If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Ephrin-A3/EphA2 axis is a critical hypoxia-regulated tumour-promoting pathway in HCC.
•
ACLY is a key functional mediator of Ephrin-A3/EphA2 signaling-induced HCC stemness.
•
Expression of EFNA3 and ACLY are associated with worse survival in patients with HCC.
•
Ephrin-A3/EphA2 axis rewires the metabolic profile of HCC via regulation of ACLY.
Background & Aims
The highly proliferative nature of hepatocellular carcinoma (HCC) frequently results in a hypoxic intratumoural microenvironment, which creates a therapeutic challenge owing to a lack of mechanistic understanding of the phenomenon. We aimed to identify critical drivers of HCC development and progression in the hypoxic microenvironment.
Methods
We performed integrative analysis of multiple transcriptomic and genomic profiles specific for HCC and hypoxia and identified the Ephrin-A3/Eph receptor A2 (EphA2) axis as a clinically relevant and hypoxia-inducible signalling axis in HCC. The functional significance and mechanistic consequences of the Ephrin-A3/EphA2 axis were examined in EFNA3- and EPHA2- knockdown/overexpressing HCC cells. The potential downstream pathways were investigated by transcriptome sequencing, quantitative reverse-transcription PCR, western blotting analysis and metabolomics.
Results
EFNA3 was frequently upregulated in HCC and its overexpression was associated with more aggressive tumour behaviours. HIF-1α directly and positively regulated EFNA3 expression under hypoxia. EFNA3 functionally contributed to self-renewal, proliferation and migration in HCC cells. EphA2 was identified as a key functional downstream mediator of EFNA3. Functional characterisation of the Ephrin-A3/EphA2 forward-signalling axis demonstrated a promotion of self-renewal ability and tumour initiation. Mechanistically, the Ephrin-A3/EphA2 axis promoted the maturation of SREBP1 and expression of its transcriptional target, ACLY, was significantly associated with the expression of EFNA3 and hypoxia markers in clinical cohorts. The metabolic signature of EPHA2 and ACLY stable knockdown HCC cells demonstrated significant overlap in fatty acid, cholesterol and tricarboxylic acid cycle metabolite profiles. ACLY was confirmed to mediate the self-renewal function of the Ephrin-A3/EphA2 axis.
Conclusions
Our findings revealed the novel role of the Ephrin-A3/EphA2 axis as a hypoxia-sensitive modulator of HCC cell metabolism and a key contributor to HCC initiation and progression.
Lay summary
Hepatocellular carcinoma (HCC) is a fast-growing tumour; hence, areas of the tumour often have insufficient vasculature and become hypoxic. The presence of hypoxia within tumours has been shown to negatively impact on the survival of patients with tumours, including HCC. Herein, we identified the Ephrin-A3/EphA2 axis as a key functional driver of tumour initiation and progression in response to hypoxia. Additionally, we showed that SREBP1-ACLY-mediated metabolic rewiring was an important downstream effector that induced cancer stemness in response to Ephrin-A3/EphA2 forward-signalling.
We read with great interest the article by Husain et al.,1 showing that the expression of EFNA3 was upregulated in hepatocellular carcinoma (HCC) and related to poorer survival rates. Husain et al. found that the expression level of EFNA3 was regulated by HIF-1α in a hypoxic microenvironment. Hypoxia-induced Ephrin-A3/EphA2 forward signaling played a vital role in initiation and progression of HCC. The authors identified the clinical significance and molecular mechanisms of EFNA3 in HCC. However, the relationship between the HCC patients’ overall survival (OS) and EFNA3 was explored only based on The Cancer Genome Atlas (TCGA) database in their study.
It is an aggressive tumour associated with poor patient survival. Its fast-growing nature frequently results in insufficient or inefficient vasculatures and lower oxygen supplies within them, giving rise to hypoxic niches. The presence of intratumoural hypoxia in tumours has been demonstrated to be strong predictor of poor patient survival and is recognised as a driver of disease progression at various stages of tumour development, including initiation and metastasis.
Hypoxia-inducible factors (HIFs) are the key transcription factors that drive adaptive changes in cells for their survival and propagation in hypoxic microenvironments. The current first-line therapeutic regimens (atezolizumab + bevacizumab) against advanced HCC have an important target, i.e. vascular endothelial growth factor pathway, which is regulated by hypoxia or HIF pathways.
There is overwhelming evidence suggesting that self-renewing tumour cells drive tumour initiation, metastasis and drug resistance, phenotypes that lead to poorer clinical outcomes.
Recent lines of evidence suggest that cancer stemness, defined by the self-renewal and tumour-initiating capacity of cells, is a plastic phenotype and hypoxia can modulate the degree of cancer stemness within a population of cells in several cancer types including HCC.
However, the precise mechanism by which hypoxia induces cancer stemness is still not fully understood.
Herein, we report that the Ephrin-A3/EphA2 axis is a hypoxia-sensitive modulator of cancer stemness in HCC cells. Ephrin-A3 (EFNA3) belongs to a family of membrane-bound ligands called Ephrins. This family of ligands corresponds to the largest receptor tyrosine kinase (RTK) family, namely Eph receptors. Ephrins and Eph receptors are highly diverse in their range of signalling modes. Ephrins can bind to various members of the Eph receptor family and participate in forward, reverse and bi-directional signalling.
Furthermore, Eph receptors can participate in ligand-dependent and -independent signalling. As a result, the functional outcome of any Ephrin or Eph receptor ought to be highly context-dependent. EFNA3 mRNA has been demonstrated to be under the regulation by hypoxia-inducible microRNAs and long non-coding RNAs, which down- and up-regulate EFNA3 mRNA, respectively.
