HMGB1-mediated DNA bending: Distinct roles in increasing p53 binding to DNA and the transactivation of p53-responsive gene promoters

Accepted Manuscript HMGB1-mediated DNA bending: Distinct roles in increasing p53 binding to DNA and the transactivation of p53-responsive gene promoters

Michal Štros, Martin Kučírek, Soodabeh Abbasi Sani, Eva Polanská PII: DOI: Reference:

S1874-9399(17)30396-6 https://doi.org/10.1016/j.bbagrm.2018.02.002 BBAGRM 1229

To appear in: Received date: Revised date: Accepted date:

13 November 2017 2 February 2018 2 February 2018

Please cite this article as: Michal Štros, Martin Kučírek, Soodabeh Abbasi Sani, Eva Polanská , HMGB1-mediated DNA bending: Distinct roles in increasing p53 binding to DNA and the transactivation of p53-responsive gene promoters. The address for the corresponding author was captured as affiliation for all authors. Please check if appropriate. Bbagrm(2017), https://doi.org/10.1016/j.bbagrm.2018.02.002

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT

HMGB1-mediated DNA bending: distinct roles in increasing p53 binding to DNA and the transactivation of p53-responsive gene promoters Michal Štros*, Martin Kučírek, Soodabeh Abbasi Sani, and Eva Polanská

RI

PT

Laboratory of Analysis of Chromosomal Proteins, Institute of Biophysics of the Czech Academy of Sciences, Brno, Czech Republic

SC

Abbreviations: HMG, high mobility group; WT, wild-type; oxiHMGB1, oxidized HMGB1; HMGB1C, HMGB1 lacking the acidic C-terminal tail; EMSA, electrophoresis mobility shift assay.

NU

Keywords: HMGB; p53; DNA bending; DNA-protein interaction; promoter transactivation. Corresponding author:

MA

* Tel. +420 517 183 or + 420 777 159 816 (mobile).

D

E-mail address: [email protected]

PT E

ABSTRACT

AC

CE

HMGB1 is a chromatin-associated protein that has been implicated in many important biological processes such as transcription, recombination, DNA repair, and genome stability. These functions include the enhancement of binding of a number of transcription factors, including the tumor suppressor protein p53, to their specific DNA-binding sites. HMGB1 is composed of two highly conserved HMG boxes, linked to an intrinsically disordered acidic Cterminal tail. Previous reports have suggested that the ability of HMGB1 to bend DNA may explain the in vitro HMGB1-mediated increase in sequence-specific DNA binding by p53. The aim of this study was to reinvestigate the importance of HMGB1-induced DNA bending in relationship to the ability of the protein to promote the specific binding of p53 to short DNA duplexes in vitro, and to transactivate two major p53-regulated human genes: Mdm2 and p21/WAF1. Using a number of HMGB1 mutants, we report that the HMGB1-mediated increase in sequence-specific p53 binding to DNA duplexes in vitro depends very little on HMGB1-mediated DNA bending. The presence of the acidic C-terminal tail of HMGB1 and/or the oxidation of the protein can reduce the HMGB1-mediated p53 binding. Interestingly, the induction of transactivation of p53-responsive gene promoters by HMGB1 requires both the ability of the protein to bend DNA and the acidic C-terminal tail, and is promoter-specific. We propose that the efficient transactivation of p53-responsive gene promoters by HMGB1 depends on complex events, rather than solely on the promotion of p53 binding to its DNA cognate sites. 1

ACCEPTED MANUSCRIPT

1. Introduction HMGB1 is an “architectural” chromatin-associated protein, a member of the high mobility group (HMG) superfamily, that has been implicated in many DNA-dependent processes at the chromatin level [1]. The association of HMGB1 with chromatin is not confined to specific sites, but it is rather highly dynamic, and the protein can scan potential binding sites and move

PT

from one chromatin site to another in a “ hit and run” fashion [2]. In addition to the function of HMGB1 in the cell nucleus, the protein can also be released outside the cell and function in

RI

cell signaling, inflammation, and the promotion of tumor development and progression (reviewed in [1, 3-5]). The importance of HMGB1 for life was revealed in knockout

SC

experiments demonstrating that the inactivation of HMGB1 in mice was lethal [6]. HMGB1 is composed of two highly conserved DNA-binding HMG boxes (domains A and B),

NU

each composed of ~80 amino acids, followed by an intrinsically disordered acidic C-terminal tail (reviewed in [1, 7]). HMGB1, like other HMGB-type proteins (HMGB2, HMGB3, and

MA

HMGB4), can act in the nucleus as DNA chaperones, by promoting the formation of complex nucleoprotein structures via protein-protein interactions and the association of distant DNA sites by bending/looping (reviewed in [1]). The function of HMGB1 as a DNA chaperone

D

may depend on the intercalating amino acids phenylalanine/isoleucine and the redox-sensitive

