insightintotheintegrase-dnarecognitionmechanism · to get deeper insight into the in/dna...

Insight into the Integrase-DNA Recognition MechanismA SPECIFIC DNA-BINDING MODE REVEALED BY AN ENZYMATICALLYLABELED INTEGRASE*

Received for publication, April 29, 2008, and in revised form, July 10, 2008 Published, JBC Papers in Press, August 12, 2008, DOI 10.1074/jbc.M803257200

Olivier Delelis1, Kevin Carayon1, Elvire Guiot, Herve Leh, Patrick Tauc, Jean-Claude Brochon,Jean-Francois Mouscadet, and Eric Deprez2

From the Laboratoire de Biologie et Pharmacologie Appliquee, CNRS, Ecole Normale Superieure Cachan, Institut d’Alembert,61 Ave. du President Wilson, 94235 Cachan, France

Integration catalyzed by integrase (IN) is a key process in theretrovirus life cycle. Many biochemical or structural humanimmunodeficiency virus, type 1 (HIV-1) IN studies have beenseverely impeded by its propensity to aggregate. We character-ized a retroviral IN (primate foamy virus (PFV-1)) that displays asolubility profile different from that of HIV-1 IN. Using varioustechniques, including fluorescence correlation spectroscopy,time-resolved fluorescence anisotropy, and size exclusion chro-matography, we identified a monomer-dimer equilibrium forthe protein alone, with a half-transition concentration of 20–30�M.We performed specific enzymatic labeling of PFV-1 IN andmeasured the fluorescence resonance energy transfer betweencarboxytetramethylrhodamine-labeled IN and fluorescein-la-beled DNA substrates. FRET and fluorescence anisotropy high-light the preferential binding of PFV-1 IN to the 3�-end process-ing site. Sequence-specific DNA binding was not observed withHIV-1 IN, suggesting that the intrinsic ability of retroviral INs tobind preferentially to the processing site is highly underestimatedin the presence of aggregates. IN is in a dimeric state for 3�-proc-essingonshortDNAsubstrates,whereas INpolymerization,medi-ated by nonspecific contacts at internal DNA positions, occurs onlonger DNAs. Additionally, aggregation, mediated by nonspecificIN-IN interactions, occurs preferentially with short DNAs at highIN/DNA ratios. The presence of either higher order complex isdetrimental for specific activity. Ionic strength favors catalyticallycompetent over higher order complexes by selectively disruptingnonspecific IN-IN interactions. This counteracting effect was notobservedwith polymerization. The synergic effect on the selectionof specific/competent complexes, obtained by using short DNAsubstrates under high salt conditions, may have important impli-cations for further structural studies in IN�DNA complexes.

Retroviral integrases (INs)3 catalyze the integration of viralDNA into the host genome ensuring its perpetuation in the host

cell. The integration process requires two catalytic steps. Dur-ing the first step, titled 3�-processing (3�-P), IN specificallyremoves two nucleotides from each viral DNA end. IN thentransfers both ends into target DNA by a one-step transesteri-fication reaction, resulting in full-site integration. IN alone iscompetent for the insertion process (1). Retroviral INs consistof three functional domains. The core domain contains theDDE catalytic triad and is flanked by the N-terminal domain(involved in multimerization) and the C-terminal DNA-bind-ing domain. The integrity of theDDE triad and ametallic cofac-tor are strictly required for enzymatic activity.Biochemical and structural studies of HIV-1 IN have been

severely impeded because of solubility. No structural data areavailable for the full-length protein (free or bound to DNA) todate, although x-ray structures of DNA-free single or doubledomains have been successfully solved (2–5). Hence, thesequence specificity of the 3�-P reaction remains poorly under-stood and little is known about the mechanism of IN/DNAsubstrate recognition. Point mutations or the use of detergentmay improve the solubility but only to a limited extent (6, 7) andmay also cause changes to IN properties. For instance, deter-gent is detrimental to 3�-P activity in the presence of the phys-iological cofactor Mg2� (8), and one soluble mutant was foundto be resistant to diketo-acid compounds (9).Endonucleolytic cleavage depends primarily on the presence

of the canonical CA sequence preceding the removed dinucle-otide. Other positions are crucial for 3�-P activity under Mg2�

conditions, despite the absence, in vitro, of IN preference/spec-ificity for the cognate U5 or U3 terminal sequence at the DNAbinding level (10–12). Thus, it is generally hypothesized that INspecificity is fully explained at the catalytic level. However, dueto poor solubility, it is a difficult task to determine which enzy-matic properties of HIV-1 IN actually correspond to intrinsicproperties of retroviral INs and which properties are related toaggregation. During the course of our study on HIV-1 IN, wehave observed significant differences relating to solubilitybetween HIV-1 IN and another retroviral IN, primate foamyvirus-1 (PFV-1) IN. PFV-1 is the prototype of foamy virusesbelonging to the retrovirus family and differ from lentivirusesin some aspects. For instance, PFV-1 reverse transcription

* This work was supported by TrioH European Project Grant FP6 503480,Agence Nationale de la Recherche Grant 06-PCVI-0015, and grants fromCNRS and Institut d’Alembert. The costs of publication of this article weredefrayed in part by the payment of page charges. This article must there-fore be hereby marked “advertisement” in accordance with 18 U.S.C. Sec-tion 1734 solely to indicate this fact.

1 Both of these authors contributed equally to this work.2 To whom correspondence should be addressed. Tel.: 33-147-40-23-94; Fax:

33-147-40-76-84; E-mail: [email protected] The abbreviations used are: IN, integrase; INT; TAMRA-labeled integrase;

FCS, fluorescence correlation spectroscopy; Fl, fluorescein; FRET, fluores-

cence resonance energy transfer; HIV-1, human immunodeficiency virus,type 1; LTR, long terminal repeat; ODN, oligodeoxynucleotide; PFV-1, pri-mate foamy virus type-1; TFA, time-resolved fluorescence anisotropy;TGase, transglutaminase; 3�-P, 3�-processing; Mo, monomeric; Di, dimeric;Te, tetrameric; TAMRA, carboxytetramethylrhodamine.

THE JOURNAL OF BIOLOGICAL CHEMISTRY VOL. 283, NO. 41, pp. 27838 –27849, October 10, 2008© 2008 by The American Society for Biochemistry and Molecular Biology, Inc. Printed in the U.S.A.

27838 JOURNAL OF BIOLOGICAL CHEMISTRY VOLUME 283 • NUMBER 41 • OCTOBER 10, 2008

by guest on March 23, 2020

http://ww

w.jbc.org/

Dow

nloaded from

http://www.jbc.org/

occurs at late stages of the replication cycle (13, 14). Interactionwith host cellular cofactors is also distinct, since LEDGF/p75only interacts with lentiviral INs (15). Furthermore, only theU5end is processed due to the asymmetric nature of the integra-tion process in the case of PFV-1 (16). However, integration isan obligatory step for replicating PFV-1, a hallmark of retrovi-rus life cycle (16). PFV-1 IN shares common structural featureswith other retroviral INs, such as three-domain organization aswell as the three major catalytic activities, 3�-P, strand transfer,and disintegration (17–19).To get deeper insight into the IN/DNA recognition mecha-

nism, we studied and compared solubility, oligomeric statusand catalytic efficiency of PFV-1 and HIV-1 IN. We used vari-ous methods, such as time-resolved fluorescence anisotropy(TFA), fluorescence correlation spectroscopy (FCS), and sizeexclusion chromatography, and found that free PFV-1 IN wassignificantly more soluble than HIV-1 IN. This enabled us tosuccessfully apply transglutaminase (TGase)-mediated TAMRAlabeling to a soluble IN at a single predefined position. Fluores-cence resonance energy transfer (FRET) experiments and aDNA-binding anisotropy-based assay were conducted to assessthe specificity of IN-DNA interactions. A specific mode ofinteraction was clearly shown for PFV-1 IN but not for HIV-1IN. DNA promotes the catalytically competent dimeric state.The dimerization process is in competition with other pro-cesses (which were found to be detrimental to activity), such asIN polymerization and aggregation, mediated by nonspecificIN-DNA and IN-IN interactions, respectively. A selection ofspecific/competent complexes may be achieved using shortDNA substrates under conditions of high ionic strength.

