Báo cáo khoa học: Structural origins for selectivity and specificity in an engineered bacterial repressor–inducer pair pdf

The overall structures of ligand-free TetRK64 L135I138 and of TetRK64L135I138 in complex with 4-ddma-atc are very similar for the chemical structure of 4-ddma-atc and related TetR ligand

engineered bacterial repressor–inducer pair

Michael A Klieber1, Oliver Scholz2,*, Susanne Lochner3, Peter Gmeiner3, Wolfgang Hillen2

and Yves A Muller1

1 Lehrstuhl fu¨r Biotechnik, Department of Biology, Friedrich-Alexander University, Erlangen-Nuremberg, Germany

2 Lehrstuhl fu¨r Mikrobiologie, Department of Biology, Friedrich-Alexander University, Erlangen-Nuremberg, Germany

3 Lehrstuhl fu¨r Pharmazeutische Chemie, Department of Chemistry and Pharmacy, Friedrich-Alexander University, Erlangen-Nuremberg, Germany

Introduction

The bacterial repression system consisting of the

effec-tor molecule tetracycline, the tetracycline-inducible

repressor protein tetracycline repressor (TetR) and the

tet operator (tetO) has proven itself to comprise a

valuable tool for studying gene expression not only in prokaryotes, but also in eukaryotes [1–3] The repres-sor protein TetR, and first and foremost its ability to adopt different conformational states upon effector

Keywords

altered inducer selectivity; altered inducer

specificity; bacterial transcription regulation;

crystal structures; tetracycline repressor

Correspondence

Y A Muller, Lehrstuhl fu¨r Biotechnik,

Department of Biology, Friedrich-Alexander

University Erlangen-Nuremberg, Im IZMP,

Henkestrasse 91, D-91052 Erlangen,

Germany

Fax: +49 0 9131 8523080

Tel: +49 0 9131 8523081

E-mail: ymuller@biologie.uni-erlangen.de

*Present address

Department of Biochemistry, University of

Zurich, Switzerland

Database

Structural data are available from the Protein

Data Bank under the accession numbers

3FK6 for TetR(K 64 L 135 I 138 ) alone and 3FK7

for the 4-ddma-atc complex

(Received 10 March 2009, revised 9 July

2009, accepted 31 July 2009)

doi:10.1111/j.1742-4658.2009.07254.x

The bacterial tetracycline transcription regulation system mediated by the tetracycline repressor (TetR) is widely used to study gene expression in prokaryotes and eukaryotes To study multiple genes in parallel, a triple mutant TetR(K64L135I138) has been engineered that is selectively induced

by the synthetic tetracycline derivative 4-de-dimethylamino-anhydrotetracy-cline (4-ddma-atc) and no longer by tetracy4-de-dimethylamino-anhydrotetracy-cline, the inducer of wild-type TetR In the present study, we report the crystal structure of TetR(K64L135I138) in the absence and in complex with 4-ddma-atc at reso-lutions of 2.1 A˚ Analysis of the structures in light of the available binding data and previously reported TetR complexes allows for a dissection of the origins of selectivity and specificity In all crystal structures solved to date, the ligand-binding position, as well as the positioning of the residues lining the binding site, is extremely well conserved, irrespective of the chemical nature of the ligand Selective recognition of 4-ddma-atc is achieved through fine-tuned hydrogen-bonding constraints introduced by the His64fi Lys substitution, as well as a combination of hydrophobic effect and the removal of unfavorable electrostatic interactions through the intro-duction of Leu135 and Ile138

Abbreviations

atc, anhydrotetracycline; 4-ddma-atc, 4-de-dimethylamino-anhydrotetracycline; dox, 6-deoxy-5-hydoxy-tetracycline; PDB, Protein Data Bank;

tc, tetracycline; tetO, tet operator; TetR, tetracycline repressor; TetR(K 64 L 135 I 138 ), TetR-BD triple mutant H64K, S135L and S138I.

binding, plays a key role in this system In the absence

of tetracycline, TetR binds with high affinity to the tet

operator tetO and thereby blocks the transcription of

any downstream genes Upon binding of the inducer

tetracycline, the repressor TetR switches conformations

and dissociates from the operator DNA As a result of

numerous functional and structural studies, the atomic

mechanism that underlies the functional switch in

TetR is now understood in significant detail [4–7]

