Controlling the enzymatic activityof a restriction enzyme by lightBenno Schierlinga, Ann-Josée Noëla, Wolfgang Wendea, Le Thi Hienb, Eugeny Volkovb, Elena Kubarevab, TatianaOretskayab, Michael Kokkinidisc, Andreas Römppd, Bernhard Spenglerd, and Alfred Pingouda,1

aInstitute of Biochemistry, Justus-Liebig University, Giessen, Germany;bChemistry Departement and A.N. Belozersky Institute of Physico-Chemical Biology,M.V Lomonosov Moscow State University, Moscow, Russia;cInstitute of Molecular Biology and Biotechnology, Foundation of Research and Technology,University of Crete, Heraklion, Crete, Greece; and dInstitute of Inorganic and Analytical Chemistry, Justus-Liebig University, Giessen, Germany

Edited by Robert Huber, Max Planck Institute for Biochemistry, Planegg-Martinsried, Germany, and approved November 16, 2009 (received for review August19, 2009)

For many applications it would be desirable to be able to controlthe activity of proteins by using an external signal. In the presentstudy, we have explored the possibility of modulating the activityof a restriction enzyme with light. By cross-linking two suitably lo-cated cysteine residues with a bifunctional azobenzene derivative,which can adopt a cis- or trans-configuration when illuminated byUV or blue light, respectively, enzymatic activity can be controlledin a reversible manner. To determine which residues when cross-linked show the largest “photoswitch effect,” i.e., difference inactivity when illuminated with UV vs. blue light, >30 variants ofa single-chain version of the restriction endonuclease PvuII wereproduced, modifiedwith azobenzene, and tested for DNA cleavageactivity. In general, introducing single cross-links in the enzymeleads to only small effects, whereas with multiple cross-linksand additional mutations larger effects are observed. Some ofthe modified variants, which carry the cross-links close to the cat-alytic center, can be modulated in their DNA cleavage activity by afactor of up to 16 by illumination with UV (azobenzene in cis) andblue light (azobenzene in trans), respectively. The change in activityis achieved in seconds, is fully reversible, and, in the caseanalyzed, is due to a change in Vmax rather than Km.

azobenzene ∣ DNA cleavage ∣ endonuclease ∣ photoswitch ∣ PvuII

Proteins exist in nature whose activity can be controlled bylight; perhaps one of the best known examples is rhodopsin,

which is regulated by the cis∕trans isomerization of its cofactorretinal. For many biological applications it would be desirableto selectively switch the activity of a protein on and off by lightin a similar manner (1). This could be accomplished by the intro-duction of a photosensitive compound into the protein of inter-est. Recent developments in photosensitive compounds such asthe azobenzene derivatives have made the scenario a reality.Azobenzene can be reversibly isomerized between the extendedtrans- and the more compact cis-configuration by illuminationwith UV (trans → cis) or blue-light (cis → trans) as well as bythermal relaxation (cis → trans) (2–4). Four generally applicableapproaches have been used to introduce azobenzene groups intopeptides or proteins: (i) incorporation during peptide synthesis(5–8), (ii) incorporation during in vitro translation (9, 10), (iii) in-corporation in vivo by using an orthogonal tRNA/aminoacyltRNA synthetase pair specific for phenylalanine-4′-azobenzene(11), and (iv) chemical modification of peptides and proteins(3, 12, 13). Another more specific approach is to use azobenzene-modified ligands (e.g., inhibitors) for proteins (14, 15). Chemicalmodification, the most widely used of these approaches, can bedone with mono- or bifunctional azobenzene derivatives. Modi-fication with monofunctional azobenzene derivatives relies onsteric effects (e.g., interference with ligand binding), whereasmodification with bifunctional azobenzene derivatives usuallyis intended to change the conformation of the target molecule,thereby changing its activity. Site-specific modification isachieved by introducing amino residue-specific reacting groups.

For example, bifunctional cross-linkers with a central azobenzenemoiety have been synthesized that contain thiol-specific couplinggroups to cross-link cysteine residues in a highly specific manner(2, 16, 17). It has been shown that peptides and proteins can bemodified with such bifunctional azobenzene derivatives and thatthese chemically modified peptides and proteins can be inducedto change their conformation and/or their activity in a reversiblemanner by illumination with light (16, 18–25). To date, this is themost promising generally applicable method to produce photo-switchable proteins, in particular when structural informationis available (26). Monofunctional azobenzene derivatives attach-ed to proteins have also been used to regulate protein function byillumination in vitro (27–30) and in vivo (31–35). Whereas fewexamples are known in which enzymes have been modified withmonofunctional azobenzene derivatives or with an azobenzene-bearing amino acid such that their catalytic activity could bemodulated by light (9, 10, 36, 37), none so far, to our know-ledge, have been controlled by bifunctional azobenzene cross-linkers.

There are a number of applications where the temporal con-trol of an enzyme activity is desirable. Reengineered site-specificendonucleases, such as homing endonucleases, zinc finger nu-cleases, and restriction endonucleases, are being used for genomeengineering (38). Although highly specific, they exhibit off-sitecleavage over time. Presumably, this phenomenon could be mini-mized by temporal control of their activity, which could beachieved by introducing a “photoswitch.” Here, we report thatwe have succeeded in using the cross-linker 4,4′-bis(maleimi-do)azobenzene to modify a restriction enzyme in order to controlits DNA cleavage activity by photo-induced isomerization of theazobenzene moiety. Our best success relies on a combination ofchemical modification and protein mutagenesis to achieve a rel-ative 16-fold change in activity.