EXPERIMENTAL PROCEDURES

Purification and IN Labeling by Guinea Pig TGase—ForDNA-binding and 3�-P assays, unlabeled wild-type HIV-1 andPFV-1 IN were purified as previously described (8). The plas-mid encoding PFV-1 IN with the specific C-terminalPKPQQFM tag for TGase-mediated labeling was constructedby PCR amplification of pET15b-PFV-IN (20) with P1 and P2primers (P1, 5�-ACA TAT GTG TAA TAC CAA AAA ACCAAA CCT GG-3�; P2, 5�-AGG ATC CTA CAT AAA CTGCTGAGGTTTTGGCTCGAGTTCATTTTTTTC-3�). Thecorresponding plasmid for taggedHIV-1 INwas constructed byamplification of pET15b-HIV-IN (8) with P3 and P4 primers(P3, 5�-CCA TAT GTT TTT AGA TGG AAT AGA TAA-3�;P4, 5�-CGG ATC CTA CAT AAA CTG CTG AGG TTT TGGGTC CTC ATC CTG TCT ACT-3�). The underlined nucleo-tides correspond to theTGase tag (PKPQQFM) located at theCterminus of IN. The resulting DNAs encoding IN-PKPQQFMwere inserted into pET15b.For labeling, IN attached to Ni2�-beads was incubated over-

night at 4 °C with 0.5 mM TAMRA cadaverine (MolecularProbes) and 1 unit of guinea pig TGase (Sigma) in the labelingbuffer (20 mMHepes, pH 7.5, 10 mM CaCl2). Beads were exten-sively washed with Tris buffer (50 mM, pH 8, 1 M NaCl) toremove free fluorophores. Elutionwas donewith 1M imidazole.TAMRA-labeled or unlabeled PFV-1 INwas further purified bycation exchange chromatography on a Mono S column (Bio-Rad) previously equilibrated with 20 mM Hepes, pH 7.5, 0.1 M

NaCl. INwas diluted before loading (1ml/min) and a linear saltgradient (0.1–1 MNaCl) was applied at 1ml/min. INwas elutedat 450–550 mM NaCl (cation exchange chromatography wasunsuccessful with HIV-1 IN for solubility reasons). It was pos-sible to further concentrate PFV-1 IN using Amicon Ultra cen-trifugal filters (Millipore). Concentrations up to 250 �M in 20mM Hepes, pH 7.5, 75 mM NaCl were obtained. Ionic strengthwas adjusted by buffer exchange during the concentration step.Dialysis was not used, since protein aggregation was reproduc-ibly observed during the dialysis step of concentrated samples(significant precipitation occurred at PFV-1 IN concentrationsof 12 mg/ml, whichever protocol was used).Size Exclusion Chromatography—100 �l of varying concen-

trations of purified PFV-1 IN (from 2 to 60 �M) were applied atroom temperature onto a Superdex 200 preparation grade col-umn (Hiload 16/60; Amersham Biosciences), using an AKTAFPLC system (Amersham Biosciences) previously equilibratedwith the protein buffer (20 mM Tris-HCl, pH 8, 0.05 or 0.5 MNaCl, 10% glycerol, 50 �M ZnSO4, and 4 mM �-mercaptoetha-nol). The column was calibrated with the following globularproteins: sweet potato�-amylase (200 kDa), yeast alcohol dehy-drogenase (150 kDa), bovine serum albumin (66 kDa), ovalbu-min (45 kDa), carbonic anhydrase (29 kDa), and horse heartcytochrome c (12.5 kDa) (MW-GF-200 kit; Sigma). Fractionswere collected at a flow rate of 0.3 ml/min, under a pressure of0.5 megapascals. Absorbance was measured at 280 nm. Theexclusion volume of the column V0 was measured by calibra-tion with blue dextran (2000 kDa).Characterization of TAMRA-labeled IN (INT)—We evalu-

ated the number of TAMRAs per labeled protein bymass spec-trometry analysis; INT and the unlabeled tagged IN were sub-jected to SDS-PAGE. After gel staining, the gel slidescontaining the corresponding proteinswere destained in 25mMNH4CO3, 50% acetonitrile. A proteolysis using sequence gradetrypsin (Promega) was then performed overnight at 37 °C. Theresulting peptides were recovered by the addition of 50% ace-tonitrile, 5% trifluoroacetic acid and dried in a SpeedVac. Theywere reconstituted in 5% acetonitrile, 0.1% trifluoroacetic acidand applied on a SEND-ID protein array according to Cipher-gen’s protocol. The mass spectra were obtained by reading theprotein chip using a SELDI mass spectrometer. The resultingspectra were compared with the in silico profile obtained withGPMAW software (Lighthouse Data).The labeling yield of purified INT was estimated by both

absorbance spectroscopy and FCS. Given the molar extinctioncoefficient of TAMRA-cadaverine at 556 nm (62,000M�1�cm�1) and the results from the Bradford protein assay, thelabeling yield was estimated to be 45% by absorbance spectros-copy. The FCS autocorrelation curve (Fig. 1E) was also used toestimate the labeling yield, since it is possible to deduce fromthe inverse autocorrelation amplitude the mean number oflabeled molecules in the detection volume. Here, g(0)�1 wasequal to 0.025 and the detection volume was 1.27 �m3 (thelateral �0 and axial z0 dimensions are equal to 0.45 and 1.5 �m,respectively) giving an estimated concentration of INT (52 nM).Considering the concentration of total IN (100 nM), thededuced labeling yield (�50%) is in good agreement with thatdetermined by absorbance spectroscopy. This calculation is

Heterogeneity and Catalytic Properties of IN�DNA Complexes

OCTOBER 10, 2008 • VOLUME 283 • NUMBER 41 JOURNAL OF BIOLOGICAL CHEMISTRY 27839


http://ww

w.jbc.org/

Dow

nloaded from

http://www.jbc.org/

evidently correct only if a single fluorophore is attached to theprotein (SELDI analysis of INT confirmed that a single TAMRAmoiety was covalently and specifically attached to the proteinC-terminal end).INT, as obtained after the labeling procedure, was also tested

for 3�-P at the U5 end (21-mer oligonucleotide) using a stand-ard assay (8) and was compared with both the untagged and theunlabeled tagged proteins. 3�-P and half-transfer products wereabout 23–27% and 3.5–4.5%, respectively, for a 3-h incubation(data not shown), and no significant difference was observedbetween the three samples, showing that neither thePKPQQFM tag nor the attached fluorophore influences thecatalytic activity.PFV-1 ODNs for DNA Binding/3�-P Anisotropy-based Assays

and FRET Experiments—All ODNs (�65-mer) were purchasedfrom Eurogentec (Liege, Belgium) and further purified by gelelectrophoresis. ODNs of various lengths (15–65-mer) weretested for 3�-P activity, all mimicking the specific (cognate) ter-minal U5 sequence: PFV-CAAT (sequence of the 65-mer,5�-GAACTACACTTATCTTAAATGATGTAACTCCTTAGGATAATCAATATACAAAATTCCATGACAAT-3�.The 3�-dinucleotide cleaved by IN is underlined). The pro-cessed strand and its complementary nonprocessed strandwere denoted A and B, respectively. Fluorescein (Fl) wasattached at the 5�- or 3�-end of strand A for DNA-bindingexperiments. Fl was attached at the 3�-end of strand A for theactivity assay. The PFV-GTAT sequence was identical, exceptthat the four 3�-term bases of strand A were GTAT instead ofthe canonical sequence CAAT. The specific HIV-1 ODN (U5sequence) is provided in Ref. 21. The three nonspecific (ran-dom) sequences used for competition experiments (Fig. 4) wereas follows: 5�-ACA TAA TCT AAA ATA ATT GCC-3� (21-mer), 5�-ACC TAT GCG CCG CTA GAT TCC-3� (21-mer),and 5�-TCA AGC TAGAAGATT ATC TCAAGTACA TAATCTAAAATAATTGCC-3� (45-mer). Annealing of oligonu-cleotides was obtained by mixing equimolar amounts of com-plementary strands in 20 mM Hepes, pH 7.5, 100 mM NaCl,heating to 85 °C for 5 min, and slow cooling to 25 °C.For FRET experiments, ODNs of various lengths (21–300-

mer) were studied, all mimicking the PFV-1 U5 end. The dou-ble-stranded ODN was Fl-labeled at the 5�-end (A or B strand)(Fig. 3A). Fl-labeled 100- and 300-mer double-stranded ODNswere obtained by PCR. All PCRs were performed using thepHSRV13 plasmid encoding the proviral PFV-1 genome.The 100-mer BFl5 and 300-mer BFl5 were obtained using, asprimers, the single-stranded 21-mer BFl5 ODN and a secondODN hybridizing to an internal sequence located, respec-tively, 100 or 300 bp from the U5 extremity. The 100-merAFl5 and 300-mer AFl5 were obtained using, as primers, thenonfluorescent single-stranded 21-mer B ODN and a 5�-Fl-labeled ODN hybridizing to an internal sequence located,respectively, 100 or 300 bp from the U5 extremity. Afteramplification, PCR products were purified on agarose gelusing the Qiagen gel extraction kit.Steady-state Fluorescence Anisotropy-based Assay—Anisot-

ropy-based assay is a quantitative method to determine bothDNA binding and catalytic parameters of IN using the samesample (21). Briefly, IN binding to Fl-labeled DNA increases

the steady-state anisotropy value (r) (measured on a Beaconinstrument (Panvera, Madison, WI)), allowing the calcula-tion of the fraction of DNA sites bound to IN. The DNA-binding step can be recorded at 25 °C using Fl-labeled ODNsat the 5�- or 3�-end of either the A or B strand. By shifting thetemperature to 37 °C, the activity-dependent decrease in ther value allows quantification of the 3�-P reaction. This occursonly if Fl is initially linked to the AT dinucleotide (thereleased PFV-1 dinucleotide). The fraction of releaseddinucleotides (Fdinu) can be calculated using the relativedecrease of r compared with the initial value obtained at theend of the DNA-binding step (rt � 0) (real time assay) (21).No decrease in the r value was observed with the PFV-GTATcontrol sequence, in contrast to the PFV-CAAT canonicalsequence (data not shown). The activity can be also calcu-lated in fixed time experiments (3�-P is stopped by SDS),Fdinu � (rNP � r)/(rNP � rdinu), where rNP and rdinu are theanisotropies for pure solutions of nonprocessed double-stranded ODN and dinucleotide, respectively (rdinu wasmeasured using the 5�-AT-3�-Fl dinucleotide).The formation of IN�DNA complexes and the subsequent

