HMGB1 is a Co-factor in Mammalian Base Excision Repair

Summary

Deoxyribose phosphate (dRP) removal by DNA polymerase β(Pol β) is a pivotal step in base excision repair (BER). To identify novel BER co-factors, especially those with dRP lyase activity, we used a Pol β null cell extract and BER intermediate as bait for sodium borohydride cross-linking. Mass spectrometry identified the high mobility group box 1 protein (HMGB1) as specifically interacting with the BER intermediate. Purified HMGB1 was found to have weak dRP lyase activity and to stimulate AP endonuclease and FEN1 activities on BER substrates. Co-immunoprecipitation experiments revealed interactions of HMGB1 with known BER enzymes, and GFP-tagged HMGB1 was found to accumulate at sites of oxidative DNA damage in living cells. HMGB1 −/− mouse cells were slightly more resistant to MMS than wild-type cells, probably due to the production of fewer strand-break BER intermediates. The results suggest HMGB1 is a BER co-factor capable of modulating BER capacity in cells.

Introduction

Base excision repair (BER) is a major DNA repair pathway for maintenance of the genome. BER corrects strand breaks and single-base DNA damage arising through endogenous and exogenous exposure to methylating and oxidizing agents among other genotoxicants (Klungland et al., 1999; Lindahl, 1982; Lindahl and Wood, 1999). The current generally accepted working model for mammalian BER is that the process is mediated through at least two sub-pathways that are differentiated by repair patch size and the enzymes involved (Biade et al., 1998; Fortini et al., 1998; Frosina et al., 1996; Klungland and Lindahl, 1997; Prasad et al., 2000). These sub-pathways are designated “single-nucleotide BER” (SN BER) and “long-patch BER” (LP BER) (2-nucleotide or more repair patch). In both BER sub-pathways, repair can be initiated by a damage-specific DNA glycosylase that recognizes and removes the damaged/altered base by hydrolysis of the N-glycosylic bond (Demple and Harrison, 1994; Mosbaugh and Bennett, 1994; Slupphaug et al., 1995). The various steps in both sub-pathways may be coordinated through protein-protein and/or DNA-protein interactions. Despite reconstitution of Pol β-mediated LP BER using various auxiliary proteins, and the demonstration of stimulation of the strand-displacement synthesis step in some cases, the mechanism by which these proteins participate in BER is uncertain. Furthermore, mechanisms of regulation for each of the known protein-protein interactions and their functional implications during BER in vivo are unclear. To understand mechanisms of BER regulation, comprehensive study to identify additional interacting factors is imperative.

Sub-pathway choice is influenced by the rate-limiting step in SN BER, i.e., removal of the 5′-dRP by the dRP lyase activity of Pol β (Horton et al., 2000; Srivastava et al., 1998). For example, if the 5′-dRP cannot be removed efficiently, continued DNA synthesis will emphasize the LP BER sub-pathway (Horton et al., 2000). Yet, both sub-pathways appear to operate simultaneously to repair most types of DNA lesions in vitro (Dianov et al., 1999; Horton et al., 2000; Prasad et al., 2000). We have shown previously that the 5′-dRP BER intermediate is the cytotoxic lesion following exposure to methylating agents, and its removal is a pivotal step in BER in vivo (Sobol et al., 2003; Sobol et al., 2000). Pol β null fibroblasts are hypersensitive to monofunctional DNA methylating agents, and here persistence of the dRP in DNA may signal downstream events such as an increase in apoptosis or necrotic cell death (Horton et al., 2005). Furthermore, oxidized abasic sites are introduced into DNA by free radicals and reactive oxygen species produced by genotoxic agents including UV and γ-irradiation, organometallic oxidants, copper-phenathroline chemical nuclease, and antitumor drugs such as neocarzinostatin (Perrin et al., 1996; Pratviel et al., 1991; von Sonntag, 1991). Such abasic site lesions include 2-deoxyribonolactone, a C1′-oxidized AP site, and are implicated in DNA breakage, mutagenesis, and formation of covalent DNA-protein cross-linked complexes with BER enzymes such as Pol β (Sung et al., 2005). In addition, a BER intermediate, 3′-[2, 3-didehydro-2, 3-dideoxyribose], generated by the AP lyase activity of bifunctional DNA glycosylases, has cytotoxic and genotoxic consequences if not repaired by BER (Loeb and Preston, 1986; Simonelli et al., 2005; Sobol et al., 2003).

Since the BER intermediate dRP group is cytotoxic, its removal is important, especially when Pol β, an enzyme that contributes most of the BER DNA synthesis and dRP lyase activities in the cell, is deficient (Sobol et al., 2003; Sobol et al., 2000). Therefore, to identify additional proteins that have dRP lyase activity or influence removal of the dRP from BER intermediates in the absence of Pol β, we used Pol β null mouse embryonic fibroblast (MEF) cell extract for sodium borohydride (NaBH₄) cross-linking of the Schiff base dRP lyase intermediate protein-DNA complex. Then, mass spectrometry analysis was used to identify a novel BER accessory factor, high mobility group box 1 (HMGB1). HMGB1 is a nuclear non-histone DNA-binding protein that belongs to the high mobility group box family of proteins (Bustin and Reeves, 1996). It has an architectural role in the assembly of nucleoprotein complexes and is highly conserved across species (Bustin and Reeves, 1996; Muller et al., 2004; Thomas and Travers, 2001). HMGB1 binds to DNA in the minor groove without sequence specificity and has the ability to transiently introduce bends or kinks into linear DNA (Bianchi, 2004; Bianchi et al., 1989; Falciola et al., 1997; Pil et al., 1993; Travers et al., 1994; Ura et al., 1996). In the current study, localization experiments with GFP-tagged proteins revealed HMGB1 recruitment to sites of cellular oxidized DNA base damage, and cell-based experiments and biochemical data suggested a role for HMGB1 in BER.

