBiotech, Nanjing, China). In short, at different time point, 10 μl of CCK8 reagent was added into each well and the cells were incubated for 2 h at 37°C. Then, the absorbance value was measured at 450 nm by using the microplate reader (Epoch, BIO-TEK). The relative viability of cells was calculated as a percentage using the formula: (mean OD450 of treated cells/mean OD450 of control cells) × 100%.

For the colony formation assay, 200 cells were seeded in 9 cm culture dishes, and After 10 days incubation at 37°C with 5% CO2 in a humidified incubator, the cells were fixed with methanol and stained with crystal violet. Visible colonies were counted.

As previously described (Gao et al., 2016), cell’s migration ability of was examined by scratch assay. In brief, cells were seeded and cultured in 6-well plate. The scratch was performed when cell density reached to 80%. Then, cells were cultured with serum-free DMEM. The scratch image were captured at 0 and 24 h. Cell migration ratio = (start distant - end distant)/start distant.

As previously described (Gao et al., 2016), invasion assay was performed in transwell chamber. Cells were seeded in matrigel coated filters, 200 μl DMEM were added to upper compartments of the chambers while 500 μl DMEM plus 10% FBS was added to lower compartments. After 24 h incubation, the cells on the upper surface of the filter was wiped off, then, the cells on the lower surface of the filters were fixed with ethanol, stained with crystal violet and counted.

Quantitative real-time RT-PCR (qRT-PCR) was carried out using FastStart Essential DNA Green Master kit (Roche) and was used to detect the mRNA expression levels. The primer sequences were list in Table 1.

Cell protein was separation with SDS-PAGE and transferred onto nitrocellulose membrane. Membranes were blocked for 2–3 h with 5% non-fat dried milk at room temperature, and then incubated with primary antibody (anti-CD73 mAb was purchased from BBI Life sciences; mAb of anti-E-Cadherin, anti-N-cadherin, anti-Flbronectin1were purchased from R&D Systems; anti-Vimentin mAb were purchased from Abcam) overnight at 4°C. After washed with TBST for three times (15 min each time), membranes were incubated with secondary antibody for 2 h. Proteins were detected with western blotting luminol reagent (Bio-Rad), β-actin or GAPDH was used as the internal standard.

Cells were treated by using cell cycle detection kit (KeyGENBiotech, Nanjing, China) according to the manufacturer’s instructions. In short, cells were fixed by ethanol and stained with propidium iodide (PI) in buffer containing 10 μg RNase A. Cell cycle distribution was assessed by using FACS calibur flow cytometer (BD Biosciences).

CD73 expression in tissues were obtained from GEPIA2 and GEO database. Survival curves were obtained from GEPIA2 and KM plotter. The difference of CD73 expression value were analyzed by using t-test. All survival results are displayed with p-values obtained using the log-rank test. The data from cell experiments are expressed as the mean ± standard error. Data comparisons were conducted by using the Student’s t-test. p < 0.05 was considered to be statistically significant.

We first analyzed the expression levels of CD73 in LUAD and LUSC based on TCGA data via GEPIA2. Compared with that in normal tissues, the expression levels of CD73 were significant higher in tumor tissues of LUAD (Figure 1A). While there were no significant difference of CD73 expression between LUSC tumor and normal control tissues (Figure 1B).

CD73 expression levels in LUAD were also confirmed based on GSE32863 (gene expression profiling of 60 LUAD and 60 matched adjacent non-tumor lung tissues) and GSE11659 dataset (transcriptome profiling of 57 LUAD samples and 11 peritumoral normal lung tissues). The results further demonstrated that CD73 were overexpression in LUAD tumor tissues (Figures 1C,D). Notably, GSE37745 dataset analysis showed that CD73 expression levels were higher in LUAD than that in LUSC tumor tissues (Figure 1E). And moreover, GSE31552 and GSE103512 datasets showed that there were no significant difference of CD73 expression level between LUSC and normal controls (Figures 1F,G).

