Tetramer opening in LacI-mediated DNA loopingDanielis Rutkauskasa, Hongli Zhanb, Kathleen S. Matthewsb,c, Francesco S. Pavonea,d, and Francesco Vanzia,e,1

aEuropean Laboratory for Nonlinear Spectroscopy (LENS) and dDepartment of Physics, University of Florence, 50019 Sesto Fiorentino (FI), Italy; bDepartmentof Biochemistry and Cell Biology and cW. M. Keck Center for Interdisciplinary Bioscience Training, Rice University, 6100 Main Street, Houston, TX 77251 andeDepartment of Evolutionary Biology, University of Florence, Via Romana 17, 50125 Firenze, Italy

Edited by Donald M. Crothers, Yale University, New Haven, CT, and approved August 18, 2009 (received for review April 28, 2009)

Lactose repressor protein (LacI) controls transcription of the genesinvolved in lactose metabolism in bacteria. Essential to optimalLacI-mediated regulation is its ability to bind simultaneously to twooperators, forming a loop on the intervening DNA. Recently, severallines of evidence (both theoretical and experimental) have suggestedvarious possible loop structures associated with different DNA bind-ing topologies and LacI tetramer structural conformations (adoptedby flexing about the C-terminal tetramerization domain). We address,specifically, the role of protein opening in loop formation by employ-ing the single-molecule tethered particle motion method on LacIprotein mutants chemically cross-linked at different positions alongthe cleft between the two dimers. Measurements on the wild-typeand uncross-linked LacI mutants led to the observation of two distinctlevels of short tether length, associated with two different DNAlooping structures. Restricting conformational flexibility of the pro-tein by chemical cross-linking induces pronounced effects. Crosslink-ing the dimers at the level of the N-terminal DNA binding head (E36C)completely suppresses looping, whereas cross-linking near theC-terminal tetramerization domain (Q231C) results in changes oflooping geometry detected by the measured tether lengthdistributions. These observations lead to the conclusion thattetramer opening plays a definite role in at least a subset ofLacI/DNA loop conformations.

gene regulation � lactose repressor

Negative feedback control by lactose repressor (LacI) providesa classic example of bacterial gene expression regulation. In the

absence of lactose, LacI blocks the transcription of genes encodinglactose-metabolizing enzymes by binding to a primary operator(O1) on the DNA, located just downstream of the RNA polymerasepromoter. Transcription repression is strongly enhanced by thepresence of two auxiliary operators, situated both upstream (O3)and downstream (O2) of the primary operator (1). This effect is dueto the fact that LacI is a homotetrameric protein assembled as adimer of dimers, with each dimer presenting a DNA-bindingdomain, enabling the tetramer to bind simultaneously to twooperators (2) and form a loop in the intervening DNA (Fig. 1 A andB). Consequently, LacI binding to one of the auxiliary operatorsincreases the proximity of the remaining free DNA-binding domainto the primary operator, thus enhancing the probability of tran-scription blocking. Moreover, upon dissociation from the primaryoperator, the repressor is likely to remain bound to one of theauxiliary operators, avoiding diffusion into the cytosol and retain-ing proximity to the primary operator to facilitate rebinding. Thus,LacI-mediated DNA looping is fundamental for the efficiency ofrepression, as demonstrated by the drastic effects of deletion ofauxiliary operators (3).

In general, DNA looping is a recurring motif among several otherDNA-binding proteins with different functions (4). The loopingprobability is quantified by the Jm factor (5), defined as the localconcentration of the LacI bound to one of the operators withrespect to the remaining free operator. This factor is governed bythe DNA-protein mechanics, as manifest in the profound depen-dence of in vivo repression (6) or in vitro cyclization efficiency (7)on interoperator spacing. Modulation of this property with theperiod of helical repeat is due to a large DNA-twisting-associatedenergetic penalty and, therefore, certain phase requirements for

joining the loose DNA ends for cyclization, as well as for LacIbinding to two operators. On the length scale of hundreds of basepairs, the interplay between the energetic and entropic costs ofDNA bending impacts the periodic dependence of looping (8–10)or cyclization efficiency: at lengths of up to approximately 300 bp(i.e., twice the bending persistence length of DNA), the decrease ofbending energy with length determines an increase of Jm, whereasat longer lengths the dominant entropic cost of looping results in Jmdecrease. A complete treatment of looping requires identificationof the configurations adopted by the repressor-DNA assembly, asa number of different loop structures with differing formationprobabilities are feasible. The LacI crystal structure exhibits aV-shape with an angle of approximately 34° between the twodimeric subunits (2). On the other hand, electron microscopy (11),FRET (12), and small angle X-ray scattering (13) indicate that theangle of tetramer opening can vary and be as large as 180°.Additionally, four DNA loop topologies are conceivable with LacIin the V-shape, depending on the orientations of the boundoperators (9, 10): antiparallel (A1 and A2) and parallel (P1 and P2),as shown in Fig. 1B. Moreover, LacI can bind inside or outside theDNA loop in the so called wrap-toward or wrap-away configu-rations, respectively (14). With the extended LacI conformation,only P1 is quantitatively considered in the literature (P1E), as itminimizes free energy (at least for small–medium, 92–153 bp,interoperator spacings) over all other loop types with the opentetramer (9, 10).

