This article was downloaded by: [New York University] On: 21 July 2015, At: 07:15 Publisher: Taylor & Francis Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: 5 Howick Place, London, SW1P 1WG
RNA Biology Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/krnb20
Cryptic RNA-binding by PRC2 components EZH2 and SUZ12 Juan G. Betancur a
ab
ab
& Yukihide Tomari
Institute of Molecular and Cellular Biosciences
b
Department of Medical Genome Sciences, The University of Tokyo, Bunkyo-ku, Tokyo 113-0032, Japan Accepted author version posted online: 15 Jul 2015.
Click for updates To cite this article: Juan G. Betancur & Yukihide Tomari (2015): Cryptic RNA-binding by PRC2 components EZH2 and SUZ12, RNA Biology, DOI: 10.1080/15476286.2015.1069463 To link to this article: http://dx.doi.org/10.1080/15476286.2015.1069463
Disclaimer: This is a version of an unedited manuscript that has been accepted for publication. As a service to authors and researchers we are providing this version of the accepted manuscript (AM). Copyediting, typesetting, and review of the resulting proof will be undertaken on this manuscript before final publication of the Version of Record (VoR). During production and pre-press, errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal relate to this version also.
PLEASE SCROLL DOWN FOR ARTICLE Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and should be independently verified with primary sources of information. Taylor and Francis shall not be liable for any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of the Content. This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http:// www.tandfonline.com/page/terms-and-conditions
Cryptic RNA-binding by PRC2 components EZH2 and SUZ12 Juan G. Betancur1,2,3, and Yukihide Tomari1,2,4 1
Institute of Molecular and Cellular Biosciences; 2Department of Medical Genome Sciences, The University of Tokyo, Bunkyo-ku, Tokyo 113-0032, Japan
3
Current address: Laboratory for Developmental Genetics, RIKEN-IMS, Tsurumi-ku,
Downloaded by [New York University] at 07:15 21 July 2015
Yokohama, Kanagawa 230-0045, Japan; 4Correspondence should be addressed to Y.T. Phone: +81-3-5841-7839. Fax: +81-3-5841-8485 E-mail: [email protected] Running title: Cryptic RNA-binding by PRC2 Keywords: PRC2, EZH2, SUZ12, lncRNA, Xist, HOTAIR
Abstract Polycomb repressive complex 2 (PRC2) is a histone-modifying complex that di/trimethylates histone H3 lysine 27 (H3K27), a mark of transcriptionally repressed chromatin. However, how PRC2 is specifically recruited to its target loci remains controversial, although it has been postulated that long non-coding RNAs (lncRNAs) can function as guides. Here we purified individual components of PRC2 from human cultured cells and found that EZH2 and SUZ12 can directly bind to RNAs. In agreement with recent evidence, our results support the notion that these two PRC2 subunits have RNA-binding activity, with general preference for longer RNAs. However, the length alone does not explain their cryptic substrate preference. Our data highlight the difficulty
of
characterizing
the
RNA-binding 1
activity
of
PRC2.
Introduction Polycomb repressive complex 2 (PRC2) is a member of the Polycomb group (PcG) protein family that di/tri-methylates histone H3 lysine 27 (H3K27), a mark of transcriptionally repressed chromatin
1
. The complex is composed of 4 core
components: EED, EZH2, RbAp46 or RbAp48 and SUZ12
2, 3
, which are orthologs of
the Drosophila Extra sexcombs (Esc), Enhancer of zeste E(z), a nuclear remodeling factor (Nurf55) and Suppressor of zeste 12 (Su(z)12), respectively 4. EZH2 is the
Downloaded by [New York University] at 07:15 21 July 2015
catalytic subunit of the complex 2, while the other subunits are required for maximum methyltransferase activity 5. In addition, EED has been reported to recognize trimethylated H3K27 6, and RbAp46/48 is a histone chaperone that binds to histone H3 and H4 7. There are also accessory proteins that co-purify with PRC2 at sub-equimolar ratios, including Polycomblike proteins (PCLs) 8, AEBP2
9
and JARID2
10
. AEBP2
9
and JARID2 11, 12 can bind DNA directly. In flies PRC2 (and other PcG complexes) binds to DNA elements known as polycomb response elements (PRE) with the aid of several different co-factors. However, many of the genomic binding sites for such factors do not overlap with PcGbound regions, but instead with active epigenetic marks, which indicates that those factors are not sufficient for PRC2 recruitment on their own
4, 13, 14
. In addition, in
mammals, no PRE-like elements have been described, although it is known that many PRC2 binding sites are C+G rich, or correspond to CpG islands,
4, 15, 16
.