3�-P catalytic reactionwere performed by incubating Fl-labeledODNs with unlabeled IN in buffer A (20 mM Hepes, pH 7.5, 1mM dithiothreitol, 10 mMMgCl2). IN, DNA, and NaCl concen-trations are explicitly mentioned in the figure legends.FRET Experiments—FRET between Fl-labeled DNA (donor)

and TAMRA-labeled IN (acceptor) was monitored with anEclipse (Varian) spectrofluorimeter. DNA concentration wasconstant (20 nM) in buffer A � 50 mM NaCl. INT was progres-sively added, and the decrease in the steady-state emissionintensity of the donor was recorded at 25 °C. The excitationwavelength was 500 nm. The donor quenching (qD) wasmeasured at 520 nm. qD was estimated using the equation,qD � 1 � (ID,[INT]/ID,0), where ID,0 and ID,[INT] represent thedonor emission intensity at 520 nm in the absence (DNAalone) and in presence of acceptor (along the titration),respectively. Excitation and emission slit widths were 10 and5 nm, respectively. For each DNA/INT mixture, a corre-sponding control experiment was done using unlabeled INto check the signal stability.Fluorescence Correlation Spectroscopy—FCS measurements

were performed under two-photon excitation (840 nm) on ahome-built system using an 80-MHz mode-locked TsunamiTi:Sapphire laser (pulse 100 fs) pumped by a Millenia solidstate laser (Spectra Physics) and a Nikon TE2000 invertedmicroscope. Before entering through the epifluorescenceport of the microscope, the laser beam was expanded with atwo-lens afocal system to fill the back aperture of the objec-tive (Nikon, Plan Apo, �100, numerical aperture 1.4, oilimmersion). The setup was optimized to obtain a diffrac-tion-limited focal spot. Measurements were typically carriedout in a 50-�l solution dropped on the coverslip (buffer A �50 mM NaCl). Fluorescence was collected by the same objec-tive, separated from the excitation by a dichroic mirror(Chroma 700DCSPXR) and focused onto an avalanche pho-todiode (SPCM-AQR-14; PerkinElmer Life Sciences). Anadditional filter was used to reject the residual excitationlight (Chroma 2P-Emitter E700SP). The detector was con-




http://ww

w.jbc.org/

Dow

nloaded from

http://www.jbc.org/

nected to a digital correlator (ALV 6000) that calculates thenormalized correlation function g(�) of fluorescence fluctu-ations according to Equation 1,

g�� I�t� � I�t � ��

I�t�2 (Eq. 1)

where I(t) is the number of detected fluorescence photons pertime unit. Assuming a three-dimensional Gaussian distributionof the excitation intensity, the FCS function for a free Browniandiffusion process is given by Equation 2,

g�� 1

N�

1

�1 ��

�D� � �1 �

�02

z02 �

�

�D

(Eq. 2)

where N is the mean number of fluorophores in the excitationvolume, and �D is the translational diffusion time. �0 and z0 arethe lateral and axial dimensions of the excitation volume,respectively. The mean translational diffusion time was deter-mined by fitting autocorrelation curves using a Levenberg-Marquardt nonlinear least-squares fitting algorithm accordingto the analytical model (Equation 2). For two-photon excita-tion, the diffusion coefficient D is then calculated according tothe equation, D � �2/8�D. The calibration of the excitationvolumewas performed using a 10 nMwater solution of TAMRA(D � 2.8 � 10�10 m2 s�1). The lateral �0 and axial z0 dimen-sions of the excitation volume were 0.45 and 1.5 �m, respec-tively. The excitation power was adjusted using a neutral den-sity filter. For our setup, we determined that an excitationpower of 10milliwattswas suitable for two-photon excitation ofboth TAMRA alone and INT (fluorescence intensity exhibits aquadratic dependence, and �D value is constant (absence ofphotobleaching) as a function of the incident power below 16milliwatts). Recording times were typically 5min (average of 10cycles of 30 s).Time-resolved Fluorescence Anisotropy—Fluorescence of

Trp or Fl was used for hydrodynamic studies of free or DNA-bound PFV-1 IN, respectively (in buffer A � 50 mM NaCl).IN�DNA complexes were obtained using unlabeled PFV-1 INand Fl-labeled 15- and 21-mer U5 PFV-1ODN (Fl at the 5�-endof strandA). Correlation times (�) were calculated from the twopolarized fluorescence decays I�(t) and I�(t), using time-corre-lated single photon counting (21, 22). Briefly, the time scalingwas 19.5 ps/channel, and 4096 channels were used. The excita-tion light pulse source was a Ti:Sapphire laser (Millennia-pumped Tsunami femtosecond laser; Spectra Physics) (repeti-tion rate, 8 MHz) associated with a second or third harmonicgenerator tuned to 490 or 296 nm for Fl or Trp, respectively.The emission monochromator (ARC SpectraPro-150) was setto 530 or 340 nm (� � 15 nm) for Fl or Trp, respectively. Thetwo polarized components were collected alternately over aperiod of 30 s (total count of I�(t) was 15 millions, which is acondition compatible with the recovery of correlation times upto �80 ns using Trp fluorescence (7)). I�(t) and I�(t) were ana-lyzed by the maximum entropy method (23, 24) to determinethe distributions of both lifetimes (�) and rotational correlationtimes (�), according to Equations 3 and 4.

I��t� �1

3 � �

i � 1

n

� ie�1/�i�1 � 2 �

j � 1

m

�je�1/�j�� (Eq. 3)

I��t� �1

3 � �

i � 1

n

� ie�1/�i�1 �

j � 1

m

�je�1/�j�� (Eq. 4)

with

�i � 1

n

� i � 1 (Eq. 5)

and

�j � 1

m

� j � r�0 (Eq. 6)

where�i represents the relative population characterized by thelifetime �i, and �j represents the initial anisotropy related to amotion characterized by the rotational correlation time �j. r�0 isthe apparent fundamental anisotropy value (typically 0.24–0.26 in the present study, which is below the fundamental ani-sotropy value, 0.3, forTrp atex� 296nm, suggesting that a fastcomponent in the PFV-1 IN anisotropy decay cannot beresolved by the instrument as previously found for many pro-teins (25)). The decay of the total fluorescence intensity, IT(t),and the fluorescence anisotropy decay, r(t), are then calculatedfrom both polarized components according to Equations 7 and8, respectively,

IT�t� � I��t� � 2G I��t� (Eq. 7)

r�t� �I��t� G I��t�

IT�t��

j � 1

m

� je�1/�j (Eq. 8)

where G represents the correction factor for the difference inthe monochromator transmission between parallel and per-pendicular polarized components.

RESULTS

TFA and FCS Reveal Distinct Solubility Properties betweenHIV-1 and PFV-1 INs—TFA and FCS are complementarymethods used to study the hydrodynamic properties of pro-teins, measuring the rotational and translational diffusion ofmolecules, respectively (25). Overall rotational diffusion orflexibility are major causes of light depolarization, and anisot-ropy decay analysis allows the determination of rotational cor-relation time (�) distribution. Long correlation time (�long) rep-resents the overall tumbling motion related to the proteinhydrodynamic volume. In FCS, the temporal behavior of fluo-rescence fluctuations within a small excitation volume,described by the autocorrelation function, allows the determi-nation of translational diffusion times.PFV-1 IN was first studied by TFA using Trp fluorescence.