In the absence of tools adequate for determining high-resolutionloop structures, inferences about the actual configurations weremade indirectly. For example, AFM does not have enough resolu-tion to decipher the details of repressor binding to DNA. However,it permits visualization of the loop shape and protein position withinthe loop (15). Similarly, LacI was deduced to mediate A1 loops ona negatively supercoiled plasmid with 197 bp interoperator distance(16). FRET measurements on bent A-tract DNA sequences dem-onstrated the possibility of tetramer opening in the loop formation(12, 17). A comprehensive delineation of various loop occurrenceswas inferred from simulations (9, 10). From these analyses, differenttopologies (Fig. 1B) can be adopted by the LacI-DNA complex,depending on experimental variables such as the degree of DNAsupercoiling, inter-operator spacing, DNA sequence, and the pres-ence or absence of DNA bending HU protein. For example,negative supercoiling favours antiparallel loops: the location ofpermanganate attack sensitive site in the wild-type 92-bp loopbetween O1 and O3 operators (18) is compatible with the A1structure. Linear DNA segments of 50–180 bp appear to loop withthe extended LacI conformation: DNaseI footprinting (loops of 52and 74 bp) and gel-mobility patterns (loops of 153, 158, 163, and 168bp) (19) are best interpreted assuming P1E configuration (9). Loopsin the size range of 65–90 bp adopt either P1 or P1E structure,depending on whether HU is, respectively, present or not (20). This

Author contributions: K.S.M., F.S.P., and F.V. designed research; D.R., H.Z., and F.V. per-formed research; D.R., H.Z., and F.V. analyzed data; and D.R., H.Z., K.S.M., F.S.P., and F.V.wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

1To whom correspondence should be addressed. E-mail: fvanzi@lens.unifi.it.

www.pnas.org�cgi�doi�10.1073�pnas.0904617106 PNAS � September 29, 2009 � vol. 106 � no. 39 � 16627–16632

BIO

PHYS

ICS

AN

DCO

MPU

TATI

ON

AL

BIO

LOG

Y

difference was attributed to the tendency of HU to induce the sharpbending of DNA required for P1. On the other hand, Zhang et al.quantitatively interpret the same repression efficiency data consid-ering solely P1E looping (10), which points to the fact that inter-pretation of the measurement is not unequivocal and depends onthe details of the model.

The aforementioned biochemical assays monitor either staticstructures or averages of different species that coexist or intercon-vert on a timescale faster than the temporal resolution of theexperiment. However, LacI-DNA complexes are dynamic and cantransition through different states. Therefore, a further refinementof the role of various loops can be accomplished only by using anensemble averaging free single-molecule approach, which enablesdynamic observation of loop formation/breakdown and, therefore,offers the potential to identify diverse and temporally separatespecies, as demonstrated in the work by Finzi and Gelles usingtethered particle motion (TPM) (21). A subsequent experiment ona DNA construct with shorter (158 bp and 153 bp) inter-operatorspacings led to the observation of two interconverting looped states,attributed to a V-shape and an open tetramer conformer (15).

Multiple loop conformations were detected by TPM also withtorsionally constrained DNA (22).

In the present study we address the significance of LacI tetramerflexibility in loop formation. Tetramer opening was previouslyconsidered in theoretical works and simulations (9, 10, 23), ob-served in solution (11, 13) or with LacI bound to a prebent DNA(12). To further explore this issue, we apply the TPM method (Fig.1C) to a DNA construct with a relatively long interoperatordistance (305 bp), which is expected to allow the formation of anumber of different loop conformers that might otherwise beenergetically unfavorable at smaller spacings. Hypothesizing thatsome of these structures could be associated with repressor open-ing, marked changes in the TPM measurements are expected as wechemically cross-link the LacI tetramer at specific positions (Fig. 1D and E) to reduce conformational flexibility and therefore pro-hibit open tetramer-associated structures.

ResultsTwo TPM Looping States and Looping/Unlooping Kinetics. The timerecordings of the bead’s average radius of mobility �R� (seeMaterials and Methods) for the wild-type and the two uncross-linkedprotein mutants (Fig. 2 A–C) all exhibit the same pronouncedthree-state character, evident upon visual inspection of the exper-imental traces (Left) and quantitatively highlighted by the �R�histograms exhibiting three distinct peaks (Right). The state with thelongest tether length corresponds to an unlooped DNA (21, 24),whereas the two states of lower bead mobility correspond todifferent looped DNA configurations (15, 22). The differences seenin the histograms of Fig. 2 derive from a combination of bead-to-bead variability, which is typical of TPM experiments, and smallpossible differences in the final concentrations of the proteins. The

