Báo cáo khoa học: Directed evolution of a histone acetyltransferase – enhancing thermostability, whilst maintaining catalytic activity and substrate specificity doc

We have developed a directed evolution protocol that allows the screening of hundreds of histone acetyltransferase mutants for histone acetylating activity, and used this to enhance the

enhancing thermostability, whilst maintaining catalytic activity and substrate specificity

Hans Leemhuis1, Karl P Nightingale1,2and Florian Hollfelder1

1 Department of Biochemistry, University of Cambridge, UK

2 Chromatin and Gene Expression Group, Medical School, University of Birmingham, UK

The post-translational modification of the histone

N-terminal ‘tails’ plays a central role in the epigenetic

regulation of gene expression [1] These processes are

highly integrated into transcriptional control

mecha-nisms, with many histone modifying enzymes being

associated with core components of the transcriptional

machinery A diverse group of these enzymes generate

various marks in chromatin by the covalent

modifica-tion of the histone tails (i.e phosphorylamodifica-tion,

methyla-tion, acetylamethyla-tion, etc.) The histone-code hypothesis

[2,3] suggests that these marks define the functional

status of the underlying DNA, leading to transcrip-tional activation or silencing via the recruitment of specific effector proteins

Histone acetylation is the process of acetylation of specific lysine residues in histones (Fig 1) by a range of histone acetyltransferase (HAT; EC 2.3.1.48) enzymes and typically leads to gene activation Histone acetyla-tion exerts funcacetyla-tional effects via two mechanisms First,

it is associated with the charge neutralization of lysine residues, thereby reducing the interaction of the lysine-rich (and thus positively charged) histone tails with

Keywords

acetylation; chromatin; enzymology;

epigentics; protein engineering

Correspondence

F Hollfelder, Department of Biochemistry,

80 Tennis Court Road, University of

Cambridge, Cambridge CB2 1GA, UK

Fax: +44 1223 766002

Tel: +44 1223 766048

E-mail: fh111@cam.ac.uk

Website: http://www.bioc.cam.ac.uk/uto/

hollfelder.html

(Received 1 August 2008, revised 5

September 2008, accepted 17 September

2008)

doi:10.1111/j.1742-4658.2008.06689.x

Histone acetylation plays an integral role in the epigenetic regulation of gene expression Transcriptional activity reflects the recruitment of oppo-sing classes of enzymes to promoter elements; histone acetyltransferases (EC 2.3.1.48) that deposit acetyl marks at a subset of histone residues and histone deacetylases that remove them Many histone acetyltransferases are difficult to study in solution because of their limited stability once purified

We have developed a directed evolution protocol that allows the screening

of hundreds of histone acetyltransferase mutants for histone acetylating activity, and used this to enhance the thermostability of the human P⁄ CAF histone acetyltransferase Two rounds of directed evolution significantly stabilized the enzyme without lowering the catalytic efficiency and substrate specificity of the enzyme Twenty-four variants with higher thermostability were identified Detailed analysis revealed twelve single amino acid mutants that were found to possess a higher thermostability The residues affected are scattered over the entire protein structure, and are different from muta-tions predicted by sequence alignment approaches, suggesting that sequence comparison and directed evolution methods are complementary strategies

in engineering increased protein thermostability The stabilizing mutations are predominately located at surface of the enzyme, suggesting that the protein’s surface is important for stability The directed evolution approach described in the present study is easily adapted to other histone modifying enzymes, requiring only appropriate peptide substrates and antibodies, which are available from commercial suppliers

Abbreviations

DSC, differential scanning calorimetry; HAT, histone acetyltransferase; HDAC, histone deacetylase.

DNA The subsequent decreased compaction of

chro-matin facilitates access of DNA binding proteins (i.e

transcription factors) Second, acetylated lysine residues

are recognized by ‘bromodomains’, a specific protein

fold found in many transcriptional regulators and

chromatin remodellers, suggesting that these proteins

are recruited to regions of acetylated chromatin, and

subsequently contribute to gene activation

HAT and histone deacetylase (HDAC) activities are

typically found in multi-subunit complexes, which are

recruited to their target loci by interactions with

tran-scriptional activators, or repressors, respectively [4]

Several classes of histone modification enzymes,

including HATs [5,6] and histone demethylases, are

less active when studied in vitro (i.e overexpressed as

an individual polypeptide in the absence of interaction

partners) A limitation of the biochemical

characteriza-tion of the HAT enzymes is their low activity and

stability in vitro It is desirable to improve such

mar-ginally stable proteins in order to be able to use and

study them in biochemical experiments

Proteins from thermophiles [7] have adapted various

strategies [8] to make them more thermostable,

includ-ing the incorporation of stabilizinclud-ing structural features,

such as an increase in the number of charged residues

and ion pairs [9,10], increased a-helical content [11],

increased structural compactness [12] and entropic

sta-bilization due to an increased lysine to arginine ratio

[13] Improved stability of proteins can be achieved by

design and library-based methods Site-directed

muta-genesis based on sequence alignments and comparison

of 3D structures has been successful in creating

pro-teins with higher thermostability [14–19], although

many designed mutations had no stabilizing effect at

all This emphasizes that our current ability to

inte-grate the lessons from naturally thermostable proteins

into engineering stable new structures is still far from

perfect [20] and has ensured that directed evolution is

increasingly being used to enhance the thermostability

of proteins This approach involves the generation of

genetic diversity in the gene encoding the protein of

interest, followed by screening for mutant proteins

with the desired properties, and has been successfully applied to change or improve enzyme function (sub-strate selectivity and activity) and expression or to enhance the stability of proteins [21–28] The general picture emerging from these studies is that just one or

