Glycometabolic reprogramming-mediated proangiogenic phenotype enhancement of cancer-associated fibroblasts in oral squamous cell carcinoma: role of PGC-1α/PFKFB3 axis

Abstract

Background

Angiogenesis is a key rate-limiting step in the process of tumour progression. Cancer-associated fibroblasts (CAFs), the most abundant component OSCC stroma, play important roles in pro-angiogenesis. Recently, the stroma “reverse Warburg effect” was proposed, and PFKFB3 has been brought to the forefront as a metabolic enzyme regulating glycometabolism. However, it remains unclear whether glycometabolism reprogramming is involved in promoting the angiogenesis of CAFs.

Methods

CAFs and paracancerous fibroblasts (PFs) were isolated from OSCC and adjacent tissues. We detected the pro-angiogenesis and glycometabolism phenotype of three pairs of fibroblasts. Targeted blockage of PFKFB3 or activation of PGC-1α signal was used to investigate the effect of glycolysis on regulating angiogenesis of CAFs in vitro and vivo.

Results

CAFs exhibited metabolic reprogramming and enhanced proangiogenic phenotype compared with PFs. Inhibition of PFKFB3-dependent glycolysis impaired proangiogenic factors (VEGF-A, PDGF-C and MMP9) expression in CAFs. Furthermore, CAFs proangiogenic phenotype was regulated by glycometabolism through the PGC-1α/PFKFB3 axis. Consistently, PGC-1α overexpression or PFKFB3 knockdown in CAFs slowed down tumour development by reducing tumour angiogenesis in the xenograft model.

Conclusion

CAFs of OSCC are characterised with glycometabolic reprogramming and enhanced proangiogenic phenotypes. Our findings suggest that activating PGC-1α signalling impairs proangiogenic phenotype of CAFs by blocking PFKFB3-driven glycolysis.

Bcakground

Tumour microenvironment (TME), a multicellular system with complex tumour-stromal interactions, is an indispensable “soil” in the process of tumour progression [1]. As one of the most abundant stromal components in the TME, fibroblasts can be activated to cancer-associated fibroblasts (CAFs) by tumour cells and have played a prominent role in cancer pathogenesis. Accumulating evidence has revealed that the cross-talk between CAFs and tumour cells is actively involved in tumorigenesis and tumour progression [2, 3]. CAFs remodel the extracellular matrix (ECM) and interact with cancer cells through paracrine signalling, allowing the tumour to invade and metastasize through the TME [2]. Hence, a new cancer treatment strategy, targeting the tumour stroma, is gradually emerging.

During the process of tumour growth, invasion and metastasis, angiogenesis is a key rate-limiting step [4]. CAFs promote angiogenesis by secreting cytokines, VEGF-A, PDGF-C and MMPs, to meet the requirement of malignant tumours [1]. Alternatively, CAFs can modulate angiogenesis indirectly by remodelling the interstitial fluid pressure, stiffness and elasticity of the tumour stroma [5]. Our previous research revealed that tumour cell-secreted exosomes can trigger the phenotype transition of fibroblasts into a more proangiogenic status, and thereby promote the melanoma angiogenesis [6]. However, the molecular insights, by which the proangiogenic ability of fibroblasts is regulated, need further study.

Cancer is a metabolic disease with reprogrammed metabolism, which is a hallmark of cancer cells [7]. Compared with normally differentiated cells, relying on the mitochondrial TCA cycle to provide energy for cellular activities, cancer cells tend to reprogramme glycometabolism away from oxidative phosphorylation (OXPHOS) towards glycolytic metabolism, even with sufficient oxygen supply termed as “Warburg effect” [8]. It is noteworthy that the newly concept of “reverse Warburg effect” emphasised the important role of stromal component in tumour metabolism [9]. “Corrupted” by cancer cells, stromal cells shift towards aerobic glycolysis and produce energy-rich metabolites to facilitate tumour growth. Our previous study proved that tumour microvesicles can induce glycometabolic reprogramming in normal fibroblasts by activating ERK1/2 signal pathway [10]. Coincidentally, endothelial cells (ECs) have a similar metabolism process, a higher glycolytic activity and relying on glycolysis rather than OXPHOS for energy production [11]. The metabolic reprogramming in ECs regulates vascular expansion and the formation of blood vessels [12]. Phosphofructokinase-2/fructose-2,6-biphosphatase3 (PFKFB3), the most abundantly expressed glycolytic enzyme in ECs, is also an oncogene-like regulatory element, which regulates part of the glycolytic flux by converting fructose-6-phosphate to fructose-2,6-bisphosphatase [13–15]. Several previous studies demonstrated that inhibiting glycolytic flux by either small molecule pharmaceutical PFK15 or gene knockdown could impair tumour cells or ECs proliferation and migration in vitro and vivo [16–18]. However, the relationship is still unclear between glycometabolism and the proangiogenic phenotype in CAFs.

Proliferator-activated receptor gamma coactivator 1 alpha (PGC-1α), regulating oxidative phosphorylation activity and mitochondrial biogenesis, is a unique member of the PPARγ transcriptional coactivators family [19]. Despite increasing evidence verified that PGC-1α has a prominent effect in tumour, there are still disagreements regarding the pro or anti-tumorigenic role of PGC-1α in different types of malignancies [20–22]. PGC-1α knockout mouse embryonic fibroblasts exhibited stronger characterisation of the metabolic adaptations and oncogenic potential [23]. In a recent study by Caporarello et al., PGC-1α knockdown in normal human lung fibroblasts reduced mitochondrial mass and function while enhancing matrix synthetic fibroblasts activation and soluble profibrotic secreting, which indicated that PGC-1α plays a key role in the regulation of metabolism and paracrine signalling [24]. So, a better understanding of the role of PGC-1α in CAFs of OSCC will be critical in evaluating whether this pathway is susceptible to therapeutic intervention.