Fig. 1. Structures of Lacl variants, looped complexes, and chemical cross-linkers.(A)LacI tetramerstructurewithDNA-bindingdomaininred,proteincore ingreenand tetramerization domain in blue. Each dimer is shown in color tones ofdifferent darkness. (B) Different LacI-DNA loop topoisomers. Arrows indicate the5�-3� orientation of operators. (C) TPM setup: unlooped and looped configura-tions highlight the difference of bead mobility. Schematic (not to scale) drawingof DNA construct indicates the lengths of different segments (interoperatordistance reported as center-to-center). (D) The E36C (Left) and Q231C (Right) LacImutants. Each protein is shown in side view (Upper) and top view (rotated by 90°about the horizontal axis, Lower), with each dimer in a different shade of gray.E36C is shown cross-linked with DPDPB and Q231C with BMOE; the cross-linker isshown in red spacefill representation. The tetramer structure shown is a model ofthe maximal opening compatible with the chosen cross-linker. (E) Chemicalstructuresofthecross-linkers.Theexternal sulfuratomsineachstructurearefromthe cysteine sulfhydryl groups of the cross-linked protein.

Fig. 2. Representative TPM recordings (Left) obtained with wild-type LacI,uncross-linked E36C, uncross-linked Q231C, and Q231C cross-linked with BMOE.All traces show �R� obtained by Gaussian filtering with � � 4 s. The Right panelsshow histograms of corresponding recordings on the Left.

16628 � www.pnas.org�cgi�doi�10.1073�pnas.0904617106 Rutkauskas et al.

full kinetic scheme of the reactions underlying the TPM measure-ment contains two looped states and interconversion between them(Fig. 3A). To access the rate constants, �R� traces are first trans-formed into traces of states (Fig. 3B), by applying a half-amplitudethreshold (see Materials and Methods). Then, a particular rateconstant is obtained by fitting a single-exponential decay to thedistribution of durations in a corresponding looped state beforetransition to a selected alternative state (Fig. 3C). From thisanalysis, we find that the rate constants of the transitions from thelong tether looped state are of similar magnitude and about twiceas large as those pertaining to the short tether looped state. Solvingthe set of equations for the partial equilibria shown in Fig. 3A,yields:

Jm1/Jm2 � kL1k21/kL2k12 [1]

Substituting for the measured rates (Fig. 3C) results in Jm1/Jm2 � 0.9 � 0.2. The indicated error margin was calculated bypropagating rate constant fitting errors. Analysis of the digitaltraces also yields the probabilities of the different transitionsbetween states. The transitions between the looped states andthe unlooped state and those between the two looped states inboth directions (about 100 transitions of each kind were ob-served) apparently occur with comparable frequency (within

approximately 20%), indicating that the looped states can in-terconvert directly, without an intermediate transition to theunlooped state, as shown also by Wong et al. (15).

Effects of Cross-linking on Operator Binding and on Looping. Therelative mobility of the dimer units within tetrameric LacI wasexplored by cross-linking of LacI protein mutants containing singlecysteine residues at positions 36 or 231 (Fig. 1D). Residue 36 wasselected based on its position at the operator-distal region of theN-terminal DNA binding domain, whereas 231 was selected as partof the small cluster of residues that make contact between twodimers within a tetramer. Both sites were expected to be tolerantof substitutions to cysteine, based on systematic LacI mutagenesisstudies [reviewed in (25)]. Three different cross-linking agents wereused (Fig. 1E): BMOE (bis-maleimidoethane), BM[PEO]2 (1,8-bismaleimidodiethylene glycol), both irreversible, and DPDPB(1,4-di-[3�-(2�-pyridyldithio)-propionamido]butane), which is re-versible through reduction with DTT. Each of these reagents hashigh specificity for cysteine residues and was reacted with proteinat concentrations that favor the intramolecular reaction. Nominalspacer arm lengths are: 8 Å for BMOE, 15 Å for BM[PEO]2, and20 Å for DPDPB, allowing a comparison of different constraints ontetramer flexibility. Note that these distances set the higher limit forthe tetramer opening as a cross-linker may not prohibit tetramerclosing; we also assume that the protein structure is unperturbed bythe reaction with the cross-linker.

The C�–C� distances, measured across dimers in the V-shapedLacI using the theoretically predicted all-atom model [1Z04, (26)],are 17.5 Å at residue 36 and 9.2 Å at residue 231. Therefore, E36Cwas cross-linked with the long or medium cross-linkers (respec-tively, DPDPB and BM[PEO]2), while Q231C was cross-linked withBM[PEO]2 or the short (BMOE) cross-linker. By modeling possiblestructures of the cross-linked proteins (as shown in Fig. 1D), themaximum tetramer opening angles [�, as described in (9)] wereestimated at 44° for E36C-DPDPB (Fig. 1D, Left), 84° for Q231C-BMOE (Fig. 1D, Right), and 108° for Q231C-BM[PEO]2. Linkingthe N-terminal domains within the tetramer (at position 36) inhibitsthe opening of the ‘‘V’’ structure to a greater degree than linkingthe C-subdomains of the core (at position 231).

Crosslinking induces no significant change in the binding of 40 bpoperator DNA, nor in the IPTG-responsiveness of E36C (Fig. 4A–C), with the exception of greater operator affinity in the presenceof IPTG for E36C-BM[PEO]2. Crosslinking of Q231C with eitherBM[PEO]2 or BMOE reduced operator affinity by about one orderof magnitude (Fig. 4 D–F) and somewhat diminished its respon-siveness to IPTG. Stoichiometric measurements indicate that thesemodified proteins remain fully active in binding to operator DNA.

