Introduction

Pancreatic cancer (PC) is the malignant tumor with the highest mortality rate among gastrointestinal tumors. It is characterized by insignificant initial symptoms, rapid deterioration, low surgical resection rate, and poor overall prognosis [31]. In this study, we found the methylation level of HDAC4 mRNA was decreased under hypoxia and the expression of HDAC4 was increased. That is, m6A modification negatively regulates HDAC4 expression. According to the different functions of “reader” proteins, we speculate that YTHDF2 may be involved in the regulation of HDAC4 expression by regulating its mRNA stability. To substantiate this hypothesis, we firstly measured the expression level of YTHDC2 and YTHDF2, and the western blot assay showed, those “readers” were not affected under hypoxic conditions (Fig. 4E). Furthermore, we established YTHDF2-silenced stable cell lines (Supplementary Fig. 4A, B). Transcriptional expression level of HDAC4 were detected using a qRT-PCR assay, we found that HDAC4 mRNA expression levels were significantly increased when YTHDF2 was impaired compared with those in control cells (Fig. 4F). Thereafter, we performed RIP assays with anti-YTHDF2 antibody, as detected, HDAC4 mRNA was significantly enriched in the YTHDF2 antibody pull down sample compared with the IgG control (Fig. 4G). Together, our findings indicated that HDAC4 was governed by m6A modification depending on ALKBH5 and recognized by YTHDF2, which receded its stability under hypoxic conditions.

HDAC4 is identified as an oncogenic driver in PC

It has been reported that histone deacetylases (HDACs) promote tumour cell migration and the inhibition of HDACs may be a promising antitumour therapy for pancreatic cancer [32, 33]. Emerging evidence also indicates that HDAC4 mediates smoking-induced pancreatic cancer metastasis [34]. In clinical studies, elevated HDAC4 expression was significantly associated with the absence of organ metastases and tumor proliferative capacity in pancreatic adenocarcinoma [35]. These studies suggest that HDAC4 may be implicated in pancreatic malignant disease progression. In this study, we decided to systematically investigate the function of HDAC4 in pancreatic cancer under hypoxia. First, we established HDAC4-overexpression PANC-1 and MIA PaCa-2cell lines, and cultured them in 20 and 1% oxygen for 48 h (Supplementary Fig. 5A, B). We found that HDAC4 overexpression resulted in the significantly increased proliferation rate of PC cells under hypoxia compared to normoxia (Fig. 5A). The migratory assays indicated that HDAC4 overexpression significantly facilitated migratory ability of PC cells in normoxia and hypoxia (Fig. 5B, Supplementary Fig. 5C, D). Besides, HDAC4 overexpression amplified the hypoxia-induced key glycolysis regulatory genes, such as LDHA, HK2, GLUT1 and GLUT3 (Fig. 5C). Moreover, HDAC4 overexpression also increased production of lactate under normoxic and hypoxic conditions (Fig. 5D, E and Supplementary Fig. 5E). The extracellular acidification rate kinetic profiles also demonstrated the increase of glycolytic activity in HDAC4-overexpression PC cells (Fig. 5F, G and Supplementary Fig. 5F). Those data strongly suggest that HDAC4 could maintain hypoxia-induced glycolysis and metastasis in PC cells.

Fig. 5: The effects of HDAC4 on PC cell metastasis and glycolysis.
figure 5

A PC cells with NC and HDAC4 overexpression were incubated under normoxia or hypoxia. CCK8 assays were used to measure cell growth, and the relative proliferation rates were calculated. B Transwell assays were used to investigate the changes in the migratory capability of PC cells after viral transfection with HDAC4, and cells were exposed to 20% O2 or 1% O2 for 48 h. C PANC-1 and MIA PaCa-2 cells with NC and HDAC4 overexpression were cultured under normoxia or hypoxia for 48 h. And the mRNA expression of glycolysis driver genes (LDHA, HK2, GLUT1 and GLUTG3) was detected by qRT-PCR assay. Error bars are mean ± SEM. (n = 3). D, E Lactate production was measured in HDAC4-overexpressing PANC-1 or MIA PaCa-2 cells incubated under normoxia or hypoxia. F, G The extracellular acidification rate (ECAR) assays were performed to measure the glycolytic metabolism level in HDAC4-overexpressing cells incubated under normoxia or hypoxia. The quantification of glycolysis and glycolytic capacity was summarized from raw data. The metabolic inhibitors were injected sequentially at different time points as indicated. n = 3-4 for each treatment group. *P < 0.05, **P < 0.01, ***P < 0.001 and ns not significant.

