Epigenetic control via histone methylation is important in transcriptional regulation and occurs in the nucleus. However, this process can be regulated spatially by a protein localized in the Golgi apparatus. Subcellular compartmentalization can therefore affect epigenetic modification.
Histone methylation at lysine residues plays important roles in many biological processes such as heterochromatin formation, X-chromosome inactivation, transcriptional regulation and cancer formation [1,2]. Methylation of histone 3 at lysine 4 (H3K4) can occur in monomethylated, dimethylated or trimethylated forms [3–5]. H3K4 trimethylation (H3K4me3) is a prevalent mark that is exclusively associated with actively transcribed genes such as HOX genes that are essential for vertebrate development .
The first identified H3K4 methyltransferase is the large macromolecular complex of proteins associated with Set1 (COMPASS) from Saccharomyces cerevisiae and the SET domain-containing protein Set1 is the catalytic subunit in the complex [7–9]. Mixed-lineage leukaemia (MLL) is predicted to be a functional homologue of Set1 in mammalian cells and was found in COMPASS-like complexes capable of methylating H3K4 in metazoan cells [10,11]. Translocations of MLL gene are associated with a variety of aggressive acute leukaemia in humans [2,10]. In fact, humans encode at least six different Set1 family members: MLL1–MLL4, SET1A and SET1B . In addition to the catalytic subunit, COMPASS complexes require a number of regulatory subunits to achieve full histone methyltransferase activity. Four regulatory subunits are shared between the metazoan COMPASS-like complexes and the yeast COMPASS complexes, including ASH2 (absent, small or homoeotic discs 2; related to Cps60 in yeast), WDR5 (WD40 repeat-containing protein 5; related to Cps30 in yeast), RBBP5 (retinoblastoma-binding protein 5; related to Cps50 in yeast) and DPY30 (related to Cps25 in yeast) . These four critical regulatory subunits form a complex coined the WRAD sub-complex (WDR5–ASH2–RBBP5–DPY30) . The WRAD sub-complex is critical for the full enzyme activity of H3K4 methyltransferase. It was reported that the minimal H3K4 methyltransferase complex required for di- and tri-methylation of H3K4 in mammalian cells includes MLL1, ASH2L, WDR5, and RBBP5 . In the absence of any of the core regulatory members such as ASH2L, WDR5 or RBBP5, the methyltransferase activity of MLL is severely lessened . WDR5 plays a critical role in the WARD sub-complex as it is required for binding of the COMPASS-like complexes to the K4-dimethylated H3 tail and essential for global H3K4me3 and HOX gene activation in human cells . In addition, WDR5 is required to maintain MLL complex integrity, including the stability of ASH2L within the complex . Recently, DPY30 was found to directly regulate H3K4 methylation in both in vitro and in vivo studies .
H3K4 methylation is a prevalent marker for transcriptionally activated genes and is implicated in developmental processes and human diseases such as cancers . A genome-wide study has revealed that active promoters are marked by trimethylation of H3K4 and enhancers are marked by monomethylation, but not trimethylation of H3K4. Such study has led to a prediction that there are over 200 promoters and 400 enhancers within the 30-Mbp region in human genome . Using a single-molecule-based sequencing technology for high-throughput profiling of histone modifications in mammalian cells, it was reported that trimethylation of H3K4 and H3K27 effectively discriminates genes that are either expressed or repressed, reflecting cell state and lineage potential . During carcinogenesis, cancer cells usually experience severe hypoxia resulting from reduced oxygen supply from blood vessels and hypoxia activates the major transcription factor hypoxia-inducible factor 1 (HIF-1), as well as other important transcription factors that are actively involved in tumour progression and metastasis. Interestingly, it was recently found that hypoxia-mediated regulation of H3K4 methylation is implicated in hypoxia-induced epithelial–mesenchymal transition (EMT) that is important for metastasis and cancer stemness . Hypoxia could induce the expression of WDR5 and histone deacetylase 3 (HDAC3) and both of them synergistically facilitate H3K4 methylation, leading to activation of mesenchymal gene expression and EMT features in tumour cells. In addition, hypoxia may alter H3K4 methylation via changes in demethylase activities. Hypoxia could induce H3K4me3 via inhibition of JARID1A demethylase in cancer cells . The histone demethylase LSD1 (lysine-specific demethylase 1) was also found to be implicated in modulation of H3K4 methylation and silencing of a DNA mismatch repair gene MLH1 in tumour cells .
