Nucleosome competition reveals processiveacetylation by the SAGA HAT moduleAlison E. Ringela, Anne M. Cieniewiczb,c, Sean D. Tavernab,c, and Cynthia Wolbergera,c,1

aDepartment of Biophysics and Biophysical Chemistry, Johns Hopkins University School of Medicine, Baltimore, MD 21205; bDepartment of Pharmacologyand Molecular Sciences, Johns Hopkins University School of Medicine, Baltimore, MD 21205; and cCenter for Epigenetics, Johns Hopkins University School ofMedicine, Baltimore, MD 21205

Edited by Carl Wu, Howard Hughes Medical Institute, Ashburn, VA, and approved September 1, 2015 (received for review April 30, 2015)

The Spt-Ada-Gcn5 acetyltransferase (SAGA) coactivator complexhyperacetylates histone tails in vivo in a manner that dependsupon histone 3 lysine 4 trimethylation (H3K4me3), a histone markenriched at promoters of actively transcribed genes. SAGA containsa separable subcomplex known as the histone acetyltransferase(HAT) module that contains the HAT, Gcn5, bound to Sgf29, Ada2,and Ada3. Sgf29 contains a tandem Tudor domain that recognizesH3K4me3-containing peptides and is required for histone hyper-acetylation in vivo. However, the mechanism by which H3K4me3recognition leads to lysine hyperacetylation is unknown, as invitro studies show no effect of the H3K4me3 modification onhistone peptide acetylation by Gcn5. To determine how H3K4me3binding by Sgf29 leads to histone hyperacetylation by Gcn5, weused differential fluorescent labeling of histones to monitor acety-lation of individual subpopulations of methylated and unmodifiednucleosomes in a mixture. We find that the SAGA HAT modulepreferentially acetylates H3K4me3 nucleosomes in a mixture con-taining excess unmodified nucleosomes and that this effect requiresthe Tudor domain of Sgf29. The H3K4me3mark promotes processive,multisite acetylation of histone H3 by Gcn5 that can account for thedifferent acetylation patterns established by SAGA at promotersversus coding regions. Our results establish a model for Sgf29function at gene promoters and define a mechanism governingcrosstalk between histone modifications.

acetyltransferase | histone crosstalk | histone modifications |transcription | coactivator

The different patterns of histone posttranslational modificationsdistributed across the genome are proposed to constitute a

“histone code” that orchestrates distinct transcriptional programsby recruiting specific effector proteins (1–4). The phenome-non of histone code “crosstalk,” whereby one type of histonemodification directs the establishment of another, or throughwhich several histone modifications are recognized in tandem, hasemerged as an important and widespread mechanism regulatingchromatin-templated processes (5, 6). The multifunctional com-plexes that activate transcription are thought to mediate crosstalkthrough “reader” domains that recognize particular chromatinmarks as well as through catalytic subunits that deposit or removehistone modifications (5, 7, 8). As a result, distinct combinations ofhistone posttranslational modifications that correlate with tran-scriptional output cluster across the genome (9–11). High levels ofhistone 3 lysine 4 trimethylation (H3K4me3) and histone H3hyperacetylation are present at the promoters of actively tran-scribed genes (9, 11, 12), and multiple lines of evidence support theexistence of regulatory mechanisms coupling the two modifica-tions. Studies using tandem mass spectrometry to sequence wholehistone tails have shown that H3K4me3 is highly correlated withhyperacetylation of the same H3 tail (13, 14). Deleting the yeastSet1 methyltransferase, which trimethylates H3K4, leads to dra-matically lower levels of histone H3 tail acetylation overall (13, 15).These results are consistent with a role for H3K4 methylation intriggering hyperacetylation of histone H3; indeed, a number of

histone acetyltransferase (HAT) complexes also contain readermodules that recognize the H3K4me3 mark (16, 17).The SAGA (Spt-Ada-Gcn5 acetyltransferase) complex is a

highly conserved transcriptional coactivator (18) that is involved inthe transcription of nearly all yeast genes (19, 20) and mediatescrosstalk between H3K4me3 and histone hyperacetylation. TheHAT activity of SAGA resides in a four-protein subcomplexknown as the “HAT module” (21, 22), which contains the catalyticsubunit Gcn5 (23–25), Ada2, Ada3 (26, 27), and Sgf29 (21, 28).The association of Gcn5 with the other HAT module subunitsmodulates Gcn5 activity and specificity. Gcn5 on its own acetylatesfree histones but not nucleosomes (22, 23, 25) and preferentiallymodifies histone H3 at K14 (25, 29). A complex containing Gcn5bound to Ada2 and Ada3 is more active overall and modifiesnucleosomal substrates (22, 25). Binding to Ada2/Ada3 alsobroadens the lysine specificity of Gcn5 to acetylate multiple lysineresidues on the histone H3 tail (22, 25). Sgf29 contains a tandemTudor domain that binds to H3K4me3 (17, 30) and is required formaintaining wild-type levels of histone H3 acetylation in vivo(30–32). However, in vitro studies comparing acetylation kineticsby the SAGA complex on peptide substrates in the presence orabsence of Sgf29 do not show a rate enhancement when theH3K4me3 modification is present (30). Because these experi-ments were limited in scope to peptide substrates and measuredacetylation using relatively insensitive end-point assays (30),the full impact of H3K4me3 on HAT module activity may

Significance

Crosstalk between histone modifications regulates transcriptionby establishing spatial and temporal relationships between his-tone marks. Despite discoveries of reader domains that physicallyassociate with chromatin-modifying enzymes, the mechanisms bywhich recognition of one modification triggers other kinds ofmodifications have remained elusive. Gcn5 is the catalytic subunitof the Spt-Ada-Gcn5 acetyltransferase (SAGA) histone acetyl-transferase (HAT) module, which also recognizes histone 3 lysine 4trimethylation (H3K4me3) through the tandem Tudor domain-containing protein Sgf29. Although previous studies could notconnect H3K4me3 recognition to differences in acetylation byGcn5, we report enhanced processivity by the HAT module onmethylated substrates using a previously unpublished histonecolor-coding assay. Our work defines a mechanism for histonecrosstalk that may account for genome-wide patterns of Gcn5-mediated acetylation.

Author contributions: A.E.R., S.D.T., and C.W. designed research; A.E.R. and A.M.C. per-formed research; A.E.R., A.M.C., S.D.T., and C.W. analyzed data; and A.E.R. and C.W.wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

Freely available online through the PNAS open access option.1To whom correspondence should be addressed. Email: cwolberg@jhmi.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1508449112/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1508449112 PNAS | Published online September 23, 2015 | E5461–E5470

BIOCH

EMISTR

YPN

ASPL

US

Dow

nloa

ded

by g

uest

on

June

10,

202

0

be much greater than previously determined. Thus, althoughmultiple lines of evidence suggest that H3K4me3 regulateshistone acetylation by SAGA in yeast and mammalian cells (17,30, 31, 33), in vitro studies have failed to explain how H3K4me3recognition by Sgf29 gives rise to different patterns of histone acet-ylation. As a result, the underlying mechanism governing crosstalkbetween H3K4 trimethylation and acetylation by SAGA remainspoorly understood.To elucidate the mechanism underlying crosstalk between HAT

module binding to H3K4me3 and histone acetylation, we devel-oped a histone color-coding assay for measuring acetylation rateson distinct histone populations within a single mixture of meth-ylated and unmodified nucleosomes. We find that the HATmodule preferentially acetylates H3K4me3 nucleosomes in amixture and that this preference depends on the Sgf29 tandemTudor domain. Compared with unmodified nucleosomes, acety-lation by the HAT module is processive on nucleosomes con-taining the H3K4me3 modification, thereby triggering histone H3hyperacetylation. These data reveal the mechanism by which theSgf29 tandem Tudor domain within the SAGA HAT module fa-cilitates crosstalk between Gcn5 acetyltransferase activity andH3K4me3 recognition and provide an explanation for how SAGAcan establish different patterns of acetylation at gene promotersversus coding regions.

