LEF-1 recognition of platinated GG sequences within double-stranded DNA. Influence of flanking bases

Available online at www.sciencedirect.com JOURNAL OF

Inorganic Biochemistry Journal of Inorganic Biochemistry 102 (2008) 242–250 www.elsevier.com/locate/jinorgbio

LEF-1 recognition of platinated GG sequences within double-stranded DNA. Influence of flanking bases Katerˇina Chva´lova´, Marie-Agne`s Sari, Sophie Bombard *, Jirˇ´ı Kozelka

*

Laboratoire de Chimie et Biochimie Pharmacologiques et Toxicologiques, Universite´ Paris Descartes, UMR 8601 CNRS, 45, rue des Saints-Pe`res, 75270 Paris, France Received 2 April 2007; received in revised form 3 August 2007; accepted 23 August 2007 Available online 5 September 2007

Abstract The lymphoid enhancer-binding factor 1 (LEF-1) recognizes a double-stranded 9 base-pairs (bp) long motif in DNA which is significantly bent upon binding. This bend is centered at two destacked adenines whose geometry closely resembles that of two adjacent guanines crosslinked by the antitumor drug cisplatin. It has been proposed that cisplatin–GG crosslinks could hijack high mobility group (HMG) box containing transcription factors such as LEF-1. In order to examine such a possibility, we used electrophoretic mobility shift assays to determine the affinity of the HMG box of LEF-1 for a series of 25 oligonucleotides containing a central GG sequence, free or site-specifically modified by cisplatin. The binding affinity of the GG-platinated oligonucleotides was 3–6-fold higher than that determined for the corresponding unplatinated oligonucleotides, however, the binding to all cisplatin-modified oligonucleotides was at least 1 order of magnitude weaker than that to the 25 bp oligonucleotide containing the recognition 9 bp motif. The binding affinity was dependent on the nature of bases flanking the cisplatin-crosslinked G*G* dinucleotide, the AG*G*T sequence displaying the strongest affinity and CG*G*T showing the strongest binding enhancement upon platination. In contrast, modification of the AGGT sequence with the third-generation platinum antitumor drug oxaliplatin did not enhance the affinity significantly. These results suggest that the cisplatin-caused bending of DNA does produce a target for LEF-1 binding, however, the cisplatinated DNA does not appear to be a strong competitor for the LEF-1 recognition sequence.  2007 Elsevier Inc. All rights reserved. Keywords: Anticancer drugs; Cisplatin; DNA binding; HMGB proteins

1. Introduction High mobility group box (HMGB) proteins contain one or more ‘‘HMG boxes’’, sequences of approximately 80 amino acids having a typical L-shaped structure [1]. This structure varies little from one HMG box to another in spite of relatively weak sequence-homology and serves to recognize specific DNA structures (structure-specific HMGB proteins) or sequences (sequence-specific HMGB proteins). The sequence-specific HMGB proteins recognize sequences of several base-pairs (bp) and include transcrip* Corresponding authors. Tel.: +33 1 4286 2086; fax: +33 1 4286 8387 (J. Kozelka). E-mail addresses: [email protected] (S. Bombard), [email protected] (J. Kozelka).

0162-0134/$ - see front matter  2007 Elsevier Inc. All rights reserved. doi:10.1016/j.jinorgbio.2007.08.006

tion factors such as lymphoid enhancer-binding factors LEF-1 and TCF-1, the sex-determining factor SRY and related Sox proteins, and fungal regulatory proteins MatMc, Mat-a1 and Rox1 [2]. The structure-specific HMGB proteins recognize DNA structures such as four-way junctions, cruciforms, supercoiled DNA and their presumed native function is to play an architectural role as DNA chaperones [2]. In 1992 Bruhn et al. have reported that the chromosomal protein HMGB1 binds specifically to DNA crosslinked by the antitumor drug cisplatin [3]. Cisplatin binds predominantly to GG sequences of DNA inducing bending towards the major groove and local unwinding of the double-helix [4]. This distortion is recognized by structure-specific HMGB proteins such as HMGB1, as exemplified in the crystal structure of the complex formed between a 16 bp DNA bearing a cisplatin–GG