PAQR3 (progestin and adipoQ receptors member 3) is a member of the PAQR family which are predicted to have seven transmembrane domains similar to G-protein-coupled receptors (GPCRs) . PAQR1 and PAQR2 function as cell-surface receptors for adiponectin, an adipocyte-secreted cytokine that regulates glucose and lipid metabolism . Although PAQR3 shares high sequence homology with PAQR1 and PAQR2, it is exclusively localized in the membrane of the Golgi apparatus rather than the plasma membrane . PAQR3 was named RKTG (Raf kinase trapping to Golgi) due to its spatial regulation of Raf kinase . Our previous studies have revealed that PAQR3 functions as a tumour suppressor due to its inhibitory activity on Raf–MEK (mitogen-activated protein kinase/ERK kinase)–ERK (extracellular-signal-regulated kinase) signalling [25–29]. Subsequent characterization of PAQR3 indicates that it also regulates GPCR signalling pathways by sequestering the Gβ-subunit to the Golgi apparatus . In addition, PAQR3 is markedly down-regulated in human gastric cancers. The expression level of PAQR3 is closely associated with the progression and metastasis of gastric cancers, indicating that PAQR3 is a new genetic signature that can predict the prognosis of the patients with gastric cancer . In the present study, we provide evidence for a unique role of PAQR3 in the regulation of H3K4 methylation via sequestration of the core regulatory subunits of the COMPASS-like complexes in mammalian cells.
MATERIALS AND METHODS
The Myc-tagged PAQR3 and GFPc1-tagged PAQR3 were described previously . The full-length human ASH2, hDPY30, RBBP5 and WDR5 cDNA were cloned by reverse transcription (RT)-PCR from total RNA isolated from human cell lines and confirmed by sequencing. These cDNAs were fused with a FLAG–epitope tag at the N-termini in p3XFlag–CMV10 plasmid (Sigma–Aldrich).
Cell culture, transfection and hypoxic treatment
Human embryonic kidney (HEK) 293T and HeLa cells were cultured in Dulbecco's modified Eagle's medium (DMEM) containing 10% FBS. AGS cells were cultured in F-12K medium containing 10% FBS. Transient transfection was performed with the polyethyleneimine (PEI) method for HEK293T and Polyjet (Invitrogen) for HeLa cells according to the manufacturer's instructions. Hypoxia was achieved by culturing the AGS cells for 24 h in an airtight chamber containing 1% oxygen, 5% carbon dioxide and 94% nitrogen using a protocol previously described .
Antibodies, immunoprecipitation and immunoblotting
The antibodies were purchased as follows: antibody against Myctag was from Santa Cruz Biotechnology; antibody against FLAG-tag was from Sigma–Aldrich; antibodies against mono-, di- and tri-methylated histone H3K4 were from Cell Signaling Technology, antibody against histone H3 was from Beyotime Institute of Biotechnology; antibodies against Golgin-97, Alexa Fluor 488-conjugated donkey anti-mouse IgG, Alexa Fluor 546-conjugated goat anti-mouse, rabbit IgG and Hoechst 33342 were from Molecular Probes; Cy5-labelled goat anti-mouse IgG was from GE Healthcare. The cells were lysed in a buffer containing 20 mM Tris/HCl (pH 7.5), 150 mM NaCl, 5 mM EGTA and 1% Nonidet P-40 with a mixture of protease inhibitors (Sigma–Aldrich) before immunoprecipitation and immunoblotting assays using protocols as previously described .
ChIP and PCR analysis
ChIP was performed using the Chromatin Immunoprecipitation Assay Kit (Upstate Biotechnology) and the anti-H3K4me3 antibody (9751) from Cell Signaling Technology. Primers used for real-time PCR detection are listed as follows: 5′-GTGCT-GGGTGCTAAGGT-3′ and 5′-CAGACGCCAGGATCCCA-3′ for HOXC8 (promoter region −600 to −400); 5′-CCAGG-CGGGTAGCTGTG-3′ and 5′-GTTTGGTTGGCGAGCTC-3′ for HOXA9 (promoter region −800 to −600).
Analysis of HOX gene expression
Total RNA was isolated using TRIzol reagent from Invitrogen. The RNA was treated with RNase-free DNase I and reverse-transcribed with oligo(dT) primer using the SuperScript First-Strand Synthesis System for RT–PCR (Invitrogen). Quantitative real-time PCR was done with the SYBR Green PCR system (Applied Biosystems), using actin as an internal control for normalization. Primers used for each gene are listed as follows: 5′-GATGGCATTGGATTATGCAG-3′ and 5′-AAGCACGGTGA-TCAGGTACA-3′ for PAQR3, 5′-GATCATTGCTCCTCCTG-AGC-3′ and 5′-ACTCCTGCTTGCTGATCCAC-3′ for actin, 5′-AGGAACCTGATGGAAACCTG-3′ and 5′-AAACAGCGA-AGGAGAGGAAG-3′ for HOXC8, 5′-CATCGATCCCAAT-AACCCA-3′ and 5′-CCCTGGTGAGGTACATGTTG-3′ for HOXA9.