Results

NaBH₄ Cross-linking of Proteins in Cell Extracts

We probed for previously unrecognized BER proteins in cell extracts that can bind specifically to BER intermediates by reacting with the aldehyde group of an abasic site deoxyribose to form a Schiff base complex. We used a BER intermediate probe (Figure 1A) and extracts from Pol β null cells and null cells expressing Flag epitope-tagged Pol β. The strategy included preparation, as the BER intermediate, of a 34-bp DNA with a ³²P-labeled uracil nucleotide at position 16 and a nick between positions 15 and 16 (Figure 1A). This DNA then was treated with uracil-DNA glycosylase (UDG) to create the dRP-containing single-nucleotide gapped molecule. The resulting DNA contained a ³²P-labeled dRP flap in a single-nucleotide gap and a 3′-biotin tag to facilitate isolation of cross-linked protein-DNA complexes (Figure 1A).

Figure 1. NaBH₄ Cross-linking of Proteins in Cell Extracts.

(A) Diagram of BER intermediate with ³²P-labeled uracil at position 16 and a nick between positions 15 and 16, and the reaction scheme. (B) Phosphorimager scan of a NuPAGE Bis-Tris gel illustrating the NaBH₄ cross-linking of proteins in various cell extracts; lanes 1-5, ³²P-labeled DNA (200 nM) was mixed with either dilution buffer (lane 1), Pol β null expressing Flag epitope-tagged Pol β MEF extract (lane 2), Pol β null MEF extract (lane 3), BTNE (lane 4), or purified Pol β (lane 5), and 5 mM NaBH₄. (C) Phosphoimager scan of a NuPAGE Bis-Tris gel illustrating the NaBH₄ cross-linking in various cell extracts; reaction conditions and extracts are as in (B), except that the DNA substrate was not pre-treated with UDG. The positions of protein markers and the cross-linked protein-DNA products are indicated. (D) Photograph of an immunoblot illustrating the presence of HMGB1 in various extracts.

When an extract from cells expressing Flag epitope-tagged Pol β was subjected to NaBH₄ cross-linking and separated by SDS-PAGE, only two proteins migrating around 50 kDa and 37 kDa were strongly labeled (Figure 1B, lane 2). This indicated the selectivity of the probe and the extraordinary binding specificity of these proteins to the BER intermediate. The electrophoretic mobility of the labeled ~50 kDa protein-DNA complex was consistent with the mass of Pol β cross-linked to the ³²P-labeled 19-mer oligonucleotide tagged with biotin and was not observed in a Pol β null extract (Figure 1B, lane 3). Further, the electrophoretic mobility of the ~50 kDa protein-DNA complex was similar to that observed with a purified Pol β-DNA complex (compare Figure 1B, lanes 2, 4, and 5). In contrast, the strong labeling of a single species in the Pol β null cell extracts corresponding to an unknown protein-DNA complex of ~37 kDa was a surprise. The preferential cross-linking of this protein reflected extraordinary specificity in light of the multitude of proteins in the cell. Extracts from null cells expressing Flag epitope-tagged Pol β and Pol β null cells contained similar levels of this protein-DNA complex (Figure 1D, lanes 2 and 3). As will be described below, the protein in this complex was found to be HMGB1. Interestingly, cross-linking of this protein-DNA complex was not observed in the bovine testis nuclear extract (BTNE), a laboratory reference (Figure 1B, lane 4) and a deficiency of this protein in the BTNE was confirmed by immunoblotting (Figure 1D, lane 4). A distinct minor band with slightly faster mobility than cross-linked Pol β was observed with the sample of purified Pol β (Figure 1B, lane 5). A band with similar mobility was also observed with the BTNE and the Pol β expressing cell extract, but not with the Pol β null extract (Figure 1B, compare lanes 2 to 5). The identity of this band was not explored here, but was probably Pol β with the dRP group radiolabeled (Pinz and Bogenhagen, 2000). A control experiment, where DNA substrate was not pre-treated with UDG, showed only a background level of cross-linking (Figure 1C). The results are consistent with an absolute requirement for the 5′-dRP group at the nick.

Isolation of the NaBH₄ Cross-linked Protein-DNA Complex

The cross-linked protein was isolated and purified using streptavidin-coated Dynabeads, as described under “Experimental Procedures” and “Supplemental Data.” Cross-linked protein was eluted from the beads in gel-loading buffer and resolved by SDS-PAGE. The gel was stained as shown in Figure 2A, and slices were excised from the gel corresponding to the radioactive band (Figure S1); proteins were then subjected to in-gel trypsin digestion and mass spectrometry.

Figure 2. Identification of the Cross-linked Protein-DNA Complex by Mass Spectrometry.