OSCC is an aggressive disease characterised by the intense fibrotic stromal response and abundant blood vessels. To clarify the role and mechanism of glycometabolism in proangiogenic phenotype of CAFs, CAFs and PFs were isolated from OSCC and paracancerous tissues. Our results demonstrated that aerobic glycolysis and proangiogenic ability were enhanced in CAFs. Inhibition of PFKFB3-dependent glycolysis impaired the expression of proangiogenic factors (VEGF-A, PDGF-C, and MMP9) in CAFs. Furthermore, we confirmed that PGC-1α displayed a low level in CAFs of OSCC and upregulation of PGC-1α partially inhibited the proangiogenic phenotype of CAF through the PGC-1α/PFKFB3 axis. In vivo, we validated that PGC-1α overexpression or PFKFB3 knockdown of CAFs slowed down tumour growth by reducing tumour angiogenesis. In conclusion, we propose a novel rationale for targeting the glucose metabolism axis: PGC-1α/PFKFB3 suppressing the pro-angiogenesis of CAFs in patients with OSCC.

Methods

Primary fibroblasts culture

CAFs and PFs were isolated from patients with oral squamous cell carcinoma who did not receive radiotherapy and chemotherapy before surgery, respectively. Fresh OSCC tissue samples were approved by the Ethics Committee of School and Hospital of Stomatology at Wuhan University (2019LUNSHENZIA70), and written informed consents for using tissue samples were provided by all patients. Specimens were dissected into tumour tissue and adjacent normal tissue (1.5–2 cm from the tumour margin [25]) for the isolation of CAFs and PFs, respectively. The isolation procedure of two kinds of fibroblasts was consistent. Briefly, after three times washing with sterile phosphate-buffered saline containing 5% penicillin–streptomycin (Hyclone, UT, USA), tissue samples were minced into 1 ×1 × 1 mm pieces in tissue culture dishes, and cultured in DMEM/HG (Hyclone, UT, USA) medium supplemented with 2.5% Dispase2 (Yeasen, Shanghai, China), 0.5%DNase1 (Biofroxx, 1121, Germany) and 5% penicillin–streptomycin, for 3 h. Then, they were seeded in a T25 culture flask with DMEM containing 10% FBS (Gibco, 10099141, Australia) at 37 °C in 5% CO₂. After attached cells reached the attachment-inhibited growth (cell confluence>80%), they were passaged and used within five passages. Mycoplasma was routinely tested. Primary cells identification and mycoplasma testing were provided in Supplementary Method.1.1 and Supplementary Fig. 1.

Bioinformatics analysis

The RNA-seq transcriptome data of patients with HNSCC was downloaded from TCGA database. Cibersort was utilised to obtain the data of cells component. The pair-wise gene expression correlation analysis was from GEPIA (gepia.cancer-pku.cn).

Measurement of glycolysis

The lactate production, glucose uptake, oxygen consumption rate (OCR) and intracellular ATP were detected with Lactate Acid Test Kit (Nanjing Jiancheng Bioengineering Institute, Nanjing, China), Glucose Uptake Kit (Cayman, Ann Arbor, Michigan, USA), ATP Assay Kit (Beyotime, Jiangsu, China) and OCR Assay Kit (BestBio, Shanghai, China) in accordance with the manufacturer’s instructions.

Western blot analysis

The experiment was processed as previously described [10]. The antibodies were as follows: α-SMA (Abcam, ab7817, 1:1000); FAP (Abcam, ab53066, 1:1000); PFKFB3 (Abcam, ab181861, 1:2000); PGC-1α (Abcam, ab106814, 1:500); PPARγ (CST, C26H12, 1:1000); PDK1 (CST, C47H1, 1:1000); β-actin (CST, 4970T, 1:2000); LDHA (Proteintech, 19987-1-AP, 1:2000); PKM2 (Proteintech, 15822-1-AP, 1:1000); MCT4 (Proteintech, 22787-1-AP, 1:2000); GLUT1 (Proteintech, 21829-1-AP, 1:2000); MMP9 (Proteintech, 10375-2-AP, 1:1000); VEGF-A (Proteintech, 66828-1-Ig, 1:1000); PDGF-C (Proteintech, 55076-1-AP, 1:500); VCAM-1 (Abcam, ab134047, 1:2000); VLA-4 (CST, 8440T, 1:1000).

Quantitative PCR

Total RNA was extracted using TRlzol (Takara, Tokyo, Japan) and reverse-transcribed to cDNA (Takara). Quantitative PCR was performed with SYBR Green (Takara) using the Applied Biosystems QuantStudio 6. The primer sequences were provided in Supplementary Method.1.2.

Multiplexed immunohistochemistry and cell immunofluorescent staining

An Opal 4-Color Automation IHC Kit (Akoya, NEL800001KT, USA) was used for tissues. Surgical specimens from patients with OSCC or paracancerous tissues were extracted and fixed in 4% paraformaldehyde for 24 h, dehydrated by ethanol, embedded in paraffin. Then, the sections were dewaxed in xylene, rehydrated in graded ethanol. Afterwards, antigen retrieval was performed using AR6 buffer (pH 6.0) with microwave treatment. After being incubated with primary antibody and secondary-HRP, all slide's fluorescence signals were stained by tyramide signal amplification buffer (TSA buffer). The microwave antibody retrieval step, another primary antibody incubation, and fluorescence signal stain were repeated for the different primary antibody. Finally, DAPI was stained for nuclear. In our study, the antibodies bound to TSA with corresponding fluorescent wavelengths were α-SMA (520 nm), PFKFB3 (570 nm), CD31 (690 nm).

For cells, fibroblasts in different groups were fixed by 4% paraformaldehyde and permeabilized with 0.5% Triton X-100 solution. Blocked with BSA, cells were incubated with primary antibodies (1:200) at 4 °C overnight. After that, cells were incubated with fluorescence-conjugated secondary IgG (Thermo Fisher Scientific, Waltham, MA). Analyses were performed with a fluorescent microscope (Biozero BZ-8000, Keyence, Osaka, Japan).

