Determination of HIF-1α degradation pathways via modulation of the propionyl mark

Abstract

The hypoxia-inducible factor-1α (HIF-1α) is a key regulator of hypoxic stress under physiological and pathological conditions. HIF-1α protein stability is tightly regulated by the ubiquitin-proteasome system (UPS) and autophagy in normoxia, hypoxia, and the tumor environment to mediate the hypoxic response. However, the mechanisms of how the UPS and autophagy interplay for HIF-1α proteostasis remain unclear. Here, we found a HIF-1α species propionylated at lysine (K) 709 by p300/CREB binding protein (CBP). HIF-1α stability and the choice of degradation pathway were affected by HIF-1α propionylation. K709-propionylation prevented HIF-1α from degradation through the UPS, while activated chaperon-mediated autophagy (CMA) induced the degradation of propionylated and nonpropionylated HIF-1α. CMA contributed to HIF-1α degradation in both normoxia and hypoxia. Furthermore, the pan-cancer analysis showed that CMA had a significant positive correlation with the hypoxic signatures, whereas SIRT1, responsible for K709-depropionylation correlated negatively with them. Altogether, our results revealed a novel mechanism of HIF-1α distribution into two different degradation pathways.

INTRODUCTION

Hypoxia (Hx) is observed in most tumor types, where the heterodimeric hypoxia-inducible factors (HIFs) play a key role in regulating cellular responses and adaptation to hypoxic stress (1). Under hypoxic conditions, HIF-1α forms a dimer with HIF-1β in the nucleus. HIF-1α/β binds to the hypoxia response elements (HREs) of target genes, recruits p300/CBP histone acetyltransferases, and regulates expression of various genes involved in metabolism, angiogenesis, metastasis, invasion, and proliferation of cancer cells (2-5).

HIF-1α protein levels are tightly regulated via the ubiquitin-proteasome system (UPS). The mechanisms of HIF-1α degradation through the UPS have been extensively studied. Under normoxic conditions, the UPS rapidly degrades HIF-1α. Prolyl hydroxylases (PHDs) hydroxylate HIF-1α at proline 402 and 564 (6, 7). Then, the von Hippel-Lindau (VHL)/Elongin-B/C E3 ubiquitin ligase complex polyubiquitinates hydroxylated HIF-1α, leading to the 26S proteasome-mediated degradation (8, 9).

Recently, the role of autophagy in regulating HIF-1α stability has emerged. Autophagy is subdivided into three types: macroautophagy (MA), chaperon-mediated autophagy (CMA), and microautophagy (10, 11). Among them, CMA is characterized by the selective transfer of target proteins into the lysosomal lumen without vesicle formation, where HSC70 plays a key role as a chaperone by recognizing the KFERQ-motif of substrates (12). HIF-1α contains the N529EFKL533-motif, a noncanonical KFERQ-motif, through which HSC70 binds to HIF-1α. Furthermore, CHIP/STUB1, but not VHL, functions as an E3 ubiquitin ligase via lysine (K) 63-linked polyubiquitination during CMA (13-15). Since HIF-1α is important for cancer progression, elucidating its proteostasis and modulation of HIF-1α levels through different degradation pathways will provide an attractive path for developing cancer treatments.

Post-translational modifications (PTMs) regulate the stability, activity, interactions, and location of proteins, provide diversity in the function of protein targets, and play important roles in various pathways (16). Ubiquitination, hydroxylation, phosphorylation, acetylation, and methylation are PTMs that play important roles in HIF-1α stability and transcriptional activity (5). Although various PTMs and their functions have been extensively studied, PTMs that specifically link to each degradation pathway of autophagy and the UPS have been insufficiently investigated.

This study aimed to elucidate the relationship between the UPS and CMA in HIF-1α proteostasis and to discover a new PTM involved in the triage of degradation pathways. We present a regulatory mechanism for the degradation pathways based on the K709-propionylation status.