ResultsEffect of H3K4me3 on Histone Peptide Acetylation. To investigatehow H3K4 trimethyl recognition may contribute to HAT moduleactivity on methylated versus unmethylated substrates, we usedrecombinant HAT module from Schizosaccharomyces pombe,which can be isolated as a stable four-protein complex con-taining Sgf29, Gcn5, Ada2, and Ada3. The full-length S. pombe

HAT module and variants lacking either the Sgf29 tandem Tudordomain or the Gcn5 bromodomain were expressed as complexesin Escherichia coli and purified to homogeneity (Fig. 1). We firstconfirmed that S. pombe Sgf29 interacts specifically withH3K4me3 peptides in a pull-down assay. As shown in Fig. 2A,streptavidin beads coated with H3K4me3 peptides efficiently re-tain Sgf29, whereas unmodified H3 peptides do not.To determine the effect of H3K4me3 on the kinetics of pep-

tide acetylation by Gcn5, we measured initial acetylation ratesby the HAT module at increasing concentrations of either un-modified or H3K4me3 peptides in the presence of saturatingamounts of acetyl-CoA. Compared with unmodified peptides,the presence of the H3K4me3 modification increases kcat/Km by3.1-fold (Fig. 2B and Table 1). This change in kcat/Km is causedsolely by a decrease in Km from 250 ± 40 μM to 77 ± 10 μM, withno corresponding change in kcat (Fig. 2B and Table 1). Deletionof the Sgf29 tandem Tudor domain (ΔTudor) eliminates thedifference in acetylation kinetics on H3K4me3 versus un-modified peptides (Fig. 2C and Table 1), as is consistent with arole for H3K4me3 binding in the observed difference in kcat/Km.We also ruled out the possibility that the H3K4me3 modificationaffects kcat or Km for acetyl-CoA (Fig. 2D and Table 2); ourfinding is consistent with the proposed Gcn5 reaction mecha-nism, in which the cofactor binds before the peptide (34, 35).To confirm that acetylation by Gcn5 is not directly affected by

truncating the Sgf29 tandem Tudor domain, we compared kcatand Km values of wild-type and ΔTudor HAT module complexesfor unmodified histone peptides and nucleosomes. Because thekinetic parameters for the two complexes are nearly the samefor peptides (Table 1) and nucleosomes (Table 3), differencesobserved with the ΔTudor HAT module must be related toH3K4me3 recognition. We note that the rate enhancement ob-served resulting from H3K4me3 differs from the results of anearlier study that had found no role for Sgf29 in governingSAGA activity on H3K4me3 versus unmodified peptides in vitro(30). However, the previously reported conclusions were drawnfrom end-point assays (30), whereas the data presented here arefit from a full steady-state titration and therefore are moresensitive to differences in kinetic parameters.

H3K4me3 Stimulates HAT Module Activity on Nucleosomes. The ex-periments described above and previously reported studies (30)used peptide substrates, leaving open the possibility that H3K4trimethylation might have a greater impact on acetylation by theSAGA HAT module in a nucleosomal context. We thereforegenerated nucleosomes containing recombinant full-length his-tone H3 that was homogeneously modified with H3K4me3. Toincorporate this modification, we performed enzyme-mediated li-gation using an engineered variant of the bacterial transpeptidase,Sortase A (F40 SrtA) (36, 37), which splices an N-terminal

Fig. 1. Domain architecture of HAT module complexes. (Left) SDS/PAGE gelof purified S. pombe HAT module complexes containing different domaintruncations in which all four subunits are present in roughly equimolarquantities. (Right) Location of the catalytic domain, tandem Tudor domain,and bromodomain within the HAT module.

Fig. 2. The presence of H3K4me3 increases overall acetylation by the HAT module on peptides. (A) Pull-down using biotinylated histone peptides containingH3K4 trimethylation and Sgf29. (B–D) Steady-state kinetic titrations using an enzyme-coupled assay comparing the activity of the wild-type HAT module onunmodified versus H3K4me3 peptides (B), the ΔTudor HAT module on unmodified versus H3K4me3 peptides (C), or the wild-type versus ΔTudor HAT module(D) using a constant amount of H3K4me3 peptide and varying the acetyl-CoA concentration.

E5462 | www.pnas.org/cgi/doi/10.1073/pnas.1508449112 Ringel et al.

Dow

nloa

ded

by g

uest

on

June

10,

202

0

synthetic peptide onto recombinant histone H3 (Fig. S1A) to yieldfully native histone H3 uniformly trimethylated at H3K4 (Fig.S1B). We confirmed the presence of the trimethyl modificationusing Western blotting with an antibody specific for the H3K4me3modification (Fig. S1C) as well as MALDI-TOF mass spec-trometry (Fig. S2). The modified histone H3 was reconstitutedwith unmodified histones H2A, H2B, and H4 into nucleosomecore particles, thus generating a pool of nucleosomes uniformlytrimethylated at H3K4.To test whether H3K4 trimethylation affects acetylation by the

HAT module in a nucleosomal context, we used a radioactivefilter-binding assay to measure steady-state rates of acetylationas a function of nucleosome concentration. We find that the Kmfor nucleosomes drops from 1.9 ± 0.3 μM for unmodified nu-cleosomes to 0.41 ± 0.05 μM for H3K4me3 nucleosomes (Fig. 3Aand Table 3), whereas kcat stays roughly the same, corresponding toan ∼3.5-fold increase in the specificity constant kcat/Km (Table 3).The Km for nucleosomes is almost two orders of magnitude lowerthan that for peptides, indicating that the HAT module bindsfar more tightly to nucleosomes than to peptides. Nonetheless,the 3.5-fold increase in kcat/Km caused by the H3K4me3 mod-ification in a nucleosomal context (Fig. 3A and Table 3) is onlymodestly greater than the 3.1-fold increase in specificity ob-served with peptides (Fig. 2B and Table 1).Because radioactive filter-binding assays report on total acetyla-

tion and do not distinguish differences among histones, we com-pared histone acetylation patterns on unmodified versus H3K4me3nucleosomes using acid–urea gels, which separate histones based onmolecular weight as well as the number of acetylated lysine residues(38). At early time points the HAT module acetylates H3K4trimethylated histone H3 in nucleosomes more quickly thanunmodified nucleosomal histone H3 (Fig. 3B). Thirty secondsafter the reaction is initiated, bands corresponding to three acet-ylation sites are evident on the H3K4me3 histone, but only twoacetylation sites are populated for unmodified histone H3 (Fig.3B). Although the HAT module also acetylates histone H4, therate of histone H4 acetylation is insensitive to H3K4 trimethyla-tion (Fig. 3B). This pattern is consistent with a mechanism in cis,whereby recognition of the methyl mark stimulates acetylation onthe same histone tail.To examine whether recognition of the H3K4me3 modifica-

tion changes the intrinsic lysine specificity of the Gcn5 subunit,which preferentially catalyzes H3K14 acetylation in vitro (29, 39)and H3K9 acetylation in vivo (20, 40), we probed the reactionproducts using antibodies that recognize specific acetylated ly-sine residues. Antibodies to the H3K9ac and H3K14ac modifi-cations recognize five bands on acid–urea gels, corresponding toone through five modification sites per histone, whereas anti-bodies to H3K18ac and H3K23ac recognize only the top fourbands (Fig. S3). This pattern reveals that singly acetylated his-tones are predominately modified at H3K9 or H3K14 on bothunmodified and H3K4me3 substrates. Although H3K4 trimeth-ylation enhances the rate of H3K9 acetylation at early timepoints, acetylation at all the other lysines tested changes little in

the presence of the methyl modification (Fig. S3). Therefore,H3K4me3 recognition does not change the intrinsic specificity ofGcn5 for catalyzing H3K9 and H3K14 acetylation, insteadstimulating overall acetylation of histone H3. In agreement withour results, deleting Sgf29 in yeast results in global reductions inacetylation at multiple positions on histone H3 (30).

Differential Labeling Technique for Studying Mixtures of Post-translationally Modified Substrates. Chromatin in vivo is hetero-geneous with respect to the set of posttranslational modificationsfound on each nucleosome. For an enzyme complex such asthe HAT module, which contains multiple chromatin-bindingdomains (16, 41), the ability to acetylate nucleosomes selectivelywith particular combinations of histone modifications may be asimportant as the overall enzymatic activity. Although multiplebiochemical assays yield bulk acetylation rates (42), there are noreported assays for selectively monitoring the acetylation kineticsof different types of substrates in the context of a mixture. Wetherefore devised an approach that combines differential fluo-rescent labeling of histones with acid–urea gel electrophoresis toquantitate HAT module activity on both modified and unmodifiednucleosomes within the same reaction mixture.To track histone-specific acetylation, we reconstituted nucle-

osomes whose histone H3 was color-coded according to whetherit contained the H3K4me3 modification. Methylated H3 wasgenerated using F40 SrtA as described above; the fluorescentlabel was incorporated by introducing a Q125C substitution inhistone H3 that could be labeled with a dye of choice. Un-modified histone H3 was labeled with Alexa Fluor 647, andH3K4 trimethylated histone H3 was labeled with Alexa Fluor488 (Fig. S4A). After the SAGA HAT module was incubatedwith the nucleosomes (Fig. S4B) and the reaction products wereresolved on acid–urea gels, the products corresponding to one orthe other color-coded histone could be visualized selectively on alaser scanner using excitation wavelengths and emission filtersfor one or the other fluorophore (Fig. S4C). In this way, acety-lation of methylated versus unmethylated H3 can be quantitatedwithin a single mixture.