Chromatin immunoprecipitation (ChIP) assay

ChIP assay was conducted according to the manufacturer’s instruction of the ChIP kit (Thermo Fisher Scientific, Rockford, USA). The protein-DNA complexes were cross-linked by 1% formaldehyde then quenched by glycine. Cells were collected and lysed in lysis buffer. Then chromatin-rich lysates were sonicated to shear DNA using a S220 focused-ultrasonicator (Covaris, Woburn, MA). The supernatant after centrifugation was diluted in IP dilution buffer for co-immunoprecipitation. After incubated with antibodies or IgG at 4 °C overnight, ChIP Grade Protein A/G Magnetic Beads were used to bind chromatin DNA. The cross-links were reversed after the immune complexes were eluted, and then the DNA was purified and subjected to qPCR assay. The primers for three regions of PFKFB3 promoter are listed in the Supplementary Method.1.4.

Luciferase reporter gene assay

The phRL-TK plasmid and the pGL3-Basic plasmid inserted with the different promoter sequence of PFKFB3 were co-transfected into CAFs using Lipofectamine3000 (MiaoLing, Wuhan, China). After 48 h, luciferase activity was detected by the Dual-Luciferase Reporter Assay Kit (Promega, Madison, USA).

Cell migration assay

The migratory ability of HUVECs was analysed using Transwell chambers (Corning, NY, USA) with a polycarbonate membrane (6.5-mm diameter, 8-μm pore size). In total, 5 × 10⁴ HUVECs were seeded in serum-free ECM in the upper chambers. CAFs or PFs (5 × 10⁴/well) were seeded in 500 μl of serum-free DMEM into the lower chambers. After 24 h, the upper chambers were removed, and the migration cells which traversed to the reverse membrane were fixed with 4% paraformaldehyde, stained with 0.1% crystal violet. Three random fields were selected to photograph. ImageJ and GraphPad software were used to count and quantify the migrated cells.

Cell proliferation assay

HUVECs in different groups (1 × 10⁴ cells/well) were seeded into 96-well plates and cultivated by the conditional medium (CM) of PFs or CAFs. In all, 10 μl/well CCK-8 reagent (Biosharp, Hefei, China) was added to each well at 0, 24, 48 and 72 h, and the plates were incubated at 37 °C for 1.5 h. The absorbance was measured at 450 nm using a microplate reader.

Tube-formation assay

The tube-formation assay was performed by direct and indirect methods. For direct co-culture, the HUVECs were pretreated with a red cell membrane fluorescent probe (Beyotime, Jiangsu, China). The untreated-CAFs were pretreated with a green cell membrane fluorescent probe (Beyotime, Jiangsu, China), and the CAFs transfected with GFP-labelled adenovirus were directly cocultured with the labelled HUVECs. The two kinds of cells were 1:1 seeded onto the Matrigel (BD Biosciences, San Jose, CA, USA) and incubated for 6 h. The indirect culture was processed as previously described [6]. Quantitative analysis of tube-like structures was measured by ImageJ.

Enzyme‑linked immunosorbent (ELISA) assay

The same density of fibroblasts (3 × 10⁵/well) in a six-well plate with FBS-free DMEM/HG medium for 24 h to normalise the conditioned medium. The concentrations of VEGF-A, PDGF-C, and MMP9 in the CM were detected with sandwich ELISA according to the instruction (4 A Biotech Co, Ltd, China).

Cell transfection and lentivirus infection

The siRNAs targeting PFKFB3 (GenePharma, Shanghai, China), PFKFB3 plasmid (MiaoLing, Wuhan, China) and PGC-1α lentivirus (GeneChem, Shanghai, China) were constructed. Lipofectamine3000 and P3000 (Invitrogen, Carlsbad, CA) were used according to the manufacturer’s protocols. Details were provided in Supplementary Method.1.3.

Tumour xenografting in nude mice

The in vivo study was in accordance with the Ethical Committee on Animal Experiments of the Animal Care Committee of Wuhan University (S07921020L). OSCC cell line CAL27 was purchased from China Center for Type Culture Collection (Shanghai, China). Short tandem repeat was performed routinely on the cell line to confirm its authenticity, and Mycoplasma was routinely tested. The Female BALB/C nude mice (18–20 g; 6–8 weeks of age) were randomly divided into A, B, C three groups (n = 5), the group A and C mice were inoculated subcutaneously siCTRL CAFs with CAL27. The mixture of siPFKFB3 CAFs and CAL27 were inoculated subcutaneously into the group B mice. After one week when subcutaneous tumours were visible, the group C mice received an intraperitoneal injection of PFK15 (20 mg/kg) four times for 1 week and the group A and B mice were injected with PBS as control. The size of the xenografts was measured every day, and the tumour volumes were calculated according to the formula (width² × length)/2 [26]. After 2 weeks, all mice were sacrificed, and tumour tissues were harvested for histologic assessment.

In the other xenograft experiment, the Female BALB/C nude mice (18–20 g; 6–8 weeks of age) were randomly divided to two groups (n = 5). CAFs infected by PGC-1α or vector lentivirus were premixed with CAL27 respectively, then inoculated into the mice subcutaneously. The measurement method of tumour volumes was consistent with the above. After three weeks, all mice were sacrificed, and tumour tissues were harvested for histologic assessment. No animals were excluded from the analysis.

Statistical analysis

All experiments results were repeated at least three times and expressed as mean ± standard error of the mean (SEM). No statistical method was used to choose a sample size. Statistical analysis was conducted with SPSS and GraphPad software. One-way ANOVA or a non-parametric test was used for statistical analysis. The variance within each group was similar. Results were considered statistically significant when the value of P < 0.05.

