Arginine methylation of HSP70 regulates retinoid acid-mediated RARβ2 gene activation

Significance

HSP70 proteins are well known as molecular chaperones involved in protein folding and quality control. Whether they also function in gene transcription on chromatin, and if so, how they are regulated, remains elusive. Here we report that HSP70 can also regulate gene transcription through its association with chromatin, distinct from its “classic” function as a molecular chaperone. The function of HSP70 in gene transcription is subject to regulation of an arginine methylation on a highly conserved residue in HSP70, which modulates the recruitment of a key component in the pre-initiation complex, and thus transcription initiation. The present study reveals an additional, previously overlooked function of HSP70 chaperone proteins, and links arginine methylation of nonhistone proteins to gene transcriptional regulation.

Abstract

Although “histone” methyltransferases and demethylases are well established to regulate transcriptional programs and to use nonhistone proteins as substrates, their possible roles in regulation of heat-shock proteins in the nucleus have not been investigated. Here, we report that a highly conserved arginine residue, R469, in HSP70 (heat-shock protein of 70 kDa) proteins, an evolutionarily conserved protein family of ATP-dependent molecular chaperone, was monomethylated (me1), at least partially, by coactivator-associated arginine methyltransferase 1/protein arginine methyltransferase 4 (CARM1/PRMT4) and demethylated by jumonji-domain–containing 6 (JMJD6), both in vitro and in cultured cells. Functional studies revealed that HSP70 could directly regulate retinoid acid (RA)-induced retinoid acid receptor β2 (RARβ2) gene transcription through its binding to chromatin, with R469me1 being essential in this process. HSP70’s function in gene transcriptional regulation appears to be distinct from its protein chaperon activity. R469me1 was shown to mediate the interaction between HSP70 and TFIIH, which involves in RNA polymerase II phosphorylation and thus transcriptional initiation. Our findings expand the repertoire of nonhistone substrates targeted by PRMT4 and JMJD6, and reveal a new function of HSP70 proteins in gene transcription at the chromatin level aside from its classic role in protein folding and quality control.

Nucleosome is the basic unit of chromatin, which consists of ∼146 bp of DNA wrapped around an octamer comprised of two copies of the core histone proteins (H3, H4, H2A, and H2B). A central mechanism for regulating chromatin structure and therefore gene expression is achieved via covalent posttranslational modifications (PTMs) on histone proteins, which include but are not limited to phosphorylation, acetylation, ubiquitylation, sumoylation, and methylation (1, 2). Histone methylation can be found on all basic residues, including arginines, lysines, and histidines (3, 4), which was originally believed to be an irreversible modification that could only be removed by histone eviction or by dilution during DNA replication; however, the discovery of histone demethylase LSD1 has changed this notion, and the subsequent discovery of the Jumonji C (jmjC)-domain–containing demethylases by a number of laboratories supports this conclusion (5, 6). Histone methylation dynamics is known to have important roles in many biological processes, including developmental control, cell-fate decision, aging, cancer, and other diseases (7, 8).

Protein arginine methylation results in the addition of one or two methyl groups to the guanidino nitrogen atoms of arginine (9), which is catalyzed by a family of enzymes called protein arginine methyltransferases (PRMTs). The PRMT family has been shown to include at least 11 members in mammalian genomes, designated as PRMT 1 to 11 based on difference in primary sequences and substrate specificity. Arginine methylation has also been found on hundreds of nonhistone proteins in both the nucleus and the cytoplasm (10, 11). In conjunction with the expanding number of proteins that are known to be arginine methylated, a growing list of biological processes is being shown to involve arginine methylation, which include transcription, RNA processing and transport, translation, signal transduction, DNA repair, and apoptosis, among others (10–14). Arginine methyltransferases have been linked to diseases, including prostate and breast cancers, cardiovascular diseases, viral pathogenesis, and spinal muscular atrophy, suggesting these enzymes might be promising drug targets (15). Arginine methylation has been considered to be an irreversible posttranslational modification until recently. Two groups showed that arginine methylation of histones can be reduced in vitro by deimination with peptidyl-arginine deiminase 4, which converts both unmodified arginine and monomethylarginine, but not dimethylarginine, to citrulline (16, 17). However, peptidyl-arginine deiminases are not real demethylases because citrulline lacks the positively charged guanidine group, and thus it cannot functionally substitute for arginine. Recently, it was shown that a JmjC-domain–containing protein named JMJD6 functions as an arginine demethylase, which specifically demethylates both H3R2me and H4R3me (18). More recently, we reported the dual enzymatic activities of JMJD6 targeting both histone (H4R3me) and RNA (methyl group on the γ-phosphate of 7SK snRNA’s first 5′ nucleotide), which are important for its function in Pol II promoter-proximal pause release of a substantial list of genes (19). Interestingly, JMJD6 also targets to arginine methylation on nonhistone proteins, such as estrogen receptor alpha (ERα) and RNA helicase A (RHA) (20, 21).

HSP70 (heat-shock protein of 70 kDa) proteins constitute an evolutionarily conserved protein family of ATP-dependent molecular chaperones (22). Humans have 13 genes encoding HSP70s, among which some are induced by heat shock and other types of stress, and others are constitutively expressed (23, 24). HSP70s share a domain architecture consisting of an amino (N)-terminal nucleotide-binding domain, followed by a substrate-binding domain, which can be further subdivided into a peptide-binding domain and a carboxyl (C)-terminal “lid” domain (25). At the extreme C-terminal end of most cytoplasmic HSP70 chaperones lies an EEVD motif that is known to interact with certain tetratricopeptide repeat-containing proteins (26, 27). In addition to their function in client protein folding, HSP70 chaperones play an important role during the translocation of a number of proteins, such as nuclear receptors and kinases, to their designated subcellular compartments (28).

Mammalian HSP70 proteins are subject to various posttranslational modifications, including phosphorylation, acetylation, ubiquitination, and methylation (29, 30). Interestingly, despite the fact that lysine and arginine methylation on HSP70 chaperones was first reported in the early 1980s, the exact residues being modified and the methyltransferases catalyzing such modification are only now being identified. A novel family of lysine methyltransferases, including METTL21A, was found to trimethylate multiple HSP70 isoforms on K561, a conserved lysine residue (31). Trimethylation of this residue was reported to affect binding of HSP70 chaperones to their client protein, the Parkinson disease-associated protein α-synuclein (32). A third group reported that the same residue was dimethylated by the SETD1A methyltransferase, which led to the translocation of HSP70 from the cytoplasm to the nucleus, and stimulated the enzymatic activity of Aurora kinase B on histone H3, thus causing proliferation of cancer cells (33).

Recent research efforts have primarily focused on HSP70 lysine methylation, but little is known about its arginine methylation and the functional consequences of such modification. Using anti-methyl-lysine or arginine antibodies combined with proteomics analysis, we have identified several lysine and arginine methylation sites on HSP70. Further investigation of methylated arginine 469 (R469), which was mediated at least partially by PRMT4 in vitro and in vivo, revealed that, in contrast to HSP70 lysine methylation reported to be indispensable for its ATPase activity or client protein binding, the R469 methylation was crucial for HSP70-regulated gene transcription, exemplified by HSP70 binding to the retinoid acid (RA) receptor β2 (RARβ2) gene promoter and inducing RARβ2 gene transcription in a R469 methylation-dependent fashion. Mechanistically, HSP70 proved to be required for the effective recruitment of TFIIH to the RARβ2 gene promoter, and thus transcription initiation through physical interaction with TFIIH occurred in a HSP70 R469 methylation-dependent manner.

Results

HSP70 Is Lysine and Arginine Methylated.