ResultsWe have chosen the single-chain variant of the naturallyhomodimeric restriction enzyme PvuII (scPvuII) for our study(39) because it allows introducing unique amino acid substitu-tions in the N- and C-terminal “halves” of the protein. TheN- and C-terminal halves of the protein correspond to the iden-tical subunits of the homodimeric PvuII, from which scPvuIIwas derived by joining the C-terminus of one subunit with theN-terminus of the other with a short four-amino-acid linker.

Author contributions: W.W. designed research; B. Schierling, A.-J.N., and M.K. performedresearch; L.T.H., E.V., E.K., T.O., and B. Spengler contributed new reagents/analytic tools;A.R. analyzed data; and A.P. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

See Commentary on page 1259.1To whom correspondence should be addressed. E-mail: Alfred.M.Pingoud@chemie.bio.uni-giessen.de.

This article contains supporting information online at www.pnas.org/cgi/content/full/0909444107/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.0909444107 PNAS ∣ January 26, 2010 ∣ vol. 107 ∣ no. 4 ∣ 1361–1366

BIOCH

EMISTR

YSE

ECO

MMEN

TARY

Dow

nloa

ded

by g

uest

on

Apr

il 21

, 202

0

Selection of Azobenzene Attachment (Cross-Link) Sites in Single-ChainPvuII. In order to generate a photoswitchable variant of scPvuII,several surface-exposed residue pairs were chosen [based on thestructure of scPvuII (40) and by using the program sGAL (41)] forcross-linking with the bifunctional azobenzene derivative 4,4′-bis(maleimido)azobenzene (acronym: azomal) (Fig. 1). The Cγ-Cγ-distances of these residue pairs (∼8� 3 Å) approximate thereportedrangeof∼5–12 Åforazobenzene in thecis-configuration,compared to approximately 17 Å in the trans-configuration (17).Residue pairs (Fig. 2) located in α-helices flanking the DNA bind-ing site (V89/Q96, R129/S133, K147/E151) or loops and β-strandsnext to the active site (T49/N62, E66/Q96, I74/E110) wereselected, in addition to residue pairs connecting the linker region(L1–L4, see Materials and Methods) between the two halves ofscPvuII and residues next to the active site (L/D61 and L/N62).The residues chosen were exchanged for cysteine, resulting inscPvuII variants with one azobenzene cross-link site (double-cysteine variants have one cross-link site) in the N-terminal halfof scPvuII and in variants with “equivalent” azobenzene cross-linksites in the N- and C-terminal halves of scPvuII (quadruple-cysteine variants have two cross-link sites). Altogether, 11 variantswith double-cysteine substitutions and 17 with quadruple-cysteinesubstitutions were produced (Table S1) andmodified with azomal.In eight of these, additional amino acid substitutions were intro-duced that lower the activity by interfering with DNA binding(H83A) (42) and catalysis (Y94F) (43).

Azomal Modification and Purification of scPvuII Variants. The mod-ification of the scPvuII variants was performed with azomal in thecis-state (i.e., under illumination with UV light), with an averageyield of 50–80% modified protein. The crude modified proteinpreparation contained, in addition to the desired cross-linkedspecies, side products in which only one cysteine residue wasmodified by a single azomal molecule (not cross-linked) and inwhich two cysteine residues were modified with one azomalmolecule each (not cross-linked). Enrichment of the desiredcross-linked species is essential to detect the magnitude of thephotoswitch effect because other species can distort the results(Table S2). The reaction mixture was therefore subjected to aseries of chromatographic steps as described in Materials andMethods. It was demonstrated by a mass spectroscopic analysis

that cross-linking by the bifunctional sulfhydryl-reactiveazobenzene derivative has been achieved (Fig. S1).

Activity of Azomal-Modified scPvuII-Variants in the cis- and trans-States of Azobenzene. The azomal-modified scPvuII variants weretested for a light-inducible change in activity via the photoswitch.Table 1 shows the ratio of the DNA cleavage activity of allazomal-modified variants measured under UV light (cis) andunder blue light (trans). As expected, the unmodified scPvuIIand scðC49C62Þ2 (i.e., the scPvuII variant carrying the T49Cand N62C substitution in both halves of the protein, see Table 1for naming convention and Fig. 2 for the position of these

Fig. 1. A scheme of the trans → cis isomerization of azomal attached tocysteine residues induced by illumination with UV light and cis → trans isom-erization with blue light. It is apparent that the sulfur–sulfur distance isconsiderably larger in the trans-state than in the cis-state.

Fig. 2. (A) The crystal structure of scPvuII (40) with the residue pairs chosenfor cross-linking with azomal: (C49 C62), ðC61 L1½-GSGC161-�Þ, (C66 C96),(C74* C110*), (C89 C96), (C129 C133), and (C147* C151*). Amino acid residuesof these residue pairs are depicted in identical colors. For clarity, residue pairsare shown mostly on the N-terminal half (Right), only those that are on the“backside” are shown in the C-terminal half (Left), denoted with an asterisk(*). Note that amino acid pairs were cross-linked in the N-terminal half, whenonly one cross-link was introduced into scPvuII, or in equivalent positions ofthe N- and C-terminal halves, when two cross-links were introduced (Table 1).Amino acids of residue pairs chosen for cross-linking were substituted bycysteines via site-specific mutagenesis. The amino acid residues of the activesite (D58, E68, and K70) are indicated in orange. H83 (green) and Y94 (blue),which in some variants were substituted by alanine and phenylalanine,respectively, are also indicated. (B) A blow up of the region comprisingβ-strands β1, β2, and β3 and the catalytic center is shown, together with amodel of azomal cross-linking C49 and C62.