RESULTS

HIF-1α is propionylated at K709 by p300

As p300-dependent PTM plays a key role in numerous biological pathways (16), we hypothesized that p300 might be involved in marking HIF-1α for different degradation fates in the UPS or CMA. In regards of recent studies showing that p300 catalyzes acylation as well as acetylation (17, 18), we attempted to identify a novel acyl modification of HIF-1α implicated in the HIF-1α degradation pathways. As an initial step, HEK293T cells were co-transfected with HA-tagged HIF-1α and p300. After immunoprecipitation (IP) of HIF-1α, we explored any p300-dependent HIF-1α modification by using antibodies specific to various pan-K-acyl groups. We showed that p300 caused HIF-1α acetylation, consistent with the previous findings (Supplementary Fig. 1A) (19). Additionally, we found that p300 induced propionylation, but not butyrylation, crotonylation, or succinylation (Fig. 1A). HIF-1α propionylation was p300-specific because co-expression of PCAF or GCN5 did not induce it (Fig. 1B and Supplementary Fig. 1B). We further confirmed that HIF-1α propionylation relied on the catalytic activity of p300 (Supplementary Fig. 1C). To identify the potential K residues propionylated by p300, we focused on K674 and K709, two highly conserved residues among species (Supplementary Fig. 1D). Previous reports have shown that p300 and PCAF acetylate K709 and K674, respectively (19, 20). By the overexpression of HA-HIF-1α with alanine substitution, we showed that p300 preferentially propionylated K709 of HIF-1α (Fig. 1C).

Fig. 1.

HIF-1α is propionylated at K709 by p300. (A) IP with anti-HIF-1α antibody from HEK293T cells expressing HA-HIF-1α and HA-p300. WB with anti-pan-K-acyl antibodies. (B) IP with an anti-HIF-1α antibody from HEK293T cells expressing HA-HIF-1α and either HA-p300 or Myc-GCN5. WB with anti-K-propionyl antibody. (C, D) HEK293T cells were transfected with HA-HIF-1α WT, K674A, or K709A together with Myc-p300 or empty vector. IP was performed with anti-HA antibody. (E) IP with anti K709-propionyl antibody. Lamin B1 was used as a loading control.

To verify this, we generated K709-propionyl-specific HIF-1α antibody (Supplementary Fig. 1E). We IPed HIF-1α using HA-tag antibody, followed by western blot (WB) with the K709-propionyl antibody. K709-propionylation was detected in WT and K674A, but not in the K709A (Fig. 1D). Additionally, when we performed a reverse IP using a K709-propionyl-specific antibody, WT HIF-1α, but not K709A, was IPed (Fig. 1E). These data strongly support our observation that HIF-1α was selectively propionylated on K709 by p300. Collectively, we identified p300-mediated K709-propionylation of HIF-1α as a novel PTM.

K709-propionylated HIF-1α is transcriptionally active and preserved from degradation

To investigate the propionylation of endogenous HIF-1α, we exposed HeLa cells to hypoxia and performed IP to show that K709-propionylation was strongly induced in HIF-1α stabilized by hypoxia (Fig. 2A). To investigate whether propionyl-modified HIF-1α is transcriptionally active, we performed chromatin immunoprecipitation (ChIP) using HIF-1α- or K709-propionyl-antibodies and analyzed the HIF-1α occupancy at target genes, such as SLC2A1, PDK, and VEGFA. Propionylated HIF-1α was highly associated to the HRE promoters in a hypoxia-dependent manner (Fig. 2B), indicating that propionylation occurs to HIF-1α with transcriptional competence. We also tested HIF-1α in normoxia. HeLa cells were treated with DMOG (an inhibitor of PHDs), Baf A1 (an inhibitor of CMA/autophagy), or MG132 (a proteasomal inhibitor) to induce HIF-1α. These inhibitors did not alter the HIF-1α mRNA levels (Supplementary Fig. 2A). Interestingly, K709 propionylation was detected in HIF-1α stabilized by DMOG or Baf A1, but not by MG132 (Fig. 2C). Propionylated HIF-1α was enriched in nucleus (Supplementary Fig. 2B, C) and associated with target genes upon Baf A1 but not MG132 (Fig. 2D). Further, our RT-qPCR analysis showed that the occupancy of propionylated HIF-1α was linked to the transcriptional output (Supplementary Fig. 2D). These data indicate that HIF-1α normally channeled into the UPS is not propionylated and transcriptionally inactive, while the one channeled into CMA is propionylated and transcriptionally active in normoxia.

Fig. 2.

K709-propionylated HIF-1α is transcriptionally active and preserved from degradation. (A, C) WB for the K709-propionyl levels under hypoxia (Hx, 1% O₂ for 4 h) or normoxia (Nx) with DMOG (1 mM), Baf A1 (400 nM), or MG132 (25 μM) for 4 h. Lamin B1 and α-Tubulin are loading controls. (B, D) Chromatin was IPed with rabbit IgG, HIF-1α, or K709-propionyl antibodies and precipitates were analyzed by RT-qPCR. (E, F) HeLa cells were treated with control or EP300-specific siRNAs for 48 h and then exposed to 1% O₂ for 4 h (Hx). All data in the graphs are presented as the means ± SD (n = 3). *P < 0.05; **P < 0.01; ***P < 0.001, and ns; not significant.