Initially, we sought to identify nonhistone proteins that are targeted for lysine or arginine methylation through immunoprecipitation (IP) with antibody against the protein of interest followed by immunoblotting (IB) with anti-methyl-lysine (Kme) or arginine (Rme) antibody. Consistent with previous reports, it was found that HSP70 was abundantly methylated on both lysine and arginine residues, among the many proteins tested (Fig. 1A). Specifically, cell lysates from HEK293T were subjected to IP using anti-HSP70 antibody followed by IB using either anti-methyl-lysine or arginine antibody. To confirm the methylation signals detected by these antibodies were specific, similar experiments were performed in cells treated with or without adenosine-2′, 3′-dialdehyde (AdOx), a methyltransferase inhibitor. It was found that HSP70 methylation was nearly blocked in the presence of AdOx (Fig. 1B). To further demonstrate the specificity of HSP70 methylation detected, HEK293T cells were transfected with a Flag-tagged HSP70, and then treated with or without AdOx followed by IP with anti-Flag antibody. It was again found that HSP70 methylation was attenuated in the presence of AdOx (Fig. 1C). Alternatively, HSP70 methylation was shown by in vivo labeling with l-[methyl-³H]-methionine in the presence of protein translation inhibitors [cycloheximide/chloramphenicol mixture, no-translation mixture (NTM)], followed by IP with anti-HSP70 antibody. It was found that a 70-kDa band was robustly labeled with ³H, signal of which decreased significantly when AdOx was added (Fig. 1D, compare lane 3 to lane 4). Inhibition of protein translation by NTM was examined through labeling of cells with [³⁵S]-methionine in parallel, finding that radioactive signal disappeared in the presence of NTM (Fig. 1D, compare lane 1 to lane 2). Therefore, the radioactive signal detected with l-[methyl-³H]-methionine labeling was a result of posttranslational methylation instead of residual translational incorporation of l-[methyl-³H]-methionine. Similarly, in vivo labeling experiments were repeated with cells transfected with Flag-tagged HSP70 and followed by IP with anti-Flag antibody, which confirmed that HSP70 indeed was methylated in cultured cells (Fig. 1E).

Fig. 1.

HSP70 methylation in cultured cells. (A) Cell lysates collected from HEK293T cells were subjected to IP with control IgG or anti-HSP70 antibody followed by IB with anti-methyl-lysine (Kme) (Left), antimethyl-arginine (Rme) (Center), or anti-HSP70 (Right) antibody. IgG H.C: IgG heavy chain. (B) HEK293T cells were treated with or without AdOx (20 μM) for 6 h before IP and IB experiments as described in A. (C) HEK293T cells were transfected with Flag-tagged HSP70 and then treated with or without AdOx (20 μM) for 6 h before IP with anti-Flag antibody, followed by IB with anti-Kme, anti-Rme, or anti-Flag antibody. Input was shown on the far left panel. N.S: nonspecific bands. (D) In vivo labeling of HSP70 proteins. HEK293T cells were maintained in DMEM without methionine for 24 h before adding NTM [cycloheximide (100 μg/mL) and chloramphenicol (40 μg/mL)] for another hour, followed by adding [³⁵S]-methionine (20 μCi/mL) (lanes 1 and 2) or l-[methyl-³H]-methionine (20 μCi/mL) (lanes 3 and 4) in the presence or absence of AdOx (20 μM) for 4 h. Cells were then lysed and subjected to IP with anti-HSP70 antibody, followed by SDS/PAGE gel and autoradiogram. (E) HEK293T cells were transfected with Flag-tagged HSP70 before in vivo labeling with l-[methyl-³H]-methionine as described in D. (F) HSP70 proteins purified from HEK293T cells were subjected to Coomassie blue staining (C.B.S). IgG H.C: IgG heavy chain; IgG L.C: IgG light chain. (G and H) Sequence alignment of the region surrounding arginine 469 (boxed) in multiple HSP70 isoforms (Homo sapiens) (G) or in paralogous HSP70 genes in various organisms (H) by using Clustal Omega. Asterisk (*) indicates positions which have fully conserved residue; Colon (:) indicates conservation between groups of strongly similar properties; Period (.) indicates conservation between groups of weakly similar properties. At, Arabidopsis thaliana; Dm, Drosophila melanogaster; Dr, Danio rerio; Hs, Homo sapiens; Mm, Mus musculus; Pf, Plasmodium falciparum; Rn, Rattus norvegicus; Sc, Saccharomyces cerevisiae; Xt, Xenopus (Silurana) tropicalis.

Next, we sought to identify the potential methylation sites in HSP70 protein purified from HEK293T cells by mass spectrometry analysis (Fig. 1F). Consistent with previous reports, tryptic peptides containing methylated lysine 561 were identified (Table 1 and Dataset S1, J and K). Meanwhile, several other lysine (K) and arginine (R) residues were found to be methylated with high confidence as well, with K77 and K526 being mono-, di-, and trimethylated, and K88, K550, and R469 being monomethylated (Table 1 and Dataset S1, A–E and G–I). In addition, K332, a lysine residue unique to HSPA1L among all isoforms of HSP70 protein family, was also targeted for mono-, di-, and trimethylation (Table 1 and Dataset S1F). In particular, R469 was found to be highly conserved among different isoforms of HSP70 proteins (Fig. 1G), and conserved during evolution (Fig. 1H), suggesting that methylation on this residue might be functionally important. In the present study, we have focused on investigating the biological function of monomethylation on R469 (R469me1) in HSP70 protein.

Table 1.

Arginine and lysine methylation sites identified through mass spectrometry analysis

Identified peptides with arginine or lysine methylation	Type of methylation	Position	Hsp70 isoforms containing the modified peptides
R.FELSGIPPAPR+14.G	Rme1	469	HSPA1A, HSPA1B, HSPA6
K.FELTGIPPAPR+14.G	Rme1	469	HSPA8
R.K+14FGDPVVQSDMK.H	Kme1	77	HSPA1A, HSPA1B
R.K+28FGDPVVQSDMK.H	Kme2
R.K+42FGDPVVQSDMK.H	Kme3
K.FGDPVVQSDMK+14.H	Kme1	88	HSPA1A, HSPA1B
R.FDDAVVQSDMK+14.H	Kme1	88	HSPA8
K.AK+14IHDIVLVGGSTR.I	Kme1	332	HSPA1L
K.AK+28IHDIVLVGGSTR.I	Kme2
K.AK+42IHDIVLVGGSTR.I	Kme3
K.YK+14AEDEVQR.E	Kme1	526	HSPA1A, HSPA1B, HSPA1L
K.YK+28AEDEVQR.E	Kme2
K.YK+42AEDEVQR.E	Kme3
K.NALESYAFNMK+14.S	Kme1	550 (HSPA1A, HSPA1B)	HSPA1A, HSPA1B, HSPA1L
		552 (HSPA1L)
K.NSLESYAFNMK+14.A	Kme1	550	HSPA8
K.SAVEDEGLKGK+14.I	Kme1	561	HSPA1A, HSPA1B
K.LQGK+42INDEDKQK.I	Kme3	561	HSPA8

Tryptic peptides containing methylated arginine or lysine residues were shown (+14, monomethylation, Rme1, or Kme1; +28, dimethylation, Kme2; +42, trimethylation, Kme3). HSP70 isoforms containing these tryptic peptides, and the positions of the methylated lysine or arginine were shown as indicated.

HSP70 Is Methylated by CARM1/PRMT4 and Demethylated by JMJD6 at R469.

To elucidate the biological function of methylation on R469 in HSP70 protein, an antibody specifically targeting R469me1 was generated and its specificity was examined in multiple ways: first, total lysates from HEK293T cells were subjected to IB with anti-R469me1 antibody preincubated with Flag-peptide (negative control) or HSP70 peptide with unmodified R469 (R469un) or monomethylated R469 (R469me1). It was found that anti-R469me1 antibody detected a specific band with a molecular weight at around 70 kDa, which presumably corresponded to HSP70 protein (Fig. 2A). Preincubation with peptide containing R469me1, but not unmodified R469, significantly decreased the chemiluminescent signal compared with control, suggesting the anti-R469me1 antibody specifically recognized the monomethylated form of HSP70 (Fig. 2A). Second, total lysates from HEK293T cells transfected with Flag-tagged wild-type or mutant HSP70 with substitution of arginine 469 to alanine (R469A) were subjected to IP with anti-Flag antibody, followed by IB with anti-R469me1 antibody. As expected, mutation of R469 completely abolished the chemiluminescent signals, further confirming the specificity of this antibody toward R469 methylated-HSP70 (Fig. 2B).

Fig. 2.

Methylation and demethylation of arginine 469 in HSP70 by PRMT4 and JMJD6, respectively. (A) Anti-HSP70R469me1 antibody was preincubated with 3XFlag (Left), HSP70R469 unmodified (Center), or HSP70R469me1 (Right) peptide for 30 min before probing HEK293T total cell lysates. (B) HEK293T cells transfected with control vector, Flag-tagged wild-type (WT) or mutant HSP70 (R469A) were subjected to IP with anti-Flag antibody, followed by IB with anti-HSP70R469me1 antibody. (C) HEK293T cells were transfected with control siRNA or siRNA specifically against each individual member of PRMT protein family for 72 h, followed by IB with anti-HSP70R469me1 (Upper) or anti-HSP70 (Lower) antibody. (D) In vitro methylation assay was performed by mixing purified bacterially expressed PRMTs with wild-type (Upper) or mutant HSP70 (R469A) (Lower) proteins. White arrows indicate automethylation of PRMT1, PRMT4, and PRMT6; Black arrows indicate methylation of HSP70. (E) In vitro methylation assay was performed by mixing purified bacterially expressed PRMT4 proteins with R469 or R469A peptide. Increased amount of each reaction was taken for dot blot as indicated and followed by autoradiogram. (F) In vitro demethylation assay was performed by mixing R469me1 peptide with or without purified bacterially expressed JMJD6 protein. Increased amount of each reaction was taken for dot blot as indicated and followed by IB with anti-HSP70R469me1 antibody. (G) HEK293T cells were transfected with control siRNA or siRNA specifically against JMJD6 for 72 h, followed by IB with antibodies as indicated. The ratio of R469me1 levels between the siCTL and siJMJD6 sample was 0.45 quantified using ImageJ.