Instead of requiring single recruiting cofactors, specific recruitment of PRC2 to its target loci is more likely the result of a combinatorial effect of several independent events. Those events might include interactions engaged by EED and RbAp48 with
2
histones
6, 7
, binding to DNA by JARID2 and AEBP2
9, 11, 12
, and as recent reports
suggest, the contribution of long non-coding RNAs (lncRNAs) 17, 18. Two of the lncRNAs that have consistently been shown to interact with PRC2 are the X-inactive-specific transcript (Xist) and Hox antisense intergenic RNA (HOTAIR). In humans Xist is a ~19 kb transcript transcribed from the X inactivation center in the X chromosome 19. It is required for X chromosome inactivation (XCI), a process in which one of the two X chromosomes is randomly selected to be silenced during early
Downloaded by [New York University] at 07:15 21 July 2015
development to equalize gene expression from the X chromosome between male and female mammals
20, 21
. XCI requires the catalytic activity of PcG proteins, in particular
the H3K27 methyltransferase activity of PRC2 22, and the H2AK119 mono ubiquitinase activity of PRC1
23
that together effectively silence gene expression from the target
chromosome. PRC2 directly binds Xist, which is tethered to the inactive X chromosome providing a platform for spreading H3K27me3 in cis
17
24, 25
,
, although alternative models
have also been proposed where the direct interaction between PRC2 and Xist is dispensable for XCI 26, 27. The region of Xist bound by PRC2 has been mapped to the 5’ end of the transcript, where there is a stretch of 8.5 repeats of a ~28 nt sequence, known as the A-repeat domain
17, 28
. Each of the 8.5 repeated sequences adopts a double stem
loop structure (Figure 1B), although there is also some evidence that only hairpin 1 of the double stem loop forms, and hairpin 2 is instead involved in the formation of interrepeat duplex interactions
29
. From in vitro experiments, it has been postulated that a
single repeat of the A-repeat domain might be enough for direct binding of PRC2 to Xist
17, 30
, and in vivo Xist with a deletion of the A-repeat domain spreads on the
3
inactive X chromosome, but fails to induce inactivation 28, which could be related to an inability to bind to PRC2. In contrast to Xist, HOTAIR acts in trans to regulate thousands of genes
18, 31-33
.
HOTAIR is proposed to directly recruit PRC2 through a 300 nt domain located at its 5’ end, which results in the tri-methylation of H3K27 and silencing of the target loci. Initial evidence of the direct interaction between Xist and HOTAIR with PRC2 came from electrophoretic mobility shift assays (EMSA) and RNA pull-down experiments Downloaded by [New York University] at 07:15 21 July 2015
with Xist and HOTAIR fragments. While some experiments concluded that SUZ12, but not EZH2 is responsible for the RNA-binding activity of the complex
30
, the opposite
has also been shown by different groups 17, 18, 34, 35. More recently, Cifuentes-Rojas et al. showed that both EZH2 and SUZ12 can bind to RNAs
36
. The source of these
contradictory results regarding the RNA-binding activity of PRC2 is unclear but may be due to the sensitivity of complex formation to the conditions used for the RNA binding reaction and electrophoresis as shown below and by others
37
. Moreover, although the
double stem loop structure of A-repeat has been postulated to be the PRC2-bound element of Xist, other lncRNAs (e.g. HOTAIR) lack such repeats, and therefore it is still an open question whether PRC2 makes use of different mechanism to bind to different RNAs or not. Here, we purified recombinant proteins of the four core component of PRC2 from human cells to warrant their integrity as much as possible, and performed a series of in vitro RNA-binding experiments. In agreement with recent reports we show that two PRC2 components, EZH2 and SUZ12, have cryptic RNA-binding activity, with general preference for longer RNAs. However, the length alone does not explain their substrate
4
preference. Our data highlight the difficulty of characterizing the RNA-binding activity of PRC2. EZH2 and SUZ12 are RNA-binding proteins To revisit which is the RNA-binding subunit of the PRC2 complex we used an RNA fragment that corresponds to the first repeat of the A-repeat domain of Xist (herein called 1 rep Wt RNA) (Figure 1A, B). This short fragment has been previously shown to bind to either EZH2 or SUZ12 by different groups
17, 30, 35
. We purified each of the 4
Downloaded by [New York University] at 07:15 21 July 2015
core PRC2 components, EED, EZH2, RbAP48 and SUZ12 from HEK293T cells overexpressing an SBP (streptavidin-binding peptide) tagged version of each protein. After a single purification step we obtained a relatively clean preparation of each protein (Supplementary Figure 1A) and verification by western blot showed that there was minimal cross contamination with endogenous PRC2 components, except for SBPSUZ12 that was co-purified with endogenous RbAp48 in our experimental condition (Supplementary Figure 1B). Electrophoretic mobility shift assays (EMSA) showed binding of SUZ12 to radiolabeled 1 rep Wt RNA but no apparent RNA-binding of other components, which agreed with at least one of the previous reports about the RNAbinding activity of PRC2 30 (Figure 1C). However, during the course of experiments, we noticed that the degree of RNA-binding by SUZ12 largely fluctuates, depending on the buffer and electrophoresis conditions of EMSA (data not shown), and could not make an unambiguous conclusion. To more directly evaluate RNA-binding by PRC2 components, we next assembled the same reaction but this time crosslinked the samples with short-wave UV light, and then loaded them into an SDS-PAGE. The molecular weight of 1 rep Wt RNA is ~9 kDa and covalent binding of the RNA is not expected to radically affect the migration 5
pattern of the RNA-protein complex. Therefore this methodology allows the identification of RNA-binding proteins according to their apparent molecular weight. In contrast to the EMSA results, in addition to a band of the expected molecular weight detected with SUZ12 we also detected a complex that corresponds to EZH2 complexed with 1 rep Wt RNA (Figure 1D). Note that SBP-EZH2 and SBP-SUZ12 migrate just above 100 kDa in an SDS-PAGE (Supplementary Figure 1A) and are slightly shifted upwards when crosslinked to 1 rep Wt (Figure 1D). No complexes were detected with
Downloaded by [New York University] at 07:15 21 July 2015
EED, RbAp48 or a GFP control. Thus, not only SUZ12 but also EZH2 can directly bind to 1 rep Wt RNA. Together these results support the findings by Lee and colleagues 36 and reconcile the previous contradictory reports that had argued by using EMSA that either EZH2 or SUZ12, but not both, possesses RNA-binding activity. Differences in the methodologies used for EMSA and/or protein expression and purification might be responsible for the previous contradictions, because RNA-binding by EZH2, which was evident in the UV crosslinking assay, could not be detected by EMSA in our experimental condition. EZH2 and SUZ12 have only a slight preference for structured 1 rep RNA It has been suggested that PRC2 specifically recognizes the double stem loop structures present in the A-repeat domain of Xist and other short ncRNAs with similar structures 17, 30, 35. We constructed a mutant version of 1 rep RNA that is predicted not to form any stable secondary structure (1 rep Mut RNA), similar to one previously used
17
.