Distributions of � are shown for two IN concentrations (Fig.1A). �long values were �25 and 53 ns for IN concentrations of 2and 125 �M, respectively. Given a molecular mass of 46.4 kDa




http://ww

w.jbc.org/

Dow

nloaded from

http://www.jbc.org/

FIGURE 1. Rotational and translational dynamics of PFV-1 IN. A, top, rotational correlation time (�) distribution of IN (gray, 2 �M; white, 125 �M) at20 °C. Fluorescence anisotropy decays were obtained by monitoring the intrinsic Trp fluorescence. One additional short correlation time (�400 ps) wasfound under both concentration conditions (not shown). Bottom, corresponding experimental fluorescence anisotropy decays (gray, 2 �M; black, 125�M) with line fits (black, 2 �M; white, 125 �M) resulting from maximum entropy method analysis (see “Experimental Procedures” for details). The lifetimedistributions of PFV-1 IN, recovered from intensity decays (not shown), were as follows: �1 � 0.28 � 0.03 ns (29%), �2 � 1.09 � 0.09 ns (28%), �3 � 2.6 �0.3 ns (20%), �4 � 6.65 � 0.25 ns (23%) for 2 �M and �1 � 0.24 � 0.01 ns (21%), �2 � 0.94 � 0.1 ns (31%), �3 � 2.4 � 0.2 ns (21%), �4 � 6.6 � 0.1 ns (27%)for 125 �M. B, long correlation time as a function of [IN]. C, analysis of PFV-1 IN by size exclusion chromatography. Experiments were performed at roomtemperature using a Superdex 200 column. Three concentrations of IN were loaded on the column. Only the monomeric species were detected using2 �M. An additional peak, corresponding to the dimeric form, appeared for the two highest IN concentrations (20 and 60 �M). No other species weredetected. V0 and Ve are the exclusion volume of the column and the elution volume of the protein, respectively. The calibration sample (black squares)is composed of globular proteins of known molecular mass (see “Experimental Procedures”). The circles indicate the positions of elution peaks forvarious samples of PFV-1 IN, in the presence of either 0.05 M (gray circles) or 0.5 M (white circles) NaCl. The concentration of the loaded sample is explicitlymentioned in the figure. D, SELDI analysis of trypsin digestion of the unlabeled tagged (top) and TAMRA-labeled (bottom) PFV-1 IN. The peptidesobtained after proteolysis were analyzed using a SEND-ID protein chip. Spectra are magnified in the region of interest, corresponding to the expectedmass for the modified C-terminal peptide (NELEPKPQQFM; molecular mass 1860.16 Da) for the TAMRA-labeled protein. The entire spectra did not showany other differences between the two profiles (data not shown). The TAMRA-labeled IN was characterized by two peaks (1860.2 and 1876.9 Da). The firstone is consistent with the expected theoretical value for the TAMRA-labeled C-terminal peptide (i.e. a single fluorophore was found as expected on theC terminus-specific tag), and the second one (�MW � 16.7 Da) corresponds to the same peptide containing an oxidized methionine. E, characterizationof TAMRA-labeled PVF-1 IN by FCS; autocorrelation analysis of a 100 nM INT solution. Inset, average diffusion time as a function of the total INconcentration. The INT concentration was constant (100 nM), and total IN concentration was varied using unlabeled IN.




http://ww

w.jbc.org/

Dow

nloaded from

http://www.jbc.org/

for monomeric PFV-1 IN, these values are consistent with amonomeric (Mo) and a dimeric (Di) form, respectively (Table1). �long increased continuously from25 to 53 ns as a function ofprotein concentration with the half-transition state occurringat 20�M (Fig. 1B). The intermediary �long values represent aver-age correlation times during the transition and the observedtransition reasonably corresponds to a Mo-Di equilibrium.This result was qualitatively confirmed by size exclusion chro-matography (Fig. 1C).These results significantly differ from those of HIV-1 IN

(Table 1). Under similar conditions (i.e. absence of detergent),HIV-1 IN was mainly tetrameric below 200 nM and stronglyaggregated above 200 nM (8, 26). This critical concentrationwas only 10-fold higher with detergent (7). MonomericHIV-1 IN (32 kDa) (characterized by �long � 19 ns (7)) wasobtained only at low submicromolar concentrations underdetergent conditions.The TGase-mediated TAMRA labeling of HIV-1 and PFV-1

INs (INT) was next performed to study their translational dif-fusion properties by FCS. This enzymatic labeling (27) occursspecifically on a Gln residue of the PKPQQFM peptide sub-strate (fused to the C-terminal IN end) and leads to IN beinglabeled at a single position (confirmed by mass spectrometryanalysis; Fig. 1D). HIV-1 INT did not display autocorrelationcurves that were satisfactory, due to the presence of brightspikes that originate in the presence of aggregates, regardless ofthe concentration used (data not shown). The number/fre-quency of bright spikes was slightly lower but remained signif-icant upon the addition of detergent, indicating that irreversi-ble aggregates were formed during the purification/labelingprocedure. Thus,HIV-1 INTwas not further used in the presentstudy. By contrast, PFV-1 INT resulted in satisfactory autocor-relation curves (Fig. 1E) (no bright spike was detected duringthe acquisition). The fit yielded a diffusion time (�D) of 565�s atsubmicromolar concentrations. The FCS acquisition wasrepeated using a constant concentration of INT and varying theconcentration of unlabeled IN up to 250 �M to study �D and itsconcentration dependence (Fig. 1E, inset). �D significantlyincreased from565 to 680�s in this concentration range (again,no bright spike was detected in the different mixtures oflabeled/unlabeled PFV-1 INs). The �D ratio was 1.2, close to thetheoretical ratio (1.3) for a Mo-Di transition. The FCS half-

transition concentration (30 �M) was consistent with thatobserved in TFA.We have previously shown that the HIV-1 IN DNA-binding

step is very slow (kon � 0.23 min�1) (28). We suggested thatthese slow kinetics originate in a prebinding transition fromhigher order multimeric states (tetramers and aggregates) tosmaller molecular species (Mo), whereas the DNA-bounddimeric form correlates with optimal 3�-P activity (21, 26, 28).Accordingly, Faure et al. (29) found that theMo�DNA precedesthe Di�DNA complexes by analyzing the time-dependent for-mation of LTR-cross-linked species. Interestingly, fluorescenceanisotropy shows that a rapid equilibrium is reached after mix-ing PFV-1 IN and DNA, as compared with the much slowerformation kinetics of HIV-1 IN�DNA complexes (Fig. 2), con-firming TFA and FCS data (lack of detection of large PFV-1 INmolecular species beforeDNAbinding). Therefore, the conver-sion of higher order multimers to Mo, responsible for the slowDNA-binding step, is a characteristic of HIV-1 IN. PFV-1,which is mainlymonomeric at lowmicromolar concentrations,is characterized by a rapid DNA-binding step.Study of PFV-1 INT-DNA Interaction by FRET Reveals a Pref-

erential DNA Binding for the 3�-End Processing Site—FRETexperiments were performed to get deeper insight into thepositioning of IN on the DNA substrate. DNA of various sizes(increasing from 21- to 300-mer) were Fl-labeled at the 5�-end,either at the processing end or at the opposite end (Fig. 3A). Fig.3B shows one example for the quenching donor (qD) plottedagainst PFV-1 INT concentration. Only the quenching donorwas measured (at 520 nm), not the acceptor sensitization (at580 nm), since the acceptor concentration was a varyingparameter along the titration. qD increased as [INT] increasedand reached a plateau corresponding to amaximumquenchingefficiency qD,max of 78% (the donor emission did not changeusing equivalent amounts of unlabeled IN). qD,max � 78% isconsistent with the theoretical value (at least 75%), assuming alabeling yield of 50% (see “Experimental Procedures”) and aminimal IN/DNA stoichiometry of 2:1 (see below). The qD,maxvalue was decreased continuously by decreasing the INT/unla-

TABLE 1Comparison of long correlation times obtained for PFV-1 andHIV-1 INs

IN Theoretical �longa

monomeric formExperimental �longmonomeric form

Experimental �longhigher order

oligomeric formsns ns ns

PFV-1 21.5 22.5–30b 47–53cHIV-1 15 18–20d 70–90 � 100e

a Assuming partial specific volume (v) of 0.73 ml/g and average hydration of 0.4g(H2O)/g(protein) (44, 45). �long is related to the volume of the rotating unit (V) by� � �V/kT� �Mv/kT, where� is the viscosity, k is the Bolztman constant,T is thetemperature (K), andM is the molecular weight.

b In the 1–5 �M range.c Monodisperse solution of dimeric species above 100 �M.d At submicromolar concentrations in the presence of detergent (7).e Mainly tetrameric (� � 70–90 ns) in the absence of detergent below 200 nM;aggregation (� 100 ns) occurs above 200 nM (8, 26).

