HDAC6 Substrate Discovery Using Proteomics-Based Substrate Trapping: HDAC6 Deacetylates PRMT5 to Influence Methyltransferase Activity

Abstract

Histone deacetylase 6 (HDAC6) is upregulated in a variety of tumor cell lines and has been linked to many cellular processes, such as cell signaling, protein degradation, cell survival, and cell motility. HDAC6 is an enzyme that deacetylates the acetyllysine residues of protein substrates, and the discovery of HDAC6 substrates, including tubulin, has revealed many roles of HDAC6 in cell biology. Unfortunately, among the wide variety of acetylated proteins in the cell, only a few are verified as HDAC6 substrates, which limits the full characterization of HDAC6 cellular functions. Substrate trapping mutants were recently established as a tool to discover unanticipated substrates of histone deacetylase 1 (HDAC1). In this study, we applied the trapping approach to identify potential HDAC6 substrates. Among the high confidence protein hits after trapping, protein arginine methyl transferase 5 (PRMT5) was successfully validated as a novel HDAC6 substrate. PRMT5 acetylation enhanced its methyltransferase activity and symmetrical dimethylation of downstream substrates, revealing possible crosstalk between acetylation and methylation. Substrate trapping represents a powerful, systematic, and unbiased approach to discover substrates of HDAC6.

Graphical Abstract

INTRODUCTION

Gene expression is regulated by chromatin remodeling factors and modification of nucleosomal histone proteins.¹ One of the histone modifications, acetylation, is regulated by histone acetyltransferase (HAT) and histone deacetylase (HDAC) proteins. HDAC proteins remove acetyl groups from ε-N-acetyl lysine amino acids of histones to alter the packaging of genomic DNA in nucleosomes and affect gene expression.² As a result of their key role in epigenetics, HDAC proteins are implicated in many disease states and represent therapeutic targets. HDAC inhibitors induce cell cycle arrest and apoptosis in various cancer cell lines.^3,4 Many HDAC inhibitors are in clinical trials as anticancer drugs, with four approved by the U.S. Food and Drug Administration for clinical use, including Vorinostat (SAHA or suberoylanilide hydroxamic acid) for the treatment of cutaneous T cell lymphoma.^5,6

The human HDAC family includes 18 members, with HDAC1-11 distinguished by a metal-dependent catalytic mechanism in contrast to Sirtuins1-7 that involve a NAD+ catalytic mechanism.⁷ Among the HDAC family, we focus here on HDAC6 due to its role in disease. HDAC6 is overexpressed in cancers of the breast, heart, liver, kidney, testis, brain, and pancreas.^8,9 HDAC6 also acts as an important target for neurodegenerative disorders.¹⁰ In addition to playing a role in disease, HDAC6 is also protective by functioning in the innate response to intracellular bacterial infections¹¹ and autophagy.¹²

In addition to links to disease, HDAC6 is structurally distinct from the other HDAC family members by containing a ubiquitin-binding zinc-finger domain and two active deacetylase domains.^13,14 Characterization of the two HDAC6 catalytic domains has been controversial. Using recombinant HDAC6, one report documented that both catalytic domains were required for active deacetylation,¹⁵ whereas another paper reported independent activities, with the C-terminal domain (catalytic domain 2 or CD2) more active than the N-terminal domain (catalytic domain 1 or CD1).¹⁶ Using immunoprecipitated HDAC6, both catalytic domains independently deacetylated histones,^17,18 whereas CD2 preferentially deacetylated the substrate tubulin,¹⁷ suggesting different substrate specificities of the two catalytic domains. Distinct from its deacetylase activity, CD1 also displayed E3 ligase activity with some substrates.¹⁹

HDAC6 deacetylates various substrates to regulate numerous processes in the cell.^20–23 As relevant examples, cytoplasmic HDAC6 deacetylates α-tubulin and cortactin to regulate cell migration. HDAC6-mediated deacetylation of peroxiredoxin regulates redox processes inside the cell. Alternative splicing of pre-mRNA is influenced by HDAC6-dependent deacetylation of Sam68 (Src-associated substrate during mitosis of 68 kDa). By characterizing substrates, the unique functions of HDAC6 in cell biology are slowly being uncovered.

In recent years, mass spectrometry (MS)-based approaches have identified thousands of acetylated proteins in the cell.^24–26 Because acetylation is dynamically regulated, HDAC proteins likely play a large and important role in acetylation-dependent cellular events. However, among the thousands of acetylated proteins, relatively few HDAC substrates have been validated, including substrates of HDAC6. The slow progress to identify HDAC substrates is largely due to the paucity of HDAC substrate discovery methods. The traditional method of immunoprecipitation is problematic for substrate isolation because enzymatically active HDAC proteins bind substrates transiently, resulting in a loss of the substrates during enrichment (Figure 1A). As an alternative to immunoprecipitation, HDAC6 selective inhibitors were used to alter acetylation levels and identify HDAC6 substrate by MS-based proteomics.²⁷ HDAC6 knockout mice cell lines were also used to identify novel substrates.²⁸ Additional simple methods to identify HDAC6 substrates would complement available approaches to allow a full characterization of HDAC6 activities in the cell.

Figure 1.

Substrate trapping strategy. A) Wild-type (WT) HDAC binds substrates transiently, which challenges purification and identification. B) Inactive mutant HDAC binds substrates with longer residence time, which facilitates purification and identification. C) An HDAC inhibitor (HDACi) competes with substrate binding to displace active site bound substrates.