Next, we aimed to elucidate the role of p300-mediated K709-propionylation in the regulation of HIF-1α stability. To this end, we used specific siRNAs to deplete p300 and confirmed that knockdown (KD) of p300 did not affect HIF-1α mRNA levels in either normoxia or hypoxia (Supplementary Fig. 3A). However, p300 depletion caused a significant reduction in HIF-1α protein and propionylation levels, accompanied by decreased target gene expression (Fig. 2E, F). KD of CBP, a closely related p300 homologue, with specific siRNAs also caused HIF-1α destabilization in hypoxia without impacting its mRNA levels (Supplementary Fig. 3B-D). These data indicate that p300/CBP is necessary to maintain the HIF-1α propionylation and its protein levels.

CMA is efficient for propionylated HIF-1α degradation and important for HIF-1α proteostasis under hypoxia

To investigate the importance of CMA for the hypoxic response in various cancers, we analyzed RNA-seq datasets from 33 different cancer types in The Cancer Genome Atlas (TCGA). Patient samples were clustered into high- and low-hypoxia tumor groups based on the expression of 45 selected HIF-1α target genes (hereafter referred to as hypoxia signatures), involved in metabolism, angiogenesis, invasion, and metastasis (3-5) (Supplementary Fig. 4A). GSEA was then performed to show that both the UPS and autophagy were associated with hypoxia signatures (Supplementary Fig. 4B-D). In particular, autophagy was related to hypoxia in 76% of cancers, indicating that it largely affects hypoxic adaptation and cancer progression.

CMA is suggested to regulate HIF-1α within a negative feedback loop during hypoxia (13). In this regard, we analyzed CMA across various cancers by exploring the correlation between CMA-related genes (e.g., LAMP2, HSPA8, HSP90AB1, GFAP, and EEF1A1) and hypoxia signatures using TCGA datasets. As expected, CMA correlated positively with hypoxia signatures in most cancers, which was in contrast to VHL (Fig. 3A). In addition, CMA, but not macroautophagy (MA), showed a positive correlation with hypoxia signatures (Fig. 3B). Indeed, expression levels of LAMP2A mRNA and protein, a CMA marker, were elevated upon hypoxia (Supplementary Fig. 5A). These results support previous reports that CMA is activated and involved in HIF-1α proteostasis to secure the hypoxic tumor progression (13, 21, 22). In an additional support of CMA activation, HIF-1α was more efficiently stabilized by Baf A1 than MG132 (Fig. 3C), whereas it was destabilized by aminonicotinamide (6-AN), a CMA activator, and resulted in the suppression of target genes, without altering HIF1A mRNAs (Supplementary Fig. 5B, C). Again, the protein and propionylation levels of HIF-1α affected by 6-AN was efficiently blocked by Baf A1 (Fig. 3D and Supplementary Fig. 5D), indicating that CMA is the prevailing degradation pathway operating under hypoxia and efficient for targeting propionylated HIF-1α.

Fig. 3.

CMA is efficient for propionylated HIF-1α degradation and important for HIF-1α proteostasis under hypoxia. (A) Pearson correlation coefficients and the corresponding P-values were calculated between hypoxia signatures and VHL or CMA-genes. (B) Correlation of the hypoxia signature levels with CMA or macroautophagy (MA) genes. n = 2773 for each group. ***P < 0.001, and ns; not significant. (C, D) HeLa cells were treated with Baf A1 (400 nM), MG132 (25 μM) for 4 h or 6-AN (50 μM) for 24 h with hypoxia (1% O₂ for last 4 h). p62 and LC3B are autophagy markers, while p53 is proteasomal inhibition marker. (E) HEK293T cells were co-transfected with HA-HIF-1α WT or K709A/Q/R and Flag-CHIP for 48 h.

Next, we asked whether CMA targets HIF-1α in a propionyl-dependent manner. CHIP/STUB1 is a key component of CMA and is required for ubiquitination and degradation of HIF-1α (14). To examine the effect of the propionyl mark on CHIP/STUB1 interaction, we co-expressed HA-HIF-1α WT and K709 mutants with Flag-CHIP/STUB1 in HEK293T cells, and IP was performed. Our data showed that CHIP/STUB1 efficiently interacted with all types of HIF-1α (Fig. 3E), indicating that K709-propionylation is dispensable for CHIP/STUB1 interaction.