EFNA3 remains poorly characterised in HCC. Herein, we demonstrated for the first time that EFNA3 mRNA is directly regulated by HIF-1α in HCC cells and exhibits various pro-tumourigenic functional phenotypes. Of these, promotion of self-renewal was found to be mediated through EphA2. To date, EphA2 remains to be investigated in-depth in the context of Ephrin-A3 and hypoxia. We provide evidence that the Ephrin-A3/EphA2 axis is involved in metabolic reprogramming in the hypoxic microenvironment and regulates tumour initiation by modulating ATP citrate synthase (ACLY) expression in HCC.
EFNA3 is a clinically relevant tumour-promoting gene under direct regulation of hypoxia
To identify genes that are directly regulated by HIFs and contribute to HCC disease progression, we performed integrative analysis of multi-domain data, where we evaluated 4 distinct datasets consisting of: i) genome-wide core hypoxia response element (HRE) motif mapping, including genes containing at least one HRE consensus sequence (5’-RCGTG-3’) at their proximal promoter regions, i.e. 1 kb upstream of their transcription start site (TSS); ii) global HIF-1 chromatin binding profile in HepG2 cells exposed to hypoxia,
for genes bound by HIF-1 under hypoxia; iii) hypoxia-induced transcriptomic profile of 6 HCC cell lines exposed to 24 h hypoxia; and 4) whole transcriptomic sequencing profile of 16 pairs of HCC and corresponding non-tumourous liver (The Hong Kong University - Queen Mary Hospital [HKU-QMH] database) (Fig. 1A). With the above selection criteria, we identified EFNA3 as the most significantly upregulated gene in HCCs compared to corresponding non-tumourous liver in both HKU-QMH and The Cancer Genome Atlas-Liver Hepatocellular Carcinoma (TCGA-LIHC) cohorts (Fig. S1A).
Fig. 1Hypoxia and CNV-mediated overexpression of EFNA3 is clinically relevant in HCC.
(A) Identification of EFNA3 with integrative analysis of datasets i.e. HRE-containing genes, direct targets of HIF-1α, genes upregulated upon hypoxia treatment, and genes significantly overexpressed in HCC. (B) EFNA3 expression in HCC and corresponding NTL tissue samples in HKU-QMH (n = 97) and TCGA-LIHC cohort (NT, n = 50; HCC, n = 373). (C) Relative EFNA3 expression in paired HCCs with NTL tissue in the HKU-QMH cohort. (D) Clinicopathological correlation analysis of EFNA3 expression. (E) Kaplan-Meier plot for overall survival in TCGA-LIHC cohort stratified according to EFNA3 median expression. (F) EFNA3 mRNA expression change upon hypoxia treatment in HCC cell lines. (G) Ephrin-A3 and HIF-1α protein levels upon hypoxia treatment by western blot. (H) Enrichment of EFNA3 promoter region upon immunoprecipitation with anti-HIF-1α antibody in hypoxic conditions by ChIP assay. (I) Luciferase reporter activity of HRE in EFNA3 promoter region upon co-transfection with HIF1A or HIF1A(del-ODD) cDNA. (J) CNV of EFNA3 in HKU-QMH cohort (n = 20) and TCGA-LIHC (n = 370). (K) Expression of EFNA3 stratified according to EFNA3 CNV in TCGA-LIHC cohort. (L) EFNA3 expression of NTL and HCCs stratified according to hypoxia and EFNA3 CNV status in TCGA-LIHC cohort. Student’s t test was used for comparing two groups, Fisher's exact test was used for clinicopathological correlation anaysis and log-rank (Mantel-Cox) test was used for survival analysis. (∗p <0.05, ∗∗p <0.01, ∗∗∗p <0.001, ∗∗∗∗p <0.0001). CNV, copy number variation; HCC, hepatocellular carcinoma; HIF, hypoxia-inducible factor; HKU-QMH, The University of Hong Kong-Queen Mary Hospital; HRE, hypoxia response element; NTL, non-tumorous liver; TCGA-LIHC, The Cancer Genome Atlas-Liver Hepatocellullar Carcinoma. (This figure appears in color on the web.)
We validated the overexpression of EFNA3 in TCGA-LIHC (n = 370, p <0.0001) and in our HKU-QMH cohort (n = 97, p <0.0001) by quantitative PCR (Fig. 1B), with the majority of cases showing over 2-fold overexpression (Fig. 1C). Overexpression of EFNA3 (≥4-fold) correlated with a more aggressive phenotype of HCC, including the presence of venous invasion, more advanced tumour stage in HKU-QMH and TCGA-LIHC cohorts (Fig. 1D and Fig. S1B) and poorer overall survival of patients in TCGA-LIHC cohort (Fig. 1E). It is of note that EFNA3 is also frequently overexpressed in the other cancer types (Fig. S1C). These results suggest that EFNA3 is clinically relevant and a potential tumour-promoting gene under the regulation of hypoxia.
EFNA3 is upregulated by HIF-1α-mediated hypoxia and copy number variation
To confirm the hypoxia-mediated induction of EFNA3, the mRNA and protein expression of EFNA3 upon 1% O2 treatment were examined in various HCC cell lines. A significant upregulation of EFNA3 was observed in MHCC97L (authenticated to have no contamination; see the supplementary information and Fig. S8), PLC/PRF/5, HepG2 and Huh7 cells (Fig. 1F,G and Fig. S1D). Next, the presence of the HRE consensus sequence on the TSS of EFNA3 was mapped by in silico analysis (Fig. S1E). Chromatin immunoprecipitation assay on HIF-1α significantly enriched for EFNA3 mRNA proximal promoter region (Fig. 1H). Moreover, HIF-1α or HIF-1α(del-ODD) [mutant insensitive to proteasomal degradation due to deletion of oxygen-dependent domain i.e. amino acids 401-603
] drove the expression of the EFNA3 promoter and was specific for the presence of the HRE sequence at the TSS of EFNA3 mRNA (Fig. 1I). Lastly, knockdown (KD) of HIF1A abolished the hypoxia-mediated upregulation of EFNA3 expression in HCC cells (Fig. S1F,G). These data strongly suggest that EFNA3 is under direct transcriptional regulation of HIF-1α in HCC.