Results

Tumour stroma glycolysis phenotype is positively correlated with tumour microvessel density

HNSCC of TCGA database was used for cells component analysis. The results showed that the richness of stroma was positively correlated with fibroblasts and endothelial cells. There was also a positive correlation between fibroblasts and endothelial cells (Fig. 1a). Next, we analysed the correlations between the levels of genes related to glucose metabolism and CAFs markers via GEPIA. The results showed a positive correlation between CAFs markers and genes related to glycolytic metabolism. Conversely, genes related to TCA cycle or OXPHOS complexes negatively correlated with the levels of CAFs in the tumour (Fig. 1b and Supplementary Table 1). Thus, CAFs in the OSCC stroma are positively correlated with ECs component and glycolytic phenotype of OSCC.

Fig. 1. Tumour stroma glycolysis phenotype is positively correlated with tumour microvessel density.

a The association between stroma, fibroblasts and ECs was analysed from the TCGA database. b The pair-wise gene expression correlation analysis was evaluated using the GEPIA database. c Representative multiplexed immunohistochemistry staining of α-SMA, PFKFB3 and CD31 in paracancerous tissues (PT) and cancer tissues (CT). Scale bars = 50 μm. And tissue segmentation map of epithelium or carcinoma, stroma, and vessels. Scale bars = 50 μm. N = 42. d Differential expression of α-SMA, PFKFB3 and CD31 in the margin of the tumour (TM) and the centre of the tumour (TC) in the OSCC tissue. Scale bars = 500 μm. e The expression of PFKFB3 in the stroma of the PT and CT. f The CD31⁺ MVD was quantified and analysed in PT and CT. g Spearman rank correlation analysis of PFKFB3 expression in stromal cells with CD31⁺ MVD. Data are presented as means ± SD. Results are representative of at least three independent experiments. *P <0.05, ns not significant.

To further investigate the correlation between glycolysis phenotype in tumour stroma and tumour microvessel density, PFKFB3, the representative glycolytic enzyme [27], CD31 and α-SMA were stained in OSCC and paracancerous tissues. As the multiplexed immunohistochemistry and tissue segmentation map showed, the CAFs of OSCC displayed remarkable higher expression of PFKFB3 (Fig. 1c, e) and the CD31⁺ MVD was more abundant in OSCC (Fig. 1c, f). The spearman rank correlation analysis indicated that the elevated expression of PFKFB3 in stroma positively correlated with CD31⁺ MVD (Fig. 1g). In addition, from a holistic perspective, we observed an explicit differential expression of α-SMA, PFKFB3 and CD31 in the margin and centre areas of the OSCC tissues (Fig. 1d). Considered together, the glycolysis level of the stroma is positively correlated with tumour microvessel density.

CAFs of OSCC promote proliferation, migration, and tube formation of HUVECs

We sought to understand the impact of primary CAFs of OSCC on pro-angiogenesis. Accordingly, we compared the ability of promoting angiogenesis of PFs and CAFs in vitro. CCK-8 and transwell assay results indicated that CAFs notably promoted HUVECs proliferation and migration, compared to PFs (Fig. 2a–c). An indirect co-culture system was constructed to evaluate the differences in HUVECs activated by PFs or CAFs (Fig. 2d). Western blot showed that CAFs significantly activated VEGFR2 expression in HUVECs compared to PFs (Fig. 2e, f). Tube-formation assay indicated that the number of the capillary-like structure was significantly increased when indirect incubated with CAFs (Fig. 2g, h). ELISA results showed that CAFs-CM concentrations of MMP9, VEGF-A and PDGF-C were remarkably higher than those of PFs (Fig. 2i). These data suggested that the ability of CAFs promoting angiogenesis is stronger than that of PFs.

Fig. 2. CAFs of OSCC promote proliferation, migration and tube formation of HUVECs.

PF1/CAF1, PF2/CAF2 and PF3/CAF3 were from patients 1, 2 and 3, respectively. a CCK-8 assay was performed to determine the proliferation ability of HUVECs after cultured with PFs-CM and CAFs-CM. b Transwell assay was performed to evaluate the migration ability of the HUVECs indirect cocultured with PFs or CAFs. Scale bars = 100 μm. c The migration cells were quantified and analysed. d Graphic representation of the indirect co-culture model system used in the study. e Western blot analysis of VEGFR2 expression in HUVECs activated by PFs or CAFs from three different patients. f Quantitative analysis for immunoblotting of (e). The relative protein expression was normalised by each pair of PFs. g Tube formation of HUVECs was assessed after indirect cocultured with PFs or CAFs using a Matrigel assay. Scale bars = 200 μm. h Quantification and analysis of the number of meshes formed. i Soluble MMP9, VEGF-A and PDGF-C in the CM of three pairs of primary cells were examined using ELISA. Data are presented as means ± SD. Results are representative of at least three independent experiments. *P <0.05, ns not significant.

CAFs of OSCC undergo glycometabolism reprogramming

Consistent with our previous study [10], primary CAFs presented higher glucose uptake and lactate generation (Fig. 3a, b), which indicated that the sum of glycometabolism flux was increased. However, the glycometabolism tendency is not clear. Further research focused on examining OCR and intracellular ATP. These results showed OCR, intracellular ATP production and PGC-1α expression (regulating mitochondrial biogenesis) [19] were significantly lower in CAFs compared with PFs (Fig. 3c–e), which implied that mitochondrial dysfunction may lead to low activity of OXPHOS in CAFs. Western blot indicated that the expression of PFKFB3, PKM2, PDK1, key glycolytic enzymes [8, 9, 12], especially PFKFB3, was upregulated in the CAFs (Fig. 3e, f). Conversely, the level of AMPK activation and the expression of SDH, especially SDHA, were reduced in the CAFs than PFs (Fig. 3e, f), indicating that the TCA cycle of CAFs was notably attenuated. The above data revealed that the glycometabolism of CAFs in OSCC has been reprogrammed, and it is manifested as an enhancement in aerobic glycolysis with a decrease in OXPHOS.

Fig. 3. CAFs of OSCC undergo glycometabolism reprogramming.