In contrast to the results with 40-bp operator DNA binding,chemical cross-linking of E36C with BM[PEO]2 or DPDPB led toa complete abolition of the repressor capacity to mediate looping:out of approximately 30 tethers satisfying the TPM selection criteria(see Materials and Methods) in the measurements with E36C-BM[PEO]2 none exhibited looping dynamics within measurementtimes of at least one hour. This observation was additionally verifiedby measuring the same DNA tethers, alternating cross-linked anduncross-linked protein (see Materials and Methods): out of 15tethers monitored, 12 exhibited a change from no looping (withcross-linked E36C) to looping (with uncross-linked E36C). Fur-thermore, the looping ability of E36C-DPDPB was restored uponcleavage of the DPDPB linkage with DTT: out of nine tethersmonitored, seven exhibited a change from the absence of dynamics(with E36C-DPDPB) to looping (with DTT-treated E36C-DPDPB).

Crosslinking of Q231C with the longer BM[PEO]2 or the shorterBMOE, on the other hand, did not abolish looping but significantlymodified the structure of the observed looped states. The twodistinguishable looping states observed with the uncross-linkedprotein (Figs. 2C and 3B) could no longer be clearly deciphered in

Fig. 3. Loopingkinetics. (A)Kinetic schemewithtwo loopedstates. [R] indicatesthe concentration of Lac repressor tetramer. (B) Uncross-linked Q231C: a trace ofstates (red) superimposedonatraceof �R� (black)withthresholds indicatedbythegreen lines. (C) Distributions of uncross-linked Q231C state durations (blacksquares) before a particular transition (indicated by an arrow connecting twostates) fitted with mono-exponential functions (red lines). These distributionsresult from the records of 20 tethers with total measurement time of about 13 h.The short and long tether looped states are arbitrarily numbered as 1 and 2,respectively. The resulting rate constants are: k12 � 0.022 � 0.003 s�1, kL1 �0.023 � 0.003 s�1, k21 � 0.041 � 0.002 s�1, and kL2 � 0.046 � 0.005 s�1.

Rutkauskas et al. PNAS � September 29, 2009 � vol. 106 � no. 39 � 16629

BIO

PHYS

ICS

AN

DCO

MPU

TATI

ON

AL

BIO

LOG

Y

individual �R� traces (Fig. 2D). Fig. 5 shows a comparison of thecumulative �R� histograms measured with uncross-linked Q231C(Fig. 5A), Q231C-BM[PEO]2 (Fig. 5B), and Q231C-BMOE (Fig.5C). Q231C cross-linked with either cross-linker displays a wide �R�distribution in the range between 110 and 160 nm, where the twolooping-associated peaks appear to be complemented with otherstates with intermediate �R� values. Fig. 5 D and E highlight thechanges of �R� probability distribution induced by cross-linking. Fig.5D shows that cross-linking with BM[PEO]2 results in a decrease of

probability of the two main looped states (centered at about 120 nmand 140 nm), in favor of a third conformation peaking at about 130nm. Fig. 5E shows that cross-linking with the shorter BMOE mainlyresults in a transfer of probability from the 120 nm peak to the 130nm intermediate loop conformation.

DiscussionLacI-mediated DNA looping provides a versatile mechanism forenhancing transcription blockade by increasing the effective localconcentration of LacI with respect to the primary operator. Tran-scription regulation via looping is made robust by the existence ofvarious loop conformers, adopted with different likelihood depend-ing on parameters such as interoperator distance, DNA supercoil-ing and/or flexibility, and protein structure/flexibility. A particularloop structure is defined by a combination of accessible proteinconformations and DNA binding topology. Here we have ad-dressed the question of the significance of LacI tetramer openingwith regard to loop formation in vitro for a DNA construct with aninteroperator spacing of 305 bp. TPM assay provides simple meansfor dynamic measurements of the effective end-to-end length ofsingle DNA molecules. We have observed that wild-type LacI, aswell as E36C and Q231C protein mutants, exhibit two pronouncedTPM looped states. Curiously, the two-looped-state structurefound in these TPM traces was not previously observed with DNAconstructs having the same interoperator spacing but longer flank-ing sequences (24). If we assume that various looping states areassociated with different DNA-protein conformations, two TPMlooping states are presumably observed because in our DNAconstruct the lengths of the flanking segments approach thecanonical DNA persistence length, so that they tend to point in thedirection of their exit from the repressor DNA-binding domains.Consequently, various loop structures (with different relative exitangles from the two DNA binding domains) result in differenteffective DNA end-to-end distances (15). The flexibility of longerflanking sequences [as in (24)] abolishes this persistence effect andreduces the sensitivity of the TPM method to different loopgeometries.