To be able to control the expression of several genes

in parallel, TetR mutants have been isolated in

elabo-rate screens that respond to novel synthetic tetracycline

analogs [8,9] One of these variants is the TetR triple

mutant TetR(K64L135I138) in which residues His64,

Ser135 and Ser138 of TetR have been replaced by Lys,

Leu and Ile, respectively [9] This mutant is selectively

induced by the synthetic inducer

4-de-dimethylamino-anhydrotetracycline (4-ddma-atc) and slightly by atc,

but no longer by tetracycline To better understand the

switch in selectivity and the acquired novel specificity

of TetR(K64L135I138) for 4-ddma-atc, we solved the

crystal structure of TetR(K64L135I138) in the presence

and absence of 4-ddma-atc at resolutions of 2.06 and

2.1 A˚, respectively We show that the effects of the

three mutations on ligand binding can be generally

thermodynamically dissected into individual

contribu-tions and that they originate from a favorable

inter-play of different physico-chemical properties, such as

solvation effects, constrained hydrogen-bonding

geom-etries and electrostatic discrimination

Results

Comparison of the 4-ddma-atc-bound and free

overall TetR(K64L135I138) structure

As for all TetRs, TetR(K64L135I138) forms a dimer

and, with respect to the monomer architecture, each

chain contains an N-terminal DNA-binding domain

(residues 1–48) and a C-terminal effector-binding

domain (residues 49–205) [10] (Fig 1A) The latter

also comprises the dimer interface The two

effector-binding sites present in the dimer are identical and,

because the binding sites are located within the protein

interface, each binding site is lined by residues from

both monomers

The overall structures of ligand-free TetR(K64

L135I138) and of TetR(K64L135I138) in complex with

4-ddma-atc are very similar (for the chemical structure

of 4-ddma-atc and related TetR ligands, see Fig 2)

The two crystals that have been used for structure

determination are highly isomorphous and only small

deviations occur in the cell axes (Table 1) They each

contain a complete dimer in the asymmetric unit Almost no differences exist between the monomers in each crystal and the monomers⁄ dimers between crys-tals The main chain atoms of the two monomers in each crystal can be superimposed with rmsd of 0.92 and 1.34 A˚ for the 4-ddma-atc-bound and effector-free TetR(K64L135I138) structures, respectively When direc-tly comparing the structure of 4-ddma-atc-bound TetR (K64L135I138) with that of effector-free TetR(K64L135

I138), it is obvious that ligand binding does not induce any considerable changes in the tertiary structures The four possible pairwise cross-superpositions between crystals yield an rmsd in the range 0.52–1.23 A˚ for a total of 770 common main chain atoms This shows that the differences between the ligand-bound and ligand-free structures are not larger than the differ-ences observed between the two monomers present in each crystal To some extent, this is not too surprising because the crystal with 4-ddma-atc bound was obtained by soaking a ligand-free crystal with the ligand However, when considering the molecular mechanism by which TetR exerts its function, then the structural similarity might be considered unexpected

A central function of TetR is its ability to adopt different conformations According to the tional switch model, TetR exists in two conforma-tions In the ligand-free structure, the DNA-binding heads are oriented such that TetR can readily bind

to the operator DNA, whereas, in effector-bound TetR, the separation of the DNA-binding domain is changed, such that TetR can no longer recognize the tetO DNA sequence (Fig 1A) [4] However, as noted above, the domain orientations in the 4-ddma-atc-bound TetR(K64L135I138) and the ligand-free TetR (K64L135I138) structure are very similar and resemble that of induced TetR more closely than that of indu-cer-free TetR (data not shown) Moreover, the main chain of loop segment 100–105 that switches con-formations upon effector binding is in an identical conformation in both structures and resembles that observed for induced TetR (Fig 1C) In this confor-mation, segment 100–105 moves towards the ligand and thereby enables residues from the segment to participate in ligand binding

The observation that effector-free TetR(K64L135I138) adopts an induced-like conformation must appear unexpected However, it can easily be rationalized if alternative models are considered (e.g the population shift model) as explanations for the allosteric behavior

of TetR [11,12] According to this model, effector-free and DNA-free TetR is able to adopt a variety of divers and freely inter-converting conformations Among these, there are also conformations that can

readily interact with DNA or that resemble the

effec-tor-bound conformation as, for example, observed in

the present study in the case of the crystal structure of

the effector-free TetR(K64L135I138) In this model,

induction can be explained by a shift of the population

towards a single DNA-binding incompetent

conforma-tion Indications that such a population shift model

might apply to TetR have recently started to emerge

[13,14]

The effector-binding site of TetR(K64L135I138) in the presence and absence of 4-ddma-atc Particularly interesting with respect to the observed specificity and selectivity of TetR(K64L135I138) for 4-ddma-atc are the interactions between the ligand and the protein in the effector-binding site (Figs 1B and 2A) Of the two binding sites that can be observed independently in the 4-ddma-atc-complex structure,