a few amino acid substitutions can be sufficient to increase the thermostability of a protein by up to tens

of degrees Celsius and that it is generally hard to pre-dict which mutations will be stabilizing

In the present study, we describe a procedure that uses random mutagenesis and subsequent screening to identify substantially more thermostable mutants of the catalytic domain of the human HAT P⁄ CAF, with-out affecting its catalytic activity or substrate specific-ity P⁄ CAF, p300 ⁄ CBP-associating factor, is a trans criptional coactivator with a variable N-terminal, a central HAT domain and a C-terminal bromodomain Several stabilizing mutations at multiple residues throughout the protein were found to generate a more thermostable HAT

Results

Generating P⁄ CAF mutants with increased thermostability

One thousand variants of the catalytic domain of

P⁄ CAF generated by error-prone PCR mutagenesis were screened for HAT activity following a mild heat challenge (22C for 2 h), as shown in Fig 2 This

Fig 1 The acetylation of a lysine residue catalysed by P ⁄ CAF

using acetyl-CoA as the acetyl donor.

Fig 2 Screening procedure used to generate thermostable P ⁄ CAF variants (A) Acetylation of lysine residues is detected by an ELISA protocol using an antibody specific for acetylated lysine residues The signal is amplified by a secondary antibody conjugated to horseradish peroxidase (B) Overview of the selection procedure Step 1: microtitre plates with liquid medium (200 lL) were inocu-lated with single transformants and grown overnight Step 2: 25 lL

of culture was transferred to a second plate containing fresh med-ium with isopropyl thio-b- D -galactoside to induce protein expres-sion Step 3: cells were harvested by centrifugation and lysed with BugBuster Following a heat challenge, the lysates were directly used to detect histone acetyltransferase activity, as shown in (A).

procedure yielded 24 P⁄ CAF mutants with an

increased resistance to thermal inactivation Even the

relatively mild screening conditions applied were able

to reduce the activity of wild-type P⁄ CAF by

approxi-mately 90%, allowing the identification of mutations

that detectably improved the thermostability Of the

24 selected variants, twelve carried a single mutation

and twelve carried two to four mutations Twenty-four

different amino acid mutations were identified in total

(Table 1) The observation that the screening of a

rela-tively small number of library members (1000) lead to

a high percentage of variants displaying higher

thermo-stability may be a reflection of P⁄ CAF being

mono-meric under the assay conditions, whereas it is likely

to be part of multi-subunit complexes in vivo Most

mutations were charge neutral Three mutations

intro-duced a charge, four mutations removed a charge and

two mutations switched a positive and negative charge

At this stage, one of the mutants (V582A) was

intro-duced into the pREST-P⁄ CAF vector, purified and shown to have an apparent melting temperature that was 3C higher than the wild-type catalytic domain of

P⁄ CAF (Fig 3), demonstrating the feasibility of creat-ing thermostable P⁄ CAF mutants by directed evolu-tion

Second round of directed evolution

To further increase the thermostability of P⁄ CAF, a second round of evolution was performed by randomly recombining all 24 selected P⁄ CAF mutants using DNA shuffling Here, a more stringent heat challenge was applied, with a 75 min incubation at 37C Screening of 700 variants yielded seven mutants with enhanced resistance to thermal inactivation compared

to the best mutant (V582A) selected in the first round

of directed evolution Sequencing revealed that four double and three triple mutants had been selected, that all seven variants were combinations of mutations selected in the first round of directed evolution (Table 1) and that no additional mutations were intro-duced during the shuffling procedure

Characterization of selected P/CAF variants

To investigate the thermostability of the selected mutants in more detail, two double and one triple mutant enzymes were expressed and purified to homo-geneity Differential scanning calorimetry (DSC) was used to measure directly the thermal denaturation of the proteins, giving the apparent melting temperature (Tm) All P⁄ CAF enzymes tested showed irreversible thermal unfolding, prohibiting the calculation of the free energy (DG) of unfolding All selected variants unfolded at higher temperatures than wild-type P⁄ CAF, with the L503P⁄ D601G ⁄ Y612C and V582A ⁄ D639E mutants having the highest apparent denaturation temperatures (Fig 3 and Table 2) At this stage, we aimed to deter-mine the contribution of the individual mutations to thermostability, and constructed the L503P, D601G and D639E mutants by site-directed mutagenesis DSC of the single mutants showed that D601G, V582A and D639E made a large contribution to the increased denaturation temperature of the selected

P⁄ CAF variants (Fig 3) The Y612C mutation repro-ducibly had the opposite effect, slightly lowering the apparent melting temperature This mutant was still selected because it is stabilizing compared to the wild-type enzyme, under the screening conditions, where there is acetyl-CoA By contrast, the DSC measure-ments are made in the absence of acetyl-CoA, explain-ing the effect of the Y612C mutation

Table 1 Mutations identified in P ⁄ CAF variants displaying higher

thermostability under screening conditions The second column

gives the relative solvent accessibility score as calculated by the

software ASA-VIEW [35].