K. Chva´lova´ et al. / Journal of Inorganic Biochemistry 102 (2008) 242–250

crosslink and the domain A of HMGB1 [5]. Interestingly, sequence-specific HMGB proteins such as LEF-1, SRY or UBF have also been reported to bind specifically to cisplatin–GG adducts of DNA [6–8]. The question arises whether these protein complexes with platinated DNA could combine recognition of structure (bent DNA) and sequence. We therefore decided to study the recognition between a sequence-specific HMGB protein and DNA bearing a cisplatin–GG crosslink, and to investigate in particular the influence of the neighboring bases on this recognition. We chose the HMG box of the lymphoid enhancer factor 1, LEF-1, whose binding to its cognate DNA sequence, d(CTTCAAAGG)–d(CCTTTGAAG), has been studied previously by NMR [9,10] and molecular dynamics simulations [11]. The recognition complex shows destacking of two adjacent adenines and bending towards the major groove strongly reminiscent of the bending and destacking seen in the cisplatin–DNA structure (see NMR work on cisplatin–oligonucleotide adducts reviewed by Marzilli et al. [12]). The hypothesis seemed therefore plausible that cisplatin-modified DNA could represent a target for LEF-1 to which the protein might bind with high affinity. We considered it thus interesting to investigate whether the DNA pre-bent with cisplatin could form a stable complex with the LEF-1 HMG domain. A second reason why we were interested in LEF-1 binding to Pt–GG crosslinks was the presence of two methionine residues at the DNA binding site. Methionines are good ligands for platinum and we wondered whether recognition of a Pt– GG crosslink by the LEF-1 HMG domain could produce covalent DNA–Pt–protein crosslinks. In this work we have examined the affinity of the HMG box of LEF-1 for a series of 25 bp oligonucleotides bearing a site-specific GG–cisplatin crosslink in the center of the sequence, and differing in the nature of the bases flanking the crosslink. We compared the relative affinities with those determined under the same conditions for the corresponding nonplatinated oligonucleotides, and with that for a 25 bp oligonucleotide bearing the cognate LEF-1 recognition sequence. For the platinated oligonucleotide forming the strongest complex, the cisplatin–GG cross-link was replaced by the third-generation platinum antitumor drug, oxaliplatin, and the recognition by the HMG domain of LEF-1 of both cross-links was compared. 2. Experimental 2.1. Materials All chemicals were purchased from Sigma (France) except when indicated otherwise. 2.2. Expression and purification of the LEF-1-HMG domain The LEF-1-HMG domain was expressed as a fusion protein in Escherichia coli strain BL21-CodonPlus-Ril (Stratagene). Fusion proteins are often more stable and sol-

243

uble compared to the native protein alone. In addition, fusion tags are useful for purification. However, fusion proteins may interfere with the functionality of the target protein; hence they are cleaved after purification. The gene construct composed of the expression vector pGEX-3X and the BamHI-EcoRI fragment of LEF-1 was kindly provided by prof. Rudolf Grosschedl (University of California, San Francisco). Cysteine 319 was mutated to serine (using the Stratagen Quick-Change site directed mutagenesis kit) in order to avoid dimerization of the protein, and the final sequence of the plasmid was verified using the Sanger method [13]. This Cys versus Ser mutation is far away from the DNA binding domain and does not affect the DNA binding [9]. Three methods of expression/purification to obtain the pure LEF-1-HMG domain were used. In the first, expression of the fusion protein was induced by 1 mM isopropyl-b-D-thiogalactopyranoside (IPTG) (Euromedex, France) for 18 h at 25 C. The bacterial cells were collected by centrifugation, re-suspended in phosphate-buffered saline (PBS) and lyzed using lysozyme. Then phenylmethylsulfonyl fluoride (PMSF) was added and cells were sonicated two times for 30 s. The fusion protein was obtained from the lysate using glutathion-sepharose affinity chromatography (Pharmacia), where the fusion protein was eluted from the beads with 10 mM glutathion. The fusion protein was concentrated and treated by the blood coagulation factor Xa (Novagen) (3 h incubation at 16 C with 500 u of factor Xa per 1 mg of GST-HMG-LEF1 fusion protein which was diluted at a concentration of 3– 5 lM). Free glutathion-S-transferase (GST) was removed using glutathion-sepharose affinity chromatography. In the second method the protein expressed as described above was purified as follows. The cells were collected by centrifugation, re-suspended in GST binding buffer (Na2HPO4 4.3 mM, KH2PO4 1.47 mM, NaCl 0.137 M, KCl 2.7 mM) and lyzed using lysozyme. Then Triton 0.1% and PMSF were added and the cells were allowed to undergo an osmotic choc in 20% glycerol. After centrifugation the fusion protein in the supernatant was allowed to bind to GST binding resin (GSTrap, Pharmacia) during 2 h at 4 C in the solution containing DTT 10 mM, urea 0.6 M (both Merck) and PMSF. The mixture was then centrifuged, resins transferred to the column and washed with GST binding buffer containing 10% of glycerol. The fusion protein was afterwards obtained by elution from beads using elution buffer (Tris–HCl pH 8, 50 mM, glutathion 10 mM, glycerol 10%). The fusion protein was concentrated and treated by factor Xa as indicated above. In the third method LEF-1-HMG was expressed with the E. coli chaperon system using co-transformation of E. coli cell with pGST-LEF and pGroESL (a plasmid expressing the GroES and GroEL E. coli chaperonin system). The protein was purified as in the first method described above. The purity of the LEF-1-HMG samples was assessed by sodium dodecyl sulfate-polyacrylamide gel electrophoresis