However, this RNA was also shifted by SUZ12 EMSA (Figure 1E) and was crosslinked to both EZH2 and SUZ12 (Figure 1F).
6
To directly assess the affinity of both proteins for 1 rep Wt and 1 rep Mut RNAs we performed competition assays with increasing concentrations of non-radiolabeled wildtype and mutant RNAs (Figure 1G). Both RNAs competed with radiolabeled 1 rep Wt and Mut RNAs for crosslinking to EZH2 and SUZ12, but the wild type RNA was a slightly stronger competitor. These data show that both proteins have a small preference for 1 rep Wt over Mut RNAs, but also suggest that the exact sequence and/or structure of 1 rep Wt RNA is not an absolute requirement for binding to EZH2 or SUZ12.
Downloaded by [New York University] at 07:15 21 July 2015
EZH2 and SUZ12 partially but not fully discriminate RNA-binding partners One of several Xist secondary structures predicts that each of the 8.5 repeats of the human A-repeat domain adopts a double stem loop structure similar to 1 rep Wt RNA 28 (Supplementary figure 2), and therefore the repeats can in theory act cooperatively to bind to EZH2 and SUZ12. For comparison, 2 rep, 4 rep and 8.5 rep Wt RNA fragments were prepared (Figure 2A) and mutations in each of the double stem loop repeats were introduced at approximately the same positions as in 1 rep Mut RNA to disrupt each double stem loop structure and produce 2 rep, 4 rep and 8.5 rep Mut RNAs (Supplementary table 2 and Supplementary figure 2). UV crosslinking is not suitable to measure binding of proteins to 2, 4, and 8.5 rep Mut RNAs because of their large molecular weights, and instead we utilized the doublefilter-binding assay. In this methodology, increasing amounts of recombinant proteins are incubated with each RNA, and samples are applied using vacuum to two superposed membranes: a nitrocellulose membrane on top retains protein-bound RNA, while free RNA passes through and is captured by a positively charged nylon membrane below (Figure 2B) 38, 39. The fraction of bound RNA is then calculated.
7
Binding of 1 rep Wt and Mut RNAs to EZH2 and SUZ12 was very weak and ~95% of the RNA remained unbound, even at the highest protein concentration. In contrast, when 2 rep Wt, 4 rep Wt and 8.5 rep Wt RNAs were used at equimolar concentration, the fraction of bound RNA greatly increased and reached near saturation in the case of 8.5 rep Wt RNA (Figure 2C, D). These results suggest that EZH2 and SUZ12 prefer to bind longer RNAs. Mutations in 4 rep and 8.5 rep RNAs did not overtly affect the affinity to EZH2 or
Downloaded by [New York University] at 07:15 21 July 2015
SUZ12, agreeing with the UV crosslinking data that suggest the exact sequence and/or structure of the A-repeat domain is not necessary for these proteins to bind. However, mutations in 2 rep RNA almost completely abolished binding of EZH2 and SUZ12, raising the possibility that they are capable of distinguishing some characteristics of the A-repeat domain in particular RNA contexts. To further explore the cryptic RNA-binding activity of EZH2 and SUZ12, we also assessed the binding of the two proteins to HOTAIR fragments of different lengths (Figure 2E). Unlike Xist, HOTAIR does not contain any apparent sequence or structure repeats, but a ~300 nt long sequence at the 5’ end (herein called HOTAIR-A) was previously postulated to contain a PRC2 binding element by protein pull-down assays by immobilized RNA fragments
18
. In contrast, a ~600 nt sequence at the 3’ end of
HOTAIR has been shown to bind to Lysin-specific demethylase 1 (LSD1) but not to PRC2
18
. In our double-filter-binding assay, full length HOTAIR (HOTAIR-FL) and
HOTAIR-A were bound by both proteins with similar affinity (Figure 2F, G). To test a variety of RNA lengths we deleted ~200 nt at the 5’ end of HOTAIR-A (the remaining 104 nt region called HOTAIR-B) which did not abolish binding to EZH2 or SUZ12, but a more shortened 26 nt fragment (HOTAIR-C) showed no detectable binding. 8
Interestingly, a 102 nt fragment at the 3’ end of the full-length transcript (HOTAIR-D) did not bind to EZH2 or SUZ12, even though its length was similar to HOTAIR-B (Figure 2F, G), again implying that these proteins may have some ability to distinguish RNAs. To further assess the RNA-binding activity of EZH2 and SUZ12, we tested a set of RNA fragments of different lengths that correspond to portions of the GFP mRNA (Figure 2H). Although shorter RNA fragments (28 nt and 102 nt) showed only little
Downloaded by [New York University] at 07:15 21 July 2015
binding, longer RNAs (318 nt and 645 nt long) bound to EZH2 and SUZ12 with affinities similar to HOTAIR (Figure 2I, J), even though such RNAs are not expected to play any regulatory role of PRC2 function. In contrast, EED, RbAP48 or a control GFP protein showed no detectable binding of Xist, HOTAIR or GFP RNA fragments (Figure 2K, L, M). Taken altogether, we could make a clear conclusion that the two PRC2 components, EZH2 and SUZ12, have intrinsic RNA-binding activities with general preference for longer RNAs and lower threshold for substrate RNA size of around 100 nt. However, their substrate preference was largely cryptic, hindering further characterization in vitro. PRC2, a dual RNA-binding complex Using purified proteins expressed in human cells the present work establishes that PRC2 has at least two components with RNA-binding activity, in agreement with a recent report by Lee and colleagues who observed similar results with EZH2 and SUZ12 prepared from insect cells
36
. A region between aminoacids 342-368 was
previously identified as the RNA binding domain of EZH2 domain of SUZ12 still remains unknown. 9
34
, but the RNA binding
In our experimental setup at the tested concentration, the interaction between EZH2 and 1 rep RNA could not be detected by EMSA. This might be an indication that the EZH2-RNA interaction was sensitive to binding and electrophoresis conditions and might be a reason why initial reports failed to identify the dual RNA-binding activity of the complex. A recent paper also reported that PRC2-RNA interactions can be largely affected by the experimental conditions 37. Therefore EMSA results should be analyzed with caution.
