Introduction
Esophageal carcinoma (ESCA) is a common type of malignant tumors, which is divided into two main subtypes: esophageal squamous cell carcinoma (ESCC) which accounts for 90% of the ESCA, while the remaining is esophageal adenocarcinoma (EA) [
1]. Due to lack of specific treatment for ESCC and methods for its early diagnosis, five-year survival rate remain below 15% [
2].
18F-fludeoxyglucose positron emission tomography/computed tomography (
18F-FDG PET/CT) is a noninvasive and preferred early diagnostic tool, which is widely employed in the assessment of response to therapy, by simultaneous assessment of tumor morphology and metabolism over time [
3]. Tumor FDG uptake is determined by the expression of glucose transporter family (GLUT1, 3 and 4) and glycolytic enzymes [
4]. The PET-related parameters include the maximal and mean standardized uptake values (SUVmax and SUVmean), metabolic tumor volume (MTV), and total lesion glycolysis (TLG), which correlates with biological factors in the tumor microenvironment (TME) [
5,
6]. Significant correlation was found between PET imaging parameters and the in vivo biological characterization of cancer lesions [
7‐
9]. Our previous data also demonstrated that
18F-FDG PET/CT parameters have important clinical value in predicting novel molecular and clinical phenotypes of malignant tumors. The phenotypes include nucleophosmin 1 (NPM1) expression in lung cancer [
10], eukaryotic translation initiation factor 2 Subunit β (EIF2S2) expression in colorectal cancer [
11], and methyltransferase 3 (METTL3) expression in ESCA [
12].
Eukaryotic initiation factor 6 (eIF6) is the first eIF associated with large 60S subunit in the nucleus, which is involved in translation initiation [
13‐
15]. Previous studies demonstrated that eIF6 is highly expressed in human cancers such as colorectal cancer, hepatocellular carcinoma, ovarian serous carcinoma, acute promyelocytic leukemia, non-small cell lung cancer, and affect lymphoma genesis and tumor progression [
16‐
20]. In addition, eIF6 translational activity regulate fatty acid synthesis and glycolysis through upregulation of lipogenic and glycolytic enzymes [
21]. Scagliola et al. [
22] showed that eIF6 induces a metabolic rewiring during progression from non-alcoholic fatty liver to hepatocellular carcinoma. However, the functions and molecular mechanisms of eIF6 in ESCA remain unclear. FDG uptake reflects glucose metabolism in malignant cells [
6]. This study evaluated the relationship between eIF6 expression and glucose metabolism using FDG-PET and examined the prognostic significance in ESCA patients.
Both metabolic and immune-mediated syndrome in the TME have been associated with malignant tumor progression and metastasis [
23,
24]. Recent studies have shown that there is overexpression of eIF6 mRNA in activated T cells and lymphoid cells [
16,
25]. Besides, eIF6 was shown to be essential in overall immune system, particularly for metabolic switch in CD4
+ T cell activation [
26]. On the other hand, potential of immune checkpoint inhibition (ICI), such as blockage of programmed cell death-1 (PD-1)/ PD-1 ligand 1 (PD-L1) pathway, has been established in several tumors [
27,
28]. However, this treatment option has not been applied to all patients, thus the need to find new targets.
In this study, we correlated the eIF6 expression with patient survival using public database [
29], as well as
18F-FDG PET parameters in ESCA patients. We then used ESCA cell lines to elucidate the functions of eIF6. In addition, we attempted to investigate the association between the eIF6 expression and gene signatures of immune cells in ESCA. Our data showed that eIF6 might be a potential diagnostic and prognostic biomarker in ESCA patients.
Materials and methods
Patient samples
The study respectively reviewed fifty-five patients who were surgically treated and had pathologically confirmed ESCC in Taihe Hospital from January 2018 to July 2020. This study included patients who underwent 18F-FDG PET/CT imaging and were analyzed using immunohistochemistry (IHC) tests. The patients had available clinicopathological data, and none of them received chemo- or radiotherapy prior to the 18F-FDG PET/CT imaging. In this study, we used thirteen paired surgically obtained samples from patients with ESCA. The fresh-frozen ESCA tissues and matched non-tumor tissues were analyzed by qRT-PCR and Western blot.