244

K. Chva´lova´ et al. / Journal of Inorganic Biochemistry 102 (2008) 242–250

(SDS-PAGE), yielding a single band. The purified protein had a mass of 14,300 ± 10 Da (average mass, determined using MALDI (Matrix Assisted Laser Desorption Ionization) or ESI (Electrospray Ionization) mass spectrometry), independently of the preparation and purification method. This mass was larger than that expected for the cleavage site of factor Xa (12,605 Da). The protein was resistant to Edman degradation, indicating that the N-terminus was blocked by a post-translational modification. We observed the formation of a small amount of the same fragment already during the lysis procedure, in absence of factor Xa, and in spite of the presence of a cocktail of protease inhibitors. This suggested that during incubation with factor Xa, the GST-HMG-LEF1 fusion protein was cleaved spontaneously and site-specifically, but not due to the expected interaction with factor Xa. The formation of the 14,300 Da fragment was reproducible and the fragment bound specifically to the LEF-1 cognate DNA, albeit with about 50-fold lower affinity than that reported by Giese et al. [14] for the 105 amino acids fragment consisting of amino acids 287–385 and the N-terminal overhang GIPPHM resulting from cleavage with factor Xa. All these observations indicated that our LEF-1-HMG had an additional 16 amino acids (FGGGDHPPKSDLIEGR) overhang at the N-terminal side with respect to the Xa cleavage site, and was modified at the N-terminal phenylalanine. An unexpected extension past the C-terminal can be excluded, since the coding strand of the plasmid terminates at the 3 0 -end with two adjacent STOP codons [15]. Common modifications of a N-terminal include formylation (+28 Da) and acetylation (+42 Da) [16,17], yielding for the N-terminal modified fragment a theoretical average mass of 14296.5 and 14310.5 Da, respectively, both in agreement with the mass of 14,300 ± 10 Da found experimentally. The 2e charge of the FGGGDHPPKSDLIEGR extension past the factor Xa cleavage site (2D and E are anionic, K and R cationic, acetylation or formylation of the N-terminus deletes one positive charge) provides a rationale for the diminished affinity towards the cognate DNA. Since according to the structure of the recognition complex [9,10] the N-terminal overhang is not expected to interfere with the recognition mechanism or to affect binding specificity, we used this 121 amino acids fragment consisting of amino acids 287–385 of LEF-1 preceded by the Ac-FGGGDHPPKSDLIEGRflGIPPHM overhang at the N-terminus (fl indicates the usual factor Xa cleavage site) in the affinity studies. We shall designate this 121 amino acids fragment subsequently LEF-1HMG121. 2.3. Oligonucleotide probes The synthetic oligodeoxyribonucleotides were purchased from Eurogentec and their purity was verified by gel electrophoresis (showing one band in each case) and by mass spectrometry. The site specific 1,2-(GpG) cisplatin intrastrand cross-link was generated by reacting a 100 lM solu-

tion of the GG-containing top strand in 50 mM NaClO4 with 2 equiv. of cis-[Pt(NH3)2(H2O)2]2+ at 37 C for 24 h. In the case of oxaliplatin the 1,2-GG crosslink was prepared by reacting the 100 lM oligonucleotide solution in 50 mM NaClO4 with 1.5 equiv. of [Pt(R,RDACH)(H2O)2]2+ (DACH = 1,2-diaminocyclohexane) at 37 C for 4 h. cis-[Pt(NH3)2(H2O)2]2+ was prepared by dissolving of cisplatin (Hereaus) in water in the presence of 1.9 equiv. of AgNO3 during 24 h at 37 C protected from light. AgCl was removed by centrifugation at 12,000 g for 10 min. [Pt(R,R-DACH)(H2O)2]2+ was obtained as a 1 mM aqueous solution by stirring overnight a sample of [Pt(NO3)2(R,R-DACH)] in water acidified to pH 4 with HClO4. A sample of each platinated single strand was radiolabeled in order to check the sites of platination. The labeling was achieved using T4-polynucleotide kinase (USB) and [c32P]-adenosine 5 0 -triphosphate (ATP) (Amersham Bioscience). The radiolabeled platinated single strand was purified using polyacrylamide gels on which the GG-Pt chelate migrated as a major band between weaker bands due to the remaining unplatinated oligonucleotide (migrating faster) and a smear of multiplatinated species (migrating more slowly) (Fig. S1). The major band containing the GG–Pt chelate was cut from the gel, eluted with aqueous 0.15 M NaCl and precipitated by addition of ethanol. Formation of the GG–Pt cross-link was confirmed using the Maxam-Gilbert sequencing reaction for guanines (showing that the cleavage by DMS/piperidine was inhibited at both guanines) [18]. The unlabeled portion of the platinated ss oligonucleotides was purified with PAGE in a similar way as the labeled one, using UV detection. The radiolabeled duplexes were prepared by mixing the labeled single-stranded oligonucleotide with increasing amounts of the unlabeled complementary strand until the duplex was formed as the only species, as checked using non-denaturing gel electrophoresis. The hybridization was achieved by heating to 90 C for 5 min in 0.1 M NaCl, and slowly cooling down to 4 C. To prepare unlabeled duplexes, we mixed the two single strands using their UV absorbance for quantification and hybridized them using the same protocol as for the labeled duplexes. 2.4. Electrophoretic mobility shift assays The protein-AAA25 binding equilibrium (see Table 1 for labelling of the oligonucleotides) was investigated using direct titration. One nanomolar solution of the labeled duplex (p[32P]AAA25) was incubated with increasing amounts of LEF-1-HMG121 in 10 ll of binding buffer (50 mM NaCl, 20 mM Tris–HCl pH 7.5, 1 mM DTT, 1 mM MgCl2, and 50 mg/ml of BSA) [14] for 1 h at 20 C. The affinity constants of the (platinated and unplatinated) GG-containing duplexes were determined using competition assays: 1 nM solutions of the labeled duplex containing the consensus sequence (p[32P]AAA25) were incubated with 200 nM LEF-1-HMG121 and an increasing amount of the unlabeled platinated or nonplatinated

K. Chva´lova´ et al. / Journal of Inorganic Biochemistry 102 (2008) 242–250

duplex oligonucleotide. The protein–DNA complexes were separated from uncomplexed DNA by electrophoresis on a native 6% polyacrylamide gel in 25 mM Tris, 190 mM glycine and 5 mM EDTA. Equilibrated samples were loaded with sample buffer (bromophenol blue and xylene cyanol in 15% ficoll) to a pre-cooled gel and electrophoresed at 4 C and 170 V for 1–2 h, dried under vacuum, exposed to a molecular imaging plate and analyzed on a Molecular dynamics phosphor imager (Storm system). The radioactivity in each band was quantified with the ImageQuant software.