EFNA3 is located on chromosome 1q21.3, a frequently amplified arm in HCC.
Indeed, over 60% of the HCC tumours in HKU-QMH and TCGA-LIHC cohorts showed copy number increase of EFNA3 (Fig. 1J). Stratification of the HCC tumours according to copy number variation (CNV) showed that higher CNV positively correlated with higher expression of EFNA3 (Fig. 1K). When TCGA-LIHC primary HCC tumour samples were stratified according to EFNA3 CNV in combination with hypoxic status, where carbonic anhydrase 9 (CA9) and SLC2A1 (also known as GLUT1) expression were used to stratify HCCs into “hypoxic” or “normoxic” status (Fig. S8B-8E) (described in the supplementary information), EFNA3 expression was highest when the 2 factors were combined (Fig. 1L). Furthermore, patients that were positive for both factors had the poorest survival outcome (Fig. S1H). These results suggest that hypoxia and EFNA3 CNV co-operate to drive overexpression of EFNA3 in patients with HCC.
EFNA3 promotes tumour initiation, tumour growth and cancer stemness in HCC
For functional characterisation of EFNA3 in HCC, we established stable EFNA3 KD cell models (short-hairpin [sh]ENFA3#1 and shEFNA3#5) in hypoxia-sensitive HCC cell lines with relatively high EFNA3 expression, including MHCC97L, PLC/PRF/5 and Huh7 cells (Fig. 2A, 2B and Fig. S2A). shEFNA3 significantly suppressed the sphere-forming ability (Fig. 2C) and proliferation rate of HCC cells in vitro (Fig. S2B). Intriguingly, recombinant Ephrin-A3-Fc treatment markedly promoted only the sphere-forming ability of HCC cells but not their proliferation rate (Fig. 2D and Fig. S2C). The suppressive effect of EFNA3 KD on migratory ability was observed to be prominent under hypoxia (Fig. S2D). Additionally, in an in vivo liver orthotopic implantation model, EFNA3 KD cells led to a significant reduction in tumour-forming ability in the liver (Fig. 2E) and resulted in a lower incidence of lung metastasis (Fig. 2F). To further assess the tumour-initiating ability of EFNA3 KD cells, a limiting dilution assay was performed, where 1x105, 1x104 and 1x103 HCC cells were injected subcutaneously into NOD/SCID mice. We observed a significantly lower tumour-initiating capacity in shEFNA3 cells relative to non-target control cells (Fig. 2G). Extreme limiting dilution analysis on the limiting dilution assay showed a 70-fold reduction in “stem-cell frequency” (p = 3.12x10-9), which estimates the frequency of self-renewing tumour-initiating cells.
(A) EFNA3 mRNA levels in knockdown of EFNA3 (shEFNA3) and NTC in HCC cells by qPCR. (B) Ephrin-A3 protein upon KD of EFNA3 in HCC cells by western blotting. (C) Sphere formation assay upon KD of EFNA3. Scale bars: 100 μm. (D) Sphere formation assay upon recombinant Ephrin-A3 stimulation. Scale bars: 100 μm. (E) Bioluminescence intensity from tumours of liver orthotopic model of luciferase-labelled MHCC97L cells upon EFNA3 KD. (F) Bioluminescence intensity of lung metastases. (G) Limiting dilution assay of MHCC97L cells upon KD of EFNA3 in NOD/SCID mice. Student’s t test was used for comparing two groups and extreme limiting dilution analysis (ELDA) was used for limiting dilution assay (∗p <0.05, ∗∗p <0.01, ∗∗∗p <0.001, ∗∗∗∗p <0.0001). HCC, hepatocellular carcinoma; KD, knockdown; NTC, non-target control. (This figure appears in color on the web.)
EphA2 is a key functional mediator of Ephrin-A3 in HCC
Since Ephrins are ligands for Eph receptors, we sought to identify the corresponding receptor for Ephrin-A3 in HCC. We screened for receptors whose kinase activities were i) induced upon Ephrin-A3 stimulation, and ii) reduced upon KD of EFNA3 in HCC cells (Fig. 3A) by utilising a human RTK phosphorylation antibody array to measure their tyrosine phosphorylation levels. RTK activity was profiled in Huh7 stimulated with recombinant Ephrin-A3 for 30 mins and MHCC97L and Huh7 with shEFNA3 (Fig. 3A). Only EphA2 demonstrated a marked induction in its tyrosine kinase activity upon Ephrin-A3 stimulation and a consistent reduction in shEFNA3 cells (Fig. 3B,C and Fig. S3A). Furthermore, abundant EPHA2 mRNA and protein expression was confirmed in HCC cell lines previously utilised to determine functional consequences of EFNA3 (Fig. 3D and Fig. S3B). HKU-QMH and TCGA-LIHC RNA-sequencing cohorts also confirmed EPHA2 is one of the Eph receptor family members to have the highest expression in human HCC (Fig. S3C,D). The Ephrin-A3-induced tyrosine kinase activity of EphA2, resulting in increased phospho-EphA2 (Tyr-588) in HCC cells, was validated by immunoprecipitation (Fig. 3E). EFNA3 KD-mediated reduction in phospho-EphA2 (Tyr588) was also confirmed by western blotting (Fig. 3F). Furthermore, endogenous EphA2 successfully pulled down with recombinant Ephrin-A3-Fc, suggesting that EphA2 is a direct binding partner of Ephrin-A3 in HCC (Fig. 3G). Hypoxia treatment could also induce phospho-EphA2 (Tyr588) (Fig. 3H). Furthermore, phospho-EphA2 (Tyr588) levels were induced and sustained for 60 mins upon Ephrin-A3 stimulation (Fig. 3I). These results revealed that Ephrin-A3 directly binds with EphA2 and induces autophosphorylation of tyrosine residues (Tyr588).