FIGURE 2. Differential DNA-binding kinetics of HIV-1 and PFV-1 INs. Fl-labeled DNA that mimics either the HIV-1 or the PFV-1 U5 LTR extremity (dou-ble-stranded 21-mer, 4 nM) was mixed with HIV-1 (�) or PFV-1 IN (f), respec-tively, in buffer A � 50 mM NaCl. Protein concentrations are indicated in thefigure. Steady-state fluorescence anisotropy values (r) were measured at25 °C as described under “Experimental Procedures.” �r(t) � r(t) � rfree DNA.




http://ww

w.jbc.org/

Dow

nloaded from

http://www.jbc.org/

beled IN ratio (Fig. 3B), confirmingthat the decrease in the donor inten-sity was actually due to resonanceenergy transfer.We found that qD was depend-

ent on the DNA length (Fig. 3C);for a given IN concentration, qDwas systematically higher for shortDNAs. This result was obtainedalthough the apparent affinitydecreases as the DNA sizedecreases (Table 2), demonstratingthat the IN positioning onDNAwasthe main factor influencing theDNA size dependence of qD. Onlypositions of INT in close proximityto donor (�60–80 Å for theFl-TAMRA pair) can significantlyinfluence the energy transfer. Thecontinuous decrease in the qD valueobtained with DNA molecules ofgreater sizes demonstrates that INmay occupymany internal positionsin DNA, leading to a “pearl chain”structure. Such a polymerization onlong DNAs is consistent with theobservation of light scattering thatoccurs for only 100- and 300-mersat IN concentrations above 600 nM(data not shown).Interestingly, for 45-, 100-, and

300-mer DNA, qD was systemati-cally higher (for a given DNA size)when the Fl donor was attached onthe terminal processing site, up to2-fold higher for 300-mer (FRETresults are summarized in Fig. 3D).No such difference was observedwith the 21-mer. Our results indi-cate a significant preference for INbinding at the processing end. Thisis consistent with the absence ofbias for the short 21-mer substrates,where a difference in the quenchingdonor between the two ends is notexpected, since the dimension ofDNA is comparable with both theoverall dimension of IN and the For-ster distance.The Anisotropy-based Assay Con-

firms a Specific DNA-binding Modefor PFV-1 but Not for HIV-1 IN—We next assessed the specific DNA-binding of PFV-1 IN (suggested byFRET) by competition experiments.Unlabeled PFV-1 IN was incubatedwith Fl-labeledDNA, 45-mer PFV-1U5 (Fig. 4A), or 45-mer random

D21-mer CAAT-3’

5’3’5’

qD 0.6850.683

45-mer CAAT-3’5’3’

5’

qD 0.7020.585

100-mer CAAT-3’5’3’

5’

qD0.5030.270

300-mer CAAT-3’5’

qD0.2330.120

A CAAT5’3’

5’X-mer AFl5

Fl --Fl

CAAT5’3’

5’X-mer BFl5

3’ 3’

0

0,2

0,4

0,6

0,8

21-AFl5

45-AFl5

100-AFl5

21-BFl5

45-BFl5

100-BFl5

300-BFl5

300-AFl5

C

0.0 0.3 0.60.00

0.25

0.50

0.75

1.00

1.0

q D,m

ax

1:1

1:3

1:9

Don

orqu

ench

ing

(qD)

[PFV-1 IN], µM

0

250

500

750

1000

510 550 590 630Fluo

rece

nce

Inte

nsity

Wavelength, nm

DNAalone

1:1

1:9

1:3

ID,[INT]

Don

orqu

ench

ing

(qD)

B

FIGURE 3. FRET between DNA substrates and PFV-1 IN. FRET between Fl-labeled DNA substrates (donor) ofvarious lengths and INT (acceptor) was measured as a function of [INT], as described under “Experimental Proce-dures.” A, the different DNA substrates used in the FRET study. The Fl donor was attached at the 5�-end of either theA or B strand, giving X-mer AFl5 or BFl5, respectively (X�21, 45, 100, or 300). B, quenching donor (qD) between 21-merAFl5 and INT (measured at 520 nm). The plateau value depends on the labeled/unlabeled IN ratio (indicated on theplot). Right, emission spectra for the various mixtures of labeled and unlabeled IN (total concentration, 1 �M) in thepresence of 20 nM 21-mer AFl5. C, comparison of qD values between the different DNA substrates. [INT] � 100 (whitesquares), 200 (gray squares), or 300 nM (black squares). D, summary of the differential quenching amplitude betweenthe two DNA ends as a function of the DNA size for 200 nM INT. All FRET experiments were performed in 20 mM Hepes,pH 7.5, containing 1 mM dithiothreitol, 10 mM MgCl2, and 50 mM NaCl.




http://ww

w.jbc.org/

Dow

nloaded from

http://www.jbc.org/

sequence (Fig. 4B) in the presence of increasing concentrationsof unlabeled DNA (45-mer PFV-1 U5 or random sequence).The decrease in the anisotropy value was then recorded. Usingthe cognate Fl-labeled DNA, unlabeled cognate and randomsequences competed with different IC50 values (1.4 and 2.5�M, respectively) (Fig. 4A). Consistently, the competition waseasier when the Fl-labeled DNAwas random, with correspond-ing values of 0.5 and 1.6�M, respectively (Fig. 4B). These resultsindicate that PFV-1 IN has a higher apparent affinity for thecognate sequence. The IC50 value found for the PFV-1 IN�45-mer U5 complex (1.4 �M at 50 mM NaCl) was used to estimatethe apparentKd value (Kd,app) according to Refs. 30 and 31) (the

Cheng-Prusoff relationship was not used here because [IN] [DNA substrate]). The calculated Kd,app from the competitionassay (0.13 �M) was then compatible with the correspondingKd,app obtained by direct titration measurement (0.1 �M inTable 2), indicating no influence or only a slight influence of theFl moiety on the DNA binding properties of IN.Preferential/specific DNA-binding of HIV-1 IN is generally

difficult to assess in vitro (10, 12, 32) despite the fact that (i)specific and nonspecific complexes are distinct in term ofradius of gyration (33), and (ii) several residues (Gln148, Tyr143,Lys156, Lys159, Lys14) appear to be involved in such specific con-tacts before and/or after the 3�-P reaction (11, 32, 34–36). Inparticular, it was recently shown that Gln148 is also involved inmaintaining the complex stability for the subsequent strandtransfer reaction by specifically contacting the 5�-LTR end ofthe donor (viral) DNA (34, 35). Competition experiments wererepeated with HIV-1 IN and confirm that no significant differ-ence can be observed between the cognate and a random45-mer DNA sequence, irrespective of the Fl-labeled DNAsequence (Fig. 4, C and D). Moreover, HIV-1 and PFV-1 INsdisplay different apparent affinities for their respective DNAsubstrate (Fig. 4, compare A and B with C and D). For 45-merDNA, the specificity and the absence of specificity for PFV-1and HIV-1, respectively, were obtained regardless of the ionicstrength (50 or 150 mM NaCl). For 21-mer DNA, PFV-1 INdisplayed differential DNA binding between cognate and ran-dom sequences at 150 mM NaCl (two different randomsequences were tested and gave similar results), whereas nospecific DNA binding was observed at 50 mM NaCl (data notshown). Again, no specificitywas observedwithHIV-1 INusing21-mer DNA at either salt concentration (data not shown).Comparative Study of Catalytic Properties of HIV-1 and

PFV-1 INs—Taking into account that solubility andDNAbind-ing properties of HIV-1 and PFV-1 INs were significantly dif-ferent, we next compared catalytic features of both enzymesusing a fluorescence 3�-P assay (21). Briefly, this assay is basedon steady-state anisotropy (r) and allows the separation ofDNA-binding and catalytic parameters of 3�-P reaction aswell as the study of real time kinetics. A simultaneous quan-titative analysis of DNA binding and dinucleotide release ispossible, since both steps strongly influence the molecularsize of the fluorescent moiety when Fl is linked to thereleased dinucleotide.For HIV-1, we have previously found that the r value, as

obtained after DNA-binding and before catalysis (rt � 0), wasfully predictive of the subsequent activity at low IN/DNAratios,according to the fractional saturation function (21). By con-trast, for higher ratios, r continued to increase but the 3�-Pactivity significantly decreased. The variable r is dependent onboth the fractional DNA saturation and the molecular size ofthe complex. Therefore, higher order INmultimers/aggregatesare detrimental to 3�-P activity and account for the character-istic bell-shaped curves when plotting activity versus IN con-centration or the rt � 0 parameter.We obtained a similar behav-ior with PFV-1 IN using the cognate 21-mer PFV-CAATsubstrate when the 3�-P activity was plotted against IN concen-tration (Fig. 5A) or rt � 0 (Fig. 5B). This suggests that, althoughHIV-1 and PFV-1 INs, free in solution, are characterized by

TABLE 2Apparent Kd values of PFV-1 IN as a function of DNA size and ionicstrengthFl-labeled ODNs of various sizes mimicking the cognate U5 DNA end (12 nM) weremixed with increasing concentrations of unlabeled PFV-1 IN in the presence of 10mM Mg2�. DNA binding was measured by recording the steady-state anisotropyvalue as described under “Experimental Procedures.”