A

C

B

Fig 1 Crystal structure of TetR(K 64 L 135 I 138 ) in complex with 4-ddma-atc (A) Ca-representation of the TetR(K 64 L 135 I 138 ) dimer (shown in dif-ferent shades of magenta) in complex with 4-ddma-atc (shown in yellow) superimposed on wild-type TetR in complex with DNA (in black, PDB entry: 1QPI) [6] According to the generally agreed induction mechanism, effector binding to TetR induces a conformational change in TetR, which alters the orientation and separation of the DNA-binding domains (indicated by two-headed arrows), so that TetR no longer binds to the operator DNA (B) Schematic representation of the interactions between 4-ddma-atc and TetR(K64L135I138) (C) Stereo represen-tation of the structure of TetR(K 64 L 135 I 138 ) in complex with 4-ddma-atc (shown in magenta and yellow) superimposed onto the effector-free TetR(K 64 L 135 I 138 ) structure (shown in grey) Of the two binding sites present in the crystal structure, binding site I (Table 2) is shown The binding sites are highly similar in the presence and absence of 4-ddma-atc The loop segment 100–105 that switches conformations in other ligand-free and ligand-bound TetR structures is only slightly displaced in ligand-free TetR(K 64 L 135 I 138 ) compared to the ligand-bound TetR structure.

both binding sites show the density for the ligand in

the initial difference-fourier electron density maps,

albeit to a different extent The ligand is well defined

in binding site I, but only poor density has been

observed for the ligand in binding site II To account

for this observation, the occupancy of the ligand in

binding site II was estimated at 50% and that in

bind-ing site I at 100% When usbind-ing these values durbind-ing the

refinement, the temperature factors of the ligands

refine to values similar to those of the surrounding

res-idues, hinting that the estimated occupancies correctly

reflect those in the crystal Inspection of the crystal packing yields a plausible explanation for the differ-ences in occupancies In the crystal, the accessibility to site II is impaired by the packing of neighboring mole-cules, whereas site I appears to be readily accessible through the solvent channels in the crystal

When superimposing the two ligand-binding sites on the basis of the coordinates of 93 residues surrounding the ligands, it is obvious that the ligand is slightly shifted in site II (Table 2) This shift is not only appar-ent with respect to 4-ddma-atc bound to site I, but also compared to various other ligand-TetR complexes (Table 2) The two 4-ddma-atc molecules differ by an rmsd of approximately 1 A˚, whereas the deviations between 4-ddma-atc bound to site I and tetracycline and 6-deoxy-5-hydoxy-tetracycline (dox) bound to TetR [15,16], as well as atc bound to revTetR [14], are

in the range 0.4–0.5 A˚ when considering 27 common ligand atoms We suspect that the positional shift of the ligand in site II is related to the 50% occupancy The concomitant occurrence of ligand and water mole-cules (also refined at 50% occupancy) at almost identi-cal positions might lead to increased coordinate errors during the refinement of the atomic positions and hence a less accurate ligand positioning in site II Accordingly, the description of the binding of the ligand in the present study is restricted to 4-ddma-atc binding to site I in TetR(K64L135I138)

4-ddma-atc binding leads to only minor rearrange-ments in the TetR(K64L135I138)-binding site (Fig 1C) Among the most notable changes are a slight shift of the entire loop segment 100–105 in the direction of the ligand, the presence of two alternative side chain confor-mations for Asn82 in the 4-ddma-atc-bound structure versus a single conformation in ligand-free TetR(K64L135I138) and, finally, the occurrence of a slightly different rotamer for Ile138 (i.e different posi-tioning of atom Cd) in the ligand-free and ligand-bound structure Overall, when considering both the main chain fold and the conformations of the side chains, the structures of ligand-free TetR(K64L135I138) and 4-ddma-atc-bound TetR(K64L135I138) are very similar This simi-larity also extends to a number of water molecules that occupy identical positions in the two structures

Specific interactions between 4-ddma-atc and TetR(K64L135I138)

TetR(K64L135I138) binds 4-ddma-atc via a number of specific interactions (Fig 3A) One of the most notable ones involves Lys64 Atom Nf of Lys64 interacts with two oxygen atoms of 4-ddma-atc, namely of the amide group attached to atom C2 and the OH group

A

B

C

D

Fig 2 Chemical structures of (A) 4-ddma-atc, (B) atc, (C)

tetracy-cline (tc) and (D) dox.

attached to atom C3 of 4-ddma-atc Furthermore,

atom Lys64-Nf is located in hydrogen-bonding

dis-tance to the amide group of Asn82 and the main chain

carbonyl oxygen atom of Tyr66 It is not possible to predict the strength of the interaction between Lys64 and 4-ddma-atc based on structural data only For

Table 1 Data collection and refinement statistics.