18F-FDG PET/CT imaging and data analysis
Glucose metabolism was assessed by
18F-FDG PET/CT imaging using the Biograph mCT (64) system (Siemens Healthcare, Germany). As described [
10,
12], the patients fasted for at least 6 h but had free access to water until the start of the imaging process. Whole-body position was monitored 50 min after intravenous administration of FDG (3.7–4.1 MBq/kg) and lasted for about 15 min. The PET/CT images were acquired following the manufacturer’s protocol. The metabolic parameters of the
18F-FDG PET images were evaluated by two experienced nuclear medicine physicians who were blinded to the final clinical diagnosis. Briefly, a region of interest (ROI) was placed around the primary tumor, and then FDG uptake in the lesions was determined from the PET images to calculate the maximum and mean SUV (SUV
max and SUV
mean). SUV is defined as tissue concentration (Bq/g) × lean body mass (g)/injected dose (MBq)). A SUV
max threshold of 2.5 was used to define the metabolic tumor volume (MTV) (cm
3), while total lesion glycolysis (TLG) is obtained by multiplying SUV
mean by MTV (TLG = ΣMTV × SUV
mean) [
30,
31]. In addition, the MTV and TLG were defined semi-automatically using an SUV-based platform as previously described [
12].
Quantitative real time PCR
Total RNA was isolated from frozen tissues and cell lines using TRIzol reagent (Thermo Fisher Scientific, USA). The RNA was used to synthesize cDNA with RT Master Mix kit (TaKara, China). The qRT-PCR experiment was performed using a TB Green Premix Ex Taq Kit (TaKaRa, China) in the Applied Biosystems ViiA TM 7 Real-time PCR system (Life Technologies, CA). We used β-actin as an internal control for normalization. The following primers were used:
-
eIF6, Forward: 5′-GGCCGACCAGGTGCTAGTAGG-3′; Reverse: 5′- CACAC-CAGTCATTCACCACCATCC-3′.
-
β-actin, Forward: 5′-TCTTCCAGCCTTCCTTCCT-3′; Reverse: 5′- AGCACTG-TGTTGGCGTACAG -3′.
Western blot
As previously described [
19], fresh tissues or cells were collected and digested with RIPA buffer containing 1% protease inhibitor on ice. The protein concentrations were measured by BCA method (Beyotime Biotechnology) and then resolved in SDS-PAGE gel. The protein samples were transferred onto PVDF membranes (Millipore, USA), followed by incubation with primary antibodies; anti-eIF6 (Thermofisher, Waltham, MA, USA), anti-GLUT1 (Abcam, Cambridge, UK), anti-Ecadherin (BD Pharmigen, USA), anti-Vimentin (Abcam), anti-Cytochrome c (Cyt c, Abcam), anti-CD45 (Proteintech, China), anti-CD11b (Abcam), anti-PD-L1 (Proteintech) or anti-GAPDH (Cell Signaling Technology, USA) at 4 °C overnight. The membranes were washed with TBST, and then incubated with secondary antibodies (Cell Signaling Technology, USA) for 1 h at room temperature. Finally, the proteins were visualized by an enhanced chemiluminescence (ECL) detection kit.
Immunohistochemical (IHC) assay
IHC staining was performed following a previously described protocol [
10]. Briefly, the samples were dissected on 5-μm-thick tissue sections embedded in paraffin, and then incubated with antibodies against eIF6 (1:400), GLUT1 (1:200), followed by secondary antibodies. Two pathologists who were blinded to clinical data independently analyzed the IHC data. The protein expression was profiled based on the staining intensity score, from 0 (negative), 1 + (weakly positive), 2 + (moderately positive) to 3 + (strongly positive). High- or low-expression was defined with scores equal to or above, and below the final IHC staining score 2, respectively.
Gene expression pattern and patient prognosis in public datasets
We downloaded the mRNA expression (HTSeq counts) and associated clinical data for different human cancers from the TCGA (
https://portal.gdc.cancer.gov/). The expression profiles were plotted in the ggplot2 R (
https://github.com/tidyverse/ggplot2). On the other hand, survival curve was demonstrated using Kaplan–Meier plots [
32]. To evaluate the predictive accuracy of eIF6 in TCGA cancers, “pROC” in R package was used to generate the time-dependent receiver operating characteristic (ROC) curve and then the area under curve (AUC) was calculated.