Collectively, our data demonstrate that multiple degradation pathways operate for the proteostasis of HIF-1α, where the UPS preferentially targets nonpropionylated HIF-1α, and CMA targets both propionylated and nonpropionylated HIF-1α.

SIRT1 mediates HIF-1α depropionylation and correlates negatively with the hypoxia signatures

Sirtuins remove acyl and acetyl groups from lysines of histones and nonhistone substrates (23, 24). To find a sirtuin, particularly implicated in HIF-1α regulation in a hypoxic context, we performed an unbiased analysis of the sirtuins’ effects on the overall survival of patients and their hypoxia-dependent prognosis was compared. Interestingly, Kaplan-Meier curves illustrated that only SIRT1 strongly depended on hypoxia with a survival benefit (Fig. 4A and Supplementary Fig. 6). Indeed, SIRT1 correlated negatively with hypoxia signatures in many cancers (Fig. 4B).

Fig. 4.

SIRT1 mediates HIF-1α depropionylation and correlates negatively with the hypoxia signatures. (A) Kaplan-Meier survival curve of patients with cancers of high- or low-hypoxia signatures according to SIRT1 levels. (B) Pearson correlation coefficients and the corresponding P-values were calculated between the hypoxia signatures and SIRT1 or CMA-genes. (C) HeLa cells were transfected with Flag-SIRT1 or empty vector for 48 h and then exposed to 1% O₂ for 4 h with or without MG132 (25 μM). (D) HeLa cells were treated with a control or SIRT1-specific siRNAs for 48 h and then exposed to 1% O₂ or DMOG (1 mM) for 4 h. (E) Comparison of hypoxia signature levels in the SIRT1-high vs. SIRT1-low groups. n = 2773 for each. ***P < 0.001.

Next, to address any implication of SIRT1 in K709-depropionylation, we transfected HeLa cells with Flag-SIRT1, -SIRT2, or -SIRT6 and found that SIRT1 significantly reduced K709-propionylation of HIF-1α stabilized by DMOG (Supplementary Fig. 7A). To verify this under hypoxia, we transiently expressed Flag-SIRT1 and showed that SIRT1 significantly downregulated both HIF-1α protein and K709-propionylation under hypoxia and the protein, but not propionyl level, was restored by MG132 (Fig. 4C), indicating that depropionylated HIF-1α was degraded through the UPS. Whereas, SIRT1 KD led to a significant increase of K709-propionylation under hypoxia or DMOG treatment (Fig. 4D). Unexpectedly, but curiously, SIRT1 KD did not result in further induction of HIF-1α protein (Fig. 4D and Supplementary Fig. 7B), indicating the possibility of HIF-1α destabilization via alternative pathways in SIRT1 defective condition. Additionally, we showed that the hypoxia signatures were significantly lower in high-SIRT1 tumor groups, demonstrating that SIRT1 is efficient to suppress cancer progression via targeting HIF-1α (Fig. 4E). In agreement, the Kaplan-Meier survival curves also indicated that patients with high-SIRT1 had better survival in the high-CMA than the low-CMA groups (Supplementary Fig. 7C). Altogether, these results demonstrate that SIRT1 is involved in modulation of HIF-1α K709-propionylation and correlates negatively with the hypoxia signatures.

DISCUSSION

This study demonstrates that the degradation route of HIF-1α is modulated by the opposing activities of p300 and SIRT1 toward K709 propionylation. HIF-1α can exist in multiple species according to PTM codes that either affects HIF-1α’s stability or transcriptional activity (16, 19, 20, 25-27). Previously, K709 was reported to be acetylated by p300, which was involved in reducing ubiquitination and increasing HIF-1α stability and activity (19). Here, we showed K709 was additionally propionylated by p300/CBP, and further detailed the role of K709 modification for leveraging the destruction pathways through the UPS or CMA, operating in both normoxia and hypoxia. HIF-1α without the K709 propionyl mark was preferentially cleared by the UPS. Whereas, CMA mediated efficient degradation of HIF-1α with a propionyl mark, which becomes particularly important under hypoxia, where the propionylated HIF-1α is prevalent (Supplementary Fig. 7D).

The UPS and autophagy are often collaborative (28). Similarly, our data propose an interplay between the UPS and CMA for HIF-1α degradation in a propionylation-sensitive way. In addition to SIRT1, SIRT2 was known to deacetylate K709 and promote HIF-1α ubiquitination via enhanced PHD binding (29). In both cases K709 modification is likely to regulate HIF-1α degradation through the UPS and CMA. Of note, SIRT1 over-expression caused depropionylation and enhanced destruction of HIF-1α under hypoxia, but a dramatic induction of HIF-1α was not observed upon SIRT1 KD, suggesting that HIF-1α proteostasis is more complicated than expected. For example, lack of SIRT1 leads to accumulation of ubiquitinated proteins in cells (30) and there exist PHD/VHL-independent mechanisms such as HIF-1α methylation by SET7/9 or ubiquitination by MDM2 for the UPS-mediated HIF-1α control (26, 31).

