Article Text

Download PDFPDF

Original article
Localisation of PGK1 determines metabolic phenotype to balance metastasis and proliferation in patients with SMAD4-negative pancreatic cancer
  1. Chen Liang1,2,3,4,
  2. Si Shi1,2,3,4,
  3. Yi Qin1,2,3,4,
  4. Qingcai Meng1,2,3,4,
  5. Jie Hua1,2,3,4,
  6. Qiangshen Hu1,2,3,4,
  7. Shunrong Ji1,2,3,4,
  8. Bo Zhang1,2,3,4,
  9. Jin Xu1,2,3,4,
  10. Xian-Jun Yu1,2,3,4
  1. 1Department of Pancreatic Surgery, Fudan University Shanghai Cancer Center, Shanghai, China
  2. 2Department of Oncology, Shanghai Medical College Fudan University, Shanghai, China
  3. 3Department of Pancreatic Surgery, Fudan University Shanghai Cancer Center, Shanghai, China
  4. 4Pancreatic Cancer Institute, Fudan University, Shanghai, China
  1. Correspondence to Professor Xian-Jun Yu, Department of Pancreatic Surgery, Fudan University Shanghai Cancer Center, Shanghai 200032, China; yuxianjun88{at}hotmail.com

Abstract

Objective Pancreatic ductal adenocarcinoma (PDAC) is the most aggressive type of GI tumour, and it possesses deregulated cellular energetics. Although recent advances in PDAC biology have led to the discovery of recurrent genetic mutations in Kras, TP53 and SMAD4, which are related to this disease, clinical application of the molecular phenotype of PDAC remains challenging.

Design We combined molecular imaging technology (positron emission tomography/CT) and immunohistochemistry to evaluate the correlation between the maximum standardised uptake value and SMAD4 expression and examined the effect of SMAD4 on glycolysis through in vitro and in vivo experiments. Furthermore, we identified the effect of SMAD4 on metabolic reprogramming by metabolomics and glucose metabolism gene expression analyses. Dual luciferase reporter assays and chromatin immunoprecipitation were performed to identify whether SMAD4 functioned as a transcription factor for phosphoglycerate kinase 1 (PGK1) in PDAC cells. Proliferative and metastatic assays were performed to examine the effect of PGK1 on the malignant behaviour of PDAC.

Results We provide compelling evidence that the glycolytic enzyme PGK1 is repressed by transforming growth factor-β/SMAD4. Loss of SMAD4 induces PGK1 upregulation in PDAC, which enhances glycolysis and aggressive tumour behaviour. Notably, in SMAD4-negative PDAC, nuclear PGK1 preferentially drives cell metastasis via mitochondrial oxidative phosphorylation induction, whereas cytoplasmic PGK1 preferentially supports proliferation by functioning as a glycolytic enzyme. The PDAC progression pattern and distinct PGK1 localisation combine to predict overall survival and disease-free survival.

Conclusion PGK1 is a decisive oncogene in patients with SMAD4-negative PDAC and can be a target for the development of a therapeutic strategy for SMAD4-negative PDAC.

  • pancreatic cancer
  • SMAD4
  • PGK1
  • metabolic phenotype
  • metastasis
View Full Text

Statistics from Altmetric.com

Significance of this study

What is already known on this subject?

  • Pancreatic cancer is one of the most lethal diseases.

  • Thus far, most agents have failed to significantly improve patient survival.

  • SMAD4 is a well-known driver gene in pancreatic ductal adenocarcinoma (PDAC), and its inactivation is prevalent in pancreatic carcinoma (50%).

  • PDAC is largely regarded as a single disease partly because of the lack of an effective classification system.

  • PDAC is a heterogeneous disease characterised by metabolic plasticity.

What are the new findings?

  • SMAD4 mutations are involved in metabolic reprogramming in PDAC.

  • SMAD4 inhibits the transcription of phosphoglycerate kinase 1 (PGK1), so PGK1 expression is deregulated in SMAD4-negative pancreatic cancer cells.

  • PGK1 translocates into the nucleus to function as a transcription factor, which inhibits the promoter activity of E-cadherin.

  • The effect of nuclear PGK1 is different from the metabolic effect of PGK1 as a glycolytic enzyme in the cytoplasm; thus, the subcellular location of PGK1 balances proliferation and metastasis in PDAC.

How might it impact on clinical practice in the foreseeable future?

  • The combination of SMAD4 and PGK1 can aid in the prediction of metastatic patterns, especially for SMAD4-negative PDAC.

  • Our results showed why SMAD4 expression may be an important factor for stratifying patients with PDAC into different treatment strategies, such as systemic chemotherapy or local control.

  • The molecular classification scheme is improved to distinguish four phenotypes with distinct prognoses and biological features.

Introduction

Pancreatic ductal adenocarcinoma (PDAC) is the most aggressive GI tumour, with a 5-year overall survival (OS) rate <8%.1 2 Compared with other tumours in the era of targeted therapy, targeted drugs for PDAC remain scarce.3 Although there has recently been great advancement in our understanding of the molecular classification of PDAC, clinical application of the molecular phenotype of PDAC remains challenging.4–7 Improving the understanding of PDAC biology is clearly and urgently needed and will contribute to the development of the molecular classification of PDAC as well as targeted therapies.