Sgf29 Mediates Selective Acetylation of H3K4me3 Nucleosomes in aHeterogeneous Mixture. To see whether the HAT module pref-erentially acetylates nucleosomes containing H3K4 trimethyla-tion, we used the color-coding method to assay histone H3 acetylationin mixtures of methylated and unmodified nucleosomes. Because

Table 1. Summary of steady-state kinetic parameters fit from titrations in which theconcentration of the histone H3 amino acids 1–21 peptide was varied and the acetyl-CoAconcentration was held constant at 100 μM

HAT module Histone H3 peptide amino acids 1–21 kcat, s−1 Km, μM kcat/Km, μM−1·s−1

Wild type Unmod 2.5 ± 0.1 250 ± 40 0.010 ± 0.002ΔTudor Unmod 2.5 ± 0.1 210 ± 30 0.012 ± 0.002Sgf29 T202A/Y205V Unmod 3.4 ± 0.2 320 ± 50 0.011 ± 0.002ΔBromodomain Unmod 1.8 ± 0.07 380 ± 40 0.0047 ± 0.001Wild type H3K4me3 2.4 ± 0.08 77 ± 10 0.031 ± 0.004ΔTudor H3K4me3 2.5 ± 0.09 190 ± 20 0.013 ± 0.001

Table 2. Summary of steady-state kinetic parameters fit fromtitrations in which the concentration of the H3K4me3 aminoacids 1–21 peptide was held at 400 μM and the acetyl-CoAconcentration was varied

HAT module Substrate kcat, s−1 Km, μM kcat/Km, μM−1·s−1

Wild type Ac-CoA 1.7 ± 0.1 3.4 ± 0.7 0.50 ± 0.1ΔTudor Ac-CoA 1.8 ± 0.1 2.8 ± 0.3 0.64 ± 0.08

Ringel et al. PNAS | Published online September 23, 2015 | E5463

BIOCH

EMISTR

YPN

ASPL

US

Dow

nloa

ded

by g

uest

on

June

10,

202

0

H3K4 trimethylation increases the kcat/Km for the HAT module byaround fourfold (Fig. 3A and Table 3), the HAT module shouldacetylate H3K4me3 and unmodified nucleosomes at the same ratein mixtures containing a fourfold molar excess of unmodifiednucleosomes (43). However, in assays of SAGA HAT moduleactivity using a 4:1 molar ratio of unmodified versus methylatednucleosomes, we unexpectedly found that the HAT modulepreferentially acetylates the methylated nucleosomes despitethe presence of a fourfold molar excess of unmodified nucle-osomes (Fig. 4A). An overlay of images corresponding to eachlabeled substrate reveals that nucleosomes with the H3K4me3modification are acetylated to completion within the first 5minutes, whereas unmodified nucleosomes still are not fullyacetylated at the last time point (Fig. 4A). For the SAGA HATmodule lacking the Tudor domain, the pattern of acetylation isthe same for both methylated and unmethylated nucleosomes(Fig. 4B). Because there is fourfold less H3K4me3 substrate in thereaction, this finding indicates that complexes lacking the tandemTudor domain have lost the ability to distinguish between meth-ylated and unmodified nucleosomes. To verify that deletion of theSgf29 tandem Tudor domain did not disrupt the overall integrityand activity of the HAT module, we introduced point mutations inthe Tudor domain predicted to disrupt binding to a methylatedpeptide based on structures of the human and yeast Sgf29 tandemTudor domains bound to H3K4me3 peptides (30). We mutatedT202 to alanine and Y205 to valine (T202A/Y205V), therebyeliminating two hydrogen bonds between Sgf29 and the H3backbone and disrupting the aromatic cage that surrounds themethyl lysine (Fig. S5A). In pull-down assays, we find that isolatedSgf29 containing the T202A/Y205V substitutions binds moreweakly to H3K4me3 peptides than does the wild-type protein (Fig.S5B). When incorporated into the HAT module, the effect ofmutating Sgf29 mimics that of a full Tudor domain truncation, inwhich mixtures of methylated and unmodified nucleosomes are

acetylated at equal rates (Fig. S5C). Together, these results showthat when nucleosomes harboring different histone marks com-pete for the HAT module active site, the interaction between theSgf29 tandem Tudor domain and the H3K4me3 modification fa-cilitates preferential acetylation of methylated nucleosomes.To rule out the possibility that H3K4me3 recognition acts in

trans by stimulating overall HAT module activity, we assayednucleosome acetylation in the presence of an added histone H3tail peptide containing the H3K4me3 modification. Peptide wasadded at a concentration 10-fold higher than the reported dis-sociation constant of methylated peptide for Sgf29 (30) andnucleosomes were held at a concentration matching the amountof H3K4me3 substrate in our experiments using nucleosomemixtures (Fig. 4). As shown in Fig. S6, the added peptide did notincrease HAT module activity on unmodified (Fig. S6 A and B)or H3K4me3 nucleosomes (Fig. S6 C and D), ruling out transstimulation of Gcn5 by H3K4me3 histone tails. To control forfluorophore-specific effects on acetylation by the HAT module,we compared HAT module activity on otherwise unmodifiednucleosomes containing histone H3 labeled with either AlexaFluor 488 or Alexa Fluor 647. The HAT module acetylates thesetwo nucleosomes at the same rate in homogenous preparations(Fig. S7 A and B) and in mixtures containing varying proportionsof the Alexa Fluor 488- or Alexa Fluor 647-labeled nucleosomes(Fig. S7 C and D), indicating that the choice of fluorescent labeldoes not influence the observed acetylation patterns.

Sgf29 Promotes Processive Acetylation by the SAGA HAT Module onH3K4me3 Nucleosomes. What mechanism links H3K4me3 recog-nition by Sgf29 at gene promoters to hyperacetylation by Gcn5?When provided with competing substrates, the HAT moduleclearly modifies H3K4me3 nucleosomes first, even under con-ditions in which one expects to see equal rates of acetylation(Fig. 4A and Table 3). We hypothesized that H3K4me3 bindingby Sgf29 might promote processive acetylation of methylatednucleosomes, with the HAT module acetylating multiple lysinesbefore dissociating from the nucleosome. To test for processiveacetylation by Gcn5, we used the reaction scheme diagrammedin Fig. 5A. We incubated the HAT module with labeled nucle-osome for a short period and then added a 20-fold molar excessof unlabeled, unmodified competitor nucleosome. As controls,either a portion of the initial reaction was quenched with aceticacid, which halts all enzymatic activity, or buffer alone was added(mock quench), which allows the reaction to proceed (Fig. 5A). Theresulting pattern of acetylation on fluorescently labeled histone H3

Table 3. Summary of steady-state kinetic parameters fit fromtitrations in which the concentration of the nucleosome wasvaried and the acetyl-CoA concentration was held constantat 25 μM

HAT module Nucleosome kcat, s−1 Km, μM kcat/Km, μM−1·s−1

Wild type Unmod 1.3 ± 0.1 1.9 ± 0.3 0.68 ± 0.1ΔTudor Unmod 1.9 ± 0.2 2.8 ± 0.5 0.68 ± 0.1Wild-type H3K4me3 1.0 ± 0.05 0.41 ± 0.05 2.4 ± 0.3

Fig. 3. The presence of H3K4me3 increases overall acetylation by the HAT module on nucleosomes. (A) Steady-state kinetic titrations comparing wild-typeHAT module activity on unmodified versus H3K4me3 nucleosome core particles (NCPs) using a radioactive filter-binding assay. Replicates are colored withdifferent shades of purple (wild-type) or green (H3K4me3). (B) HAT reaction time course on unmodified versus H3K4me3 NCPs resolved on an acid–urea geland stained for total protein using SyproRuby stain.

E5464 | www.pnas.org/cgi/doi/10.1073/pnas.1508449112 Ringel et al.