First round (22 C)

N504I ⁄ Q519L ⁄ V572A ⁄ V582A 100-14-8-94

Second round (37 C)

a Some mutations were found more than once: L503P ⁄ D601G (·2),

V582A (·4), Y612C (·2) and K627R ⁄ D639E (·2) b

Pro655 is not vis-ible in the structure.

Thermal inactivation

Resistance to thermal inactivation of the P⁄ CAF

mutants was determined by heating protein samples

for 5 min at various temperatures and measuring

resi-dual activity In this case, the reduction in activity

corresponds to the percentage of enzyme molecules

undergoing irreversible inactivation We found that the

temperature at which enzymes lost 50% of their activ-ity (T50) was increased for all mutants examined (Fig 4 and Table 2), with the triple mutant L503P⁄ D601G ⁄ Y612C having the highest T50 value Note that these T50 temperatures are close to the denaturating temperatures for the corresponding enzymes, and that the stabilizing effect (as measured in

C) for both methods is similar (Table 2) Interestingly, the Y612C mutant displayed an increased T50 tempera-ture, consistent with selection in the thermostability screen, despite its lower apparent Tmtemperature Next, we investigated whether the cofactor acetyl-CoA stabilizes P⁄ CAF against thermal inactivation, as observed for two other HATs [29,30], and evaluated whether the effect is comparable for wild-type and mutant enzymes Inactivation experiments were repeated in the presence of saturating acetyl-CoA (100 lm), revealing a large stabilizing effect for acetyl-CoA (5.4–8.7C) for all enzymes examined (Fig 4 and Table 2) Comparison of the T50 and T50AcCoA values indicated that acetyl-CoA binding is particularly stabi-lizing for the V582A, Y612C and V582A⁄ D639E mutants

The time taken for temperature-dependent inactiva-tion was assessed by measuring the activity half-life (t1⁄ 2, 48 C), which was determined at 48C All mutants showed longer activity half-lives than wild-type P⁄ CAF (Table 2), with the most stable mutants having 60- to 70-fold larger t1⁄ 2, 48 C values, broadly following the trend seen for the T50 and Tm tempera-tures The stabilizing effects of single mutations are

Fig 3 Thermal denaturation traces mea-sured by differential scanning calorimetry (A) Multiple P ⁄ CAF mutants and (B) single

P ⁄ CAF mutants (compared to wild-type, wt) Conditions: 20 l M P ⁄ CAF protein in

50 m M sodium phosphate (pH 7.5), 150 m M NaCl and a scan rate of 1 CÆmin)1.

Table 2 Stability parameters of purified P ⁄ CAF and selected

mutants Measurements were performed under the following

con-ditions: 50 m M sodium phosphate (pH 7) and 150 m M NaCl T m ,

apparent melting temperature; T50, temperature at which half of

the initial activity is lost in 5 min; T 50AcCo, temperature at which half

of the initial activity is lost in 5 min in the presence of acetyl-CoA;

t1⁄ 2, 48 C, activity half-life at 48 C Errors for T m and T50values are

less than 0.5 C.

Tmb, c (C)

T50 (C)

T50AcCoA (C)

t1⁄ 2, 48 C (min)

L503P ⁄ D601G ⁄ Y612C a 54.2 55.4 61.4 315 ± 12

V582A ⁄ D639E a

a

Mutants selected in second round of directed evolution. bFor

thermal inactivation curves, see Fig 4 c Measured by differential

scanning calorimetry.

approximately additive in the double and triple mutants for all stability measurements, as expected for mutations that are far apart in the structure Overall, the inactivation experiments clearly demonstrate that the selected P⁄ CAF mutants have a strongly enhanced resistance towards thermal inactivation

Catalytic properties and specificity of the P/CAF enzymes

The kinetic parameters of the wild-type and mutant

P⁄ CAF enzymes for acetylation of the histone H3 and H4 peptides were determined using a continuous assay and synthetic peptide substrates Wild-type P⁄ CAF has

kcatand KMvalues of 12 min)1and 53 lm with the H3 peptide and 0.29 min)1 and 194 lm with the H4 pep-tide, whereas the KM for acetyl-CoA was 0.28 lm (Table 3) Note that the H3 peptide is a much better substrate for the enzyme, with kcat⁄ KMof 3774 s)1Æm)1 versus 25 s)1Æm)1 for the H4 peptide, in agreement with histone H3 being the physiological substrate for