A variety of tumour types exhibit allelic loss of chromosome 18q to mediate SMAD4 inactivation.8 9 Notably, whereas SMAD4 inactivation is prevalent in PDAC (50%–60%), it is distinctly uncommon (<10%) in other tumour types, suggesting that SMAD4 is a specific tumour suppressor gene for PDAC.8 Indeed, in addition to a variety of normal cellular processes in normal epithelial cells,10 SMAD4 is involved in the transforming growth factor-β (TGF-β) signalling pathway and translocates to the nucleus as a transcriptional cofactor to facilitate the transcription of TGF-β-responsive genes in cancer cells.11

Aerobic glycolysis is now considered the main metabolic pathway in PDAC, and PDAC has been termed a ‘glycolytic PDAC’.12 Phosphoglycerate kinase 1 (PGK1), the first ATP-generating enzyme in the glycolytic pathway, catalyses the generation of 3-phosphoglycerate (3 PG).13 PGK1 upregulation is pervasive in human PDAC.14 Recent studies have identified important dual roles for PGK1 as a glycolytic enzyme and protein kinase in the mutual regulation of cell metabolism and autophagy.15 16

PDAC is a heterogeneous tumour that exhibits metabolic plasticity.4 12 To date, few studies have examined the association between SMAD4 expression and the metabolic reprogramming of PDAC. Here, we demonstrate that PGK1 is a decisive oncogene for patients with SMAD4-negative PDAC and can be a target for the development of the therapeutic strategy for SMAD4-negative PDAC.

Materials and methods

Detailed methods are enclosed in the online supplementary materials and methods.

Patients and samples

The clinical tissue samples were obtained from patients diagnosed with PDAC who underwent operations at Fudan University Shanghai Cancer Center (FUSCC) from 2010 to 2012. Patients who were lost to follow-up or could not be evaluated were excluded. In total, 232 PDAC samples were performed strict pathological diagnoses and postoperative follow-ups at FUSCC. The pathological grading was performed by two independent pathologists at our centre. Clinical information regarding the samples is presented in online supplementary table S1. Part of each fresh sample was immediately placed in liquid nitrogen for next-generation sequencing and metabolite profiling, and another part of the sample was fixed in 4% paraformaldehyde to make paraffin-embedded tissue sections for histopathological examination.

Cell culture

The human pancreatic cancer cell lines SW1990, PANC-1, CFPAC-1, BxPC-3 and MiaPaca-2 were obtained from the American Type Culture Collection. The cell culture media were supplemented with 10% fetal bovine serum. BxPC-3 cells were cultured in RPMI 1640 medium, and CFPAC-1 cells were cultured in Iscove's Modified Dulbecco's Medium (IMDM). The other cells were cultured in Dulbecco's Modified Eagle Medium (DMEM). The cell lines were authenticated by DNA fingerprinting in 2016 and passaged in our laboratory fewer than 6 months after their receipt.

TCGA data set analysis

The Cancer Genome Atlas (TCGA) data are presented in online supplementary table S2. Detailed information is provided in the online supplementary materials and methods.

Gene expression analysis by quantitative real-time PCR

Quantitative real-time PCR was performed as described previously.17 Detailed information is provided in the online supplementary materials and methods. The primers are listed in online supplementary table S3.

In vivo animal model

BALB/c-nu mice (female, 4–6 weeks of age, 18–20 g; Shanghai SLAC Laboratory Animal) were housed in sterile filter-capped cages. The left and right flanks of the mice were transplanted subcutaneously with 3×106 cells. Tumour volume was measured twice per week using callipers and calculated using the following formula: (length × width2)/2. At day 40 post-tumour cell injection, the mice were prepared for positron emission tomography (PET)/CT scanning using an Inveon MicroPET/CT (Siemens Medical Solution). For the xenograft model with transformed human pancreatic ductal epithelial (HPDE) cells, we directly injected adeno-associated virus-mediated silencing PGK1 (AAV-shPGK1) or negative control virus particles (AAV-NC; 3×1011 plaque forming unit (pfu)/mL; 2 sites, 10 µL/site) into xenografts after 10 days of subcutaneous transplantation. At the same time, all mice were intraperitoneally injected with either 2-deoxy-D-glucose (2-DG; 500 mg/kg) or phosphate-buffered saline (PBS; 100 µL) daily for 14 days. For some studies, mice were anaesthetised by injecting 2% pentobarbital sodium (Sigma, St Louis, Missouri, USA) to perform the surgery. The tail of the pancreas was exposed for injection with 3×106 cells, and the incision was subsequently closed to establish the orthotopic implantation tumour model.

We successfully established PDAC patient-derived xenograft (PDX) mouse models by using 10 freshly isolated consecutive pancreatectomy samples as previously described.18 Briefly, every PDAC sample was isolated in two parts. One part was subjected to immunohistochemistry (IHC) staining for PGK1 and SMAD4 expression. The remaining sample was cut into five equal blocks of approximately 10 mm3 for subcutaneous transplantation into flanks of non-obese diabetic severe combined immune-deficient mice. Next, according to the IHC results, we defined the 10 samples as four groups: SMAD4-positive and PGK1-high group (n=10), SMAD4-positive and PGK1-low (n=10), SMAD4-negative and PGK1-high (n=25), and SMAD4-negative and PGK1-low (n=5). Eventually, 36 PDXs were successfully established. Mice with the same patient samples from two PGK1-high groups were randomly assigned to the PGK1 silencing group or the control group. When the PDX reached a mean volume of 100 mm3, AAV-shPGK1 or AAV-NC (3×1011 pfu/mL) was administered by direct injection into PDXs (2 sites, 10 µL/site). One month after the PDXs reached 100 mm3, the mice were prepared for MicroPET/CT scanning and histopathological examination.

Statistical analysis

The SPSS software program (version 22.0; IBM Corporation) was used for all statistical analyses of group comparisons of normally distributed data with the independent Student’s t-test or one-way analysis of variance (ANOVA). For multiple comparisons, the Tukey-Kramer honestly significant difference test was applied following ANOVA. The Fisher’s exact or χ22 tests were used to compare the categorical variables. Pearson’s correlation analysis was used to determine the correlation between two indicated molecular expression. Kaplan-Meier analysis and log-rank test were used to analyse OS and disease-free survival (DFS). Statistical differences were considered significant at p<0.05, p<0.01 and p<0.001.