Protein acylation occurs through nucleocytosolic acyl-CoAs. We showed that HIF-1α is propionylated as well as acetylated. This could be partly related to the relative abundance of intracellular acyl-CoAs. Indeed, it was shown that acetyl-CoA is most abundant in HeLa cells, followed by propionyl-CoA (32). Interestingly, large amounts of acetyl-CoA in tumors are rapidly consumed to increase energy utilization and lipid metabolism. Particularly, hypoxic tumors deplete acetyl-CoAs very quickly by up-regulation of fatty acid synthase (FASN) and fatty acid synthesis for tumor growth and survival (33-35). Furthermore, a recent study has shown that the nuclear level of propionyl-CoA was substantially induced by hypoxia, up to approximately equimolar to acetyl-CoA (36). Therefore, increased propionyl-CoA/acetyl-CoA balance might be partially exploited to induce HIF-1α propionylation under hypoxia as a major PTM implicated in the dynamic modulation of HIF-1α stability. As K709 could be both propionylated and acetylated by p300, their common and distinct roles, additional proteins involved in deacetylation and depropionylation, and any physiological condition or cancer type preferred to each modification should be investigated in the future.

In conclusion, we identified a novel modification of HIF-1α, K709-propionylation, which was involved in the UPS and CMA triage choice for HIF-1α proteostasis in both normoxia and hypoxia. These results expand our understanding of how the PTM codes dynamically regulate HIF-1α and hypoxic downstream targets.

MATERIALS AND METHODS

Cell culture

HeLa and HEK293T cells were obtained from the American Type Culture Collection and cultured in Dulbecco’s modified Eagle medium (DMEM; WELGENE) with 10% fetal bovine serum (FBS; Gibco) and 1% penicillin/streptomycin (PS; WELGENE) at 37°C in a 5% CO₂ incubator. For the hypoxic condition, cells were incubated in a hypoxia chamber (1% O₂, 5% CO₂, and balanced N₂) at 37°C.

Transfection and quantitative real-time PCR (RT-qPCR)

Cells were transfected with plasmids using FuGENE HD (Promega) or Lipofectamine 2000 (Invitrogen) following the manufacturer’s protocol. For RNA interference, HeLa cells were transfected with either non-targeting siRNAs or specific siRNAs using Lipofectamine RNAiMAX (Invitrogen). RT-qPCR was performed with specific primers, using THUNDERBIRD SYBR qPCR Master Mix (TOYOBO) and CFX96 (Bio-Rad Laboratories) real-time PCR detector. The siRNA and primer sequences are listed in Supplementary Table 1 and 2.

Western blot (WB) and immunoprecipitation (IP)

Cells were washed with cold PBS and lysed with NP-40 lysis buffer (50 mM Tris HCl pH 7.4, 150 mM NaCl, 5 mM EDTA, 1% NP-40) containing protease inhibitors. The samples were analyzed by SDS-PAGE. After electrophoresis, proteins were transferred to a nitrocellulose membrane, which was incubated with a blocking solution consisting of Tris-buffered saline containing 0.1% Tween 20 and 5% nonfat dry milk at room temperature for 1 h, followed by incubation with primary antibodies at 4°C overnight. HRP-conjugated secondary antibodies (Bethyl Laboratories) were used. The signal was developed using a WB detection kit (AbFrontier).

Statistical analyses

Means and standard deviations were analyzed using GraphPad Prism (v8). P-values were calculated based on a two-tailed, unpaired Student’s t-test. The difference between means was considered significant for P < 0.05. All experiments were performed three to five times.

Supplementary materials and methods

Plasmid constructs, biochemical fractionation, antibodies including K709-propionyl-specific antibody, chemicals, ChIP, and TCGA data analysis are described in the supplementary information.

Funding Statement

ACKNOWLEDGEMENTS This work was supported by National Research Foundation of Korea (NRF) grants funded by the Korean government (MSIT) (NRF-2022R1A2C2003505, 2021R1F1A1049941, and NRF-2019 R1A5A2027340 to E.-J.C., NRF-2022R1C1C2005612 to J.C., and NRF2022R1A5A102641311 to H.-D.Y.).