The prognostic value of PKM2 and its correlation with tumour cell PD-L1 in lung adenocarcinoma

This was a retrospective study with a cohort of 173 primary lung adenocarcinoma patients with histologically confirmed and surgically resected tumours at the Thoracic Surgery Department from January 2010–June 2014 at Jiangxi Cancer Hospital (Nanchang, China). Patients who had received neoadjuvant therapy before surgery, including chemotherapy, radiotherapy or immunotherapy, were excluded. All patients underwent lobectomy with additional radical lymph node dissection. Approval was obtained from the Institutional Ethical Board of Jiangxi Cancer Hospital. Verbal informed consent was provided by each patient before their tumour tissue samples were obtained. All procedures involving human samples were conducted according to the Declaration of Helsinki.

Slides (3 slides per sample) from formalin-fixed paraffin-embedded lung adenocarcinoma tissues were used for the histologic evaluation of PKM2 and PD-L1 expression by IHC. Briefly, 4-μm tissue sections were deparaffinized and subjected to heat-induced antigen retrieval. The slides were incubated overnight at 4 °C with a rabbit anti-human PKM2 mAb (1:100; D78A4, Cell Signaling Technology, Danvers, MA, USA) or rabbit anti-human PD-L1 mAb (1:100; E1L3N, Cell Signaling Technology). After incubation with horseradish peroxidase-conjugated secondary antibody (Envision™ Detection Kit, GK500705, Genetech), diaminobenzidine was used to visualize the staining.

Staining for PKM2 was scored based on a 12-point semi-quantitative scoring system [18, 19]. PKM2 expression was quantified after evaluating both the intensity of the staining (0 = negative, 1 = weakly positive, 2 = moderately positive, and 3 = strongly positive) and the distribution of positively stained cells (0 = 0%, 1 = 1 to 10%, 2 = 11 to 50%, 3 = 51 to 80%, and 4 = 81 to 100%). The final histological score was determined as the staining intensity × distribution, with a range from 0 to 12. The cut-off point was the median (score < 8, negative expression and ≥ 8, positive expression).

The scores corresponding to PD-L1 expression on tumour cells (TC-PD-L1) were evaluated as a percentage of positively stained cells in the overall section with tumour cells as described in previous studies [20, 21]. Consistent with many other studies [22, 23], PD-L1 positivity was considered membranous PD-L1 expression in ≥5% of tumour cells. Two independent investigators assessed all slides without prior knowledge of the clinical outcome. When any discrepancies occurred between the two investigators, a consensus was reached by discussion.

The human lung adenocarcinoma A549 cell line was purchased from American Type Culture Collection. Cells were routinely maintained in RPMI-1640 basic medium supplemented with 10% FBS (Gibco, Life Technologies), 100 units/mL penicillin and 100 mg/mL streptomycin at 37 °C in a humidified atmosphere with 5% CO₂.

Lentiviruses expressing PKM2-specific shRNAs (designated PKM2-shRNAs) or the negative control vector (designated scramble-shRNA) were purchased from GenePharma Inc. (Shanghai, China, http://​www.​genepharma.​com). A549 cells were cultured at 1 × 10⁶ cells per well in 6-well plates and infected with PKM2-shRNA and scramble-shRNA lentiviruses in accordance with the manufacturer’s instructions. Target sequences with high PKM2 knockdown efficiency (PKM2-shRNA-1: 5′-CCGGCTACCACTTGCAATTATTTGACTCGAGTCAAATAATTGCAAG-TGGTAGTTTTTG-3′ and PKM2-shRNA-2: 5′-CCGGGCTGTGGCTCTAGACACTAAACTCGAGTTTAGTGTCTAGAGCCACAGCTTTTTG-3′) and the scrambled sequence without any effect on PKM2 levels (5′-CCGGGAGGCTTCTTATAAGTGTTTACTCGAGTAAACACTTATAAGAAGCCTCTTTTTG-3′) were adopted as described previously [6]. After 72 h, stable PKM2 knockdown cells were selected with puromycin (1 μg/mL).

Cellular PKM2 and PD-L1 protein levels in lung adenocarcinoma cell lines were evaluated in total cell extracts by Western blot analysis. A total of 30 μg of protein was fractionated using SDS/PAGE and then transferred onto polyvinylidene fluoride membranes (Immobilon P; Millipore, Bedford, MA, USA). After blocking with 5% non-fat milk and washing with PBS, the membranes were incubated with rabbit anti-PKM2 mAb (1:1000; D78A4, Cell Signaling Technology, Inc., USA), rabbit anti-PD-L1 mAb (1:1000; E1L3N, Cell Signaling Technology, Inc., USA) and rabbit anti-Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) mAb (1:5000, Cell Signaling Technology, Inc., USA), followed by the corresponding secondary antibodies (1:10000, sc-2004, Santa Cruz Biotechnology). The protein bands were visualized via Chemidoc Touch (Bio-Rad) and quantified relative to GAPDH expression using ImageJ software (NIH, Bethesda, MD, USA).