Fig. 3EphA2 is a key functional mediator of Ephrin-A3 in HCC.
(A) Screening strategy for identifying candidate Eph receptors corresponding to Ephrin-A3 in HCC. (B) Immunoblots of phospho-RTK array from Huh7 cells supplemented with/without EphrinA3-Fc. Signal intensity heatmaps showing relative phosphor-RTK levels. (C) Phospho-Tyr of Eph receptors in Huh7 (upper) supplemented with 1 ug/ml Ephrin-A3-Fc, shEFNA3 MHCC97L (middle) and shEFNA3 Huh7 (lower) relative to controls. (D) Expression profile of Eph receptors in HCC cell lines by transcriptome sequencing. (E) Immunoprecipitation of EphA2 and detection of phospho-EphA2 (Tyr588) levels upon Ephrin-A3-Fc treatment in Huh7 cells. (F) Relative levels of phospho-EphA2 (Tyr588) and total EphA2 upon KD of EFNA3. (G) Physical binding of recombinant Ephrin-A3-Fc to endogenous EphA2 in Huh7 cells. (H) Expression of EphA2 and phospho-EphA2 (Tyr588) in Huh7 cells upon hypoxic or normoxic treatment. (I) Phospho- EphA2 (Tyr 588) levels upon Ephrin-A3-Fc treatment in 60 minutes. (J) Sphere formation assay upon KD of EPHA2 in HCC cells. Scale bars: 100 μm. (K) Subcutaneous injection model upon KD of EPHA2 in Huh7 cells at experimental endpoint. (L) Limiting dilution assay of Huh7 cells upon KD of EPHA2 in NOD/SCID mice. Student’s t-test was used for comparing two groups and extreme limiting dilution analysis (ELDA) was used for limiting dilution assay (∗p <0.05, ∗∗p <0.01, ∗∗∗p <0.001, ∗∗∗∗p <0.0001). HCC, hepatocellular carcinoma; IP, immunoprecipitation; KD, knockdown; NTC, non-target control; RTK, receptor tyrosine kinase; Tyr, tyrosine residue. (This figure appears in color on the web.)
Next, we tested the functional role of EphA2 in HCC using shEPHA2 Huh7 and PLC/PRF/5 cells (Fig. S3E). We observed functional overlap between EPHA2 and EFNA3, where the stable shEPHA2 demonstrated significantly suppressed sphere-forming ability (Fig. 3J) and proliferation rate (Fig. S3F). Migration rate was modestly reduced in shEPHA2 cells (Fig. S3G). Concordantly, subcutaneous injection of shEPHA2 HCC cells in nude mice led to a significant reduction in tumour growth and incidence rate (Fig. 3K). A limiting dilution assay further confirmed that shEPHA2 cells had significantly lower tumour-initiating capacity, with a 6.4-fold (p = 0.048) reduction in “stem-cell frequency” (Fig. 3L). These results suggest that EphA2 is a functionally relevant Eph receptor in HCC.
Hypoxia and Ephrin-A3-mediated EphA2 tyrosine kinase activity is prominent in HCC development
Since induction of EphA2 tyrosine kinase activity is a key event in Ephrin-A3-stimulated signalling, we queried if EphA2 tyrosine kinase activity in Ephrin-A3/EphA2 forward-signalling is essential for its pro-tumourigenic function in HCC. We utilised phospho-EphA2 (Tyr588) levels as a readout for tyrosine kinase activity of the EphA2 receptor. In Huh7 cells, Ephrin-A3 ligand treatment only induced the phosphorylation of phospho-EphA2 (Tyr588) but not phospho-EphA2 (Ser 897), i.e. phosphorylation site independent of Ephrin-A3/EphA2 forward-signalling. We also generated MHCC97L cells overexpressing wild-type (EPHA2) or kinase-dead, i.e. K646M mutant forms (EPHA2 [K464M]) to further assess the signalling of this axis.
Point mutation of the tyrosine kinase domain in EPHA2 K646M resulted in a significant reduction of phospho-EphA2 (Tyr588) levels while maintaining phospho-EphA2 (Ser897) (Fig. 4A).
This was also observed in the immunohistochemistry staining of liver orthotopic MHCC97L xenograft tumours with EPHA2 and EPHA2 (K646M) overexpression (Fig. 4B and Fig. S4). It was noted that the level of EphA2 (Tyr588) phosphorylation was more prominent in the hypoxic regions of the tumours as indicated by positive staining of CA9 in tumour tissue, suggesting that the EphA2 tyrosine kinase activity induced by Ephrin-A3/EphA2 forward-signalling was upregulated in hypoxic niches (Fig. 4B). Furthermore, functionally, mutation of the tyrosine kinase activity in EPHA2 K646M resulted in the abolishment of EPHA2-induced self-renewal ability in MHCC97L cells (Fig. 4C). Treatment with Ephrin-A3, hypoxia, or both also enhanced the self-renewal ability, and KD of EPHA2 was able to desensitise HCC cells to the self-renewal promoting response for both Ephrin-A3 and hypoxia (Fig. 4D). The ephrin-A3/EphA2 axis also stimulated induction of the cancer stem-cell marker CD13 (Fig. S3H). These results suggest that the activation of EphA2 tyrosine kinase activity induced by Ephrin-A3/EphA2 forward-signalling is critical in hypoxia-mediated induction of self-renewal in HCC cells. In addition, immunohistochemistry on tissue microarrays from the HKU-QMH cohort (n = 54) revealed that approximately 37% of HCC tumours had elevated phospho-EphA2 (Tyr588) levels (Fig. 4E and 4F). Taken together, the hypoxia-mediated Ephrin-A3/EphA2 forward-signalling axis is a prominent feature of HCC tumours, which can induce self-renewal ability of HCC cells.