PF1/CAF1, PF2/CAF2 and PF3/CAF3 were from patients 1, 2 and 3, respectively. a Glucose uptake assay was carried on PFs and CAFs from three different patients. Quantification and analysis of relative glucose uptake was detected by the fluorescence intensity of 2-NDGB. b Lactate production assay was conducted on PFs and CAFs from three different patients with a lactate production assay kit. c Intracellular ATP was measured on PFs and CAFs from three different patients of using luminescent assay. d Oxygen consumption rate (OCR) was measured on PFs and CAFs from three different patients using BBoxiProbe. e Western blot analysis of PFs and CAFs in three different patients. f Quantitative analysis for immunoblotting of (e). The relative protein expression was normalised by each pair of PFs. Data are presented as means ± SD. Results are representative of at least three independent experiments. *P <0.05, ns not significant.

Inhibition of PFKFB3-dependent glycolysis can impair proangiogenic phenotype of CAFs

To explore the potential relationship between aerobic glycolysis and proangiogenic phenotype of CAFs, we tried to suppress the glycolytic flux. Given the critical role of PFKFB3 in glycolysis, PFK15 was used to block PFKFB3 in CAFs [13, 27, 28]. The results indicated that the glucose uptake, lactate generation and intracellular ATP were reduced respectively in a dose-dependent manner in response to PFK15 (Fig. 4a–c). However, the inhibition of PFKFB3 had no significant effect on OCR (Fig. 4d). Western blot showed that the expression of MCT4 and GLUT1 was decreased (Fig. 4ei, f). A similar tendency was observed in the expression of PKM2 and LDHA (Fig. 4ei, f). These results indicated that PFK15 dramatically inhibits the glycolytic activity of CAFs.

Fig. 4. Inhibition of PFKFB3-dependent glycolysis can impair proangiogenic phenotype of CAFs.

a–n CAFs were exposed to gradient concentration of PFK15. a Glucose uptake was analysed in CAFs after PFK15 treatment. b Lactate production of CAFs treated with PFK15 was detected. c Intracellular ATP generation of CAFs treated with PFK15 was measured. d OCR of CAFs treated with PFK15 was measured. e (i) Western blot analysis of PGC-1α, glycolytic rate-limiting enzymes (PKM2, LDHA), MCT4 and GLUT1 of CAFs treated with PFK15. (ii) Western blot analysis of proangiogenic factors (MMP9, VEGF-A and PDGF-C) of CAFs after PFK15 treatment. f Quantitative analysis for immunoblotting of (e). The relative protein expression was normalised by the expression of the interior reference β-actin. g Transwell assay was taken to evaluate the migration ability of the HUVECs indirect cocultured with PFK15- treated CAFs. Scale bars = 100 μm. h The migration cells were quantified and analysed. i The proliferation ability of HUVECs was detected after being cultured with the supernatant of PFK15- treated CAFs. j Soluble MMP9, VEGF-A and PDGF-C in the CM of CAFs treated with PFK15 were detected using ELISA. k Tube formation of HUVECs was evaluated after direct cocultured with PFK15-treated CAFs. Scale bars = 200 μm. l Quantification and analysis of the number of meshes formed. m Western blot analysis of VLA-4 in CAFs and VCAM-1 in HUVECs. n Quantitative analysis for immunoblotting of (m). The relative protein expression was normalised by the expression of the interior reference β-actin. Data are presented as means ± SD. Results are representative of at least three independent experiments. *P<0.05, ns not significant.

Next, we investigated whether the proangiogenic ability of CAFs can be changed in the case of glycolytic flux reduction. CCK-8 and transwell assay showed that the ability of CAFs treated with PFK15 promoting the migration and proliferation of HUVECs was weakened (Fig. 4g–i). Interestingly, fluorescence co-localisation showed that CAFs were firstly found to stretching and surrounding the HUVECs, which suggested that CAFs may establish a supporting niche to sustain the capillary-like structure (Fig. 4k). Similarly, this kind of structure was disintegrated and CAFs-HUVECs adhesion mediated by VCAM-1/VLA-4 axis was weakened when CAFs treated with PFK15 (Fig. 4k–n). Consistently, Western blot and ELISA showed that the expressions of PDGF-C, VEGF-A and MMP9 were visibly decreased (Fig. 4eii, f, j). To conclude, the inhibition of glycolysis activity by suppressing PFKFB3 can impair the proangiogenic phenotype of CAFs.

PGC-1α/PFKFB3 axis regulates the proangiogenic phenotype in CAFs

Given the role of PGC-1α in mitochondrial biogenesis and activity [29], we explored whether PGC-1α modulates the glycolysis of CAFs. CAFs were transfected with PGC-1α lentivirus or PFKFB3 siRNAs. Western blot and qPCR showed that PGC-1α overexpression significantly suppressed the level of PFKFB3 (Fig. 5a–c). On the contrary, PFKFB3 knockdown or inhibition did not have clear impact on PGC-1α in CAFs (Figs. 5g, h and 4e, f), revealing that PFKFB3 was a potential downstream target gene of PGC-1α signalling. PPARγ is a primary nuclear reporter of PGC-1α-coactivation gene [19], Western blot and qPCR showed the expression of PPARγ is activated by PGC-1α (Fig. 5a–c). Immunofluorescence showed that, compared with control group, the expression of PPARγ in the nuclear was significantly enhanced while the cytoplasmic expression of PFKFB3 was dramatically weakened in PGC-1α overexpression group (Fig. 5e, f). Furthermore, to determine whether PGC-1α/PPARγ regulates transcription of PFKFB3 directly, we predicted the possible binding sites of PFKFB3 promoter (Fig. 5d) and carried out the ChIP assay. The qPCR results showed that positive amplification products were detected at the P2 site (Fig. 5i). The mutation of P2 site in the PFKFB3 promoter significantly rescued PGC-1α/PPARγ-induced PFKFB3 promoter-luciferase activity (Fig. 5j). Taken together, the results indicated that PGC-1α/PPARγ complexes are bound to the promoter of the PFKFB3 gene and suppress its transcription in CAFs.