Results

SMAD4 deficiency induces metabolic reprogramming in PDAC

Sequencing analysis for Kras mutation and IHC staining for CDKN2A, TP53 and SMAD4 were performed in the FUSCC cohort. Kras mutations were identified in 211 patients (90.9%). Loss of CDKN2A and TP53 mutation was identified in 182 (78.4%) and 160 (69.0%) patients, respectively. Loss of SMAD4 expression was also prevalent (64.2%; online supplementary table S1). Due to limitations in tissue sample acquisition, we were expected to perform IHC instead of genetic sequencing to examine SMAD4 status. Thus, we analysed genetic SMAD4 mutations from 60 patients in the FUSCC cohort and 177 patients from the TCGA cohort and compared the results with those of IHC analysis. As shown in online supplementary table S4, the SMAD4 expression levels in both FUSCC and TCGA cohorts were significantly correlated with SMAD4 genetic status.

18F-fluorodeoxyglucose (18F-FDG) PET/CT has been widely used for visualisation of the metabolic activity of viable tumour cells. The maximum standardised uptake value (SUVmax) has been used as a surrogate marker for the prognosis of PDAC,19 20 and the volumetric parameters of PET/CT, metabolic tumour volume (MTV) and total lesion glycolysis (TLG) comprehensively reflect both metabolic activity and tumour volume.19 As expected, patients with PDAC with negative SMAD4 expression exhibited high SUVmax, MTV and TLG levels, and similar results were obtained in patients with SMAD4 mutation (figure 1A–C; online supplementary figure S1A, S1B).

Figure 1

SMAD4 deficiency induces metabolic reprogramming in PDAC. (A) Representative 18F-FDG PET/CT imaging of patients with PDAC with positive or negative SMAD4 expression (scale bar, 200 µm; magnification scale bar, 40 µm). (B) Analysis of the SUVmax of patients with PDAC in the SMAD4Positive (n=45) and SMAD4Negative (n=56) groups. Data are expressed as mean±SD. (C) Analysis of the SUVmax of patients with PDAC in the wild-type SMAD4 (SMAD4WT; n=13) and mutant SMAD4 (SMAD4MUT; n=19) groups. (D) Western blot analysis of the indicated human pancreatic cancer cell lines. The HPDE cell line was included as a positive control for the detection of endogenous SMAD4 expression, and β-actin was used as a loading control. (E) Analysis of the viability of PDAC cell lines after treatment with the glycolytic inhibitor 2-DG for 48 hours. Cell numbers were determined as the indicated concentration of 2-DG using the Cell Counting Kit-8 reagent, and the ratios were compared with that at 0 µM. Data are expressed as mean±SD of three independent experiments and analysed using a one-way analysis of variance. *P<0.05, **p<0.01 vs PANC-1 cells. (F) Analysis of viability of PDAC cell lines cultured with medium with glucose deprivation for the indicated times. The ratios were compared with that at time point 0. *P<0.05, **p<0.01 vs PANC-1 cells. (G) SMAD4-induced metabolic reprogramming in BxPC-3 and CFPAC-1 cells treated with 5 ng/mL TGF-β for 24 hours, as reflected by the ECAR and OCR. (H) TGF-β/SMAD4-induced metabolic reprogramming in BxPC-3 and CFPAC-1 cells, as assessed by glucose uptake and lactate production. *P<0.05 vs control group without treatment. (I) Representative 18F-FDG MicroPET/CT imaging of tumour-bearing mice. The xenografts with BxPC-3 or BxPC-3-SMAD4 cells are indicated with white arrows. (J) The ratios of the xenograft SUVmax in the SMAD4 group (n=5) and the parent group (n=5). ** P<0.01. 2-DG, 2-deoxy-D-glucose; 18F-FDG, 18F-fluorodeoxyglucose; ECAR, extracellular acidification rate; HPDE, human pancreatic ductal epithelial cells; OCR, oxygen consumption rate; PDAC, pancreatic ductal adenocarcinoma; PET, positron emission tomography; SUVmax, maximum standardised uptake value; TGF-β, transforming growth factor-β.

Next, we examined SMAD4 expression levels in HPDE cells and pancreatic cancer cell lines with different mutational backgrounds. Both MiaPaca-2 and PANC-1 cells are two SMAD4 wild-type (WT) cells, exhibiting intact SMAD4 expression, with Kras, p53 and CDKN2A mutations, whereas CDKN2AWT CFPAC-1 and KrasWT BxPC-3 cells are two SMAD4-mutant cell lines with loss of SMAD4 expression (figure 1D and online supplementary table S5). Interestingly, we treated these cells with 2-DG, a glycolysis inhibitor, or with glucose deprivation, and found that SMAD4-negative cells were more dependent on glucose uptake to support their growth than were SMAD4-positive cells (figure 1E and F).

Furthermore, we established BxPC-3 and CFPAC-1 cells with ectopic SMAD4 expression (online supplementary figure S2A). Indeed, SMAD4 expression combined with TGF-β-mediated stimulation significantly decreased the extracellular acidification rate (ECAR) and increased the oxygen consumption rate (OCR; figure 1G). Similar results were also obtained by examining glucose uptake and lactate production (figure 1H). Moreover, we silenced SMAD4 expression in MiaPaca-2 and PANC-1 cells, which increased glycolytic capacity and decreased mitochondrial function (online supplementary figure S2B-S2D). Furthermore, we transfected Flag-KrasG12D to transform HPDE cells and subsequently silenced their SMAD4 expression (online supplementary figure S2E and S2F). Examinations of glucose uptake and lactate production showed that silencing SMAD4 could alleviate the TGF-β-induced repression of glycolysis in transformed HPDE-KrasG12D cells (online supplementary figure S2G). Next, we subcutaneously implanted SMAD4-overexpressing BxPCP-3 cells into nude mice and observed that ectopic SMAD4 significantly inhibited tumour growth and 18F-FDG uptake in vivo, as expected (figure 1I and J).