Triple mutant TetR(K 64 L 135 I 138 )

Triple mutant TetR(K64L135I138) in complex with 4-ddma-atc

Data collection statistics

Unit cell parameters

Refinement statistics

Number of protein atoms, solvent

molecules and ligand atoms

rmsd B-factors bonded atoms: main chain,

side chains (A˚2)

Percentage of residues in most favored regions,

additional allowed, generously allowed and

disallowed regions of the Ramachandran plotb

95.1, 4.9, 0.0, 0.0 95.5, 4.2, 0.3, 0.0

a Values in parentheses refer to the highest resolution shell b According to PROCHECK [27].

Table 2 Superposition of the ligand-binding sites and ligand positions in selected TetR complexes.

No ligand TetR (K64L135I138) Ia

No ligand TetR (K64L135I138) II

4-ddma-atc TetR (K64L135I138) Ia

4-ddma-atc TetR (K64L135I138) II

atc revTetR (PDB code:

2VKV)

tc TetR(D) (PDB code:

2VKE)

dox TetR(D) (PDB code: 2O7O)

No ligand

TetR(K64L135I138) I

– 0.905, 1.396 b 0.532, 0.843 0.830, 1.360 0.854, 1.386 0.851, 1.385 0.870, 1.356

No ligand

TetR(K 64 L 135 I 138 ) II

4-ddma-atc

TetR(K 64 L 135 I 138 ) I

4-ddma-atc

TetR(K 64 L 135 I 138 ) II

a Because the crystals contain two molecules in the asymmetric unit, two separate binding sites (I and II) are present in each of the two TetR(K64L135I138) crystal structures.bAbove the diagonal the rmsd (A ˚ ) between structures of a selection of 93 residues surrounding the ligand-binding site is reported (first number, rmsd obtained upon superposition of all main chain atoms of the selection; second number, superposition of all atoms) c Below the diagonal: rmsd (A ˚ ) calculated between the ligands (27 common ligand atoms) in the different com-plexes after optimal superposition of the structures based on the main chain atoms of 93 residues surrounding the binding site.

example, the structure does not allow a distinction of

whether the side chain of Lys64 is protonated or not

Uncertainties also arise with respect to the correct

orientation of the amide group attached to atom C2 of

4-ddma-atc, as well as the amide group of amino acid

Asn82, because the amide oxygen and nitrogen atoms

are indistinguishable at the resolutions of the solved

crystal structures Because atom Nf of Lys64 can only

participate in three hydrogen bonds, it should be noted

that, of the four potential hydrogen bond acceptors⁄

donors, atom O from the amide group of 4-ddma-atc

is the furthest apart (3.1 A˚ versus 2.5–2.8 A˚ for the

other hydrogen bond acceptors⁄ donors) However, all

four potential hydrogen bond partners are

geometri-cally quite favorably placed In all four cases, almost ideal linear hydrogen bond geometries can be antici-pated because the angle formed between atoms Lys64-Ce, Lys64-Nf and the potential acceptor⁄ donor atoms are in the range 108–120, in line with the tetrahedral positioning of the hydrogen atoms attached

to Lys64-Nf

TetR(K64L135I138) forms an extended hydrophobic contact with the ligand at the ‘backside’ of 4-ddma-atc 4-ddma-atc directly contacts the mutated resi-dues 135 (Ser135Leu) and 138 (Ser138Ile) of TetR(K64L135I138) Because of the hydrophobic nature

of the Leu and Ile side chains and because of their increased size compared to the serine residues in

A

C

B

Fig 3 (A) Close-up view on the ligand-binding site of 4-ddma-atc in complex with TetR(K 64 L 135 I 138 ) and (B) tetracycline in complex with TetR [15] (PDB entry: 2VKE) The residues that differ between the two structures are underlined The hydrogen-bonding network in which residue

64, namely Lys64 in TetR(K 64 L 135 I 138 ) (A) and His64 in TetR (B), participates, is indicated by dashed lines Water molecules present in the two structures between residue 135 and the ‘backside’ of the ligand are depicted; all other water molecules are omitted (C) Stereo repre-sentation of the superimposed structures of 4-ddma-atc bound to TetR(K64L135I138) (shown in yellow and magenta), atc bound to revTetR [14] (PDB entry: 2VKV; shown in cyan) and tetracycline bound to TetR [15] (PDB entry: 2VKE; shown in grey and blue) The stereo represen-tation shows that the binding position of the ligands and the orienrepresen-tations of the side chains lining the binding sites are highly conserved in the different complexes In wild-type TetR, this also extends to water molecules surrounding residue 135.