In this study, we applied the “substrate trapping” approach to discover unanticipated substrates of HDAC6. In prior work with HDAC1, the failure of immunoprecipitation to identify substrates due to transient binding (Figure 1A) was overcome using inactive mutants, which bind to substrates with longer residence time without deacetylation (Figure 1B).²⁹ As a helpful control to ensure active site binding of proteins isolated by the inactive mutant, an HDAC inhibitor was included in the mutant immunoprecipitation as a binding competitor (Figure 1C). With HDAC1, trapping identified lysine demethylase 1 (LSD1) and Eg5 (KIF11) as novel substrates.^29,30 More recently, trapping was combined with proteomics-based mass spectrometry to identify many substrates simultaneously and create substrate profiles.^31,32

Given the high sequence similarity of catalytic domains in the metal-dependent HDAC family, we hypothesized that substrate trapping could be expanded to identify novel substrates of other HDAC family members, including HDAC6. Alanine mutation of position H611 in the C-terminal catalytic domain of HDAC6 generated a mutant that isolated substrates, which were identified using mass spectrometry (MS) analysis. Many known HDAC6 substrates were identified, confirming the value of the trapping method. In addition, one new substrate, protein arginine methyl transferase 5 (PRMT5), was validated, which suggests possible crosstalk between acetylation and methyltransferase activity. In total, this work documents successful use of substrate trapping to identify substrates of HDAC6.

RESULTS AND DISCUSSION

Prior trapping with HDAC1 utilized catalytic site mutants F150A, H141A, and C151A as substrate traps.^29,33 Based on sequence alignment (Figure S1), the homologous amino acids for H141A and C151A of HDAC1 are conserved in both catalytic domains of HDAC6, although HDAC1 F150A was not. As a result, only residues corresponding to H141A and C151A in HDAC1 in both catalytic domains of HDAC6 were mutated separately to generate four possible trapping mutants: H216A, C226A, H611A, and C621A (Figure 2A).

Figure 2.

HDAC6 trapping mutants. A) HDAC6 functional domains are shown, along with the sites of mutation in the two catalytic domains (CD1 and CD2) indicated by arrows. B) HEK293 cells were untreated (NP – no protein) or transfected with expression plasmids for Flag-tagged wild-type (WT) and mutant HDAC6 proteins. After recovery and lysis, proteins were immunoprecipitated via the Flag tag and used in HDAC-Glo I/II luminescence assays. The mean percent deacetylase activity of three independent trials compared to wild-type HD6 (set at 100%) is shown with standard error. The Flag protein levels (HDAC6-F) after immunoprecipitation are also shown (bottom gel). Repetitive trials of quantified data are shown in Table S1. C) After transfection and 24-h recovery, cells were incubated with SAHA (10 μM) for 24 h to induce robust acetylation. After cell lysis, HDAC6 proteins in the lysates were immunoprecipitated with anti-Flag-agarose beads in the absence (DMSO) or presence of SAHA (100 μM), separated by SDS-PAGE, and immunoblotted with Flag (α-Flag) or acetyl-tubulin (α-AcTub) antibodies. The level of Actubulin in the input lysate before immunoprecipitation is shown as an expression control. Repetitive trials are shown in Figure S2.

Trapping of the Known Substrate Tubulin by HDAC6 Mutants.

With mutants of HDAC6 in hand, assays were performed to determine their deacetylase activities. Wild-type or mutant HDAC6 proteins were expressed as Flag-tagged proteins in HEK293 cells. After cell lysis, HDAC proteins in the cell lysates were immunoprecipitated via the Flag tag, and deacetylase activity was assessed with the HDAC-Glo I/II assay. HDAC6 H216A and C226A mutants from CD1 were relatively active (Figure 2B, columns 3 and 5), maintaining 86% ± 1 and 65% ± 10 activities, respectively, compared to wild type (Figure 2B, column 2). In contrast, HDAC6 H611A and C621A mutants from CD2 were comparatively less active (Figure 2B, columns 4 and 6), maintaining 39% ± 2 and 19% ± 2 activities, respectively, compared to wild type. Consistent with these results, prior work documented that alteration of the histidine in the second catalytic domain of HDAC6 resulted in a loss of enzyme activity.^16–18

To test whether the HDAC6 mutants displayed substrate trapping abilities, the ability of the four HDAC6 mutants to trap tubulin, the first identified substrate of HDAC6,²⁰ was explored. HEK293 cells were transfected with expression constructs for Flag-tagged wild-type or mutant HDAC6 and then treated with pan HDAC inhibitor SAHA to increase acetylation of cellular proteins. After lysis, overexpressed HDAC proteins in the lysates were immunoprecipitated via the Flag tag in the absence or presence of SAHA, separated by SDS-PAGE, and immunoblotted with Flag and acetyl-tubulin antibodies (Figure 2C). Among the four mutants tested, the CD2 histidine mutant, H611A, immunoprecipitated the highest levels of acetyl-tubulin compared to wild type or the other mutants (Figure 2C, compare lane 4 to lanes 2–3 and 5–6). In addition, bound tubulin was significantly reduced in the presence of competitive SAHA (Figure 2C, compare lanes 4 and 9), consistent with active site binding. Based on the results, the CD2 H611A mutant was used in subsequent trapping studies.

Substrate Discovery Using the HDAC6 H611A Trapping Mutant.