Photograph of a SYPRO Ruby stained gel (A) of the purified cross-linked protein-DNA complexes. The gel was scanned for imaging (“Supplemental Data”) and stained with SYPRO Ruby fluorescent dye before excising the protein bands designated as ³²P-labeled proteins, “1-3.” (B) MALDI mass spectrum from the in-gel digestion of Band 1 in (A). The ions labeled as 842.51, 1045.56, and 2211.11 correspond in mass to autolysis products of trypsin. The ions labeled with an asterisk (*) correspond in mass to tryptic peptides of HMGB1 protein. (C) Observed masses in MALDI mass spectrum correspond in mass to theoretical tryptic peptides of HMGB1. “X” indicates the ion 1520.84 analyzed in (B). The inset shows the mass range of 900-1750. (D) MALDI/MS/MS data of the (M+H)⁺ ion of m/z 1520.84 shown in (B). Fragment ions observed confirm the sequence of this peptide as tryptic peptide T28-29 (aa113-127) of the HMGB1 protein.

Identification of Protein in the NaBH₄ Cross-linked Complex by Mass Spectrometry

The MALDI mass spectrum (Figure 2) from the tryptic digestion of Band 1 in Figure 2A showed several ions corresponding in mass to autolysis products of trypsin (e.g. 842.51, 1045.56, and 2211.11). In addition, the inset illustrates ions observed in the mass range of 900-1750. Eleven of these ions corresponded to peptides in HMGB1 (Figure 2B and C), and the (M+H)⁺ ion of m/z 1520.84 was selected automatically during the data dependent acquisition for MS/MS analysis. The resulting MS/MS spectrum (Figure 2D) showed a series of ions that were the result of fragmentations along the peptide backbone. The values from both the peptide masses and the MS/MS fragment ion masses were used in a database search. The protein in Band 1 was identified as HMGB1 (accession #123371) with a Mowse-based score (Pappin et al., 1993) of 102 and a protein score confidence interval of 99.995%.

The MS data were manually interrogated to validate the protein identification in Band 1. Ions that correspond in mass to tryptic peptides of HMGB1 are labeled with an asterisk in Figure 2. Based on these peptide mass fingerprint data, 32% sequence coverage of HMGB1 was observed. Manual interpretation of the MS/MS spectrum shown in Figure 2D revealed a series of fragment ions that corresponds to b- and y-series ions (Biemann, 1988; Roepstorff and Fohlman, 1984). These data confirmed the sequence of this peptide as IKGEHPGLSIGDVAK, which corresponds to a tryptic peptide T28-29 (residue 113-127) of the mouse HMGB1 protein. We conclude that HMGB1 was the protein cross-linked to the BER intermediate probe.

Recruitment of HMGB1 at Sites of DNA Damage in Cells

To examine the response of HMGB1 to DNA damage in living cells, we expressed GFP-tagged human HMGB1 in HeLa cells. These cells were irradiated with 405 nm laser light and the real time response of GFP-HMGB1 was analyzed. Accumulation of GFP-HMGB1 at laser-irradiated sites was compared with those of two DNA glycosylases, OGG1 and NTH1, and two proteins involved in double-strand break repair, KU70 and RAD52, all of which were tagged with GFP at their amino termini. GFP-HMGB1 accumulated after irradiation with 500 scans (50-75 μJ/mm²) (Figure 3), producing base damage and double-strand breaks in addition to single-strand breaks (Lan et al., 2005). However, GFP-HMGB1, like other GFP-tagged proteins shown here, did not accumulate at irradiated site after 100 scans or less, which produces mainly single-strand breaks (not shown). GFP-OGG1 and GFP-NTH1 also accumulated at the irradiated sites (Figure 3), as did GFP-KU70 and GFP-RAD52, involved in non-homologous end joining and homologous recombination, respectively, although with less efficiency. Although 8-MOP is known to produce mono-adducts and DNA crosslinks after irradiation with UVA light ~360 nm UVA light (Cimino et al., 1985), it is also a photosensitizer producing 8-hydroxy-2-deoxyguanosine by absorbing UVA light (Orimo et al., 2006). Addition of 8-MOP before irradiation with only 10 scans of 405 nm laser light, resulted in significant accumulation of GFP-OGG1 and GFP-NTH1 (Figure 3, lower panels), indicating that both purinic and pyrimidinic bases were damaged by the photosensitization. In contrast to the glycosylases, DNA double-strand break repair proteins, KU70 and RAD52, either did not or weakly accumulated under this condition. However, accumulation of HMGB1 was enhanced significantly by the photosensitization, strongly suggesting that HMGB1 is recruited to sites of base damage in living cells.

Figure 3. Accumulation of GFP-tagged Proteins at Sites of DNA Damage in Living Cells.

GFP-tagged human OGG1, NTH1, HMGB1, KU70, and RAD52 were transfected into HeLa cells as described under “Experimental Procedures.” Cells were irradiated with 405-nm scanning laser micro-irradiation system combined with a confocal microscope. Either 500 or 10 scans of 405 nm laser light were applied without or with photosensitizer (8-MOP, 100 μM), respectively. Two representative cells 3 min after irradiation (10 min for RAD52) are shown. These time periods are necessary to obtain the maximum accumulation of each protein. Scanned lines are indicated with arrows.