Full size image

Hypoxia-induced glycolysis and metastasis were reversed by silenced HDAC4 in a m6A dependent manner

To confirm that the observed phenotypes were mediated by the dysregulation of ALKBH5- HDAC4 axis, we generated PC cells interfered with HDAC4 (Supplementary Fig. 6A, B). As expected, knockdown of HDAC4 significantly abolished the increased mobility ability induced by hypoxia in ALKBH5-overexpressing cells (Fig. 6A and Supplementary Fig. 6C). The Seahorse Extracellular Flux analysis showed hypoxia enhanced glycolytic capacity of ALKBH5-overexpressing cells, which could be reverted by HDAC4 silencing (Fig. 6B, C). HDAC4 silencing also abolished hypoxia-induced mRNA upregulation of key glycolysis regulatory gene in ALKBH5-overexpressing cells (Fig. 6D, E). And analogous results could be verified in ALKBH5-overexpressing cells, which were treated with tasquinimod (an inhibitor targeting HDAC4) under hypoxia (Fig. 6F–H and Supplementary Fig. 6D). These several functional rescue assays suggested that effects of hypoxia were reversed by silenced HDAC4.

Fig. 6: HDAC4 blockade reversed the effects of hypoxia in an m6A dependent manner.
figure 6

A ALKBH5-overexpressing PC cells transfected with the shNC or sh-HDAC4 were incubated under hypoxic conditions. Transwell assay was used to investigate the cell migration abilities. B, C ALKBH5-overexpressing PC cells transfected with the shNC or sh-HDAC4 were cultured under hypoxic conditions. The ECAR was measured to determine the glycolytic metabolism of treatment cells. Glycolytic variations (right) of glycolysis and glycolytic capacity were summarized from raw data. The metabolic inhibitors were injected sequentially at different time points as indicated. n = 3-4 for each treatment group. D, E ALKBH5-overexpressing PC cells transfected with the shNC or sh-HDAC4 were incubated under hypoxic conditions. The expression of a panel of glucose metabolism-related genes was detected by qRT-PCR. Error bars are mean ± SEM. (n = 3). F ALKBH5-overexpressing PANC-1 and MIA PaCa-2 cells treated with Tasquinimod under hypoxic conditions. The cell migration ability was assessed by Transwell assay. G ALKBH5-overexpressing PANC-1 and MIA PaCa-2 cells treated with Tasquinimod under hypoxic conditions. The glycolytic metabolism of treatment cells was measured by Seahorse XF24 system. Glycolytic variations (right) of glycolysis and glycolytic capacity were summarized from raw data. The metabolic inhibitors were injected sequentially at different time points as indicated. n = 3-4 for each treatment group. H The expression of a panel of glucose metabolism-related genes in PANC-1 and MIA PaCa-2 cells was detected by qRT-PCR. Error bars are mean ± SEM. (n = 3). *P < 0.05, **P < 0.01, ***P < 0.001 and ns not significant.

Full size image

HDAC4 enhanced the stability of HIF1α in hypoxic pancreatic cancer cells

HDAC4 possesses histone deacetylase activity and represses transcription when tethered to a promoter. Recent studies show that HDAC4-HIF1α axis as is an important signal pathway regulating glycolysis, apoptosis and autophagy in ovarian tumor cells during hypoxic adaptation [9, 36]. HDAC4 inhibits HIF1a ubiquitination and enhances its stability through deacetylation. Congruently, in this study, we aimed to demonstrate whether HDAC4 could enhance the stability of HIF1a in hypoxic pancreatic cancer cells. To confirm the regulatory role of HDAC4 on HIF1α, we treated HDAC4-overexpression PANC-1 and MIA PaCa-2 cell lines with Cobalt (II) chloride (CoCl2) to mimic hypoxia condition. Overexpression of HDAC4 significantly delayed HIF1α protein degradation (Fig. 7A). Conversely, knockdown of endogenous HDAC4 facilitated HIF1α protein degradation in hypoxic pancreatic cancer cells (Fig. 7B). We then generated a HIF1α mutant in which the lysine sites affected by HDAC4. The proteasome-specific inhibitor MG132 (10 µM) rescued the HIF1α protein from degradation in HDAC4-knockdown cells. Silenced HDAC4 caused a greater increase in acetylation of the wildtype HIF1α (HIF1α-wt) than in that of the mutant HIF1α (HIF1α-mut) (Fig. 7C). These results suggest that HDAC4 could enhance HIF1α protein stability.