DNA size �NaCl� Kd,app

bp mM �M

15 50 0.56150 3.6200 NDa

21 50 0.13150 2.1200 3.5

45 50 0.1150 0.4200 1.6

65 50 0.06150 0.39200 0.97

100 50 0.05150 0.29200 ND

300 50 0.06150 0.37200 ND

a ND, not determined.

B

0 750 1500 22500.04

0.08

0.12

0.16

0.20

r

[unlabeled DNA], nM

C

A

0 750 1500 2250

0.04

0.08

0.12

0.16

r

[unlabeled D NA], nM

D

0 150 300 450

0.04

0.08

0.12

0.16

0.20

r

[unlabeled DNA], nM0 150 300 450

0.00

0.04

0.08

0.12

0.16

[unlabeled DNA], nM

r

FIGURE 4. Comparison of HIV-1 and PFV-1 IN DNA binding properties. IN(A and B, PFV-1; C and D, HIV-1) was incubated with 4 nM Fl-labeled 45-merDNA (A and C, cognate sequence; B and D, random sequence) and varyingconcentrations of unlabeled 45-mer DNA competitor (f, cognate sequence;E, random sequence) in buffer A � 50 mM NaCl (See “Experimental Proce-dures” for ODN sequences). Fluorescence anisotropy was measured asdescribed under “Experimental Procedures.” �r � rcomplex � rfree DNA. [IN] �700 nM (PFV-1) or 400 nM (HIV-1).




http://ww

w.jbc.org/

Dow

nloaded from

http://www.jbc.org/

distinct self-association states, they display comparable enzy-matic features for the 3�-P catalytic reaction. HIV-1 IN waspreviously shown to work very slowly as a single-turnoverenzyme for 3�-P (rate constant � 0.004 min�1) (28). Here,PFV-1 IN possesses similar slow single-turnover kinetics (Fig.5D), suggesting an intrinsic single-turnover property of retro-viral INs that is clearly independent of the initial aggregation

state. Interestingly, the 3�-P activity at the top of the bell-shapedcurve (Actmax) was significantly better (from 50 to 75%) withgreater ionic strength (from 50 to 200 mM NaCl) (Fig. 5A),although more IN was required to obtain optimal activity,because the overall IN�DNA affinity is weakened by ionicstrength (Table 2). This result suggests that large complexes,responsible for the activity drop off, were more sensitive toionic strength than catalytically competent complexes.Furthermore, the 3�-P activity is better at higher ionic

strength for a given rt � 0 value in the decreasing phase (Fig. 5B),confirming that rt � 0 accounts for both the number of 3�-P-competent complexes and the less active large complexes. Theenhanced activity in high salt concentration at the optimal rt � 0value confirms that ionic strength differentially affects the twotypes of complexes and delays the formation of largecomplexes.We obtained bell-shaped curves for all DNA lengths tested

(Fig. 5C). However, we found that the optimal activity (Actmax)was strongly length-dependent, and better 3�-P activity wasobtained using shorter ODNs (3-fold higher for 15-mer thanthat of 65-mer) (confirmed by complete kinetics study in Fig.5D). Again, Actmax did not occur at similar IN concentrations.The optimal concentration depends on DNA size, and this isconsistent with apparent Kd values showing a better IN affinityfor long DNAs (Table 2). Interestingly, the activities in thedecreasing phase reached a plateau (Actplat) (Fig. 5C). Thedecreasing phase was more pronounced for longer DNA sub-strates, as evidencedwhen plotting the Actmax/Actplat ratio ver-sus DNA size (Fig. 5E). These results indicate that shorteningthe DNA substrate improves the number of catalytically com-petent complexes over the number of less active large com-plexes. The stimulating effect of ionic strength on Actmax (asdescribed in Fig. 5A for the 21-mer) was reproduced with the15-mer DNA substrate (Actmax � 90% at 200 mM NaCl; datanot shown). Surprisingly, no such effect was observed for thelonger 45- and 65-mer DNAs (data not shown). This result willbe further discussed below.Fluorescence of Fl was next used for the TFA study of the

hydrodynamics of PFV-1 IN�DNA complexes using 15- and21-mer Fl-labeled ODNs (Table 3). At the optimal IN concen-tration for activity (Actmax), �long values were consistent withdimeric forms of PFV-1 IN bound to short DNAs (53–63 ns).Above this concentration, we detected large complexes (�long 100 ns), consistent with the bell-shaped curves shown in Fig.5A. Therefore, the dimeric form appears to be themost catalyt-ically competent form for 3�-P, and higher order multimers or

FIGURE 5. Influences of DNA size and ionic strength on the PFV-1 IN 3�-Pactivity. A and B, ionic strength effect. 3�-P activity was determined using the21-mer DNA substrate as a function of [IN] (A) or the �rt � 0 value (B). [NaCl] �50 mM (f), 150 mM (E), or 200 mM (F). �rt � 0 � rt � 0 � rfree DNA. rt � 0 repre-sents the r value after the DNA-binding step (before the start of the reaction).C, DNA size effect. The response of the 3�-P activity to IN concentration wasstudied with 15-mer (E), 21-mer (f), 45-mer (�), and 65-mer (Œ), mimickingthe U5 LTR extremity. [NaCl] � 50 mM. Activities reported on the y axis (A–C)correspond to a 420-min incubation with 12 nM Fl-labeled DNA at 37 °C andwere calculated as described under “Experimental Procedures.” D, timecourse of 3�-P for 15– 65-mer (symbols are as in C). Experimental conditionscorrespond to optimal conditions in C for each DNA size. E, Actmax/Actplat ratioas a function of the DNA length. Actmax and Actplat are defined under “Results”and in C.

TABLE 3Long correlation times of PFV-1 IN�DNA complexesTFA experiments were performed with 12 nM DNA in buffer A � 50 mM NaCl at20 °C, as described under “Experimental Procedures.”

DNA size �PFV-1 IN� �longa

bp nM ns15 1200 63 � 621 400 53 � 821 600 61 � 521 1200 100

a The correlation time distributions of freeODNs as a function of theDNA sizewerepreviously published (21). �long � 6 and 9.5 ns at 20 °C for freeODN15-mer and freeODN21-mer, respectively.




http://ww

w.jbc.org/

Dow

nloaded from

http://www.jbc.org/

aggregates, obtained for high IN/DNA ratio, led to suboptimalactivity. These results strongly parallel those previouslyobtained with HIV-1 (21).

DISCUSSION

Weused various biophysical techniques to address the prob-lem of IN solubility and found that HIV-1 and PFV-1 INs arecharacterized by different solubility properties. PFV-1 IN, freein solution, was found to bemore soluble thanHIV-1 IN.More-over, fluorescence anisotropy and FRET experiments highlightsequence-specific IN�DNA recognition that is measurable onlywith PFV-1 IN. However, most of the enzymatic features aresimilar for both types of INs (slow single turnover kinetics,dimeric form responsible for optimal 3�-P). In addition,althoughmore soluble, PFV-1 IN can lead to higher order mul-timers or aggregates when bound to its DNA substrate underconditions of high IN/DNA ratio (leading to suboptimal activ-ity), as previously found for HIV-1 IN (21). Finally, DNA sizeand ionic strength were found to be important parameters thatmodulate the number of higher order complexes. In addition,we have demonstrated the feasibility of the enzymatic labelingof a retroviral integrase with a low molecular weight fluoro-phore. In contrast to conventional chemical labeling methods,such a strategy leads to a site-specific labeling of the protein ofinterest, avoiding multiply labeled proteins and heterogeneoussamples, and then is particularly well suited for FRET and FCSstudies. Regrettably, bright spikes were detected by FCS withHIV-1 INT, suggesting the presence of aggregates, whereasPFV-1 INT resulted in satisfactory autocorrelation curves withno bright spike detected, reinforcing the idea that HIV-1 IN hasa higher propensity for aggregation than PFV-1 IN.In previous studies, HIV-1 IN was found mainly tetrameric,