We were able to quantitatively analyze the dynamics and to assessthe rate constants of the transitions between the different states,finding that the Jm values and, therefore, the energies of loopformation of the two looped states, are equal within the error. Thisresult, in combination with similar lifetimes of the two looped states

Fig. 4. Operator binding. O1 binding measured at equilibrium in the absence (filled symbols) or in the presence of IPTG (1 mM, open symbols), for E36C (A–C) andQ231C (D–F) uncross-linked or cross-linked as indicated. In B, C, E and F the dotted line shows the binding curve of uncross-linked protein for reference. In C the orangedata show the effect of pretreating E36C-DPDPB with DTT. The values of KD (expressed in pM) obtained by fitting the data are as follows: 14 � 2 (E36C); 22 � 3(E36C-BM[PEO]2); 12 � 1 (E36C-DPDPB); 23 � 2 (Q231C); 320 � 50 (Q231C-BM[PEO]2); 240 � 30 (Q231C-BMOE). For comparison, KD for wild-type LacI is 15 � 4 (30).

Fig. 5. Aggregate distributions of �R� for Q231C uncross-linked (A) and cross-linked with BM[PEO]2 (B), or BMOE (C). Histograms are composed from therecords of 16 tethers (15 h) for Q231C, 21 tethers (18 h) for Q231C-BM[PEO]2 and23 tethers (19 h) for Q231C-BMOE. To minimize the effects of bead-to-beadvariability, timerecords, irrespectiveof thenumberof loopingstates theyexhibit,werechosenwiththeunloopedstatewithinanarrow(10nm) intervalaroundthemaximum (170 nm) of the distribution of the unlooped states of a larger popu-lation of tethers (approximately 100). D and E highlight the changes in thedistributions induced by cross-linking with BM[PEO]2 (D) or BMOE (E). The histo-gram in gray shows the uncross-linked data (the same as in A) for reference; thered bars show the difference between cross-linked and uncross-linkedhistograms.

16630 � www.pnas.org�cgi�doi�10.1073�pnas.0904617106 Rutkauskas et al.

(Fig. 3C), is consistent with approximately equal areas of theirprobability distributions (Fig. 5A). Furthermore, the kinetic analysisof TPM traces revealed that transitions between different statesoccur with comparable probabilities and, in particular, the twolooped states interconvert without necessarily passing through theunlooped state. This observation implies that the switching of thelooped state cannot be always associated with a change of the DNAbinding topology, since this would require intermediate unlooping.As some transformation of the DNA-protein arrangement is re-quired to alter the observable tether length, the data point to thenecessity of a protein conformational change: for example, tetrameropening. Therefore, the two observable TPM looping states haveto be accounted for by a DNA topoisomer assembling both with theV-shaped and open tetramer conformations. Interconversion be-tween two looped states was previously observed on DNA mole-cules with 158 bp interoperator distance (15). ComplementaryAFM data were interpreted in terms of DNA wrapping aroundLacI in a P2 configuration, and, from a simple estimate of tetherlengths, it was concluded that wrap-toward P2 with the open andV-shape tetramer gives rise to the short and long tether states,respectively (15). Due to different experimental conditions andinteroperator distance, this interpretation cannot be directly ap-plied to our results. Computational work (9, 10) on wrap-away loopsindicates that, for short interoperator distances, P1 is energeticallyfavorable with respect to P2; further, P1 with the open tetramer(P1E) is more favorable than any other loop type (P1, P2, A1, andA2). The computed energy difference between P1E and P1 decaysas a function of interoperator distance (9, 10): extrapolation to 305bp suggests a negligible difference between P1E and P1. Consistentwith this observation, we find that the energies of forming short-and long-tether loops are essentially equal. Therefore, it is temptingto suggest that direct interconversion between our TPM loopingstates involves the transition between P1E and P1. However,analysis of loop structures responsible for the observed TPM statesis complicated by potential degeneracy of the TPM recording,where different loop structures may correspond to the same ap-parent tether length. Calculations by Towles et al. (27), for example,show that wrap-away P2, A1, and A2 all result in similar values of�R�, corresponding to a short-tether looped state, and antiparallelloops are energetically advantageous. The long tether, on the otherhand, is attributed to P1, which, at 305 bp, features an energyintermediate between P2 and the antiparallel loops. These calcu-lations enable interpretation of the measured tether length distri-butions entirely in terms of loop topoisomers with V-shapedtetramer but do not account for the fact of direct interconversionbetween the two TPM looping states. Given these uncertainties, theassessment of the actual role of tetramer opening in DNA loopingrequires a more direct approach.

We used TPM to compare the dynamics of looping mediated bywild-type LacI, E36C, and Q231C, with those of the cross-linkedvariants that are characterized by diminished conformational flex-ibility. Crosslinking of E36C led to an overwhelming loss of itscapacity to mediate looping. However, chemical modification didnot have any effect on the E36C binding affinity to 40 bp O1 DNAfragments. In concert, these data imply that cross-linking impededthe possibility of structural readjustment of repressor required toaccommodate the two operators simultaneously in a looped con-figuration. That the integrity of the protein is not otherwise affectedis confirmed by the fact that E36C looping ability was restored uponcleavage of DPDPB linkage with DTT. Possible implications of theobserved cross-linking effect range from restricting the fine orient-ing of the headpieces to prohibiting large scale conformationalchanges associated with the tetramer opening (see Fig. 1D, Left).Further experiments will be required to address these distinctions,but it should be noted that both BM[PEO]2 and DPDPB spacerarms are sufficiently long to cross-link the tetramer in the classicV-shape; therefore, our data indicate that both TPM looping statesand, thus, all of the underlying loop conformations involve some