Characterization of Purified HMGB1

HMGB1 purified from Hela cells was examined for dRP lyase activity on a BER intermediate substrate (Figure 4). To study dRP release, i.e., dRP lyase activity, the DNA substrate was prepared by pre-treating ³²P-labeled duplex oligonucleotide containing a U:G mismatch with UDG and APE. The resulting DNA substrate with a dRP at the 5′-end and ³²P-label at the 3′-end of the same strand (Figure 4A) was incubated with HMGB1, or with Pol β as a positive control. HMGB1 released the dRP from the substrate in a time-dependent manner (Figure 4B), suggesting that purified HMGB1 had dRP lyase activity. Yet, the dRP lyase activity was much weaker than that of Pol β (~600-fold, Figure 4F, and Prasad et al., 1998) (compare Figure 4B, lanes 3 and 4 with lanes 5 and 6, respectively). Interestingly, when the lyase reaction was performed with Pol β in the presence of purified HMGB1, the dRP lyase activity of Pol β was inhibited (compare Figure 4B, lanes 3 and 4 with lanes 7 and 8, respectively). These results suggest that HMGB1 may compete with Pol β for binding to the DNA substrate or mask the dRP from cleavage.

Figure 4. Characterization of HMGB1 from HeLa Cells.

(A) Schematic representation of the dRP lyase substrate (19-mer with 5′-sugar phosphate) generated by pre-treatment of the ³²P-labeled 34-bp oligonucleotide duplex with UDG and APE and the expected product formed as a result of dRP lyase (19-mer without sugar phosphate). (B) The dRP lyase reaction was performed with Pol β and HMGB1, or both, as indicated. The positions of the substrate and the product are indicated. (C) A portion of the purified HMGB1 (lanes 2 and 3, 0.45 and 0.9 μg, respectively) was renatured (lane 4, ~0.2 μg) and analyzed by NuPAGE Bis-Tris gel electrophoresis. The renatured protein was examined for (D) NaBH₄ cross-linking and (E) dRP lyase activity. (D) The reaction mixtures in lanes 1-4 contained DNA alone (lane 1), 500 nM purified HMGB1 (lane 2), or 140 (lane 3) and 280 nM renatured HMGB1 (lane 4). (E) The dRP lyase reaction was performed as in (B) with ~420 nM renatured HMGB1, and samples were withdrawn at 20, 40, and 60 min (lanes 2, 3, and 4, respectively), as indicated. A control reaction in lane 1 was incubated with substrate alone for 60 min. (F) The dRP lyase activities of purified HMGB1 (B) and the renatured HMGB1 (E) were quantified. Data were plotted as product formed (%) against incubation time and fitted to a straight-line equation. The initial rates of purified HMGB1 and renatured HMGB1 were 0.008 and 0.0008/min, respectively. (G) Effect of HMGB1 was evaluated in a standard BER system that contained purified human proteins, UDG (10 nM), APE (1 nM), Pol β (10 nM), and DNA ligase I (200 nM). The reaction mixtures were supplemented with either no HMGB1 (lane 1) or 10 to 200 nM HMGB1 (lanes 2-6) and incubated for 6 min at 37 °C. The positions of HMGB1, BSA, cross-linked HMGB1, dRP lyase substrate and product, and the ligated and unligated BER products are indicated. All enzyme dilutions were made in a buffer containing 50 mM HEPES, pH 7.5, 100 mM NaCl, 1 mM DTT, 0.1 mM EDTA, 20% glycerol, and 1.5 μM BSA. Control reactions contained the same volume of dilution buffer as enzyme solutions.

Since the dRP lyase activity of HMGB1was very weak compared to that of Pol β, we next asked whether the dRP lyase and NaBH₄ cross-linking activities were due to contaminants present in our purified HMGB1. A sample of purified HMGB1 was separated by SDS-PAGE and the stained gel region representing the center of the HMGB1 band was excised. The protein was then eluted from the gel slice, renatured in the presence of BSA, as described under “Supplemental Data,” and tested for cross-linking and dRP lyase activity, as shown in Figure 4D and E. The results demonstrated that this renatured HMGB1 could form a covalent protein-DNA cross-linked product and also could remove the dRP group from a BER intermediate. The recovery of dRP lyase activity in the renatured sample was ~10% (Figure 4F). We concluded that purified human HMGB1 was functional as a dRP lyase and had NaBH₄ cross-linking activity with a BER intermediate. In view of the inhibitory effect of HMGB1 on Pol β dRP lyase, we examined the effect of HMGB1 on a BER system reconstituted with purified human proteins. HMGB1 was found to be inhibitory for BER (Figure 4G).

Stimulation of APE Strand Incision Activity by HMGB1

Since HMGB1 exhibited binding affinity for the model AP site-containing BER intermediate, we examined the effect of HMGB1 on the strand incision activity of APE. Under steady-state conditions, APE activity was stimulated by HMGB1 more than 10-fold (Figure 5A and B). This stimulation could have implications for BER under conditions of limiting APE activity. To illustrate this effect, we reconstituted uracil-DNA BER in vitro using purified human proteins, including only 0.05 nM APE (Figure 5C). HMGB1 stimulated BER product formation.

Figure 5. Stimulation of APE and FEN1 Incision Activities by HMGB1.