The radiolabeled platinated duplex p[32P]AG*G*T prepared as described above (10 nM) was incubated with 200 nM LEF-1-HMG121 overnight at room temperature in a buffer containing 20 mM Tris–HCl pH 7.5, 50 mM NaCl, 1 mM MgCl2, adding 10 mM 2-nitro-5-thiocyanobenzoic acid which serves as an inhibitor of DNase. As controls we incubated the radiolabeled unplatinated duplex p[32P]AGGT with LEF-1-HMG121, the cognate oligonucleotide p[32P]AAA25 with LEF-1-HMG121, and the radiolabeled platinated p[32P]AG*G*T without LEF-1-HMG121. The ability to form DNA–Pt–protein ternary complexes was assessed by 10% SDS-PAGE after mixing the samples with the loading buffer (50 mM Tris pH 6.8, 100 mM DTT, 2% SDS, 0.1% bromophenol blue, 10% glycerol) and denaturing by heat at 90 C for 5 min. Gels were electrophoresed for 1–2 h at 140 V, dried and visualized by using Molecular dynamics phosphor imager (Storm system). The radioactivity in each band was quantified with the ImageQuant software. 3. Results 3.1. LEF-1-HMG121 binds specifically to 25 bp DNA containing its consensus sequence First we determined the dissociation constant between LEF-1-HMG121 and its consensus sequence within a 25 bp oligonucleotide (p[32P]AAA25). For this purpose

0.8 0.7 0.6 0.5

θ

2.5. Search for covalent crosslinks between platinated DNA and LEF-1-HMG121

245

0.4 0.3 0.2 0.1 0 -10 10

10

-9

-8

10

-7

10

10

-6

[LEF-1 HMG]/M

Fig. 1. Gel mobility shift assay of the titration of [32P]AAA25 (1 nM) with LEF-1-HMG121 (top). Plot of the fraction of bound oligonucleotide (squares) with a fit to Eq. (1) (solid line) (bottom).

direct binding experiments were performed. One of the four experiments is shown in Fig. 1. The dissociation constant Kd was determined from the plot of h, the fraction of bound oligonucleotide probe, against the protein concentration, P, by least-square fits to Eq. (1) [19], as 45 ± 15 nM (weighted average from 4 experiments). The value of Kd was similar for all the preparation methods of LEF-1-HMG121 used, including the third method using the E. coli chaperon system (see Section 2). The latter method was employed in order to check whether misfolding of the protein could affect the Kd value. This was obviously not the case. One referee pointed out that the maximum fraction of DNA we observed to bind to protein is 75% (Fig. 1), and perhaps levels off at this value and does not approach 100%. Incomplete DNA binding and leveling off at less than 100% has been in fact observed by many authors (see e.g. Fig. 5 in [8], Fig. 3 in [20], Fig. 2 in [21]). The origin

Fig. 2. Competition gel mobility shift assays. LEF-1-HMG121 (200 nM) was mixed with radiolabeled p[32P]AAA25 (1 nM) and varying concentrations of unlabeled competitors in the DNA binding buffer (see Section 2.1).

K. Chva´lova´ et al. / Journal of Inorganic Biochemistry 102 (2008) 242–250

246

Table 1 Top strands of DNA duplexes used in this work Duplex label

Sequence of top strand

AAA25 AGGA AGGC AGGT CGGT TGGT

5 0 -GGG AGA GCT TCA AAG GGT GCC CTA C-3 0 5 0 -TTA TTC CTC TAG GAC TTC TAT TAT T-3 0 5 0 -TTA TTC CTC TAG GCC TTC TAT TAT T-3 0 5 0 -TTA TTC CTC TAG GTC TTC TAT TAT T-3 0 5 0 -TTA TTC CTC TCG GTC TTC TAT TAT T-3 0 5 0 -TTA TTC CTC TTG GTC TTC TAT TAT T-3 0

The GG platinum binding site is marked in bold letters.

of this incomplete binding is poorly understood. One conceivable explanation is that at large protein concentrations, protein–protein interactions impede DNA binding. h¼

P P þ Kd

ð1Þ

Table 2 Relative affinities (dimensionless numbers) for binding of LEF-1-HMG121 to platinated and unplatinated oligonucleotides, determined from the slopes of the straight lines shown in Fig. 3 Oligonucleotide

AGGA AGGC AGGT CGGT TGGT

Relative affinitya Platinated

Unplatinated

7.6 ± 1.1 5.4 ± 0.7 16.8 ± 1.4 5.9 ± 0.8 8.8 ± 0.7

3.0 ± 0.4 2.0 ± 0.2 3.9 ± 0.9 1.0 ± 0.3 2.9 ± 0.8

Affinity enhancement due to platination 2.5 2.7 4.3 5.9 3.0

a The standard deviations of the relative affinities were considered to be proportional to those of the slope k determined for each regression line y = kx + 1 in Fig. 3. The standard deviations of the slopes k, r(k), were calculated assuming classical error propagation, and considering that the errors of the individual yi values were composed of the deviation of their mean value from the regression line, plus the individual standard deviation pP ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi2ffi x2 ½jy kxi 1jþrðy i Þ i i i P 2 of the value yi, i.e., rðkÞ ¼ :

x i i

Fig. 3. Competition for LEF-1-HMG121 (200 nM) between labeled p[32P]AAA25 (1 nM) and varying concentrations of unlabeled competitors. Plots of the ratio of fraction of bound p[32P]AAA25 in the absence of unlabeled competitor and the fraction bound at the indicated concentrations of unlabeled competitor, as a function of concentration of unlabeled competitor. The straight lines were obtained from linear regression under the constraint to pass through (0,1).

