Purification of Nuclear Proteins That Bind to Cisplatin-damaged DNA IDENTITY WITH HIGH MOBILITY GROUP PROTEINS

The biochemical processes responsible for the rec- ognition and repair of cisplatin-damaged DNA in human cells are not well understood. We have developed a damaged DNA affinity precipitation technique that allows the direct visualization and characterization of cellular proteins that bind to cisplatin-damaged DNA. The method separates damaged DNA-binding proteins from complex radiolabeled cell mixtures and further resolves them into individual polypeptides by sodium dodecyl sulfate-polyacrylamide gel electrophoresis. This technique is complementary to gel retardation and Southwestern blotting analyses that have been previously used to identify cellular components that specif- ically bind to cisplatin-damaged DNA. Using this technique, we have characterized a set of HeLaS3 nuclear proteins of 26.5, 28, 90, and 97 kDa that specifically bind to cisplatin-DNA adducts. Competition studies with soluble cisplatin-damaged DNA confirmed these findings. The major cisplatin-damaged DNA-binding proteins of 26.5 and 28 kDa recognized adducts of DNA modified with cisplatin but not with its trans-isomer or with UV radiation. These proteins were pu- rified 450-fold to near homogeneity by ion-exchange and cisplatin-damaged DNA affinity chromatography. Amino-terminal sequence analysis showed that the 26.5- and 28-kDa proteins were identical to high mo- bility group (HMG) proteins HMG-2 and HMG-l, re-spectively. ~

~ Cis-Diamminedichloroplatinum (11) or cisplatin has emerged as one of the most widely prescribed chemotherapeutic agents used in combination with other antineoplastic drugs and with ionizing radiation to treat patients with locoregionally advanced malignancies (1)(2)(3)(4). Cellular DNA is generally thought to be the critical biologic target for cisplatin-mediated cell killing (5). A number of different intrastrand and interstrand DNA adducts are formed following cisplatin exposure of mammalian cells (6)(7)(8)(9)(10). Although correlations between cell survival and DNA adduct removal have been reported (11)(12)(13)(14)(15), the biochemical mechanisms responsible for recognition and processing of these lesions have not been wellcharacterized.
An important step in elucidating these processes has been the recent discovery of intracellular factors that bind to * This work was supported by Clinical Oncology Career Development Award 90-170 from the American Cancer Society (to E. N. H) and National Institutes of Health Grant CA-45734 (to P. C. B.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
$ To whom correspondence should be addressed Charlotte Radiation Care Inc., 300 Billingsley Rd., Suite 102, Charlotte, NC 28211. damaged DNA restriction fragments. By use of gel retardation analysis, Chu and Chang (16) identified an XPE-binding factor as a nuclear protein that recognizes UV-irradiated DNA, undamaged ss'DNA, cisplatin-damaged DNA, and undamaged dsDNA with decreasing affinities. XPE-binding factor expression is increased in cisplatin-resistant HeLa and HT1080 cells. Chu and Chang (17) also described a cisplatincross-linked-(CCD) binding factor present in the cytoplasm and nucleus of HeLa and HT1080 cells. In contrast to XPEbinding factor, CCD binding expression is not increased in cisplatin-resistant cells. Toney et al. (18) used Southwestern blotting to identify HeLa cell cytosolic factors of 28 and 100 kDa that selectively bind to DNA damaged by chemotherapeutically active platinum (11) compounds, but not to undamaged DNA or DNA modified with clinically ineffective platinum compounds. The 100-kDa protein apparently binds to 1,2-intrastrand d(GpG) and d(ApG) cross-links formed by cisplatin (19). By use of the same techniques, Andrews and Jones (20) identified nuclear proteins of 26, 28, and 97 kDa that bind to cisplatin-damaged and 1,2-diaminocyclohexane platinum (11)-damaged DNA; the expression of these proteins was similar in sensitive and resistant human ovarian carcinoma cells. The correspondence between the proteins identified by Southwestern and gel retardation analysis is unknown.