Next, we sought to identify the PRMT, which might be responsible for modifying this residue in cultured cells by transfecting short interfering RNA (siRNA) specifically targeting each individual member of the PRMT family (PRMT1 to -11), followed by examination of R469me1 levels. It was found that knock-down of either coactivator-associated arginine methyltransferase 1 (CARM1)/PRMT4 or PRMT7 led to a significant reduction of R469me1 levels compared with control sample, whereas other members exerted no apparent effects, suggesting that CARM1/PRMT4 or PRMT7 may methylate R469 (Fig. 2C). The knock-down efficiency of each PRMT was examined through quantitative RT-PCR (RT-qPCR) (Fig. S1). To test which member of the PRMT family can methylate HSP70 directly in vitro, an in vitro methylation assay was performed mixing purified bacterially expressed wild-type or mutant HSP70 (R469A) with individual PRMT (PRMT1 to -11). It was found that PRMT1, -4, -6, and -8 were capable of methylating HSP70 in vitro (Fig. 2D). Consistent with previous reports, automethylation of PRMT1, -4, and -6 was also observed. Interestingly, the methylation of HSP70 mediated by PRMT4 was nearly abolished when arginine 469 was substituted to alanine, suggesting R469 was the primary site methylated by PRMT4 in vitro (Fig. 2D). The expression of wild-type and mutant HSP70 (R469A) was comparable when examined by Coomassie blue staining (Fig. S2A). The expression of all PRMTs was also shown (Fig. S2B). Of note, although members in the PRMT family, except PRMT10 and PRMT11, have been shown to possess methyltransferase activity (10), only PRMT1, -3, -4, -6, -8, and -9 seemed to be active when purified from bacterial cells and histones were serving as substrates under current conditions tested (Fig. S2 C and D). Because many of the histone-modifying enzymes purified from bacterial cells are inactive, Flag-tagged PRMTs were overexpressed and purified from HEK293T cells, followed by in vitro methylation assay using histones, HSP70 (wild-type) or HSP70 (R469A) as substrates. It was found that all PRMTs except PRMT10 were capable of methylating histones, indicating they were active (Fig. S2E). Consistent with what we observed by using PRMT proteins purified from bacterial cells, PRMT1, -4, -6, and -8 purified from overexpressed HEK293T cells were capable of methylating HSP70 in vitro (Fig. S2F). In addition, PRMT7 was capable of methylating HSP70 as well (Fig. S2F). However, the observed methylation signal in the presence of PRMT4 was completely abolished when R469 was substituted to alanine, whereas there was no significant change for PRMT1, -6, -7, and -8 when comparing HSP70 (wild-type) to HSP70 (R469A) (Fig. S2F). The expression of all Flag-tagged PRMTs was shown (Fig. S2G). To further confirm that PRMT4 is capable of methylating HSP70 at R469, short-peptide containing arginine 469 (R469), or with substitution of arginine 469 to alanine (R469A) was incubated with or without PRMT1 or PRMT4, finding that PRMT4 methylated the R469 peptide robustly, but not the R469A peptide (Fig. 2E). Meanwhile, PRMT1 displayed no significant activity toward either R469 or R469A peptide (Fig. S2H). Taking these data together, despite several members in the PRMT protein family being capable of methylating HSP70 in vitro, we find that R469 was the primary site in HSP70 methylated by PRMT4. PRMT4 appeared to be at least partially responsible for R469 methylation in cultured cells. Meanwhile, although it targeted HSP70 at sites other than R469 in vitro, PRMT7 influenced HSP70R469me1 levels in cultured cells. We speculate that there is a cross-talk between PRMT7-mediated methylation and R469 methylation. Alternatively, PRMT7 might also target to HSP70 R469 directly despite it displayed no such activity in vitro, which certainly remains as an intriguing topic for future investigation.

Fig. S1.

Knock-down efficiency of siRNAs targeting PRMTs. Knock-down efficiency of siRNA targeting to each individual member of PRMT protein family as shown in Fig. 2C was examined through RT-qPCR. Of note, the expression level of PRMT8 was too low to be detected in HEK293T cells. Data shown is the relative fold-change, as indicated, compared with control siRNA transfected samples after normalization to actin (±SEM).

Fig. S2.

HSP70 methylation mediated by PRMTs. (A) In vitro purified, bacterially expressed His-tagged wild-type and mutant HSP70 (R469A) were examined by Coomassie blue staining (C.B.S). (B) In vitro purified, bacterially expressed GST-tagged PRMTs were examined by Coomassie blue staining. Of note, PRMT3 was purified separately. (C) Histones purified from HeLa cells (Sigma) were examined by Coomassie blue staining. (D) In vitro methylation assay was performed by mixing PRMTs with histones as described in B and C, respectively. Of note, in vitro methylation assay with PRMT3 was performed separately. (E) In vitro methylation assay was performed by mixing Flag-tagged PRMTs purified from overexpressed HEK293T cells with histones as described in C. (F) In vitro methylation assay was performed by mixing Flag-tagged PRMTs purified from overexpressed HEK293T cells with wild-type (Upper) or mutant HSP70 (R469A) (Lower) proteins. (G) The expression of Flag-tagged PRMTs in HEK293T cells was examined by IB with anti-Flag antibody. Of note, the expression of PRMT11 was low under current conditions tested. (H) In vitro methylation assay was performed by mixing PRMT1 protein as described in B, with R469 or R469A peptide. Increased amount of each reaction was taken for dot blot as indicated and followed by autoradiogram.

JMJD6, which belongs to the JmjC-domain–containing demethylase family, is the only member in this family possessing arginine demethylase activity reported so far. It can demethylate both mono- and di-methylated arginine on both histones and nonhistone proteins (18–21). We tested directly whether HSP70R469me1 might serve as a substrate for JMJD6 by performing in vitro demethylation assay mixing short-peptide containing monomethylated R469 (R469me1), with or without in vitro purified bacterially expressed JMJD6, followed by either immunoblotting with anti-R469me1 antibody (Fig. 2F) or MALDI-TOF MS analysis (Fig. S3). It was found that R469me1 signal decreased dramatically in the presence of JMJD6 based on IB (Fig. 2F). MALDI-TOF MS analysis further revealed that the decrease of R469me1 signal was because of the removal of one methyl-group by JMJD6 (Fig. S3). To demonstrate JMJD6 demethylation of HSP70R469me1 in cultured cells, HEK293T cells were transfected with control siRNA or siRNA specific against JMJD6 followed by IB with anti-R469me1 antibody, finding that knock-down of JMJD6 consistently led to a mild yet significant increase of R469me1 levels (Fig. 2G). Lack of dramatic effects when knocking down of JMJD6 might be because it only functions locally to demethylate HSP70R469me1, or because there are other arginine demethylases yet to be discovered (19).

Fig. S3.

Demethylation of HSP70 R469me1 by JMJD6. In vitro demethylation assay was performed by mixing R469me1 peptide with (Lower) or without (Upper) purified bacterially expressed JMJD6 proteins, followed by MALDI-TOF MS analysis.

R469 Methylated-HSP70 Is Localized in Both Cytoplasm and Nucleus, with an Implication in Nuclear Function.

Most of the molecular chaperones are believed to localize predominantly in the cytoplasm, whereas others are more concentrated in specific cellular compartments, such as endoplasmic reticulum (ER)-resident HSPA5/GRP78, and mitochondrial HSPA9/GRP75 and HSPA14/HSP60 (23, 24, 28). We examined cellular localization of HSP70 through cellular fractionation experiments in three independent cell lines, HEK293T, HeLa, and MCF7, and found that it distributed equally in both cytoplasm and nucleus, whereas other chaperones, such as HSP90 and HSP27, were predominantly localized in the cytoplasm, suggesting that HSP70 might play a distinct role in the nucleus (Fig. 3A and Fig. S4A). Similarly, R469 monomethylated-HSP70 was found in both the cytoplasm and nucleus (Fig. 3A). Poly(ADP-ribose) polymerase 1 (PARP1) and HSP60 served as purity controls of nuclear and cytoplasmic fractions, respectively (Fig. 3A). The localization of HSP70 and its methylated form were further confirmed in the other three cell lines examined (Fig. S4B).