*G TG *T G T

TG

*G CG *T G T

CG

*G A *T G G T

G A

*G A *C G G C

G A

G

*G A *A G G A

20 18 16 14 12 10 8 6 4 2 0

A

To determine the affinity of LEF-1-HMG121 for oligonucleotides containing one specific 1,2-(GpG) cisplatin intrastrand cross-link we used competitions experiments, in which the labeled p[32P]AAA25 duplex was displaced from the complex with LEF-1-HMG121 using the examined platinum-crosslinked oligonucleotides as unlabeled competitors. The affinities of the unplatinated oligonucleotides were determined as well. A typical gel is shown in Fig. 2. The 25-bp duplexes used in this study are listed in Table 1. The top strands had the general sequence 5 0 -TTATTCCTCTXG*G*YCTTCTATTATT-3 0 , where asterisks denote the cisplatin binding site, and X and Y represent

Relative affinity

3.2. LEF-1-HMG121 recognizes one site-specific GG-Pt crosslink in a flanking bases-dependent manner

Fig. 4. Diagrammed representation of the relative affinities from Table 2. Hatched bars: platinated oligonucleotides. Black bars: unplatinated oligonucleotides. The error bars indicate the standard deviations.

dA, dC or dT. In a first series of experiments we examined the oligonucleotides where X = dA was kept constant and Y was varied (Y = dA, dC or dT). The strongest binding was found for the sequence AG*G*T. In the second series, Y = dT was therefore fixed and X was varied. The results were analyzed as shown in Fig. 3 as plots of the ratio between the fraction of the labeled p[32P]AAA25 bound without competitor to that bound with competitor, as a function of the competitor concentration. These plots could be fit as straight lines passing through (0,1). From the theory [22], it can be deduced that the slopes of the straight lines are proportional to the affinity constant between the protein and the given competitor oligonucleotide. The relative affinities are listed in Table 2. A diagrammed representation of the relative affinities, highlighting the effect of cisplatin–GG crosslinks, is shown in Fig. 4. The data from Figs. 3 and 4, and Table 2 show that in all cases the affinity of LEF-1-HMG121 for the duplex bearing the cisplatin cross-link is higher than that for the unplatinated duplex. Thus, LEF-1-HMG121 is able to specifically recognize the cisplatin–GG crosslink within duplex DNA. However, none of the platinated oligonucleotides showed

K. Chva´lova´ et al. / Journal of Inorganic Biochemistry 102 (2008) 242–250

247

higher affinity for LEF-1-HMG121 than for the LEF-1 consensus sequence. 3.3. Affinity of LEF-1-HMG121 for oxaliplatin- versus cisplatin-modified p[32P]AG*G*T The duplex AGGT whose cisplatin-adduct showed the highest affinity towards LEF-1-HMG121 was modified with the third-generation platinum antitumor drug, oxaliplatin, and its affinity was compared to that of the cisplatin-modified oligonucleotide (Fig. 5). The relative affinity determined for the oxaliplatin-modified duplex was 5.6 ± 0.6, i.e., within error limits almost the same as that determined for the unmodified duplex (3.9 ± 0.9). Thus, LEF-1HMG121 does not specifically recognize a GG–oxaliplatin crosslink in the AGGT context. 3.4. Search for covalent crosslinks between AG*G*T and LEF-1-HMG121 Covalent cross-links of DNA with proteins mediated by a DNA adduct of a drug might represent lethal lesions which irreversibly sequester a vital protein such as a transcription factor or a repair protein [23,24]. It was suggested that these ternary complexes may play a role in the mechanism of novel platinum antitumor complexes trans[PtCl2(E-iminoether)2] [25], cis-trans-cis-[PtCl2(CH3COO)2(NH3)(1-adamantylamine)] [26], and of trifunctional binuclear platinum complexes [27]. A recognition complex between a protein and platinum-modified DNA may place nucleophilic sites of the protein in the vicinity of the platinum center and thus favor the formation of a covalent bond to platinum. Therefore, we searched for DNA–

Fraction of DNA bound (competitor)

Fraction of DNA bound (no competitor)/

8 AAA25

7

AG*G*T cisplatin AG*G*T oxaliplatin AGGT

6 5 4 3 2 1 0

500

1000

1500

2000

[competitor DNA] (nM)

Fig. 5. Plot of the ratio of fraction of bound labeled p[32P]AAA25 in the absence of unlabeled competitor and the fraction bound at the indicated concentrations of unlabeled competitor, as a function of concentration of unlabeled competitor. AG*G*T modified by cisplatin (squares), AG*G*T modified by oxaliplatin (triangles), AGGT (crosses) and AAA25 (diamonds). Data are the averages from four measurements with error bars showing the standard deviation.