In this report we describe an assay that allows the direct visualization and characterization of cellular proteins that specifically bind to damaged DNA and/or interact with a protein-nucleic acid complex. This damaged DNA affinity precipitation assay is analogous to immunoprecipitation/gel electrophoretic techniques and is complementary to current molecular biologic techniques that identify sequence-specific transcription factors and damaged DNA-binding factors (21,22). By use of this assay, we have characterized a set of HeLaS3 nuclear proteins that specifically bind to cisplatindamaged DNA but not to undamaged DNA. The major cisplatin-damaged DNA-binding proteins of 26.5 and 28 kDa were preparatively purified to near homogeneity by ion-exchange and cisplatin-damaged DNA affinity chromatography. Sequence analysis of 21 NH2-terminal amino acids showed that the 26.5-kDa protein is identical to high mobility group (HMG) protein -2 while the 28-kDa protein is identical to HMG-1. These results suggest that HMG-1 and -2 proteins preferentially recognize CDDP-damaged DNA. Hence, these proteins may participate in the repair of CDDP-damaged DNA, in addition to their many postulated functions which include general chromatin assembly proteins and specific transcriptional activators (23)(24)(25)(26)(27).
Cell Culture Trchniques-The HeLaS3 line was obtained from the American Type Culture Collection (Rockville, MD). Monolayer cell cukures were grown in Dulhecco's modified Eagle's medium supplemented with 10% fetal bovine serum a t 37 "C in a 95% air, 5% CO, humidified atmosphere. Suspension cultures were grown in Joklik's modification of minimal essential medium supplemented with 5% fetal bovine serum. The HeLaS3 cell line was rout,inely found free of Mycoplasma contamination (28).
Cell Labeling and Preparation of Extracts-Subconfluent monolayers of cells were washed three times each with 25 ml of warm Hanks' balanced salt solution and incubated for 2 h a t 37 "C with 4 ml of Eagle's minimal essential medium lacking methionine hut containing 20 mM HEPES, pH 7.4, and 125 pCi/ml [:"S]methionine (>lo00 Ci/mmol, Du Pont-New England Nuclear). Subsequently, the cells were washed twice in cold Hanks'halanced salt solution. Nuclear and cytoplasmic extracts were prepared by the method of Dignam et al. (29) except that buffer A contained 0.5% Nonidet P-40 and 1 mM phenylmethylsulfonyl fluoride. Instead of the Dounce homogenization step, cells were agitated by vortex for 10 s. Protein was determined hy the bicinchoninic acid method (30). Nuclear and cytoplasmic extracts contained approximately 0.3 and 1.2 mg/ml protein, respectively, with a specific activity of 1 X 10' cpm/mg.
Preparation of Ckplatin-damaged DNA Cellulose-Aliquots of calf thymus ds-or ssDNA cellulose were suspended in 1 mM sodium phosphate, pH 7.4, 3 mM NaCI, and different concentrations of cisplatin. The DNA concentration of the cellulose suspension was 1.05 X lo-:' M nucleotide-phosphate as measured by the method of Alherts and Herrick (31). The initial cisplatin concentrations ranged from 1 X M to 1 X 10" M. Following incubation with gentle rocking a t 37 "C for 24 h in the dark, the DNA cellulose suspensions were washed five times in 20 mM Tris HCI, pH 8.0, 1 mM EDTA, 500 mM NaCI, and 0.5 mM dithiothreitol. The DNA cellulose suspensions were washed once in 20 mM Tris-HCI, pH 8.0, and 50 mM NaCI, and stored a t 4 "C. Experiments with tran.+diamminedichloroplatinum were performed under identical conditions. The molar ratio of free cisplatin to nucleotide-phosphate at the onset of incubation, R,, determines the molar ratio of hound platinum t o nucleotide-phosphate (32). The concentrations of cisplatin and nucleotide-phosphate for typical experiments yielded an R, value of 0.03, unless otherwise indicated.
Cisplatin modification of calf thymus ds-and ssDNA was performed as described (18). The DNA preparations were sheared prior to use in competition assays by the method of Chu and Berg (33).