P⁄ CAF [31] The mutations had no significant effect

on the kcatH3 values but the V582A and Y612C muta-tions increased the KMH3 value by ten-fold (Table 3)

An overlay of the P⁄ CAF structure and the Tetrahy-mena HAT, with a bound H3 peptide shows that Tyr612 is located in the binding groove for the peptide substrate, which may explain the higher KMH3 observed with the Y612C mutant By contrast, the basis of the high KMH3of the V582A mutant is unclear because Val582 is far away from the peptide binding groove The Y612C mutation reduced the kcatH4 by three-fold, whereas the other mutations had no signifi-cant effect on kcatvalues (Table 3)

Histone acetylation at distinct histone residues is thought to have variable functional effects We there-fore examined whether the thermostable P⁄ CAF mutants affected the specificity of HAT activity Recombinant histone H4 [32,33] (containing minimal endogenous post-translational modifications) was incu-bated with wild-type and four single mutant P⁄ CAF enzymes and the specificity of acetylation was assessed

by western blotting using antibodies specific for

A

B

D

C

Fig 4 Thermal inactivation curves of P ⁄ CAF and selected mutants In the absence (A, B) and presence (C, D) of 100 l M ace-tyl-CoA, used to determine T50(Table 2) Conditions: 10 l M P ⁄ CAF protein in 50 m M sodium phosphate (pH 7.5) and 150 m M NaCl was incubated for 5 min at various temperatures before measuring the residual activity by ELISA The activity measured without a heat challenge was set to 100% The values derived from these curves are given in Table 2.

acetylation at distinct histone H4 lysines (K5, K8, K12

and K16) Figure 5 shows the specificity profiles of

wild-type P⁄ CAF and its mutants, normalized using

the H4K16ac (i.e the most physiologically abundant

acetyl isoform) Broadly, the wild-type and mutant

enzymes yield very similar patterns of lysine specificity,

indicating that the individual mutations do not impact

on substrate recognition This may be expected

because the immediate sequence environments of many

of these residues are similar [e.g for H4K5, 8 and 12

GK(ac)-G]

Discussion

Location of mutations in the P/CAF structure

Analysis of the location of the thermostabilizing

mutations identified, and comparison with

homolo-gous HAT sequences, shows that they are scattered

throughout the enzyme structure (Fig 6), largely

occurring at nonconserved amino acid residues, with

only two conserved residues being mutated (see Fig S1) Close examination of the P⁄ CAF crystal structure [34] reveals that almost all of these stabiliz-ing mutations are on the surface of the enzyme This observation is reinforced by comparison of the distri-butions of solvent accessible residues in the entire proteins versus the residues selected by directed evolu-tion The proportion of exposed residues (calculated with the programme asa-view) [35] is markedly increased in the selected set (see Fig S2 and Table S1) Together with the analysis of the individual mutations, this suggests that the raised thermostablity

is brought about by optimization of surface residues

Table 3 Kinetic parameters of the wild-type and mutant P ⁄ CAF enzymes at 25 C and pH 7.5 in 150 m M Mes buffer.

k catH3(min)1) K MH3(l M ) k catH4(min)1) K MH4(l M ) K MAcCoA(l M ) a

a KMAcCoA was determined using the H3 peptide.

Fig 5 Lysine specificity of wild-type (wt) and mutant P ⁄ CAF

enzymes on histone H4 Acetylation of lysines 4, 8, 12 and 16 of

histone H4 was measured by western blotting using site specific

anti-acetyl sera The degree of acetylation was normalized to that

detected at lysine 16, which was given a value of 1.0.

Fig 6 Location of thermostabilizing mutations Cartoon represen-tation of the catalytic domain of P ⁄ CAF (crystal structure 1CM0 of the Protein Data Bank) [34] Residues mutated in variants with higher thermostability are indicated by sticks Magenta indicates the residues shown to be stabilizing as a single mutant and grey indicates that the residue was found mutated only in combination with one or more other mutations The catalytic glutamate residue

is shown in green and the bound coenzyme A molecule is shown

in black.