Downloaded by [New York University] at 07:15 21 July 2015
Most initial studies of the RNA-binding activity of PRC2 focused on the interaction of PRC2 subunits with 1 rep RNA
17, 30, 35
. This led to the general assumption that the
sequence and structural elements required for specific binding to PRC2 could be contained within such a small RNA. This was further reinforced with the finding that short RNAs (50–200nt long) that resemble the double stem loop structure of 1 rep RNA are transcribed from many PRC2 target genes, are bound by SUZ12 and cause gene repression in cis
30
. However, in our crosslinking and competition experiments (Figure
1D, F, G), the double stem loop structure of 1 rep RNA was not an absolute requirement for binding to EZH2 and SUZ12, and in fact both proteins barely bound RNAs of such short length, regardless of their origin (26–28 nucleotide fragments in Figure 2). Instead, in the case of A-repeat affinity increased gradually with the addition of repeats, with maximum binding observed with 8.5 rep RNAs (Figure 2C, D). However, the effect of increasing RNA length was less clear with the HOTAIR RNA series. Therefore, as the number of repeats increases, an as-yet-unidentified RNA element may act in an additive manner to alter the structure of A-repeat, resulting in enhanced binding by EZH2 and SUZ12; or alternatively A-repeat may serve as a platform for cooperative protein
10
binding. Nonetheless, additional investigations are required to understand what makes the A-repeat domain of Xist special. Our finding that EZH2 and SUZ12 are also capable of binding to unrelated GFP mRNA is in agreement with other reports that EZH2 and SUZ12 PRC2
37, 40, 41
36
and assembled
bind RNAs promiscuously. In fact, the Kd of the PRC2–A-repeat
interaction is only within one order of magnitude smaller than the Kd for unrelated RNA substrates from sources ranging from bacteria to insects
37
. In addition, CLIP
Downloaded by [New York University] at 07:15 21 July 2015
experiments for EZH2 demonstrated that in vivo this protein binds to introns and to the 5’ end of nascent RNAs in a promiscuous manner, with no apparent consensus RNA motif 41, 42. Nevertheless, it has been reported that EED can modulate the RNA-binding activity of EZH2 and SUZ12 to make it more specific
36
, and our data show that the
isolated PRC2 components retain a certain level of substrate preference, as evidenced by the ability of EZH2 and SUZ12 to distinguish between the 2 rep Wt and Mut RNAs as well as the ~102 nt HOTAIR RNA sequences inside or outside of the previously defined PRC2-binding region at the 5’ end 18. However, narrowing down the specificity determinants is inherently challenging due to their cryptic nature and inability to bind short, fragmented RNAs. Finally, although the significance of the RNA-binding by two different components of PRC2 awaits further investigation, it might be an indication of the complexity of signals that are required for the timely and precise deposition of H3K27me2/3. RNAs may be transferred between EZH2 and SUZ12 or both subunits might bind cooperatively to target RNAs, or even to different classes of RNAs simultaneously. For precise recruitment fully assembled PRC2, additional co-factors and a specific intranuclear microenvironment might be required. In fact, recent studies have illustrated the 11
complexity of the PRC2-RNA interplay. On one hand, EED modulates the RNAbinding activity of EZH2 and SUZ12
36
, and JARID2, a PRC2 partner protein with
RNA-binding activity, can compete with both proteins for RNA partners to modulate the methylatransferase activity of PRC2
36, 43
. However, a binding site of JARID2 on
Xist has been mapped to a non-overlapping region downstream of the A-repeat domain 44
and therefore how the multiple PRC2-associated RNA-binding proteins cooperate
and/or compete for binding to Xist still remains to be investigated. On the other hand,
Downloaded by [New York University] at 07:15 21 July 2015
RNA-mediated recruitment of PRC2 is enhanced by the recognition of pre-existing H3K27me3 marks by EED at repressed loci, further activating the methylatransferase activity of PRC2 and amplifying the repression signal
6, 40, 41
. All these elements add
additional layers of complexity to the regulation of gene expression mediated by PRC2 and highlight the difficulty of characterizing the RNA-binding activity of PRC2 in vitro.
Materials and methods Protein purification HEK293T cells were transfected with pCAGEN-SBP-EZH2, -SUZ12, pCAGENSBP-ps-EED, -RbAp48 or pEFh-SBP-GFP to over express streptavidin binding protein (SBP)-tagged proteins. Nuclear lysates were prepared and proteins were purified using Streptavidin Sepharose High Performance beads (GE Healthcare) and eluted with biotin. Proteins were concentrated using VivaSpin 2 ml columns (Sartorius) and quantified by CBB staining using BSA standard curves. See supplemental material for details. RNAs Primers for synthesis of DNA templates and for in vitro transcription, as well as the sequences of all RNAs used in this study are available as supplemental material. DNA 12
templates were in vitro transcribed using T7-scribe standard RNA IVT kit (Cellscript) and gel purified. Short RNAs 3).
