The gene-encoding HMO2 was amplified from yeast genomic DNA using primers NHP10F 5′-CACATAAGCATATGTCAGTTGAA-3′ and NHP10R 5′-CGTCTTACCATATGTTCAAAGAA-3′ (NdeI sites underlined) and cloned into the NdeI site of pET14b, resulting in expression of HMO2 with an N-terminal His₆-tag. pET14b-HMO2 was transformed into Escherichia coli Rosetta Blue and expression of HMO2 was induced by addition of 1 mM isopropyl-β-d-thiogalactopyranoside (IPTG) at A₆₀₀ of 0.2 and induction was carried out for 4 h. Cells were pelleted and then resuspended in lysis buffer pH 7.5 (50 mM Na_xH_yPO₄, 300 mM NaCl, 10 mM imidazole, 20% glycerol, 1 mM 2-mercaptoethanol and 1 mM phenylmethylsulfonyl fluoride (PMSF)). Cells were disrupted by sonication. The nucleic acids were digested by adding 2 μl of 2000 U/ml of DNaseI (New England Biolabs) and incubated on ice for 1 h. To the lysate, 0.01% of Triton X-100 was added and the solution was incubated on ice for 1 h and then centrifuged at 6000 × g for 20 min. The lysate was incubated with Ni-NTA agarose beads for 1 h on ice. The nickel beads were equilibrated using lysis buffer prior to use. The beads were washed using wash buffer pH 7.5 (50 mM Na_xH_yPO₄, 300 mM NaCl, 20 mM imidazole, 20% glycerol, 1 mM 2-mercaptoethanol and 1 mM PMSF). Proteins were eluted by gravity flow using elution buffer pH 7.5 (50 mM Na_xH_yPO₄, 300 mM NaCl, 250 mM imidazole, 20% glycerol, 1 mM 2-mercaptoethanol and 1 mM PMSF). Pure HMO2 was quantitated using Coomassie blue stained SDS–PAGE gels using bovine serum albumin as standard as well as by Micro BCA Protein Assay Kit (Pierce). At least three independent preparations of HMO2 were used, indicating reproducibility of the preparations. HMO1 was obtained as described earlier (37).

One hundred nanograms of supercoiled or linear pGEM5 was incubated at 4°C with different concentrations of HMO2 in 10 μl reaction buffer (20 mM Tris pH 8.0, 0.1 M EDTA, 15 mM NaCl, 0.1 mM DTT and 0.01% BRIJ 58). Complexes were resolved using 1% TBE agarose gels using 0.5× TBE buffer (50 mM Tris borate, 1 mM EDTA) and visualized by ethidium bromide staining.

Supercoiled pUC18 was nicked using Nt.BstNBI (New England Biolabs). One hundred nanograms of nicked DNA was incubated with varying concentration of HMO2, at room temperature for 1 h. Reactions were initiated using 1 μl of 80 U/μl of T4 DNA ligase (New England Biolabs) in presence of ligase buffer and incubated at room temperature for 1 h. Reactions were terminated using 1 μl stop buffer (5 mM EDTA, 1.1% glycerol and 0.2 mg/ml proteinase K final concentration) and incubated at 37°C for 1 h. Samples were loaded on 1% TBE agarose gels and electrophoresed for 14 h followed by staining with ethidium bromide. For determination of the direction of DNA supercoils produced by HMO2, electrophoresis was performed in the presence of 0.3 µg/ml chloroquine, using Bacillus subtilis HU as a control.

All oligonucleotides used for preparation of DNA constructs were purchased and purified by denaturing gel electrophoresis. Sequences are available in Supplementary Table S1. Electrophoretic Mobility Shift Assays (EMSAs) were performed using DNA duplex and DNA with loops, nicks, gaps and overhangs. A common 37-nt oligonucleotide was used to generate the different constructs. The top strand was 5′-end labeled using T4 polynucleotide kinase (New England Biolabs) and [γ³²P] ATP. The sequences of duplex as well as DNA with nick, gap, loops and overhangs were as described in (39) or presented in Supplementary Table S1. Top and bottom strands were annealed by heating at 90°C and slowly cooling to room temperature. Similarly, a 50 bp DNA duplex and duplex with either a single or tandem loops was generated by extending the 37 bp construct at both ends. Reactions contained 5 fmol of DNA substrate and varying concentration of HMO2 in 10 μl reaction buffer (20 mM Tris pH 8.0, 0.1 mM EDTA, 5 mM NaCl, 0.1 mM DTT, 0.01% BRIJ 58 and 0.1% Triton X-100) and incubated at room temperature for 1 h. Experiments designed to compare different DNA constructs were performed side-by-side with the same protein preparation. Samples were loaded on a prerun 8% polyacrylamide gel (39 : 1 acrylamide:bisacrylamide) and electrophoresed using 0.5× TBE at 175 V for 2 h. Complexes were visualized using Amersham Biosciences Storm Phosphorimager. Reactions were performed at least in duplicate.