The high rate of surface mutations is consistent with

the likely scenario that P⁄ CAF is part of larger

multi-subunit complexes in vivo Directed evolution of the

monomeric P⁄ CAF under the conditions described

here then replaces the residues originally responsible

for potential binding interactions to partner proteins

by stabilizing residues

The mutations that are most effective in raising the

thermostability of the catalytic domain of P⁄ CAF

(V582A, D601G and D639E) are discussed in more

detail below Mutation V582A introduced a smaller

alanine residue at the position of Val582, a residue

strictly conserved among HATs This residue is in the

vicinity of the coenzyme A molecule bound to

the enzyme and is solvent exposed, indicating that the

V582A mutation increases thermostability by lowering

the hydrophobicity of the protein’s surface The

analo-gous V582D mutation is also likely to derive its

stabilizing effect as a result of reduced surface

hydro-phobicity Alternatively, the effect of this mutation

may be explained by modulating the surface charge

distribution, electrostatic interactions, or a

combina-tion of these effects Similarly, the thermostabilizing

mutation D601G is located in a short loop connecting

an a-helix and a b-strand at the surface of the protein,

with the Asp601 side-chain solvent being exposed The

observation that the D601G mutation is stabilizing is

noteworthy because the introduction of glycines is

usu-ally not considered to be stabilizing [20] Asp601 is not

conserved among GCN5 HAT enzymes, and a glycine

at the equivalent position is found in a few other

HATs (see Fig S1) The source of stabilization is

likely to arise from optimization of the surface charge

distribution by removing the negatively charged

aspar-tate side-chain Alternatively, the release of

conforma-tional strain allowed by the extra flexibility of the

glycine residue may lead to better protein packing and

increased thermostability The latter mechanism of

sta-bilization has been discussed in more detail by Veille

and Zeikus [10]

A D639E mutation significantly increased the

ther-mostability of P⁄ CAF This aspartate residue is found

in most GCN5 HATs (see Fig S1) and is part of a

loop region at the surface of the enzyme with its side

chain solvent being exposed Different conformations

have been observed for this loop in crystal structures

of GCN5 HAT enzymes and a recent study

demon-strated that the movement of this loop correlates with

the different stages of the acetylation reaction [36] A

possible explanation for the stabilizing effect of D639E

is that the longer glutamate side chain provides more

opportunities for salt bridge formation on the surface

of the protein

Implications for protein thermostability engineering

Generally, only a few mutations are sufficient to increase the thermostability of a protein by as much as tens of degrees Celsius [37–40] but the prediction of the these mutations proves to be a challenge One might assume (e.g for structure-based computational approaches that are especially advantageous when no straightforward screening or selection systems are available) that surface residues should not be targeted

to increase thermal stability because their interactions with the solvent were expected to be similar in the native and unfolded state of the protein Instead, inter-nal residues have been preferred targets as in the com-putational optimization of protein structures [41–43] Our study suggests that surface residues play a more important role in P⁄ CAF: such mutations occurred in residues with a clearly higher solvent accessibility rela-tive to the average residue of this protein: 51.7% ver-sus 34.1%; scale 0–100%; as calculated with asa-view [35] For the single mutations shown to be stabilizing

in the present study (eleven in total, excluding P655R, which showed no electron density in the crystal struc-ture), the solvent accessibility was even higher at 59.6% (see Fig S2 and Table S1) Other directed evo-lution studies aimed at the generation of thermostable proteins also frequently identify surface mutations [27]; seven out of eight in the case of an esterase [44], three out of three for an a-glucan phosphorylase [24] and ten out of 12 for a phosphite dehydrogenase [45] Computational optimization of surface charge–charge interactions was able to predict site-directed mutants with increased thermostabilities for five small proteins (with < 100 amino acid residues) [46] This preference for surface mutations may be explained by (a) the rela-tively high percentage of residues located at the protein surface, (b) surface residues forming proportionally fewer interactions than ‘internal’ residues, and there-fore being less likely to cause detrimental effects that offset gains by mutation, or (c) that the surface is rela-tively important to thermostability of proteins In addition, (d) protein surfaces have been suggested to

be important for thermostability of proteins showing irreversible unfolding [47] as in the case with P⁄ CAF This scenario may involve partial unfolding of the protein surface structure that would be addressed by surface mutations

Sequence comparison approaches in the design of proteins with enhanced stability are based on the assumption that there is a positive correlation between the conservation of a residue and its contribution to the stability of the protein Thus, substituting poorly

conserved residues for more conserved residues is

expected to increase the thermostability of a protein,

as shown for GroEL minichaperone [14],

immuno-globulin domains [48], phytase [49] and a few other

proteins [20,50,51] To investigate whether the

stabi-lized P⁄ CAF mutants carry mutations predicted by

sequence comparison, the sequences of 15 homologous

HAT proteins, sharing approximately 25% sequence

identity, were aligned (see Fig S3) We then matched

the stabilizing single mutants with the sequence

align-ment The sequence comparison showed that 31

resi-dues out of 167 (or 19%) in P⁄ CAF, were different

from the amino acid most frequently found at this

position Thus, according to the sequence comparison

approach, these 31 residues are the most likely to be

mutated However, of the twelve stabilizing mutations

found by directed evolution, only the D601G

muta-tion was marginally predominant (with five glycine

versus three aspartate residues at this position) All

other stabilizing mutations introduced less conserved

residues (H600R, F605L, D639E and P655R) or

resi-dues not seen in any of the HATs (Q519R, H524Q,

V582A, H592R, E611K, Y621C and E649D)