The next goal was to identify the enriched proteins after trapping to discover possible new substrates of HDAC6. HEK293 cells were transfected with Flag-tagged wild-type and H611A mutant HDAC6 expression plasmids, treated with SAHA to induce protein acetylation, and lysed, followed by immunoprecipitation of HDAC6 proteins in the presence or absence of SAHA, as described earlier. Before liquid chromatography-tandem MS (LC-MS/MS) analysis of the bound proteins, the levels of wild-type and mutant H611A after immunoprecipitation were assessed to ensure equal protein capture (Figure S3). The samples were then subjected to LC-MS/MS analysis after trypsin digestion to identify bound proteins. A total of 749 proteins were observed in four independent trials after label-free quantitation using Maxquant (Table S2).³⁴ One trial showed dramatically lower levels of HDAC6 in the H611A mutant sample and was removed from further analysis. Using the remaining three trials, a fold enrichment ratio was generated by dividing the intensity of each peptide in the mutant H611A sample by the intensity of that same peptide in the wild-type HDAC6 sample (Table S3). Using a ≥2.0-fold enrichment ratio in at least two out of three trials as the minimal threshold, 120 candidate substrates were identified (Table S3). To ensure active site binding of potential substrates, hit peptides were also required to have reduced intensity in mutant H611A samples treated with SAHA compared to without SAHA. Among the hit list, five known substrates of HDAC6 were identified, including cortactin (CTTN),²¹ sam68 (KHDRBS1),²³ peroxiredoxin 1 (PRDX1),²² peroxiredoxin 2 (PRDX2),²² and α-tubulin (TUBA1A).²⁰ The presence of multiple known HDAC6 substrates in the hit list established trapping as a viable discovery strategy.

To further assess the efficacy of trapping, the 120 identified putative HDAC6 substrates were compared to previous proteomics studies.^27,28 In the first study where SILAC labeling and HDAC6-inhibition by tubacin in HeLa cells identified 169 differentially acetylated proteins, four hits were in common to the HDAC6 trapped hits, including known HDAC6 substrate cortactin (CTTN), eukaryotic translation initiation factor 4B (EIF4B), heat shock 70 kDa protein 4 (HSPA4), and lupus La protein (SSB). In the second study using the liver of HDAC6 knockout mice, along with acetyllysine immunopurification and LC-MS/MS analysis, 107 proteins with elevated acetylation were identified.²⁸ Four hits were in common with the HDAC6 trapped hits, including the known HDAC6 substrate peroxiredoxin 1 (PRDX1),²² heterogeneous nuclear ribonucleoprotein F (HNRNPF), nucleolar protein 58 (Nop58), and polyadenylate-binding protein 1 (Pabpc1). In addition, all three studies isolated different isoforms of RNA-dependent RNA helicases (DDX1 and DDX17 vs DDX3X) and elongation factors (EEF1D vs EEF1A1 vs EEF2). Despite the use of different organisms, cell lines, or methods of purification, many common hits were identified among the three proteomics studies, which documents the complementarity of all three methods.

To further analyze the 120 candidate substrates, gene ontology (GO) analysis using DAVID Bioinformatics Resources 6.8 software revealed cellular processes, molecular function, and cellular compartment of the hits. Consistent with the known roles of HDAC6 in tubulin and ubiquitin regulation, cotranslational membrane targeting, microtubule cytoskeleton organization, ER-associated and ubiquitin-dependent protein catabolism, and protein sumoylation were observed among the biological processes (Figure 3). Several processes linked to epigenetic activities were also identified, such as gene expression and methylation. Unexpectedly, many biological processes involving RNA-related activities were observed, including translation RNA catabolism, RNA processing, RNA splicing, RNA stability, and RNA localization. Consistent with these biological process analysis, molecular functions identified among the 120 hit proteins included RNA binding, protein binding, cytoskeleton structure, and actin binding (Figure S4A). Finally, the predominant cytosolic, cytoplasmic, and nucleus locations of the hit proteins (Figure S4B) are consistent with prior work documenting shuttling of HDAC6 between the nucleus and the cytoplasm (Figure S4B)³⁵ and our observations in HEK293 cells (Figure S5).

Figure 3.

Functional analysis of potential HDAC6 substrates. The 120 enriched proteins from substrate trapping were classified according to biological processes using DAVID Bioinformatics Resources 6.8 software. The inverse log of the p value is shown. Additional analysis by DAVID is shown in Figure S4.

The 120 hit proteins were also analyzed for known physical interactions with HDAC6 using the GeneMANIA application in Cytoscape.^36,37 HDAC6 has reported interactions with six of the hit proteins (Figure 4, yellow), including the two substrates, α-tubulin (TUBA1A) and cortactin (CTTN). HDAC6 also has indirect interactions with 78 of the hit proteins (Figure 4, outer circles). Considering both direct and indirect interactions, 70% (84 proteins) of the hit list is found in the HDAC6 interactome, suggesting that these proteins are in proximity of HDAC6 for possible deacetylation.

Figure 4.

Interactome of the 120 possible HDAC6 substrates. The 120 proteins enriched in HDAC6 trapping studies were analyzed using the GeneMANIA application in Cytoscape for known physical interactions. HDAC6 (red) interacts with six proteins directly (yellow) and many other proteins indirectly through one (green), two (blue), or three (purple) other proteins. Many hit proteins have no known interaction (gray). Proteins in the high confidence (circle) and higher confidence (square, Figure S7) hit lists are indicated. Finally, the five known substrates of HDAC6 are starred in black, with the new substrate, PRMT5, identified and validated in this work starred in red.

Finally, the cellular abundances of the 120 hit proteins in HEK293 cells using the PAXdb database were considered to assess the success of the affinity-based enrichment, which should be independent of protein abundance. The abundance range for all 10,225 observed proteins in HEK293 cells was 5.12 to 448 ppm, with a median of 135 ppm.³⁸ The abundance of the 120 hit proteins ranged from 15.1 to 448 (Figure S6), with 71% (85 proteins) of average abundance (100–200 ppm), 16% (19 proteins) of relatively high abundance (>200 ppm), and 7% (8 proteins) in the low abundance range (<100 ppm). Based on the abundance analysis, proteins were trapped regardless of their relative abundance in the cell.