Total RNA was isolated from the cell lines using TRIzol reagent (Invitrogen, Carlsbad, CA, USA) and reverse transcribed using a PrimeScript RT™ Master Mix (Catalogue no: RR036A; Takara Biotechnology Co., Ltd.), followed by quantitative polymerase chain reaction (qPCR) with GoScript qPCR Master Mix (Promega; Madison, WI, USA) according to the manufacturer’s protocol.

The primers for PKM2 were forward 5′-ATTATTTGAGGAACTCCGCCGCCT-3′ and reverse 5′- ATTCCGGGTCACAGCAATGATGG -3′.

The primers for GAPDH were forward 5′-AAGGTCATCCCTGAGCTGAA-3′ and reverse 5′-TGACAAAGTGGTCGTTGAGG-3′.

GAPDH was used as an internal control, and the relative levels of mRNA were calculated by the 2[−∆∆CT] method.

A total of 1 × 10⁶ cells were harvested and then stained with PE-conjugated anti-PD-L1 antibody (Cat# 557924, BD Biosciences) or with the corresponding isotype-matched controls (Cat# 555749, BD Biosciences) for 30 min at room temperature. Cells were run on a Gallios Flow Cytometer (Beckman Coulter) and analysed using FlowJo software (TreeStar).

Data regarding mRNA expression, mutational status and survival time from The Cancer Genome Atlas for Lung Adenocarcinoma (TCGA, Provisional) were downloaded from cBioPortal (http://​cbioportal.​org). RNA-Seq V2 RSEM was utilized to analyse mRNA expression. The cut-off for high/low PKM2 based on mRNA expression for survival analysis was determined by Cutoff Finder. For the CD274 (PD-L1) gene, the cut-off for the PD-L1 high/low group was recorded according to the positive rate of PD-L1 expression by IHC and then checked for an association with survival. The Mann-Whitney test was used to compare PD-L1 levels between tumours with high PKM2 levels and tumours with low PKM2 levels, as PD-L1 levels were not normally distributed (P < 0.05 from the Kolmogorov-Smirnov normality test). Moreover, we compared survival differences between the high PKM2/high PD-L1 group and the low PKM2/low PD-L1 group.

Statistical analyses were performed using SPSS 22.0 (IBM Corp., Armonk, NY, USA) and GraphPad Prism 6.07 (GraphPad Software, Inc., La Jolla, CA, USA). Continuous variables were presented as the mean ± standard deviation and compared by unpaired Student’s t tests. The chi-square test or Spearman’s rank correlation analysis was used for the analysis of PKM2 and PD-L1 expression levels in tumour cells as categorical or continuous variables, respectively. The log-rank test was applied to test the relationship between mRNA expression and overall survival (OS)/disease-free survival (DFS), and the Kaplan-Meier method was used to depict survival curves. Cox proportional hazards regression models were used to perform univariate and multivariate survival analyses, and only variables whose P-values were less than 0.1 in the univariate analysis were included in the multivariate analysis. All statistical significance was set at P < 0.05 (two-sided).

The baseline clinicopathological characteristics of 173 primary lung adenocarcinoma tissue samples by IHC are listed in Table 1. The average age was 47 y (range 27–67 y). There were 124 (71.7%) female patients, and the majority of patients were never smokers (68.2%). Fifty-one (29.5%) patients were diagnosed with stage I, 66 (38.2%) with stage II, 43 (24.9%) with stage III, and 13 (7.5%) with stage IV disease. For the major driver mutational status, mutations in epidermal growth factor receptor (EGFR) and Kirsten rat sarcoma viral oncogene homologue (KRAS) were detected in 20 (11.6%) and 25 (14.4%) patients, respectively. Anaplastic lymphoma kinase (ALK) translocations were found in eight patients (4.6%) and were mutually exclusive with mutations in EGFR and KRAS. Immunohistochemical analysis of lung adenocarcinoma tumour tissues showed that PKM2 was primarily expressed in the cytoplasm of tumour cells, whereas PD-L1 was expressed both on the membrane and in the cytoplasm of tumour cells (Fig. 1). Considering that PD-L1 is mainly expressed on the cell surface, we only evaluated the membrane expression of PD-L1 and termed it TC-PD-L1. Finally, 65 (37.6%) cases displayed positive TC-PD-L1 staining (Fig. 2C). We next explored the correlation between PKM2 expression and tumour biology and observed that PKM2 expression was associated with N classification (P = 0.006) and TNM stage (P = 0.005); however, no significant correlation was observed between PKM2 expression and T classification (P = 0.114), M classification (P = 0.232) (Table 1). Nevertheless, no strong relationships were detected between PKM2 expression and other clinicopathological characteristics, including patient age, sex and smoking history.