Dow

nloa

ded

by g

uest

on

June

10,

202

0

then was visualized using acid–urea gels. For a distributive (non-processive) enzyme, the acetylation pattern in the presence ofcompetitor should mimic the chemical quench, because the enzymeshould dissociate readily from the labeled substrate and acetylatethe unlabeled competitor instead. A processive enzyme should re-main bound to the labeled substrate, even in the presence of addedcompetitor, thus allowing further acetylation of the labeled nucle-osome. As a result, the acetylation pattern for a processive enzymein the presence of competitor should mimic the mock-quench.When the SAGA HAT module was incubated with nucleo-

somes containing unmethylated histone H3 fluorescently labeledwith Alexa Fluor 647, the addition of unlabeled nucleosomescompeted for HAT module activity and prevented labeled histoneH3 from becoming fully acetylated (compare competitor and mock-quenched samples in Fig. 5B). In contrast, the addition of unlabeledcompetitor nucleosomes had little effect on the acetylation offluorescently labeled H3K4me3 nucleosomes, as can be seen by thesimilarity in the intensity profiles of the competitor and mock-quench samples (Fig. 5C). This pattern is characteristic of processiveacetylation of H3K4me3 histone H3, presumably mediated bybinding of the HAT module subunit, Sgf29, to the methylatedN-terminal tail of histone H3 via the Sgf29 tandem Tudor domain(30). Consistent with the role of H3K4me3 binding in processivity,the addition of H3K4me3 peptide at the start of the reaction, at aconcentration that should saturate binding to the Sgf29 tandemTudor domain (30), abrogates processive acetylation by the HATmodule (Fig. 5D). Even though the addition of the H3K4me3

peptide has little effect on overall rates of acetylation by the HATmodule (Fig. S6D), HAT module processivity is reduced signifi-cantly. This behavior indicates that Sgf29 promotes processiveacetylation by the SAGAHATmodule on H3K4me3 nucleosomes.To probe whether processive acetylation of H3K4me3 nucle-

osomes depends on the interaction between the Sgf29 tandemTudor domain and the H3K4 trimethyl modification, we re-peated the experiment using a HAT module lacking the Sgf29Tudor domain (ΔTudor). Unlike the intact HAT module, theΔTudor complex can be competed off both unmodified (Fig. 5E)and H3K4me3 nucleosomes (Fig. 5F). These results, along withthe fact that the wild-type HAT module is efficiently competedoff H3K4me3 nucleosomes in the presence of the H3K4me3peptide (Fig. 5D), are consistent with a role for the Sgf29 Tudordomain in engaging the methylated histone H3 tail.Because the Gcn5 bromodomain recognizes acetyl lysine (44,

45) and regulates the lysine specificity of Gcn5 (39), we askedwhether the bromodomain contributes to processive acetylation bythe HAT module. Truncating the bromodomain of Gcn5 whileleaving the Sgf29 tandem Tudor domain intact (Fig. 1) causes atwofold decrease in kcat/Km for peptide substrates compared withthe wild-type HAT module (Table 1). As has been observed pre-viously, deletion of the Gcn5 bromodomain changes the overallpattern of acetylation catalyzed by the HAT module (39), partic-ularly by delaying accumulation of the most highly acetylated states(Fig. S8 A and B). However, processive acetylation is clearlyimpaired on H3K4me3 nucleosomes (compare competitor and

Fig. 4. The HAT module preferentially acetylates nucleosomes harboring H3K4me3 out of a mixture. (A) Fluorescent images of a HAT reaction containing a 4:1molar ratio of differentially labeled unmodified versus H3K4me3 nucleosomes resolved on an acid–urea gel. (Left) Merged image with unmodified nucleosomescolored green and H3K4me3 nucleosomes colored red. (Right) Individual images used to generate the merged image. (B) Fluorescent images of a ΔTudor HATreaction containing a 4:1 molar ratio of differentially labeled unmodified versus H3K4me3 nucleosomes resolved on an acid–urea gel. (Left) Merged image withunmodified nucleosomes colored green and H3K4me3 nucleosomes colored red. (Right) Individual images used to generate the merged image.

Ringel et al. PNAS | Published online September 23, 2015 | E5465

BIOCH

EMISTR

YPN

ASPL

US

Dow

nloa

ded

by g

uest

on

June

10,

202

0

mock-quenched samples in Fig. 5 G and H), although not to thedegree seen with the ΔTudor complex (Fig. 5 E and F). Theseresults suggest that, although processivity is governed primarily bybinding of the Sgf29 tandem Tudor domain to H3K4me3 marks innucleosomes, there is some contribution by the Gcn5 bromodo-main, presumably through binding to acetylated lysine residues.Under the conditions used to monitor processive acetylation, in

which the HAT module and labeled nucleosome are presentequimolar ratios, we observe six acetylation events per histone tail(Fig. 5 B–H; also see Fig. S10). However, we only observe fiveacetylation events per histone in assays containing higher con-centrations of nucleosomes (Figs. 3B and 4 and Figs. S3, S5C, S6,and S7). The catalytic efficiency of Gcn5 for the various lysineresidues on the H3 tail varies by two orders of magnitude (29).Thus, we attribute the extra acetylation site observed at low con-centrations of nucleosomes to the different specificity of Gcn5 foreach lysine and would anticipate six acetylation events per histoneat long time points in all our assays.

DiscussionAlthough crosstalk between histone modifications is a well-docu-mented feature of the histone code (4, 5), the underlying mech-anisms have remained obscure. Although studies in vivo haveconnected the presence of H3K4me3 to transcriptional activationby the SAGA complex (30, 31, 46), the way by which SAGAtransduces H3K4me3 recognition into hyperacetylation by Gcn5

was not known. In this study we uncover a mechanism by whichthe HAT subcomplex of the SAGA transcriptional coactivatorcouples histone hyperacetylation to H3K4 trimethylation, a uni-versal mark associated with promoter regions of actively tran-scribed chromatin in organisms ranging from yeast (9, 47) tohumans (48). We show that H3K4 trimethylation modestly ac-celerates acetylation by the HAT module on peptides and nucle-osomes (Figs. 2B and 3A and Tables 1 and 3). Using differentialhistone labeling, we find that the HAT module is highly selectivefor H3K4me3 histones in mixtures of modified and unmodifiednucleosomes (Fig. 4A). This selectivity is a consequence of processiveacetylation of nucleosomes containing the H3K4me3 modifi-cation (Fig. 5 B and C), with the HAT module acetylating eachH3K4me3-containing nucleosome multiple times before dis-sociating. These data establish a paradigm for Sgf29 function atgene promoters and provide a mechanistic explanation for theobservation that acetylation by Gcn5 in vivo is governed by thepresence of H3K4 trimethylation (13, 30).Processive, multisite acetylation on H3K4me3 nucleosomes also

explains how the HAT module might establish different acetyla-tion patterns at promoters versus coding regions of transcribedgenes. Although Gcn5 is broadly enriched across highly tran-scribed genes in yeast (49), Gcn5-dependent acetylation patternsdiffer between gene promoters and coding regions (20, 49). Forexample, both H3K9 and H3K14 acetylation can be catalyzed byGcn5 in vivo, but high-resolution ChIP-sequencing data show that

Fig. 5. Sgf29 promotes processive acetylation against H3K4me3 nucleosomes. (A) Schematic of the competition assay to test for processive behavior.(B–H) Results from competition assays containing: the wild-type HAT module and unmodified nucleosomes (B), the wild-type HAT module and H3K4me3nucleosomes (C), the wild-type HAT module, H3K4me3 nucleosomes, and H3K4me3 peptide (amino acids 1–8) added at the beginning of the reaction (D), theΔTudor HAT module and unmodified nucleosomes (E), the ΔTudor HAT module and H3K4me3 nucleosomes (F), the Δbromodomain HAT module and un-modified nucleosomes (G), or the Δbromodomain HAT module and H3K4me3 nucleosomes (H). (Upper Row) Acid–urea gel of competition assay visualizedusing a Typhoon scanner. Background subtraction was performed using a sliding paraboloid in ImageJ. (Lower Row) Quantitation of acid–urea gel in ImageJ.

E5466 | www.pnas.org/cgi/doi/10.1073/pnas.1508449112 Ringel et al.

Dow

nloa

ded

by g

uest

on

June

10,

202

0

H3K9 acetylation is narrowly limited to promoter regions (20),whereas H3K14 acetylation spreads across gene bodies as well(49). How can the same enzyme generate different acetylationpatterns? Our results are consistent with a model in which H3K4trimethylation, which is enriched at gene promoters comparedwith coding regions (9, 50), promotes a pattern of acetylation inpromoter regions different from that in gene bodies (Fig. 6). In theabsence of H3K4 trimethylation, the HAT module likely acety-lates histones according to the intrinsic lysine specificity of Gcn5,which vastly favors modification at the H3K14 position (29, 39,51). At promoters, however, recognition of the H3K4me3 mark bySgf29 promotes processive, multisite acetylation by the HATmodule. The bromodomain of Gcn5 can bind to these newlyacetylated residues, providing a second feedback mechanism thatpromotes retention of the complex and, consequently, localhyperacetylation (Fig. 6). Together, the actions of the tandemTudor domain and the bromodomain provide a mechanism bywhich the reader functions of SAGA might fine-tune the activityof Gcn5 along different genic regions, thus establishing diffusepatterns of H3K14 acetylation punctuated by narrow regions ofhyperacetylation that colocalize with H3K4 trimethylation. Theseresults also explain why depleting Sgf29, a noncatalytic subunitcontaining a tandem Tudor domain that recognizes H3K4me3,reduces levels of H3K9, H3K14, and H3K18 acetylation in yeastand human cells (30) despite having little effect on overall SAGAactivity (30). This model leaves open the possibility that otheracetyltransferases, particularly those with H3K4me3-binding do-mains, also might contribute to acetylation at actively transcribedgenes. For example, the yeast HAT complex NuA3 contains anoncatalytic subunit, Yng1, which binds to H3K4 trimethylationand enhances H3K14 acetylation by NuA3 on H3K4me3 peptidesin vitro (52).The effect of H3K4me3 on acetylation by Gcn5 is clearly more

complicated than simply providing additional binding energy formethylated substrate. Binding of Sgf29 to methylated peptideis not proportionally reflected in a lower Km, as methylationchanges the affinity of isolated Sgf29 for peptides by ∼50-fold (30),whereas we observe only a fourfold change in Km (Tables 1 and 3).We note that, in addition to the Sg29 tandem Tudor domain, the