Validation of Trapping Hits.

To create a higher confidence list from which to select a candidate substrate to validate, 28 proteins enriched above 2.0-fold in all three trials were identified (Table S3), which included the known HDAC6 substrate, cortactin. The known physical protein–protein interactions among 28 trapped high confidence hits revealed direct and indirect binders of HDAC6, as well as proteins with no known interactions (Figures 4 and S7, squares). With this higher confidence list, methylation and methyltransferase activities were observed among the biological processes and molecular function categories (Figure S8A,B). In addition, analysis of cellular compartment revealed localization in the methylosome (Figure S8C). Given these clear connections to methylation, methylosome-localized protein arginine methyl transferase 5 (PRMT5) was selected for further validation.

PRMT5 is an epigenetic modifier, catalyzing transfer of one or two methyl groups from S-adenosylmethionine (AdoMet, SAM) to the guanidino nitrogen(s) of arginine. Similar to HDAC6, PRMT5 is a drug target for a variety of cancer types³⁹ and also affects numerous cellular processes, such as gene transcription, differentiation, genomic organization, and cell cycle control.⁴⁰ Given the similar functions of HDAC6 and PRMT5, validation studies were performed to reveal possible crosstalk between methylation and acetylation. As the first step to validate PRMT5 as an HDAC6 substrate, we first confirmed that PRMT5 was trapped by HDAC6. Here, trapping experiments were performed, as described, but with analysis by Western blotting to observe trapped PRMT5. As expected based on the LC-MS/MS data, the H611A HDAC6 mutant bound more PRMT5 than wild-type HDAC6 after immunoprecipitation (Figure 5A, compare lane 3 to lane 2), which was reduced in the presence of a competitive inhibitor SAHA (Figure 5A, compare lane 5 to lane 3). The Western blot analysis corroborated the LC-MS/MS results and showed that PRMT5 bound to the HDAC6 active site.

Figure 5.

HDAC6 deacetylates PRMT5. A) HEK293 cells were untreated (NP, no protein) or transfected with expression plasmids for wild-type (WT) or mutant H611A HDAC6. After 24-h recovery, cells were incubated with SAHA (10 μM) for 24 h to induce robust acetylation. HDAC6 in the cell lysates was immunoprecipitated with anti-Flag-agarose beads with or without SAHA (100 μM), separated by SDS-PAGE, and probed with Flag and PRMT5 antibodies. GAPDH and PRMT5 were used as input controls to show equal protein expression and gel loading. Repetitive trials are shown in Figure S9. B) Endogenous HDAC6 and PRMT5 were separately immunoprecipitated from HEK293 cells. Immunoprecipitated HDAC6 was preincubated in the absence or presence of SAHA (2 mM), mixed with immunoprecipitated PRMT5, and then incubated for 1 h at 37 °C with rocking. After washing, proteins were separated by SDS-PAGE and probed with acetyllysine (AcK), PRMT5, and HDAC6 antibodies. C) Quantified data from three independent trials (Figure S10A and part B), with mean and standard error shown (Figure S10B). D) HEK293 cells were untreated (NP, no protein) or transfected with expression plasmids for Flag-tagged wild-type or H611A HDAC6. After lysis, PRMT5 was immunoprecipitated, washed, separated by SDS-PAGE, and probed with acetyllysine (AcK), PRMT5, and HDAC6 antibodies. E) Quantified data from three independent trials (Figure S11A and part D) with mean and standard error shown (Figure S11B). F) HEK293 cells were untreated (DMSO vehicle) or treated with SAHA (10 μM), TubA (10 μM), and SHI 1:2 (10 μM) for 24 h. After lysis, PRMT5 was immunoprecipitated, separated by SDS-PAGE, and visualized with acetyllysine (AcK) and PRMT5 antibodies. G) Quantified data from three independent trials (Figure S12A and part F) with mean and standard error shown (Figure S12B). ***p ≤ 0.0005, **p ≤ 0.005, *p ≤ 0.05, ns = not significant.

PRMT5 Is a Substrate of HDAC6.

To further confirm that PRMT5 is a physiological substrate of HDAC6, in vitro deacetylation assays were performed using immunoprecipitated HDAC6 and PRMT5. Briefly, PRMT5 and HDAC6 were separately immunoprecipitated from HEK293 cells. After washing, immunoprecipitated HDAC6 was preincubated in the absence or presence of SAHA and combined with the immunoprecipitated PRMT5. After incubation, proteins were then separated by SDS-PAGE and probed with an acetyllysine antibody to monitor changes in the acetylation state. The basal acetylation of PRMT5 decreased in the presence of HDAC6 (Figure 5B, compare lane 2 to lane 1), which is consistent with HDAC6-dependent deacetylation. In addition, inhibition of HDAC6 by SAHA rescued PRMT5 acetylation (Figure 5B, compare lane 3 to lane 2). Quantified data from three independent trials showed a significant decrease in acetylation in the presence of HDAC6 (Figure 5C). The deacetylation data provide evidence that PRMT5 is a substrate of HDAC6 in vitro.