Fig. 7: ALKBH5, HDAC4 and HIF1α regulate each other in pancreatic cancer cells.
figure 7

A, B PANC-1 and MIA PaCa-2 cells with HDAC4 overexpression or HDAC4 knockdown were treated with 200 µM CoCl2. Then, the cells were treated with 10 µg/mL cycloheximide (CHX) for the indicated time course (0, 1, 2, 3 h). The protein levels of HIF1α and HDAC4 were analyzed by western blot. C PC cells with or without silencing of HDAC4 expression were transfected with a wild-type Flag-HIF-1α (WT) or mutant Flag-HIF-1α (mut) plasmids for 24 h and exposed to hypoxia for 48 h. Then, the cells were treated with 5 μM MG132 for another 6 h. Cell lysates were immunoprecipitated with an anti-Flag antibody, and then immunoblotted for Flag and the acetylation form of HIF1α. D, E PC cells with HIF1α knockdown (siHIF1α) or control (NC) were treated with or without 200 µM CoCl2, and the expression of HDAC4 and HIF1α was detected using qRT-PCR and western blot. F Predicted HRE sequences in ALKBH5 promoter by JASPAR and the diagram for constructing for constructing of luciferase reporters. The wild-type or HRE mutant sequence of ALKBH5 promoter was inserted into a pcDNA3.1 vector. G Dual luciferase reporter assays showing the effect of HIF1α on ALKBH5 mRNA reporters with either wild-type or mutated HRE sequence under hypoxia. *P < 0.05, **P < 0.01, ***P < 0.001 and ns not significant.

Full size image

ALKBH5 was transcriptionally induced by HIF1α under hypoxic conditions

It has been reported that ALKBH5 is a nuclear 2-oxoglutarate dependent oxygenase and is a direct target of HIF1α [23, 31], whereas another report showed hypoxia favored m6A labelling in HCC cells [39]. The role of m6A in transcriptome reprogramming during hypoxic response in PC cells are still unclear. In this study, we systematic analyzed the alterations of m6A and its target genes in hypoxic pancreatic cancer cells by the MeRIP-seq. We found that hypoxia reduced m6A modification level of total mRNAs in a ALKBH5 dependent manner. Unexpectedly, our data suggested that ALKBH5 may act as an oncogenic driver to promote cell metastasis in hypoxia instead of serving as a tumor suppressor in normoxia. Nevertheless, the mechanism by which ALKBH5 acts inversely in normoxia and hypoxia is still unclear.

Glycolysis plays a critical role in the regulation of the tumour microenvironment, affecting biological processes such as inflammatory factor secretion, immune evasion and tumor angiogenesis [40,41,42]. Glycolysis is one of the primary metabolic signatures in cancers and supports the energetic requirements of sustained proliferation and metastasis [43]. Apart from DNA and histone modifications, RNA m6A modification has been recently proposed to be another important layer of epigenetic regulation in energy metabolism regulation [44, 31, 47]. HIFs are composed of the stable HIF1β subunit and the oxygen-sensitive HIF1/2α subunit [48]. HIF1α is a crucial regulator of oxygen homeostasis and is degraded by the ubiquitin–proteasome pathway [49, 50]. Under hypoxic conditions, HIF1α is stabilized and binds with HIF1β to promote the transcription of downstream gene [51, 52]. Ling et al. found that the stability and transcriptional activity of HIF1α were enhanced by USP22 depended its deubiquitination activity [50]. Consistent with a previous report [9], we detected HDAC4-mediated deacetylation enhanced HIF1α stability under hypoxia. Recently, ALKBH5 was shown to be directly targeted by HIF1α and regulated by HIF2α in breast cancer cells [38]. Our study also confirmed that overexpressing HIF1α promoted ALKBH5 transcription in pancreatic cancer cells. Therefore, we demonstrated a positive feedback loop between ALKBH5, HDAC4 and HIF1α, which contributes to the maintenance of PC glycolysis under hypoxic microenvironment (Fig. 9).

Fig. 9: The proposed model of hypoxia-related m6A modification induced pancreatic cancer cell glycolysis and metastasis via the ALKBH5-HDAC4-HIF1α-positive feedback loop.
figure 9

The expression of ALKBH5 was increased under hypoxic conditions which led to increased stability of HDAC4 mRNA via m6A modification and in a YTHDF2 dependent manner. Increased HDAC4 enhanced HIF1a protein stability. And then overexpressed HIF1a promoted transcription of ALKBH5, LDHA, HK2, GLUT1 and GLUT3. Therefore, ALKBH5/HDAC4/HIF1α form a positive feedback loop, which then induces a more glycolytic metabolism and migration of PC cells under hypoxia.

Full size image

Altogether, our study revealed the mechanism of m6A modification and the ALKBH5/HDAC4/HIF1α positive feedback loop in promoting glycolysis of PC. Based on our findings, the hypoxia-directed treatment maybe a novel therapeutic option for PC patients.