Downloaded by [New York University] at 07:15 21 July 2015
Final concentration of radiolabeled RNA in all experiments: 1 nM.
24
RNA Biology Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/krnb20
Cryptic RNA-binding by PRC2 components EZH2 and SUZ12 Juan G. Betancur a
ab
ab
& Yukihide Tomari
Institute of Molecular and Cellular Biosciences
b
Department of Medical Genome Sciences, The University of Tokyo, Bunkyo-ku, Tokyo 113-0032, Japan Accepted author version posted online: 15 Jul 2015.
Click for updates To cite this article: Juan G. Betancur & Yukihide Tomari (2015): Cryptic RNA-binding by PRC2 components EZH2 and SUZ12, RNA Biology, DOI: 10.1080/15476286.2015.1069463 To link to this article: http://dx.doi.org/10.1080/15476286.2015.1069463
Disclaimer: This is a version of an unedited manuscript that has been accepted for publication. As a service to authors and researchers we are providing this version of the accepted manuscript (AM). Copyediting, typesetting, and review of the resulting proof will be undertaken on this manuscript before final publication of the Version of Record (VoR). During production and pre-press, errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal relate to this version also.
PLEASE SCROLL DOWN FOR ARTICLE Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and should be independently verified with primary sources of information. Taylor and Francis shall not be liable for any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of the Content. This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http:// www.tandfonline.com/page/terms-and-conditions
Cryptic RNA-binding by PRC2 components EZH2 and SUZ12 Juan G. Betancur1,2,3, and Yukihide Tomari1,2,4 1
Institute of Molecular and Cellular Biosciences; 2Department of Medical Genome Sciences, The University of Tokyo, Bunkyo-ku, Tokyo 113-0032, Japan
3
Current address: Laboratory for Developmental Genetics, RIKEN-IMS, Tsurumi-ku,
Downloaded by [New York University] at 07:15 21 July 2015
Yokohama, Kanagawa 230-0045, Japan; 4Correspondence should be addressed to Y.T. Phone: +81-3-5841-7839. Fax: +81-3-5841-8485 E-mail: [email protected] Running title: Cryptic RNA-binding by PRC2 Keywords: PRC2, EZH2, SUZ12, lncRNA, Xist, HOTAIR
Abstract Polycomb repressive complex 2 (PRC2) is a histone-modifying complex that di/trimethylates histone H3 lysine 27 (H3K27), a mark of transcriptionally repressed chromatin. However, how PRC2 is specifically recruited to its target loci remains controversial, although it has been postulated that long non-coding RNAs (lncRNAs) can function as guides. Here we purified individual components of PRC2 from human cultured cells and found that EZH2 and SUZ12 can directly bind to RNAs. In agreement with recent evidence, our results support the notion that these two PRC2 subunits have RNA-binding activity, with general preference for longer RNAs. However, the length alone does not explain their cryptic substrate preference. Our data highlight the difficulty
of
characterizing
the
RNA-binding 1
activity
of
PRC2.
Introduction Polycomb repressive complex 2 (PRC2) is a member of the Polycomb group (PcG) protein family that di/tri-methylates histone H3 lysine 27 (H3K27), a mark of transcriptionally repressed chromatin
1
. The complex is composed of 4 core
components: EED, EZH2, RbAp46 or RbAp48 and SUZ12
2, 3
, which are orthologs of
the Drosophila Extra sexcombs (Esc), Enhancer of zeste E(z), a nuclear remodeling factor (Nurf55) and Suppressor of zeste 12 (Su(z)12), respectively 4. EZH2 is the
Downloaded by [New York University] at 07:15 21 July 2015
catalytic subunit of the complex 2, while the other subunits are required for maximum methyltransferase activity 5. In addition, EED has been reported to recognize trimethylated H3K27 6, and RbAp46/48 is a histone chaperone that binds to histone H3 and H4 7. There are also accessory proteins that co-purify with PRC2 at sub-equimolar ratios, including Polycomblike proteins (PCLs) 8, AEBP2
9
and JARID2
10
. AEBP2
9
and JARID2 11, 12 can bind DNA directly. In flies PRC2 (and other PcG complexes) binds to DNA elements known as polycomb response elements (PRE) with the aid of several different co-factors. However, many of the genomic binding sites for such factors do not overlap with PcGbound regions, but instead with active epigenetic marks, which indicates that those factors are not sufficient for PRC2 recruitment on their own
4, 13, 14
. In addition, in
mammals, no PRE-like elements have been described, although it is known that many PRC2 binding sites are C+G rich, or correspond to CpG islands,
4, 15, 16
.
Instead of requiring single recruiting cofactors, specific recruitment of PRC2 to its target loci is more likely the result of a combinatorial effect of several independent events. Those events might include interactions engaged by EED and RbAp48 with
2
histones
6, 7
, binding to DNA by JARID2 and AEBP2
9, 11, 12
, and as recent reports
suggest, the contribution of long non-coding RNAs (lncRNAs) 17, 18. Two of the lncRNAs that have consistently been shown to interact with PRC2 are the X-inactive-specific transcript (Xist) and Hox antisense intergenic RNA (HOTAIR). In humans Xist is a ~19 kb transcript transcribed from the X inactivation center in the X chromosome 19. It is required for X chromosome inactivation (XCI), a process in which one of the two X chromosomes is randomly selected to be silenced during early
Downloaded by [New York University] at 07:15 21 July 2015
development to equalize gene expression from the X chromosome between male and female mammals
20, 21
. XCI requires the catalytic activity of PcG proteins, in particular
the H3K27 methyltransferase activity of PRC2 22, and the H2AK119 mono ubiquitinase activity of PRC1
23
that together effectively silence gene expression from the target
chromosome. PRC2 directly binds Xist, which is tethered to the inactive X chromosome providing a platform for spreading H3K27me3 in cis
17
24, 25
,
, although alternative models
have also been proposed where the direct interaction between PRC2 and Xist is dispensable for XCI 26, 27. The region of Xist bound by PRC2 has been mapped to the 5’ end of the transcript, where there is a stretch of 8.5 repeats of a ~28 nt sequence, known as the A-repeat domain
17, 28
. Each of the 8.5 repeated sequences adopts a double stem
loop structure (Figure 1B), although there is also some evidence that only hairpin 1 of the double stem loop forms, and hairpin 2 is instead involved in the formation of interrepeat duplex interactions
29
. From in vitro experiments, it has been postulated that a
single repeat of the A-repeat domain might be enough for direct binding of PRC2 to Xist
17, 30
, and in vivo Xist with a deletion of the A-repeat domain spreads on the
3
inactive X chromosome, but fails to induce inactivation 28, which could be related to an inability to bind to PRC2. In contrast to Xist, HOTAIR acts in trans to regulate thousands of genes
18, 31-33
.