Fig. 4Hypoxia and Ephrin-A3-mediated EphA2 tyrosine kinase (Tyr588) activity is prominent in HCC.
(A) Expression of EphA2, phospho-EphA2 (Tyr588) and phospho-EphA2 (Ser897) upon recombinant Ephrin-A3 treatment in Huh7 cells, and upon stable overexpression of EPHA2 and EPHA2 (K646M) in MHCC97L cells. (B) IHC detection of EphA2, phospho-EphA2 (Tyr588) and CA9 in EPHA2 WT and EPHA2 K646M mutant OE MHCC97L liver orthotopic xenograft tumour. Scale bars: 500 μm (C) Sphere formation assays upon stable OE of EPHA2 and EPHA2 (K646M) in MHCC97L cells. Scale bars: 100 μm (D) Sphere formation assay upon KD of EPHA2 treated with 2.5 μg/ml recombinant Ephrin-A3 and/or hypoxia in Huh7 cells. Scale bars: 100 μm. (E) Summary of relative staining intensity of phospho-EphA2 (Tyr588) in HCC and corresponding NTL tissues in TMA of HKU-QMH cohort. (F) Representative images of TMA Student’s t test was used for comparing two groups (∗p <0.05, ∗∗∗∗p <0.0001). EV, empty vector control; HCC, hepatoceullar carcinoma; NTL, non-tumourous liver; Tyr, tyrosine residue. (This figure appears in color on the web.)
Ephrin-A3/EphA2 axis causes SREBP1 transcriptional activation via its maturation
To uncover the underlying mechanism of the promoting effect of tumour initiation and cancer stemness on the Ephrin-A3/EphA2 axis in HCC cells, transcriptome sequencing was performed on Huh7 and PLC/PRF/5 HCC cells with either EFNA3 or EPHA2 KD (Fig. 5A). The transcriptomic changes upon KD of EFNA3 or EPHA2 of 2 HCC cell lines were integrated and subjected to gene-set enrichment analysis for canonical pathways (v7.1).
(A) Transcriptome sequencing strategy to identify pathways regulated by both EFNA3 and EPHA2. (B) Volcano plot of GSEA pathways for shEFNA3 and shEPHA2 clones. (C) GSEA enrichment plot of “REACTOME ACTIVATION OF GENE EXPRESSION BY SREBF/SREBP” gene set. Expression of SREBP1 target genes in (D) EPHA2 KD cells, (E) upon 6 h stimulation with 0, 1 or 2.5 μg/ml recombinant Ephrin-A3, (F) upon knockdown of EFNA3 under hypoxia in Huh7 cells, and (G) upon overexpression of EPHA2 or EPHA2 (K646M) under hypoxia in MHCC97L cells. Expression of pre-mature and mature SREBP1 in (H) shEFNA3 Huh7 cells, (I) Ephrin-A3-Fc -/+ shEPHA2-treated Huh7 cells and (J) upon overexpression of EPHA2 or EPHA2 (K646M) in MHCC97L cells. (K) Free fatty acid levels upon KD of EPHA2 in Huh7 cells. (L) ORO staining on orthotopic xenografts with overexpression of EPHA2, and EPHA2 (K646M). (M) Sphere formation assay of siSREBP1 PLC/PRF/5 and Huh7 cells in combination with 2.5 μg/ml recombinant Ephrin-A3. Scale bars: 100 μm. Student’s t test was used for comparing two groups (∗p < 0.05, ∗∗p <0.01, ∗∗∗p <0.001, ∗∗∗∗p <0.0001). EV, empty vector control; NTC, non-target control; ORO, oil red O. (This figure appears in color on the web.)
The majority of these were metabolic pathways, with lipid-related pathways being the most prominently affected. Of these, transcriptional activity of sterol regulatory element-binding proteins (SREBPs), master regulators of lipid metabolism, as denoted by “REACTOME Activation of Gene Expression by SREBF/SREBP” was one of the top 5 significant negatively enriched pathways, with a normalised enrichment score of -2.021 (nominal p value = 0.01, FDR q-value = 0.022) (Fig. 5C). This gene set comprises SREBP1 and SREBP2 target genes, with genes transcribed by SREBP1, i.e. stearoyl-CoA desaturase (SCD), fatty acid synthase (FASN) and acetyl-CoA carboxylase (ACACA), demonstrating a stronger downregulation (Fig. S5B). Hence, SREBP1 was considered for further characterisation by qPCR assessment of ACACA, FASN, SCD, ACSS2 (acyl-CoA synthetase 2) and ACLY, known SREBP1 target genes. Stable KD of EPHA2 significantly reduced the expression of SREBP1 target genes (Fig. 5D), while recombinant Ephrin-A3 treatment significantly induced them in HCC cells (Fig. 5E). Moreover, under hypoxic conditions, shEFNA3 significantly reduced the expression of SREBP1 target genes (Fig. 5F), while EPHA2 overexpression increased their expression, specific to its kinase activity (Fig. 5G), suggesting that the endogenous Ephrin-A3/EphA2 axis promotes SREBP1 transcription under hypoxia.
Maturation of SREBP1 is a critical step in its transition from a transcriptionally inactive to active state.
