Introduction
Liver hepatocellular carcinoma (LIHC) is the main histological subtype of primary liver cancer. LIHC is highly aggressive and therapeutic options are limited. Thus, the prognosis for patients with LIHC is very poor [
1,
2]. The current drugs for treating LIHC include sorafenib, lenvatinib, and regorafenib [
3], which are multitarget tyrosine kinase inhibitors, and atezolizumab, pembrolizumab, nivolumab, and ipilimumab [
4], which are immunotherapeutic agents. Although these drugs have achieved significant success in the treatment of LIHC, treatment benefits are limited to a small subset of patients [
5,
6]. LIHC is an extremely heterogeneous tumor, which limits the efficacy of cancer therapies [
7]. Therefore, new effective diagnostic, prognostic, and therapeutic biomarkers based on single-cell analyses are urgently needed to develop personalized therapeutic strategies against LIHC.
Adenosine monophosphate-activated kinase (AMPK), a serine/threonine protein kinase, consists of AMPKα (catalytic core; α1 or α2), AMPKβ and AMPKγ (regulatory units; β1 or β2, and γ1, γ2, or γ3) [
8]. AMPK activates or inhibits metabolic-related pathways in response to changes in intracellular AMP/ATP ratios [
9]. Theoretically, human AMPK can form 12 different isoforms, depending on the combination of subunit subtypes. The expression of heterotrimeric complexes of AMPK varies widely in mammalian eukaryotic cells [
10]. AMPK subunits are distributed differently in tissues and organs, and the distribution may be related to the regulation of tissue-specific target molecules. In addition, many AMPK substrates are distributed in cells and tissues [
11]. However, the tissue and substrate specificity of AMPK complexes are unclear.
AMPKα2 is encoded by the
PRKAA2 gene.
PRKAA2 plays important roles in both tumor initiation and progression, including the regulation of mTOR kinase activity and the maintenance of NADPH levels [
12,
13]. Changes in
PRKAA2 expression have been linked to the occurrence, development, and prognosis of multiple tumor types, including breast cancer, ovarian cancer, gastric cancer, kidney cancer, and liver hepatocellular carcinoma [
14‐
16]. Thus,
PRKAA2 is a potential target for new therapeutic strategies.
PRKAA2 may influence tumor immunity in some cancer types. For instance, Zhang et al. constructed a risk model using
PRKAA2 and eight other genes. The resulting risk scores were closely linked to immunotherapy responses in patients with head and neck squamous cell carcinoma [
17]. Stromal cells, which are the most active cell type in the tumor microenvironment (TME), predominantly consist of endothelial cells, epithelial cells, fibroblasts, and immune cells, including T cells, B cells, neutrophils, and macrophages [
18]. Dynamic interactions between these cells are a major determinant of tumor pathophysiology [
19]. Therefore, single-cell analysis is needed to explore the functional roles of
PRKAA2 in the TME.
We demonstrated that AMPK subunits exhibit tissue-specific expression patterns and determine substrate specificity and physiological function. PRKAA2, which codes for the catalytic core of AMPK, was expressed at significantly high levels in LIHC, and expression of PRKAA2 was associated with poor prognosis. Thus, our studies focused on PRKAA2. The roles of PRKAA2 in LIHC and the relationship between PRKAA2 and the LIHC tumor immune microenvironment and malignant cell metabolism were explored. PRKAA2 may contribute to LIHC progression by inducing metabolic reprogramming of malignant cells and promoting immune escape of tumor cells. Moreover, patients with high PRKAA2 expression displayed an immune cold phenotype, while tumors with low PRKAA2 expression exhibited the opposite immune characteristics. Our results provide a theoretical foundation for developing PRKAA2-based strategies for individualized treatment of patients with LIHC.