HAT module contains a number of other chromatin-interactingdomains that are not sensitive to the H3K4me3 modification.With the potential for multiple histone-binding interactions thatmay further impact both binding and catalysis, there is not asimple correspondence between Sgf29-H3K4me3 affinity and Kmfor the HAT complex.Our results highlight the benefit of using defined, homoge-

neous chromatin templates containing native modifications tostudy the mechanistic basis of crosstalk between histone modi-fications. For many chromatin-modifying complexes, the mech-anism by which reader domains translate the histone code intofunctional outcomes remains an open question that is difficult toaddress biochemically with recombinant unmodified histones orwith heterogeneously modified chromatin isolated from cells. Ourhistone color-coding assay, combined with established methods forgenerating uniformly modified histones (36, 53), makes it possibleto monitor acetylation on different pools of nucleosomes withinthe same reaction, thereby better approximating the heterogeneityof chromatin inside the cell. This differential labeling assay alsomakes it possible to observe substrate competition directly, ratherthan making inferences from independent measurements of spec-ificity. In the present study, we compared activity on unmodifiedand H3K4me3 nucleosomes; however, the number of independentsubpopulations that can be interrogated simultaneously usingcolor-coding is limited only by the need to avoid significant spectraloverlap between the fluorophores used. This technique can beadapted to study tail-specific acetylation on asymmetrically modi-fied nucleosomes as well (54), for example those in which one copyof histone H3 contains the H3K4me3 mark and the other copy isunmodified. By incorporating more than one label into the sameoctamer, different histones within the same nucleosome could bemonitored independently. Our assay using fluorescent histone la-beling, along with other recently developed techniques such asfluorescent labeling of nucleosomal DNA (55) and DNA-barcodednucleosome libraries (56), are particularly suited to addressing themechanistic basis of histone crosstalk.

Materials and MethodsCloning, Expression, and Purification of Proteins. SAGA HAT module subunitswere PCR-amplified from S. pombe genomic DNA using KOD polymerase(EMD Millipore) and were cloned into vectors for bacterial expression usingthe In-Fusion Cloning kit (Clontech). Because S. pombe Sgf29 contains anintron, the two exons were PCR-amplified using primers containing a 15-bpinternal overlap and were used together in a single In-Fusion reaction. The HATmodule subunits were cloned into three vectors, which were compatible forcoexpression in E. coli. Ada3 was cloned into pET32a, The Gcn5/ Gcn5Δbro-modomain was cloned into the first multiple cloning site of CDFduet, Ada2was cloned into the second multiple cloning site of CDFduet, and Sgf29/Sgf29ΔTudor was cloned into pRSF. Full-length Sgf29 for pull-down assays wascloned into pET32a. Histone H3 from Xenopus laevis containing C110A/Q125Cmutations and/or having the first 32 amino acids truncated (H3Δ32) for sortasereactions were generated using QuikChange site-directed mutagenesis(Agilent Technologies).

Plasmids containing the wild-type HAT module and deletion constructslacking the Sgf29 Tudor domain (ΔTudor corresponding to Sgf29 amino acids1–95) or the Gcn5 bromodomain (Δbromodomain corresponding to Gcn5amino acids 1–340) were cotransformed into Rosetta2-(DE3) E. coli cells (EMDMillipore) and coexpressed as intact complexes bearing a hexahistidine tag onthe Ada3 subunit. Ada3 was expressed as a thioredoxin fusion in pET32a; Gcn5and Ada2 were provided on CDFduet; and Sgf29 was supplied on pRSF. Startercultures cotransformed with all three plasmids were grown overnight at 37 °Cin Terrific broth (TB) containing 50 μg/mL carbenicillin, 31 μg/mL chloram-phenicol, 50 μg/mL streptomycin, and 25 μg/mL kanamycin. The saturatedcultures then were diluted by 100-fold into 4–12 L of TB and grown at 37 °Cuntil reaching an OD600 of 0.6–0.8. The flasks were transferred to an ice bathfor 45 min, and the shaker temperature was lowered to 16 °C. The cells wereinduced with the addition of 0.25 mM isopropyl β-d-1-thiogalactopyranosideovernight (16–18 h), harvested by centrifugation, and resuspended in lysisbuffer containing 40 mM Hepes, pH 7.6, 500 mM NaCl, 10% glycerol, 20 mMimidazole, pH 8.0, and 5 mM β-mercaptoethanol (BME).

Fig. 6. Model depicting posttranslational modification-dependent acetylationby SAGA. (Left) H3K4me3 enrichment at gene promoters stimulates processive,multisite acetylation by the SAGA HAT module, resulting in hyperacetylation ofhistone H3. (Right) Within gene bodies, Gcn5 catalyzes low levels of acetylationbased on its intrinsic lysine specificity, preferentially modifying H3K14.

Ringel et al. PNAS | Published online September 23, 2015 | E5467

BIOCH

EMISTR

YPN

ASPL

US

Dow

nloa

ded

by g

uest

on

June

10,

202

0

To purify the HAT module, cell pellets were thawed, lysed using a Micro-fluidizer (Microfluidics Corp.), and clarified by centrifugation at 32,000 × g. Thesoluble fraction was loaded onto a HisTrap HP column (GE Life Sciences)equilibrated in lysis buffer. The HAT module was eluted with a 20–400 mMimidazole gradient over 15 column volumes and dialyzed overnight at 4 °Cinto 20 mM Hepes, pH 7.6, 100 mM NaCl, 10% glycerol, and 5 mM BME in thepresence of 1 mg recombinant Tobacco Etch Virus (TEV) protease per 10 mgpurified protein to cleave the thioredoxin tag. Following dialysis, imidazolewas added to the cleavage reaction at a final concentration of 20 mM, andthen the mixture was run over the HisTrap HP column (GE Life Sciences), whichretained the hexahistidine-tagged thioredoxin while allowing the HAT mod-ule to pass through. Solid ammonium sulfate was added to the flow-throughover 10 min with stirring at 4 °C, to a final concentration of 0.7 M. Precipitatedprotein was removed using a 0.4-μm filter, and the supernatant was applied toa Phenyl HP column (GE Life Sciences) equilibrated in buffer A containing20 mM Hepes, pH 7.6, 200 mM NaCl, 0.7 M ammonium sulfate, 10% glycerol,and 5 mM BME. The column was stepped to 30% buffer B containing 20 mMHepes, pH 7.6, 40 mM NaCl, 10% glycerol, and 5 mM BME. The column wasdeveloped with a 30–100% gradient over five column volumes. Fractionscontaining the HAT module were concentrated and loaded onto a HiPrep26/60 Sephacryl S-400 (GE Life Sciences) gel filtration column equilibratedin 20 mM Hepes, pH 7.6, 100 mM NaCl, and 200 μM Tris(2-carboxyethyl)phosphine (TCEP). The purified HAT module was concentrated to 4–10 mg/mL,flash-frozen in liquid nitrogen, and stored at −80 °C until use. Extinction co-efficients were calculated using the ProtParam tool from the ExPASy Bio-informatics Resource Portal (www.expasy.org).

Cells expressing Sgf29 were grown and lysed as described above for the HATmodule except that 50 μg/mL carbenicillin and 31 μg/mL chloramphenicol wereused as antibiotics. Clarified lysate was loaded onto a HisTrap HP column (GELife Sciences) equilibrated in buffer containing 20 mM Hepes, pH 7.6, 500 mMNaCl, 20 mM imidazole, pH 8, and 5 mM BME. After washing with 10 columnvolumes of buffer, the proteins were eluted with 300 mM imidazole and thenwere dialyzed overnight into 20 mM Hepes, pH 7.5, 100 mM NaCl, and 5 mMBME. The thioredoxin/hexahistidine tags were removed by adding 1 mg of TEVprotease per 10 mg recombinant protein during dialysis. The dialyzed sampleswere loaded onto the HisTrap HP column, and the tag was separated using agradient of 0–200 mM imidazole over 10 column volumes. Fractions containingpure protein were dialyzed overnight at 4 °C against 20 mM Hepes, pH 7.6,100 mMNaCl, and 200 μM TCEP. The proteins were concentrated to ∼5 mg/mL,flash-frozen in liquid nitrogen, and stored at −80 °C.

Sortase was purified as previously described (36). Purified protein was di-alyzed overnight at 4 °C into 50 mM Hepes, pH 7.5, 150 mM NaCl, and 1 mMDTT, concentrated to 35–48 mg/mL, flash-frozen in liquid nitrogen, and storedat −80 °C.