KD of EFNA3 significantly reduced mature SREBP1 levels (Fig. 5H). Recombinant Ephrin-A3 treatment induced SREBP1 maturation in Huh7 cells, which was reduced upon KD of EPHA2 (Fig. 5I). Higher levels of mature SREBP1 in the EPHA2 but not in the EPHA2 (K646M) overexpression model further substantiated the role of EphA2 tyrosine kinase activity in regulating SREBP1 maturation (Fig. 5J). To further dissect whether the transcriptional activation of SREBP1 by the Ephrin-A3/EphA2 axis affected downstream de novo lipogenesis, the total intracellular free fatty acid contents were quantified and found to be significantly lower in EPHA2 KD HCC cells (Fig. 5K). This was further supported by liver orthotopic tumours, where EPHA2-overexpressing HCC cells showed more oil red O staining compared to empty vector controls and EPHA2 (K646) mutants (Fig. 5L). This indicated that EphA2-mediated fatty acid accumulation depends on the tyrosine kinase activity of EphA2 in vivo. We further assessed if SREBP1 acted as a functional mediator of the Ephrin-A3/EphA2 axis. Indeed, Ephrin-A3-induced sphere-forming ability was abolished upon small-interfering (si)SREBP1 treatment (Fig. 5M and Fig. S5C). Taken together, it is evident that the Ephrin-A3/EphA2 axis mediated the maturation of SREBP1 to transcriptionally induce the expression of enzymes involved in de novo lipogenesis and resulting in cancer stemness in HCC cells.
SREBP1-regulated ACLY is a critical downstream driver gene of the Ephrin-A3/EphA2 axis in HCC development
We asked if any specific SREBP1-regulated lipogenic enzyme played a more functional role in tumour initiation or acted as a functional mediator of the hypoxia-regulated Ephrin-A3/EphA2 axis. We first looked at the SREBP1-regulated gene expression profile in hypoxic tumours compared to normoxic tumours from the TCGA-LIHC and HKU-QMH cohorts. ACLY is the only lipogenic enzyme with a significantly higher expression in hypoxic tumours compared to normoxic tumours in both HKU-QMH RNA-sequencing and TCGA-LIHC datasets (Fig. 6A and 6B). ACLY also positively correlated with the expression of EFNA3 in the TCGA-LIHC cohort (Fig. 6C,D). A significant positive correlation between ACLY and EPHA2 was also observed in hypoxic HCCs (Fig. S6A). ACLY expression is also significantly higher in HCC tumours with higher EFNA3 and EPHA2 expression (Fig. S6B). Liver orthotopic xenografts of MHCC97L showed higher ACLY expression in EPHA2 but not in EPHA2 (K646M) models (Fig. 6E). These results suggest that the hypoxia-induced Ephrin-A3/EphA2 axis may regulate ACLY expression in HCC.
Fig. 6ACLY is a critical SREBP1 target gene under the regulation of Ephrin-A3/EphA2 axis.
(A) Volcano plot of SREBP1 target gene expression in hypoxic HCCs relative to median normoxic condition in TCGA-LIHC and HKU-QMH cohorts. (B) Relative ACLY expression in hypoxic vs. normoxic HCCs in TCGA-LIHC and HKU-QMH cohorts. (C) Correlation analysis summary between SREBP1 target genes and EFNA3 in TCGA-LIHC cohort. (D) Scatter plots of ACLY vs. EFNA3 expression levels in HCCs in TCGA-LIHC cohort. (E) IHC of ACLY in liver orthotopic MHCC97L tumours upon OE of EPHA2 or EPHA2 (K646M). Scale bars: 500 μm (F) Summary of expression differences of SREBP1 target genes in subpopulations enriched in LCSC markers, EpCAM, CD24 and CD13, from patients’ tumours by FACS (G) ACLY expression in sorted LCSC marker (EpCAM, CD13 and CD24)-high vs. -low populations in HKU-QMH cohort. (H) ACLY expression in HCCs and NTLs in TCGA-LIHC and HKU-QMH cohorts. (I) Kaplan-Meier plot for overall survival of patients with HCC and high or low expression of ACLY in HCCs of TCGA-LIHC cohort. Student’s t test was used for comparing two groups, Pearsons and Spearman's rank was used to test correlation between two groups, and Wilcoxon's matched-pairs signed rank was used to compare paired clinical samples (∗p < 0.05, ∗∗p <0.01, ∗∗∗p <0.001, ∗∗∗∗p <0.0001). EV, empty vector control; HKU-QMH, The University of Hong Kong-Queen Mary Hospital; NTL, non-tumourous liver; T, tumour; TCGA-LIHC, The Cancer Genome Atlas-Liver Hepatocellular Carcinoma. (This figure appears in color on the web.)
To further examine if ACLY is critical in mediating the tumour-initiating effect of the Ephrin-A3/EphA2 axis, the expression of ACLY was examined in HCC subpopulations and was found to be higher in liver cancer stem-cell marker-enriched subpopulations (epithelial cellular adhesion molecule [EpCAM], CD13 and CD24) (Fig. 6F,G). Moreover, ACLY was significantly overexpressed in HCC in TCGA-LIHC and HKU-QMH cohorts (Fig. 6H), and its higher expression correlated with poorer overall survival of patients (Fig. 6I). These data suggest that ACLY plays a critical role in promoting the tumour-forming ability and cancer stemness of HCC cells.
The Ephrin-A3/EphA2 axis regulates HCC stemness via SREBP1-ACLY metabolic rewiring
Our data suggests that the Ephrin-A3/EphA2 axis can drive self-renewal and tumour initiation of HCC cells, and SREBP1 and ACLY are potential critical downstream drivers of this phenotype. We then asked if the Ephrin-A3/EphA2 axis participated in metabolic rewiring by regulating ACLY expression in HCC. To address this, we generated stable ACLY KD clones (Fig. S7A) to perform targeted metabolomics and assessed their metabolic profile overlap with that of EPHA2 KD cells. ACLY was previously shown to regulate the conversion of citrate into acetyl-Coenzyme A (acetyl-CoA) which is the main substrate for de novo lipogenesis and cholesterol biosynthesis.
Phosphorylation of recombinant human ATP:citrate lyase by cAMP-dependent protein kinase abolishes homotropic allosteric regulation of the enzyme by citrate and increases the enzyme activity. Allosteric activation of ATP:citrate lyase by phosphorylated sugars.