Conditions of the protein competition assays were otherwise idrntical to the cisplatin-damaged DNA affinity precipitation assay. ('nits of activity in individual samples were calculated as the reciprocal of the serial dilution that yielded complete inhihition of 26.5-and 28-kI)a protein hinding.
Protein Purification-Nuclear extracts from 3 X 10" HeLaS3 were prepared by the method of Dignam rt al. (29). The nuclear protein extract was diluted to a final concentration of 16 mM HEI'RS. pH 7.9, 350 mM NaCI, 1.2 mM M g CI,, 0.16 mM EDTA, 0.4 m u dithiothreitol, and 20% glycerol, and applied to a 1.5 X IO-cm column of 1'-11 phosphocellulose equilibrated in the same huffer. The P-11 column was washed with 110 ml of 10 mM Tris-HCI, pH 7.5, and 350 mM NaCI, and developed stepwise with 60 ml each of 10 mM Tris-HCI. pH 7.5, containing 450. 465. and 550 mM NaCl. respectivelv. Chlumn fractions containing cisplatin-damaged DNA binding activity were pooled, diluted 6-fold with 20 mM Tris-HCI. pH 7.5, and applied to a  ["SI Methionine-laheled proteins precipitated by each preparation were analyzed on 10% polyacrylamide grls as descrihed under "Experimental Procedures." Imnr I, 1.3 X 10' cpm of crude nuclear extract: lone 2, proteins precipitated with undamaged dsI)NA cellulose: lnnr 3.

Identification of Human Cell Proteins
That Rind to Cisplatin-damaged DNA-The interaction of HeLaS3 cellular proteins with genomic DNA damaged by cisplatin was st.udied by use of the damaged DNA affinity precipitation technique. Fig.  1 shows the pattern of HeLaS9 nuclear proteins that bound to different cellulose matrices containing native and damaged DNA. Comparison of these complex protein patterns defined five classes of binding proteins: 1) major proteins of 2fi.5 and 28 kDa substantially enriched by binding to cisplatin-damaged DNA; 2) minor proteins of 90 and 97 kDa that bound to cisplatin-damaged DNA, as well as a group of lower molecular weight proteins that migrated near the dye front; Dependence of Specific Protein Binding on Extent of Cisplatin-DNA Damage-We observed a dose-dependent increase in the 26.5-and 28-kDa proteins with increasing levels of cisplatin-DNA adduct formation (Fig. 2). For these studies, dsDNA cellulose was incubated with different concentrations of cisplatin ranging from 0.001 to 0.1 mM. Binding of the 26.5and 28-kDa proteins was detected at an initial cisplatin concentration of 0.01 mM. The 90-and 97-kDa proteins were detected at the same level of DNA damage.
A set of minor proteins ranging in size from 30 to 42 kDa also hound to cisplatin-damaged DNA. These proteins were not specific for cisplatin-damaged DNA as they were found to hind to DNA damaged by a number of other alkylating agents, including dimethyl sulfate, methylnitronitrosoguanidine, and methylmethanesulfate (data not shown).
The 30-42-kDa proteins appeared to bind nonspecifically to chemically damaged DNA.
Specificity of the Cisplatin-damaged DNA-binding Proteins-In order to study the specificity of 26.5-and 28-kDa protein binding, ["%]methionine-labeled HeLaS3 nuclear extracts were incubated with different DNA cellulose matrices. Only ds-and ssDNA cellulose modified by cisplatin specifically hound the 26.5-and 28-kDa proteins (Fig. 3). No binding was ohserved with dsDNA cellulose damaged by treatment with equimolar or a 10-fold greater concentration of tramdiamminedichloroplatinum (11) or by treatment with UV radiation.