PGK1 is a glycolytic enzyme that is correlated with SMAD4 expression in PDAC

We selected eight SMAD4-positive and eight SMAD4-negative PDAC samples to perform liquid chromatography-mass spectrometry metabolomics analysis, investigating the mechanism through which SMAD4 deficiency induced aerobic glycolysis (online supplementary figure S3A and online supplementary table S6). The mutational background of these samples were also determined, and the differences in Kras, TP53 and CDKN2A mutations between SMAD4-positive and SMAD4-negative PDAC were not statistically significant (online supplementary table S7). Thus, SMAD4 was mainly responsible for the alterations in the metabolic profile, showing 42 statistically significant differential metabolites. Among these metabolites, 3 PG showed the most significant increase in the SMAD4-negative samples compared with that in the SMAD4-intact samples (figure 2A and B). Furthermore, we examined the expression of glycolytic genes by PCR array in BxPC-3 cells overexpressing SMAD4 (figure 2C and D). Ingenuity pathway analysis indicated that glucose metabolism was significantly upregulated in SMAD4-negative PDAC cells (figure 2E). Given the metabolomics data, PGK1, the glycolytic enzyme that catalyses the synthesis of 3 PG, might be a target of SMAD4 in PDAC (figure 2F; online supplementary figure S3B). Moreover, TGF-β/SMAD4 could repress the mRNA and protein levels of PGK1 in vitro (figure 2G, online supplementary figure S3C and S3D), and there was a negative relationship between PGK1 expression and SMAD4 expression in xenografts (figure 2H) and in the PDAC samples from the TCGA and FUSCC cohorts (figure 2I and J; online supplementary table S1 and online supplementary figure S6). Additionally, mutant SMAD4 was associated with a high level of PGK1 expression in the TCGA and FUSCC cohorts (figure 2K and L).

Figure 2

PGK1 is a glycolytic enzyme that is correlated with SMAD4 expression in PDAC. (A) Metabolites extracted from human SMAD4-positive (n=8) and SMAD4-negative (n=8) PDAC samples were analysed by LC-MS analysis. The volcano plot indicates the different metabolite levels. A two-tailed Student’s t-test was used to evaluate the data. The red points represent the metabolites with p<0.05 and FD >1.5. (B) A heat map showing differentially altered metabolites in SMAD4-positive patients compared with those in SMAD4-negative patients. (C) A human glucose metabolism PCR array was performed to identify the differentially expressed genes in the control cells compared with BxPC-3-SMAD4 cells. (D) The PCR array data are displayed as a volcano plot. The red and green points represent the genes with p<0.05 and FD >1.5. (E) Ingenuity pathway analysis indicates the activation of metabolic pathways. (F) The Venn diagram indicates that PGK1, as a glycolytic enzyme, correlates with SMAD4 in PDAC. (G) qRT-PCR and western blot analyses of PGK1 mRNA levels (top panels) and protein levels (bottom panels), respectively, after manipulating the expression of SMAD4 and treating with TGF-β. **P<0.01 vs the control group with no treatment. (H) IHC analysis of the correlation between PGK1 and SMAD4 expression in samples from mouse xenografts. (I) Data from the TCGA cohort show that PGK1 expression is negatively correlated with SMAD4 expression. (J) Representative IHC staining of SMAD4 and PGK1 in human PDAC samples in the FUSCC cohort (scale bar, 200 µm). (K) Level of PGK1 expression in patients with PDAC with SMAD4WT and SMAD4MUT from the TCGA cohort. (L) Level of PGK1 expression in patients with PDAC with SMAD4WT and SMAD4MUT from the FUSCC cohort. *P < 0.05, **P<0.01. FD, fold change; FUSCC, Fudan University Shanghai Cancer Center; IHC, immunohistochemistry; LC-MS, liquid chromatography-mass spectrometry; PDAC, pancreatic ductal adenocarcinoma; PGK1, phosphoglycerate kinase 1; qRT-PCR, quantitative real-time PCR; SMAD4MUT, mutant SMAD4; SMAD4WT, wild-type SMAD4; TGF-β, transforming growth factor-β.

PGK1 expression is deregulated in SMAD4-negative pancreatic cancer cells

We constructed a vector with SMAD4 containing a mutation (SMAD4ΔNLS) that disrupts SMAD4 translocation into the nucleus, and this vector was transfected into SMAD4-negative cells (figure 3A and B). ECAR and OCR analyses showed that SMAD4ΔNLS had little capacity to induce reprogramming of glycolytic metabolism (figure 3C). Interestingly, TGF-β/SMAD4WT, but not TGF-β/SMAD4ΔNLS, significantly inhibited PGK1 promoter activity and PGK1 expression (figure 3D–3F). Indeed, analysis of the PGK1 promoter sequence indicated that a conserved putative SMAD4 binding element (SBE) was found in the PGK1 promoter (figure 3G). SMAD4 silencing alleviated TGF-β-induced repression of the PGK1 promoter activity in HEK-293T and HPDE-KrasG12D cells (online supplementary figure S5A-S5D). The SBE was mutated to establish a mutant PGK1 promoter construct (MUT; figure 3H). After treatment with TGF-β, cotransfection with the SMAD4 vector repressed the luciferase activity of the WT PGK1 promoter construct but not that of the vector with mutant SBE (figure 3I). Furthermore, SMAD4 occupancy of the PGK1 promoter was greatly increased following treatment with TGF-β (figure 3J).