Fig. 5. PGC-1α/PFKFB3 axis regulates the proangiogenic phenotype in CAFs.

a Western blot analysis of protein expression in PFs, CAFs and CAFs infected with PGC-1α lentivirus. b Quantitative analysis for immunoblotting of (a). The relative protein expression was normalised by the expression of the interior reference β-actin. c Quantitative PCR was used to measure mRNA expression of PGC-1α, PPARγ and PFKFB3 in CAFs infected with PGC-1α lentivirus. d Schematic representation of the predicting sites on PFKFB3 promoter bound with complexes. e The expression of PGC-1α, PPARγ and PFKFB3 was detected by immunofluorescence in CAFs after PGC-1α lentivirus treatment. Scale bars = 50 μm. f Quantitative analysis for immunofluorescence of (e). The relative fluorescence intensity of nuclear or cytoplasm was normalised to CTRL. g The protein expression of PGC-1α and PFKFB3 in CAFs transduced with CTRL or PFKFB3 siRNAs. h Quantitative analysis for immunoblotting of (g). The relative protein expression was normalised by the expression of the interior reference β-actin. i ChIP-qPCR analysis was conducted to detect the enrichment of PFKFB3 at the different predicting binding sites using PGC-1α antibody or PPARγ antibody or IgG antibody in CAFs. j PFKFB3 luciferase reporter assay was performed after overexpressing PGC-1α and PPARγ in CAFs. k Western blot analysis of PGC-1α and PFKFB3 protein expression in CAFs, which treated with PGC-1α lentivirus and vector or PFKFB3 plasmid. l Quantitative analysis for immunoblotting of (k). The relative protein expression was normalised by the expression of the interior reference β-actin. m–v CAFs were treated with PGC-1α lentivirus and vector or PFKFB3 plasmid. m Glucose uptake, n lactate production and o intracellular ATP generation in CAFs treated with PGC-1α lentivirus or PFKFB3 plasmid in combination with PGC-1α lentivirus. p OCR of CAFs after indicated treatment was measured. q Transwell assay was performed to evaluate the migration ability of the HUVECs indirect cocultured with CAFs after indicated treatment. Scale bars = 100 μm. r The migration cells were quantified and analysed. s The proliferation ability of HUVECs was detected after cultured with the supernatant of treated CAFs. t Tube formation of HUVECs was assessed after direct cocultured with treated CAFs. Scale bars = 100 μm. u Quantification and analysis of the number of meshes formed. v Soluble MMP9, VEGF-A and PDGF-C in the CM of CAFs after indicated treatment were detected using ELISA. Data are presented as means ± SD. Results are representative of at least three independent experiments. *P <0.05, ns not significant.

Next, we examined whether PGC-1α regulates proangiogenic phenotype of CAFs. PFKFB3 overexpression plasmid was co-transfected into PGC-1α overexpression CAFs (Fig. 5k, l). The results exhibited that the uptake of glucose and generation of lactate were reduced (Fig. 5m, n), On the contrary, OCR and intracellular ATP were dramatically increased after CAFs transfected with PGC-1α lentivirus (Fig. 5o, p), which indicated that the aerobic respiration increases, and glycolysis decreases in CAFs. As expected, the inhibition of PGC-1α overexpression on glycolysis was partially reversed by the overexpression of PFKFB3 (Fig. 5m–p). We further found that the PGC-1α overexpression glycometabolism mode, in which glycolysis was inhibited, significantly impaired the proangiogenic phenotype of CAFs. Transwell assay, CCK-8, tube-formation assay and ELISA showed that the proangiogenic ability of CAFs was significantly impaired by PGC-1α overexpression (Fig. 5q–v). Consistently, proangiogenic phenotype was reversed partly by the ectopic expression of PFKFB3 (Fig. 5q–v). Collectively, PGC-1α/PFKFB3 axis is involved in the regulation of glycometabolism pattern, and changes in this pattern may affect proangiogenic phenotype in CAFs.

PGC-1α overexpression or PFKFB3 inhibition suppressed proangiogenic phenotype of CAFs in vivo

To examine the effect of PGC-1α and PFKFB3 on proangiogenic phenotype of CAFs, xenograft models were established (Fig. 6ai, gi) [10, 17]. The tumour growth curve and tumour volumes showed that siPFKFB3 CAFs significantly slowed the development of tumour and PFK15 almost blocked the tumour development at 20 mg/kg concentration (Fig. 6aii, b, c). The immunohistochemistry results showed that siPFKFB3 CAFs or PFK15 administration notably decreased MVD of xenograft (Fig. 6d, f). However, PFK15 administration did not significantly reduce the expression of PFKFB3 in the stroma (Fig. 6d, e).

Fig. 6. PGC-1α overexpression or PFKFB3 inhibition suppressed proangiogenic phenotype of CAFs in vivo.

a–f The nude mice were injected with CAL27 + siCTRL CAFs, CAL27 + siPFKFB3 CAFs or CAL27 + siCTRL CAFs+PFK15, respectively. a (i) Schematic representation of the co-inoculation model in which CAFs were transfected with a siRNA targeting PFKFB3, or the control vector (siCTRL), and then subcutaneously co-inoculated with CAL27 in BALB/C nude mice. (ii) Photographs of the nude mice and xenograft tumours dissected from the nude mice. b Volumes of the xenograft tumours on the nude mice were measured every day after indicated treatments. c Weights of xenograft tumours excised from the nude mice were measured after indicated treatments. d The expression of α-SMA, PFKFB3 and CD31 was detected by immunohistochemistry. e Quantification expression level of PFKFB3 in the stroma. f The density of CD31⁺ microvessel was quantified and analysed. g–l The nude mice were injected with CAL27 + CTRL CAFs or CAL27 + PGC-1α CAFs, respectively. g (i) Schematic representation of the co-inoculation model in which CAFs were treated with PGC-1α lentivirus, or the control vector (CTRL), and then subcutaneously co-inoculated with CAL27 in BALB/C nude mice. (ii) Photographs of the nude mice and xenograft tumours harvested from the nude mice after indicated treatment. h Volumes of the xenograft tumours on the nude mice were measured after indicated treatments. i Weights of xenograft tumours excised from the nude mice were measured after indicated treatments. j The expression of α-SMA, PGC-1α, PFKFB3 and CD31 was detected by immunohistochemistry. k Quantification expression level of PGC-1α and PFKFB3 in the stroma. l The CD31⁺ MVD was quantified and analysed. m Schematic representation of the glycometabolic reprogramming-mediated proangiogenic phenotype in CAFs. Data are presented as means ± SD. Results are representative of at least three independent experiments. *P <0.05, ns not significant.