1362 ∣ www.pnas.org/cgi/doi/10.1073/pnas.0909444107 Schierling et al.

Dow

nloa

ded

by g

uest

on

Apr

il 21

, 202

0

residues in the structure of scPvuII) and scðC49C62Þ2ðF94Þ2 havethe same activity under UV and blue-light illumination. On theother hand, several of the azomal-modified variants show a higheractivity when the azobenzene moiety adopts the cis-configurationas compared to the trans-configuration. In general, larger effectsare seen when the azomal modification is in both halves of thescPvuII protein. For example, scðC49C62Þ2Azo has an approxi-mately 7-fold higher activity in the cis-state than in the trans-state;in contrast, the corresponding variant with only one half modified[scðC49C62ÞAzo] showed only an approximately 2-fold increase(Table 1). It is also noteworthy that the azomalmodifications closeto the active site have a larger impact on the cis∕trans activity ratiothan azomal modifications in α-helices flanking the DNA bindingsite. In an effort to increase the cis∕trans activity ratio, additionalamino acid substitutions were introduced independently: H83Aand Y94F, which lower the cleavage activity of wtPvuII (42, 43).The scðC49C62Þ2ðF94Þ2Azo (i.e., the scðC49C62Þ2Azo variantwith the additionalY94F substitution in both halves of the protein)has an approximately 16-fold higher activity in the cis- than in thetrans-state of the azobenzene moiety, the largest effect that we

observed so far for a photoswitchable scPvuII variant (Table 1).A similar photoswitching effect was observed for scðC61L4ÞðC49� C62�ÞðF94Þ2Azo (i.e., the scPvuII variant with a cross-linkbetween C61 and a cysteine in the linker L4, and an additionalC49* C62* cross-link in the C-terminal half, together withY94F substitutions in both halves of the protein).

Activity of Azomal-Modified Homodimeric PvuII Variants in the cis-and trans-States of Azobenzene. Given that the modified scPvuIIvariants with identical modifications in both the N- and C-term-inal halves showed larger photoswitch effects than the variantsthat were only modified in one half, we wondered whether anal-ogous homodimeric PvuII variants when modified in both sub-units would show a similar photoswitch effect. We have analyzedthis for two variants and found that the homodimeric PvuIIvariants showed a lower photoswitch effect (Table 1). Most likelythe homodimeric variant is more flexible than the sc variant andcan adapt better to conformational strains imposed by the cis vs.trans configurations of the azobenzene moiety.

Table 1. Photoswitching effect of different scPvuII variants cross-linked with 4,4′-bis(maleimido)azobenzene

Type of modification Variant

Relativeactivitycis/trans

Unmodified sc 1.0 ± 0.1scðC49C62Þ2 1.0 ± 0.1

scðC49C62Þ2ðF94Þ2 1.0 ± 0.0One azomal

modificationscðC49C62ÞAzo 2.2 ± 0.0scðC74C110ÞAzo 1.1 ± 0.2scðC66C96ÞAzo 2.3 ± 0.4

scðC147C151ÞAzo 1.5 ± 0.0scðC89C96ÞAzo 1.1 ± 0.1

scðC129C133ÞAzo 1.1 ± 0.0Two azomal

modificationsscðC49C62Þ2Azo 7.0 ± 0.3scðC74C110Þ2Azo 3.0 ± 1.0scðC66C96Þ2Azo 3.4 ± 1.0

scðC147C151Þ2Azo 1.1 ± 0.2Two azomal

modifications andadditionalmutation(s)

scðC49C62Þ2ðA83ÞAzo 6.2 ± 0.5scðC49C62Þ2ðA83Þ2Azo 10.0 ± 2.9scðC49C62Þ2ðF94ÞAzo 15.9 ± 2.3scðC49C62Þ2ðF94Þ2Azo 15.5 ± 2.4

scðC49C62Þ2ðA83Þ2ðF94ÞAzo 5.2 ± 1.1scðC49C62Þ2ðA83Þ2ðF94Þ2Azo 2.6 ± 0.2

One azomalmodificationinvolving thelinker

scðC61 L1ÞAzo 3.2 ± 0.3scðC62 L1ÞAzo 1.0 ± 0.1scðC61 L2ÞAzo 0.8 ± 0.2scðC61 L3ÞAzo 1.9 ± 0.1scðC61 L4ÞAzo 2.2 ± 0.4

Two azomalmodifications, oneinvolving thelinker

scðC61 L1ÞðC49� C62�ÞAzo 3.3 ± 0.4scðC62 L1ÞðC49� C62�ÞAzo 3.6 ± 0.5scðC61 L2ÞðC49� C62�ÞAzo 1.8 ± 0.9scðC61 L3ÞðC49� C62�ÞAzo 4.0 ± 0.2scðC61 L4ÞðC49� C62�ÞAzo 6.3 ± 0.9

scðC61 L4ÞðC49� C62�ÞðF94ÞAzo 15.7 ± 2.0scðC61 L4ÞðC49� C62�ÞðF94Þ2Azo 16.3 ± 1.3