Recombinant X. laevis histones, including the mutants used for differentiallabeling and the truncated version of histone H3 for enzyme-mediated liga-tion in which the first 32 amino acids were deleted (H3Δ32), were purified aspreviously described (57). DNA for recombinant mononucleosomes was puri-fied from an EcoRV digest of pST55-16xNCP601, a kind gift from Song Tan(Pennsylvania State University, State College, PA) as described previously(58). Nucleosomes were assembled using the salt gradient dialysis method(57) and then were dialyzed into low-salt buffer containing 10 mMTris·HCl, pH 7.5, 5 mM KCl, and 1 mM DTT for storage at 4 °C, where theywere used within 6 wk.

Enzyme-Mediated Ligation of Methylated Peptide to Histone H3 with Sortase.Peptide for enzyme-mediated ligation with the sequence H2N-ART(K-me3)QTARKSTGGKAPRKQLATKAARKSAPATGGK-NH2 was purchased at >85%purity (United Peptide). The peptide was solubilized in MilliQ water at aconcentration of 20 mM and was dialyzed against three changes of MilliQwater at 4 °C using cellulose ester dialysis tubing with a molecular masscutoff of 100–500 kDa (Spectrum Labs). After dialysis, peptide concentra-tions were determined with a bicinchoninic acid (BCA) assay using BSA as astandard (Thermo Scientific) read in a POLARstar Omega plate reader (BMGLabtech), and then stored at −20 °C. Full-length histone H3 containing tri-methylated H3K4 and either the wild-type H3 sequence or a C110A/Q125Cdouble mutant were prepared as previously described (36) with the fol-lowing modifications: Ligation reactions were run in assay buffer containing50 mM Hepes, pH 7.6, 150 mM NaCl, 10 mM CaCl2, and 1 mM DTT. Eachreaction contained 500 μM 35-mer H3 peptide, 90 μM H3Δ32, and 300 μMsortase enzyme. Because the globular form of histone H3 (missing the first32 amino acids) has poor solubility in water, lyophilized protein was firstdissolved at a concentration of 8–10 mM in solubilization buffer [6 Mguanidine·HCl, 20 mM Hepes (pH 7.6), and 10 mM DTT] before diluting itinto the reaction. The reactions were incubated at 30 °C for 16–18 h, during

which the globular and ligated histones gradually precipitated while thesortase enzyme and histone peptide remained soluble. The soluble and in-soluble portions of the sortase reaction were separated by spinning at21,000 × g in a microcentrifuge for 10 min at room temperature. The pelletwas resuspended in five reaction volumes of buffer containing 7 M deion-ized urea, 20 mM Hepes, pH 7.5, 5 mM BME, and 0.5 mM EDTA disodium saltand was loaded onto a MonoS HR 5/5 column (GE Life Sciences) equilibratedin the same buffer. To separate full-length histone from unligated substrate,the column was developed with a 0.1–1 M NaCl gradient over 40 columnvolumes. Fractions containing the full-length H3K4me3 histone were pooledand dialyzed at 4 °C against three changes of MilliQ water containing0.1 mM PMSF and 5 mM BME. Dialyzed histone was lyophilized and resus-pended in a small amount of MilliQ water, and the concentration was de-termined spectrophotometrically using the same extinction coefficient(4040 M−1·cm−1) as for full-length H3. Equal amounts of purified H3K4me3histone H3 and full-length bacterially expressed histone H3 ranging from0.5–10 μg were run on an SDS/PAGE gel and stained with Coomassie BrilliantBlue. To correct the extinction coefficient of purified H3K4me3 for thepresence of contaminants, band intensities were compared in ImageJ be-tween the samples to calculate a correction factor. Typically, the extinctioncoefficient for recombinant histone H3 overestimated the concentration ofH3K4me3 by a factor of 1.1–1.5, and this correction factor was applied whencalculating the H3K4me3 concentration at the octamer reconstitution step.The protein was lyophilized for long-term storage at −20 °C.

Fluorescent Labeling of Histones and Nucleosomes. Histone H3 C110A/Q125Cwas used for fluorescent labeling with cysteine-reactive dyes because theendogenous cysteine in Xenopus histone H3 (C110) points to the interior ofthe histone octamer and is not in a good position to accommodate a fluo-rescent dye, whereas Q125 is located on a solvent-exposed helix. Lyophilizedhistones were resuspended and reduced in 6 M guanidine·HCl, 20 mMHepes, pH 7.5, and 400 μM TCEP for 30–60 min at room temperature at aconcentration of 4 mg/mL. Meanwhile, 1 mg of Alexa Fluor 488 C5 maleimideand 1 mg of Alexa Fluor 647 C2 maleimide (Life Technologies) wereeach dissolved in 100 μL of DMSO at final concentrations of 13.9 mM and7.7 mM, respectively. Histones bearing different posttranslational modifi-cations were labeled with unique colors; unmodified H3 was labeled withAlexa Fluor 647, and H3K4me3 was labeled with Alexa Fluor 488. Unfolded/reduced histones were diluted to a concentration of 2 mg/mL with labelingbuffer and were incubated with a fivefold molar excess dye for 2 h at roomtemperature with gentle agitation and protection from light. Reactions werequenched with the addition of 1 mM BME and were combined immediatelywith equimolar amounts of the other three histones to reconstitute histoneoctamers as previously described (57). Refolded octamers were purified by gelfiltration, which also removed excess dye that remained after dialysis. Nucle-osomes were reconstituted with differentially labeled octamers as previouslydescribed (57). Labeled nucleosomes were dialyzed into nucleosome storagebuffer [50 mM KCl, 10 mM Tris·HCl (pH 7.5), and 1 mM DTT], concentrated to>5 μM for storage at 4 °C, and were used within 6 wk.

Steady-State Kinetic Assays. Steady-state kinetic measurements using peptidesubstrates were done with a continuous spectrophotometric assay as pre-viously described (42), with a few minor modifications. Briefly, each 50-μLreaction contained 5 mM MgCl2, 1 mM DTT, 0.2 mM thiamine pyrophos-phate (Sigma), 0.2 mM NAD+, 100 mM Hepes, pH 7.6, 50 mM NaCl, 2.5 mMpyruvate, 1 μL of 0.45 U/mg at 13 mg/mL pyruvate dehydrogenase (Sigma),0.5–100 μM acetyl-CoA, 25–50 nM HAT module, and 0–1,400 μM histone H3peptide containing the sequence H2N-ARTKQTARKSTGGKAPRKQLA-COOH(purchased at >90% purity from United Peptide). Peptide concentrationswere determined with a BCA assay using BSA as a standard (Thermo Scien-tific). For titrations in which the peptide concentration was varied, theacetyl-CoA concentration was held at 100 μM. For titrations in which theacetyl-CoA concentration was varied, the peptide concentration was held at400 μM. All the reaction components, except for acetyl-CoA, were assembledin 384-well plates (Greiner Bio-One) and incubated at 30 °C for 5 min. Re-actions were initiated by the addition of acetyl-CoA, and the absorbance at340 nm was monitored continuously using a POLARstar Omega plate reader(BMG Labtech) for 5–60min. Absorbance at 340 nmwas converted into themolarconcentration of NADH using Beer’s Law, assuming e340nm = 6,220 M−1·cm−1. Tocalculate initial rates, NADH production was plotted as a function of time and fitto a line where initial velocity conditions were satisfied, typically within the first3 min. A blank reaction containing the HAT module and acetyl-CoA but nosubstrate was performed for each titration, and the rate was subtracted asbackground from the other reactions. Initial rates were measured in triplicate,normalized to the enzyme concentration, and plotted as a function of substrate

E5468 | www.pnas.org/cgi/doi/10.1073/pnas.1508449112 Ringel et al.

Dow

nloa

ded

by g

uest

on

June

10,

202

0

concentration. The resulting curve was fit to the Michaelis–Menten equationusing nonlinear least squares regression implemented in GraphPad Prism 5.

Steady-state titrations with nucleosomal substrates were performed usinga radioactive filter-binding assay (42). Nucleosome concentrations were de-termined spectrophotometrically at 260 nm using the extinction coefficientof the 147-bp DNA fragment: e260nm = 2,346,045 M−1·cm−1. Briefly, samplescontaining 100 mM Hepes, pH 7.6, 1 mM DTT, 0–10 μM unmodified NCP or0–1 μM H3K4me3 NCP, 50 mM NaCl, and 50 nM HAT module were incubatedat 30 °C for 5 min. Reactions were initiated with the addition of acetyl-CoAat 25 μM, containing a 3:1 molar ratio of unlabeled:titrated acetyl-CoA(PerkinElmer) and were quenched by spotting 25 μL onto P81 filter paper.Because proteins and peptides stick to P81 filter but free acetyl-CoA doesnot, acetylation was monitored by scintillation counting with the filter pa-per. At first, six time points were collected at 15-s intervals to determine therange in which product formation was linear with time. Under all conditionstested, initial rate conditions were satisfied for the first minute of the re-action. Subsequent experiments were performed by collecting three timepoints at 15, 30, and 45 s. An unwashed filter was reserved at each substrateconcentration to convert counts per minute to a molar concentration ofacetyl-CoA. Initial rates were determined by plotting product concentrationas a function of time and fitting a line to the data, using the rate of acetyl-transfer in the absence of substrate as a reference. Each experiment wasperformed in duplicate, and kinetic parameters were fit as described for theenzyme-coupled assay earlier in this paper.