Fig. 3.

HSP70 association with chromatin in gene transcriptional regulation. (A) Cellular fractionation was done in three independent cell lines, HEK293T, HeLa, and MCF7, followed by IB with various antibodies, as indicated. (B) HSP70 ATPase activity assay was performed by mixing PRMT4 (0.04 μg/μL), HSP70 (wild-type or R469A mutant) (0.05 μg/μL), SAM (1 mM), and ATP (100 μM), as indicated. (C) HEK293T cells were treated with or without RA (10⁻⁷ M) for 1 h and then subjected to ChIP with control IgG, anti-HSP70, RXRα, or RARα antibody. Binding of HSP70, RXRα, and RARα on the RARβ2 gene promoter region were examined through qPCR. ChIP signals were presented as fold-enrichment over that of IgG (±SEM, **P < 0.01, ***P < 0.001). (D and E) MCF7 cells were treated with or without E₂ (10⁻⁷ M) for 1 h and then subjected to ChIP with control IgG, anti-HSP70 (D) or ERα (E) antibody. Binding of HSP70 (D) and ERα (E) on TFF1/pS2 or GREB1 gene promoter regions were examined through qPCR (±SEM, *P < 0.05, **P < 0.01). (F and G) HEK293T (F) or MCF7 (G) cells were transfected with control siRNA or siRNA specifically targeting HSP70 for 72 h, followed by treatment with RA (10⁻⁷ M) (F) or E₂ (10⁻⁷ M) (G), respectively, for 6 h before RT-qPCR analysis to examine mRNA levels of genes as indicated. Data shown was the relative fold-change compared with control siRNA transfected samples after normalization to actin (±SEM, *P < 0.05, **P < 0.01, ***P < 0.001). (H) Cell lysates as described in F and G were subjected to IB with antibodies, as indicated. (I) HEK293T cells were pretreated with or without PES (20 μM) or AMI (100 μM) for 1 h before treatment with or without RA (10⁻⁷ M) for 6 h, followed by RT-qPCR analysis to examine the mRNA levels of RARβ2. Data shown is the relative fold-change compared with control sample after normalization to actin (±SEM, ***P < 0.001).

Fig. S4.

Cellular localization of HSP70 and its R469-methylated form. (A) HEK293T cells were subjected to immunostaining with anti-HSP70 antibody. HSP70 proteins are shown in green. (B) Cellular fractionation was done in three independent cell lines, HCT116, SW480, and LNCaP, followed by IB with various antibodies as indicated.

HSP70 facilitates the folding process by transient association of its substrate-binding domain with short hydrophobic peptide segments within their “client proteins.” The client protein binding and release cycle is driven by ATP binding and hydrolysis, which requires the ATPase activity from HSP70 nucleotide-binding domain (28). To assess whether R469me1 plays a functional role in regulating ATP hydrolysis rate, HSP70 ATPase activity was assayed in the presence or absence of PRMT4. Notably, it was found that ATP hydrolysis was not affected by adding PRMT4, suggesting that R469me1 was not involved in modulating HSP70 ATPase activity (Fig. 3B). Meanwhile, ATPase activity of mutant HSP70 (R469A) was very similar to that of wild-type HSP70 (Fig. 3B). Considering its unique nuclear localization compared with most other molecular chaperones, and its presence in a wide range of protein complexes involved in histone modifications and chromatin remodeling, we sought to examine whether HSP70, and hence R469 monomethylation, might exert a functional role in gene transcriptional regulation at the chromatin level.

HSP70 Associates with Chromatin and Regulates Gene Transcription.

The binding of HSP70 with promoter regions of target genes, such as RA-induced RARβ2 gene in HEK293T cells and homeotic (Hox) genes in NTera2 cells, 17β-Estradiol (estrogen, E2)-induced trefoil factor 1 precursor (TFF1/pS2) and growth regulation by estrogen in breast cancer 1 (GREB1) genes in MCF7 cells, androgen [5α-Dihydrotestosterone (DHT)]-induced prostate specific antigen (PSA/KLK3) and NK3 homeobox 1 (NKX3.1) genes in LNCaP cells, and TNF-α–induced IL6 and IL8 genes in HeLa cells, was examined using ChIP assay. Interestingly, RA-dependent recruitment of HSP70 to RARβ2 and selected Hox, and E₂-dependent recruitment to TFF1/pS2 and GREB1 gene promoters were observed, whereas HSP70 displayed no binding upon DHT or TNF-α treatment (Fig. 3C, Left, Fig. 3D, and Fig. S5A). The recruitment of HSP70 to the RARβ2 gene promoter was accompanied with that of retinoid X receptor α (RXRα) and RARα, the two nuclear receptors that drive the activation of RARβ2 gene (Fig. 3C, Center and Right). Similarly, the recruitment of HSP70 to TFF1/pS2 and GREB1 gene promoter regions was accompanied with that of ERα, which mediates the activation of these genes (Fig. 3E). These data suggested that HSP70 might involve in gene transcriptional activation, such as RARβ2, Hox, and E₂-target genes. Indeed, knock-down of HSP70 attenuated RA-induced RARβ2 and selected Hox, and E₂-induced TFF1/pS2 and GREB1 gene activation (Fig. 3 F and G and Fig. S5B). Consistent with the lack of HSP70 binding on promoter regions, the induction of DHT- or TNF-α target genes was not affected by HSP70 knock-down (Fig. S5 C and D), indicating that knock-down of HSP70 did not induce a nonspecific defect in gene transcription. The knock-down efficiency of HSP70 was shown by qPCR and IB with primers and antibody specifically targeting HSP70, respectively (Fig. 3 F–H and Fig. S5). To determine whether regulation of transcription by HSP70 is coupled with its ATPase activity, a key component of HSP70 protein chaperone function, we measured the effects of 2-phenylethynesulfonamide (PES), a HSP70 chaperone inhibitor (34), on RA-induced RARβ2 gene activation. It was found that PES exhibited no apparent effects, suggesting that HSP70’s function in gene transcription is distinguished from its role in protein folding and quality control (Fig. 3I, the four columns on the left).

Fig. S5.

Transcriptional regulation by HSP70. (A) NTera2 cells were treated with or without RA (10⁻⁷ M) for 1 h and then subjected to ChIP with control IgG or anti-HSP70 antibody. Binding of HSP70 on selected Hox gene promoter regions as indicated was examined through qPCR. ChIP signals were presented as fold-enrichment over that of IgG (±SEM, **P < 0.01, ***P < 0.001). (B) NTera2 cells were infected with control shRNA or shRNA specifically targeting HSP70 lentiviral particles for 72 h, followed by treatment with or without RA (10⁻⁷ M) for 6 h before RT-qPCR analysis to examine mRNA levels of genes as indicated. Data shown was the relative fold-change compared with control shRNA transfected samples after normalization to actin (±SEM, *P < 0.05, **P < 0.01, ***P < 0.001). (C and D) LNCaP (C) or HeLa (D) cells were transfected with control siRNA or siRNA specifically targeting HSP70 for 72 h, followed by treatment with or without DHT (10⁻⁷ M) (C) or TNF-α (20 ng/mL) (D), respectively, for 6 h before RT-qPCR analysis to examine mRNA levels of genes as indicated. Data shown was the relative fold-change compared with control siRNA-transfected samples after normalization to actin (±SEM). (E) Knock-down efficiency of HSP70 shRNA or siHSP70 as described in B–D were examined through IB, as indicated.