Methods
Data acquisition
RNA-seq data for 31 normal tissues and LIHC samples were obtained from the Genotype-Tissue Expression (GTEx,
https://gtexportal.org/home/) and the Cancer Genome Atlas (TCGA,
https://www.cancer. gov/tcga) databases, respectively. The normal tissue types included adipose, adrenal gland, bladder, blood, blood vessel, bone marrow, brain, breast, cervix uteri, colon, esophagus, fallopian tube, heart, kidney, liver, lung, muscle, nerve, ovary, pancreas, pituitary, prostate, salivary gland, skin, small intestine, spleen, stomach, testis, thyroid, uterus, and vagina. Single-cell RNA-seq (scRNA-seq) data for LIHC were collected from the China National Genebank Database (CNGBdb,
https://db.cngb.org/search/project/CNP0000650); accession code: CSE0000008).
Correlation analysis
Based on prior reports in the literature, 106 AMPK substrates were extracted [
11]. The expression correlation between AMPK subunits and AMPK substrates in 31 normal tissues was determined by Pearson correlation analysis (Correlation coefficients > 0.2).
Differential expression analysis of AMPK subunits
To explore differences in the expression of AMPK subunits between tumor samples and their matched normal tissue controls in pan-cancer, we used the gene set cancer analysis (GSCA) online tool (
http://bioinfo.life.hust.edu.cn/GSCA/). A false discovery rate-adjusted
p-value less than 0.05 indicated a significant difference.
Survival and immune infiltration analyses
Based on the median PRKAA2 level, patients were divided into high- and low-expression subgroups. The R packages “survival” and “survminer” were employed for Kaplan–Meier survival analyses. The abundance of each infiltrating immune cell type was analyzed using a single-sample gene set enrichment analysis. Wilcoxon signed-rank test was performed and p less than 0.05 implied a significant difference.
Single-cell RNA-seq data processing
The scRNA-seq matrices were collected for more than 300 transcripts/cell and more than 3 cells/gene condition. The NormalizeData function from the R package “Seurat” was applied to normalize scRNA-seq data. The clustering analysis was conducted based on the integrated joint embedding generated by the Harmony algorithm. The top 15 principal components and the top 2000 variable genes were selected for subsequent analysis. Cell clusters were detected using the FindClusters function in Seurat (resolution = 0.6), and cell clustering results were visualized by uniform manifold approximation and projection or t-distributed stochastic neighbor embedding (t-SNE) analysis.
Cell–cell communication analysis
We performed cell communication analysis using CellphoneDB [
20]. Average expression levels were calculated based on the annotated ligand-receptor pairs attained from the STRING database. The ligand-receptor pairs with
p < 0.05 values were identified, and the interactions between two cell types using these identified pairs were analyzed. Cytokines play a critical role in cell communication; thus, cytokine signaling, based on the transcriptomic profiles, was determined using CytoSig.
Pseudotime analysis
To detect the association between
PRKAA2 expression and T-cell evolution during tumor progression in single cells, a Pseudotime analysis was performed using R packages “
monocle2” [
21]
. The monocle subject was built by applying the function “newCellDataSet”. Trajectory analysis was performed based on the differentially expressed genes determined by the R package “
Seurat”. The “reduceDimension” function was used for dimensionality reduction, and cells were placed on Pseudotime trajectories using the “orderCells” functions.
Single-cell metabolic activity was evaluated using the R package “scMetabolism” based on a previously reported method [
22]. Differences in the metabolic pathway scores between subgroups with high or low
PRKAA2 expression were analyzed using Wilcoxon signed-rank test, and
p less than 0.05 implied a significant difference.
Functional enrichment analysis
Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) analyses were conducted using Metascape. Genes enriched more than 3-fold and p less than 0.05 were considered significantly different.