wild-type TetR, a number of water molecules that

bridge between the serines and the effector in other

effector complexes are absent in the 4-ddma-atc

TetR(K64L135I138) complex (Fig 3)

Although the TetR(K64L135I138) crystals were soaked

with 4-ddma-atc without the addition of any extra

magnesium, a partially occupied magnesium ion can be

observed at a position identical to that observed in other

TetR effector complexes With the exception of the

Lys64 interaction and the extended hydrophobic

inter-face introduced by Leu135 and Ile138, all other ligand

protein interactions are highly similar to those observed

in other TetR-ligand complexes (see also below)

Discussion

4-ddma-atc binding to TetR(K64L135I138) compared

to tetracycline, atc and dox binding to TetR

Various crystal structures of TetR in complex with

tet-racycline, dox and atc have already been solved to

high resolution Comparing these structures among

themselves and to TetR(K64L135I138) promises to

pro-vide insight into why TetR(K64L135I138) specifically

recognizes 4-ddma-atc and why wild-type TetR is

selectively induced by tetracycline, dox and atc and

not by 4-ddma-atc [9,17] Upon superposition of these

structures, it is immediately obvious that, in all these

structures, the effector-binding sites are highly similar

both with respect to the positioning of the ligands and

the conformations of the residues lining the binding

site (Table 2) In an initial comparison of the ligand

positions of tetracycline, dox and atc in TetR, these

ligands superimpose with an average rmsd of 0.26 A˚

for 27 common ligand atoms In the case of the

4-ddma-atc complex, the ligand appears to be slightly

displaced with respect to the other ligands (4-ddma-atc

bound to site I, average rmsd of 0.47 A˚ compared to

the other ligands; Table 2) However the difference is

small and only slightly exceeds the estimates for the

coordinate errors in the different crystal structures

A major difference between TetR(K64L135I138) in

complex with 4-ddma-atc and all other effector TetR

complexes is seen for the interaction between residue

64 and the various effectors As noted above, in

TetR(K64L135I138) Lys64 is involved in a number of

specific interactions with 4-ddma-atc and it appears

that, in the wild-type TetR complexes, histidine is able

to participate in similar interactions because the

posi-tion of atom Ne of His64 coincides almost exactly with

that of atom Nf of Lys64 (atom displacement of

0.9 A˚; Fig 3C) Compared to Lys64, however, a

histi-dine residue can participate in fewer hydrogen bonds

In wild-type TetR, it appears that a putative hydrogen atom attached to Ne of His64 is poised to interact with the oxygen atom attached to atom C3 present in all tetracycline derivatives This oxygen is positioned in plane with the imidazole ring, and a linear almost ideal hydrogen bond can be anticipated for this interaction

In comparison, an additional interaction often dis-cussed as being important for ligand binding [16], namely the interaction between Ne of His64 and the amide group attached to atom C2 of tetracycline, appears less favorable because the amide group is con-siderably displaced out of the plane of the imidazole ring

The presence of a histidine at position 64 compared

to a lysine appears to affect a neighboring asparagine residue As noted above, Asn82 is hydrodrogen-bonded to Lys64 in TetR(K64L135I138) In all other TetR structures with a histidine at position 64, no such interaction exists because the amide group of Asn82 participates in a bidental interaction with all tetracycline derivatives possessing a 4-amino-group, namely with the nitrogen of the dimethyl-amino-group and the oxygen attached to atom C3 (Figs 2 and 3)

A further notable difference between 4-ddma-atc-bound TetR(K64L135I138) and other complexes com-prises the number of water molecules in the interface between the ligand and the protein By contrast to TetR(K64L135I138), where Ser135 and Ser138 are replaced by Leu and Ile, water molecules are attached

to the serines in all other TetR structures and fill a cleft between the ligand and the protein Close inspec-tion of these water molecules shows that their posi-tions are largely conserved in the complexes formed between TetR and tetracycline, dox or atc (Fig 3C)

The specificity and selectivity of the 4-ddma-atc TetR(K64L135I138) interaction

The structural investigations reported in the present study aimed to gain insight into the mechanism

by which effector selectivity is switched in TetR(K64L135I138) compared to wild-type TetR Previ-ously published data on induction efficiencies and inducer affinities of TetR(K64L135I138), as well as for all corresponding single and double mutant variants, are highly valuable for discussing the origins of speci-ficity and selectivity [8,9] (Table 3) When analyzing the changes in the free binding energies of the mutants for the different ligands, it is apparent that these changes are often additive For example, the sum of the changes in the binding affinities (DDG) observed in the three single site mutants His64Lys, Ser135Leu and