As additional confirmation of HDAC6-dependent deacetylation of PRMT5, two cell-based assays were performed. First, the level of PRMT5 acetylation was monitored in the presence of overexpressed wild-type or H611A mutant HDAC6. The expectation was an increase in PRMT5 acetylation in the presence of overexpressed inactive HDAC6 mutant due to a shift in equilibrium between acetyltransferase and deacetylase activities. Cells overexpressing Flag-tagged wild-type or H611A mutant HDAC6 were lysed before immunoprecipitation of PRMT5, separation by SDS-PAGE, and monitoring of PRMT5 acetylation. As expected, overexpression of the inactive H611A HDAC6 increased PRMT5 acetylation compared to untransfected or wild-type HDAC6 expressing cells (Figure 5D, compare lane 3 to lanes 1 and 2). Quantified data from three independent trials showed a significant increase in acetylation in the presence of HDAC6 H611A mutant compared to wild-type HDAC6 (Figure 5E), indicating that HDAC6 deacetylated PRMT5 in cells.

As a second cell-based method, the level of PRMT5 acetylation was monitored in the presence of HDAC inhibitor drugs. Again, the expectation was enhanced PRMT5 acetylation upon HDAC6 inhibition. Cells were incubated with broad spectrum SAHA, HDAC6/HDAC10-selective Tubastatin A,⁴¹ or HDAC1/2-selective SHI1:2,⁴² followed by immunoprecipitation of PRMT5 from the resulting cell lysates, separation by SDS-PAGE, and monitoring of PRMT5 acetylation. As expected, an increase in PRMT5 acetylation was observed with SAHA- or Tubastatin A-treated cells (Figure 5F, compare lanes 2 and 3 to lane 1). No increase in PRMT5 acetylation was observed in the presence of the HDAC1/2-selective inhibitor, SHI1:2 (Figure 5F, compare lane 4 to lane 1). Quantified data from three independent trials showed a significant increase in PRMT5 acetylation only with HDAC6-targeting SAHA and Tubastatin A (Figure 4G). Taken together, the trapping, in vitro, and cell-based studies confirm that PRMT5 is a substrate of HDAC6.

Effect of Acetylation on PRMT5 Activity.

Next, our goal was to test the influence of HDAC6-dependent acetylation on the methyltransferase activity of PRMT5. Symmetrical dimethylation of histone H4 arginine 3 (H4R3), a known substrate of PRMT5, was used to measure PRMT5 activity.⁴³ HEK293 cells were transfected with an expression plasmid for the Myc-PRMT5 fusion protein and treated with DMSO, SAHA, TubA, or SHI1:2 to alter acetylation. Myc-PRMT5 was then immunoprecipitated, and acetylation levels were monitored. Consistent with the data from endogenous PRMT5 (Figure 5F), overexpressed Myc-PRMT5 showed increased acetylation after treatment with SAHA and Tubastatin A (Figure 6A, compare lanes 3 and 4 to lane 2) but not SHI1:2 (Figure 6A, compare lane 5 to lane 2). To assess the methyltransferase activity of the acetylated form of PRMT5, immunoprecipitated PRMT5 from the inhibitor-treated cells was used in a commercial H4R3 methyltransferase assay. Elevated H4R3 dimethylation was observed with PRMT5 from SAHA and Tubastatin A-treated cells compared to untreated (Figure 6B), which suggests that acetylation enhances PRMT5 enzymatic activity. In contrast, an insignificant decrease of H4R3 dimethylation was observed with PRMT5 from SHI1:2 treated cells (Figure 6B), confirming HDAC6 dependence. The results suggest that acetylation increases PRMT5 enzymatic activity in an HDAC6-dependent manner.

Figure 6.

Acetylation affects PRMT methyltransferase activity. A) HEK293 cells expressing the PRMT5-Myc fusion protein were untreated (DMSO, NT) or incubated with SAHA (10 μM), TubA (10 μM), or SHI 1:2 (10 μM) for 24 h. After lysis, PRMT5-Myc from the lysate was immunoprecipitated with the Myc antibody, eluted using a Myc-peptide, separated by SDS-PAGE, and visualized with acetyllysine (AcK) and PRMT5 antibodies. Repetitive trials were shown in Figure S13A. B) Dimethylation activities of immunoprecipitated Myc-PRMT5 from part A were determined using a commercial H4R3 methyltransferase assay, with signals from Myc-PRMT5 compared among SAHA, TubA, or SHI1:2-treated cells and untreated (DMSO) cells (set to 100%). The mean and standard error from three independent trials are shown in Figure S13B. C) Flag-tagged WT and mutant HDAC6 were overexpressed in HEK293 cells, which were treated with thapsigargin (1 μM) for 5 h to induce ER stress alone or with the PRMT5-specific inhibitor EPZ015666 (5 μM) for 12 h, before harvesting. Endogenous p53 in the cell lysates was immunoprecipitated, separated by SDS-PAGE, and probed with symmetric N^G,N^G-dimethylarginine (sDMA) or p53 antibodies. Repetitive trials were shown in Figure S14A. D) Quantified data from three independent trials from part C, with mean and standard error shown in Figure S14B. E,F) Flag-tagged WT and mutant H611A HDAC6 were overexpressed in HEK293 cells and either untreated (E) or treated with thapsigargin (1 μM) for 5 h to induce ER stress (F). After cell lysis, endogenous PRMT5 and endogenous G3BP1 were immunoprecipitated separately. After washing, proteins from each immunoprecipitate were separated by SDS-PAGE and probed with acetyllysine (AcK), PRMT5, HDAC6, sDMA, or G3BP1 antibodies. Repetitive trials were shown in Figures S15A and S16A. G.H) Quantified data from three independent trials from parts E (G) and F (H), with mean and standard error shown in Figures S15B and S16B. ****p ≤ 0.0001, ***p ≤ 0.0005, **p ≤ 0.005, *p ≤ 0.05, ns–not significant.