Figure 3

PGK1 expression is deregulated in SMAD4-negative pancreatic cancer cells. (A) Nuclear and cytoplasmic proteins were extracted from SMAD4-negative cells transfected with empty vector (EV), the vectors with wild-type SMAD4 (SMAD4WT) and a mutation (SMAD4ΔNLS) that disrupted SMAD4 translocation into the nucleus to perform western blot analysis with the indicated antibodies. β-actin was used as the loading control for the cytoplasmic extracts, and histone H3 was used as the loading control for the nuclear extracts. (B) Immunofluorescence analysis of the subcellular distribution of SMAD4. DAPI was used to detect nuclei (scale bar, 5 µm). (C) The effect of TGF-β/EV, TGF-β/SMAD4WT and TGF-β/SMAD4ΔNLS on the ECAR or OCR. (D) The effect of TGF-β/SMAD4 on PGK1 promoter activity was examined by dual luciferase reporter assay. **P<0.01 vs the control group with no treatment. (E) The level of PGK1 mRNA in cells with SMAD4ΔNLS and SMAD4WT after treatment with 5 ng/mL TGF-β. **P<0.01 vs the TGF-β/EV group. (F) The level of PGK1 protein in cells with SMAD4ΔNLS and SMAD4WT after treatment with 5 ng/mL TGF-β. (G) The position of the putative SMAD4 binding element (SBE) in the human PGK1 promoter. (H) Sequences of the wild-type (WT) and mutated (MUT) SBE in the luciferase reporter constructs. The asterisks indicate mutations in the SBE. (I) TGF-β/SMAD4 decreased the WT promoter activity of PGK1 but had no effect on MUT promoter activity. *P<0.05 vs the WT control group with no treatment. (J) TGF-β stimulation increased SMAD4 occupancy on the PGK1 promoter as reflected by the ChIP assay. **P<0.01. 2-DG, 2-deoxy-D-glucose; ECAR, extracellular acidification rate; MUT, mutant; OCR, oxygen consumption rate; PGK1, phosphoglycerate kinase 1; TGF-β, transforming growth factor-β; WT, wild-type.

Roles of SMAD4 loss in PDAC development in a PGK1-dependent manner

Data from the TCGA cohort indicated that high expression of PGK1 was a predictor of poor OS in patients with PDAC with SMAD4MUT but not in those with SMAD4WT (online supplementary figure S6A). Moreover, analysis of our cohort data showed that SMAD4-negative PDAC with a low level of PGK1 had lower SUVmax, MTV and TLG levels (online supplementary figure S6B-S6D). We constructed SMAD4-negative PDAC cells with silenced PGK1 expression (online supplementary file S6E, S6F). In vivo and in vitro assays indicated that PGK1 was an important oncogene in SMAD4-negative PDAC for metabolic reprogramming (online supplementary figure S6G-S6J). Additionally, the glycolytic capacity repressed by TGF-β/SMAD4 could be recovered by re-expressing PGK1 in BxPC-3 and CFPAC-1 cells (online supplementary figure S6J).

To further confirm the causal link between SMAD4 loss, PGK1 upregulation and metabolic reprogramming, xenografts with HPDE-KrasG12D cells were directly injected by AAV-shPGK1 or AAV-NC. At the same time, all mice were subjected to intraperitoneal injection of 2-DG (500 mg/kg) or PBS (figure 4A). IHC staining confirmed the decreased level of PGK1 expression by PGK1 silencing and the increased level of PGK1 expression by SMAD4 silencing in xenografts with HPDE-KrasG12D cells (figure 4B). Compared with the AAV-NC subgroup, the decreased xenograft volume and SUVmax values induced by PGK1 silencing were higher in HPDE-KrasG12D cells with sh-SMAD4 group than those in the scramble groups (figure 4C and D; online supplementary figure S6K). Moreover, the xenograft volumes with AAV-shPGK1 treated with 2-DG or without 2-DG were not significantly different (figure 4C and D). This result might be because PGK1 silencing significantly suppressed glycolysis, which concealed the effect of 2-DG on glycolysis, suggesting that PGK1 silencing inhibited xenograft growth mainly by glycolysis inhibition. PDX mouse models were also established, and in vivo infection with AAV-shPGK1 into PDXs with high PGK1 expression showed that tumour growth was significantly inhibited by PGK1 silencing, similar to those of low PGK1 expression (figure 4E and F). Notably, tumour volume and SUVmax in SMAD4-negative PDAC with PGK1 silencing were similar to those in SMAD4-positive PDAC with low PGK1 expression levels (figure 4F and G). Thus, the roles of SMAD4 loss in PDAC development and metabolic reprogramming were likely PGK1-dependent. This causal link was also confirmed by cell proliferation assays in vitro, as shown in online supplementary figure S6L.

Figure 4

There is a causal link between SMAD4 loss, PGK1 upregulation and metabolic reprogramming. (A) Representative image of xenografts formed by injection of HPDE-KrasG12D cells into nude mice. (B) IHC staining confirmed the level of PGK1 expression in xenografts formed by injection of HPDE-KrasG12D cells (scale bar, 50 µm; left panel). IHC staining of PGK1 from the mouse xenograft model with HPDE-KrasG12D cells was quantified using the Q score (right panel). (C) The xenograft growth curve formed by injection of HPDE-KrasG12D cells. Data are expressed as mean±SEM (n=5/group). (D) The xenograft SUVmax values in the xenografts formed by HPDE-KrasG12D cell injection. Data are expressed as mean±SD. *P<0.05, **p<0.01 vs group treated with AAV-NC/2-DG(−). (E) Representative images (scale bar, 5 mm) of SMAD4-negative (n=21) and SMAD4-positive (n=15) PDXs treated with AAV-shPGK1 (n=9 for SMAD4-negative PDXs; n=4 for SMAD4-positive PDXs) or AAV-NC (n=8 for SMAD4-negative PDXs; n=4 for SMAD4-positive PDXs). IHC staining confirmed the level of PGK1 expression in the tissues of the PDXs (scale bar, 50 µm). (F) The volumes of the PDXs were calculated with the following formula: (length × width2)/2. **P<0.01 vs the control group treated with AAV-NC. (G) The xenograft SUVmax ratios in the SMAD4-negative and SMAD4-positive PDXs. ns, not significant, *p<0.05, **p<0.01 vs the control group treated with AAV-NC. *P < 0.05; **P<0.01. 2-DG, 2-deoxy-D-glucose; AAV-NC, adeno-associated negative control virus; AAV-shPGK1, adeno-associated virus-mediated silencing PGK1; HPDE, human pancreatic ductal epithelial cells; IHC, immunohistochemistry; PDX, patient-derived xenograft; PGK1, phosphoglycerate kinase 1; SUVmax, maximum standardised uptake value.