degree of protein opening with respect to the canonical V-shape.On the other hand, we found that E36C can be efficiently cross-linked with the short-arm BMOE cross-linker (causing a significantdistortion away from the classic V-shape), and this modification hadno effect on 40 bp O1 binding, demonstrating that LacI tetramer canundergo significant dimer-to-dimer angle changes without effectson operator binding. The fact that cross-linking with BMOE cantake place in solution also demonstrates the underlying thermalfluctuations of the tetramer.

Crosslinking of Q231C did not abolish its ability to mediatelooping but modified the structure of the observable looped states(Fig. 5). It should be noted that the TPM experiments are carriedout at a protein concentration of 100 pM, so that the measuredreduction in O1 affinity due to cross-linking with either BM[PEO]2or BMOE (see Fig. 4 E and F) would at most have an effect ofreducing double operator occupancy (RO-OR in Fig. 3A). Indeed,this effect, leading to a lower occupancy of the TPM unlooped state,is observable in the reduction of unlooped probability (Fig. 5C). Inthe average �R� histograms, chemical modification with either longBM[PEO]2 (Fig. 5B) or short BMOE (Fig. 5C) introduced statesintermediate between the two clearly distinguishable short and longtether loops observed with the uncross-linked protein. These in-termediate states could be interpreted in terms of loop structureswith intermediate degrees of tetramer opening. Thus, the cross-linker conceivably modifies the conformational energy landscape ofthe LacI-DNA assembly in such a way as to stabilize configurationsintermediate between the protein in the V-shape and open tetramerto an extent sufficient for TPM observation. In Q231C, where thecross-linker connects to the protein close to the apex of its V-shape(see Fig. 1D, Right), residual flexibility may potentially be attainedthrough flexing of the portion of the protein N-ward of thecross-link point. This part of the protein includes the bindingdomains for inducer and for DNA: some distortion of the structuremight explain the decrease of operator affinity and of IPTG-responsiveness observed upon cross-linking (Fig. 4 E and F).

In conclusion, LacI flexibility and the opening of the tetramericstructure to a variety of angles play an important role in loopformation under the conditions of our experiment. Our observa-tions indicate that the protein can adopt different structures toaccommodate different loop geometries. This structural flexibilityis very likely to play an important role in looping in vivo undervarying DNA topological conditions.

Materials and MethodsCrosslinkable Protein Mutants, Preparation of Proteins and DNA. The DNAconstruct, end-labeled with biotin and digoxigenin, was obtained as describedpreviously (24). The lengths of the different DNA segments are shown in Fig. 1C.The three native cysteine residues in LacI (at positions 107, 140, and 281) werereplaced with alanine; E36C or Q231C were each introduced by site-specificmutagenesis (Quickchange, Stratagene) of the Cys-less LacI gene. Previous datademonstrate that these mutants exhibit no phenotypic effect on LacI function(25). Wild-type LacI and protein mutants were expressed and purified, as de-scribed previously (28), except that the purification of mutant proteins wascarried out in the presence of 2 mM DTT.

Crosslinking. Before cross-linking, complete reduction of cysteines was ensuredby incubation (15 min at room temperature) in cross-linking buffer (CB: 100 mMsodium phosphate buffer, pH 7.15, 150 mM NaCl, 5 mM EDTA) with 15 mM�-mercaptoethanol. In the subsequent steps, CB buffer deoxygenated by N2

bubbling was used. The protein was transferred to CB buffer using a PD-10column (GE Healthcare). BMOE, BM[PEO]2, or the cleavable DPDPB (Pierce) weredissolved inDMSOandaddedinatwo-foldmolarexcesswithrespect tothedimerconcentration. The cross-linking reaction was carried out for 1 h at room tem-perature. The final oligomerization state of cross-linked proteins was verified tobe tetramer by running the cross-linked products on Superdex200, using LacIdimer and tetramer as references. Crosslinking was verified by SDS-polyacryl-amide gel electrophoresis that demonstrated formation of SDS-resistantoligomer after cross-linking; where the reaction was reversible with DTT, resto-ration of monomer was observed. Cleavage of DPDPB, when required, was

Rutkauskas et al. PNAS � September 29, 2009 � vol. 106 � no. 39 � 16631

BIO

PHYS

ICS

AN

DCO

MPU

TATI

ON

AL

BIO

LOG

Y

obtained by treating the protein with 50 mM DTT in CB buffer for 20 min. Allsamples were stored at �20 °C.

The extent of geometrical constraint induced upon cross-linking of E36C withDPDPB and Q231C with either BM[PEO]2 or BMOE was estimated by modelingtetramer opening as a rigid body rotation of each dimer about an axis running inthe center of the tetramerization four-helix bundle and perpendicular to theplaneof theclassicV-shapeviewofLacI (theplaneof thepage inFig.1D). Foreachprotein mutant/cross-linker pair, the extent of opening required to bring the twocysteine residues’ sulfhydryl groups at a distance corresponding to the full lengthof the cross-linker was calculated and the value of the tetramer aperture angle �

[as defined in (9)] was measured.