below 200 nM in the absence of detergent, and significant pro-tein aggregationwas observed above 200 nM (8, 26). In the pres-ence of detergent, this critical concentration was found in thelow micromolar range. Here, PFV-1 IN was found mainlymonomeric in the lowmicromolar range, in the absence of anydetergent and DNA. Using PFV-1 IN, either unlabeled (in sizeexclusion chromatography and TFA experiments) or TGase-mediated TAMRA-labeled (in FCS experiments), we haveclearly identified aMo7Di equilibrium for the protein, free insolution. TFA and FCS gave approximately the same half-tran-sition concentration, 20–30 �M. PFV-1 IN is therefore charac-terized by a Mo 7 Di equilibrium, and aggregation does notsignificantly occur in the absence of DNAduring the transition.The reason for better solubility thanHIV-1 IN remains unclear,since no evident differences appear when comparing the globalhydrophobic profiles of both INs (not shown). Mutating only afew surface residues significantly improves HIV-1 IN solubility.For instance, the singlemutation F185K improves the solubilityof the entire protein (6). It is then possible that only a smallnumber of key residues may account for enhanced PFV-1 INsolubility. Due to the poor sequence similarity between bothINs (�15%), a greater understanding of solubility determinantsrequires further structural studies.Interestingly, using the TAMRA-labeled PFV-1 IN and fluo-

rescein-labeled DNA substrates, we found that FRET was sys-tematically more efficient when fluorescein was attached at the

processing as compared with the nonprocessing end. This dif-ference was observed only with long DNA substrates (from 45-to 300-mer) and not with a short 21-mer DNA substrate, withsize comparable with the Forster distance. This result high-lights a preference of DNA binding for the processing end andwas confirmed by competitive DNA-binding experiments.Altogether, these results suggest that, using PFV-1 IN, it is pos-sible to distinguish between specific and nonspecific com-plexes, not only at the catalytic level (as typically found forHIV-1 (11, 21)) but also at the DNA binding level. In our opin-ion, these results do not reveal a differential DNAbinding prop-erty of both INs but rather, most likely, that the aggregativeproperties and high propensity of HIV-1 IN to establish non-specific contacts mask in vitro an intrinsic ability of the proteinto bind preferentially to its cognate sequence. However, FRETand competition experiments suggest that the specificity,although significant, is modest for PFV-1 IN. This could be dueto the inherent nonspecific DNA-binding mode (responsiblefor the binding to the target DNA) that can be mainly ascribedto the C-terminal domain, as for the HIV-1 protein (12),although we cannot exclude the possibility that this property isalso mediated by the N-terminal region in the case of PFV-1 IN(18). Furthermore, the catalytic domain, which ensures 3�-Pand the subsequent joining reaction into a large variety of targetDNA sequences, should be also in part responsible for nonspe-cific DNA-binding.FRET analysis, as a function of DNA size, also suggests a

nonspecific DNA binding mode at internal positions onto longDNA substrates. In addition, we found that the 3�-P catalyticprocess was stimulated by (i) shortening the DNA size and (ii)increasing the ionic strength. Nevertheless, the stimulatingeffect of ionic strength was not observed for long DNA sub-strates such as 45- and 65-mer. This result, together with theFRET data, suggests the existence of two types of higher ordercomplexes. IN polymerization, mediated by nonspecific inter-nal IN-DNA interactions, is the main phenomenon occurringon long DNAs and is strongly detrimental to 3�-P activity. Con-sistently, the Actmax/Actplat ratio (as defined in Fig. 5) is lowerfor 15- and 21-mer DNAs, since no polymerization is expectedto occur on short DNAs for evident steric reasons. However,their slight decreasing phases may be assigned to the presenceof aggregates, mediated by nonspecific IN-IN interactions athigh IN/DNA ratios. Only a differential effect of high salt con-centration on nonspecific IN-DNAand IN-IN interactionsmayexplain why ionic strength only stimulates Actmax of shortDNAs. Ionic strength disrupts more efficiently nonspecificIN-IN interactions than nonspecific IN-DNA interactions.This also explains why the specific interaction between PFV-1IN and the short 21-mer DNA was only evidenced under highsalt conditions (see above), since aggregates on DNA underes-timate detection of specific complexes at low ionic strength.Altogether, our results show that two types of large complexesare present at high IN/DNA ratios, and both are detrimental forIN activity. The polymerization process is mainly influenced bythe DNA length, whereas aggregation is more sensitive to theionic strength. Consequently, an additive/synergic positiveeffect on the IN activity is observed by combining both param-




http://ww

w.jbc.org/

Dow

nloaded from

http://www.jbc.org/

eters (i.e. shortening the DNA length and increasing the ionicstrength).FRET analysis indicates that IN on long DNAs can be posi-

tioned at internal DNA sites due to its nonspecific DNA-bind-ing mode, decreasing the 3�-P efficiency. The DNA lengthdependence of 3�-P activity raises the question of the particu-larly low 3�-P efficiency, which can be expected for long DNAs,such as physiological viral DNAs (about 9 and 13 kbp long forHIV-1 and PFV-1, respectively). To date, the mechanism bywhich IN is specifically positioned at the processing site in vivois still unclear. Although IN prefers to bind to its cognate DNAsequence above other sequences, the relative specificity of INdescribed in this study may not be sufficient to ensure its selec-tive in vivo positioning at the processing ends. It has been pre-viously proposed forHIV-1 that the nucleocapsid protein bindstightly to viral DNA, except at theDNA ends (37). Additionally,this protein stimulates IN activity in vitro (38) and may indi-rectly assist the correct positioning of IN onto the processingends. It is important to note that a nucleocapsid, characterizedby a typical two canonical cysteine-histidinemotifs, is not pres-ent in PFV-1, although the PFV-1 C terminus region of Gagcontains three glycine-argininemotifs called “GRboxes,”whichare supposed to display similar properties (39, 40).The PFV-1 IN�DNA complexes, as obtained under optimal

conditions for 3�-P activity (Actmax in Fig. 5), were further stud-ied by time-resolved fluorescence anisotropy. Using shortDNAsubstrates (15- and 21-mer), long rotational correlation timeswere found to be compatible with dimeric forms (Table 3), sug-gesting that Di is the most catalytically competent form for3�-P, as was previously found for HIV-1 IN (21, 29), whereaslarge complexes are characterized by suboptimal 3�-P activity.DNA primarily promotes IN dimerization, since Di are presentat concentrations in which free PFV-1 IN is monomeric (lowmicromolar range).Most likely, a tetramer (Di ofDi) is requiredfor integration, as shown for HIV-1 when two LTR ends aresimultaneously present in the context of the synaptic complex(41) or when IN is bound to a three-way junction DNA sub-strate mimicking an integration intermediate (42). The abilityof the tetrameric species to perform 3�-P is still controversial(29, 43), and the present data cannot clearly address the proper3�-P activity of the Te, since the decreasing phase is certainlyrelated to an heterogeneous mixture of large complexes,including tetrameric forms. Recently, a kinetic study of ASVMn2�-dependent 3�-P suggests that a Di 7 Te equilibriumaccounts for the biphasic behavior of the single-turnover 3�-P,with Te responsible for the fast phase (43). We found amonophasic behavior for both PFV-1 (this study; Fig. 5D) andHIV-1 (21) Mg2�-bound INs, with rate constants compatiblewith the slow ASV phase. Apparent discrepancies betweenHIV-1, PFV-1, and ASVmay reflect intrinsic or metal-depend-ent differences for Di7Te or/and Te7 aggregates transitionsontoDNA.However, HIV-1 and PFV-1 INs, which display sim-ilar monophasic behaviors, are characterized by different sin-gle-turnover rate constants (for a given DNA size, i.e. 21-merDNA substrate at 50 mM NaCl, kHIV-1 � 0.004 min�1 (21) andkPFV-1 � 0.0028 min�1 (this study)). Apparent affinities forDNAsubstratewere also significantly different (Kd,app� 40 and130 nM for HIV-1 (21) and PFV-1 (Table 2), respectively, for

21-mer DNA substrate at 50 mM NaCl). Although most of theenzymatic properties described here are qualitatively similarbetween HIV-1 and PFV-1 INs, the two enzymes display sub-stantial quantitative differences; PFV-1 IN has lower affinity forthe DNA substrate and lower catalytic efficiency as comparedwith HIV-1 IN; the reasons for these differences remain to beelucidated. Although the sequence homology between HIV-1and PFV-1 INs is low, our data suggest that both proteins,which display similar enzymatic properties, at least qualita-tively, have structural homology. Considering its solubilitycharacteristics, PFV-1 IN could therefore represent a goodmodel for further structural studies of retroviral INs.