HAT Assays for Analysis by Acid–Urea Gel Electrophoresis. Nucleosomes wereacetylated in buffer containing 20mMHepes, pH 7.6, 50mMNaCl, 1 mMDTT,50 μM acetyl-CoA, 20 μg/mL BSA, and 0.2 μM or 1 μM nucleosome coreparticle. Reactions were incubated in buffer at 30 °C for 5 min, initiated bythe addition of 50 nM HAT module, quenched at different time points byflash-freezing in liquid nitrogen, and then lyophilized for analysis by acid–urea gel electrophoresis.

Acid–urea gels were assembled and run as previously described (38). His-tones were visualized with either SYPRO Ruby protein stain (Life Technologies)or by Western blotting. For Western blotting, proteins were transferred toPVDF membrane as previously described (38). Membranes were blockedovernight in 5% nonfat milk at 4 °C and were washed in TBS. Primary anti-bodies were diluted in 1% nonfat milk in TBS supplemented with 0.1% Tween20 (TBST) as follows: anti-H3 (Abcam ab1791, 1:25,000), anti-H3K9ac (ActiveMotif 39917, 1:5,000), anti-H3K14ac (07-353, EMD Millipore, 1:5,000), anti-H3K18ac (EMD Millipore 07-354, 1:7,500), anti-H3K23ac (EMD Millipore 07-355, 1:5,000), or anti-H3K4me3 (Abcam ab8580, 1:5,000). Each primary anti-body was applied for 1 h at room temperature followed by washing in TBST.Goat anti-rabbit IgG-HRP secondary antibody (Amersham Biosciences) wasdiluted to 1:5,000 in 1% nonfat milk and TBST, applied for 1 h at room tem-perature, and washed in TBST. Blots were developed using Pierce ECL WesternBlotting Substrate (Thermo Scientific) and were exposed using film.

Pull-Down Assays. To study the interaction between Sgf29 and the H3K4me3modification, 25 μL of M-280 Streptavidin DynaBeads (Life Technologies) werewashed into histone binding buffer [20 mM Hepes (pH 7.6), 50 mM NaCl,0.02% Tween-20, and 1% BSA] and then were incubated for 1 h at roomtemperature with 1 μg of biotinylated histone H3 21-mer peptide ± H3K4me3,which was generously provided by S.D.T. The beads were washed twice withhistone binding buffer and then twice in complex binding buffer [20 mMHepes (pH 7.6), 50 mM NaCl, 0.02% Tween 20]. Recombinant Sgf29 was added

to the beads at a final concentration of 10 μM and allowed to bind at roomtemperature for 1 h. The beads were washed three times with complexbinding buffer, eluted by boiling with 2× SDS/PAGE loading buffer, run on anSDS/PAGE gel, and stained with Coomassie brilliant blue.

Differential Labeling Assay. Assays using differentially labeled nucleosomeswere run as described in HAT Assays for Analysis by Acid–Urea Gel Electro-phoresis. Control reactions containing only one kind of labeled nucleosomewere performed at substrate concentrations of 0.2 μM and 1 μM. Nucleosomemixtures were prepared by combining 0.2 μM H3K4me3 nucleosome/AlexaFluor 488 with 0.8 μM unmodified nucleosome/Alexa Fluor 647 and runningthe same time-course experiments. Quenched time points were resolved on28-cm acid–urea gels and visualized using a Typhoon Imager (GE Life Sciences).Reactions containing Alexa Fluor 488 were excited using the 473-nm laser andimaged using the 520BP40 filter, and reactions containing Alexa Fluor 647were excited using the 635-nm laser and imaged using the 670BP30 filter.Native gels containing mixtures of the labeled nucleosomes were imaged inthe same way to ensure that there was no spectral overlap between the twodyes (Fig. S9). Merged images were prepared using Adobe Photoshop.

Processivity Assay. Three reactions were prepared for each combination ofnucleosome type/HAT module complex containing 20 mM Hepes, pH 7.6,50 mM NaCl, 1 mM DTT, 50 μM acetyl-CoA, 20 μg/mL BSA, and 0.05 μM fluo-rescently labeled nucleosome in a total volume of 25–35 μL. Reactions con-taining peptide were performed with 50 μM H3K4me3 amino acids 1–8peptide (ART-Kme3-QTAR), which does not contain any free lysine residuesthat can be acetylated by Gcn5. Reactions were incubated for 5 min at 30 °C,initiated with the addition of enzyme to a final concentration of 50 nM, andmixed vigorously by pipetting up and down. After the reaction was allowed toproceed for 10–60 s, reactions were quenched by the addition of s one-fifthvolume acetic acid, mock-quenched with the addition of one-fifth volumenucleosome storage buffer, or subjected to competition by the addition of aone-fifth volume of unlabeled nucleosomes, bringing the concentration of theunlabeled competitor to 1 μM. After the reactions were incubated for a totalof 1–10 min at 30 °C, the samples were flash-frozen in liquid nitrogen, ly-ophilized, resuspended in 10 μL of acid–urea sample buffer, and run on a28-cm acid–urea gel (38). Gels were scanned using a Typhoon imager (GE LifeSciences) with the same excitation wavelengths/emission filter combinations asused in the differential labeling assay. ImageJ was used for background sub-traction and to quantitate band intensities. To show that the concentration ofunlabeled competitor used in the processivity assay acts as an effective chasefor methylated nucleosomes, we measured acetylation by the HAT module onreactions containing 50 nM H3K4me3 nucleosomes labeled with Alexa Fluor488, with 1 μM unlabeled nucleosome added at the beginning of the reaction.Reactions were monitored for the same length of time as the processivity assayusing the wild-type HAT module with methylated nucleosomes. The additionof unlabeled nucleosomes effectively slows acetylation by the HAT module(Fig. S10B) compared with acetylation in the absence of competitor (Fig. S10A).

ACKNOWLEDGMENTS. We thank Greg Bowman (Johns Hopkins University)for providing the X. laevis histone expression plasmids, Song Tan (Pennsyl-vania State University) for providing the pST55-16xNCP601 plasmid, and PhilCole (Johns Hopkins University School of Medicine) for helpful discussions.This work was supported by National Institute of General Medical SciencesGrants R01GM098522 (to C.W.) and R01GM106024 (to S.D.T.) and by Na-tional Science Foundation Grant DGE-1232825 (to A.M.C.).

1. Jenuwein T, Allis CD (2001) Translating the histone code. Science 293(5532):1074–1080.2. Strahl BD, Allis CD (2000) The language of covalent histone modifications. Nature

403(6765):41–45.3. Turner BM (2000) Histone acetylation and an epigenetic code. BioEssays 22(9):836–845.4. Fischle W, Wang Y, Allis CD (2003) Histone and chromatin cross-talk. Curr Opin Cell

Biol 15(2):172–183.5. Musselman CA, Lalonde ME, Côté J, Kutateladze TG (2012) Perceiving the epigenetic

landscape through histone readers. Nat Struct Mol Biol 19(12):1218–1227.6. Suganuma T, Workman JL (2008) Crosstalk among Histone Modifications. Cell 135(4):

604–607.7. Lalonde ME, Cheng X, Côté J (2014) Histone target selection within chromatin: An

exemplary case of teamwork. Genes Dev 28(10):1029–1041.8. Taverna SD, Li H, Ruthenburg AJ, Allis CD, Patel DJ (2007) How chromatin-binding

modules interpret histone modifications: Lessons from professional pocket pickers.

Nat Struct Mol Biol 14(11):1025–1040.9. Pokholok DK, et al. (2005) Genome-wide map of nucleosome acetylation and meth-

ylation in yeast. Cell 122(4):517–527.10. Kurdistani SK, Tavazoie S, Grunstein M (2004) Mapping global histone acetylation

patterns to gene expression. Cell 117(6):721–733.