Muta-tion V582A even substituted the completely conserved

Val582 These results emhasize the effectiveness of

directed evolution in the identification of

thermostabi-lizing mutants that cannot be predicted by rational

protein design Because the effects of stabilizing

muta-tions are often additive, a combination of sequence

comparison and directed evolution may create even

more stable proteins by combining the advantages of

both methods

Conclusions

Proteins have evolved to be stable and biologically

active under conditions imposed by the cellular

envi-ronment and the habitat of the organism; however,

there is little evolutionary advantage in being more

stable than required Because the majority of

muta-tions are destabilizing, most proteins will be only

mar-ginally stable [52–54] This implies that there is ample

sequence flexibility available to stabilize proteins,

whilst retaining their catalytic capacity In the present

study, we have shown that mutations stabilizing the

catalytic domain of the human histone

acetyltransfer-ase P⁄ CAF are readily identified by directed

evolu-tion These mutations have either no or minimal

effects on the catalytic properties and specificity of

this enzyme The directed evolution method described

is able to generate stable histone modifying enzymes

in a straightforward procedure Furthermore, histone

modifying enzymes are potential targets for drug

screening in ‘epigenetic therapies’ for a range of dis-eases, notably cancers and acute myeloid leukaemia [55–58] A more thermostable target enzyme with identical catalytic properties will facilitate high-throughput drug screening and allow biochemical experiments to be carried out in the absence of poten-tial binding partners that are stabilizing the protein

in vivo

The ELISA screening approach described in the present study is easily adapted to investigate other histone modifying enzymes simply by using another appropriate histone substrate and antibody, which are widely available from commercial suppliers This there-fore represents a general protocol to improve the sta-bility of a wide range of histone modification enzymes for in vitro applications

Experimental procedures

Antibodies and peptides Polyclonal rabbit antibodies specific for the recognition of acetylated lysine residues in histone tails (H4K5ac, H4K8ac, H4K12ac and H4K16ac) were kindly provided by Bryan Turner (University of Birmingham, UK) Goat anti-(rabbit IgG) serum conjugated to horseradish peroxidase was purchased from Sigma (St Louis, MO, USA) Histone tail peptides were synthesized by the PNAC facility (Cam-bridge University, UK): H3, ARTKQTARKSTGGKAPR KQLC, and H4, SGRGKGGKGLGKGGAKRHRKV GGK The H4 peptide is biotinylated at its C-terminus to allow attachment to streptavidin coated microtitre plates

Expression and purification All (mutant) P⁄ CAF proteins (20 kDa) were expressed in Escherichia coli BL21(DE3) from pRSET-P⁄ CAF in liquid medium (10 gÆL)1 NaCl, 10 gÆL)1 yeast extract and

20 gÆL)1 bacto tryptone) with 100 mgÆL)1 ampicillin at

25C This vector expresses the catalytic domain of

P⁄ CAF The proteins were purified by cation exchange and size exclusion chromatography, as described previously [59]

In the latter column, P⁄ CAF eluted as a monomer (at con-centrations above those used in the stability and catalytic assays), suggesting that any effects of mutations on the oligomerization state can safely be excluded, thus simplify-ing the analysis of stability-conferrsimplify-ing mutations

DNA manipulation The gene encoding the catalytic domain (amino acids 492– 658) of P⁄ CAF [60] was amplified from pRSET-P ⁄ CAF using the primers P⁄ CAF-for-BamHI and P ⁄ CAF-rev-PstI The resulting PCR product was restricted with BamHI and