Reaction conditions and products analyses are described under “Experimental Procedures.” (A) Incision activity of APE was measured on ³²P-labled UDG-pretreated DNA substrate with 0.1 nM APE in the absence (lane 2) or presence (lanes 3-5) of increasing concentrations of HMGB1 (10-50 nM). Lanes 1 and 6-8 represent DNA and HMGB1 alone controls, respectively. Incision products in controls were subtracted from the APE products for quantification. PhosphorImage of PAGE illustrating the APE incision products. (B) Quantification of APE activity (fold-increase) plotted as a function of HMGB1 concentration. (C) Effect of HMGB1 in a reconstituted BER system was evaluated under limiting APE concentration. The reaction mixture contained purified human proteins, UDG (10 nM), APE (0.05 nM), Pol β (10 nM), and DNA ligase I (200 nM). Reaction mixtures in lanes 1-6 also contained either no HMGB1 (lane 1) or 10 to 200 nM HMGB1 (lanes 2-6) and were incubated at 37 °C for 6 min. HMGB1 concentrations, and ligated and unligated BER products are indicated. FEN1 cleavage activity was measured on either 3 nt-nicked-THF flap (D and F) or nicked-THF flap (E and G) substrate in the absence (lanes 1-2) or in the presence (lanes 3-7) of increasing concentrations of HMGB1 (5-50 nM), respectively, and a constant concentration of FEN1 (5 nM) (lanes 3-7). Control reactions (without FEN1) in lanes 1 and 2 contained either DNA alone (lane 1) or 50 nM HMGB1(lane 2), respectively. All reactions were performed in a buffer containing 50 mM HEPES, pH 7.5, 50 mM KCl, 5 mM MgCl₂, 0.1 mg/ml BSA, and 0.1 mM EDTA. Fold-increase in FEN1 activity in (D) and (E) was quantified and plotted as a function of HMGB1 concentration (F and G). The substrates are schematically depicted above the gel images. The positions of the substrate and FEN1 cleavage products are indicated.

Stimulation of FEN1 Flap Incision Activity by HMGB1

Since the flap generated by Pol β strand-displacement synthesis during LP BER is removed by APE and FEN1 and HMGB1 exhibited binding affinity for a flap-containing model substrate (Figure S4 and Liu et al., 2005), we asked if HMGB1 also could influence FEN1 cleavage on model flap intermediates of LP BER. Both “nicked-3 nt-THF flap” and “nicked-THF flap” substrates were employed to measure FEN1 endonucleolytic cleavage activities in the presence of HMGB1 (Figure 5D-G). With increasing concentrations of HMGB1, from 5 to 50 nM, the amount of FEN1 cleavage product was increased (Figure 5D and E, lanes 4-7). No cleavage product resulting from incubation with HMGB1 or BSA alone was detected (Figure 5D and E, lanes 1-2).

Interaction of HMGB1 with BER Proteins in MEF Extract Expressing Flag Epitope-tagged Pol β

Since HMGB1 has binding specificity for BER intermediates and stimulates FEN1 cleavage activity, we examined the possibility of a direct physical interaction of HMGB1 with BER enzymes, APE, Pol β, and FEN1. We initially used extract from cells expressing Flag epitope-tagged Pol β. Flag epitope or human HMGB1 antibodies and non-immune rabbit IgG were used to co-immunoprecipitate proteins from the MEF extract as described (see “Supplemental Data”), and immunoblots were developed with anti-APE, anti-Pol β, anti-HMGB1, or anti-FEN1 antibodies (Figure 6). The results indicated that the Flag antibody not only immunoprecipitated Pol β, as expected, but also co-immunoprecipitated APE, HMGB1, and FEN1 (Figure 6A, Panels I-IV, lane 2). When a similar experiment was performed with Pol β −/− MEF extract, neither Pol β nor HMGB1 (Figure S5) were immunoprecipitated. In a reciprocal experiment, antibody against HMGB1 co-immunoprecipitated APE, Pol β, and FEN1 (Figure 6B, Panels I-IV, lane 2), whereas non-immune IgG failed to co-immunoprecipitate these enzymes (Figure 6A and B, Panels I-IV, lane 1). When a similar co-immunoprecipitation experiment was performed with a mixture of purified Pol β, APE, HMGB1, and FEN1, both anti-Pol β and anti-HMGB1 antibodies immunoprecipitated HMGB1 and Pol β, respectively, and also pulled down APE and FEN1 (Figure 6C and D, Panels I-III, lane 1). These results illustrate interaction between HMGB1 and key BER enzymes.

Figure 6. Co-immunoprecipitation Demonstrating Interaction of HMGB1 with BER Proteins.

(A and B) extract from MEF cells overexpressing Flag epitope-tagged Pol β was immunoprecipitated (IP) with either non-immune IgG (Panels I-IV, lane 1), anti-Flag (A; Panels I-IV, lane 2) or anti-HMGB1 (B; Panels I-IV, lane 2) antibodies, respectively, and the immunoblots (IB) were developed with the antibodies, as indicated. Lanes 3 in all panels represent a positive control with 1/50^th of the cell extract processed directly in a 4-12% NuPAGE Bis-Tris gel. (C and D) to the mixture of Pol β, HMGB1, APE, and FEN1 (1.5 μM each) in a final volume of 50 μl either anti-Pol β (C; Panels I-III, lane 1), anti-HMGB1 (D; Panels I-III, lane 1) or non-immune IgG (lane 2) were added, and the proteins were IP as described in “Supplemental Data.” Lane 3 represents a positive control where the proteins mixture was processed directly. The immunoblots were developed with respective antibodies, as indicated. Protein identity is shown on the right-hand side of the photographs.