PGK1 mediates epithelial-to-mesenchymal transition via repression of E-cadherin in SMAD4-negative cells

Epithelial-to-mesenchymal transition (EMT) is a process that appears to be associated with SMAD4 loss,21 22 and SMAD4 and E-cadherin were positively correlated in PDAC samples (online supplementary table S1 and online supplementary figure S4). Thus, we first determined the effect of PGK1 on SMAD4 regulated migration. PGK1 overexpression could recover the migration capacity, although the presence of SMAD4 significantly inhibited migration (figure 5A). Second, the expression of EMT markers was examined after PGK1 silencing, and the fold change in the expression of the epithelial marker, E-cadherin, was the most obvious (figure 5B). Western blot analysis also demonstrated the repressive effect of PGK1 on E-cadherin expression (figure 5C). Furthermore, analysis of the E-cadherin promoter showed that PGK1 might bind to putative sites in the core promoter region of E-cadherin (figure 5D). Indeed, PGK1 silencing decreased the nuclear fraction of PGK1 and its enrichment on the E-cadherin promoter (figure 5E). Moreover, dual luciferase assays indicated that PGK1 inhibited E-cadherin promoter activity in a dose-dependent manner (figure 5F and G). IHC analysis of mouse xenografts and human PDAC samples also indicated that PGK1 was negatively correlated with E-cadherin expression (figure 5H; online supplementary figure S1).

Figure 5

PGK1 mediates EMT via repression of E-cadherin in SMAD4-negative cells. (A) PGK1 increased the cell migration capacities of BxPC-3 and CFPAC-1 cells. (B) qRT-PCR analysis of EMT markers in BxPC-3 and CFPAC-1 cells with PGK1 silencing. *P<0.05, **p<0.01 vs the scramble group. (C) Western blot analysis of E-cadherin expression in BxPC-3 and CFPAC-1 cells with PGK1 silencing. (D) Analysis of the core promoter region of human E-cadherin. (E) PGK1 silencing decreased PGK1 occupancy on the E-cadherin promoter as reflected by ChIP assay. (F) Ectopic PGK1 expression decreased the promoter activity of E-cadherin in HEK-293T cells. **P<0.01 vs control cells with transfecting EV. (G) PGK1 regulated E-cadherin promoter activity in a dose-dependent manner. *P<0.05, **p<0.01 vs the control group without transfecting PGK1-expressing vector. (H) IHC staining for PGK1 and E-cadherin in mouse xenograft samples with PGK1 silencing. EMT, epithelial-to-mesenchymal transition; EV, empty vector; IHC, immunohistochemistry; PGK1, phosphoglycerate kinase 1; TS, transcription start-site; qRT-PCR, quantitative real-time PCR.

Subcellular localisation of PGK1 confers metabolic phenotypes in SMAD4-negative PDAC cells

We used CRISPR/Cas9 genome editing knock-in technology to replace endogenous PGK1 with PGK1MUT, PGK1ΔNLS and PGK1WT in SMAD4-negative PDAC cells (figure 6A). PGK1MUT was one mutation that exhibited loss of glycolytic enzymatic activity, as indicated by the inability to synthesise 3 PG (figure 6A and B). PGK1ΔNLS was another mutation that blocked PGK1 translocation into the nucleus (figure 6A and online supplementary figure S7). ECAR and OCR analyses indicated that four cell lines had distinct metabolic phenotypes, with different degrees of ECAR and OCR (figure 6C and D). Furthermore, we performed glucose metabolism flux analysis with uniformly labelled C13-glucose (figure 6E) to confirm the metabolic phenotypes, revealing a significant increase in glucose-derived tricarboxylic acid cycle metabolites (citrate) in cells with pWPI and PGK1MUT, as well as glycolytic metabolites (lactate) in cells with PGK1ΔNLS and PGK1WT (figure 6F and G).

Figure 6

Subcellular localisation of PGK1 confers metabolic phenotypes in SMAD4-negative PDAC cells. (A) Nuclear and cytoplasmic proteins were extracted from SMAD4-negative PDAC cells with PGK1MUT, PGK1ΔNLS and PGK1WT by CRISPR/Cas9 genome editing knock-in technology to perform western blot analysis with the indicated antibodies. The innocuous protein GFP was used as a control. (B) Determination of the 3 PG content in BxPC-3 and CFPAC-1 cells with different PGK1 statuses. **P<0.01 vs the control group (pWPI plasmid). (C) Determination of the effect of PGK1 status on metabolic reprogramming, reflected by ECAR and OCR. (D) Metabolic phenotypes were stratified by ECAR and OCR. (E) Glucose metabolism flux analysis as illustrated by carbon flux (grey) from uniformly labelled 13C-glucose. (F) Representative data depict the abundance of selected labelled metabolites following LC-MS analysis. (G) 13C-glucose labelled lactate (m+3) and citrate (m+2) by glucose metabolism profiles were calculated as a percentage of the total metabolite pool. *P<0.05, **p<0.01 vs the control group (pWPI plasmid). 2-DG, 2-deoxy-D-glucose; 3 PG, 3-phosphoglycerate; ECAR, extracellular acidification rate; LC-MS, liquid chromatography-mass spectrometry; OCR, oxygen consumption rate; PDAC, pancreatic ductal adenocarcinoma; PGK1, phosphoglycerate kinase 1.