The confocal analyses with overexpression experiments were performed as previously reported . The Golgi localization was determined by immunostaining with antibodies against either Golgin-97 or GM130, two well-characterized Golgi markers.
Preparation of nuclear extracts and H3K4 methyltransferase activity assay
Nuclear extracts were prepared from HEK293T cells 48 h after transfection by using EpiQuik™ Nuclear Extraction Kit I (Epigentek) and stored at −80°C until being analysed. The nuclear extracts were used to analyse H3K4 methyltransferase activity using EpiQuik™ Histone Methyltransferase Activity/Inhibition Assay Kit (H3-K4; Epigentek) according to the manufacturer's instructions and detected by a microplate reader at 450 nm.
The unpaired and two-tailed Student's t test is used for all the statistical analyses.
PAQR3 is able to regulate trimethylation of H3K4
To address whether PAQR3 had a global effect on the steady-state level of H3K4 methylation, HEK293T cells were transfected with various amounts of Myc-tagged PAQR3. Whole-cell extracts were prepared 3 days after the transfection and analysed by immunoblotting with antibodies against different forms of H3K4 methylation. The global levels of H3K4 methylation were significantly reduced in PAQR3-overexpressing cells, shown as an apparent decrease in mono- and tri-methylation of H3K4 (Figure 1A). However, the global level of H3K4 dimethylation (H3K4me2) was unaffected (Figure 1A).
PAQR3 negatively regulates trimethylation of H3K4
We next employed confocal analysis to investigate H3K4me3 at the single-cell level. HeLa cells were transfected with GFP-fused PAQR3 and subjected to immunofluorescence analysis to detect H3K4me3 levels. The cells with PAQR3 overexpression had a reduced level of H3K4me3 compared with adjacent cells without PAQR3 overexpression (Figures 1B and 1D). We also investigated the H3K4me3 level under the condition of PAQR3 down-regulation. Knockdown of PAQR3 was performed by a specific shRNA that had a high efficiency to silence the expression of endogenous PAQR3 . The cells with down-regulation of PAQR3 had a profound increase in H3K4me3 signals (Figures 1C and 1D). Collectively, these results indicate that PAQR3 is able to negatively regulate H3K4me3.
PAQR3 modulates hypoxia-induced H3K4 trimethylation
To further investigate the potential function of PAQR3 on H3K4me3, we analysed whether PAQR3 could alter H3K4me3 under hypoxia as a previous study has revealed that hypoxia was able to elevate the cellular H3K4me3 level . In AGS gastric cancer cells, the H3K4me3 level was markedly increased by hypoxia (Figures 1E and 1F). Knockdown of PAQR3 could increase the H3K4me3 level under normoxia and further facilitate hypoxia-induced H3K4me3 (Figure 1E). In contrast, overexpression of PAQR3 reduced the H3K4me3 level under normoxia and abolished the increment of H3K4me3 under hypoxia (Figure 1F). These data therefore indicate that PAQR3 could regulate hypoxia-induced H3K4me3 in cancer cells.
PAQR3 regulates HOX gene expression
The transcription of the HOX gene family has been well characterized to be tightly regulated by H3K4 methylation during embryonic development . We examined whether HOX gene expression could also be regulated by PAQR3. Quantitative RT–PCR was performed with HEK293T cells under the condition of PAQR3 overexpression or PAQR3 down-regulation. We found that the mRNA levels of HOXC8 and HOXA9 were significantly reduced by PAQR3 overexpression (Figure 2A). In contrast, the mRNA levels of these two genes were significantly elevated by PAQR3 knockdown (Figure 2B).
PAQR3 regulates HOXC8 and HOXA9 gene expression and alters the H3K4me3 level in the promoter region
Next, we used a ChIP assay to directly analyse the status of H3K4me3 at the promoters of HOXC8 and HOXA9 genes. Immunoprecipitated genomic DNA was subjected to PCR using probes specific for HOXC8 and HOXA9 genes. As shown in Figure 2(C), PAQR3 overexpression resulted in an obvious decrease in H3K4me3 at the HOXC8 promoter, whereas there was a slight reduction in H3K4me3 at the HOXA9 promoter (Figure 2C). Collectively, our results demonstrate that PAQR3 has a regulatory effect on the expression of HOX genes that are controlled by H3K4 methylation.