Sensitivity of HMGB1 +/+ and HMGB1 −/− MEFs to MMS

We evaluated a mouse fibroblast cell line with a disrupted HMGB1 gene (HMGB1 −/−), along with an isogenic cell line carrying the HMGB1 gene (HMGB1 +/+). Sensitivity to the methylating agent MMS has been used routinely for evaluating the BER capacity of mammalian cells. The results revealed a slight increase in MMS resistance for the HMGB1 −/− cell line, as compared with the control HMGB1 +/+ line (Figure 7A). Although the difference between these cell lines was only modest, we repeated the experiment several times and consider this difference significant. Yet, the absence of HMGB1 expression did not have a strong influence on the response to the MMS-induced lesions in these cells. The identity of these cell lines was confirmed by immunoblotting using an HMGB1-specific monoclonal antibody (inset, Figure 7A). In addition, the level of Pol β expression in the two cell lines was found to be similar.

Figure 7. Biological and Biochemical Analyses of HMGB1 +/+ and HMGB1 −/− MEFs after treatment with MMS.

(A) HMGB1 +/+ (red filled circles) and HMGB1 −/− (blue open circles) MEFs were treated for 1 h with MMS and cell sensitivity was determined by growth inhibition experiments as described under “Experimental Procedures.” Data represent the mean ± S.E. of six independent experiments. (Panel A, inset) Immunoblot analysis of HMGB1 +/+, HMGB1 −/− and Pol β +/+ MEF extracts with specific antibodies to either (i) HMGB1 or (ii) Pol β. Purified HMGB1 and Pol β proteins were loaded as positive controls as indicated. (B) HMGB1 +/+ (red filled symbols) and HMGB1 −/− (blue open symbols) were treated for 1 h with MMS in the absence (circles) or presence of MX (30 mM) for 1 h (squares) or 4 h (diamonds). Data represent the mean ± S.E. of three independent experiments. (C) Comet assay was performed as described in “Supplemental Data.” HMGB1+/+ and HMGB1−/− cells were treated with MMS (0.5 mM for 30 min) or a combination of MMS and MX (30 mM), and harvested for analysis at 2 h. Medium (control) represents the results from untreated HMGB1+/+ and HMGB1−/− cells. Images were obtained with a fluorescence microscope and image analysis system using 20× magnification. (D) Quantification of Olive Tail Moment (OTM) from (C) in HMGB1 +/+ (red bars) and HMGB1 −/− cells (blue bars). Data represent the mean ± S.E. of three independent experiments; for each treatment, 150 cells were scored. (E) Measurement of incision activity of APE on MX-adducted DNA substrate. Reaction conditions and product analysis are described in “Supplemental Data.” ³²P-labeled MX-adducted duplex oligonucleotide (50 nM) was incubated with 10 nM APE in the presence of increasing concentrations (25 to 200 nM) of HMGB1 (lanes 2-6) for 10 min at 37 °C. The positions of substrate and incised product and the HMGB1 concentration are indicated. (F) The incision activity of APE on MX-adducted substrate was quantified and plotted as arbitrary PhosphorImager units as a function of HMGB1 concentration.

The increased sensitivity to MMS observed in Figure 7A could reflect the increased accumulation of cytotoxic BER intermediates in the HMGB1 +/+ cells than in the HMGB1 deficient cells, since it is well known that accumulation of BER intermediates can correlate with MMS sensitivity in mouse cell lines (Horton et al., 2003). HMGB1 may stimulate production of strand breaks after base lesion removal, an effect expected to be more prominent under conditions of limiting APE activity. Therefore, we tested the effect of treatment of the HMGB1 +/+ and −/− cells with the combination of MMS and methoxyamine (MX), an agent known to react with abasic sites in vivo rendering them much less susceptible to APE incision than the normal abasic site (Horton et al., 2000; Liuzzi and Talpaert-Borle, 1985; Rosa et al., 1991). A stimulatory effect of HMGB1 on APE strand incision activity may be more evident in such a cellular background with MX-adducted AP sites. HMGB1 +/+ cells were considerably more sensitive to MMS than the HMGB1 −/− cells in the presence of MX (Figure 7B).

To confirm the expectation that strand breaks were more abundant in the HMGB1 +/+ cells than the HMGB1 −/− cells, cells were analyzed using the comet assay (Figure 7C). More strand breaks were observed in the HMGB1 +/+ cells than in the HMGB1 −/− cells treated with MMS and MX simultaneously (Figure 7C, bottom panels, and 7D). Single strand breaks under these conditions reflect APE activity, since MX-adducted AP sites are known to be resistant to the alkaline conditions used in the comet assay (Liuzzi and Talpaert-Borle, 1985). To confirm the stimulatory effect of HMGB1 on APE activity against an MX-adducted AP site substrate, an MX-adducted oligonucleotide AP site substrate was incubated with a relatively high concentration of APE in the presence of increasing concentrations of HMGB1. The results revealed HMGB1 stimulation of the APE activity against this substrate (Figure 7E and F). The increased sensitivity of HMGB1 proficient cells to MMS combined with MX is consistent with HMGB1-induced stimulation of single strand break formation and APE strand incision activity.