Furthermore, we overexpressed PGC-1α in CAFs and established xenograft models via combined injection of CAFs and CAL27 subcutaneously. The tumour growth curve and tumour volumes indicated that PGC-1α overexpression slowed down tumour growth (Fig. 6gii, h, i). The immunohistochemistry results were observed the same inhibitory effect in MVD of tumours after PGC-1α/PFKFB3 axis activated (Fig. 6j–l). To sum, PGC-1α overexpression or PFKFB3 knockdown could slow down tumour development by reducing tumour angiogenesis. Together, both in vitro and in vivo data implied that PGC-1α/PFKFB3 axis-mediated glycometabolic reprogramming promotes the proangiogenic phenotype of CAFs (Fig. 6m).

Discussion

Although the prognosis for cancer patients has been gradually improved by the treatment model of tumour cells targeting, the intractable TME is still a determinant of cancer cell behaviour and disease progression [1, 30]. The imbalance of angiogenesis and metabolism plays important roles in tumorigenesis and tumour progression [7, 31]. CAFs, a dominant component of the TME, have received renewed attention in recent years for their function on tumour initiation, progression, metastasis, and resistance to therapies. However, the contribution of CAFs to angiogenesis and metabolism mechanism has not been fully elucidated. Here, we further confirm the changes in the pro-angiogenesis and glucometabolic levels of CAFs and clarify the internal correlation in vitro and in vivo.

CAFs have been reported to exacerbate or stimulate the angiogenic programming of neoplastic tissues through paracrine or ECM remoulding [5, 32, 33]. Our previous studies showed that fibroblasts activated in vitro exhibited proangiogenic phenotype transformation [6, 34]. In this work, we further identified the difference in proangiogenic phenotype between the primary CAFs and PFs of OSCC. Our results indicated that primary CAFs promoted the migration, proliferation and tube formation of HUVECs in vitro and exhibited upregulated secretion of VEGF-A, PDGF-C, and MMP9, which are potent proangiogenic factors intrinsically linked. Clinical studies have confirmed VEGF-A and its receptors are key targets of anti-angiogenic agents [35]. PDGF regulates angiogenesis indirectly by inducing VEGF transcription and secretion [36]. MMP9 increases the bioavailability of VEGF by degradation of extracellular components, such as collagen type IV [37]. In brief, our results illustrated that primary CAFs of OSCC play an important role in tumour angiogenesis.

Next, we further tried to explore the mechanism underlying the proangiogenic phenotype of CAFs. Metabolic reprogramming is a hallmark of cancer [7]. Cancer cells tend to reprogramme glycometabolism away from OXPHOS towards glycolysis to meet the energy requirements of tumour cells even with sufficient oxygen supply [8]. The new proposal of “reverse Warburg effect” shifted our attention to the stroma glycometabolism [9]. Similarly, endothelial cells also exhibit metabolic reprogramming, which not only regulates the formation of blood vessels, but also fuels vascular expansion [12]. Thus, whether the proangiogenic switch of CAFs is triggered by metabolic reprogramming deserves to explore. Our results indicated that CAFs of OSCC undergo metabolic reprogramming, which exhibit high glycolytic flux and low OXPHOS activity caused by mitochondrial dysfunction. Although glycolysis was activated, the total ATP production of CAFs was reduced, which was partly due to the dysfunction of mitochondrial aerobic respiration. Proliferating cells do not require excessive ATP for cellular activity [38]. The bulk of the glucose cannot be committed to carbon catabolism for ATP production, and a shift toward to increase glycolytic intermediates, which can funnel into anabolic side pathways to support de novo synthesis of nucleotides, lipids and amino acids needed to support cell proliferation [8]. Moreover, the lower energy supply efficiency of glycolysis does not compensate for the higher energy supply efficiency of glucose aerobic respiration in mitochondria through OXPHOS [12]. Zhang et al. revealed a decrease in activation of AMPK in normal oral fibroblasts after cocultured with OSCCs, which resulted in a decrease of ATP production [39]. Similarly, we detected that CAFs also exhibited a state of AMPK hypoactivation, which further supported the conclusion. Recent research demonstrated the metabolism of CAFs was reprogrammed by OSCC inducing unidirectional mitochondrial exchange in normal fibroblasts via mPTP opening [39]. Interestingly, the compressive stress of tumour growth could also activate CAFs glycolysis through a mechanotransduction pathway [40]. Our previous study clarified TMVs can also mediate the glucometabolic reprogramming via ERK1/2 [10]. Therefore, the upstream inducing the glucometabolic reprogramming of CAFs needs further research.