The sequence of the four-way junctions (4WJ) was described in (39) and shown in Supplementary Table S1. 4WJ were obtained by 5′-end labeling one strand and annealing the other three strands, followed by purification using native polyacrylamide gels. Complete annealing of all strands was determined by comparing the electrophoretic mobility to that of annealed constructs lacking one of the unlabeled strands. EMSA was performed as described above using 5 fmol of DNA and varying concentration of HMO2 in reaction buffer. Final volume of the reaction was 10 μl.

Five fmol 50-bp duplex DNA with loops and 4-μM of HMO2 was incubated at room temperature for 1 h in reaction buffer. Supercoiled or linear pGEM5 (0.5–50 fmol) was added and incubated at room temperature for 1 h; pGEM5 was linearized with NdeI or AatII. Samples were electrophoresed on prerun 8% polyacrylamide gels. Complexes were viewed using Storm Phosphorimager.

Supercoiled pGEM5 was digested with NdeI or NaeI to obtain DNA with a 2-nt 5′ overhang or blunt ends, respectively. One hundred nanograms of linearized pGEM5 was incubated with varying concentrations of HMO2 or HMO1 at room temperature for 1 h. To this reaction, 1 μl of 400 U/μl of T4 DNA ligase was added and incubated at room temperature for 1 h. Samples were treated with exonuclease III (100 U/μl) at room temperature for 1 h. Reactions were terminated by adding 2 μl stop buffer (5 mM EDTA, 1.1% glycerol and 0.2 mg/ml proteinase K). Alternatively, supercoiled pET5a was digested with BspHI, which creates a 4-nt 5′ overhang. One hundred nanograms of the linearized pET5a was incubated with 2 µM HMO2. To this reaction 1 µl of 400 U/µl of T4 DNA ligase was added and incubated at room temperature for 1 h. Reactions were terminated by adding 2 µl of stop buffer. Samples were electrophoresed on 1% TBE agarose gels and visualized by ethidium bromide staining.

Using forward primer 5′-TGGGGTGCGAATTCTAATGAGT-3′ and reverse primer 5′-GGCTTTACACTTTATGCTTCCG-3′, an extra EcoRI site was introduced into plasmid pUC18 by PCR amplification. The pUC18 variant was then digested with EcoRI to obtain 105 bp DNA duplex with 5′-AATT extensions. The 105 bp DNA was 5′-end labeled using T4 polynucleotide kinase (New England Biolabs). One hundred femtomoles of 105 bp DNA was incubated with varying concentrations of HMO2 or 100 nM B. subtilis HU at room temperature for 1 h. To this reaction, 1 µl of 8 U/µl T4 DNA ligase was added and incubated at room temperature for 1 h. Samples were treated with exonuclease III (100 U/µl) at room temperature for 1 h. Reactions were terminated by adding 2 µl stop buffer and 1 µl of 10% SDS and then phenol chloroform extracted and ethanol precipitated. Samples were loaded on a prerun 5% polyacrylamide gel (39:1 acrylamide:bisacrylamide) and electrophoresed using 0.5× TBE at 175 V for 2 h. Complexes were visualized using Amersham Biosciences Storm Phosphorimager.

Supercoiled pGEM5 was linearized with NdeI or NaeI to yield DNA with 2-nt 5′ overhangs or blunt ends, respectively. pET5a was digested with BspHI to produce 4-nt 5′ overhang, pUC18 was digested with EcoRI, which also produces a 4-nt 5′ overhang, and a pcDNA3-derivative was digested with ApoI, which likewise creates a 4-nt 5′ overhang. pET5a was digested with ScaI and then PvuI (which cuts supercoiled DNA only poorly), to yield a blunt end and 2-nt 3′ overhang, respectively. pRAD1 was digested with DraI to produce blunt ends. Fifty to one hundred nanograms of linearized pGEM5, pET5a, pUC18, pcDNA3-derivative and pRAD1 was incubated at room temperature for 1 h with 2 μM HMO2 or HMO1 previously dialyzed using dialysis buffer (50 mM Tris pH 8.0, 50 mM NaCl and 20% glycerol). To each reaction, 1 μl of 100 U/μl exonuclease III was added and incubated at room temperature for 1 h. Reactions were terminated by addition of 2 μl 10% SDS or by phenol-extraction and ethanol precipitation (the latter in cases where shorter DNA fragments are produced that may co-migrate with SDS on the gels). Samples were electrophoresed on 1% TBE agarose gels. Gels were stained with ethidium bromide. All reactions were performed at least in duplicate, and pGEM5 digested with NdeI to produce 5′-TA overhangs was included for comparison to other DNA constructs.