Fig. 6. Search for covalent LEF-1-HMG121 crosslinks. PAGE of 25 bp oligonucleotide (10 nM) having a radiolabeled top strand, with or without LEF-1-HMG121 (200 nM) in 20 mM Tris–HCl pH 7.5, 50 mM NaCl, 1 mM MgCl2, and 10 mM 2-nitro-5-thiocyanobenzoic acid after overnight incubation at room temperature. Lane 1: p[32P]AG*G*T without LEF-1-HMG121; lane 2: p[32P]AAA25 with LEF-1-HMG121; lane 3: p[32P]AG*G*T with LEF-1-HMG121; lane 4: p[32P]AGGT (unplatinated) with LEF-1-HMG121.

protein crosslinks after incubation of the platinated duplex p[32P]AG*G*T with LEF-1-HMG121, in conditions where more than 50% of the DNA were complexed, using SDS-PAGE as detection method. The unmodified duplex p[32P]AGGT and the cognate oligonucleotide p[32P]AAA25 were used as controls. A third control was the platinated duplex p[32P]AG*G*T without protein. After incubation, the gel of the platinated duplexes showed two weak bands retarded compared with that of free probe, with migration time corresponding to a DNA–protein cross-link (Fig. 6). We did not observe any such bands in either control experiment. It is not clear why two bands are observed. One band may result from a cleavage of the covalently bound LEF-1-HMG121 domain during incubation or denaturing. Each of the slow-migrating bands accounted for 0.1% of the overall radioactivity. 4. Discussion A number of HMG box proteins have been shown to recognize specifically the GG–Pt crosslink formed as the major adduct after reaction of cisplatin or its derivatives with DNA [28]. The lymphoid-enhancer binding factor 1 (LEF-1) [6], the sex-determining region Y protein (SRY) [8], and the human upstream binding factor (hUBF) [7] are HMGB proteins with important cell functions that

248

K. Chva´lova´ et al. / Journal of Inorganic Biochemistry 102 (2008) 242–250

have been shown to bind specifically to GG sites of DNA bearing a cisplatin residue. For the HMG domain of LEF-1, the structure of the complex formed with the cognate DNA sequence has been elucidated using NMR [9,10]. LEF-1 is a sequence-specific HMGB protein which bends the DNA towards the major groove, and the bend locus features two destacked adenines. Lippard et al. have pointed out that these destacked adenines assume an orientation that is very similar to that found in the complex formed between cisplatin and a GG sequence of a DNA duplex (Fig. 2 in [5]). It appears therefore conceivable that the cisplatinated GG sequence could represent a binding target for LEF-1. LEF-1, an intrinsically sequence-recognizing protein, could in this case combine structure-specific and sequence-specific elements. In order to examine this possibility, we have determined in this work the relative affinities for the complexes formed between the LEF-1HMG121 domain and the series of oligonucleotides shown in Table 1, and compared them with that determined for an oligonucleotide of the same length, containing the recognition sequence. Comparing the binding affinities for platinated DNA sequences with those for nonplatinated ones (Fig. 3), we observe 3–6-fold binding enhancements upon platination. These enhancements are similar to the 5-fold enhancement found by Trimmer et al. for binding of the HMG domain of the hSRY protein to GG-containing 20 bp long DNA duplexes [8]. However, in that case the protein bound with a similar affinity to the GG-platinated oligonucleotides and a 20 bp duplex containing the native hSRY recognition sequence. On the other hand, the full-length hSRY protein bound to the recognition sequence 8 times more strongly than to the GG–cisplatin crosslink. Conversely, our LEF1-HMG121 fragment discriminated by a factor of 10 between the LEF-1 recognition sequence and a GG–cisplatin crosslink within a 25 bp DNA, whereas the LEF-1 HMG domain probed by Chow et al. bound with similar affinities to a 92 bp DNA containing one GG–cisplatin crosslink and to a 100 bp DNA containing the LEF-1 recognition domain [6]. These data show that binding selectivity can vary with varying length of both DNA and protein. Variations in selectivity as a function of protein and DNA length were also observed in another paper by Chow et al., who emphasized the impact of the net charges of both protein and DNA on the Kd values [29]. In any case, our data do confirm that cisplatin binding to a GG sequence creates a specific target for the LEF-1 HMG domain. Although we tested only 5 from the 16 possible XG*G*Y combinations of bases flanking the platinated G*G* sequence, the results allow the effect of the X and Y base (and of their complementary bases) to be elucidated under the assumption that their influences are independent, since we have data for the sets AG*G*Y (Y = A, T, C) and XG*G*T (X = A, T, C). We have not studied the influence of a flanking guanine from either the 5 0 or the 3 0 side. Platination of GGG is not trivial [30]. The relative affinities summarized in Fig. 4 indicate that a thymine at the 3 0 -