11. Liu CL, et al. (2005) Single-nucleosome mapping of histone modifications in S. cerevisiae.

PLoS Biol 3(10):e328.12. Millar CB, Grunstein M (2006) Genome-wide patterns of histone modifications in

yeast. Nat Rev Mol Cell Biol 7(9):657–666.13. Jiang L, et al. (2007) Global assessment of combinatorial post-translational modifica-

tion of core histones in yeast using contemporary mass spectrometry. LYS4 trimeth-

ylation correlates with degree of acetylation on the same H3 tail. J Biol Chem 282(38):

27923–27934.14. Taverna SD, et al. (2007) Long-distance combinatorial linkage between methylation

and acetylation on histone H3 N termini. Proc Natl Acad Sci USA 104(7):2086–2091.15. Noma K, Grewal SIS (2002) Histone H3 lysine 4 methylation is mediated by Set1 and

promotes maintenance of active chromatin states in fission yeast. Proc Natl Acad Sci

USA 99(Suppl 4):16438–16445.16. Lee KK, Workman JL (2007) Histone acetyltransferase complexes: One size doesn’t fit

all. Nat Rev Mol Cell Biol 8(4):284–295.17. Vermeulen M, et al. (2010) Quantitative interaction proteomics and genome-wide

profiling of epigenetic histone marks and their readers. Cell 142(6):967–980.18. Koutelou E, Hirsch CL, Dent SYR (2010) Multiple faces of the SAGA complex. Curr Opin

Cell Biol 22(3):374–382.

Ringel et al. PNAS | Published online September 23, 2015 | E5469

BIOCH

EMISTR

YPN

ASPL

US

Dow

nloa

ded

by g

uest

on

June

10,

202

0

19. Huisinga KL, Pugh BF (2004) A genome-wide housekeeping role for TFIID and a highlyregulated stress-related role for SAGA in Saccharomyces cerevisiae.Mol Cell 13(4):573–585.

20. Bonnet J, et al. (2014) The SAGA coactivator complex acts on the whole transcribedgenome and is required for RNA polymerase II transcription. Genes Dev 28(18):1999–2012.

21. Lee KK, et al. (2011) Combinatorial depletion analysis to assemble the network archi-tecture of the SAGA and ADA chromatin remodeling complexes. Mol Syst Biol 7:503.

22. Balasubramanian R, Pray-Grant MG, Selleck W, Grant PA, Tan S (2002) Role of theAda2 and Ada3 transcriptional coactivators in histone acetylation. J Biol Chem277(10):7989–7995.

23. Grant PA, et al. (1997) Yeast Gcn5 functions in two multisubunit complexes to acet-ylate nucleosomal histones: Characterization of an Ada complex and the SAGA (Spt/Ada) complex. Genes Dev 11(13):1640–1650.

24. Saleh A, Lang V, Cook R, Brandl CJ (1997) Identification of native complexes con-taining the yeast coactivator/repressor proteins NGG1/ADA3 and ADA2. J Biol Chem272(9):5571–5578.

25. Grant PA, et al. (1999) Expanded lysine acetylation specificity of Gcn5 in nativecomplexes. J Biol Chem 274(9):5895–5900.

26. Horiuchi J, Silverman N, Marcus GA, Guarente L (1995) ADA3, a putative transcrip-tional adaptor, consists of two separable domains and interacts with ADA2 and GCN5in a trimeric complex. Mol Cell Biol 15(3):1203–1209.

27. Candau R, Berger SL (1996) Structural and functional analysis of yeast putativeadaptors. Evidence for an adaptor complex in vivo. J Biol Chem 271(9):5237–5245.

28. Han Y, Luo J, Ranish J, Hahn S (2014) Architecture of the Saccharomyces cerevisiaeSAGA transcription coactivator complex. EMBO J 33(21):2534–2546.

29. Kuo YM, Andrews AJ (2013) Quantitating the specificity and selectivity of Gcn5-mediated acetylation of histone H3. PLoS One 8(2):e54896.

30. Bian C, et al. (2011) Sgf29 binds histone H3K4me2/3 and is required for SAGA complexrecruitment and histone H3 acetylation. EMBO J 30(14):2829–2842.

31. Schram AW, et al. (2013) A dual role for SAGA-associated factor 29 (SGF29) in ER stresssurvival by coordination of both histone H3 acetylation and histone H3 lysine-4 tri-methylation. PLoS One 8(7):e70035.

32. Shukla A, Lahudkar S, Durairaj G, Bhaumik SR (2012) Sgf29p facilitates the re-cruitment of TATA box binding protein but does not alter SAGA’s global structuralintegrity in vivo. Biochemistry 51(2):706–714.

33. Clouaire T, Webb S, Bird A (2014) Cfp1 is required for gene expression-dependent H3K4trimethylation and H3K9 acetylation in embryonic stem cells. Genome Biol 15(9):451.

34. Tanner KG, et al. (1999) Catalytic mechanism and function of invariant glutamic acid173 from the histone acetyltransferase GCN5 transcriptional coactivator. J Biol Chem274(26):18157–18160.

35. Tanner KG, Langer MR, Kim Y, Denu JM (2000) Kinetic mechanism of the histoneacetyltransferase GCN5 from yeast. J Biol Chem 275(29):22048–22055.

36. Piotukh K, et al. (2011) Directed evolution of sortase A mutants with altered substrateselectivity profiles. J Am Chem Soc 133(44):17536–17539.

37. Popp MW-L, Ploegh HL (2011) Making and breaking peptide bonds: Protein engi-neering using sortase. Angew Chem Int Ed Engl 50(22):5024–5032.

38. Shechter D, Dormann HL, Allis CD, Hake SB (2007) Extraction, purification and analysis

of histones. Nat Protoc 2(6):1445–1457.39. Cieniewicz AM, et al. (2014) The bromodomain of Gcn5 regulates site-specificity of

lysine acetylation on histone H3. Mol Cell Proteomics 13(11):2896–2910.40. Jin Q, et al. (2011) Distinct roles of GCN5/PCAF-mediated H3K9ac and CBP/p300-

mediated H3K18/27ac in nuclear receptor transactivation. EMBO J 30(2):249–262.41. Spedale G, Timmers HT, Pijnappel WW (2012) ATAC-king the complexity of SAGA

during evolution. Genes Dev 26(6):527–541.42. Berndsen CE, Denu JM (2005) Assays for mechanistic investigations of protein/histone

acetyltransferases. Methods 36(4):321–331.43. Cornish-Bowden A (1984) Enzyme specificity: Its meaning in the general case. J Theor

Biol 108(3):451–457.44. Hudson BP, Martinez-Yamout MA, Dyson HJ, Wright PE (2000) Solution structure and

acetyl-lysine binding activity of the GCN5 bromodomain. J Mol Biol 304(3):355–370.45. Owen DJ, et al. (2000) The structural basis for the recognition of acetylated histone H4

by the bromodomain of histone acetyltransferase gcn5p. EMBO J 19(22):6141–6149.46. Kurabe N, et al. (2007) Deregulated expression of a novel component of TFTC/STAGA

histone acetyltransferase complexes, rat SGF29, in hepatocellular carcinoma: Possible

implication for the oncogenic potential of c-Myc. Oncogene 26(38):5626–5634.47. Santos-Rosa H, et al. (2002) Active genes are tri-methylated at K4 of histone H3.

Nature 419(6905):407–411.48. Barski A, et al. (2007) High-resolution profiling of histone methylations in the human

genome. Cell 129(4):823–837.49. Johnsson A, et al. (2009) HAT-HDAC interplay modulates global histone H3K14

acetylation in gene-coding regions during stress. EMBO Rep 10(9):1009–1014.50. Wang Z, et al. (2008) Combinatorial patterns of histone acetylations and methylations

in the human genome. Nat Genet 40(7):897–903.51. Kuo MH, et al. (1996) Transcription-linked acetylation by Gcn5p of histones H3 and H4

at specific lysines. Nature 383(6597):269–272.52. Taverna SD, et al. (2006) Yng1 PHD finger binding to H3 trimethylated at K4 pro-

motes NuA3 HAT activity at K14 of H3 and transcription at a subset of targeted ORFs.

Mol Cell 24(5):785–796.53. Chatterjee C, Muir TW (2010) Chemical approaches for studying histone modifica-

tions. J Biol Chem 285(15):11045–11050.54. Li S, Shogren-Knaak MA (2008) Cross-talk between histone H3 tails produces co-

operative nucleosome acetylation. Proc Natl Acad Sci USA 105(47):18243–18248.55. Ranjan A, et al. (2013) Nucleosome-free region dominates histone acetylation in

targeting SWR1 to promoters for H2A.Z replacement. Cell 154(6):1232–1245.56. Nguyen UT, et al. (2014) Accelerated chromatin biochemistry using DNA-barcoded

nucleosome libraries. Nat Methods 11(8):834–840.57. Luger K, Rechsteiner TJ, Richmond TJ (1999) Preparation of nucleosome core particle

from recombinant histones. Methods Enzymol 304:3–19.58. Dyer PN, et al. (2004) Reconstitution of nucleosome core particles from recombinant

histones and DNA. Methods Enzymol 375:23–44.

E5470 | www.pnas.org/cgi/doi/10.1073/pnas.1508449112 Ringel et al.

Dow

nloa

ded

by g

uest

on

June

10,

202

0