First, we evaluated the correlation between PKM2 and PD-L1 based on mRNA expression using TCGA lung adenocarcinoma (LUAD, 506 samples) dataset. A positive correlation between PKM2 and PD-L1 was detected in TCGA LUAD cohort (r = 0.132, P = 0.003) (Fig. 2A). Subsequently, we compared PD-L1 expression between the PKM2 low (n = 441) and high (n = 65) expression groups and found that PD-L1 levels were significantly lower in patients with low PKM2 levels than in patients with high PKM2 levels (P < 0.001) (Fig. 2B). Moreover, we evaluated the correlation between PKM2 and PD-L1 protein expression in 173 primary lung adenocarcinoma tissue samples by IHC. We found that the protein expression of PKM2 was significantly correlated with TC-PD-L1 expression, assessed as categorical variables (Fig. 2C, P = 0.001). A similar positive association was observed when PKM2 and PD-L1 levels were assessed as continuous variables. Spearman’s rank correlation analysis suggested a positive correlation between PKM2 and PD-L1 expression levels (r = 0.287, P < 0.001; Fig. 2D). Therefore, there is a positive correlation between PKM2 and PD-L1 expression in lung adenocarcinoma patients.

To further understand the relationship between PKM2 and PD-L1 in lung adenocarcinoma tumour cells, we generated stable PKM2 knockdown cells by transfection with PKM2-specific short hairpin RNAs (shRNAs). The levels of PKM2 expression in stable PKM2 knockdown cells were confirmed by both qRT-PCR and Western blotting (Fig. 3A-B). We detected PD-L1 protein expression in stable PKM2 knockdown cells via Western blotting analysis and found that the protein level of PD-L1 was substantially downregulated (Fig. 3C). Moreover, we investigated cell surface PD-L1 protein levels (by flow cytometry) in the PKM2-shRNA and scramble-shRNA groups. The results of flow cytometry analysis revealed that the expression level of PD-L1 was more dramatically reduced in the PKM2-shRNA group compared to that in the scramble-shRNA group (Fig. 3D).

Kaplan-Meier survival analysis was performed in TCGA LUAD cohort to compare OS and DFS according to PKM2 and PD-L1 expression. Among the 605 patients included in the survival analysis, positive EGFR mutations, KRAS mutations and ALK translocations were detected in 34 (6.71%), 74 (14.6%) and 32 (6.32%) patients, respectively. To explore the possible relationships between survival and PKM2 and PD-L1 levels, we divided TCGA LUAD cohort into percentiles based on mRNA expression and identified cut-off points for low and high PKM2 at 0.871 and 0.955 for OS and DFS analyses, respectively. Patients with high PKM2 expression had a significantly shorter OS and DFS than patients with low PKM2 expression (P < 0.001 and P = 0.050, respectively; Fig. 4A-B). Moreover, the cut-off point was 0.624 and 0.628 for low and high PD-L1 expression for OS and DFS analyses, respectively. We found no statistically significant difference in OS among patients with high vs. low PD-L1 (P = 0.154), although patients with high PD-L1 expression seemed to have poorer OS than patients with low PD-L1 expression (Fig. 4C). However, higher PD-L1 expression remained associated with shorter DFS (P = 0.018, Fig. 4D). Next, we tested whether the combined expression of PKM2 and PD-L1 could improve survival differences between groups (high PKM2 & high PD-L1 vs. low PKM2 & low PD-L1) (Fig. 4E-F). Interestingly, we observed that the separation based on OS and DFS between these two groups (P < 0.001 for both) was more apparent than the separation between groups according to PKM2 expression alone (Fig. 4A-B, respectively). To make our results more conclusive, we used random number generators to randomly assign the available set of samples into two groups (a training set and a validation set) in a 1:1 ratio and validated the findings in the two groups (Fig. 5). Finally, we performed univariate and multivariate Cox regression analyses for OS and DFS. All significant clinical factors (P-values less than 0.1) in the univariate analysis were included in the multivariate Cox regression analysis (Table 2). As shown in Table 2, high expression of PKM2 and PD-L1 still powerfully and independently predicted poor OS and DFS, indicating that the concomitant PKM2 and PD-L1 expression could be a useful biomarker of response to therapy.