To further confirm the functional significance of HSP70 in RA-induced RARβ2 gene activation, wild-type (hsp70.1^+/+) and hsp70.1-deficient (hsp70.1^−/−) mouse embryonic fibroblasts (MEFs) were treated with or without RA for 4, 12 or 24 h, followed by examining RARβ2 expression level. Surprisingly, RA-induced RARβ2 expression was comparable in wild-type and hsp70.1^−/− MEFs at 4 h of treatment, with the induction in hsp70.1^−/− MEFs a little weaker than that in wild-type (Fig. 4A). However, further activation of RARβ2 with RA treatment for 12 or 24 h was observed in a time-dependent manner in wild-type MEFs, which was not detectable in hsp70.1^−/− MEFs (Fig. 4A). We reasoned that the other members of the hsp70 protein family might compensate for the loss of hsp70.1 initially, whereas full activation of RARβ2 required the presence of hsp70.1. Because all four RARβ isoforms are inducible by RA (35), we examined their expression by using a primer set common to all, with similar effects as seen for RARβ2 observed when hsp70.1 was not present (Fig. 4B). The RAR family consists of three receptor subtypes, α, β, and γ, which exhibit functional redundancy (36). We then tested if hsp70 is required for RARα and RARγ gene expression, finding that the expression of both receptors was induced by RA to a much less extent compared with RARβ, which was abrogated in hsp70.1^−/− MEFs (Fig. 4 C and D). Finally, the functional significance of hsp70 in RA-induced RARβ2 gene activation was demonstrated by transfecting hsp70.1^−/− MEFs with a control vector or vectors expressing wild-type HSP70 followed by treatment with or without RA. In accordance with its role in RAR gene activation, overexpression of wild-type HSP70 led to a dramatic increase in the fold-induction by RA for all RAR receptor subtypes (Fig. 4 E–H, four columns on the left).

Fig. 4.

HSP70 is required for RA-induced RAR gene transcriptional activation. (A–D) Wild-type (hsp70.1^+/+) and hsp70-deficient (hsp70.1^−/−) MEFs were treated with or without RA (10⁻⁷ M) for 4, 12, and 24 h, followed by RT-qPCR analysis to examine the mRNA levels of RARβ2 (A), RARβ (B), RARα (C), and RARγ (D). Data shown was the relative fold-change compared with control sample in wild-type MEFs after normalization to actin (±SEM, *P < 0.05, **P < 0.01, ***P < 0.001). (E–H) hsp70.1^−/− MEFs were transfected with control vector, Flag-tagged wild-type or mutant HSP70 (R469A) for 48 h and then treated with or without RA for 4 h, followed by RT-qPCR analysis to examine the mRNA levels of RARβ2 (E), RARβ (F), RARα (G), and RARγ (H) (±SEM, **P < 0.01, ***P < 0.001). (I) Expression of wild-type and mutant HSP70 (R469A) as described in E–H was examined through IB with anti-Flag antibody. GAPDH was served as loading control.

HSP70R469me1 Is Involved in RA-Induced RARβ2 Gene Activation.

Next, using RA-induced RARβ2 gene activation as a model, we began to test whether PRMT4-mediated R469me1 on HSP70 might play a functional role in gene transcription. First, arginine N-methyltransferase inhibitor (AMI), a general PRMT inhibitor, was shown to inhibit RARβ2 promoter- and retinoid acid response element (RARE)-driven luciferase activity, as well as RARβ2 gene expression, in the presence of RA, suggesting that PRMT-mediated arginine methylation might be important for RA-induced RARβ2 promoter activity, and hence RARβ2 gene activation (Fig. 3I, the two columns on the right, and Fig. 5 A and B). To test whether HSP70 R469 is involved in RA-induced RARβ2 gene activation, HEK293T cells were transfected with control siRNA or siRNA specifically targeting HSP70 in the presence or absence of wild-type or mutant HSP70 (R469A), followed by treatment with or without RA. It was found that wild-type, but not mutant HSP70 (R469A), rescued the activation of RARβ2 gene despite they bound to the RARβ2 gene promoter equally well, indicating HSP70 R469 is important for RARβ2 gene activation (Fig. 5C and Fig. S6). As shown above, overexpression of wild-type HSP70 in hsp70.1^−/− MEFs led to a dramatic increase in the fold-induction by RA for all RAR receptor subtypes (Fig. 4 E–H, four columns on the left), whereas the R469A mutant displayed no such effects (Fig. 4 E–H, two columns on the right). The expression of wild-type and mutant HSP70 (R469A) was comparable, as examined by IB (Fig. 4I).

Fig. 5.

R469me1 is involved in RA-induced RARβ2 gene transcriptional activation. (A and B) HEK293T cells were transfected with luciferase reporter construct containing retinoid acid response element (βRARE-luc) (A) or mouse RARβ2 gene promoter sequence [RARβ2(P)-luc] (B) for 24 h, and pretreated with or without AMI (100 μM) for 1 h before RA treatment (10⁻⁷ M) for another 12 h, followed by luciferase reporter activity measurement (±SEM, ***P < 0.001). (C) HEK293T cells were transfected with control siRNA or siRNA specifically against HSP70 together with or without control vector, wild-type, or mutant HSP70 (R469A) before RA (10⁻⁷ M) treatment for 6 h, followed by RT-qPCR analysis to examine mRNA levels of RARβ2 (±SEM, *P < 0.05, ***P < 0.001). It should be noted that siRNA against HSP70 here was targeting to the 5′UTR region of HSP70. (D) HEK293T cells were treated with or without RA (10⁻⁷ M) for 1 h and then subjected to ChIP with control IgG or anti-PRMT4 antibody. Binding of PRMT4 on the RARβ2 gene promoter region was examined through qPCR (±SEM, ***P < 0.001). (E and I) HEK293T cells were transfected with control siRNA or siRNA specifically against PRMT4 (E) or JMJD6 (I) for 24 h before transfection with luciferase reporter construct, βRARE-luc (Left) or RARβ2(P)-luc (Right) for another 48 h, followed by RA treatment for 12 h. Cells were then subjected to luciferase reporter activity measurement (±SEM, ***P < 0.001). (F) HEK293T cells were transfected with control siRNA or siRNA specifically against PRMT4 together with or without control vector, wild-type or enzymatically dead mutant (E267Q) PRMT4 (mouse) before RA (10⁻⁷ M) treatment for 6 h, followed by RT-qPCR analysis to examine mRNA levels of RARβ2 (±SEM, *P < 0.05). (G) HEK293T cells were transfected with control or PRMT4 expression vector together with or without luciferase reporter constructs, βRARE-luc (Left) or RARβ2(P)-luc (Right), for 48 h, followed by RA treatment for 12 h. Cells were then subjected to luciferase reporter activity measurement (±SEM, ***P < 0.001). (H) HEK293T cells were transfected with control or PRMT4 expression vector for 48 h, followed by treatment with RA (10⁻⁷ M) for 6 h before RT-qPCR analysis to examine mRNA levels of RARβ2 (±SEM, ***P < 0.001). (J) HEK293T cells were transfected with control siRNA or siRNA specifically targeting JMJD6 for 72 h, followed by treatment with RA (10⁻⁷ M) for 6 h before RT-qPCR analysis to examine mRNA levels of RARβ2 (±SEM, **P < 0.01).

Fig. S6.

Wild-type or mutant HSP70 (R469A) binding with chromatin. HEK293T cells transfected with Flag-tagged wild-type or mutant HSP70 (R469A) were treated with or without RA (10⁻⁷ M) for 1 h before ChIP with IgG or anti-Flag antibody. Binding of HSP70 on RARβ2 gene promoter region was examined through qPCR (±SEM, **P < 0.01, ***P < 0.001).

Consistently, PRMT4, the arginine methyltransferase capable of catalyzing R469me1, was found to increase significantly on RARβ2 promoter region upon RA treatment when examined by ChIP in HEK293T cells (Fig. 5D). Furthermore, knock-down of PRMT4 attenuated RARβ2 promoter- and RARE-driven luciferase activity as well as RARβ2 gene expression in the presence of RA (Fig. 5 E and F). More importantly, introducing a vector-expressing wild-type mouse PRMT4, but not its enzymatically dead mutant (E267Q), rescued the expression of RARβ2, suggesting that PRMT4 enzymatic activity was crucial for RA-induced RARβ2 expression (Fig. 5F). PRMT4 levels were examined by RT-qPCR (Fig. S7A). Conversely, RARβ2 promoter- and RARE-driven luciferase activity, as well as RARβ2 gene expression, was further stimulated when cells were transfected with a PRMT4 expression vector (Fig. 5 G and H). Taken together, these data suggested that PRMT4 is required for RA-induced RARβ2 gene activation, and is consistent with the fact that PRMT4 is capable of methylating HSP70 R469.

Fig. S7.

The expression levels of PRMT4 and JMJD6 examined by RT-qPCR. The expression levels of PRMT4 (A) and JMJD6 (B) in samples described in Fig. 5 F and J, respectively, were examined through RT-qPCR.