HOTAIR is proposed to directly recruit PRC2 through a 300 nt domain located at its 5’ end, which results in the tri-methylation of H3K27 and silencing of the target loci. Initial evidence of the direct interaction between Xist and HOTAIR with PRC2 came from electrophoretic mobility shift assays (EMSA) and RNA pull-down experiments Downloaded by [New York University] at 07:15 21 July 2015
with Xist and HOTAIR fragments. While some experiments concluded that SUZ12, but not EZH2 is responsible for the RNA-binding activity of the complex
30
, the opposite
has also been shown by different groups 17, 18, 34, 35. More recently, Cifuentes-Rojas et al. showed that both EZH2 and SUZ12 can bind to RNAs
36
. The source of these
contradictory results regarding the RNA-binding activity of PRC2 is unclear but may be due to the sensitivity of complex formation to the conditions used for the RNA binding reaction and electrophoresis as shown below and by others
37
. Moreover, although the
double stem loop structure of A-repeat has been postulated to be the PRC2-bound element of Xist, other lncRNAs (e.g. HOTAIR) lack such repeats, and therefore it is still an open question whether PRC2 makes use of different mechanism to bind to different RNAs or not. Here, we purified recombinant proteins of the four core component of PRC2 from human cells to warrant their integrity as much as possible, and performed a series of in vitro RNA-binding experiments. In agreement with recent reports we show that two PRC2 components, EZH2 and SUZ12, have cryptic RNA-binding activity, with general preference for longer RNAs. However, the length alone does not explain their substrate
4
preference. Our data highlight the difficulty of characterizing the RNA-binding activity of PRC2. EZH2 and SUZ12 are RNA-binding proteins To revisit which is the RNA-binding subunit of the PRC2 complex we used an RNA fragment that corresponds to the first repeat of the A-repeat domain of Xist (herein called 1 rep Wt RNA) (Figure 1A, B). This short fragment has been previously shown to bind to either EZH2 or SUZ12 by different groups
17, 30, 35
. We purified each of the 4
Downloaded by [New York University] at 07:15 21 July 2015
core PRC2 components, EED, EZH2, RbAP48 and SUZ12 from HEK293T cells overexpressing an SBP (streptavidin-binding peptide) tagged version of each protein. After a single purification step we obtained a relatively clean preparation of each protein (Supplementary Figure 1A) and verification by western blot showed that there was minimal cross contamination with endogenous PRC2 components, except for SBPSUZ12 that was co-purified with endogenous RbAp48 in our experimental condition (Supplementary Figure 1B). Electrophoretic mobility shift assays (EMSA) showed binding of SUZ12 to radiolabeled 1 rep Wt RNA but no apparent RNA-binding of other components, which agreed with at least one of the previous reports about the RNAbinding activity of PRC2 30 (Figure 1C). However, during the course of experiments, we noticed that the degree of RNA-binding by SUZ12 largely fluctuates, depending on the buffer and electrophoresis conditions of EMSA (data not shown), and could not make an unambiguous conclusion. To more directly evaluate RNA-binding by PRC2 components, we next assembled the same reaction but this time crosslinked the samples with short-wave UV light, and then loaded them into an SDS-PAGE. The molecular weight of 1 rep Wt RNA is ~9 kDa and covalent binding of the RNA is not expected to radically affect the migration 5
pattern of the RNA-protein complex. Therefore this methodology allows the identification of RNA-binding proteins according to their apparent molecular weight. In contrast to the EMSA results, in addition to a band of the expected molecular weight detected with SUZ12 we also detected a complex that corresponds to EZH2 complexed with 1 rep Wt RNA (Figure 1D). Note that SBP-EZH2 and SBP-SUZ12 migrate just above 100 kDa in an SDS-PAGE (Supplementary Figure 1A) and are slightly shifted upwards when crosslinked to 1 rep Wt (Figure 1D). No complexes were detected with
Downloaded by [New York University] at 07:15 21 July 2015
EED, RbAp48 or a GFP control. Thus, not only SUZ12 but also EZH2 can directly bind to 1 rep Wt RNA. Together these results support the findings by Lee and colleagues 36 and reconcile the previous contradictory reports that had argued by using EMSA that either EZH2 or SUZ12, but not both, possesses RNA-binding activity. Differences in the methodologies used for EMSA and/or protein expression and purification might be responsible for the previous contradictions, because RNA-binding by EZH2, which was evident in the UV crosslinking assay, could not be detected by EMSA in our experimental condition. EZH2 and SUZ12 have only a slight preference for structured 1 rep RNA It has been suggested that PRC2 specifically recognizes the double stem loop structures present in the A-repeat domain of Xist and other short ncRNAs with similar structures 17, 30, 35. We constructed a mutant version of 1 rep RNA that is predicted not to form any stable secondary structure (1 rep Mut RNA), similar to one previously used
17
.