Stable KD of EPHA2 or ACLY resulted in a reduction of acetyl-CoA (Fig. 7A,B), products of de novo lipogenesis (i.e. saturated fatty acids including capric acid [C10:0], lauric acid [C12:0], myristic acid [C14:0], palmitic acid [C16:0] – assessed by targeted metabolomics and lipidomics [Fig. 7C,D]), and total cholesterol levels (measured by Filipin staining [Fig. 7E,F]). Furthermore, quantification of polar metabolites revealed an acute accumulation of citric acid for both ACLY and EPHA2 KD cells. Other metabolites involved in the tricarboxylic acid (TCA) cycle such as isocitric acid, fumaric acid and malic acid demonstrated a significant overlapping signature of EPHA2 and ACLY KD cells (Fig. 7G,H). Besides, higher intracellular reactive oxygen species (ROS) levels, a consequence of higher TCA cycle flux likely due to accumulation of citric acid, was also detected by ROS assays in both ACLY and EPHA2 KD cell models (Fig. 7I,J). However, metabolites further away from the pathway involving citric acid or acetyl-CoA, such as unsaturated fatty acids (Fig. S7B,C), show limited overlap between the 2 types of KD cells. The extent of metabolic signature overlap of EPHA2 KD and ACLY KD suggests that ACLY expression significantly contributes to the metabolic reprogramming caused by the Ephrin-A3/EphA2 axis in HCC.
Fig. 7Ephrin-A3/EphA2 axis exerts its function via ACLY-mediated metabolic reprogramming.
Relative abundance of acetyl-CoA upon knockdown of (A) ACLY and (B) EPHA2 in Huh7 cells by targeted metabolomics. Relative abundance of saturated fatty acids upon KD of (C) ACLY and (D) EPHA2 in Huh7 cells by targeted lipidomics. Total cholesterol levels upon KD of (E) ACLY and (F) EPHA2 in Huh7 by Filipin staining. Scale bars: 52.44 μm. Relative abundance of metabolites in TCA cycle upon KD of (G) ACLY and (H) EPHA2 in Huh7 by targeted metabolomics. Intracellular ROS levels by DCFDA/H2DCFDA staining upon KD of (I) ACLY and (J) EPHA2 in Huh7. (K) Sphere formation assay of shACLY and NTC with or without recombinant Ephrin-A3 supplementation in Huh7 and PLC/PRF/5 cells. (L) Limiting dilution assay of Huh7 cells upon KD of ACLY in NOD/SCID mice. Student’s t test was used for comparing two groups and extreme limiting dilution analysis (ELDA) was used for limiting dilution assay (∗p <0.05, ∗∗p <0.01, ∗∗∗p <0.001, ∗∗∗∗p <0.0001). KD, knockdown; NTC, non-target control; ROS, reactive oxygen species; TCA, tricarboxylic acid. (This figure appears in color on the web.)
ACLY KD cells were further assessed to test if the Ephrin-A3/EphA2 axis regulated HCC stemness through ACLY-mediated metabolic rewiring. Stable KD of ACLY caused a significant reduction in the sphere-forming ability of HCC cells and completely abolished the Ephrin-A3-mediated induction of self-renewal in HCC cells (Fig. 7K). ACLY KD cells also demonstrated suppressed proliferation rates (Fig. S7D,E). A limiting dilution assay further confirmed that KD of ACLY resulted in a significantly lower tumour-initiating capacity of HCC cells in vivo with a 3.4-fold reduction in “stem-cell frequency” (p = 0.0046) (Fig. 7L). Taken together, these data suggest that SREBP1-ACLY metabolic rewiring pathways act as a mediator of the self-renewal and tumour-initiating response induced by the Ephrin-A3/EphA2 axis.
Discussion
The present study demonstrated the role of the Ephrin-A3/EphA2 axis in HCC development and identified it as a clinically relevant and hypoxia-sensitive axis in regulating the tumour-initiating properties and cancer stemness of HCC. Ephrin-A3 is upregulated by hypoxia, in a HIF-1α-dependent manner, and CNV in human HCC samples. Frequent overexpression of EFNA3 in HCC tumours significantly associates with poorer overall survival outcomes and has been shown to play pleiotropic functional tumour-promoting roles in HCC, suggesting that it plays a causal role in driving poorer survival. Importantly, we identified EphA2 as the corresponding effector receptor to Ephrin-A3 or hypoxia stimulation in inducing self-renewal and tumour-initiating ability in HCC cells. Activation of Ephrin-A3/EphA2 forward-signalling upregulated the expression of ACLY via SREBP1 maturation, altering the metabolic profile of cells and consequentially leading to the induction of self-renewal and cancer stemness properties.
The presence of hypoxic niches in solid tumours has a strong predictive link with poorer clinical features and survival outcomes.
Clinical observation of heterogeneity goes hand-in-hand with the framework of cancer stemness, where “stem-like” tumour cells are posited to act as sources of heterogeneity.
We have recently demonstrated that HCC stemness is a plastic phenotype and hypoxia is a cell-extrinsic factor that can drive cancer stemness via HIF-1α.
We identified EFNA3 to be a critical component within the underlying alterations during this phenotypic change. We found HIF-1α directly transcribes EFNA3 mRNA via an HRE sequence upstream of the EFNA3 TSS in HCCs. Since others had previously demonstrated that hypoxia-sensitive micro RNAs and long non-coding RNAs also indirectly regulated the expression of EFNA3, it ought to be an important factor in the context of hypoxia.
Other Ephrin family members, which have considerable functional overlap with EFNA3, i.e. EFNA4 and EFNA5, also correlate with poorer overall survival of patients in TCGA-LIHC cohort (data not shown).