PT E

cysteine residues within the protein ([8-13]; reviewed in [1] and refs. therein). p53 is a potent transcription factor that plays a central role in the regulation of cellular growth. p53 is activated by a number of DNA damaging agents, leading to cell cycle arrest (at the

CE

G1/S phase checkpoint) or to the induction of apoptosis [14]. The inactivation of p53 occurs in a wide variety of human cancers and has been correlated with the presence of p53 mutants

AC

incapable of DNA binding and transcription activation [15]. Thus, sequence-specific DNA binding and transactivation are the key properties that control the biological functions of p53. HMGB1 can directly interact with p53 to facilitate p53 binding (or the binding of another member of the p53 family, p73) to sequence-specific DNA sites in vitro [16-22]. The fact that HMGB1 can enhance the sequence-specific binding of p53 to linear, but not to circular, DNA was proposed as indicative of a requirement for HMGB1-mediated DNA bending to enhance p53 binding to DNA [21]. In this study, we have used a number of HMGB1 mutants to reinvestigate the importance of DNA bending by HMGB1 in enhancing p53 binding to consensus DNA sites in vitro, and for 2

ACCEPTED MANUSCRIPT

the transactivation of two major p53-regulated human genes: Mdm2 (encoding a ubiquitin ligase targeting p53 for degradation) and p21/WAF1 (encoding a protein that inhibits the cell cycle). We report that the in vitro HMGB1-mediated increase in the specific binding of p53 to DNA duplexes is dependent on the acidic C-terminal tail and/or the redox state of HMGB1, rather than on HMGB1-mediated DNA bending. On the other hand, the enhancement of sequence-specific transactivation at p53-responsive gene promoters strongly depends on the

PT

ability of HMGB1 to bend DNA and also requires the acidic C-terminal tail of HMGB1 in a promoter-specific manner. Our results suggest that HMGB1 may augment the transactivation

RI

of p53-responsive gene promoters via a more complex mechanism, rather than a simple

SC

enhancement in p53 binding to DNA cognate sites.

MA

2.1. Plasmids and competitor DNA

NU

2. Materials and methods

The DNA probe used in the electrophoresis mobility shift assay (EMSA) assay was a 58-bp

the

following

oligos

PT E

annealing

D

DNA (containing two p53 binding sites from the human Mdm2 promoter), prepared by (referred

to

as

“p53-DNA

probe”):

top,

5ʹ-

GGTCAAGTTGGGACACGTCCGGCGTCGGCTGTCGGAGGAGCTAAGTCCTGACATG TCT-3ʹ;

low,

5ʹ-AGACATGTCAGGACTTAGCTCCTCCGA-

CE

CAGCCGACGCCGGACGTGTCCCAACTTGACC-3ʹ, as described in [23]. The top and low oligonucleotides for the DNA duplexes containing p53 binding sites, using the p21/WAF1 or GADD45 promoters, were as described in [23] or [21], respectively. Before annealing, the top 32

P-labeled at the 5ʹ-end using T4 polynucleotide kinase. Unlabeled competitor

AC

oligos were

DNA was a 680 bp DNA duplex, derived from the digestion of the pBR322 plasmid with RsaI.

2.2. Purification of p53 Recombinant full-length human p53 or GST-p53 (residues 1-90) were expressed in the BL21 E. coli strain using a two-step induction at 18 °C, to limit protein aggregation. The lysate was loaded onto a 5 ml HiTrap–heparin column (Pharmacia, Stockholm, Sweden) and repurified 3

ACCEPTED MANUSCRIPT

using ammonium sulfate precipitation on a Superdex 200 column (HR 10/30), as detailed in [18]. Purified full-length p53 was kindly provided by Marie Brázdová (Institute of Biophysics, Brno). 2.3. Antibodies A rabbit polyclonal anti-HMGB1 (Abcam, Cambridge, UK, ab18256, ChIP grade) antibody

PT

was specific for the region encompassing residues 150-215 of human HMGB1. The mouse monoclonal anti--actin antibody was purchased from Sigma (St. Louis, MO, A5441). The

RI

monoclonal anti-p53 antibody (ICA-9, epitope encompassing amino acids 388-393) was

NU

2.4. Expression and purification of HMGB1

SC

kindly provided by Marie Brázdová (Institute of Biophysics, Brno).

Untagged or His-tagged isolated HMGB1 A or B domains and mutants were expressed in E.