REFERENCES1. Sinha, S., and Grandgenett, D. P. (2005) J. Virol. 79, 8208–82162. Chen, J. C., Krucinski, J., Miercke, L. J., Finer-Moore, J. S., Tang, A. H.,

Leavitt, A. D., and Stroud, R. M. (2000) Proc. Natl. Acad. Sci. U. S. A. 97,8233–8238

3. Wang, J. Y., Ling, H., Yang, W., and Craigie, R. (2001) EMBO J. 20,7333–7343

4. Goldgur, Y., Dyda, F., Hickman, A. B., Jenkins, T. M., Craigie, R., andDavies, D. R. (1998) Proc. Natl. Acad. Sci. U. S. A. 95, 9150–9154

5. Cherepanov, P., Ambrosio, A. L., Rahman, S., Ellenberger, T., and En-gelman, A. (2005) Proc. Natl. Acad. Sci. U. S. A. 102, 17308–17313

6. Jenkins, T. M., Engelman, A., Ghirlando, R., and Craigie, R. (1996) J. Biol.Chem. 271, 7712–7718

7. Deprez, E., Tauc, P., Leh, H., Mouscadet, J. F., Auclair, C., and Brochon,J. C. (2000) Biochemistry 39, 9275–9284

8. Leh, H., Brodin, P., Bischerour, J., Deprez, E., Tauc, P., Brochon, J. C.,LeCam, E., Coulaud, D., Auclair, C., andMouscadet, J. F. (2000) Biochem-istry 39, 9285–9294

9. Marchand, C., Johnson, A. A., Karki, R. G., Pais, G. C., Zhang, X., Cowan-sage, K., Patel, T. A., Nicklaus, M. C., Burke, T. R., Jr., and Pommier, Y.(2003)Mol. Pharmacol. 64, 600–609

10. Agapkina, J., Smolov, M., Barbe, S., Zubin, E., Zatsepin, T., Deprez, E., LeBret, M., Mouscadet, J. F., and Gottikh, M. (2006) J. Biol. Chem. 281,11530–11540

11. Esposito, D., and Craigie, R. (1998) EMBO J. 17, 5832–584312. Engelman, A., Hickman, A. B., and Craigie, R. (1994) J. Virol. 68,

5911–591713. Moebes, A., Enssle, J., Bieniasz, P. D., Heinkelein, M., Lindemann, D.,

Bock,M.,McClure,M.O., andRethwilm,A. (1997) J. Virol. 71, 7305–731114. Yu, S. F., Sullivan, M. D., and Linial, M. L. (1999) J. Virol. 73, 1565–157215. Cherepanov, P. (2007) Nucleic Acids Res. 35, 113–12416. Enssle, J., Moebes, A., Heinkelein, M., Panhuysen, M., Mauer, B., Schwei-

zer, M., Neumann-Haefelin, D., and Rethwilm, A. (1999) J. Gen. Virol. 80,1445–1452

17. Pahl, A., and Flugel, R. M. (1995) J. Biol. Chem. 270, 2957–296618. Lee, H. S., Kang, S. Y., and Shin, C. G. (2005)Mol. Cells 19, 246–25519. Oh, Y. T., and Shin, C. G. (1999) Biochem. Mol. Biol. Int. 47, 621–62920. Delelis, O., Petit, C., Leh, H., Mbemba, G., Mouscadet, J. F., and Sonigo, P.

(2005) Retrovirology 2, 31–4921. Guiot, E., Carayon, K., Delelis, O., Simon, F., Tauc, P., Zubin, E., Gottikh,

M., Mouscadet, J. F., Brochon, J. C., and Deprez, E. (2006) J. Biol. Chem.281, 22707–22719

22. Laboulais, C., Deprez, E., Leh, H., Mouscadet, J. F., Brochon, J. C., and LeBret, M. (2001) Biophys. J. 81, 473–489

23. Brochon, J. C. (1994)Methods Enzymol. 240, 262–31124. Livesey, A. K., and Brochon, J. C. (1987) Biophys. J. 52, 693–70625. Lakowics, J. (2006) in Principles of Fluorescence Spectroscopy, 3rd Ed.,

Springer-Verlag New York Inc., New York26. Deprez, E., Tauc, P., Leh, H., Mouscadet, J. F., Auclair, C., Hawkins, M. E.,

and Brochon, J. C. (2001) Proc. Natl. Acad. Sci. U. S. A. 98, 10090–1009527. Taki, M., Shiota, M., and Taira, K. (2004) Protein Eng. Des. Sel. 17,

119–12628. Smolov, M., Gottikh, M., Tashlitskii, V., Korolev, S., Demidyuk, I., Bro-




http://ww

w.jbc.org/

Dow

nloaded from

http://www.jbc.org/

chon, J. C.,Mouscadet, J. F., andDeprez, E. (2006) FEBS J. 273, 1137–115129. Faure, A., Calmels, C., Desjobert, C., Castroviejo,M., Caumont-Sarcos, A.,

Tarrago-Litvak, L., Litvak, S., and Parissi, V. (2005) Nucleic Acids Res. 33,977–986

30. Linden, J. (1982) J. Cyclic Nucleotide Res. 8, 163–17231. Deprez, E., Barbe, S., Kolaski,M., Leh,H., Zouhiri, F., Auclair, C., Brochon,

J. C., Le Bret, M., and Mouscadet, J. F. (2004)Mol. Pharmacol. 65, 85–9832. Jenkins, T. M., Esposito, D., Engelman, A., and Craigie, R. (1997) EMBO J.

16, 6849–685933. Baranova, S., Tuzikov, F. V., Zakharova, O. D., Tuzikova, N. A., Calmels,

C., Litvak, S., Tarrago-Litvak, L., Parissi, V., and Nevinsky, G. A. (2007)Nucleic Acids Res. 35, 975–987

34. Johnson, A. A., Santos, W., Pais, G. C., Marchand, C., Amin, R., Burke,T. R., Jr., Verdine, G., and Pommier, Y. (2006) J. Biol. Chem. 281, 461–467

35. Dicker, I. B., Samanta, H. K., Li, Z., Hong, Y., Tian, Y., Banville, J., Remil-lard, R. R., Walker, M. A., Langley, D. R., and Krystal, M. (2007) J. Biol.Chem. 282, 31186–31196

36. Zhao, Z., McKee, C. J., Kessl, J. J., Santos,W. L., Daigle, J. E., Engelman, A.,Verdine, G., and Kvaratskhelia, M. (2008) J. Biol. Chem. 283, 5632–5641

37. Lapadat-Tapolsky, M., de Rocquigny, H., Van Gent, D., Roques, B., Plas-terk, R., and Darlix, J. L. (1993) Nucleic Acids Res. 21, 831–839

38. Carteau, S., Batson, S. C., Poljak, L., Mouscadet, J. F., de Rocquigny, H.,Darlix, J. L., Roques, B. P., Kas, E., and Auclair, C. (1997) J. Virol. 71,6225–6229

39. Delelis, O., Lehmann-Che, J., and Saib, A. (2004)Curr. Opin. Microbiol. 7,400–406

40. Yu, S. F., Baldwin, D. N., Gwynn, S. R., Yendapalli, S., and Linial, M. L.(1996) Science 271, 1579–1582

41. Li, M., Mizuuchi, M., Burke, T. R., Jr., and Craigie, R. (2006) EMBO J. 25,1295–1304

42. Ren, G., Gao, K., Bushman, F. D., and Yeager, M. (2007) J. Mol. Biol. 366,286–294

43. Bosserman, M. A., O’Quinn, D. F., and Wong, I. (2007) Biochemistry 46,11231–11239

44. Zhou, X. Z. (1995) Biophys. J. 69, 2298–230345. Harding, S. E., Horton, J. C., Jones, S., Thornton, J. M., and Winzor, D. J.

(1999) Biophys. J. 76, 2432–2438




http://ww

w.jbc.org/

Dow

nloaded from

http://www.jbc.org/

Brochon, Jean-François Mouscadet and Eric DeprezOlivier Delelis, Kevin Carayon, Elvire Guiot, Hervé Leh, Patrick Tauc, Jean-Claude

INTEGRASEDNA-BINDING MODE REVEALED BY AN ENZYMATICALLY LABELED

Insight into the Integrase-DNA Recognition Mechanism: A SPECIFIC

doi: 10.1074/jbc.M803257200 originally published online August 12, 20082008, 283:27838-27849.J. Biol. Chem.

10.1074/jbc.M803257200Access the most updated version of this article at doi:

Alerts:

When a correction for this article is posted•

When this article is cited•

to choose from all of JBC's e-mail alertsClick here

http://www.jbc.org/content/283/41/27838.full.html#ref-list-1

This article cites 44 references, 22 of which can be accessed free at


http://ww

w.jbc.org/

Dow

nloaded from

http://www.jbc.org/lookup/doi/10.1074/jbc.M803257200

http://www.jbc.org/cgi/alerts?alertType=citedby&addAlert=cited_by&cited_by_criteria_resid=jbc;283/41/27838&saveAlert=no&return-type=article&return_url=http://www.jbc.org/content/283/41/27838

http://www.jbc.org/cgi/alerts?alertType=correction&addAlert=correction&correction_criteria_value=283/41/27838&saveAlert=no&return-type=article&return_url=http://www.jbc.org/content/283/41/27838

http://www.jbc.org/cgi/alerts/etoc

http://www.jbc.org/content/283/41/27838.full.html#ref-list-1

http://www.jbc.org/

insightintotheintegrase-dnarecognitionmechanism · to get deeper insight into the in/dna...

Documents