Subcellular localisation of PGK1 determines malignant behaviours in SMAD4-negative PDAC cells

Next, we determined whether PGK1-induced metabolic reprogramming supported different aggressive behaviours. Loss of PGK1 enzymatic activity did not affect the migration capacity, whereas loss of nuclear PGK1 significantly repressed the migration capacity (online supplementary figure S8A). To further characterise the specific and independent role of nuclear PGK1 in PDAC metastasis, we determined whether targeting glycolysis by silencing fructose-bisphosphate aldolase (ALDOA), a glycolytic enzyme responsible for metastasis,23 would phenocopy the metastatic outcome observed for PGK1 alone (online supplementary figure S8B). The inhibitory effects of silencing ALDOA or PGK1 on glycolysis and migration capacities were alleviated with increasing ALDOA or PGK1 expression. This recovery of metastatic potential by PGK1 overexpression was most significant, whereas the recovery of glycolysis by PGK1 and ALDOA overexpression was not statistically significant (online supplementary figure S8C-S8E), suggesting that nuclear PGK1 alone was likely sufficient to drive metastasis by repressing E-cadherin expression (online supplementary figure S8F and S8G).

Moreover, PGK1MUT decreased the proliferative rate to the level of pWPI (online supplementary figure S9A-S9C). PGK1MUT inhibited cell cycle progression, arrested the cell cycle at S phase and significantly increased the population of apoptotic cells compared with the PGK1ΔNLS and PGK1WT populations (online supplementary figure S9D-S9G). Furthermore, we established an orthotopic implantation tumour mouse model by injection of BxPC-3 cells with pWPI, PGK1MUT, PGK1ΔNLS or PGK1WT into the pancreas (online supplementary figure S9H and S9I). PGK1MUT mice had smaller tumours and more metastatic foci in the liver, whereas the PGK1ΔNLS mice showed opposite changes (online supplementary figure S9J and S9K).

PGK1 defines subtypes of pancreatic cancer with poor prognosis

Kaplan-Meier analysis indicated that both cytoplasmic and nuclear PGK1 were predictive factors of poor prognosis in SMAD4-negative PDAC (figure 7A and B). Considering the effect of nuclear PGK1, we reanalysed the IHC results from SMAD4-negative PDAC. Based on the subcellular locations of PGK1, patients were classified into nuclear PGK1 positive (NP) and negative (NN) groups, and based on the cytoplasmic PGK1 intensity patients were classified into high-level cytoplasmic PGK1 (CH) and low-level cytoplasmic PGK1 (CL) groups (figure 7C). As shown in online supplementary table S8, Kras, CDKN2A and TP53 mutations were not enriched for any subgroup of SMAD4-negative PDAC, suggesting that the difference in prognosis is driven independently of Kras, CDKN2A and TP53 mutations. Moreover, the CH and NP subgroups had the highest SUVmax, MTV and TLG levels, whereas the CL and NN subgroups and the CL and NP subgroups exhibited low levels of metabolic burden (figure 7D and online supplementary figure S10).

Figure 7

PGK1 defines the pancreatic cancer subtypes with a poor prognosis. (A) Kaplan-Meier analysis of the overall survival rate of all patients with pancreatic cancer, according to cytoplasmic PGK1 expression (left panel) and of SMAD4-negative patients (right panel). (B) Kaplan-Meier analysis of the overall survival rate of all patients with pancreatic cancer, according to nuclear PGK1 expression (left panel) and of SMAD4-negative patients (right panel). (C) Representative IHC imaging of PGK1, E-cadherin and Ki-67 staining in samples from patients with PDAC. Based on the PGK1 localisation, patients were classified into nuclear PGK1-positive (NP) and PGK1-negative (NN) groups. Based on the intensity of cytoplasmic PGK1, patients were classified into high-level cytoplasmic PGK1 (CH) and low-level cytoplasmic PGK1 (CL) groups. (D) Analysis of the SUVmax of the CH and NP, CL and NN, CL and NP, and CH and NN subgroups. *P<0.05, **p<0.01 vs the subgroup of CH and NP. (E) Measurement of the tumour volumes (maximum diameter) of the CH and NP, CL and NN, CL and NP, and CH and NN subgroups. *P<0.05 vs the subgroup of CH and NP. (F) Determination of the percentage of patients with relapse and metastasis in the CH and NP, CL and NN, CL and NP, and CH and NN subgroups. (The numbers refer to the rates of relapse and metastasis in each group.) (G) Overall survival of patients with SMAD4-negative PDAC from the four groups. (H) Disease-free survival of patients with SMAD4-negative PDAC from the four groups. (I) Proposed model of the mechanism of PGK1-mediated regulation of aggressive behaviours in SMAD4-negative PDAC cells. IHC, immunohistochemistry; OXPHOS, oxidative phosphorylation; PDAC, pancreatic ductal adenocarcinoma; PGK1, phosphoglycerate kinase 1; SUVmax, maximum standardised uptake value.

The presence of nuclear PGK1 was always associated with low E-cadherin expression and high TGF-β1 expression (figure 7C and online supplementary table S8). The high level of cytoplasmic PGK1 was a determinant of the high percentage of Ki-67-positive cells, indicating that the carcinoma was highly proliferative (figure 7C and online supplementary table S8). Notably, the CH and NP subgroups always had the largest tumours with high metastatic potential, the CL and NN subgroups had small tumours with low metastatic potential, the CL and NP subgroups always had small tumours with high metastatic potential, and the CH and NN subgroups had large tumours with low metastatic potential (figure 7E and F; online supplementary table S8). Finally, we compared the OS and DFS times of the four groups and found that the CL and NN subgroups had the best prognosis, and that the CL and NP subgroups and the CH and NP subgroups had short OS and DFS times (figure 7G and H).