The prognostic value of CD73 in patients with LUAD and LUSC was analyzed by GEPIA2 and KM plotter. In GEPIA2, the results showed that a higher CD73 expression level was correlated with poor overall survival-OS in both LUAD [HR:1.4; logrank p = 0.031] (Figure 2A) and LUSC [HR: 1.5; logrank p = 0.018] (Figure 2B). However, in KM plotter, the results showed that a high CD73 expression correlated with better OS in LUAD (HR: 0.66; logrank p < 0.001) (Figure 2C), while correlated with poor OS in LUSC (HR: 1.56; logrank p = 0.007) (Figure 2D). These results indicated a significant association between high CD73 expression and poor prognosis of LUSC. Notably, the conflicting observation for prognostic significance of CD73 in LUAD, might be due to the different source data for GEPIA2 (TCGA database) and KM plotter (combined cohort of GEO datasets).

Furthermore, we analyzed the association of CD73 with patients OS in different GEO datasets separately. For LUAD (Figure 2E; Supplementary Figure S1A), high CD73 expression was related to a better OS (HR < 1) in four datasets, while high CD73 was related to a poor OS (HR > 1) in three datasets (HR > 1). Notably, among these datasets, the correlation between CD73 and OS with statistical significance were only found in GSE3141 (p = 0.002), While there were no statistical significance in other datasets (p > 0.05). For LUSC (Figure 2F; Supplementary Figure S1B), most of the GEO datasets (five in six) showed high CD73 as a poor prognostic marker (HR > 1), which were agreement in the results of GEPIA2 and KM plotter. Taken together, further verification for the potential prognostic value of CD73 in LUAD is needed.

To further investigate the effect of CD73 on biological behavior of LUAD cancer cells. CD73 overexpression cell model was constructed and identified by q-RT-PCR and western-blot, which named as A549-CD73 (Figures 3A,B). By using CCK8 assay, we found that proliferation rate of A549-CD73 were significant higher than that of control cells (Figure 3C). While there were no significant difference of colony formation ability between A549-CD73 and control (Figure 3D). Taken together, the CCK8 and colony formation results suggested that CD73 overexpression could enhance A549 cells viability, while have no effect on stemness of A549 cells. Furthermore, flow cytometry analysis showed that, compared with A549-NC cells, the ratio of G1 phase cells in A549-CD73 were lower (53.40 vs 56.12%; p = 0.015), while, the ratio of S phase cells were higher with no statistical differences (33.02 vs 30.59%; p = 0.055), which might be involved in the promoted effect of CD73 on cells proliferation (Supplementary Figure S2). Furthermore, by using scratch assay and transwell assay, we found that CD73 overexpression enhanced migration and invasion ability of A549 cells (Figures 3E,F). These results suggested that the potential function of high CD73 expression in LUAD progression and metastasis. Notably, 100 μM APCP (a specific inhibitor of CD73 enzymatic activity. APCP do not affect the CD73 expression) treatment have no significant influence on proliferation and migration of A549-CD73 (Supplementary Figure S3), which suggested that CD73 might promote LUAD cancer cells proliferation in a enzymatic activity independent manner.

Furthermore, to investigate the effect of CD73 overexpression on LUSC cells, CD73 overexpression cell model was constructed, which named as NCI-H520-CD73 (Supplementary Figures S4A,B). As shown in Supplementary Figures S4C,D there were no significant effect of CD73 overexpression on NCI-H520 cells proliferation and migration. Notably, there were no significant difference of cell cycle between NCI-H520-CD73 and NCI-H520-NC cells (Supplementary Figure S4E). In addition, CD73 overexpression did not affect the expression levels of epithelial to mesenchymal transition (EMT) markers (Supplementary Figure S4F). Taken together, these results suggested the different roles of CD73 in LUAD and LUSC.

Epithelial to mesenchymal transition has been considered as an important phenomenon and mechanism during cancer metastasis, which are characterized by loss of epithelial features while increase of mesenchymal features. To further explore the potential mechanism of CD73 overexpression involved in LUAD cells migration. In this study, we investigated the relation between CD73 overexpression and EMT. Our results showed that the morphology of A540-CD73 cells was different from A549-NC cells, A549-CD73 cells become somewhat elongated (Figure 4A). Along with the morphological alterations, the expression level of epithelial marker—E-cadherin was decreased in A549-CD73, while mesenchymal markers (N-cadherin and Vimentin) were increased (Figures 4B,C). Notably, by using EMT-inhibitor (SB431542) treatment, the migration rate of A549-CD73 were significantly decreased (Figure 4D). These results indicated that CD73 overexpression enhanced migration ability of A549 cells via facilitating EMT progression.