We previously reported that JMJD6 is capable of demethylating mono- and di-methylated arginines, both symmetrical (s) and asymmetrical (a), and that it functions in both gene activation and repression (19). Particularly, it was found that JMJD6 regulates transcriptional pause release, and thus gene activation partially through its demethylase activity toward H4R3me2(s), a repressive mark associated with gene repression. However, the role of JMJD6 in gene repression and the potential substrates, through which it works, remain to be explored. As JMJD6 demethylase activity toward HSP70R469me1 was observed in vitro and in cultured cells, its role in RARβ2 gene expression was tested. It was found that knock-down of JMJD6 further stimulated RARβ2 promoter- or RARE-driven luciferase activity as well as RARβ2 gene expression (Fig. 5 I and J). Knock-down efficiency of JMJD6 was examined through RT-qPCR (Fig. S7B). Taken together, these findings show that PRMT4 and JMJD6, and possibly their-regulated HSP70R469me1 are important for RA-induced RARβ2 gene activation.

HSP70 Is Essential for Transcription Initiation Through Regulating TFIIH Recruitment.

Next, we sought to understand the molecular mechanism underlying HSP70 regulation of RA-induced RARβ2 gene activation. It was reported that the sequential events occurring on RARβ2 promoter and involved in RARβ2 gene activation include but are not limited to: (i) upon ligand binding, recruited-RARα undergoes a conformation change, leading to the dismissal of corepressor complexes and the recruitment of coactivator complexes, such as histone-modifying enzymes and chromatin remodeling complexes, which prepare chromatin for transcriptional activation; (ii) the MED1/TRAP220 subunit of mediator complexes, in conjunction with RARα, PARP1, and other subunits in the mediator complexes, triggers the release of the repressive CDK8 module, thus converting mediator into transcriptionally active state; (iii) change in the conformation of mediator leads to a further change in preinitiation complex (PIC) containing RNA Pol II, TFIIA, TFIIB, TFIID, TFIIE, TFIIF, resulting in the recruitment of TFIIH, and subsequent phosphorylation of Pol II at serine 5 and transcriptional initiation (37). To investigate the molecular mechanism underlying HSP70’s regulation of RA-induced RARβ2 gene activation, the binding of transcription factors (RARα/RXRα heterodimer), corepressor complexes (represented by NCoR), chromatin modifying enzymes/coactivator complexes (represented by p300), mediator complexes (represented by MED1/TRAP220 and MED6), and PIC (represented by Pol II, TFIIB, TFIID, and TFIIH) on the RARβ2 promoter region were examined through ChIP in the presence or absence of RA in wild-type or hsp70.1^−/− MEFs. As reported previously, binding of both RARα and RXRα on the RARβ2 promoter was observed even without RA treatment, which were further increased in the presence of RA in wild-type MEFs (Fig. 6A). Upon ligand binding, RARα undergoes a conformational change, resulting in the dissociation of corepressor complexes and the recruitment of coactivator complexes (38, 39). Accordingly, it was found that NCoR was displaced from the promoter, concomitant with the recruitment of the histone acetyltransferase p300 (Fig. 6B). The binding of representative components in the mediator complex including MED1/TRAP220 and MED6 were also examined, with a binding pattern similar to that observed for RARα and RXRα (Fig. 6C). Interestingly, hsp70.1^−/− MEFs exhibited similar binding profiles as wild-type MEFs for all factors tested above (Fig. S8 A–C). Next, we explored the binding profiles of PIC, with Pol II found to be present in the absence of RA, which increased with RA treatment in both wild-type and hsp70.1^−/− MEFs (Fig. 6D and Fig. S8D, Left). As expected, the factors required for Pol II recruitment including TFIIB and TFIID (TBP) displayed a similar binding profiles as Pol II (Fig. 6D and Fig. S8D, Center and Right). However, RA-induced binding of TFIIH (ERCC3), which is dispensable for Pol II recruitment but essential for phosphorylation of Pol II carboxyl-terminal domain at serine 5 (Pol II Ser5 pho) and thus transcription initiation, was only observed in wild-type, but not hsp70.1^−/− MEFs, suggesting HSP70 function in the step of initiation during RARβ2 transcriptional activation (Fig. 6E).

Fig. 6.

HSP70 is required for the recruitment of TFIIH during RA-induced RARβ2 gene transcriptional activation. (A–D) Wild-type (hsp70.1^+/+) MEFs were treated with or without RA (10⁻⁷ M) for 1 h and then subjected to ChIP with control IgG or RARα, RXRα (A); NCoR, p300 (B); MED6, MED1 (C); Pol II, TFIIB, TFIID (TBP) (D) antibody. Binding of these factors on the RARβ2 gene promoter region was examined through qPCR (±SEM, **P < 0.01, ***P < 0.001). (E) Wild-type (hsp70.1^+/+) and hsp70-deficient (hsp70.1^−/−) MEFs were treated with or without RA (10⁻⁷ M) for 1 h and then subjected to ChIP with control IgG or TFIIH (ERCC3) antibody. Binding of TFIIH (ERCC3) on the RARβ2 gene promoter region was examined through q-PCR (±SEM, ***P < 0.001).

Fig. S8.

Binding of receptors, corepressors, coactivators, mediators, and preinitiation complex in hsp70-deficient (hsp70.1^−/−) MEFs. hsp70.1^−/− MEFs were treated with or without RA (10⁻⁷ M) for 1 h, followed by ChIP with control IgG or RARα, RXRα (A); NCoR, p300 (B); MED6, MED1 (C); Pol II, TFIIB, TFIID (TBP) (D) antibody. Binding of these factors on RARβ2 gene promoter region was examined through qPCR (±SEM, *P < 0.05, **P < 0.01, ***P < 0.001).

HSP70 R469 Methylation Is Involved in TFIIH Recruitment.

The requirement of HSP70 for TFIIH recruitment to the RARβ2 promoter promoted us to test whether an interaction occurs between HSP70 and PIC. HEK293T cell lysates were subjected to IP with control IgG or anti-HSP70 specific antibody, followed by IB with antibodies specific against components in PIC, including TFIIB, TFIID (TBP), and TFIIH (ERCC3). Surprisingly, HSP70 specifically pulled down TFIIH, but not TFIIB or TFIID (Fig. 7A). Given the potential role of R469me1 in RARβ2 gene transcriptional activation, we further tested whether the interactions detected between HSP70 and TFIIH is R469 methylation-dependent. HEK293T cell lines stably expressing Flag-tagged wild-type or mutant HSP70 (R469A) were established (Fig. S9), and lysates collected from these cells as well as parental cells were subjected to affinity purification with anti-Flag antibody followed by IB with TFIIB, TFIID (TBP), and TFIIH (ERCC3). As expected, wild-type HSP70 specifically interacted with TFIIH (Fig. 7B). More importantly, the observed interaction seemed to be regulated by R469 as R469A mutation attenuated the interaction between HSP70 and TFIIH, suggesting an important role of HSP70 R469 in the recruitment of TFIIH during RARβ2 gene activation (Fig. 7B). Consistent with a role of R469me1 in the recruitment of TFIIH, knock-down of PRMT4 in HEK293T, which led to a decrease of HSP70R469me1 levels (Fig. 2C), resulted in a decrease of TFIIH (ERCC3) binding induced by RA (Fig. 7C). Finally, hsp70.1^−/− MEFs were transfected with a control vector or vectors expressing wild-type or mutant HSP70 (R469A), treated with or without RA followed by ChIP for TFIIH (ERCC3). It was found that ectopic expression of wild-type HSP70, but not the R469A mutant, induced the binding of TFIIH (ERCC3) to the RARβ2 promoter upon RA treatment (Fig. 7D). Wild-type and mutant HSP70 (R469A) were expressed equally well examined by IB (Fig. 7E).

Fig. 7.

HSP70 R469 methylation involved in TFIIH recruitment. (A) Cell lysates collected from HEK293T cells were subjected to IP with control IgG or anti-HSP70 antibody followed by IB with antibodies as indicated. (B) Cell lysates collected from HEK293T cells stably expressing control vector, Flag-tagged wild-type, or mutant HSP70 (R469A) were subjected to IP with anti-Flag antibody followed by IB with antibodies as indicated. (C) HEK293T cells were transfected with control siRNA or siRNA specifically against PRMT4 for 72 h and then treated with RA (10⁻⁷ M) for 1 h, followed by ChIP with TFIIH (ERCC3) antibody. Binding of TFIIH (ERCC3) on the RARβ2 gene promoter region was examined through qPCR. ChIP signals were presented as fold-induction by RA compared with control after normalization to input (±SEM, ***P < 0.001). (D) hsp70.1^−/− MEFs were transfected with control vector, Flag-tagged wild-type or mutant HSP70 (R469A) for 48 h, and then treated with RA (10⁻⁷ M) for 1 h, followed by ChIP with TFIIH (ERCC3) antibody. Binding of TFIIH (ERCC3) on RARβ2 gene promoter region was examined through qPCR (±SEM, ***P < 0.001). (E) The expression of HSP70 vectors as described in D was examined by IB. (F) Model: HSP70 arginine methylation in RA-induced RARβ2 gene transcriptional activation. Upon RA stimulation, the binding of receptors (RARα and RXRα), mediator complex (such as MED1 and MED6), preinitiation complex (PIC, such as Pol II, TFIIA, TFIIB, TFIID, TFIIE, TFIIF, and TFIIH), HSP70, and PRMT4 were further increased on RARβ2 gene promoter, accompanied by the exchange of corepressor complexes (such as NCoR) and coactivator complexes (such as p300). During this process, PRMT4-mediated monomethylation of R469 in HSP70 was required for the recruitment of TFIIH and transcription initiation. For simplicity, other factors shown to be essential for RA-induced RARβ2 gene transcriptional activation were not included.