PstI and cloned into pMalC2x (New England Biolabs,

Beverly, MA, USA), yielding plasmid pMal-P⁄ CAF This

construct expresses P⁄ CAF as C-terminal fusion to the

maltose binding protein (see Fig S4) DNA sequencing was

carried out with the T7 or Seq oligonucleotides L503P,

D601G and D639E mutations were introduced in

pRSET-P⁄ CAF using the QuickChange system (Stratagene, La

Jolla, CA, USA) Oligonucleotide sequences are provided in

the Table S2

First generation library construction

In the first round of directed evolution, the catalytic

domain of P⁄ CAF was expressed as C-terminal fusion to

maltose binding protein to minimize the effects of the

random mutations on the amount of protein expressed

Random mutations were introduced in the gene encoding

P⁄ CAF by error-prone PCR amplification in the presence

of manganese chloride Additional details on directed

evo-lution procedures are provided elsewhere [61] Optimal

error-prone PCR conditions were determined by sequencing

P⁄ CAF genes amplified at various MnCl2 concentrations

As expected, the number of nucleotide mutations increased

with increasing MnCl2 concentrations with a clear

prefer-ence for T > C and A > G substitutions The library

con-structed with 0.075 mm MnCl2 was used for screening

because it had one to two nucleotide mutations per gene,

and two to four changes per 1000 bp The gene for P⁄ CAF

was amplified from plasmid pMal-P⁄ CAF using primers

pMal-forII and pMal-revII (see Table S2), which anneal at

the BamHI and PstI restriction sites used for cloning; for a

schematic view of the gene and its flanking regions, see the

Fig S4 PCRs (50 lL) contained: 1· Taq DNA polymerase

buffer, 0–0.5 mm MnCl2, 0.4 mm of each dNTP, 10 ng of

pMal-P⁄ CAF template, primers at 0.5 lm and 4 U Taq

DNA Polymerase (New England Biolabs) The PCR

pro-gram used was: 1 min at 94C; 30 cycles each comprising

45 s at 94C, 45 s at 54 C and 45 s at 72 C; followed by

5 min at 72C PCR products were digested with BamHI

and PstI, purified with a QIAquick PCR purification Kit

(Qiagen, Valencia, CA, USA) and cloned into pMalC2x by

incubating overnight with T4 DNA ligase (New England

Biolabs) at 16C Ligated plasmids were transformed into

E coli DH5a cells (Invitrogen, Carlsbad, CA, USA) and

plated onto LB agar with 100 mgÆL)1 ampicillin Plasmids

were purified from pooled transformants and stored at

)20 C

Second generation library construction

Plasmid DNA of all 24 P⁄ CAF variants selected in the first

round of directed evolution were mixed and used as

tem-plates in a PCR with Pfu-DNA polymerase (Stratagene)

with primers pMal-for and pMal-rev These primers anneal

approximately 100 bp upstream and downstream of the

gene for P⁄ CAF (see Fig S4) The PCR product, purified with a QIAquick PCR Purification Kit (Qiagen) was digested with DNAse (3 U; Promega, Madison, WI, USA)

at 15C for 3, 4 and 5 min Digestion was stopped by add-ing EDTA to 10 mm Followadd-ing separation on an agarose gel (2%), fragments between 50 and 100 bp were isolated with the Qiagen II gel extraction kit (Qiagen) A re-assem-bly reaction (50 lL) contained 100 ng of DNA fragments, 0.4 mm of each dNTP, 1· Pfu reaction buffer and 2.5 U Pfu DNA polymerase The re-assembly reaction (1 min at

94C; 70 cycles each comprising 40 s at 94 C, 30 s at

50C, 30 s at 94 C; followed by 4 min at 72 C) yielded a product of expected size ( 750 bp) plus a smear of smaller and larger products Aliquots of this re-assembly reaction were used as PCR templates to amplify the shuffled P⁄ CAF genes using Pfu-DNA polymerase and primers pMal-forII and pMal-revII, obtaining a product of the expected size ( 550 bp) This fragment was digested with BamHI and PstI and cloned into pRSET (Invitrogen) Ligated plasmids were transformed into E coli DH5a cells, plated onto LB agar with 100 mgÆL)1ampicillin, and plasmids were purified from the pooled transformants to yield the second genera-tion library

Screening procedure The plasmid libraries were transformed into E coli BL21(DE3), plated on LB agar supplemented with

100 mgÆL)1 ampicillin and resulting colonies were trans-ferred to 200 lL of LB medium in 96-well plates A sche-matic view of the screening procedure is provided in Fig 2 After overnight incubation at 37C and shaking at

750 r.p.m., 25 lL was transformed to a fresh 96-well plate containing 200 lL of LB and 0.25 mm isopropyl thio-b-d-galactoside followed by incubation at 37C (750 r.p.m shaking speed) for 4 h Cells were harvested by centrifugation (4000 g) To release the P⁄ CAF activity, cells were broken by resuspension in a mixture of

BugBust-er (Novagen, Madison, WI, USA) and deminBugBust-eralized watBugBust-er (40 : 60, v⁄ v; total volume of 20 lL) and incubated at room temperature (20–25C) for 15 min The plates were subsequently incubated for 2 h at 4 and 22C, respec-tively, or for 75 min at 37C The remaining HAT activity after exposure to elevated temperatures was determined by ELISA, as described below Plasmid DNA of clones expressing a more stable P⁄ CAF was isolated and used to re-transform E coli DH5a cells to verify the apparent increase in thermal stability

ELISA procedure ELISA plates were prepared by incubating clear 96-well streptavidin coated microtitre plates (SigmaScreen; Sigma) with biotinylated H4 peptide (0.5 lg in 50 lL NaCl⁄ Piper well) at room temperature for 30 min After rinsing with