end and an adenine at the 5 0 -end enhance the affinity of LEF-1-HMG121 for the crosslink. How can this preference be interpreted? A similar study has been reported previously for the recognition of XG*G*Y adducts of cisplatin by the two domains of the chromosomal protein HMGB1 [20,30]. Domain A showed a strong preference for A Æ T base-pairs flanking the cisplatin crosslink, and it was suggested that the larger flexibility due to the smaller number of Watson–Crick base-pairs could facilitate the recognition [20]. A similar effect could contribute to the preference for 5 0 -A and 3 0 -T that is observed here. A second factor proposed by Dunham and Lippard as possibly favoring HMGB-protein recognition of GG–Pt crosslinks flanked by AT pairs involved hydrogen bonding from a polar amino acid near the recognition site to a thymine O2 or an adenine N3 atom of a base near the bend locus [20]. Such a hydrogen bond exists in the recognition complex between the HMG domain A of HMGB1 and a 16 bp DNA duplex bearing a cisplatinated TG*G*A site where the hydrogen atom donor is serine Ser41 and the acceptor is the N3 atom of the adenine flanking the destacked G*G* dinucleotide at the 3 0 -side [5]. A similar hydrogen bond was observed in the recognition complex of the LEF-1-HMG domain with its cognate DNA, where the bend locus occurs in the middle of a CAAA sequence, between asparagine Asn34 (according to residue numbering in ref. [10]) and the thymine complementary to the 3 0 adenine [9]. It is conceivable that a similar hydrogen bond also enhances the stability of the recognition complexes that LEF-1-HMG121 forms with the platinated XG*G*Y sequences. From the two main intercalation sites used by HMG proteins, LEF-1-HMG has only one, namely Met11 [31], and it is therefore very probable that this residue is also used in the recognition of G*G*-cisplatinated DNA to intercalate between the two destacked platinated guanines. This would place Asn34 in a favorable position to form a hydrogen bond to the O2(py) or N3(pu) atom of the 3 0 -neighbor of the G*G* site or its counterpart. An AT pair has only two Watson–Crick hydrogen bonds and can more easily reorient itself to accept another hydrogen bond than a GC pair. This could possibly contribute to the preference of LEF-1-HMG121 for T at the site 3 0 to the G*G*–Pt crosslink. An oxaliplatin–GG crosslink within the AG*G*T duplex did not enhance LEF-1-HMG121 binding to this sequence, whereas the cisplatin–GG crosslink did. Weaker recognition of oxaliplatin–GG crosslinks with respect to cisplatin–GG crosslinks by both HMG domains of the HMGB1 protein [32] as well as by the whole protein [33] was reported previously. Spectator ligands can, of course, interfere with recognition, either directly by contacts with the protein (rather unlikely in this case where the platinum complex binds in the major groove and the protein contacts the minor groove), or indirectly, by affecting the crosslink geometry and/or flexibility [32,21,34]. A salient feature of DNA sequence-specific recognition by LEF-1 is intercalation of a methionine residue

K. Chva´lova´ et al. / Journal of Inorganic Biochemistry 102 (2008) 242–250

(Met11) between two adenines whose destacking creates the bend in the double-helix. Another methionine (Met14) is located near the intercalation site [9–11]. Cisplatin binding to a GG sequence destacks the two adjacent guanines in a way very similar to that in which LEF-1 destacks the two adenines. If LEF-1 recognizes the GG– cisplatin crosslink by intercalating Met11 between the two crosslinked guanines, a nucleophilic attack by Met14 on platinum would then conceivably yield a covalent protein–DNA crosslink. After an overnight incubation at room temperature of the most tightly binding platinated duplex complexed with LEF-1-HMG121 we found indeed two very weak bands on denaturing gels, slowly migrating as expected for DNA bound to protein. The weakness of the bands has prevented further characterization of the crosslinked species and the identification of the binding sites. 5. Conclusion We have shown here that cisplatin–GG adducts within double-stranded DNA are weak but specific recognition targets for the HMG domain of the lymphoid enhancer factor 1 to which the factor binds with 3–6-fold affinity compared with the corresponding unplatinated sequence. The nature of the bases flanking the GG site has a significant influence on both the relative affinity and the enhancement due to platination. Thus, to some extent LEF-1 does combine structure-specific (i.e., recognizing a pre-bent structure) and sequence-specific (base-specific contacts) elements when binding to Pt-GG crosslinks. The binding of LEF-1-HMG121 to all GG–cisplatin adducts was however about one order of magnitude weaker than that to the native recognition sequence. It appears therefore unlikely that in the lymphoid cells where LEF-1 is expressed this enhancer binding factor could be detoured from its normal DNA target by Pt–GG crosslinks. Acknowledgements We are indebted to Prof. Rudolf Grosschedl for kindly providing the gene construct of the expression vector pGEX-3X and the BamHI-EcoRI fragment of LEF-1, and to Prof. Giovanni Natile for a sample of [Pt(NO3)2(R,R-DACH)]. We thank Dr. Jana Kasparkova for technical assistance with the protein–DNA crosslinking experiments and Dr. Gerard Bolbach for measuring the MALDI MS spectra. We acknowledge preliminary experiments performed by Dr. Markus Drumm, Marie-Curie postdoctoral fellow (Contract QLGA-CT200-52018) of the European Commission. Financial support from the Association for International Cancer Research (AICR) (grant No. 00-321) is gratefully acknowledged. We also acknowledge a gift of cisplatin from Hereaus GmbH and support from COST (Projects D20/003/00 and D39/004/ 06), enabling scientific exchange with other European research groups.