Interaction of PAQR3 with the core regulatory subunits of COMPASS-like complexes
The metazoan COMPASS-like complexes are composed of a SET-domain containing catalytic subunit (including MLL1–MLL4, SET1A, SET1B) and four core regulatory subunits including WDR5, ASH2, RBBP5 and DPY30 (the WRAD sub-complex). Our earlier studies indicated that PAQR3 was able to interact with the Gβ-subunit of GPCRs . As Gβ is a WD40-containing protein, we postulated that PAQR3 might also interact with other WD40-containing proteins. Within the WRAD sub-complex, there are two WD40-containing proteins, WDR5 and RBBP5. We used co-immunoprecipitation assays to determine the potential interaction of PAQR3 with WDR5 and RBBP5. When both PAQR3 and WDR5 were overexpressed, immunoprecipitation of PAQR3 could pull down WDR5 (Figure 3A). Meanwhile, when both PAQR3 and RBBP5 were overexpressed, immunoprecipitation of RBBP5 could pull down PAQR3 (Figure 3B). As expected, the WD40 domains of both WDR5 and RBBP5 could also interact with PAQR3 by co-immunoprecipitation assays (results not shown).
Interaction of PAQR3 with WDR5, RBBP5, ASH2 and DPY30
We also analysed the potential interaction of PAQR3 with two other subunits of the WRAD sub-complex. To our surprise, PAQR3 could also interact with ASH2 by the co-immunoprecipitation assay (Figure 3C), but did not bind DPY30 (Figure 3D). However, in the presence of overexpressed WDR5, immunoprecipitation of PAQR3 could pull down DPY30 (Figure 3E), indicating that PAQR3 might indirectly interact with DPY30 through WDR5. Collectively, these data indicate that PAQR3 could interact with the four core regulatory subunits of the COMPASS-like complexes directly or indirectly in mammalian cells.
PAQR3 tethers the WRAD sub-complex to the Golgi apparatus
As PAQR3 is a membrane protein specifically localized in the Golgi apparatus , we next investigated whether PAQR3 could change the subcellular localization of the subunits of the COMPASS-like complexes. When WDR5, RBBP5, ASH2 and DPY30 were overexpressed in HeLa cells, all of them were localized in the nucleus (Figures 4A–4D). Intriguingly, when co-expressed with PAQR3, the four regulatory subunits were almost completely mobilized to the Golgi apparatus (Figures 4A–4D). In contrast, overexpression of PAQR3 had no effect on the distribution of endogenous MLL1, one of the catalytic subunits of COMPASS-like complexes (Figure 4E). These data therefore indicate that PAQR3 could tether the WRAD sub-complex, but not MLL1, to the Golgi apparatus, leading to reduced distribution these subunits in the nucleus.
PAQR3 tethers members of WRAD sub-complex to the Golgi apparatus
PAQR3 inhibits H3K4 methyltransferase activity and reduces the interaction of WDR5 with MLL1
Based on our studies, we hypothesized that PAQR3 might negate H3K4 methyltransferase activity by reducing complex formation between the four subunits with the SET domain-containing catalytic subunit. To test this hypothesis, we analysed the effect of PAQR3 on the interaction of WDR5 with the SET domain of MLL1, as WDR5 is the major component that mediates the interaction of the WRAD sub-complex with the catalytic subunit . The full-length MLL1 contains 3969 amino acid residues with a conserved SET domain at the C-terminus together with other conserved domains including a Win (WDR5-interaction) motif [35,36]. Previous studies have shown that the C-terminus of MLL1 (C-ter) is the minimal domain required for the assembly of a full MLL1 core complex in vitro . As measured by a co-immunoprecipitation assay, the C-terminal domain of MLL1 was able to interact with WDR5 (Figure 5A). Interestingly, overexpression of PAQR3 could dose-dependently compete off the binding of WDR5 with the C-ter (Figure 5A).
PAQR3 decreases the interaction between WDR5 with MLL1 and reduces H3K4 methyltransferase activity
As PAQR3 could interfere with the interaction of WDR5 with MLL1, we postulated that such action of PAQR3 should directly inhibit the enzyme activity of H3K4 methyltransferase. We examined the histone methyltransferase activity in HEK293T cells with PAQR3 overexpression. Compared with the control group, PAQR3 overexpression could significantly reduce the enzyme activity of H3K4 methyltransferase (Figure 5B).