Discussion

We used chemical cross-linking of DNA-protein complexes in MEF extracts to identify novel mammalian proteins that interact with BER intermediates. The cross-linking technique relied on the ability of a DNA-bound protein to form a Schiff base involving the abasic site deoxyribose C1′-aldehyde and an amino group in the protein (i.e., β-elimination intermediate). The Schiff base was covalently trapped with NaBH₄, and mass spectrometry was used to identify HMGB1 as the major BER intermediate interactive protein in the cell extract other than Pol β. HMGB1 appeared to have strong specificity for binding BER intermediates. This specificity of HMGB1 was confirmed by the photoaffinity labeling method used previously (Lavrik et al., 2001) and by a DNA/protein binding assay (Liu et al., 2005) using purified HMGB1. Immunofluorescence experiments with GFP-tagged HMGB1 in living cells demonstrated rapid accumulation of the protein at sites of DNA damage. Experiments with purified HMGB1 revealed that while it had only weak dRP lyase activity, it was capable of stimulating the strand incision activities of APE and FEN1 on BER related substrates. HMGB1 also was found to interact with known BER enzymes. Finally, a modest increase in MMS sensitivity was observed with HMGB1 proficient MEFs compared with HMGB1 null cells.

In further characterizing the importance of HMGB1 in BER, we focused on the stimulatory effect of HMGB1 on the incision activity of APE, and used a system in which APE activity is limiting for BER. This effect could have important biological implications since APE is a low-abundance nuclear enzyme known to be limiting in certain BER sub-pathways (Wiederhold et al., 2004). Co-treatment with MX results in MX adduction of the AP sites generated after MMS treatment of cells. APE is known to be ~3 orders of magnitude less active against the MX-adducted AP site than the natural AP site (Horton et al., 2000). Also, once APE strand incision occurs, the BER intermediate containing the reduced AP site sugar is blocked for SN BER because the Pol β dRP lyase cannot remove the sugar, a step required for subsequent DNA ligation. Hence, treatment of cells with the combination of MMS and MX renders APE activity as a limiting step in BER in vivo and also causes build up of the incised BER intermediate triggering cell death. Following treatment with MMS combined with MX, we observed strong hypersensitivity in the HMGB1 proficient cells (Figure 7). These results are consistent with production of more strand breaks in the HMGB1-expressing cells, and this interpretation was confirmed by comet assay (Figure 7).

In view of the abundant expression of HMGB1 in MEFs (found here to be ~150-fold higher than Pol β; data not shown) and its binding specificity for BER intermediates, it is likely that the initial BER intermediate (i.e., AP site containing DNA) becomes HMGB1 bound immediately upon formation. The various ensuing transactions of BER are likely to occur in close proximity to HMGB1 and this could influence the enzymology of repair. For example, under conditions of limiting APE activity in cells, HMGB1 may stimulate repair because of its stimulatory effect on APE (Figure 5). In contrast, in the presence of higher levels of APE activity, HMGB1 may reduce SN BER activity through its inhibitory effect on the dRP lyase activity of Pol β (Figure 4). Finally, by virtue of its effect on FEN1 activity, HMGB1 may stimulate LP BER in a cell background where LP BER strand-displacement DNA synthesis is proficient. Further experimental work will be required to evaluate these possibilities.

HMGB1 has already been implicated in mismatch repair, and in the nucleotide excision repair of cisplatin-damaged DNA (Yuan et al., 2004; Zhang et al., 2005). HMGB1 inhibited removal of cisplatin DNA adducts by shielding them from repair, and HMGB1 was suggested as an antitumor agent for use in combination with cisplatin (Huang et al., 1994; Ohndorf and Lippard, 2006; Zamble and Lippard, 1995). Our current findings on HMGB1 interactions with BER proteins and intermediates demonstrate another example of HMGB1 involvement in a DNA repair process. HMGB1 is known to have DNA bending capacity (Pil et al., 1993; Thomas and Travers, 2001). It will be interesting to explore the role of HMGB1-induced BER intermediate bending on the effects of HMGB1 on strand incision by FEN1 and APE.

In conclusion, we applied an affinity labeling technique to screen for cellular proteins that interact with a BER intermediate, and this allowed us to identify HMGB1 as a possible accessory factor in BER. HMGB1 may influence BER by interacting with its DNA intermediates and regulating key enzymes such as APE, Pol β, and FEN1.

Experimental Procedures

Materials

Synthetic oligodeoxyribonucleotides were from Oligos Etc, Inc. (Wilsonville, OR), The Midland Certified Reagent Co. (Midland, TX), Inc., and Integrated DNA Technology (San Jose, CA). [α-³²P]dCTP and [α-³²P]ddATP (3000 Ci/mmol) and [γ-³²P]ATP (7000 Ci/mmol) were from GE HealthCare (Piscataway, NJ) and Biomedicals (Irvine, CA), respectively. Optikinase and terminal deoxynucleotidyl transferase were from USB Corp. (Cleveland, OH) and Fermentas Inc. (Hanover, MD), respectively. Anti-Pol β affinity purified polyclonal and monoclonal (18S) and anti-APE polyclonal antibodies have been described previously (Lavrik et al., 2001; Singhal et al., 1995; Srivastava et al., 1995). Anti-FEN-1 monoclonal antibody and anti-HMGB1 were from GeneTex, Inc. (San Antonio, TX). Recombinant human Pol β was overexpressed and purified as described previously (Beard and Wilson, 1995). Human APE, UDG with 84 amino acids deleted from the amino-terminus, and DNA ligase I were purified as described previously (Slupphaug et al., 1995; Strauss et al., 1997; Wang et al., 1994).