Two modifications inthe homodimer

wtðC49C62Þ2Azo 5.2 ± 0.5wtðC49C62Þ2ðF94Þ2Azo 6.3 ± 0.2

All variants produced for this study are indicated with their cysteinesubstitutions (see Table S1 for further information). The ratio of activitiesof the azomal-modified proteins in the trans- and cis-state (� standard de-viation) are given, as derived from plasmid cleavage assays after blue-lightand UV-light preillumination, respectively, under otherwise identical condi-tions. The nomenclature of the variants is as follows: sc(C49 C62) carries cys-teine residues in positions 49 and 62 in the N-terminal half of scPvuII, sc(C49*C62*) carries cysteine residues in the equivalent positions 49 and 62 in theC-terminal half of scPvuII, scðC49C62Þ2 carries cysteine residues in positions49 and 62 in the N- and C-terminal halves, and sc(C49 C62)Azo is the sc(C49C62) variant with an azomal cross-link between the two cysteine residues.

Fig. 3. (A) Activity changes of scðC49C62Þ2Azo induced by preilluminationwith blue light → UV light → blue light → UV light as determined by asubsequent DNA cleavage assay (4 nM DNA, 0.5 nM enzyme) under illumina-tion with blue or UV light, respectively. On the Top the gel electrophoreticanalysis of samples withdrawn from the incubation mixture at defined timeinterval is shown, on the Bottom the corresponding activity vs. time of illu-mination profile. According to the quantitative analysis, scðC49C62Þ2Azo isseven times more active with the azobenzene moiety in the cis- than in thetrans-configuration. (B) Reversible photoswitching of the DNA cleavage ac-tivity of scðC49C62Þ2Azo observed in situ. After preilluminating the enzymewith blue light, DNA cleavage (4 nM DNA, 0.24 nM enzyme) was followed insitu with variation of the illumination during cleavage (blue light 0–6min, UVlight 6–11 min, blue light 11–17 min, UV light 17–22 min). sc, oc, and li denotethe supercoiled, open circular, and linear forms, respectively, of the plasmidDNA substrate.

Schierling et al. PNAS ∣ January 26, 2010 ∣ vol. 107 ∣ no. 4 ∣ 1363

BIOCH

EMISTR

YSE

ECO

MMEN

TARY

Dow

nloa

ded

by g

uest

on

Apr

il 21

, 202

0

Reversibility of Photoswitching.Todemonstrate thatphotoswitchingis reversible, the scðC49C62Þ2Azo variant, which shows acis∕trans activity ratio of approximately 7, was subjected to ablue → UV → blue → UV light illumination protocol, and theactivity was measured under these conditions. As shown inFigs. 3A and Fig. S2, the activity changes are fully reversible.The reversibility was also observed during the cleavage reactionby using alternating illumination of the reaction mixture with blueand UV light (Fig. 3B). Experiments similar to those shown inFig. 3B were also carried out with the scðC49C62Þ2ðF94Þ2Azovariant, which shows a cis∕trans activity ratio of approximately16 (Fig. S3). For this variant the thermal relaxation rate wasmeasured by using a DNA cleavage experiment. Incubation inthe dark after a period of UV-light illumination leads to thethermal relaxation of the enzyme-attached azobenzene moietyfrom the cis- to the trans-state with a half-life of approximately6 h at 4 °C (Fig. 4). It is noteworthy that the relaxation fromthe more active cis state to the less active trans state suppressesoff-site cleavage (Fig. S4).

Mechanistic Analysis of the Photoswitching Effect. Since the largestphotoswitching effect was observed with variants that had themodification close to the active site, we assumed that the effectwas due to differences in Vmax rather than in Km. The Michaelis–Menten analysis showed that the scðC49C62Þ2Azo variant has aVmax of 10.3� 0.9 nM min−1 and a Km of 25.4� 2.9 nM in thecis-state, and a Vmax of 2.2� 0.1 nM min−1 and a Km of 16.7�1.3 nM in the trans-state, meaning that the effect observed ismainly due to differences in Vmax (Fig. 5).

DiscussionWe have been interested in being able to control the activity ofrestriction and homing endonucleases for gene targeting in vivoto prevent off-site cleavage before or after the specific DNAcleavage event has taken place. We focused our studies on therestriction enzyme PvuII, for which we had shown previously thattargeting to specific sites in genomic DNA in vitro can beachieved by fusing it to a triple-helix forming oligonucleotide(TFO) and thereby extending its six-base-pair recognition by

16 base pairs (44). However, for in vivo purposes it is necessaryto control the nuclease activity such that the scPvuII-TFO fusioncan be activated only after a stable triple helix has been formedand inactivated after double-strand cleavage has taken place.Temporal control of nuclease activity would be of interest alsofor other types of meganucleases, i.e., engineered homing endo-nucleases and zinc finger nucleases, to minimize the problem ofoff-site cleavage after long exposure of cells to these enzymes.

Azobenzene has been widely used for reversible switching bylight between alternative conformations of peptides and proteins.Different approaches have been used to incorporate azobenzeneinto peptides and proteins. We have chosen the chemical mod-ification approach using bifunctional azobenzene derivatives tocross-link selected amino acid residues (16).