Enrichment analysis and functional networks in ESCA
The eIF6 co-expression genes in the TCGA-ESCA were analyzed by DESeq2 R package. We then performed correlation analysis of the different variables using the Spearman’s or Pearson’s correlation test and Fisher’s exact test. Gene ontology (GO) biological process and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis were performed using the clusterProfiler in R package (p.adj < 0.1 & qvalue < 0.2). Thereafter, top 200 protein coding genes significantly correlated with eIF6 were screened out for construction of PPI networks. Each network node was determined and visualized with CytoHubba plugin in Cytoscape. The top 10 genes with most connections were selected and considered to be ‘hub’ genes.
Next, the TCGA-ESCA samples were classified into two groups based on the eIF6A expression status. The edgeR package in Bioconductor was used to define differentially expressed genes (DEGs) between the eIF6 high and low ESCA samples (|log2FC|> 1.5, p value < 0.05). In addition, we used the GSEA to investigate meaningful biological processes. The hallmark gene sets and glycolysis signatures (REACTOME_GLYCOLYSIS, KEGG_GLYCOLYSIS_GKYCIBEIGENES, and HALLMARK_GLYCOLYSIS) were downloaded from the Molecular Signatures Database (MSigDB). The false discovery rate (FDR) cutoff of 0.25, and an adjusted p < 0.05 were considered significance. Besides, 282 glycolytic genes in three glycolysis signatures were downloaded from the MSigDB database. Pearson correlation coefficients of the eIF6 expression and glycolytic genes were calculated using TIMER database and DESeq2 R package in the TCGA database.
Cell culture and transfection
Human epithelial cell line (HET1A cells) and ESCA cell lines (Eca109, KYSE30, KYSE150, CEC2) were obtained from the Cell Bank of Chinese Academy of Sciences (Shanghai, China). The cells were transfected with eIF6 siRNA using Lipofectamine 8000 transfection reagent (Beyotime, China), following the manufacturer’s instructions. The target sequences were: 5ʹ-CTGCTTTGCCAAGCTCACCAA-3ʹ for siRNA-1 (sieIF6-1) and 5ʹ-CTGGTGCATCCCAAGACTTCA-3ʹ for siRNA-2 (sieIF6-2).
Cell viability analysis using MTS assay and EdU proliferation assay
The ESCA cells were seeded into 96-well plates and transfected with siRNA. After transfection for 0, 24, 48 or 72 h, we added MTS (CellTiter 96 Aqueous One Solution Cell Proliferation Assay, Promega) to each well and then incubated for 2 h. The optical density (OD) was then measured at 490 nm using a microplate reader (SpectraMax M3). Another proliferation assay was performed with the usage of an EdU kit (ClickTM, EDU 488, Beyotime) according to the manufacturers instruction. In brief, the transfected cells were incubated with diluted EdU (10 uM) for 2 h. Subsequently, cells were fixed and stained by Click Reaction Mixture and DAPI. Images were captured from five random fields under a fluorescent microscope. Finally, percentage of EdU positive cells was quantitated using Image J software.
Flow cytometry analysis of apoptosis
Cell apoptosis was quantified by an Annexin V-FITC Apoptosis Detection Kit (BD Pharmigen, USA). The treated cells were harvested and washed twice with PBS. Afterwards, the cells were resuspended and then 3 µl FITC-conjugated annexin V and 10 µl propidium iodide staining solution were added. The rate of apoptosis was immediately measured using a FACS Calibur flow cytometer (BD Bioscience, USA).
Wound healing assay
The cells were plated overnight and scratched by a 10 μL pipette tip. Subsequently, the cells were washed with sterile PBS and then refilled with complete medium. After transfection with siRNA, the wounded cells were analyzed every day by a microscope and then analyzed using Image J software.