Given the important role of PFKFB3 in pathological angiogenesis, blockage of PFKFB3 was considered as a promising target for tumour therapy [13, 31]. In our work, PFKFB3 was found to be one of the most significantly increased isozymes in CAFs. This result is consistent with previous studies which clarified PFKFB3 bears an oncogene-like regulator element [14, 15]. Accumulating research verified that endothelial cells knocked down PFKFB3 could impair ECs proliferation, migration and vessel formation even induce tumour vessel normalisation [13, 31]. Here, targeting PFKFB3 by selective inhibitor PFK15, we suppressed glycolytic flux of CAFs. The results showed the glycolytic activity of CAFs was inhibited in a dose-dependent manner in response to PFK15. Furthermore, the proangiogenic phenotype of CAFs was weakened in vitro and vivo. However, PFK15 was observed to have no effect on the expression of PFKFB3 (Supplementary Fig. 2) [17]. We consider that this was due to PFK15 is a competitive inhibitor and cannot impact the expression level of CAFs. Interestingly, we also observed PFK15 had significant differences in the glycolysis inhibitory effect of CAFs from different patients (Supplementary Fig. 2). Recent research defined a CAFs subset, CD10⁺GPR77⁺, was correlated with chemoresistance and poor survival in lung and breast cancer patients [41]. Whether the differential sensitivity of CAFs from different sources to PFK15 is caused by the metabolic heterogeneity of CAFs subset deserves further study. Under direct co-cultivation conditions, we originally found CAFs were stretched and surrounded the HUVECs, implying that CAFs may establish a supporting niche to sustain the capillary-like structure. This special feature could also be found in OSCC tissues fluorescence. However, this structure disintegrated in the presence of PFK15. Previous research had verified that VCAM-1/VLA-4 axis induces cancer-endothelial cell fusion and cancer-CAFs adhesion [34, 42]. We found that blockage of aerobic glycolysis of CAFs could also inhibit CAFs-HUVECs adhesion mediated by the VCAM-1/VLA-4 axis, which may explain this structure disintegrated phenomenon. However, the downstream signalling pathway through which PFKFB3 regulates proangiogenic phenotypes has not been clarified in this study. Given the NF-κB signalling pathway plays an important role in angiogenesis [43], we speculate that PFKFB3 regulates the expression of downstream proangiogenic factors via the NF-κB pathway. This hypothesis is supported by recent studies, showing PFK15 treatment led to a reduction in IKBα phosphorylation and p65 nuclear accumulation [31, 44].

Warburg hypothesised tumours adopt unconventional glycometabolism was due to a defect in mitochondria. Similar phenomena can be observed in CAFs. The normal fibroblasts treated with TMVs exhibited an apparent swelling of the mitochondria [10]. Moreover, CAFs can export impaired mitochondria towards OSCCs through direct contact or indirect mechanism [39]. Prostate cancer increased OXPHOS and executed EMT by hijacking CAF-derived indispensable mitochondria through the formation of cellular bridges [45]. These studies indicated that the low OXPHOS and TCA cycle activity of CAFs are correlated to mitochondrial dysfunction. Consistent with previous research, our study found PGC-1α was low expressed in CAFs. Overexpression of PGC-1α led to a dramatic change of CAFs metabolic state which presented enhanced aerobic respiration and reduced glycolysis, suggesting that PGC-1α is essential for glycometabolism of CAFs. Our further study proved that PGC-1α negatively regulated PFKFB3, and the ectopic expression of PFKFB3 partially reversed the suppressive effect of PGC-1α overexpression on glycolysis and enhanced the effect of CAFs on pro-angiogenesis. These implied PGC-1α overexpression may impair the proangiogenic ability of CAFs by reducing PFKFB3-mediated glycolysis. Interestingly, we found that the ectopic expression of PFKFB3 in PGC-1α overexpressed CAFs resulted in an increase in ATP without significant changes in OCR, which reflected that the elevated glycolysis rate was not accompanied by a compensatory reduce in mitochondrial respiration. The study by Feng et al. suggested that the YAP/PFKFB3 axis is involved in hypoxia-induced glycolysis, but not oxidative phosphorylation and oxidative stress in endothelial cells [16]. In other words, PFKFB3 overexpression does not affect the maximal respiration rate of endothelial cells. Similarly, Jin Suk Park et al. revealed TRIM21 can target PFK and direct it to degradation, thus reducing glycolysis. However, the reduced glycolysis rate was not accompanied by a compensatory increase in mitochondrial respiration, indicating a decoupling between the two main energy-producing pathways [46]. Combined with our study, it can be considered that the regulation of mitochondrial aerobic respiration is upstream of the regulation of glycolysis. Meanwhile, attenuation of mitochondrial function leads to accumulation of oncometabolites [47]. The accumulation of lactate inside the cells inhibits the activity of PHD2 using O₂ to hydroxylate HIF1α, which leads to an increase in HIF1α level, thus promoting angiogenesis of HIF1α effect [48]. Intracellular lactate can also directly activate NDRG3, lactate receptor, to enhance angiogenesis via binding to c-Raf and Raf-ERK pathway [49]. This part of work will be further explored in follow-up studies.

Overall, in the TME of OSCC, the glycometabolism of CAFs closely ties to the phenotype of pro-angiogenesis. Although tumour vascularisation is regulated by multiple components, the novel role of stroma metabolism is proposed. CAFs of OSCC exhibit stronger glycometabolic reprogramming and proangiogenic phenotype. Manipulation of PGC-1α signalling impairs proangiogenic phenotype of CAFs by blocking PFKFB3-driven glycolysis. Targeting PGC-1α/PFKFB3 may provide a novel therapeutic intervention for the anti-angiogenic therapy of OSCC treatment.

Author contributions

The authors contributed in the following way: designed and performed the research: XL; write the manuscripts: XL, EJ and HZ; data analysis: YX and YC; perform experiments: CF and JL; supervised the research: ZS.

Funding

This study was supported by grants from the National Natural Science Foundation of China (Grant number: 81972547 to ZS); Chinese Fundamental Research Funds for the Central Universities (2042021kf0184 to EJ).

Data availability

The datasets generated during the current study are available from the corresponding author upon reasonable request.

Competing interests

The authors declare no competing interests.

Ethics approval and consent to participate

Study of OSCC tissue samples was in accordance with the Ethics Committee of School and Hospital of Stomatology at Wuhan University (2019LUNSHENZIA70). Written informed consents were obtained from all patients participated. Animal works were approved by the Ethical Committee on Animal Experiments of the Animal Care Committee of Wuhan University (S07921020L).

Consent to publish

Not applicable.

Supplementary information

The online version contains supplementary material available at 10.1038/s41416-022-01818-2.