Ser138Leu for the ligand dox corresponds almost

exactly to the change observed in the triple mutant

TetR(K64L135I138) (6.09 versus 5.86 kcalÆmol)1)

com-pared to wild-type TetR (Table 3) In the case of the

ligands atc and 4-ddma-atc, only near additivity is

achieved in the mutants (7.93 versus 6.61 kcalÆmol)1

for atc and)5.71 versus )3.60 kcalÆmol)1 for

4-ddma-atc binding)

In many cases, it is also possible to formulate almost

perfect thermodynamic cycles For example, the free

binding energy difference observed for the binding of

the ligands atc and 4-ddma-atc to wild-type TetR

(DDG = 7.63 kcalÆmol)1) corresponds exactly to the

sum of the changes observed for atc binding to the

mutant His64Lys (4.7 kcalÆmol)1), the difference in

binding energies for the ligands atc and 4-ddma-atc to

the same mutant (1.28 kcalÆmol)1) and the

differ-ence in binding energies observed for 4-ddma-atc

bind-ing to wild-type TetR and to the His64Lys mutant

(1.65 kcalÆmol)1)

Juxtaposition of the binding affinities to the

induc-tion efficiencies suggests that free binding energies in

excess of approximately )12 kcalÆmol)1 are required

for efficient induction (Table 3) If this is indeed the

case, then the switch in induction specificity in

TetR(K64L135I138) can be explained as the result of an

increase in the free binding energies (DG) above

)12 kcalÆmol)1for the ligands dox or atc and the con-comitant lowering of the free binding energy to )12.44 kcalÆmol)1 for 4-ddma-atc This appears to hold true for all the variants, with the exception of the ligand dox in combination with the mutant Ser138Ile Only very little induction is observed for this mutant with the ligand dox [9] (Table 3), although the free binding energy is lower than )12 kcalÆmol)1 It should

be noted, however, that the binding affinity data from which the free energies have been calculated are not free of errors The standard deviations have been estimated to be in the range 10–40% of the reported values [9] When translated to DG, this corresponds to approximately 0.25 kcalÆmol)1(Table 3)

The structures that we have determined in the pres-ent study and the comparison of these structures with previously solved crystal structures are in agreement with the proposed additivity or near additivity of the free binding energies In all the structures, the effector molecule binds at almost exactly the same position, and the introduction of mutations and⁄ or changes in the ligand does not lead to any notable changes in the side chain or backbone conformations of the residues lining the binding site Although the structures of each single and double mutant have not been solved, it is reasonable to assume that structure conservation also extends to these mutants As a result of the structural

Table 3 Induction efficiencies and free binding energies of TetR and mutants for various tetracycline analogs Data are compiled from Henssler et al [9].

Induction efficiencies

Free binding energies (kcalÆmol)1) b,c

a Induction efficiencies: Less than 20% induction efficiency in a b-galactosidase reporter assay +, ++, +++, ++++: between 20–40%, 40–60%, 60–80%, and 80–100% induction efficiency, respectively.bFree binding energies derived from the experimentally determined binding affinities reported in Henssler et al.[9] and calculated according to DG = )RT lnK (t = 298.15 K) In parentheses: DDG = DG(mutant) ) DG(wild-type) c The standard deviations of the affinities reported in Henssler et al [9] have been estimated to be in the range 10–40% Assuming a standard error propagation model with dDG = )RT (dK ⁄ K), this translates into 0.05–0.25 kcalÆmol )1as an error estimate for DG.

conservation and the near additivity of the DG values,

it is possible to discuss the observed selectivity in light

of individual changes introduced in the binding site by

the various mutations

With respect to single changes, the most drastic

dif-ferences in the free binding energies are observed for

the His64Lys mutation and for the removal of the

dimethyl-amino-group attached to atom C4 of the

tetracycline derivatives in wild-type TetR or in mutants

in which His64 is retained Replacing the histidine by

a lysine causes all tetracycline derivatives to be

recog-nized with almost identical binding affinities (Table 3)