water (·4), plates were blocked with BSA in NaCl ⁄ Pi(3%,

w⁄ w) at room temperature for 1 h Plates were then rinsed

with water (·4) and HAT reactions were started by the

addition of reaction buffer [5· reaction buffer: 250 mm

Tris⁄ HCl (pH 8.0), 5 mm dithithreitol, 0.5 mm EDTA and

50% glycerol (v⁄ v), 200 lm acetyl-CoA (Sigma)] and either

purified enzyme or lysate of E coli cells expressing P⁄ CAF

protein Reactions were continued for 10–45 min at room

temperature and stopped by washing (·5) with water The

degree of acetylation was determined using an antibody

specific for acetylated K8 in the H4 peptide, by diluting the

antibodies 200-fold in NaCl⁄ Piwith 3% BSA and

incubat-ing at room temperature for 1 h Plates were washed with

NaCl⁄ Pi⁄ 0.5 m NaCl ⁄ 1% Tween-20 (·2), NaCl ⁄ Pi⁄ 1%

Tween-20 (·2) and water (·3) The secondary antibody, an

anti-(rabbit IgG) horseradish peroxidase conjugate, was

added at a dilution of 1000-fold in NaCl⁄ Piwith BSA (3%,

w⁄ w) and incubated for 30 min Following washing with

NaCl⁄ Pi⁄ 1% Tween-20 (·3) and water (·4), 100 lL of

peroxidase substrate (tetramethylbenzidine; KPL, Silver

Spring, MD, USA) was added and incubated for 5–10 min

This peroxidase reaction was stopped by the addition of

sulfuric acid (50 lL, 1 m) and A450 was measured Under

the conditions described above, the increase in absorbance

obtained from the ELISA is linear (for the wild-type

with-out the exposure to the high temperature) for the time

between 0 and 30 min for acetylation of K8 in the H4

peptide

Continuous HAT assay

Progress curves as a measure of HAT activity of wild-type

and mutant P⁄ CAFs were followed using a coupled enzyme

assay [62] HAT activity generates CoA as a byproduct that

is converted back to acetyl-CoA by pyruvate

dehydroge-nase, which is accompanied by NAD+ reduction to

NADH, thus increasing A340 An example of a progress

curve is provided in the Fig S3 Reactions were carried out

in 96-well plates at 25C Reaction mixtures (300 lL)

contained Mes buffer (pH 7.5, 100 mm), H3 peptide (0–

500 lm) or H4 peptide (0–700 lm), acetyl-CoA (0–

1000 lm), NAD+ (0.2 mm), thiamine pyrophosphate

(0.2 mm), MgCl2 (5 mm), dithiothreitol (1 mm), pyruvate

(2.4 mm) and pyruvate dehydrogenase (0.3 UÆmL)1; Sigma)

Reactions were initiated by the addition of P⁄ CAF enzyme

(i.e the catalytic domain as expressed from

pREST-P⁄ CAF) at a concentration of 0.8–7 lm Linear curves of

reaction progress over time were measured at 340 nm,

con-verted using a value of 5296 AbsÆM)1for e*d The value of

e*d for NADH was determined from a calibration curve

with pure NADH in our microtitre plate setup The time

traces gave initial rates that were fitted to the Michaelis–

Menten equation The measured rates were corrected for

the spontaneous acetyl-CoA hydrolysis (obtained in the

absence of enzyme)

Stability assays Resistance to thermal inactivation was determined by incu-bating the P⁄ CAF protein in sodium phosphate (pH 7.5,

50 mm), NaCl (150 mm), dithithreitol (1 mm) and EDTA (1 mm) buffer All stability assays were performed with

P⁄ CAF protein expressed from pRSET-P ⁄ CAF Protein samples (25 lL, 10 lm P⁄ CAF) were incubated at various temperatures in the range 0–67C for 5 min in a PCR machine followed by incubation on ice for at least 30 min Residual activity was measured by ELISA The activity measured without a heat challenge was set to 100% We define the T50temperature as the temperature at which half

of the initial activity is lost in 5 min The activity half-life

t1⁄ 2, 48 Cof the proteins was determined at 48C by incu-bating the proteins for 0–1800 min at 48C, followed by incubation on ice Residual activity was determined by ELISA and the activity half-life (t1 ⁄ 2, 48 C) is defined as the time in which the activity is reduced by 50%

DSC Thermal unfolding was measured using the MicroCal VP-DSC microcalorimeter (MicroCal Inc., Northampton,

MA, USA), with a cell volume of 0.5 mL Experiments were carried out at a scan rate of 1CÆmin)1from 2–75C

at a constant over-pressure of 1.79 bar (26 psi) Samples were degassed prior to the scan and contained 20 lm (0.40 mgÆmL)1) P⁄ CAF protein (expressed from

pRSET-P⁄ CAF) in sodium phosphate (pH 7.5, 50 mm), NaCl (150 mm), dithiothreitol (1 mm) and EDTA (1 mm) buffer Samples were dialyzed against this buffer and the reference cell contained the dialysis buffer of the last dialysis step The apparent melting temperature (Tm) is defined as mid-point of thermal unfolding as seen in a DSC thermogram

Assessment of wild-type and mutant P/CAF substrate specificity

Substrate specificity of wild-type and single substitution

P⁄ CAF mutants (expressed from pRSET-P ⁄ CAF) was assessed by western blotting using antibodies specific for acetylation at distinct histone H4 lysines (H4K5ac, H4K8ac, H4K12ac and H4K16ac) Histone acetyltransfer-ase assays were performed using 20 lg of E coli expressed histone H4 substrate [32,33] and enzyme in reaction buffer [50 mm Tris⁄ HCl (pH 8.0), 1 mm dithiothreitol, 0.1 mm EDTA, 200 lm acetyl-CoA] at 30C for 3 h The reaction was halted by addition of SDS loading buffer, and sepa-rated on 15% acrylamide : bis-acrylamide (30 : 1) gel, prior

to transfer to nitrocellulose (Hybond-C; Amersham Phar-macia, Piscataway, NJ, USA), western blotting and detec-tion using a fluorescently tagged anti-rabbit secondary (Licor, 800 nm)