If one wants to modulate the activity of an enzyme by attachingan azobenzene moiety to two positions and thereby cross-linkingthem, sites must be selected that are either close to the substratebinding site or the catalytic center, or are located at critical re-gions in the protein, which are likely to react in a sensitive mannerwhen modified. Previous studies with azobenzene derivativeshave shown that cross-linking two residues within an α-helix leadsto a light-induced change in helical content (3, 13, 16, 18, 20, 25).Following this strategy, we introduced cross-linking sites in threeα-helices, two of them located in the DNA binding subdomain(C89-C96 and C129-C133, respectively) and one of them inthe catalytic subdomain (C147-C151) (Fig. 2A). A similar strategywas followed by choosing azobenzene cross-linking sites that in-volve β-sheets and loop regions adjacent to the active site (D58,E68, and K70; see Fig. 2A). Residue pairs C49-C62, C74-C110,and C66-C96 were chosen to test this strategy. Cysteines werealso introduced in the linker between the two halves of scPvuIIat different positions as anchor points for azobenzene attachmentto be cross-linked with C61 and C62, located close to the activesite residue D58. Cross-linking these sites with azomal couldsupport catalysis with azobenzene in the cis-state as suggestedby inspection of the scPvuII structure, whereas isomerizationto the trans-state could “push” the β-strands harboring the cata-lytic center apart, which may result in disruption of the integrityof the catalytic center and cause inactivation of the enzymaticfunction.

The DNA cleavage activity of the azobenzene-modifiedscPvuII variants was measured under UV- and blue light to eval-uate the photoswitch effect (Table 1). Variants with modificationof α-helices close to the DNA binding site showed no or only

Fig. 4. Activity changes of scðC49C62Þ2ðF94Þ2Azo. scðC49C62Þ2ðF94Þ2Azowas illuminated with blue light → UV light → blue light → UV light, inter-rupted by a long incubation in the dark. DNA cleavage activity was deter-mined by a DNA cleavage assay (4 nM DNA, 12 nM enzyme) underillumination with blue or UV light and during incubation in the dark, respec-tively. On the Top the gel electrophoretic analysis is shown, and on theBottom the plot of the absolute activities as determined in the respectiveillumination states is shown.

Fig. 5. Steady-state kinetic analysis of DNA cleavage by scðC49C62Þ2Azo.The Upper curve represents the initial rates of DNA cleavage measured underUV-light illumination, and the Lower curve represents the initial ratesmeasured under blue-light illumination. The drawn-out curve is the resultof a nonlinear regression analysis: Whereas scðC49C62Þ2Azo (cis) has a Km

of 25.4� 2.9 nM and a Vmax of 10.3� 0.9 nM min−1, scðC49C62Þ2Azo (trans)has a Km of 16.7� 1.3 nM and a Vmax of 2.2� 0.1 nM min−1 (each with �standard deviation).

1364 ∣ www.pnas.org/cgi/doi/10.1073/pnas.0909444107 Schierling et al.

Dow

nloa

ded

by g

uest

on

Apr

il 21

, 202

0

small changes in activity upon specific illumination (cis∕trans ac-tivity ratio <1.5). Compared to small helical proteins (19, 20, 23),the “mechanical stress” due to the azobenzene isomerization in alarge protein may not be sufficient to distort an α-helix embeddedin a complex tertiary structure, or the conformational change inthe protein induced by the azobenzene isomerization is small andcompatible with preserving the activity of the enzyme. In contrast,the variants that are supposed to influence catalytic site forma-tion show a larger effect (up to approximately 2-fold) in cleavageactivity upon specific illumination, which could be increased(up to approximately 7-fold) by cross-linking also the equivalentpositions in the C-terminal half of the protein (Table 1). Weconclude from these results that regions close to the active siteare sensitive to slight conformational distortions induced bythe azobenzene isomerization.

It could be shown that additional amino acid substitutions thatlower the activity of PvuII can increase the photoswitch effect (upto approximately 16-fold). Amino acid substitutions that affectthe enzyme activity in binding (H83A) or catalysis (Y94F) seemto influence the cleavage rate in the cis- and the trans-state of theazobenzene moiety differently, i.e., the activity in the trans-state ismore affected than the activity in the cis-state, thus the overallphotoswitching effect is increased.

Introducing symmetric substitutions in both halves of the en-zyme seem to give the best results, as in scðC49C62Þ2Azo andscðC49C62Þ2ðF94Þ2Azo. The photoswitching effects in the corre-sponding homodimeric wtPvuII variants (wtðC49C62Þ2Azo andwtðC49C62Þ2ðF94Þ2Azo) are not as pronounced as in thesingle-chain variant; presumably the homodimer might escapethe catalytic site disturbance of the tethered azobenzene moietymore easily due to higher flexibility.

With two of the variants we demonstrated complete revers-ibility of the photoswitching: it could be shown that with ablue → UV → blue → UV, etc. light illumination protocol the in-crease in activity could be repeatedly obtained. Photoswitching isvery fast, as can be seen when the change in the wavelength ofillumination is carried out in the presence of DNA (i.e., duringcatalysis) and the DNA cleavage activity monitored in situ.

We wondered, whether the cis∕trans effect is due to a change inKm or Vmax: Steady-state kinetic experiments demonstrate thatthe Km value in contrast to the Vmax value is largely unchangedupon switching from the cis- to the trans-state. This was expectedfor a variant that carries the cross-link close to the catalyticcenter.