This is the result of a drastic decrease in the affinities

for dox and atc, and a significant increase in affinity for

4-ddma-atc Because the free binding energies exceed

)12 kcalÆmol)1, the His64Lys mutation is not induced

by any of the four ligands anymore Accordingly, it is

apparent that histidine is particular well suited to

recog-nize the dimethyl-amino-group attached to atom C4

Inspection of the crystal structures shows that

recogni-tion occurs indirectly and that Asn82 most likely makes

a key contribution to this discrimination In all

com-plexes with ligands containing a dimethyl-amino-group

attached to C4, the amide group of Asn82

partici-pates in a bidental interaction with the tcs, namely

with the dimethyl-amino-group and the oxygen atom

attached at position C3 (Fig 3) The amide group of

Asn82 does not directly interact with His64 but only

indirectly because both His64 and Asn82 interact with

the oxygen attached at C3 By contrast, when His64

is replaced by a lysine residue, a direct interaction

occurs between Nf of Lys64 and the carbonyl group

of Asn82 Because all ligands are now recognized

with similar affinities, it is apparent that the

Lys64-Asn82 interaction disrupts any favorable interaction

between the dimethyl-amino-group and Asn82 This

disruption could, for example, be caused by a flip in

the orientation of the amide group, which leads to an

exchange of the positions of the nitrogen and oxygen

atoms As a result, a less favorable interaction could

occur between the amide-NH2 group and the

dimethyl-amino-group of dox and atc that is assumed

to be protonated in the TetR complex [16] The

importance of Asn82 for ligand discrimination is

fur-ther highlighted by the fact that randomization of

position 82 does not allow the identification of any

additional residue with 4-ddma-atc specificity [9]

Upon mutation of residue Ser135 to leucine, the

affinity of TetR for almost all ligands increases

(Table 3) This holds true for all the variants into

which this mutation is introduced The only exception

appears to be the binding of dox to the single mutant

Ser135Leu for which a small decrease in affinity can

be observed compared to wild-type TetR (DDG = +0.08 kcalÆmol)1) Introducing the mutation Ser135Leu to any other variant also enhances the binding affinity of dox The amounts by which the affinities increase differ for the various mutants and the ligands The most significant increase is observed for 4-ddma-atc binding Inspection of the crystal struc-tures suggests that this increase in affinity is a direct consequence of increased hydrophilic interactions and the associated hydrophobic effect As noted above, res-idue 135 interacts with the largely hydrophobic D ring Whereas, in most crystal structures, a number of highly conserved water molecules bridges between the serine at position 135 and the tetracycline derivative, the water molecules are expelled from this interface in TetR(K64L135I138) in complex with 4-ddma-atc, in line with an acquired entropic advantage for this complex Because of the removal of the dimethyl-amino-group, 4-ddma-atc represents the most hydrophobic com-pound of all the derivatives discussed in the present study Consequently, we expect the hydrophobic effect

to be the largest for this tetracycline derivative and therefore the largest increase in affinity should be observed for 4-ddma-atc binding to any mutant in which Ser135 is mutated to leucine

The main role played by the Ser138Ile mutation appears to be that of a negative selection filter because,

in most variants, the introduction of this mutation causes a significant reduction in affinity for dox and atc and, at the same time, only slightly improves the affinity for 4-ddma-atc (Table 3) This behavior can be easily explained by considering that the dimethyl-amino group present in dox and atc, which interacts with Ser138 in wild-type TetR, is assumed to be posi-tively charged in the TetR-bound effector [16] Because

of the polar nature of its side chain, a serine is better suited to stabilize an adjacent positive charge than an isoleucine Vice versa, the increased hydrophobicity of the isoleucine residue in the Ser138Ile mutants matches the increased hydrophobicity of 4-ddma-atc compared

to atc, possibly explaining the slight improvement in affinity for 4-ddma-atc in variants containing this sub-stitution However, it has been observed that space requirements are equally important for achieving selec-tivity at this position because only a serine to isoleu-cine substitution leads to the observed shift in specificities and no other hydrophobic residue is toler-ated at this position [9] The structure hints that an isoleucine fits perfectly between the protein and the tetracycline A and B rings

The results obtained in the present study show that the observed specificity and selectivity in TetR(K64L135I138) for 4-ddma-atc can be explained