To further confirm that acetylation impacts PRMT5 enzymatic activity, the symmetrical dimethylation of PRMT5 substrate, p53 was tested. PRMT5 methylates p53 at R333, R335, and R337, resulting in changes in oligomerization, nuclear location, and DNA binding specificity.⁴⁴ Moreover, DNA damage led to elevated p53 dimethylation. A separate report documented that PRMT5 was required for p53 expression and cell proliferation.⁴⁵ Finally, oncogenic stress led to p53-mediated responses that were dependent on PRMT5-mediated dimethylation.⁴⁶ With a role for arginine dimethylation in regulation of p53 activities, we hypothesized that PRMT5 acetylation influences p53-dependent functions, particularly in response to DNA damage and cell stress.

To test the influence of PRMT5 acetylation on p53 dimethylation, endogenous p53 was immunoprecipitated from cells overexpressing wild-type and inactive mutant H611A HDAC6 under thapsigargin-induced ER stress conditions. The PRMT5-specific inhibitor EPZ015666 was also included to ensure PRMT5 dependence. Elevated p53 dimethylation was observed in both untransfected and H611A HDAC6 expressing cells (Figure 6C, lanes 1 and 3) compared to HDAC6 wild-type expressing cells (Figure 6C, lane 2). The presence of EPZ015666 reduced p53 dimethylation in all samples (Figure 6C, lanes 4–6), documenting the PRMT5-dependence of p53 dimethylation. Quantification from three independent trials showed significant increases in p53 dimethylation in the presence of both untransfected and inactive H611A mutant HDAC6, as well as a significantly reduced dimethylation in the presence of EPZ015666 (Figure 6D). The data documented that HDAC6 influenced the dimethylation of p53 in a PRMT5-dependent manner.

To confirm the elevated enzymatic activity of acetylated PRMT5 using a second cellular substrate, the symmetrical dimethylation of Ras-GAP SH3-binding protein 1 (G3BP1)⁴⁷ was monitored. Prior data reported that PRMT5-mediated G3BP1 methylation impaired formation of stress granules (SG), which are stress-induced cytoplasmic condensates that form after impaired translational initiation to prevent pre-mRNA degradation.⁴⁷ Conversely, demethylation of G3BP1 promoted SG assembly.⁴⁸ Prior work also documented that both HDAC6 and G3BP1 colocalized to SG where HDAC6 directly deacetylates G3BP1 to augment RNA binding and SG assembly. Moreover, the ability to form stress granules was impaired in HDAC6-deficient or HDAC-inhibited cells.⁴⁹ Given the evidence that HDAC6 promotes SG assembly through G3BP1, we tested whether HDAC6-deacetylated PRMT5 affects G3BP1 dimethylation.

To monitor dimethylation, endogenous G3BP1 was immunoprecipitated from cells overexpressing wild-type or H611A mutant HDAC6-Flag fusion proteins to alter PRMT5 acetylation. SDS-PAGE separation and visualization of symmetrical dimethylarginine documented an increase in symmetrical dimethylation of G3BP1, as well as an increase in PRMT5 acetylation, in cells overexpressing the inactive H161A mutant (Figure 6E, lane 3) compared to cells expressing wild-type HDAC6 or no protein (Figure 6E, lanes 1 and 2). Quantification from three independent trials showed significant increases in both PRMT5 acetylation and G3BP1 dimethylation only in the presence of inactive H611A mutant HDAC6 (Figure 6G). Consistent with the H4R3 assay and p53 data, the presence of mutant H611A HDAC6 increased PRMT5 activity through elevated acetylation, resulting in increased dimethylation of G3BP1. Taken together, the results indicate that PRMT5 acetylation enhances its methyltransferase activity to increased dimethylation of downstream substrates.

Given that HDAC6, G3BP1, and PRMT5 play a role in cell stress responses, including SG assembly,^47,49 the influence of ER stress on HDAC6-dependent PRMT5 acetylation and enzymatic activity was assessed. The symmetrical dimethylation of G3BP1 was again monitored in the presence of variably acetylated PRMT5 under ER stress induced by thapsigargin, similar to the p53 experiment. In contrast to the elevated PRMT5 acetylation observed only in H611A mutant HDAC6 expressing cells without stress conditions (Figure 6E, lanes 3), under stress conditions, elevated PRMT5 acetylation was observed in both untransfected and H611A mutant HDAC6 expressing cells (Figure 6F, lanes 1 and 3) but not wild-type HDAC6 expressing cells (Figure 6F, lane 2). These results suggested that ER stress elevated basal PRMT5 acetylation, which is consistent with previous reports where some proteins show increased acetylation under stress conditions.^50,51 Separately, endogenous G3BP1 was immunoprecipitated from the same lysates and probed for symmetrical dimethylation. Consistent with the PRMT5 acetylation status, elevated G3BP1 dimethylation was observed in both untransfected and H611A mutant HDAC6 expressing cells (Figure 6F, lanes 1 and 3) but not with the wild-type HDAC6 expressing cells (Figure 6F, lane 2), similar to the p53 data. Quantification from three independent trials showed significant increases in both PRMT5 acetylation and G3BP1 dimethylation in untreated and H611A HDAC6 expressing cells, compared to wild-type HDAC6 expressing cells (Figure 6H). The full data suggest that basal PRMT5 acetylation is elevated in stress conditions to increase downstream substrate dimethylation, which is dependent on HDAC6.