Competition for Specific Protein Binding with Soluble Cisplatin-damaged DNA-A series of competitive binding experiments was performed to study further the specificity of binding of the 26.5-, 28-, 90-, and 97-kDa proteins. For these experiments, limiting amounts of ["'Sjmethionine-labeled HeLaS3 nuclear extract were incubated with a fixed amount of cisplatin-damaged DNA cellulose in the presence of increasing amounts of soluble competitor DNA. Specific protein binding to cisplatin-damaged dsDNA cellulose was effectively competed by increasing quantities of soluhle dsor ssDNA modified by cisplatin to the same extent (Fig. 4, A and I?). The 90-and 97-kDa proteins were poorly resolved in the minigel system. T o obtain a quantitative estimate of the amount of protein bound, the autoradiograms were scanned with a densitometer. Binding of the 26.5-and 28-kDa proteins was inhibited by 50% a t a greater than 5-fold molar excess of competitor dsDNA modified hy cisplatin, while complete inhibition of 26.5-and 28-kDa protein hinding was ohserved at a 100-fold excess of competitor cisplatin-damaged DNA (Fig.  4, A and B). Likewise, 26.5-and 28-kDa protein binding to cisplatin-damaged ssDNA cellulose was specifically inhibited by increasing quantities of soluble dsor ssDNA modified hy ing protein on Q-Sepharose. The 450 and 465 mM NaCl phosphocellulose column eluates were pooled, diluted, and applied to a Q-Sepharose column. The column was washed and eluted as described under "Experimental Procedures." Fractions of 2 ml each were collected. Panel R, electrophoretic analysis. Fractions from Q-Sepharose chromatography were analyzed as described in Fig. 6. The samples were from the corresponding column fractions. cisplatin (Fig. 4, C and D). In contrast, undamaged competitor DNA had much less effect on protein binding. The specific binding of the 26.5-and 28-kDa proteins was only partially inhibited at a 100-fold excess of undamaged competitor DNA (Fig. 4, A-D).
Cellular Localization of the Cisplatin-damaged DNA-binding Proteins-In HeLaS3 cells, the 26.5-, 90-, and 97-kDa proteins were localized in the nucleus and absent in the cytoplasm (Fig. 5, lanes 2 and 6). In contrast, the 28-kDa protein appeared to be equally distributed in both nucleus and cytoplasm (Fig. 5). This distribution was independent of whether the cisplatin-damaged DNA affinity precipitation assay was normalized for protein content or ["S]methionine incorporation. The cytosolic 28-kDa cisplatin-damaged DNA-binding protein had the same binding specificity as the 28-kDa nuclear protein. It remains to he established whether the proteins found in the nuclear and cytosolic extracts are identical.
Purification of the 26.5-and 28-kDa Proteins-Preparative precipitation of HeLaS3 nuclear proteins with the cisplatin- damaged DNA affinity technique revealed that the 26.5-and 28-kDa proteins were the major polypeptide components (data not shown). Chromatography with phosphocellulose was a critical step in which the 26.5-and 28-kDa proteins were enriched 130-fold from the crude HeLaS3 nuclear extract. The cisplatin-damaged DNA-hinding proteins of 26.5 and 28 kDa were eluted with 450, 465, and 550 mM NaCl (Fig. 6). The individual column fractions were analyzed hy gel electrophoresis. Fractions of interest were pooled, concentrated by negative pressure dialysis, and analyzed with the protein competition assay, as descrihed under "Experimental Procedures.'' The fractions that eluted with 450 and 465 mM NACI washes were combined and subjected to ion-exchange chromatography on Q-Sepharose. This step resulted in further purification as well as separation of the 26.5-and 28-kDa binding proteins (Fig. 7). Final purification of the 26.5-and 28-kDa proteins in their native state was achieved by binding to cisplatin-damaged DNA cellulose followed hy elution with KI. Gel electrophoretic analysis of the proteins eluting with KI demonstrated major polypeptides of 26,500 and 28,000 daltons (Fig. 8). There was minor cross-contamination of the 28-kDa fraction with the 26.5-kDa protein. A minor proteolytic degradation fragment of 25 kDa was occasionally found in the 26.5-kDa protein fraction. We estimated that the overall recovery of the 26.5-and 28-kDa proteins was approximately 65%; thus, each cisplatin-damaged DNA-binding protein has been purified about 450-fold from the nuclear extract (Table I).