However, this RNA was also shifted by SUZ12 EMSA (Figure 1E) and was crosslinked to both EZH2 and SUZ12 (Figure 1F).
6
To directly assess the affinity of both proteins for 1 rep Wt and 1 rep Mut RNAs we performed competition assays with increasing concentrations of non-radiolabeled wildtype and mutant RNAs (Figure 1G). Both RNAs competed with radiolabeled 1 rep Wt and Mut RNAs for crosslinking to EZH2 and SUZ12, but the wild type RNA was a slightly stronger competitor. These data show that both proteins have a small preference for 1 rep Wt over Mut RNAs, but also suggest that the exact sequence and/or structure of 1 rep Wt RNA is not an absolute requirement for binding to EZH2 or SUZ12.
Downloaded by [New York University] at 07:15 21 July 2015
EZH2 and SUZ12 partially but not fully discriminate RNA-binding partners One of several Xist secondary structures predicts that each of the 8.5 repeats of the human A-repeat domain adopts a double stem loop structure similar to 1 rep Wt RNA 28 (Supplementary figure 2), and therefore the repeats can in theory act cooperatively to bind to EZH2 and SUZ12. For comparison, 2 rep, 4 rep and 8.5 rep Wt RNA fragments were prepared (Figure 2A) and mutations in each of the double stem loop repeats were introduced at approximately the same positions as in 1 rep Mut RNA to disrupt each double stem loop structure and produce 2 rep, 4 rep and 8.5 rep Mut RNAs (Supplementary table 2 and Supplementary figure 2). UV crosslinking is not suitable to measure binding of proteins to 2, 4, and 8.5 rep Mut RNAs because of their large molecular weights, and instead we utilized the doublefilter-binding assay. In this methodology, increasing amounts of recombinant proteins are incubated with each RNA, and samples are applied using vacuum to two superposed membranes: a nitrocellulose membrane on top retains protein-bound RNA, while free RNA passes through and is captured by a positively charged nylon membrane below (Figure 2B) 38, 39. The fraction of bound RNA is then calculated.
7
Binding of 1 rep Wt and Mut RNAs to EZH2 and SUZ12 was very weak and ~95% of the RNA remained unbound, even at the highest protein concentration. In contrast, when 2 rep Wt, 4 rep Wt and 8.5 rep Wt RNAs were used at equimolar concentration, the fraction of bound RNA greatly increased and reached near saturation in the case of 8.5 rep Wt RNA (Figure 2C, D). These results suggest that EZH2 and SUZ12 prefer to bind longer RNAs. Mutations in 4 rep and 8.5 rep RNAs did not overtly affect the affinity to EZH2 or
Downloaded by [New York University] at 07:15 21 July 2015
SUZ12, agreeing with the UV crosslinking data that suggest the exact sequence and/or structure of the A-repeat domain is not necessary for these proteins to bind. However, mutations in 2 rep RNA almost completely abolished binding of EZH2 and SUZ12, raising the possibility that they are capable of distinguishing some characteristics of the A-repeat domain in particular RNA contexts. To further explore the cryptic RNA-binding activity of EZH2 and SUZ12, we also assessed the binding of the two proteins to HOTAIR fragments of different lengths (Figure 2E). Unlike Xist, HOTAIR does not contain any apparent sequence or structure repeats, but a ~300 nt long sequence at the 5’ end (herein called HOTAIR-A) was previously postulated to contain a PRC2 binding element by protein pull-down assays by immobilized RNA fragments
18
. In contrast, a ~600 nt sequence at the 3’ end of
HOTAIR has been shown to bind to Lysin-specific demethylase 1 (LSD1) but not to PRC2
18
. In our double-filter-binding assay, full length HOTAIR (HOTAIR-FL) and
HOTAIR-A were bound by both proteins with similar affinity (Figure 2F, G). To test a variety of RNA lengths we deleted ~200 nt at the 5’ end of HOTAIR-A (the remaining 104 nt region called HOTAIR-B) which did not abolish binding to EZH2 or SUZ12, but a more shortened 26 nt fragment (HOTAIR-C) showed no detectable binding. 8
Interestingly, a 102 nt fragment at the 3’ end of the full-length transcript (HOTAIR-D) did not bind to EZH2 or SUZ12, even though its length was similar to HOTAIR-B (Figure 2F, G), again implying that these proteins may have some ability to distinguish RNAs. To further assess the RNA-binding activity of EZH2 and SUZ12, we tested a set of RNA fragments of different lengths that correspond to portions of the GFP mRNA (Figure 2H). Although shorter RNA fragments (28 nt and 102 nt) showed only little
Downloaded by [New York University] at 07:15 21 July 2015
binding, longer RNAs (318 nt and 645 nt long) bound to EZH2 and SUZ12 with affinities similar to HOTAIR (Figure 2I, J), even though such RNAs are not expected to play any regulatory role of PRC2 function. In contrast, EED, RbAP48 or a control GFP protein showed no detectable binding of Xist, HOTAIR or GFP RNA fragments (Figure 2K, L, M). Taken altogether, we could make a clear conclusion that the two PRC2 components, EZH2 and SUZ12, have intrinsic RNA-binding activities with general preference for longer RNAs and lower threshold for substrate RNA size of around 100 nt. However, their substrate preference was largely cryptic, hindering further characterization in vitro. PRC2, a dual RNA-binding complex Using purified proteins expressed in human cells the present work establishes that PRC2 has at least two components with RNA-binding activity, in agreement with a recent report by Lee and colleagues who observed similar results with EZH2 and SUZ12 prepared from insect cells
36
. A region between aminoacids 342-368 was
previously identified as the RNA binding domain of EZH2 domain of SUZ12 still remains unknown. 9
34
, but the RNA binding
In our experimental setup at the tested concentration, the interaction between EZH2 and 1 rep RNA could not be detected by EMSA. This might be an indication that the EZH2-RNA interaction was sensitive to binding and electrophoresis conditions and might be a reason why initial reports failed to identify the dual RNA-binding activity of the complex. A recent paper also reported that PRC2-RNA interactions can be largely affected by the experimental conditions 37. Therefore EMSA results should be analyzed with caution.