Altogether, the presence of ER stress resulted in elevated basal PRMT5 acetylation that was lost upon overexpression of HDAC6 (Figure 6D), suggesting reduced HDAC6 and/or elevated acetyltransferase activity with cell stress. In addition to ER stress, prior data showed arsenite-induced cell stress also resulted in elevated G3BP1 acetylation, which was HDAC6-dependent.⁵² Previous studies also showed G3BP1 dynamically shuttles between the cytoplasm and SG,⁵³ with acetylated G3BP1 localizing outside or on the periphery of SG assemblies.⁵² The combined data suggests a model where dynamic acetylation and/or methylation regulates movement of G3BP1 between the cytoplasm and SG during cell stress.

In summary, this study showed the successful application of trapping to HDAC6 substrate identification. Substrate discovery from two HDAC isoforms establishes that mutant trapping is a powerful, systematic, and unbiased substrate identification method for the HDAC family. Future substrate trapping with the HDAC family will uncover unanticipated roles of HDAC proteins in cell biology.

METHODS

Substrate Trapping, Immunoprecipitation, and Gel Analysis.

Prior to immunoprecipitation, anti-Flag M2 affinity agarose beads (20 μL bead slurry) were washed twice with cold TBS (Tris Buffered Saline, 500 μL; 20 mM Tris-Cl at pH 8.0, 150 mM NaCl). Lysates (1 mg total protein) from transfected cells containing either overexpressed wild-type or mutant HDAC6-Flag proteins were incubated with the prewashed anti-Flag M2 agarose beads at 4 °C overnight with rocking. When inhibitor competition was included, immunoprecipitations were performed in the presence of either a DMSO vehicle (<2%) or SAHA (100 μM in <2% DMSO). After immunoprecipitation, beads were washed three times by adding a lysis buffer (1 mL; 50 mM Tris-Cl at pH 8.0, 150 mM NaCl, 10% glycerol, and 0.5% triton-X100) containing high salt (500 mM NaCl) and centrifuging (5000 rcf, 1 min) to collect the beads. Proteins were assayed for deacetylase activity, as described below, or analyzed by gel methods after elution, as described here. Bound proteins were eluted with the SDS buffer (20 μL; 100 mM Tris-Cl at pH 6.8, 4% SDS, 20% glycerol, 0.008% bromophenol blue) by boiling at 95 °C for 5 min. The eluted proteins were mixed with 10% v/v β-mercaptoethanol and boiled at 95 °C for 2 min. As controls, lysates without immunoprecipitation (50 μg) were denatured with SDS loading dye (5 μL) by boiling at 95 °C for 2 min. Proteins were separated by 10% SDS-PAGE, transferred to a PVDF membrane (Immobilon P, Fischer Scientific), and immunoblotted with the monoclonal Flag M2, PRMT5, acetyl-tubulin, and GAPDH antibodies.

HDAC Activity Assay.

After immunoprecipitation as described, the deacetylase activities of immunoprecipitated wild-type and mutant HDAC6 were measured using the HDAC-Glo I/II Assay and Screening System according to company protocols. Briefly, HDAC-Glo I/II Substrate + Developer Reagent solution (1000:1, 2 μL) and HDAC-Glo I/II Buffer (48 μL) were incubated with the immunoprecipitated proteins on beads for 30 min with rocking at 700 rpm at RT. The luminescence signal was measured using a GeNiOS plus plate reader (Tecan) and a 96-well white plate. The mean percent deacetylase activity of the mutants compared to wild-type HDAC6 (set as 100%) was calculated from three independent trials with standard error.

Substrate Trapping for LC-MS/MS Analysis.

Substrate trapping for LC-MS/MS analysis was performed as previously described with minor changes.³¹ HEK293 cell lysates (4 mg of total protein) containing either wild-type or H611A mutant HDAC6-Flag proteins were immunoprecipitated in the presence of DMSO (<2%) or SAHA (100 μM in <2% DMSO) in the lysis buffer. Immunoprecipitation, washing, and elution were done as described.³¹ Proteins were desalted using 10% SDS-PAGE with electrophoresis for only 10 min at 100 V, which allowed the proteins to enter the first 1 cm of the separating layer. After Sypro Ruby staining (Biorad), proteins from each lane (1 cm × 2 cm) were excised and cut into 1 mm small pieces. Gel pieces were destained, reduced, and alkylated as previously described,³⁰ using the volume of 100 μL for each buffer.

LC-MS/MS analysis was performed at the Wayne State University and Karmanos Cancer Center Proteomics Core facility, as previously described.³¹ Label-free quantitation was performed using Maxquant software.³⁴ Enrichment values were calculated by dividing the intensity of peptide signals in the samples with the H611A mutant by the intensity value of that particular peptide of the wild-type sample for each independent trial. Potential substrates were then selected when peptides showed at least 2.0-fold enrichment in 2 out of 3 trials or all three trials (Figure S4) in the H611A mutant compared to the wild-type samples. To ensure active site binding of the hits, peptides were also removed that displayed higher intensity in the H611A mutant sample with SAHA compared to the H611A mutant sample without SAHA (DMSO only). The localization of hit proteins was assessed manually through the uniprot.org database.

Bioinformatics Analysis.

Functional annotation was performed using DAVID Bioinformatics Resources 6.8 software (https://david.ncifcrf.gov). The paxdb database (https://pax-db.org/) was used for abundance analysis. Interactome analysis was performed with the GeneMANIA application in Cytoscape 3.6.0 to understand the known physical protein–protein interactions.^36,37

In Vitro Deacetylation Assay.