Amino Acid Sequence Analysis-To establish the identity of the 26.5-and 28-kDa cisplatin-damaged DNA-hinding proteins (purified hy ion-exchange chromatography followed hy binding to cisplatin-damaged DNA cellulose, see ahove) NH2terminal amino acid sequence analysis of the purified proteins was performed. The sequence of these proteins is shown in Fig. 9. A search of GenRank revealed that the 26.5-kDa protein is identical to HMG protein 2 and the 28-kDa protein to HMG-1. The primary structures of HMG-1 and -2 have been determined previously (36, 37). There was a sipificant decrease in yield in sequencing cycles 13 and 14, suggesting  that an unusual amino acid feature (e.g. phosphorylation) was present at these serine residues.

DISCUSSION
A set of HeLaS3 nuclear proteins that bind to cisplatindamaged DNA has been characterized by use of a damaged DNA affinity precipitation technique. Major proteins of 26.5 and 28 kDa and minor proteins of 90 and 97 kDa were found to bind specifically to adducts of cisplatin-damaged DNA. Specificity for the 26.5-and 28-kDa protein binding to cisplatin-damaged DNA was established by several criteria: 1) competition for binding was observed with cisplatin-damaged DNA, but not undamaged DNA; 2) a correlation between the extent of protein binding with level of DNA adduct formation; and 3) these proteins did not recognize DNA adducts formed by treatment with trans-diamminedichloroplatinum (11) or UV irradiation. The lack of binding to DNA modified by the trans-isomer is particularly important as trans-diamminedichloroplatinum (11) cannot form intrastrand cross-links between adjacent nucleotides and is chemotherapeutically ineffective (35). Similar levels of 26.5-and 28-kDa protein binding to ds-and ssDNA cellulose damaged by cisplatin was observed, suggesting that these proteins specifically recognize intrastrand adducts. Amino acid sequence analysis of the purified polypeptides revealed that the 26.5-and 28-kDa cisplatin-damaged DNA-binding proteins are identical to HMG-2 and -1, respectively. Interestingly, cisplatin has been shown to selectively cross-link HMG-1 and -2 to DNA in micrococcal nuclease accessible regions of chromatin in isolated nuclei (38).
By use of molecular biologic techniques, several investigators have identified human cellular factors that recognize cisplatin-damaged DNA. Chu and colleagues (16,39) identified XPE-binding factor as a nuclear protein of unspecified size that is absent in some, but not all, xeroderma pigmentosum group E cells. Biochemical and genetic studies showed that XPE-binding factor is homologous to yeast photolyase (40). Hirschfeld et al. (41) found a protein with similar properties that is synthesized at higher levels in UV-irradiated primate cells. Although expression of XPE-binding factor is increased in the cisplatin-resistant HeLa-R1 and HeLa-R3 cell lines, it preferentially recognizes UV-damaged DNA (17). These findings were confirmed by Chao et al. (42). The binding specificity of XPE factor distinguishes it from the major 26.5-kDa (HMG-2) and 28-kDa (HMG-1) binding proteins found in the HeLaS3 nuclear compartment. In addition, preliminary studies have demonstrated similar expression of the 26.5-and 28-kDa proteins in xeroderma pigmentosum groups A, B, C, D, E, and H lymphoblastoid cells.' Chu and colleagues (17) and Fujiwara et al. (43) also identified cisplatin-cross-linked DNA-(CCD) binding factor by gel mobility shift assays with a cisplatin-damaged DNA probe. The major 28-kDa cisplatin-damaged DNA-binding protein (HMG-1) had a number of features in common with CCD-binding factor, including equivalent distribution in nuclear and cytoplasmic subcellular fractions, preferential binding to DNA damaged with cisplatin but not UV radiation, and similar expression in cisplatin-resistant HeLa-R1 and HeLa-R3 The 26.5-, 28-, 90-, and 97-kDa proteins are likely the same as those proteins in the 28-100 kDa range which have been identified with Southwestern blotting in HeLa cells (18) and the 26-, 28-, and 97-kDa proteins in human ovarian carcinoma cells (20). Whether the 90-kDa protein represents a proteolytic degradation product of the 97-kDa protein or a distinct gene product remains to be determined. In contrast to these studies, the 26.5-and 