HMGB1 Cell Culture

HMGB1 +/+ and HMGB1 −/− SV40-immortalized MEFs (Calogero et al., 1999) were purchased from HMGBiotech (Milan, Italy) and grown at 37 °C in a 10% CO₂ incubator in Dulbecco's modified Eagle's medium supplemented with L-glutamine (Invitrogen, Carlsbad, CA) and 10% fetal bovine serum (HyClone, Logan, UT). MMS-induced cytotoxicity was determined by growth inhibition assays (Horton et al., 2003). Some studies were conducted in the presence of MX as described previously (Horton et al., 2000).

Accumulation of GFP-tagged Proteins at Sites of DNA Damage in Living Cells

Laser micro-irradiation was carried out as previously described (Lan et al., 2005; Lan et al., 2004) and as detailed under “Supplemental Data.”

Cell Extract Preparation

Cell extracts were prepared as previously described (Biade et al., 1998), and the details are presented in “ Supplemental Data.”

Analytical and Preparative NaBH₄Cross-linking and Purification of the Cross-linked Protein-DNA Complex

To identify proteins from the extracts that react with 5′-dRP, NaBH₄ trapping technique was utilized (Piersen et al., 1996). For a preparative scale cross-linking experiment, the reaction mixture (2 ml) contained 50 mM HEPES, pH 7.4, 20 mM KCl, 1 mM EDTA, 2 mM DTT, 5 μM duplex oligonucleotide (pre-treated with UDG) comprising 1 μM ³²P-labeled DNA to monitor the cross-linked complex, 5 mg Pol β null extract, and freshly prepared 5 mM NaBH₄. The reaction mixture was incubated for 1 h on ice and 10 min at room temperature. Following NaBH₄ cross-linking, the reaction mixture was incubated with an equal volume (2 ml) of streptavidin-coated magnetic beads (~3 × 10⁸ beads) and the protein-DNA complexes were purified using the manufacturer's suggested protocol (see “Supplemental Data”).

In-gel Tryptic Digestion

The protein bands were excised from the gel, cut into small pieces, and transferred into a 96-well microtiter plate. Gel pieces were subjected to automatic tryptic digestion and followed by MALDI analyses as described in “Supplemental Data.”

dRP Lyase Activity Assay

dRP lyase activity was performed as described previously (Prasad et al., 1998).

Preparation of Normal AP and MX-adducted DNA Substrates

³²P-labeled normal AP and MX-adducted DNA substrates were prepared as described previously (Horton et al., 2000).

AP Endonuclease Activity Assay

Incision activity of APE was assayed in a reaction buffer (10 μl) containing 50 mM HEPES, pH 7.5, 0.5 mM EDTA, 2 mM DTT, 20 mM KCl, 5 mM MgCl₂, and 50 nM ³²P-labeled normal AP or MX-adducted AP site DNA. The reactions were assembled on ice with or without the indicated concentrations of HMGB1. The incision reaction was initiated by adding purified APE at the concentrations indicated and incubating at 37 °C. After 10 min, the reactions were terminated by the addition of an equal volume (10 μl) of DNA gel-loading buffer, and the products were separated by 15% denaturing PAGE. Typhoon PhosphorImager was used for gel scanning and imaging. Data were quantified by using ImageQuant software.

Co-immunoprecipitation and Immunoblotting

To examine HMGB1 interaction with BER proteins, Pol β null cells expressing Flag epitope-tagged Pol β were utilized, and co-immunoprecipitation and immunoblotting experiments were performed essentially as described (“Supplemental Data,” and Kedar et al., 2002).

FEN1 Incision Activity Assay

FEN1 endonucleolytic cleavage activity was measured in the absence and presence of HMGB1. FEN1 (1 nM) was mixed with 50 nM nicked-THF flap or 3 nt-nicked-THF flap substrate and increasing concentrations of HMGB1 (5-50 nM). Reactions were performed in buffer containing 50 mM HEPES, pH 7.5, 50 mM KCl, 5 mM MgCl₂, 0.1 mg/ml BSA, and 0.1 mM EDTA. After incubation for 10 min at 37 °C, substrates and products were separated by 15% denaturing PAGE and detected by PhosphorImager.

Table 1.

Oligonucleotide Sequences

Oligonucleotide	Length nt	Sequencea
DS1	31	5′-CTGCAGCTGATGCGCFGTGCGGATCCGGTGC-3′
DS2	31	5′-CTGCAGCTGATGCGCC GTGCGGATCCGGTGC-3′
Downstream (D)b
Dgap/nick	15	5′-pGTGCGGATCCGGTGC-3′
DTHF flap	15	5′-pFGTGCGGATCCGGTGC-3′
DTHF-3 nt-flap	18	5′-pFTTTCGTGCGGATCCGGTG-3′
Template (T)
T	31	3′-GACGTCGACTACGCGGCACGCCTAGGCCACG-5′
Upstream (U)c
Ugap	16	5′-CTGCAGCTGATGCGC-3′
Unick	16	5′-CTGCAGCTGATGCGCC-3′

^aUnannealed residues are in boldface. “F” denotes THF and “U” denotes deoxyuridine. “p” stands for a phosphate group.

^bThe subscripts describe the annealed product. For example, D_uridine is used to create a one-nucleotide-gapped or nicked DNA with a 5′-deoxyurine residue, whereas D_gap/nick generates a one-nucleotide gapped or nicked DNA.

^cThe subscripts describe the annealed product. U_gap is the upstream oligonucleotide for generating a one-nucleotide-gapped DNA, whereas U_nick is the one for generating a nicked DNA.