HEK293 cells without transfection were harvested and lysed, as described. Immunoprecipitation was performed with lysate (1 mg of the total protein) and the HDAC6 antibody (5 μL) or lysate (500 μg of the total protein) and the PRMT5 antibody (1 μL) for 1 h at 4 °C with rocking. Protein A/G agarose beads (20 μL of bead slurry) were prepared by washing twice with cold TBS (Tris Buffered Saline, 500 μL; 20 mM Tris-Cl at pH 8.0, 150 mM NaCl). Then, each lysate–antibody mixture was separately added to the beads and incubated further overnight at 4 °C. After immunoprecipitation, each bead mixture was separately washed three times with the lysis buffer (1 mL). Immunoprecipitated HDAC6 was resuspended in the HDAC assay buffer (100 μL; 50 mM Tris-Cl, pH 8.0, 137 mM NaCl, 2.7 mM KCl, 1 mM MgCl₂) and preincubated in the absence (<2% DMSO) or presence of SAHA (2 mM in <2% DMSO) for 20 min at 30 °C while shaking at 750 rpm. Immunoprecipitated PRMT5 was resuspended in the HDAC assay buffer (50 μL), mixed with the preincubated HDAC6 immunoprecipitate, and then incubated at 30 °C for 1 h while shaking at 750 rpm. Finally, SDS-eluted proteins were separated by 10% SDS-PAGE, transferred to a PVDF membrane, and immunoblotted with acetyl lysine, HDAC6, and PRMT5 antibodies.

In Cell Deacetylation Assays.

For inhibitor treatment studies, HEK293 (2 × 10⁷) cells were treated with DMSO (0.02%) as the control, or SAHA (10 μM in 0.02% DMSO), Tubastatin A (10 μM in 0.02% DMSO), or SHI-1:2 (10 μM in 0.02% DMSO) for 24 h before harvesting and lysis, as described. Treated lysates (1 mg of the total proteins) or lysate containing either wild-type or H611A mutant HDAC6-Flag proteins (1 mg of the total proteins) were incubated with the PRMT5 antibody (1 μL) for 1 h at 4 °C with rocking to immunoprecipitate endogenous PRMT5. Then, the cell lysate–antibody mixture was added to prewashed Protein A/G agarose beads (20 μL bead slurry) and incubated further overnight at 4 °C with rocking. After immunoprecipitation, beads were washed three times with the lysis buffer (1 mL), and then proteins were eluted. Proteins were separated by 10% SDS PAGE, transferred to a PVDF membrane, and immunoblotted with acetyl lysine, PRMT5, and Flag antibodies.

PRMT5 Activity Assay.

HEK293 cells were transfected with the pCS2Myc-PRMT5 plasmid, as described. After 48 h of growth, cells were incubated with 0.02% DMSO as the control, or SAHA (10 μM in 0.02% DMSO), Tubastatin A (10 μM in 0.02% DMSO), or SHI-1:2 (10 μM in 0.02% DMSO) for an additional 24 h before harvesting and lysis, as described. Lysate (2 mg total proteins) was incubated with the Myc antibody (2.5 μg, 5 μL) for 2 h at 4 °C with rocking. The antibody–lysate mixture was incubated further with prewashed Protein A/G agarose beads (20 μL bead slurry) overnight at 4 °C with rocking. After washing three times with the lysis buffer (1 mL), the proteins were eluted using Myc-peptide (100 μg/mL, 40 μL) for 40 min at 25 °C with 550 rpm shaking. PRMT5 activity assays were performed using a chemiluminescence assay according to company protocols. Briefly, eluted protein from each immunoprecipitate (20 μL) in the PRMT5 buffer (7.5 μL) and water (17.5 μL) was added to the H4 peptide-bound wells in the precoated plate and incubated at RT for 2 h. After washing three times with the TBST buffer (200 μL) and blocking with the blocking buffer (100 μL, supplied by the manufacturer) for 10 min at RT, the diluted (1:100) primary antibody recognizing methylated R3 of Histone H4 (H4R3) in the blocking buffer was added (100 μL) to each well and incubated for 1 h at RT with shaking. After washing and blocking again, the diluted secondary HRP-labeled antibody (1:1,000) in the blocking buffer (100 μL) was added to each well and incubated for 30 min at RT with gentle rotation. After washing and blocking again, HRP chemiluminescent substrate A (50 μL) and HRP chemiluminescent substrate B (50 μL) were mixed on ice and added (100 μL) to each well. Chemiluminescence was monitored immediately using the GeNiOS plus plate reader.

Symmetrical Dimethylation of Downstream Substrates by PRMT5.

HEK293 cells were transfected with Flag-tagged HDAC6 and H611A mutant expression plasmids, as described. When ER stress was induced, cells were treated with thapsigargin (1 μM final concentration) for 5 h before harvest and lysis, as described. Lysates (1 mg of the total protein) were incubated with the PRMT5 antibody (1 μL), G3BP1 antibody (20 μL), or p53 antibody (20 μL) at 4 °C for 1 h. PRMT5 inhibition before p53 immunoprecipitation was performed by treating the cells with EPZ015666 (5 μM) for 12 h. Then, each lysate–antibody mixture was separately added to prewashed Protein A/G agarose beads (20 μL bead slurry) and incubated overnight at 4 °C with rocking. After immunoprecipitation, beads were washed three times with the lysis buffer (1 mL). Proteins were eluted, separated by 10% SDS-PAGE, and transferred to a PVDF membrane, as described. The membranes were immunoblotted with acetyl lysine, PRMT5, SDMA, G3BP1, and p53 antibodies.

Quantification and Statistical Analysis.

All the data are presented as mean ± standard error from three independent trials. Data were plotted using KaleidaGraph (Version 4.1), and statistical analysis was performed using GraphPad Prism software (Version 7.0a).