In conclusion, we have succeeded in producing a photo-switchable restriction enzyme, which upon illumination exhibitsan up to 16-fold increase in DNA cleavage activity by switchingthe configuration of the incorporated azobenzene moieties fromtrans to cis. The fact that we were able to increase the cis∕transeffect by combination of several cross-links and introduction ofadditional amino acid substitutions suggests that it could bepossible to eventually obtain a restriction enzyme that can be con-trolled by light, within the limits imposed by the photochemistryof azobenzene derivatives (45), which should make conjugates ofrestriction enzymes with additional DNA recognition modulesincluding zinc fingers more suitable for in vivo gene targeting.

Materials and MethodsProtein Expression and Purification. The mutated scPvuII-gene with the codingsequence for a C-terminal His6-tag was cloned into the expression vector pRIZ′ (39). pRIZ′was used to transform the Escherichia coli strain XL10-Gold, whichwas previously transformed with the pLGM plasmid containing the gene cod-ing for the PvuII methyltransferase. All genetic constructs were confirmed byDNA sequencing over the entire coding region. Cultures were grown at 37 °Cto OD600 ≈ 0.7 and protein expression was induced by the addition of 1 mMIPTG. After 3 h of induction, cells were harvested by centrifugation, andthe pellet was stored at −20 °C. The cell pellet was resuspended in 30 mMK-phosphate pH 7.4, 500 mM NaCl, 0.01% wt/vol Lubrol, 20 mM imidazol,1 mM PMSF and lysed by sonication. Cell debris was removed by centrifuga-

tion (>17; 000 g) for 30 min at 4 °C. The recombinant His-tagged proteinswere purified by affinity chromatography over Ni-NTA agarose (Qiagen)using 30 mM K-phosphate pH 7.4, 500 mM NaCl, 0.01% wt/vol Lubrol,5 mM EDTA, 200 mM imidazol, 5 mM DTT. Fractions containing pure proteinwere dialyzed overnight at 4 °C against 20 mM K-phosphate buffer pH 7.4,containing 150 mM KCl, 50% vol/vol glycerol and stored at −20 °C. Proteinpurification was monitored by SDS-PAGE analysis, and protein concentrationwas determined by absorbance measurements at 280 nm. The accessibility ofthe introduced cysteines was checked by modification of the scPvuII variantswith PEG-maleimide (Pierce), which results in a 2.3-kDa mass shift in SDS-PAGE analysis. DNA cleavage activity was analyzed by using supercoiledpAT153 tri plasmid DNA as substrate, which has a single PvuII site.

Linker Variations. The four-amino-acid linker sequence (-GSGG-) in theoriginal scPvuII was exchanged for another four-amino-acid linker(L1 ≡ -GSGC-) or for different eight-amino-acid linkers (L2 ≡ -GSGCGTGG-,L3 ≡ -GSGTGCGG-, and L4 ≡ -GSGTGSGC-Þ, each with a cysteine residue forazomal attachment.

Protein Labeling with Azomal and Purification of Modified scPvuII Variants. ThescPvuII variants, diluted to 5 μM protein concentration with 50 mMK-phosphate pH 7.4, 500mMNaCl, were incubated with a 2-fold molar excessof azomal (dissolved in DMSO and kept in the cis-configuration by illumina-tion with UV light) over cysteine residues for 10 min at ambient temperature.Cross-linking yield was monitored by additional modification with a 50-foldmolar excess (over cysteines) of PEG-maleimide to an aliquot of the reactionmixture. Purification of protein species with two cysteine residues cross-linked with azomal from protein species in which two neighboring cysteineresidues each carry an azomal group was performed by adding a 125-foldmolar excess (over cysteine residues) of thiolactic acid (which adds extra neg-ative charges and thereby allows removing scPvuII variants with unreactedmaleimide groups by anion exchange chromatography), incubation for10 min at ambient temperature and subsequent centrifugation (5 min,20,000 g, 4 °C), desalting [HiTrap Desalting 5-mL column (GE Healthcare)],and anion exchange chromatography [ResourceQ 1-mL column (GEHealthcare)]. The first azobenzene-containing peak was collected, dialyzedagainst dialysis buffer (see above), and analyzed for photoswitching effectsby an activity assay under specific illumination. Protein concentration wasestimated by SDS-PAGE analysis with silver staining using unlabeled proteinas reference.

Photoisomerization of Azobenzene. Photoisomerization of azobenzene to thetrans-state was achieved by illumination with a blue-light source based on alight-emitting diode (470 nm� 10 nm; 3-μmol photons m−2 s−1) for 2 min at adistance of 15 cm. For isomerization to the cis-state, the sample was illumi-nated for 4 min with a UV lamp (365 nm, 6 W) at a distance of 9 cm.

Activity Assay Under Specific Illumination. Azomal-modified protein prepara-tions were incubated with 4 nM supercoiled pAT153 tri plasmid DNA in cleav-age buffer (10 mM Tris-HCl pH 7.2, 50 mM NaCl, 10 mM DTT, 0.1 mg∕mLbovine serum albumin, 1 mM MgCl2) at 37 °C. Specific illumination was per-formed either by preillumination of the azomal-modified protein prepara-tion followed by a cleavage reaction under specific illumination or byvariation of blue- or UV-light illumination directly during the cleavagereaction. The linearized cleavage product was analyzed by agarose gel elec-trophoresis; gels were stained with ethidium bromide, and the fluorescencewas measured with a BioDocAnalyze system (Biometra) and quantitated.