Operator Binding Assay. NitrocellulosefilterbindingassayswereusedtomeasureDNA binding affinity and stoichiometry, as described previously (28). Briefly, 40bp lac operator DNA was labeled with [32P]ATP (using polynucleotide kinase) andpurified. Protein at varying concentrations was mixed with the DNA, and sampleswere passed through nitrocellulose filters; the retention of radiolabel was de-tected by phosphorimaging. The data were fit to equations of binding to deter-mine the operator DNA binding constant (where DNA concentration was �10-fold below the Kd) and stoichiometry (where DNA concentration was �100-foldabove the Kd).

Preparation of Tether Assay, Data Collection, and Analysis. We used a custom-made flow chamber (approximately 20 �L volume) constructed from a micro-scope coverslip and a Plexiglas™ element with attached tubing, separated by aspacer of silicone elastomer sheet with a cut-out flow channel (0.020‘‘ thickSE30-GF Silicone 30 Duro, Stockwell Elastomerics). The samples were prepared asdescribed previously (24) with some modifications. Briefly, a dilute suspension of1.54 �m diameter �-casein-coated silica beads (SS04N, Bangs Laboratories) inphosphate buffer saline (PBS) [2.7 mM KCl, 13.7 mM NaCl, 5.4 mM Na2HPO4, 1.8mMKH2PO4 (pH7.0)]was incubateduntila fewbeadsperfieldofviewwerestuckonto the coverslip. After removal of unbound beads, the chamber was incubatedwith 100 �g/mL anti-Digoxigenin (Roche) in PBS solution for 30 min, followed bywashing with 10 � 200 �L of LBB buffer [10 mM Tris-HCl (pH 7.4), 200 mM KCl, 0.1mM EDTA, 5% (vol/vol) dimethyl sulfoxide (DMSO), 0.2 mM DTT, 0.1 mg/mL�-casein]. DNA (approximately 100 pM in LBB) was then incubated for 1 h. Afterfive 200 �L washes with LBB� buffer (LBB lacking DTT and DMSO), streptavidin-coated polystyrene beads (440 nm diameter, Indicia Biotechnology) were incu-

bated for 10 min. Unbound beads were removed with 5 � 200 �L of LBB buffer.Video recordings of tethered particle motion were acquired on the setup de-scribed in (24), with addition of an active xyz stage drift compensation by movingthe xy piezo stage (Physik Instrumente) to correct for changes of xy centroidposition and z-sensitive diffraction ring of a silica bead selected within the regionof interest (ROI). A transient movie of a whole field of view was analyzed by asoftwareroutinetofindROIscontainingtethersandthexypositionsofthebeads’centroids were recorded at 25 Hz. Typically 1 h records were obtained beforechanging field of view. The raw x and y traces were selected for further analysisif they satisfied the following criteria: 1) traces of the x and y coordinates of thetether anchoring point (obtained by averaging each coordinate over N frames)exhibit monotonous and correlated evolution. A different behavior is suspect ofspurious bead/surface interactions; 2) the bead diffusion cloud is symmetric, asdescribed by Blumberg et al. (29) to ensure single DNA tethering of the bead;3) a trace of

�r� �1N �

i�j

jN

�xi � x��2 � yi � y��2

displays transitions between different states.The selected x and y traces were filtered with a Butterworth high-pass filter

(cut-off frequency 0.1 Hz) to position the origin of the bead diffusion cloud at (0,0). Then, the instantaneous radius R � �x2 y2 was calculated and convolutedwith a Gaussian function with � � 4 s to obtain a trace of the average radius ofmobility (�R�), examplesofwhichareshowninFig.2.Digital traces (seeFig.3B, redtrace), were obtained by applying a half-amplitude threshold procedure withthresholds determined for each trace by fitting a Gaussian function to each of thethree peaks in the histogram and setting the thresholds midway between thepeak positions found. States with durations shorter than 1.5� (i.e., 6 s) weremerged with successive states. This correction is essential for elimination of falsetransitions due to the finite dead time of the Gaussian filter.

ACKNOWLEDGMENTS. We thank Dr. Davide Normanno for discussion on themanuscriptandMr.RiccardoBallerini for technicalhelp.Thisworkwas supportedby European Union Contract MTKD-CT 2004-509761, ASI (project MoMa), Na-tional Institutes of Health Grant GM22441, and Robert A. Welch FoundationGrant C-576.

1. Reznikoff WS, Winter RB, Hurley CK (1974) The location of the repressor binding sitesin the lac operon. Proc Natl Acad Sci USA 71:2314–2318.

2. Lewis M, et al. (1996) Crystal structure of the lactose operon repressor and its complexeswith DNA and inducer. Science 271:1247–1254.

3. Oehler S, Eismann ER, Kramer H, Muller-Hill B (1990) The three operators of the lacoperon cooperate in repression. EMBO J 9:973–979.