Fig. S9.

Generation of HEK293T cells stably expressing Flag-tagged wild-type or mutant HSP70 (R469A). HEK293T cells were transfected with Flag-tagged wild-type (Upper) or mutant HSP70 (R469A) (Lower) expression vector and then selected with puromycin (2.5 mg/mL). The resultant single colonies were harvested, separated by Bis-Tris NuPAGE gel and subjected to IB with anti-HSP70 antibody. Colonies with stably expressed exogenous HSP70 were highlighted in red. Colony number two of both wild-type and mutant HSP70 (R469A) were selected for further experiments.

Discussion

Emerging evidence suggest that histone methyltransferases and demethylases use both histones and nonhistone proteins as substrates, exerting functional roles in a multitude of biological processes (40–44). HSP70 proteins have been known to be methylated on both arginine and lysine residues for decades, but the exact methylation sites, their methyltransferases and demethylases, and the biological functions of these modifications remain incompletely understood (29). In the present study, we provide evidence that a highly conserved arginine residue in HSP70 proteins is methylated, which can be reciprocally regulated by PRMT4 and JMJD6 both in vitro and in cultured cells. Further functional studies have suggested that HSP70 involves in gene transcription through its binding on chromatin, revealing a novel function of HSP70 aside from its classic role in protein folding and quality control. Specifically, HSP70 is indispensable for RA-induced RARβ2 gene activation, during which PRMT4 and potentially PRMT4-mediated HSP70 R469me1 are required for the recruitment of TFIIH, and thus transcription initiation. Mechanistically, HSP70 interacts with TFIIH in an R469me1-dependent manner.

Methylation: An Emerging Piece of the “Nonhistone Code” Involving Chaperone Proteins.

In the past few years, it seemed evident that PTMs found on chaperones become an important regulatory strategy in terms of modulating their ATPase activity, affinity with cochaperone, translocation, and stability, among other activities (29, 31–33, 36, 45–50). Interestingly, in the present study it was found that HSP70 was targeted for methylation at a couple of lysine and arginine residues, which have never been revealed or studied before, including lysine methylation on K77, K88, K332 (unique to HSPA1L), K526, and K550, and arginine methylation on R469. The high conservation of R469 among different isoforms of HSP70 proteins and during evolution prompted us to pursue the functional significance of methylation on this residue, revealing HSP70 has the potential to regulate gene transcription, such as RA-induced RARβ2 gene activation, through its association with chromatin, with R469 methylation being indispensable for transcription initiation. R469 methylation was regulated, at least partially, by PRMT4 and JMJD6 both in vitro and in vivo. Despite the controversial role of JMJD6 functioning as arginine demethylase (51–53), we and others found this enzyme can target to methylated arginine in both histone and nonhistone proteins (18–21). The biological meaning of the multiple methylated lysine residues identified here—and the cross-talk among the PTMs found on HSP70 so far—certainly remains as a great area to explore. Overall, given the large number of PTMs detected on chaperone proteins and their potential to be dynamically regulated by enzymes called “writer” (enzymes adding modifications) and “eraser” (enzymes erasing modifications), we envision that, similar to what has been proposed for histones, there is a “nonhistone” code existing on chaperone proteins. Functionally, we further propose that such a code may define the specificity of chaperone proteins toward their substrates and cofactors, and regulate the diversified cellular processes in which chaperone proteins are involved.

HSP70 Function in Gene Transcription.

Unlike many other chaperones, several isoforms in the HSP70 protein family were found to localize in both nucleus and cytosol of cells. In the cytosol of cells, it serves as a guardian for protein folding and quality control through its binding with stretches of hydrophobic residues in a variety of proteins. In the present study, we provide evidence showing that HSP70 is recruited to the RARβ2 gene promoter region in the presence of RA, and is required for the recruitment of TFIIH. Given its presence in a large number of protein complexes involved in gene regulation, we speculate that HSP70 might also regulate other transcriptional events, such as E2-induced ERα target genes and RA-induced Hox genes, using the same mechanism. However, the molecular basis underlying the observed R469me-dependent interaction between HSP70 and TFIIH remains elusive. It could be either that one of subunits in TFIIH directly reads R469me1 as a variety of functional domains are present in these proteins. Alternatively, an unknown R469me1 “reader” mediates the interaction between HSP70 and TFIIH. Both possibilities certainly remain as intriguing topics for future investigation.

Implications of HSP70 Arginine Methylation.

HSP70 is often seen to be overexpressed in various types of cancers, and its expression correlates with increased tumor grade, poor prognosis, and drug resistance (54). Whereas the precise mechanisms linking HSP70 expression to tumor development are largely unknown, currently available data indicate that elevated HSP70 levels can alter multiple cell signaling and survival pathways, including apoptosis, senescence, autophagy, and HSP90 chaperone activity. HSP70 is therefore believed to play a causal role in cancer initiation (54). Consistently, di-methylated-K561 of HSP70 was found to be significantly enriched in various types of human cancers compared with corresponding normal tissues (33). Given that HSP70 was also required for the transcription of E₂-induced target genes, it will be valuable to investigate whether HSP70 R469me1 is enriched and therefore serving as a biomarker in breast cancer, and if so, what the molecular mechanism underlying HSP70 function in breast cancer initiation and progression is. Finally, although initially greatly anticipated to have clinical effects, the development and clinical efficacy of HSP70 inhibitors have been generally disappointing, which we suggest is largely because of the fact that these inhibitors were designed based on HSP70 chaperone activity (54). As proposed above, targeting PRMT4, the methyltransferase modifying R469 of HSP70 might represent a more promising therapeutic strategy for cancer treatment.

Materials and Methods

The in vitro methylation assay was performed by mixing purified bacterially expressed wild-type or mutant HSP70 (R469A) with PRMT proteins in methylation buffer (50 mM Tris⋅HCl, pH 8.0, 20 mM KCl, 5 mM DTT, 4 mM EDTA) in the presence of 2 μCi l-[methyl-³H]-methionine at 37 °C for 1 h. The reaction was stopped by adding SDS sample buffer followed by SDS/PAGE gel and autoradiogram. For in vitro methylation assay using peptides as substrates, enzymes (PRMTs) were removed by adding glutathione agarose before dot assay and autoradiogram. For in vitro demethylation assay, purified bacterially expressed JMJD6 protein was incubated with HSP70 R469me1 peptide in demethylation buffer (50 mM Hepes, pH 8.0, 20 μM [NH₄]₂Fe[SO₄]₂, 1 mM α-ketoglutarate, and 500 μM ascorbic acid) for 4 h at 37 °C. Similarly, enzyme (JMJD6) was removed by adding Ni-NTA agarose before dot assay and autoradiogram.

Methods for other experiments are detailed in SI Materials and Methods.

SI Materials and Methods

Plasmids and Cloning Procedures.