Steady-State Kinetic Analysis. Michaelis–Menten analysis was performed byusing a [α-32P]-labeled 229-base-pair PCR product as substrate. SubstrateDNA (1.8 to 121.8 nM) was incubated with a 2.4-nM concentration of azomal-modified scPvuII-variants in cleavage buffer containing 10 mM MgCl2 underUV-light and blue-light illumination, respectively, at 37 °C. The reaction prog-ress was analyzed after 0, 0.5, 1, 1.5, 2, 2.5, 3, 5, and 8 min by electrophoresison 10% polyacrylamide gels; substrate and product bands were visualizedwith an InstantImager system (Packard) and quantified by using theInstantImager software. Initial rates were plotted vs. substrate concentrationand analyzed in terms of Km and Vmax by nonlinear regression analysis.

ACKNOWLEDGMENTS. This work has been supported by the German ResearchFoundation and the Russian Foundation for Basic Research by the DFG-RFBRprogram “International Research Training Groups” (Grants GRK 1384,08-04-91974) and the European Union (FP6-2006-15509-NEST). We thankDr. Woolley for fruitful discussions, Dr. Silva and Dr. Friedhoff for criticalreading of the manuscript, and Chun Mei Li for technical assistance.

Schierling et al. PNAS ∣ January 26, 2010 ∣ vol. 107 ∣ no. 4 ∣ 1365

BIOCH

EMISTR

YSE

ECO

MMEN

TARY

Dow

nloa

ded

by g

uest

on

Apr

il 21

, 202

0

1. Gorostiza P, Isacoff EY (2008) Optical switches for remote and noninvasive control ofcell signaling. Science, 322(5900):395–399.

2. Pozhidaeva N, Cormier ME, Chaudhari A, Woolley GA (2004) Reversible photocontrolof peptide helix content: Adjusting thermal stability of the cis state. BioconjugateChem, 15(6):1297–1303.

3. Renner C, Moroder L (2006) Azobenzene as conformational switch in model peptides.ChemBioChem, 7(6):868–878.

4. Sadovski O, Beharry AA, Zhang F, Woolley GA (2009) Spectral tuning of azobenzenephotoswitches for biological applications. Angew Chem Int Ed Engl, 48(8):1484–1486.

5. Dong SL, et al. (2006) A photocontrolled beta-hairpin peptide. Chemistry, 12(4):1114–1120.

6. James DA, Burns DC, Woolley GA (2001) Kinetic characterization of ribonucleaseS mutants containing photoisomerizable phenylazophenylalanine residues. ProteinEng, 14(12):983–991.

7. Liu D, Karanicolas J, Yu C, Zhang Z, Woolley GA (1997) Site-specific incorporation ofphotoisomerizable azobenzene groups into ribonuclease S. Bioorg Med Chem Lett,7(20):2677–2680.

8. Ueda T, Murayama K, Yamamoto T, Kimura S, Imanishi Y (1994) Photo-regulation ofhydrolysis activity of semisynthetic mutant phospholipases A2 replaced by non-naturalaromatic amino acids. J Chem Soc Perk T, 1(2):225–230.

9. Muranaka N, Hohsaka T, Sisido M (2002) Photoswitching of peroxidase activity byposition-specific incorporation of a photoisomerizable non-natural amino acid intohorseradish peroxidase. FEBS Lett, 510(1-2):10–12.

10. Nakayama K, Endo M, Majima T (2004) Photochemical regulation of the activity of anendonuclease BamHI using an azobenzene moiety incorporated site-selectively intothe dimer interface. Chem Commun (Cambridge, UK)(21):2386–2387.

11. Bose M, Groff D, Xie J, Brustad E, Schultz PG (2006) The incorporation of aphotoisomerizable amino acid into proteins in E. coli. J Am Chem Soc, 128(2):388–389.

12. Mayer G, Heckel A (2006) Biologically active molecules with a “light switch”. AngewChem Int Ed Engl, 45(30):4900–4921.

13. Woolley GA (2005) Photocontrolling peptide alpha helices. Acc Chem Res, 38(6):486–493.

14. Pearson D, Alexander N, Abell AD (2008) Improved Photocontrol of alpha-Chymotryp-sin Activity: Peptidomimetic Trifluoromethylketone Photoswitch Enzyme Inhibitors.Chemistry, 14(24):7358–7365.

15. Westmark P, Kelly J, Smith B (1993) Photoregulation of enzyme activity. Photochromic,transition-state-analog inhibitors of cysteine and serine proteases. J Am Chem Soc,115(9):3416–3419.

16. Kumita JR, Smart OS, Woolley GA (2000) Photo-control of helix content in a shortpeptide. Proc Natl Acad Sci USA, 97(8):3803–3808.

17. Umeki N, et al. (2004) Incorporation of an azobenzene derivative into the energytransducing site of skeletal muscle myosin results in photo-induced conformationalchanges. J Biochem, 136(6):839–846.

18. Flint DG, Kumita JR, Smart OS, Woolley GA (2002) Using an azobenzene cross-linker to either increase or decrease peptide helix content upon trans-to-cisphotoisomerization. Chem Biol, 9(3):391–397.