Cell culture and gene expression assays
The human LIHC cell line, HepG2, was acquired from the American Type Cell Culture Collection. Cells were grown in Dulbecco Modified Eagle Medium (Gibco, Thermo Fisher Scientific, Waltham, MA, USA) containing 10% fetal bovine serum (FBS, BI) at 37 °C and 5% CO2. PRKAA2 knockout cells were generated by transfecting cells with the lentiviral-based short hairpin RNA (shRNA) vector pGPU6/GFP/Neo (Genechem, Shanghai, China). The shRNA sequences were as follows: shRNA-1: GTGGCTTATCATCTTATCATT and shRNA-2: GTCATCCTCATATTATCAAAC. PRKAA2 mRNA levels were measured using quantitative real-time PCR (qPCR). Total RNA was extracted using an RNA extraction reagent (Takara, 9108, Japan), and 2X Super SYBR Green qPCR Master Mix (ES Science, Guangzhou, China) was used to detect PRKAA2 levels. The qPCR primer sequences for PRKAA2 and GAPDH were as follows: PRKAA2 forward, 5′-CGGGTGAAGATCGGACACTA-3′;PRKAA2 reverse, 5′-TCCAACAACATCTAAACTGCGA′; GAPDH forward, 5′-GACCTGACCTGCCGTCTA-3′;and GAPDH reverse, 5′-AGGAGTGGGTGTCGCTGT-3′.
Cell proliferation, clonogenic, migration, and invasion assays
Cells (8000/well) were seeded into 96-well plates (Servicebio, WuHan, China), and proliferation was evaluated using Cell Counting Kit-8 (Biosharp Biotechnology, Beijing, China) reagent. The optical density was read at 450 nm. For clone formation assays, cells were cultured for 14 days, fixed with 4% paraformaldehyde for 15 min, and stained with 0.1% crystal violet for 15 min (Meilunbio, Dalian, China). For scratch assays a linear wound was created in a monolayer of serum-starved cells using a 10-μL pipette tip, and cell coverage across this line was determined. For Transwell migration assays, cells were seeded into the upper chamber (8 μm; BIOFIL, Guangzhou, China) and incubated with serum-free medium. The lower chamber contained medium with 10% FBS. After 24 hours, cells were fixed with 4% paraformaldehyde and stained with 0.1% crystal violet.
Statistical analysis
Continuous variables were compared using Student’s T-test (parametric) analyzed with GraphPad Prism 8 software or Wilcoxon rank sum test (nonparametric) analyzed with R software package. In vitro experiments were repeated three times, and all results are presented as means ± standard deviation. A p-value less than 0.05 indicates a significant difference.
Discussion
Individualized cancer therapy is paramount to improving the clinical outcomes of patients [
23]. The current study reveals that AMPK plays pivotal roles in the progression and metastasis of multiple tumor types and is a potential therapeutic target for LIHC. AMPKα2, encoded by
PRKAA2, functions as the catalytic core of AMPK; however, the role of AMPKα2 in the LIHC TME is unclear.
Our results show that AMPK subunits exhibited tissue-specific expression patterns and could be substrate-specific. PRKAA2 was highly expressed in LIHC and was associated with poor prognosis. In vitro experiments show that PRKAA2 knockdown inhibited the proliferation, migration, invasion, and metastasis of LIHC cells. Furthermore, PRKAA2 expression was significantly associated with the tumor immune microenvironment. Patients with high PRKAA2 expression lacked immune cell infiltration, indicative of an immune cold phenotype. Single-cell transcriptome atlas analysis showed that PRKAA2 contributes to tumor progression. This conclusion is supported by the following evidence: 1) TGF-β signaling and EMT are enhanced in malignant cells; 2) metabolic reprogramming is induced in malignant cells; 3) CD8+ T-cell exhaustion and the formation of CD4+ Treg cells is promoted in T cells; and 4) Treg cell activation and T-cell depletion is driven by changes in the interactions between malignant cells and T cells in the TME. Taken together, our study demonstrates that the development of PRKAA2-based treatment strategies for LIHC holds great promise.