Full-length HSP70, PRMT1, PRMT2, PRMT3, PRMT4, PRMT5, PRMT6, PRMT7, PRMT8, PRMT9, PRMT10, and PRMT11 were PCR-amplified by using KOD Hot Start DNA Polymerase (Novagen) and then cloned into p3XFLAG-CMV-10 (Sigma), pET-28a (+) (Novagen) or pGEX-6P-1 (Promega) expression vectors. Mutant HSP70 (R469A) and enzymatically dead mutant PRMT4 (E267Q) were generated by using QuikChange Lightning Site-Directed Mutagenesis Kit (Stratagene). Luciferase reporter constructs containing mouse RARβ2 gene promoter sequence [RARβ2 (P)-luc] or retinoid acid response element (tk-βRARE-luc) were made in pGL2 vector (Promega).

siRNAs, Antibodies, Peptides, and Compounds.

siRNAs specifically against HSP70, JMJD6, PRMT1, PRMT2, PRMT3, PRMT4, PRMT5, PRMT6, PRMT7, PRMT8, PRMT9, PRMT10, or PRMT11 were purchased from Qiagen. Anti-HSP70 (sc-24), anti-PARP1 (sc-8007), anti-HSP60 (sc-13115), anti-HSP27 (sc-13132), anti-HSP90 (sc-7947), anti-RXRα (sc-553), anti-RARα (sc-551), anti-ERα (sc-543), anti-GAPDH (sc-25778), anti-PRMT4 (SC-33176), anti-P300 (SC-585), anti-MED6 (sc-9434 and sc-9433), anti-MED1 (sc-74475), anti-Pol II (sc-899), anti-TFIIB (sc-274), anti-TFIID (TBP) (sc-273), and anti-TFIIH (ERCC3) (sc-293) antibodies were purchased from Santa Cruz Biotechnology; anti-Flag (F1804) antibody was purchased from Sigma; anti-JMJD6 (ab10526), antimethylated arginine (Rme) (ab412), and antimethylated lysine (Kme) (ab23366) antibodies were purchased from Abcam; anti-NCoR antibody was described previously (55); anti-HSP70 R469me1 antibody was generated by Abgent [CRFELSGIPPAPR(me1)GVPQIE]. HSP70 peptides used in this study were custom synthesized: HSP70 R469 (unmodified) (FELSGIPPAPRGVPQIEVTF), HSP70 R469A (FELSGIPPAPAGVPQIEVTF), and HSP70 R469 (me1) [Biotin-GIPPAPR(me1)GVPQIE]; Flag peptide was purchased from Sigma. AdOx (A7154), PES (P0122), all-trans-RA (R2625), β-estradiol (estrogen, E₂) (E8875), DHT (A8380), and histones (H9250) were purchased from Sigma; TNF-α (human recombinant) was purchased from BD Bioscience; AMI-1 (539209) was purchased from Calbiochem.

siRNA Transfection, RNA Isolation, and RT-qPCR.

siRNA transfections were performed using Lipofectamine 2000 (Invitrogen) according to the manufacturer’s protocol. Transfection efficiency was determined by IB or RT-qPCR. Total RNA was isolated from HEK293T cells using RNeasy Mini Kit (Qiagen) following the manufacturer’s protocol. First-strand cDNA synthesis from total RNA was carried out using iScript cDNA Synthesis Kit (Bio-Rad), followed by qPCR using Mx3005 machine (Stratagene). All RT-qPCRs were repeated at least three times and representative results were shown.

Protein IP and IB.

Protein IP and IB were performed following the protocol described previously (19, 56).

Protein Purification from Bacterial Cells or HEK293T Cells.

His-tagged JMJD6 and HSP70 or GST-tagged PRMT1 to 11 were expressed in BL21 (DE3) bacterial cells (Stratagene) and purified by using Ni-NTA agarose (Qiagen) or Glutathione agarose (Sigma), respectively, following the protocol described previously (19, 56). Flag-tagged proteins were expressed in HEK293T cells and cells were lysed in lysis buffer containing 50 mM Tris⋅HCl, pH 7.4, 150 mM NaCl, 1 mM EDTA, 1% Triton X-100. Flag-tagged proteins were then affinity-purified by using anti-Flag M2 agarose and washed extensively with washing buffer containing 50 mM Tris⋅HCl, pH 7.4, 450 mM NaCl, 1 mM EDTA, 1% Triton X-100 before elution with 3XFlag peptides (Sigma). For purifying JMJD6 proteins, buffer exchange to PBS was performed by using PD-10 desalting column (GE Healthcare) after imidazole or flag peptide elution, so that any residual elution buffer would be removed to capture JMJD6 demethylase activity.

In Vivo Labeling.

HEK293T cells were maintained in DMEM without methionine for 24 h before adding NTM [cycloheximide (100 μg/mL) and chloramphenicol (40 μg/mL)] for another hour, followed by adding [³⁵S]-methionine (20 μCi/mL; PerkinElmer) or l-[methyl-³H]-methionine (20 μCi/mL; PerkinElmer) in the presence or absence of AdOx (20 μM) for 4 h. Cells were then lysed in 1% Nonidet P-40 buffer (50 mM Tris⋅HCl, pH 8.0, 150 mM NaCl, 1% Nonidet P-40) supplemented with protease inhibitor mixture (Roche) before immunoprecipitation with anti-HSP70 or anti-Flag antibody, followed by SDS/PAGE gel and autoradiogram.

Mass Spectrometry Analysis.

HEK293T cells were lysed in lysis buffer (50 mM Tris⋅HCl, pH 7.4, 150 mM NaCl, 1 mM EDTA, 1% Triton X-100) and precleared with protein G beads (Sigma) for 1 h before incubation with anti-HSP70 antibody overnight at 4 °C, followed by adding protein G beads for another 2 h before washing five times with washing buffer (same as lysis buffer except salt concentration was increased to 420 mM). Immunoprecipitates were boiled in SDS sample buffer and resolved by 4–12% Bis-Tris NuPAGE gel (Invitrogen), followed by Coomassie blue staining. Band corresponding to HSP70 was subjected to in gel digestion and mass spectrometry analysis following the protocol as described previously (19). The resulting RAW files were converted to mzXML using the program ReAdW (tools.proteomecenter.org/software.php) and analyzed with InsPecT (57) (proteomics.ucsd.edu/Software/Inspect/) (P ≤ 0.3).

For MALDI-TOF MS analysis, demethylation reactions were subjected to desalting using ZipTip (Millipore) following the manufacturer’s protocol. Standard MALDI-TOF analysis was then performed to examine demethylation events.

Cellular Fractionation.

Cells were washed with ice-cold PBS, collected, spun down and resuspended in ice-cold buffer I (10 mM Hepes, pH 8.0, 1.5 mM MgCl₂, 10 mM KCl, 1 mM DTT) supplemented with protease inhibitor mixture, followed by incubation for 15 min on ice to allow cells to swell. Nonidet P-40 was then added at a final concentration of 1% (use 10% stock solution) followed by vortexing for 10 s. Nuclei were collected by centrifuging 2∼3 min at maximum speed (∼21,100 × g). The resultant supernatant was cytosolic fraction. Nuclei were lysed in ice cold buffer II [20 mM Hepes, pH 8.0, 1.5 mM MgCl₂, 25% (vol/vol) glycerol, 420 mM NaCl, 0.2 mM EDTA, 1 mM DTT] supplemented with protease inhibitor mixture followed by vigorous rotation at 4 °C for 30 min and centrifugation 15 min at maximum speed. The resultant supernatant was nuclear fraction.

Measurement of ATPase Activity.

Measurement of ATPase activity was performed as described previously (32). Briefly, reactions were performed with PRMT4, HSP70 (wild-type or R469A mutant), SAM (1 mM) and ATP (100 μM), as indicated, in ATPase assay buffer (25 mM Hepes, pH 7.4, 1 mM DTT, 5 mM MgCl₂, and 50 mM KCl) in a total volume of 50 μL in a 96-well plate at 37 °C for 60 min. ATPase activity was measured by addition of 100 μL of BIOMOL Green reagent (Enzo Life Sciences) and incubated at room temperature for 30 min. The absorbance at 630 nm was then measured.

ChIP.

ChIP was performed following the protocol described previously (19, 56). Briefly, cells were fixed with 1% formaldehyde (Sigma) for 10 min at room temperature. Fixation was stopped by adding glycine (0.125 M) and incubating for 5 min at room temperature, followed by washing with PBS twice. Chromatin DNA was sheared to 300∼500-bp average in size through sonication. Resultant was immunoprecipitated with control IgG or specific antibodies overnight at 4 °C, followed by incubation with protein G magnetic beads (Invitrogen) for an additional 2 h. After washing and elution, the protein–DNA complex was reversed by heating at 65 °C overnight. Immunoprecipitated DNA was purified by using QIAquick spin columns (Qiagen) and analyzed by qPCR using Mx3005 machine (Stratagene). All ChIP-qPCRs were repeated at least three times and representative results were shown.

Luciferase Reporter Assay.

HEK293T cells were transfected with first round of siRNA for 24 h before a second round of siRNA transfection together with luciferase reporter and Renilla luciferase control reporter vectors (Promega) for another 48 h, followed by treatment with RA (10⁻⁷ M) for 12 h. Cells were then washed with PBS buffer twice and lysed with 1× passive lysis buffer (Promega) before measurement of luciferase reporter activities by using Envision Multimode Plate Reader (PerkinElmer).