4. Matthews KS (1992) DNA looping. Microbiol Rev 56:123–136.5. Shimada J, Yamakawa H (1984) Ring-closure probabilities for twisted wormlike chains.

Application to DNA. Macromolecules 17:689–698.6. Muller J, Oehler S, Muller-Hill B (1996) Repression of lac promoter as a function of

distance, phase and quality of an auxiliary lac operator. J Mol Biol 257:21–29.7. Shore D, Langowski J, Baldwin RL (1981) DNA flexibility studied by covalent closure of

short fragments into circles. Proc Natl Acad Sci USA 78:4833–4837.8. Zhang Y, McEwen AE, Crothers DM, Levene SD (2006) Statistical-mechanical theory of

DNA looping. Biophys J 90:1903–1912.9. Swigon D, Coleman BD, Olson WK (2006) Modeling the Lac repressor-operator assem-

bly: The influence of DNA looping on Lac repressor conformation. Proc Natl Acad SciUSA 103:9879–9884.

10. Zhang Y, McEwen AE, Crothers DM, Levene SD (2006) Analysis of in-vivo LacR-mediatedgene repression based on the mechanics of DNA looping. PLoS ONE 1:e136.

11. Ruben GC, Roos TB (1997) Conformation of Lac repressor tetramer in solution, boundand unbound to operator DNA. Microsc Res Tech 36:400–416.

12. Edelman LM, Cheong R, Kahn JD (2003) Fluorescence resonance energy transfer overapproximately 130 basepairs in hyperstable lac repressor-DNA loops. Biophys J84:1131–1145.

13. Taraban M, et al. (2008) Ligand-induced conformational changes and conformationaldynamics in the solution structure of the lactose repressor protein. J Mol Biol 376:466–481.

14. Semsey S, Virnik K, Adhya S (2005) A gamut of loops: Meandering DNA. Trends BiochemSci 30:334–341.

15. Wong OK, Guthold M, Erie DA, Gelles J (2008) Interconvertible lac repressor-DNA loopsrevealed by single-molecule experiments. Plos Biology 6:2028–2042.

16. Virnik K, et al. (2003) ‘‘Antiparallel’’ DNA loop in gal repressosome visualized by atomicforce microscopy. J Mol Biol 334:53–63.

17. Morgan MA, Okamoto K, Kahn JD, English DS (2005) Single-molecule spectroscopicdetermination of lac repressor-DNA loop conformation. Biophys J 89:2588–2596.

18. Borowiec JA, Zhang L, Sasse-Dwight S, Gralla JD (1987) DNA supercoiling promotesformation of a bent repression loop in lac DNA. J Mol Biol 196:101–111.

19. Kramer H, Niemoller M, Amouyal M, Revet B, von Wilcken-Bergmann B, Muller-Hill B(1987) lac repressor forms loops with linear DNA carrying two suitably spaced lacoperators. EMBO J 6:1481–1491.

20. Becker NA, Kahn JD, Maher LJ, 3rd (2005) Bacterial repression loops require enhancedDNA flexibility. J Mol Biol 349:716–730.

21. Finzi L, Gelles J (1995) Measurement of lactose repressor-mediated loop formation andbreakdown in single DNA molecules. Science 267:378–380.

22. Normanno D, Vanzi F, Pavone FS (2008) Single-molecule manipulation reveals super-coiling-dependent modulation of lac repressor-mediated DNA looping. Nucleic AcidsRes 36:2505–2513.

23. Villa E, Balaeff A, Schulten K (2005) Structural dynamics of the lac repressor-DNAcomplex revealed by a multiscale simulation. Proc Natl Acad Sci USA 102:6783–6788.

24. Vanzi F, Broggio C, Sacconi L, Pavone FS (2006) Lac repressor hinge flexibility and DNAlooping: Single molecule kinetics by tethered particle motion. Nucleic Acids Res34:3409–3420.

25. Pace HC, et al. (1997) Lac repressor genetic map in real space. Trends Biochem Sci22:334–339.

26. Balaeff A, Mahadevan L, Schulten K (2004) Structural basis for cooperative DNAbinding by CAP and lac repressor. Structure 12:123–132.

27. Towles KB, Beausang JF, Garcia GH, Phillips R, Nelson PC (2009) First-principles calcu-lation of DNA looping in tethered particle experiments. Physical Biology 6:025001.

28. Zhan H, Swint-Kruse L, Matthews KS (2006) Extrinsic interactions dominate helicalpropensity in coupled binding and folding of the lactose repressor protein hinge helix.Biochemistry 45:5896–5906.

29. Blumberg S, Gajraj A, Pennington MW, Meiners JC (2005) Three-dimensional charac-terization of tethered microspheres by total internal reflection fluorescence micros-copy. Biophys J 89:1272–1281.

30. Zhan H, Sun Z, Matthews KS (2009) Functional impact of polar and acidic substitutionsin the lactose repressor hydrophobic monomer-monomer interface with a buriedkysine. Biochemistry 48:1305–1314.

16632 � www.pnas.org�cgi�doi�10.1073�pnas.0904617106 Rutkauskas et al.