One hundred nanograms of supercoiled or linearized pGEM5 with blunt ends or overhangs were incubated at room temperature for 1 h with 1 μM HMO2 in 10 μl reaction buffer. One microliter of 0.1 U/μl DNaseI (New England Biolabs) was added in presence of DNase I buffer and incubated for 1, 2.5 and 5 min. Reactions were terminated by adding 2 μl of stop buffer and incubating at 37°C for 30 min, followed by addition of 1 μl of 10% SDS. Samples were electrophoresed on 1% TBE agarose gels. Gels were stained with ethidium bromide.

While HMO2 has been proposed to interact with a damage-induced histone H2A variant, conservation of its Box B HMG-domain predicts direct DNA interaction. To investigate a potential role in DNA interaction, HMO2 was therefore cloned from yeast genomic DNA and overexpressed in E. coli. HMO2 was judged to be >95% pure using Coomassie blue stained SDS–PAGE gels (Supplementary Figure S1). Consistent with the presence of the classical Box B HMG domain, HMO2 binds both supercoiled and linear DNA (Figure 1) suggesting non-sequence-specific binding as seen for other chromatin-associated HMGB proteins. While the mobility of HMO2-associated linear DNA is gradually reduced as a function of [HMO2] (Figure 1B), the mobility of supercoiled DNA is enhanced at the highest [HMO2] (Figure 1A), suggesting DNA compaction. Note also that ethidium bromide staining of HMO2-bound DNA is enhanced compared to free DNA, perhaps reflecting a DNA unwinding by HMO2 that promotes dye-binding. Figure 1.

Interaction of HMO2 with plasmid DNA. (A, B) Agarose gel retardation of 100 ng plasmid DNA titrated with HMO2. (A) Reactions with supercoiled pGEM5. Lane 1, DNA only, lanes 2–7 with 1.0–6.0 μM HMO2. (B) Reactions with linearized pGEM5. Lane 1, DNA only, lanes 2–6 with 1.0–5.0 μM HMO2. (C) HMO2 supercoils relaxed DNA. Lane 1, 100 ng supercoiled pUC18 DNA. Lane 2, nicked pUC18. Lane 3, nicked pUC18 and T4 DNA ligase. Lanes 4–8, nicked DNA and T4 DNA ligase with 100, 500, 1000, 2000 and 3000 nM HMO2.

HMGB proteins have been reported to promote DNA end-joining in vitro (26,27). The in vivo correlate would be that during DSB repair, HMGB might bring two DNA ends in closer proximity in preparation for either HR or NHEJ. HMO2, however, did not promote the formation of end-joined products in the presence of T4 DNA ligase. Instead, HMO2 prevented end-joining of DNA with a 5′-TA extension (Figure 4A) or DNA with blunt ends (Figure 4B). In Figure 4A and B, lane 2 shows 100 ng (∼4 nM) linearized DNA ligated in presence of T4 DNA ligase. But in presence of HMO2 (0.1–4 µM; lanes 3–8) there are few if any end-joined products. By comparison, 100 ng plasmid DNA (∼2 nM) may be efficiently ligated in the presence of mammalian HMGB1, with optimal ligation efficiency seen on addition of ∼1 µM HMGB1 (26). Notably, lane 9 suggests that the DNA is resistant to exonuclease III digestion in the presence of HMO2 and T4 DNA ligase. The failure to observe ligation products in presence of HMO2 as well as the inability of exonuclease III to digest the DNA suggests that HMO2 may be binding to the ends of the DNA duplex to prevent access to both T4 DNA ligase and exonuclease III. Also note that failure to join DNA ends is not due to HMO2 merely interacting with the ligase to prevent its activity, as evidenced by the activity of T4 DNA ligase on an internal DNA nick in the presence of HMO2 (Figure 1C). Figure 4.

HMO2 prevents ligation of DNA by T4 DNA ligase. (A) DNA with overhangs (5′-TA extensions). (B) DNA with blunt ends. Lanes 1, 100 ng of DNA (∼4 nM, corresponding to ∼8 nM DNA ends). Lane 2, DNA and T4 DNA ligase. Lanes 3–8, DNA, T4 DNA ligase with 100, 500, 1000, 2000, 3000 and 4000 nM HMO2. Lane 9, DNA, T4 DNA ligase, 4000 nM HMO2 and exonuclease III.