19. Guerrero L, et al. (2005) Photochemical regulation of DNA-binding specificity ofMyoD. Angew Chem Int Ed Engl, 44(47):7778–7782.

20. Kumita JR, Flint DG, Woolley GA, Smart OS (2002) Achieving photo-control of proteinconformation and activity: Producing a photo-controlled leucine zipper. Faraday Dis-cuss, 122:89–103 discussion 171–190.

21. Kusebauch U, et al. (2006) Photocontrolled folding and unfolding of a collagen triplehelix. Angew Chem Int Ed Engl, 45(42):7015–7018.

22. Kusebauch U, Cadamuro SA, Musiol HJ, Moroder L, Renner C (2007) Photocontrol ofthe collagen triple helix: Synthesis and conformational characterization of bis-cysteinylcollagenous peptides with an azobenzene clamp. Chemistry, 13(10):2966–2973.

23. Woolley GA, et al. (2006) Reversible photocontrol of DNA binding by a designedGCN4-bZIP protein. Biochemistry, 45(19):6075–6084.

24. Zhang Z, Burns DC, Kumita JR, Smart OS, Woolley GA (2003) A water-solubleazobenzene cross-linker for photocontrol of peptide conformation. BioconjugateChem, 14(4):824–829.

25. Kumita JR, Flint DG, Smart OS, Woolley GA (2002) Photo-control of peptide helix con-tent by an azobenzene cross-linker: Steric interactions with underlying residues arenot critical. Protein Eng, 15(7):561–569.

26. Zhang F, et al. (2009) Structure-based approach to the photocontrol of protein folding.J Am Chem Soc, 131(6):2283–2289.

27. Jog PV, Gin MS (2008) A light-gated synthetic ion channel.Org Lett, 10(17):3693–3696.28. Loudwig S, Bayley H (2006) Photoisomerization of an individual azobenzene molecule

in water: An on-off switch triggered by light at a fixed wavelength. J Am Chem Soc,128(38):12404–12405.

29. Muramatsu S, Kinbara K, Taguchi H, Ishii N, Aida T (2006) Semibiological molecularmachine with an implemented “AND” logic gate for regulation of protein folding.J Am Chem Soc, 128(11):3764–3769.

30. Shishido H, Yamada MD, Kondo K, Maruta S (2009) Photocontrol of calmodulininteraction with target peptides using azobenzene derivative. J Biochem, 146(4):581–590.

31. Banghart M, Borges K, Isacoff E, Trauner D, Kramer RH (2004) Light-activated ionchannels for remote control of neuronal firing. Nat Neurosci, 7(12):1381–1386.

32. Gorostiza P, et al. (2007) Mechanisms of photoswitch conjugation and light activationof an ionotropic glutamate receptor. Proc Natl Acad Sci USA, 104(26):10865–10870.

33. Szobota S, et al. (2007) Remote control of neuronal activity with a light-gatedglutamate receptor. Neuron, 54(4):535–545.

34. Volgraf M, et al. (2006) Allosteric control of an ionotropic glutamate receptor with anoptical switch. Nat Chem Biol, 2(1):47–52.

35. Numano R, et al. (2009) Nanosculpting reversed wavelength sensitivity into aphotoswitchable iGluR. Proc Natl Acad Sci USA, 106(16):6814–6819.

36. Yamada MD, Nakajima Y, Maeda H, Maruta S (2007) Photocontrol of kinesin ATPaseactivity using an azobenzene derivative. J Biochem, 142(6):691–698.

37. Willner I, Rubin S, Riklin A (1991) Photoregulation of papain activity throughanchoring photochromic azo groups to the enzyme backbone. J Am Chem Soc,113(9):3321–3325.

38. Pingoud A, Silva GH (2007) Precision genome surgery. Nat Biotechnol, 25(7):743–744.39. Simoncsits A, Tjornhammar ML, Rasko T, Kiss A, Pongor S (2001) Covalent joining of

the subunits of a homodimeric type II restriction endonuclease: Single-chain PvuIIendonuclease. J Mol Biol, 309(1):89–97.

40. Meramveliotaki C, et al. (2007) Purification, crystallization, x-ray diffraction analysisand phasing of an engineered single-chain PvuII restriction endonuclease. Acta Crys-tallogr Sect F Struct Biol Cryst Commun, 63(10):836–838.

41. Woolley GA, Lee ES, Zhang F (2006) sGAL: A computational method for finding surfaceexposed sites in proteins suitable for Cys-mediated cross-linking. Bioinformatics,22(24):3101–3102.

42. Nastri HG, Evans PD, Walker IH, Riggs PD (1997) Catalytic and DNA binding propertiesof PvuII restriction endonuclease mutants. J Biol Chem, 272(41):25761–25767.

43. Spyridaki A, et al. (2003) Structural and biochemical characterization of a newMgð2þÞ binding site near Tyr94 in the restriction endonuclease PvuII. J Mol Biol,331(2):395–406.

44. Eisenschmidt K, et al. (2005) Developing a programmed restriction endonuclease forhighly specific DNA cleavage. Nucleic Acids Res, 33(22):7039–7047.

45. Borisenko V, Woolley GA (2005) Reversibility of conformational switching in light-sensitive peptides. J Photochem Photobiol A, 173(1):21–28.

1366 ∣ www.pnas.org/cgi/doi/10.1073/pnas.0909444107 Schierling et al.

Dow

nloa

ded

by g

uest

on

Apr

il 21

, 202

0