Materials and methods

RNA m6A dot blot assay

Poly(A)+ mRNA was enriched using GenElute messenger RNA (mRNA) Miniprep Kit (Sigma‐Aldrich). The concentration and purity of RNA were measured by an Agilent Bioanalyzer 2100. Poly(A)+ mRNA samples were denatured at 70 °C for 5 min and then diluted equally to 400 ng, 200 ng and 100 ng in equal volumes. After denaturation, equal volume of diluted mRNA was added into a nylon membrane (GE Healthcare). The membranes were cross-linked at 245 nm UV for three times under auto-cross-linking mode (UVP analytik jena, CL-1000M). Then, the membranes were blocked with 5% BSA for 2 h at room temperature and incubated with anti‐m6A antibody (Synaptic System, 202003) overnight at 4 °C. Membranes were washed with PBST for three times. Horseradish peroxidase-conjugated anti-rabbit immunoglobulin G (CST, 7074P2) were diluted 1:5000 and incubated with the membranes for 1 h at room temperature. Membranes were washed with PBST and the signals were detected by standard analysis of HRPO-induced chemiluminescence (ECL, Millipore). The same gradient-diluted Poly(A)+ mRNA was also added into the nylon membrane, stained with methylene blue (Solarbio, G1300) for 2 h, and washed with ribonuclease-free water.

m6A sequencing and m6A-RNA immunoprecipitation assay

MeRIP sequencing was performed by Jiayin Biomedical Technology company (shanghai, China). In brief, total RNAs were isolated from PANC-1 cells exposed to hypoxia or normoxia, and then chemically fragmented into 100 ~ 200 nt. The fragmented RNA was incubated with m6A antibody (milipore) for immunoprecipitation. The enrichment of m6A containing mRNA was sent for high-throughput NGS and validated by quantitative RT–PCR. For NGS, purified RNA fragments from m6A-MeRIP were used for library construction (removing rRNA). Sequencing reads were aligned to the human genome hg38_gencode with STAR program. The m6A modification peaks were identified using software package of MetPeak. Motifs enriched with m6A peaks were identified by MEME. The R software package Guitar was used to calculate the frequency of each sample’s peak at each site in the mRNA transcript region and draw the frequency distribution map. The distribution of peaks in the functional area covered is drawn by R software package ChiIPseeker. Finally, GO and Pathway enrichment analysis of m6A peaks associated genes were carried out.

Pimonidazole (PIMO) staining assay

2 mg pimonidazole (Hypoxyprobe) was injected intravenously into each tumor-bearing mouse and was left to circulate for 1 h before tumour resection. The obtained tumor tissue was sectioned by LEICA CM1950 (Leica). Staining of tumour sections were performed with a Hypoxyprobe RedAPC Kit (HP8-100) according to manufacturer’s instructions. Specifically, tumour sections were fixed in 4% paraformaldehyde at room temperature for 30 min. After removing fixing solution, the sections were washed 3 times with PBS and then incubated with blocking solution (goat sreum) for 1 h at room temperature. The sections were incubated with primary antibody and PIMO at 4 °C overnight. After wash, sections were incubated with corresponding secondary antibodies for 1 h at room temperature. Finally, sections were incubated with DAPI (Beyotime, China) for 5 min at room temperature, and then mounted with mounting medium containing anti-fluorescence quencher was added to mount the slides. Fluorescent images were captured using a confocal microscope (Zeiss, Germany).

Glycolysis metabolic measurements

PC cells were seeded on Seahorse XF24 culture cell plates at a concentration of 5.0 × 104 cells per well, and cells were incubated for 24 h under 20% or 1% O2. The probe plates were pretreated with 1 ml XF Calibrant in a non-CO2 incubator at 37 °C overnight. On the second day, the treatment media were removed and cultured cells were washed twice with assay medium Seahorse XF DMEM Medium (PH = 7.4) which is supplemented with 2 mM glutamine. Before detection, the culture cell plates were incubated for 1 h in a non-CO2 incubator at 37 °C. The final concentration of inhibitors and activators are 10 mM glucose, 1 μM Oligomycin, and 50 mM 2-DG. Cells were then assayed with an XFe24 extracellular flux analyzer (Seahorse XFe/XF). All the consumables were purchased from Agilent.

Statistical analysis

Data were analyzed with GraphPad Prism V9.0 and SPSS V19.0. The significance of differences between groups were assessed using Student’s t test or χ2 test as appropriate. The results are presented as mean ± standard error of mean (SEM) from three independent experiments. P value <0.05 was considered to be statistically significant. More information of the materials and methods is in the Supplementary Materials.