The occluded DNA site size for an HMG domain is ∼10 bp (19–23). Since HMO2 has two domains, Box A and Box B, it would be predicted to require at least a 20 bp DNA site; indeed, the yeast HMGB homolog HMO1, which also features two HMG-like domains, binds poorly to 18 bp DNA, but with significantly higher affinity to 26 bp DNA (37). However, our data suggest that HMO2 binds only weakly to 37 bp DNA with loops (data not shown) and with much greater affinity to 50 bp looped DNA (Figure 2). For the Box B domain, which has high homology to mammalian HMGB proteins, interaction with a 10 bp site is a strong prediction. Although the Box A domain differs from consensus HMG sequences, its similarity to the HMO1 Box A domain, which also differs from consensus sequences, yet interacts with an ∼10 bp site, likewise predicts an ∼10 bp site. Based on these assumptions about the site sizes for Box A and Box B, a 50 bp DNA construct would be of sufficient length to accommodate a dimer; this interpretation does not necessarily imply that HMO2 likewise binds perfect duplex DNA as a dimer, as the loops may potentially impose a constraint, such as preferred binding to the 4-nt loop by only one of the two protein domains.

HMGB proteins have been previously shown to bind DNA with loops (40,45). The binding preference for DNA with flexure points is a functional correlate of the DNA bend introduced on interaction with an HMG domain, and the prediction is that HMO2 likewise bends its DNA site. HMGB proteins also bind preferably to four-way DNA junctions (42–44); while HMO2 does bind four-way DNA junctions in preference to a linear DNA duplex (compare Figure 2A and E), the preference is modest. For mammalian HMGB1, binding to four-way DNA junctions is mediated by the Box A domain (44). This selectivity arises from using a hydrophobic residue from helix II of the HMG fold as a bending wedge; the greater preference for distorted DNA is due to stacking of this residue on an exposed base pair. In contrast, mammalian HMGB1 Box B has bending wedges in both helices I and II of the HMG fold and is primarily responsible for DNA bending, whereas its affinity for distorted DNA is only modestly higher than that for perfect duplex DNA. By comparison, HMO2 is also predicted to feature a bending wedge (Leu) in helix II of its Box A domain, suggesting that this domain contributes the binding preference for distorted DNA. For HMO2 Box B, a leucine in helix I would be predicted to be the DNA intercalating residue.

Like other HMGB proteins, HMO2 binds preferentially to supercoiled DNA compared to linear DNA (Figures 1 and 3). The DNA–protein complex with highest [HMO2] migrates faster in the gel (Figure 1, lanes 6 and 7), suggesting that HMO2 might play a role in DNA compaction. This observation is consistent with the ability of HMO2 to introduce supercoils and with the enhanced staining of HMO2-bound DNA by ethidium bromide that may also reflect DNA unwinding by HMO2. While nuclear HMGB proteins have not been shown to compact DNA in vitro as determined by changes in electrophoretic mobility, Abf2p, a protein closely related to the HMG family, has been shown to compact mitochondrial DNA (46). Long-range DNA looping created by protein–protein contacts, however, has been shown to produce more compact DNA structures (47).

Consistent with its preferred binding to negatively supercoiled DNA, HMO2 produces DNA supercoils, due either to DNA underwinding or out-of-plane bending. And as noted above, its preferred binding to DNA loops likely reflects the energetic advantage to bending DNA with flexible joints. In these interactions, HMO2 emulates properties of mammalian HMGB homologs. What is unique about HMO2 is its remarkable ability to protect certain DNA constructs from exonucleolytic cleavage and to prevent their end-joining in presence of T4 DNA ligase, suggesting preferred binding to these DNA ends. DNA with blunt ends is protected efficiently, while in case of DNA with overhangs, the sequence of the single stranded overhang significantly affects HMO2 binding (Figures 5A, B, and 6A). Considering the preferred binding of HMO2 to distorted DNA sites, its interaction with DNA ends may likewise be a manifestation of preferred binding to more deformable duplex regions. Mammalian HMGB1 has been reported to bind DNA ends, but not to protect DNA ends from exonuclease III, perhaps reflecting a more dynamic association with the DNA ends (26). The ability of HMO2 to associate more stably with DNA ends is perhaps attributable to the divergent Box A domain which bears little resemblance to mammalian Box A.