249

Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.jinorgbio. 2007.08.006. References [1] M. Bustin, R. Reeves, Prog. Nucleic Acid Res. Mol. Biol. 54 (1996) 35–100. [2] R. Grosschedl, K. Giese, J. Pagel, Trends Genet. 10 (1994) 94–100. [3] S.L. Bruhn, P.M. Pil, J.M. Essigman, D.E. Housman, S.J. Lippard, Proc. Natl. Acad. Sci. USA 89 (1992) 2307–2311. [4] J.R. Jamieson, S.J. Lippard, Chem. Rev. 99 (1999) 2467–2497. [5] U.M. Ohndorf, M.A. Rould, Q. He, C.O. Pabo, S.J. Lippard, Nature 399 (1999) 708–712. [6] C.S. Chow, J.P. Whitehead, S.J. Lippard, Biochemistry 33 (1994) 15124–15130. [7] D.K. Treiber, X. Zhai, H.-M. Jantzen, J.M. Essigmann, Proc. Natl. Acad. Sci. USA 91 (1994) 5672–5676. [8] E.E. Trimmer, D.B. Zamble, S.J. Lippard, J.M. Essigman, Biochemistry 37 (1998) 352–362. [9] J.J. Love, X. Li, D.A. Case, K. Giese, R. Grosschedl, P.E. Wright, Nature 376 (1995) 791–795. [10] J.J. Love, X. Li, J. Chung, H.J. Dyson, P.E. Wright, Biochemistry 43 (2004) 8725–8734. [11] M. Drumm, S. Teletche´a, J. Kozelka, J. Biomol. Struct. Dyn. 23 (2005) 1–11. [12] S.O. Ano, Z. Kuklenyik, L.G. Marzilli, in: B. Lippert (Ed.), Chemistry and Biochemistry of a Leading Anticancer Drug, Verlag Helvetica Chimica Acta, Zu¨rich, 1999, pp. 247–291. [13] F. Sanger, S. Niclen, A.R. Coulson, Proc. Natl. Acad. Sci. USA 74 (1977) 5463–5467. [14] K. Giese, A. Amsterdam, R. Grosschedl, Genes Dev. 5 (1991) 2567– 2578. [15] W.G. Kaelin Jr., W. Krek, W.R. Sellers, J.A. DeCaprio, F. Ajchenbaum, C.S. Fuchs, T. Chittenden, Y. Li, P.J. Farnham, M.A. Blanar, et al., Cell 70 (1992) 351–364. [16] A. Gattiker, W.V. Bienvenut, A. Bairoch, E. Gasteiger, Proteomics 2 (2002) 1435–1444. [17] D.C. Chamrad, G. Ko¨rting, H. Scha¨fer, C. Stephan, H. Thiele, R. Apweiler, H.E. Meyer, K. Marcus, M. Blu¨ggel, Proteomics 6 (2006) 5048–5058. [18] S. Redon, S. Bombard, M.-A. Elizondo-Riojas, J.-C. Chottard, Biochemistry 40 (2001) 8463–8470. [19] K.M. Weeks, D.M. Crothers, Biochemistry 31 (1992) 10281– 10287. [20] S.U. Dunham, S.J. Lippard, Biochemistry 36 (1997) 11428–11436. [21] J. Malina, J. Kasparkova, G. Natile, V. Brabec, Chem. Biol. 9 (2002) 629–638. [22] S.-Y. Lin, A.D. Riggs, J. Mol. Biol. 72 (1972) 671–690. [23] B. Lambert, J.-L. Jestin, P. Bre´hin, C. Oleykowski, A. Yeung, P. Mailliet, C. Pre´tot, J.-B. Le Pecq, A. Jacquemin-Sablon, J.-C. Chottard, J. Biol. Chem. 270 (1995) 21251–21257. [24] S.M. Rink, K.J. Yarema, M.S. Solomon, L.A. Paige, B.M. TadayoniRebek, J.M. Essigmann, R.G. Croy, Proc. Natl. Acad. Sci. USA 93 (1996) 15063–15068. [25] O. Novakova, J. Kasparkova, J. Malina, G. Natile, V. Brabec, Nucleic Acids Res. 31 (2003) 6450–6460. [26] J. Kasparkova, O. Novakova, O. Vrana, F. Intini, G. Natile, V. Brabec, Mol. Pharm. 70 (2006) 1708–1719. [27] M. Kloster, K.H.R. Zaludova, J. Malina, J. Kasparkova, V. Brabec, N. Farrell, Biochemistry 43 (2004) 7776–7786. [28] D.B. Zamble, S.J. Lippard, in: B. Lippert (Ed.), Chemistry and Biochemistry of a Leading Anticancer Drug, Verlag Helvetica Chimica Acta, Zu¨rich, 1999, pp. 73–110.

250

K. Chva´lova´ et al. / Journal of Inorganic Biochemistry 102 (2008) 242–250

[29] C.S. Chow, C.M. Barnes, S.J. Lippard, Biochemistry 34 (1995) 2956– 2964. [30] S.M. Cohen, Y. Mikata, Q. He, S.J. Lippard, Biochemistry 39 (2000) 11771–11776. [31] J.O. Thomas, A.A. Travers, Trends Biochem. Sci. 26 (2001) 167– 174.

[32] M. Wei, S.M. Cohen, A.P. Silverman, S.J. Lippard, J. Biol. Chem. 276 (2001) 38774–38780. [33] A. Vaisman, S.E. Lim, S.M. Patrick, W.C. Copeland, D.C. Hinkle, J.J. Turchi, S.G. Chaney, Biochemistry 38 (1999) 11026–11039. [34] O. Delalande, J. Malina, V. Brabec, J. Kozelka, Biophys. J. 88 (2005) 4159–4169.