Estimation of immune cell characteristics
TCGA-ESCA patients were classified into low-expression or high-expression group based on the eIF6 expression. The xCell in R package was employed to evaluate immune and stroma cell abundance in the two groups. The relative immune signature score was estimated using the R GSVA package [
33,
34]. In addition, the Tumor Immune System Interactions (TISIDB) database was used to study immune-related gens and immune subtypes, based on the eIF6 expression [
35]. On the other hand, the Tumor Immune Estimation Resource (TIMER,
https://cistrome.shinyapps.io/timer/) was used to define the correlation using the spearman correlation coefficient of a pair of genes and then estimated statistical significance, as well as tumor purity in ESCA. In addition, the correlation between the eIF6 expression and different immune signatures in ESCA samples in the GEPIA2 databases (
http://gepia2.cancer-pku.cn) was calculated using Pearson’s correlation test.
Statistical analysis
SPSS package (version 16.0, SPSS for Windows, 2007) and R package version (4.0.3) were used for statistical analyses. A p < 0.05 was used to define statistical significance. In this study, the survival and ROC analysis were carried out in R or corresponding R packages which included survival, survminer and pROC. For bivariate analysis, χ2 test, χ2 test or Fishers exact test were used. Spearman’s rho test was used to evaluate the correlation between the PET parameters and eIF6 expression levels in the ESCA patients. In addition, we used the greatest Youden index (sensitivity + specificity – 1) to define optimal cutoff value for the ROC curve. The T-test and Wilcox tests were used to compare differences between two groups, while comparisons among multiple groups were performed using the Tukey and the Wilcox tests. Statistical analysis and visualization were performed using the R Statistical Package and ggplot2 package.
Discussion
The eukaryotic translation initiation factor eIF6 plays an essential role in cell growth and transformation, apoptosis, mitochondrial respiration, as well as lipogenic and glycolytic process [
15,
21,
22,
36]. Although overexpression and oncogenic functions of eIF6 have been documented in other cancers, the role and biological functions of eIF6 in ESCA remains poorly understood. Herein, our results demonstrated that eIF6 was up regulated both in ESCA tissues and cell lines, and high eIF6 expression led to poor prognosis in EA patients, with a favorable diagnostic reference value in ESCA. Moreover, knockdown of eIF6 significantly suppressed cell proliferation and migration, and induced cell apoptosis in the ESCA cells. The PET parameters have previously shown a remarkable potential for predicting gene expression status in cancers [
37‐
39]. Our IHC analysis revealed that tumor eIF6 expression was positively correlated with FDG PET parameters in ESCA tissues. In addition, SUV
max and SUV
mean might act as suitable predictors of eIF6 expression in patients with ESCA. On the other hand, the bioinformatics analyses suggested that eIF6 is involved in metabolic pathways and tumor immune infiltration in ESCA. These findings emphasize the oncogenic role of eIF6 and its therapeutic potential in the ESCA.
Our bioinformatics analyses demonstrated that high eIF6 expression predicted worse prognosis in ESCA patients. This was in sync with observations for human colorectal cancer, biliary tract cancer, hepatocellular carcinoma and lung adenocarcinoma [
17,
19,
20,
40]. A previous study showed that eIF6 knockout mice impacted Myc-induced lymphomagenesis and tumor progression by modulating p53 [
15]. Another study showed that high eIF6 expression is significantly associated with clinicopathological features, such as lymph node metastases in ovarian serous carcinoma [
18]. Through IHC staining analysis, Gantenbein et al. [
17] demonstrated significant differences in the eIF6 expression between higher and lower grade lung adenocarcinoma, but not in lung squamous cell carcinoma. In our study, eIF6 knockdown in ESCA cells inhibited cell migration via regulation of EMT marker genes. However, our data showed that eIF6 expression was not significantly associated with lymph node metastasis and p stages in ESCA tumors, probably due to different tumor-specific roles of eIF6 or small sample size.