Our studies reveal for the first time that H3K4 methylation is regulated by PAQR3 in a spatial manner by alterations of the subcellular compartmentalization of the WRAD sub-complex of the COMPASS methyltransferase (Figure 5C). PAQR3 is a transmembrane protein specifically localized in the Golgi apparatus. It can directly or indirectly interact with the four core regulatory subunits of the COMPASS-like complexes and sequestrate them to the Golgi apparatus (Figures 2 and 3). In the absence of PAQR3 overexpression, the WRAD sub-complex members including ASH2, WDR5, RBBP5 and DPY30 were all localized in the nucleus. However, overexpressed PAQR3 could tether most portions of these four subunits to the Golgi apparatus (Figure 3). As COMPASS-like complexes-mediated H3K4 methylation can only occur in the nucleus, the reduced availability of the WRAD sub-family would render a decrease in the H3K4 methyltransferase activity in the nucleus, leading to reduced level of H3K4me3. Consistently, we found that the H3K4me3 level, HOX gene expression and H3K4 methyltransferase activity are all reduced by PAQR3 overexpression (Figures 1, 2 and 5A). Importantly, we also found that down-regulation of endogenous PAQR3 had an effect on H3K4me3 (Figure 1C). Knockdown of PAQR3 also facilitated hypoxia-induced H3K4me3 (Figures 1D and 1F). These results indicate that the basal level of H3K4me3 is dependent on the expression level of PAQR3 in a particular cell. In other words, the endogenous PAQR3 level may set the tone for the degree of H3K4me3. Collectively, our studies have pinpointed a new mode of regulation on the activity of mammalian COMPASS-like complexes, leading to alterations in H3K4me3 levels in the cells. It is noteworthy that H3K4 monomethylation (H3K3me1) is also affected by PAQR3 (Figure 1A), consistent with a previous report showing that down-regulation of WDR5 and RBBP5 severely reduced the tri- and mono-methylation of H3K4 but with little or no significant effect on H3K4me2 .
In addition to its critical roles on activation of development-related genes such as the HOX gene family, the mammalian COMPASS-like complexes play an important role in cancer formation. It is known that EMT is critical for cancer metastasis. It was recently discovered that hypoxia, a condition that commonly occurs during cancer progression, is able to induce EMT and metastatic phenotypes through modulation of H3K4 methylation . Under hypoxia, HDAC3 is able to interact with hypoxia-induced WDR5 and then recruit the histone methyltransferase complex to increase H3K4 methylation, leading to activation of mesenchymal gene expression. Meanwhile, knockdown of WDR5 could abolish mesenchymal gene activation under hypoxia . In the present study, we found that hypoxia could induce the H3K4me3 level in AGS cancer cells, and PAQR3 knockdown further increased hypoxia-induced H3K4me3 (Figures 1D and 1F). Such findings are consistent with our previous observation that PAQR3 has a negative effect on EMT and tumour metastasis [29,31]. Deletion of the PAQR3 gene in the mice is associated with an increase in EMT features in collaboration with loss of p53 . In human gastric cancers, reduced expression of PAQR3 is highly correlated with increased EMT markers and accelerated tumour progression and metastasis . It is likely that the reduced expression of PAQR3 in human tumours could relieve the inhibitory effect of PAQR3 on H3K4 methylation, leading to facilitation of hypoxia-induced H3K4 methylation and activation of mesenchymal gene expression. Therefore, it is worth investigating the functional roles of PAQR3 regulation on H3K4 methylation and the associated EMT and cancer development in the future.
absent, small or homoeotic discs 2
complex of proteins associated with Set1
C-terminus of MLL1
histone H3 lysine 4
human embryonic kidney
progestin and adipoQ receptors member
retinoblastoma-binding protein 5
WD40 repeat-containing protein 5
Chunchun Liu, Yuxue Zhang, Yongfan Hou, Liqiang Shen, Yinlong Li, Weiwei Guo, Daqian Xu, Gaigai Liu, Zilong Zhao and Kaiyang Man performed the experiments. Yi Pan and Zhenzhen Wang provided critical advice. Yan Chen planned the experiments. Chunchun Liu, Yuxue Zhang and Yan Chen analysed the results and wrote the paper.
This work was supported by the Ministry of Science and Technology of China [grant numbers 2012CB524900 (to Y.C.) and 2013BAI04B03 (to Z.W.)]; and the National Natural Science Foundation of China [grant numbers 81130077, 81390350 and 81321062] to Y.C.
These authors contributed equally to this work.