28-kDa proteins were the major species identified by the damaged DNA affinity precipitation technique. The recovery of bound proteins was complete, as treatment of the protein-DNA complexes with DNase I prior to sodium dodecyl sulfate-polyacrylamide gel electrophoreses resulted in identical findings (data not shown). The different results obtained between the Southwestern blotting method and the cisplatin-damaged DNA affinity precipitation technique may be attributed to several factors including: 1 altered rates of specific protein turnover; 3) preferential binding of native versus denatured/renatured protein to cisplatindamaged DNA; 4) differential electrotransfer of the 90-and 97-kDa proteins onto nitrocellulose membranes; 5) use of genomic versus restriction fragment length DNA; 6) cooperative binding as a result of native protein-protein or protein-DNA interactions; and 7) the ability of the 26.5-and 28-kDa protein binding to withstand elution with buffers of ionic strength equivalent to 500 mM NaCl.
The structure, function, and intracellular localization of high mobility group proteins have been extensively studied (23). HMG-1 and -2 have been shown to be associated with nucleosomes of active genes (24), to participate in nucleosome assembly (25), and to stimulate transcription of class I1 and I11 genes (26,27). These properties are consistent with their capacity to bind more strongly to single-stranded than doublestranded DNA (51), to destabilize or unwind double-stranded DNA (52), and to induce negative supercoiling in relaxed plasmids (53). The ability of HMG-1 and -2 to preferentially recognize cisplatin-modified native DNA suggests a role for such proteins in excision repair. For example, the binding of HMG-1 and -2 to cisplatin-damaged DNA may result in local melting with displacement of histones allowing increased accessibility of DNA repair enzymes. By analogy, HMG-1 and -2 have been shown to stimulate transcription in a nonspecific manner, perhaps by alteration of the DNA template in order to permit increased accessibility by RNA polymerases I1 and I11 as well as other transcription factors (26). The ability of HMG-1 and -2 to preferentially bind to single-stranded DNA may be important in this regard (51). Another single-stranded DNA-binding protein, SSB (also called RP-A and RF-A), has been shown to be required for human DNA excision repair in a cell-free system (54). A role for DNA-binding proteins in the damage recognition step that precedes incision has been suggested by HBlbne and co-workers (55) on the basis of the relative affinities of T4 gene 32 protein for unmodified native DNA and DNA modified with chemical adducts, including cisplatin. The T4 gene 32 protein is also involved in DNA replication, recombination, and the repair of UV damage (56). Alternatively, the clinical efficacy of cisplatin may be associated with the sequestration of HMG-1 and -2 by the preferential binding of these proteins to cisplatin-damaged DNA, thereby restricting their participation in replication and/or transcription.
In summary, the development of a damaged DNA affinity precipitation assay has led to the identification and characterization of a set of human intracellular proteins that specifically bind to cisplatin-damaged DNA. The major proteins recognizing such damage were identified as HMG-1 and -2 by amino-terminal sequence analysis of the purified proteins, suggesting a possible role for these abundant proteins in DNA repair. Based upon our protein purification results (Table I), we estimated that there were 6 X lo5 HMG-1 molecules/cell capable of binding to cisplatin-damaged DNA. A similar number of HMG-2 molecules/cell was also calculated, and the findings are in keeping with previously reported determinations of HMG-1 and -2 amounts of 105-106/cell (23). General application of the method provides a new and powerful tool for studies of protein-damaged DNA interactions. First, considerable biochemical and cell biological information about individual proteins in their native state can be obtained without resorting to antibody production or gene cloning. Second, damaged DNA affinity chromatography can be used for the preparative isolation of binding factors for further biochemical, immunologic, and genetic studies.