The tissue specificity of AMPK isoform distribution is the basis for a variety of AMPK biological functions [
24]. AMPK can exert antitumor or tumor-promoting effects depending on the cellular context [
25,
26]. AMPK activation leads to cell cycle arrest and inhibition of tumor growth, which contribute to the prevention of multiple cancer types, including lung, colorectal, and breast cancers [
27‐
29]. In contrast, under conditions of oncogenic stress or hypoxia and nutrient deficiency in the TME, cancer cells exhibit an increased dependence on AMPK function to promote cancer cell survival [
30,
31]. Our data show that functional differences between AMPK subunits could be mediated by tissue-specific expression patterns and the high substrate specificity of AMPK subunits. High expression of
PRKAA2, which drives metabolic reprogramming and immune escape of tumor cells, contributes to LIHC development.
Our results demonstrate that the TME is closely linked to tumor heterogeneity and regulates antitumor immune responses. Among the immune cell types in the TME, T cells play a dominant role in immune regulation and antitumor activity [
32]. Treg cells are an important component of immune homeostasis. Treg cells maintain immune self-tolerance and inhibit anticancer immunity [
33], while CD8+ T cells are cytotoxic T lymphocytes that kill tumor cells [
34]. CD8+ T cells may become exhausted during tumor progression. Our findings suggest that
PRKAA2 promotes immune escape of tumor cells via CD8+ T-cell depletion and Treg cell generation, eventually leading to tumor progression. Furthermore, malignant cells that express high levels of
PRKAA2 evade immune suppression via IFN-γ/JAK/STAT-mediated loss of MHC-I molecules [
35,
36]. Altogether,
PRKAA2-mediated tumor immune escape may be due to the activation of immune escape mechanisms in malignant cells and the formation of an immunosuppressive tumor microenvironment.
Metabolic reprogramming is a crucial pathway for the proliferation and metastasis of tumor cells [
37]. Thus, understanding the features of malignant cell and non-malignant cell metabolism is important in developing a foundation for LIHC patient therapy. Our results show metabolic heterogeneity in LIHC, and malignant cells have significantly higher metabolic activity than non-malignant cells. Metabolic pathways, including oxidative phosphorylation, the TCA cycle, and glycolytic signaling, were enhanced in malignant cells that express high levels of
PRKAA2, indicating that high
PRKAA2 expression enhances energy metabolism in malignant cells. Oxidative phosphorylation was remarkably upregulated in malignant cells, which is consistent with previous single-cell studies [
38]. Interestingly, oxidative phosphorylation signaling significantly and positively correlated with hypoxia. Oxidative phosphorylation, as a sensor of oxygen availability, modulates the responses to hypoxia by stabilizing hypoxia-inducible factor [
39,
40]. In support of this observation, highly dynamic interactions between oxidative phosphorylation and hypoxia were detected in another tumor study using scRNA-seq. Consequently, the positive association between oxidative phosphorylation and hypoxia-mediated by
PRKAA2 may be a unique signature of malignant cells within the TME. Collectively, the data suggest that
PRKAA2 plays an important role in the metabolic reprogramming of malignant cells.
Patients who were split into two groups according to the PRKKA2 expression exhibited different tumor immune microenvironments, suggesting that these patients may respond to distinct treatment strategies. Tumors with low PRKAA2 expression displayed an immune hot phenotype, characterized by a high abundance of tumor immune cell infiltrates. In contrast, patients with high PRKAA2 expression exhibited the opposite immune characteristics, indicative of an immune cold phenotype. Therefore, inhibiting PRKAA2 expression may convert poorly immunogenic (cold) tumors into highly immunogenic and invasive (hot) tumors. Blocking PRKAA2 expression may restore antitumor immunity by enhancing the antitumor response of T cells and reshaping the tumor immunosuppressive microenvironment. The stratification of PRKAA2 expression patterns can be used for developing personalized treatment approaches, contributing to the establishment of precision medicine for LIHC.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.