Previously, Scagliola et al. [
22] postulated that eIF6 depletion delays the liver disease progression from non-alcoholic fatty liver disease to hepatocellular carcinoma in vivo. Mechanistically, eIF6 depletion regulated mitochondrial respiration by targeting the mTORC1-eIF4F-YY1 translational axons. Cancer cells preferentially use aerobic glycolysis for stimulate ATP generation and lactate production (Warburg effect) [
41]. Interestingly, inhibition of eIF6 could reduce cell growth by impairing lactate and ATP production in Malignant Pleural Mesothelioma [
42]. Moreover, a previous systematic review demonstrated that eIF6 promoted glycolytic flux and fatty acid synthesis and increased tumor viability. But the mechanism through which mTOR or Myc regulate the activity of eIF6 need to be further investigated [
43]. Our GSEA results demonstrated that the highly expressed eIF6 in ESCA patients regulated several metabolic pathways such as HALLMARK_MYC_TARGETS, HALLMARK_OXIDATIVE_PHOSPHORYLATION, REACTOME_GLYCOLYSIS, and KEGG_GLYCOLYSIS_GKYCIBEIGENES. In recent years,
18F-FDG PET/CT has emerged as a noninvasive diagnostic tool for evaluating tumor glycolytic activity, diagnosing and staging of various malignant tumors [
44,
45]. Molecular imaging can be applied to reveal the molecular profile of cancers [
38,
46,
47]. This study found that high eIF6 expression was positively correlated with FDG uptake (in terms of SUV
max and SUV
mean) and the expression of GLUT1 in ESCA tissue. These results are in agreement with data from previous studies [
12]. In addition, it was found that the SUV
max cutoff for PET/CT parameter was 18.2, which was more effective compared to the SUV
mean, TLG, and MTV in predicting eIF6 expression. For these reasons, noninvasive methods, such as molecular imaging, could be used for predicting eIF6 status have great clinical relevance.
Evidence from previous studies has shown that immune cells or immune-related biomarkers in the TME can predict the survival outcomes and influence response to immune checkpoint therapy [
48,
49]. By combining contrast-enhanced CT images and RNA-seq genomic data from tumour biopsies, Roger et al. demonstrated that imaging biomarker could be useful in estimating CD8 cell count and predicting clinical outcomes of patients treated with anti-PD-1 and PD-L1[
49]. The dual roles of mutation burden and MS-indels was identified in predicting outcomes of central nervous system and synchronous cancers following immune checkpoint inhibitors (ICIs) treatment [
50]. Thus, analysis of the cancer-specific immune biomarkers may reveal novel molecular targets for ESCA treatment. Here, using the xCell algorithm analysis, eIF6 was found to be negatively correlated with CD4 + T cells, hematopoietic stem cell, NK cell, T helper cells, macrophages, T cells, Th2 cells, Th1 cells, iDCs, among others. Moreover, the expression of eIF6 were negatively correlated with immune marker genes, such as genes of monocytes, TAMs, M2 macrophages, T cell, Th1, and Tfh phenotype. These result demonstrate that eIF6 expression may be negatively correlated with immune cells including macrophages, T cells, and Th1 cells. We subsequently verified that showed that eIF6 silencing increased the expression of macrophage markers CD45 and CD11b in ESCA cells. Previous studies suggested that cancer cells also acquired immune regulatory membrane proteins such as PD-L1, CD4, CD45, CTLA4 and Tim3 expressed in lymphocytes, which in turn contribute to the development of the immunosuppressive tumor microenvironment [
51‐
53]. Whether CD45 or CD11b -positive cancer cells could regulate the ESCA tumor microenvironment remains further exploration. Consistently, a previous study revealed that eIF6 overexpression increase the number of activated T cells [
16,
25]. Moreover, eIF6 overexpression induced the metabolic switch in CD4
+ T cells [
26], and was negatively correlated with most immunomodulators (chemokines, MHC-s, immune stimulators) in ESCA. Similarly, we found a negative correlation between eIF6 expression and immune checkpoints, including CTLA4, and HAVCR2, in ESCA. It has been reported that CTLA4 activates CD4 + and CD8 + T cells, whereas TIM-3 functions as a negative regulator of T cell activation and also involved in exhaustion, Th1 responses [
54,
55]. Inhibitors of CTLA4 and TIM-3 can be applied as immunotherapeutic targets in the treatment of cancer patients [
56,
57]. This suggests that eIF6 might serve as a potential biomarker of immune cell infiltration in ESCA.
There are several limitations to this study. First, although we found that high expression of eIF6 was associated with poor prognosis of ESCA patients, this finding was not experimentally validated using clinical samples. Second, although metabolic and genomic signatures were combined to investigate potential biomarkers and mechanisms, more samples and additional validation studies were needed. Third, further in-depth clinical research should be conducted to clarify the functional role of eIF6 in the tumor immunosuppressive microenvironment.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.