through a defined set of contributions Whereas the

His64Lys mutation abolishes the selectivity present in

wild-type TetR, and the Ser135Leu mutation improves

the binding of all three effectors, the Ser138Ile

muta-tion selectively disfavors effector molecules containing

a dimethyl-amino group and, at the same time, only

slightly improves 4-ddma-atc binding The

physico-chemical contributions of the individual residues

appear to be finely balanced and include geometrically

constrained hydrogen-bonding networks, electrostatic

interactions and solvation and dissolvation effects

TetR(K64L135I138) was identified using extensive

muta-tional screens and, in light of the structures presented

here, it is obvious that it would have been difficult to

construct such a highly specific repressor–inducer pair

employing a rational design and using structural

infor-mation only Conversely, however, the structures

pre-sented here, when taken together with the previously

available biochemical characterization, represent a

challenging benchmark data set for testing and

validat-ing computational models aimed at predictvalidat-ing and

designing the specificity and selectivity of protein–

ligand complexes

Experimental procedures

Protein expression, purification and

crystallization

For protein production, Escherichia coli strain RB791 was

transformed with the plasmid pWH610, which encodes for

the triple mutant TetR(K64L135I138) [9] In this construct,

the chimeric TetR variant TetR(BD) [8], which contains the

DNA-binding domain (residues 1–50) from TetR variant B

and the effector-binding domain (residues 51–208) from

TetR variant D, is further modified through the

introduc-tion of three single site mutaintroduc-tions, namely His64fi Lys,

Ser135fi Leu and Ser138 fi Ile Transformed E coli cells

were grown in LB medium at 28C and induced with

1 mm isopropyl thio-b-d-galactoside after an A600 of 0.8

was reached in the cell cultures After further incubation

for 4 h, cells were harvested by centrifugation, and the

pel-let dissolved in 30 mL of 20 mm sodium phosphate buffer

(pH 6.8) containing 5 mm EDTA and 1 mm Pefabloc

(Roche Diagnostics, Mannheim, Germany) before the cell

walls were disrupted by sonication After centrifugation

for 1 h at 100 000 g, the supernatant was purified

employ-ing a four-step chromatographic protocol The supernatant

was first applied onto a weak cation-exchange column

(SP-Sepharose FF; GE Healthcare Bio-Sciences, Uppsala,

Sweden) and subsequently onto two strong anion-exchange

columns (Resource Q and Mono Q; GE Healthcare

Bio-Sciences) Whereas, in the case of the cation-exchange

column, the buffer was identical to the buffer used during

the sonication step, the two anion-exchange columns were equilibrated with 50 mm NaCl, 20 mm Tris (pH 8.0) The proteins were eluted from all three columns using a stan-dard NaCl gradient (20 mm to 1 m) As a final chromato-graphic step, a gel filtration chromatography run was performed (Superdex 75; GE Healthcare Bio-Sciences) using a buffer consisting of 200 mm NaCl and 50 mm Tris (pH 8.0)

Initial crystallization conditions for TetR(K64L135I138) without 4-ddma-atc were identified using standard factorial screens (Hampton Research, Aliso Viejo, CA, USA) and a protein solution containing 10 mgÆmL)1 protein in 200 mm NaCl and 50 mm Tris (pH 8.0) buffer X-ray quality crys-tals were grown using the hanging drop method and mixing

1 lL of protein solution with 1 lL of reservoir solution (1 m K2HPO4, 200 mm NaCl, 50 mm Tris, pH 8.0) The droplets were equilibrated over 1 mL of reservoir solution until the crystals reached their final sizes of approximately

150· 70 · 50 lm after 10 days of incubation at 19 C For completeness, it should be noted that, in addition, tetracy-cline at a concentration of 1 mm was present in the crystal-lization droplets, although it was known from biochemical experiments that TetR(K64L135I138) is not induced by tetra-cycline [9] Careful inspection of the initial and final elec-tron density maps of the 4-ddma-atc-free TetR(K64L135I138) structure did not provide any hints for the density of tetra-cycline bound to the effector-binding site

Crystals of TetR(K64L135I138) with 4-ddma-atc were obtained upon soaking the previously grown ligand-free crystals with a saturated solution of the poorly soluble 4-ddma-atc compound The yellowish coloring of the crys-tals indicated the successful incorporation of the ligand

X-ray structure analysis and validation Diffraction data sets of TetR(K64L135I138) in complex with 4-ddma-atc and without ligand were collected at BESSY Synchrotron at the beamlines of the Protein Structure Fac-tory of Free University Berlin Before cryo-cooling, crystals were soaked for a few seconds in a cryo-protectant solution consisting of 20% ethylenglycol and 80% reservoir solu-tion All data sets were reduced using the software xds and scaled with xscale [18] The crystal parameters and the data collection statistics are reported in Table 1 The struc-tures were solved by molecular replacement using amore [19,20] As a search model for the TetR(K64L135I138) struc-ture without ligand, the Protein Data Bank (PDB) entry 2TCT was used [5,21] The solution indicated the presence

of two monomers in the asymmetric unit, and after rigid body refinement (8 to 3 A˚ resolution) the patterson correla-tion coefficient increased to 48.9% During refinement, the model was manually inspected with the software o [22] and coot [23] and automatically refined with refmac [24] and cns [25] until the refinement converged at an Rwork

of 21.6% and an Rfree of 26.2% at 2.1 A˚ (Table 1) The