However, these members were not identified in our integrative analysis, and this indicates that they may be responder genes to stimuli other than hypoxia.
Since EFNA3, EFNA4 and EFNA5 are all linked with poorer survival outcomes in patients with HCC, we reasoned that the overlapping biological features are more likely to be conserved at a protein level. As Ephrins interact with various members of the Eph receptor family, Eph receptors are an important context in determining their net functional outcome. We systematically identified EphA2 as the direct functional mediator of Ephrin-A3. EphA2 is reported to have ligand-dependent and -independent modes of signalling which are associated with different functional outcomes.
Moreover, prior to this study, EphA2 has not been carefully characterised in the context of intratumoural hypoxia. Herein, we report that within the hypoxic niches of HCC tumours, Ephrin-A3 and EphA2 play a unidirectional pro-tumourigenic role, where ligand-stimulated activation of the EphA2 tyrosine kinase and its downstream cascade result in hypoxia-induced cancer stemness in HCC. Although not clinically relevant to hypoxic HCC, Ephrin-A1 could also induce similar phenotypic changes in HCC cells (Fig. S9).
Our transcriptomic approach identified activation of SREBP1 transcriptional activity as one of the most significant downstream consequences of the Ephrin-A3/EphA2 axis. Activation of STAT3 (signal transducer and activator of transcription 3) was also found to contribute upon testing pro-tumourigenic pathways reported downstream of EphA2 (Fig. S10).
This was shown to mediate the promoting effect on self-renewal and cancer stemness in HCC. Only ACLY, a transcriptional target of SREBP1, was consistently upregulated in hypoxic HCC tumours and cancer stemness marker (EpCam, CD13 and CD24)-enriched subpopulations. Due to the unique position of ACLY within metabolic networks, i.e. by converting cytoplasmic citrate into acetyl-CoA, it regulates de novo lipogenesis, cholesterol biosynthesis, TCA cycle, and intracellular ROS. Previously, lipid metabolism has been implicated in the modulation of HCC stemness.
Our data suggests that modulation of the Ephrin-A3/EphA2 axis, which subsequently alters the expression of ACLY, affects the synthesis of fatty acids and cholesterol, and changes the intracellular ROS levels. It indicates that these metabolic changes are important regulatory factors in determining the cancer stemness of HCC cells. The nature of interaction among these metabolic pathways in the context of hypoxia and cancer stemness will provide further insights into key factors that determine the “stem-like” state of tumour cells.
Our study provides mechanistic insights into the interplay between HIF-1α and SREBP1, known master regulators of cellular phenotypes. We demonstrated the critical role of the Ephrin-A3/EphA2 axis in this interplay, how it acts as a responder to low oxygen stimulus, utilises SREBP1-mediated ACLY transcription to promote metabolic reprogramming and consequentially induces higher self-renewal and tumour-initiating capacity in HCC cells, which results in poorer survival outcomes in patients with HCC.
Abbreviations
ACACA, acetyl-CoA carboxylase; ACLY, ATP citrate lyase; CA9, carbonic anhydrase 9; CNV, copy number variation; EFNA, Ephrin-A; EpCam, epithelial cellular adhesion molecule; EPHA, Eph receptor A; FASN, fatty acid synthase; FDR, false discovery rate; HCC, hepatocellular carcinoma; HIF, hypoxia-inducible factor; HKU-QMH, The University of Hong Kong-Queen Mary Hospital; HRE, hypoxia response element; KD, knockdown; ROS, reactive oxygen species; RTK, receptor tyrosine kinase; SCD, stearoyl-CoA desaturase; sh, short-hairpin; si, small-interfering; SREBP, sterol regulatory element-binding protein; TCA, tricarboxylic acid; TCGA-LIHC, The Cancer Genome Atlas-Liver Hepatocellular Carcinoma; TSS, transcription start site.
Financial support
The study was supported by Theme-based Research Scheme (T12-704/16-R), Hong Kong Health and Medical Research Fund (06172886), Innovation and Technology Commission grant for State Key Laboratory of Liver Research, University Development Fund of The University of Hong Kong, Seed Fund for Basic Research of The University of Hong Kong (201611159075), and Loke Yew Endowed Professorship award. I.O.L. Ng is Loke Yew Professor in Pathology.
Authors' contributions
A.H., Y.T.C and I.O.N. provided the study concept and design. A.H., Y.T.C. and I.O.N. interpreted and analyzed the data. A.H, Y.T.C, K.M.F.S, Y.M.T., E.M.S.S., V.X.Z, L.K.C, E.L., C.C.L.W. and C.Y.S.C. performed the experiments. E.L., J.M.F.L and T.T.C. collected the patients' samples. A.H., Y.T.C. and I.O.N. wrote the manuscript. All authors approved the final version of the manuscript.
The authors declare no conflicts of interest that pertain to this work.
Please refer to the accompanying ICMJE disclosure forms for further details.
Acknowledgements
The authors would like to thank for the assistance and technical support provided by the Genomics Core, Bioinformatics Core and the Imaging and Flow Cytometry Core of the Center for PanorOmic Sciences at The University of Hong Kong, Hong Kong.
Supplementary data
The following are the supplementary data to this article:
Phosphorylation of recombinant human ATP:citrate lyase by cAMP-dependent protein kinase abolishes homotropic allosteric regulation of the enzyme by citrate and increases the enzyme activity. Allosteric activation of ATP:citrate lyase by phosphorylated sugars.
00125-8&id=ga1.jpg){kind=link}
00125-8&id=gr1.jpg){kind=link}
00125-8&id=gr2.jpg){kind=link}
00125-8&id=gr3.jpg){kind=link}
00125-8&id=gr4.jpg){kind=link}
00125-8&id=gr5.jpg){kind=link}
00125-8&id=gr6.jpg){kind=link}
00125-8&id=gr7.jpg){kind=link}