Downloaded by [New York University] at 07:15 21 July 2015
Most initial studies of the RNA-binding activity of PRC2 focused on the interaction of PRC2 subunits with 1 rep RNA
17, 30, 35
. This led to the general assumption that the
sequence and structural elements required for specific binding to PRC2 could be contained within such a small RNA. This was further reinforced with the finding that short RNAs (50–200nt long) that resemble the double stem loop structure of 1 rep RNA are transcribed from many PRC2 target genes, are bound by SUZ12 and cause gene repression in cis
30
. However, in our crosslinking and competition experiments (Figure
1D, F, G), the double stem loop structure of 1 rep RNA was not an absolute requirement for binding to EZH2 and SUZ12, and in fact both proteins barely bound RNAs of such short length, regardless of their origin (26–28 nucleotide fragments in Figure 2). Instead, in the case of A-repeat affinity increased gradually with the addition of repeats, with maximum binding observed with 8.5 rep RNAs (Figure 2C, D). However, the effect of increasing RNA length was less clear with the HOTAIR RNA series. Therefore, as the number of repeats increases, an as-yet-unidentified RNA element may act in an additive manner to alter the structure of A-repeat, resulting in enhanced binding by EZH2 and SUZ12; or alternatively A-repeat may serve as a platform for cooperative protein
10
binding. Nonetheless, additional investigations are required to understand what makes the A-repeat domain of Xist special. Our finding that EZH2 and SUZ12 are also capable of binding to unrelated GFP mRNA is in agreement with other reports that EZH2 and SUZ12 PRC2
37, 40, 41
36
and assembled
bind RNAs promiscuously. In fact, the Kd of the PRC2–A-repeat
interaction is only within one order of magnitude smaller than the Kd for unrelated RNA substrates from sources ranging from bacteria to insects
37
. In addition, CLIP
Downloaded by [New York University] at 07:15 21 July 2015
experiments for EZH2 demonstrated that in vivo this protein binds to introns and to the 5’ end of nascent RNAs in a promiscuous manner, with no apparent consensus RNA motif 41, 42. Nevertheless, it has been reported that EED can modulate the RNA-binding activity of EZH2 and SUZ12 to make it more specific
36
, and our data show that the
isolated PRC2 components retain a certain level of substrate preference, as evidenced by the ability of EZH2 and SUZ12 to distinguish between the 2 rep Wt and Mut RNAs as well as the ~102 nt HOTAIR RNA sequences inside or outside of the previously defined PRC2-binding region at the 5’ end 18. However, narrowing down the specificity determinants is inherently challenging due to their cryptic nature and inability to bind short, fragmented RNAs. Finally, although the significance of the RNA-binding by two different components of PRC2 awaits further investigation, it might be an indication of the complexity of signals that are required for the timely and precise deposition of H3K27me2/3. RNAs may be transferred between EZH2 and SUZ12 or both subunits might bind cooperatively to target RNAs, or even to different classes of RNAs simultaneously. For precise recruitment fully assembled PRC2, additional co-factors and a specific intranuclear microenvironment might be required. In fact, recent studies have illustrated the 11
complexity of the PRC2-RNA interplay. On one hand, EED modulates the RNAbinding activity of EZH2 and SUZ12
36
, and JARID2, a PRC2 partner protein with
RNA-binding activity, can compete with both proteins for RNA partners to modulate the methylatransferase activity of PRC2
36, 43
. However, a binding site of JARID2 on
Xist has been mapped to a non-overlapping region downstream of the A-repeat domain 44
and therefore how the multiple PRC2-associated RNA-binding proteins cooperate
and/or compete for binding to Xist still remains to be investigated. On the other hand,
Downloaded by [New York University] at 07:15 21 July 2015
RNA-mediated recruitment of PRC2 is enhanced by the recognition of pre-existing H3K27me3 marks by EED at repressed loci, further activating the methylatransferase activity of PRC2 and amplifying the repression signal
6, 40, 41
. All these elements add
additional layers of complexity to the regulation of gene expression mediated by PRC2 and highlight the difficulty of characterizing the RNA-binding activity of PRC2 in vitro.
Materials and methods Protein purification HEK293T cells were transfected with pCAGEN-SBP-EZH2, -SUZ12, pCAGENSBP-ps-EED, -RbAp48 or pEFh-SBP-GFP to over express streptavidin binding protein (SBP)-tagged proteins. Nuclear lysates were prepared and proteins were purified using Streptavidin Sepharose High Performance beads (GE Healthcare) and eluted with biotin. Proteins were concentrated using VivaSpin 2 ml columns (Sartorius) and quantified by CBB staining using BSA standard curves. See supplemental material for details. RNAs Primers for synthesis of DNA templates and for in vitro transcription, as well as the sequences of all RNAs used in this study are available as supplemental material. DNA 12
templates were in vitro transcribed using T7-scribe standard RNA IVT kit (Cellscript) and gel purified. Short RNAs 3).
Downloaded by [New York University] at 07:15 21 July 2015
Final concentration of radiolabeled RNA in all experiments: 1 nM.
24