It is also evident that HMO2 requires DNA of a suitable length to bind and protect (Figure 6C). Since HMO2 forms stable complex with plasmid DNA, but not with 50 bp perfect duplex, and since it appears to bind only as a dimer to 50 bp looped DNA, complex stability is likely a consequence of both DNA structure and protein–protein interactions. Such interactions are likely the reason for protection only of DNA long enough for assembly of several protomers at each DNA end. Self-association was also reported for the yeast HMGB homolog HMO1 as well as for mammalian HMGB1, for which residues in the box B domain critical for its ability to stimulate DNA end-joining were also important for box B self-association on supercoiled DNA and the formation of large nucleoprotein complexes (26,31,37).

If HMO2 were to protect DNA from exonuclease III digestion largely by occluding essentially the entire DNA, except for a limited number of base pairs at the ends, then protection from endonuclease digestion might also be seen. However, no such protection is observed (Figure 7). For this interpretation to be considered, one would also expect that all DNA fragments, regardless of sequence of the DNA overhangs, would be protected comparably. This is also not observed, as evidenced by the failure of HMO2 to protect DNA with C- or G-containing overhangs (Figure 6A and B). We also note that ligation at an internal nick occurs in the presence of HMO2 (Figure 1C). For HMO2 to bend internal sites, thereby restricting access to exonuclease III and DNA ligase is likewise unlikely; first, attenuated access to endonuclease would again be expected, as would protection of DNA fragments regardless of sequence of DNA overhangs. If significant out-of-phase bending occurred upon HMO2 binding, this could stiffen DNA and prevent net DNA bending, resulting in preferential generation of linear ligation products in preference to circular products, an effect likely to be most pronounced using shorter DNA fragments as opposed to plasmid DNA. Indeed, facilitated cyclization of short DNA fragments in presence of a DNA-bending protein is commonly used as an assay for DNA bending, without internal DNA bends restricting access of DNA ligase to DNA ends (27,37). In-phase bending might lead to compaction; such compacted DNA may be produced by both HMGB homologs and the bacterial histone-like HU proteins, again without restricting access of DNA ends to DNA ligase (48). Based on these observations and considerations, we therefore conclude that HMO2 exerts its protection against exonuclease III by binding to the DNA ends.

If DNA with complementary overhangs were to anneal, a duplex with nicks would be produced. However, it can also be ruled out that HMO2 merely binds preferentially to such nicked DNA constructs to prevent their ligation, as short DNA with equivalent overhangs may be ligated (Figure 6D). Nicked plasmid DNA may likewise be ligated in the presence of HMO2 (Figure 1C). Indeed, the enhanced ligation of 105 bp DNA in the presence of HMO2 suggests that HMO2 may also bind the ends of the shorter DNA, albeit less stably, as reflected in its inability to protect such duplexes from exonucleolytic digestion. That HMO2 promotes DNA end-joining in the presence of DNA ligase may also be related to the occasional appearance of more slowly migrating DNA species following incubation of linear DNA with both HMO2 and exonuclease III (Figure 6A, lane 8); since these reactions are deproteinized and no ligase is present, we surmise that long single-stranded overhangs generated by exonuclease may invade a complementary duplex in presence of HMO2 to generate a merged structure that is stable to the subsequent manipulations. This possibility is currently under investigation.

Recently, HMO2 was reported to be a part of the ATP-dependent chromatin-remodeling complex INO80 and to be important for recruitment of the INO80 complex to a DSB (3). Arrival of the INO80 complex is predicted to be a part of the early response to DSB formation. In both mammalian and yeast cells, a complex containing the Ku heterodimer and DNA ligase is capable of rejoining DNA DSBs with compatible ends via NHEJ, while DNA with damaged ends requires additional factors (2,4). Furthermore, NHEJ and HR events interfere with each other; for example, a Rad51 filament forms on single-stranded DNA in preparation for HR and impedes NHEJ (49). It is conceivable, therefore, that HMO2, in addition to its role in INO80 recruitment, also contributes to the protection of DNA ends that is essential to preserve the genetic information before DSB repair can be completed. In its complex with INO80 components, HMO2 was reported to interact with phosphorylated histone H2A, indicating that it is not buried in the complex, but accessible for interaction with chromatin. It is therefore likely that it would also be accessible for binding to naked DNA. The preferred binding to DNA with blunt ends or with A+T-containing overhangs may indicate differential functions depending on the type of DNA ends and the DNA repair pathway; for example, yeast Rad51 binds more slowly to single-stranded DNA with AT dinucleotide repeats compared to C- or G-containing dinucleotide repeats (50), perhaps requiring accessory proteins to protect such A+T-rich sequences.