Discussion

In contrast to Kras and TP53, SMAD4, a major driver gene in PDAC, has seldom been reported to correlate with tumour metabolism. Reportedly, SMAD4 is involved in matricellular fibrosis.24 This microenvironmental stress induces the metabolic reprogramming of cancer cells to support their survival and growth.12 In addition to extrinsic factors, SMAD4 deficiency is responsible for the enhanced glycolytic capacity of cancer cells with upregulated glucose transporter expression.25 Previously, we also demonstrated that PDAC with mutant SMAD4 exhibits a high metabolic tumour burden.26 In this study, we assumed that SMAD4 loss might be the major event in inducing metabolic reprogramming in CDKN2AWT CFPAC-1 and KrasWT BxPC-3 cells and interpreted the mechanism by which SMAD4 inhibits PGK1 expression to repress glycolysis. Thus, SMAD4-positive PDAC cells might be metabolically quiescent, and their energy levels make these cells unable to engage in numerous biological responses, thereby causing them to present as low-grade malignancies with good prognoses.27

Although our study and another study reported that IHC analysis of SMAD4 reflected genetic status in PDAC,28 IHC results are categorised as a dichotomous output, namely positive or negative, and cannot distinguish subtle changes in SMAD4 expression, including heterozygous SMAD4 deletion. Thus, unsurprisingly, positive SMAD4 expression alone has not been widely used for predictive purposes.29 Nevertheless, we characterised the roles of PGK1 in SMAD4-negative PDAC to provide novel insights into pancreatic cancer progression.

Notably, PGK1 could play various roles in tumour progression, independent of its activity as a glycolytic enzyme.15 16 30 31 We also revealed the essential dual roles of PGK1 for tumourigenesis as a glycolytic enzyme and transcription factor in the integrated regulation of cell metabolism and biological behaviours. The previous observation that PGK1 regulates the β-catenin expression highlights its key role in regulating tumour metastasis.30 31 Here, a novel mechanism for PGK1 underlies the PDAC metastasis in addition to the induction of glycolysis to provide energy support. Nuclear PGK1 functions as a transcription factor to repress E-cadherin expression. E-cadherin induces the redistribution of cytoplasmic β-catenin to the cell membrane, which increases tumour cell–cell adhesions,32 suggesting that nuclear PGK1 is linked to Wnt signalling and the upregulation of cell migration, which synergistically contribute to a malignant phenotype with high metastatic potential.

Several studies have attempted to correlate genetic alterations with clinical features to stratify PDAC into subgroups; however, various typing approaches are still in the exploration stage.33–35 Iacobuzio-Donahue et al35 reported that locally destructive PDAC with no documented metastatic disease uncommonly showed loss of SMAD4 expression compared with that in PDAC with distant metastasis, in which the rates of SMAD4 loss approached 75%. However, SMAD4 expression alone has shown a lack of predictive benefit.29 Furthermore, Whittle et al36 proposed a potential solution using Runx3 expression in tumour cells for predicting the behaviour of SMAD4-positive tumours. For SMAD4-negative PDAC, we selected PGK1 as the decisive gene to stratify PDACs into four subtypes (figure 7). Cytoplasmic PGK1 determined the metabolic phenotype of PDAC cells. PDAC cells with the glycolytic phenotype always had a higher proliferative ability than those with the oxidative phosphorylation phenotype, likely because glycolysis can more efficiently produce macromolecules and energy for proliferation. Therefore, cytoplasmic PGK1 might be responsible for primary tumour growth, and nuclear PGK1 determined the metastatic potential of PDAC cells. The subcellular location of PGK1 balanced the proliferation and metastasis in PDAC, and the four subgroups exhibited different biological features. Among them, the CH and NP subgroups of SMAD4-negative PDAC with a high metabolic burden, malignant potential and TGF-β expression might be similar to the squamous tumours stratified by Bailey et al.7 Notably, there are some patients with primary carcinomas that are relatively confined to the pancreas but with a significant metastatic burden.37–39 Iacobuzio-Donahue et al considered this observation to correlate with the genetic status of SMAD4.35 Our classification scheme was used to further interpret this observation. The CL and NP subgroups could distinguish small primary carcinomas with high metastatic potential from SMAD4-negative PDACs. However, a subtype of SMAD4-negative PDAC (CL and NN subgroups) was similar to SMAD4-positive PDACs, which comprise locally destructive tumours with limited metastatic disease burden. Furthermore, the prognosis of this group was not superior to that of patients with SMAD4-positive PDACs (data not shown), likely because loss of SMAD4 induces deregulation of numerous genes that mediate malignancy, and some genes might be more critical than PGK1 to maintaining malignancy in these patients, leading to an OS time inferior to that of patients with SMAD4-positive PDACs.

For clinical applications, these findings need to be further confirmed in follow-up prospective studies and extended for verification in multiple retrospective data sets. Even so, the combination of SMAD4 and PGK1 can aid in the prediction of metastatic patterns and provide guidance towards the therapeutic strategy for PDAC.

References

View Abstract

Footnotes

  • CL, SS and YQ contributed equally.

  • Contributors CL, SS and X-JY conceived and designed the strategies. CL and YQ performed the in vitro experiments with cell lines, analysed the data and wrote the paper. XJ-Y supervised the project. BZ, SJ and QM collected tissue samples. JH and QH performed IHC-related assays. CL and QH performed the flow cytometry analysis, while JX carried out the TCGA data set analysis.

  • Funding This study was jointly funded by the National Science Foundation for Distinguished Young Scholars of China (no 81625016), the National Natural Science Foundation of China (no 81602085 and 81902428) and the Shanghai Sailing Program (no 17YF1402500 and 19YF1409400).

  • Competing interests None declared.

  • Patient consent for publication Not required.

  • Ethics approval This study was approved by the Institutional Research Ethics Committee of Fudan University Shanghai Cancer Center (FUSCC), and written informed consent was obtained from all patients prior to the investigation. All animal experiments were performed in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of Fudan University, and were approved by the Institutional Animal Care and Use Committee of Fudan University.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Data availability statement All data relevant to the study are included in the article or uploaded as supplementary information.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.