MA

coli BL21(DE3) cells from plasmids encoding rat cDNAs (the amino acid sequence of the expressed rat HMGB1 is identical to that of the human protein) [20, 24]. Untagged wild-type HMGB1

(residues

1-214),

HMGB1

mutants

(HMGB1(F37A),

D

full-length

PT E

HMGB1(F102A/I111A) or HMGB1(F37A/F102A/I111A)), as well as truncated HMGB1 proteins (HMGB1C(residues 1-185) or HMGB1C(F37A)) were expressed from the pETite-N-His-SUMO plasmid prepared using the “Expresso T7 SUMO cloning and expression system” (Lucigen, Middleton, WI). All mutations were introduced using the “Site-

CE

directed mutagenesis kit” (Thermo Fisher, Waltham, MA). The HMGB sequences and the introduced mutations were verified using dideoxy-sequencing of the final plasmid constructs

AC

on both DNA strands. Purified HMGB1, mutants, or the isolated domain A were oxidized under mild conditions to promote disulfide bond formation, as detailed in [12]. Briefly, proteins were dialyzed overnight against DB buffer (0.15 M NaCl, 20 mM HEPES pH 7.9, 1 mM PMSF) containing 5 M CuCl2, followed by another dialysis cycle against DB buffer lacking CuCl2. In some experiments, the oxidized HMGB1 protein was fully reduced by treatment with 10 mM DTT at 30 °C for 30 min. The concentrations of reduced or oxidized HMGB1 proteins or domains were determined from SDS-PAGE gels using the Coomassie brilliant blue G-250 protein assay (Bio-Rad, Hercules, CA) and BSA as a standard (to determine the protein concentration, all analyzed samples were boiled in SDS-sample buffer containing 10 mM DTT). Concentrations of HMGB1 and some of the deletion mutants 4

ACCEPTED MANUSCRIPT

determined by the above method were also compared with those obtained from amino acid analyses (on average < 10% difference upon normalization, unpublished).

2.5. MALDI-TOF mass spectrometry

PT

The formation of a disulfide bond between cysteine residues 23 and 45 within HMGB1 upon mild oxidation was confirmed by MALDI-TOF mass spectrometry, as previously reported

SC

RI

[12, 13].

2.6. DNA circularization assay

NU

DNA circularization assays were performed as previously described [12, 13, 24, 25]. Briefly, the 32P-labeled 123-bp (AvaI ends) DNA duplexes (~1 nM) were ligated using T4 DNA ligase

MA

(0.05 units, New England Biolabs, Hitchin, UK) in the presence or absence of HMGB1 proteins (the exact HMGB1 concentrations are indicated in the Legends to Figures) at 30 °C for 40 min. Termination of ligation and proteinase K treatment was performed as previously

D

described [24]. Protein-free DNA samples were resolved on pre-run 5% polyacrylamide gels

PT E

in 0.5x TBE buffer (at 250 V, for 4 h at 4 °C) and DNA was visualized and quantified from dried gels on a Typhoon SLA9000 PhosphorImager (GE, Boston, MS).

CE

2.7. DNA-p53 binding studies

AC

EMSA was carried out using ~10 nM of a

32

P-labeled 58-bp DNA duplex, derived from the

human Mdm2 promoter [23]. In some EMSA experiments, we have also used ~5 nM

32

P-

labeled DNA duplexes derived from the GADD45 [17] or p21 promoters [23]. Binding experiments contained 20 ng of purified recombinant human p53 in the presence of absence of HMGB1 or its mutants (typically 0-30 M, the exact protein concentrations are indicated in the Legends to Figures) in a total volume of 20 μl in 1 x EMSA buffer (25 mM KCl, 10 mM HEPES pH 7.6, 0.5 mM EDTA, 0.01% Triton X-100, 100 μg/ml acetylated BSA, 10% glycerol, and 0.5 mM DTT). For EMSA experiments using oxidized HMGB1, 50 M DTT was included in the EMSA buffer, with no detectable reduction in oxidized HMGB1. Reaction mixtures containing

32

P-labeled DNA, unlabeled linear 680-bp competitor dsDNA 5

ACCEPTED MANUSCRIPT

(200 ng per reaction) and proteins were preincubated on ice for 20 min. The reaction mixtures were finally loaded on pre-run 5% polyacrylamide gels (29:1 acrylamide/N, Nʹmethylenebisacrylamide) in 0.5x TBE at 125 V for 4-9 h (4 °C). Following electrophoresis, the gels were dried, and the DNA was visualized and quantified using the Typhoon SLA9000 PhosphorImager (GE). The scoring of the effects of the HMGB1 constructs on p53 DNA binding in the EMSA experiments was performed from at least 4 different gels for each of the

RI

PT

mutants.

SC

2.8. Luciferase reporter assays and western blotting

Human osteosarcoma SKOV-3 (p53−/−) cells were grown in McCoy's 5A medium,

NU

supplemented with 15% heat inactivated fetal calf serum and antibiotics. Cells were maintained at 37 °C, in a humidified atmosphere containing 5% CO2. Transient transfections were carried out in 24-well plates, with cells at a density of 5 x 104 cells/well (~90%

MA

confluence) using 1 l of X-tremeGENE™ HP DNA transfection reagent (Roche, Basel, Switzerland) per well. For each well, 300 ng of the phCMV1 vectors encoding HMGB1,

D

HMGB1(F37A), HMGB1(F37A/F102A), HMGB1(F37A/F102A/I111A), HMGB1C, or

PT E

HMGB1C(F37A) were mixed with 30 ng of p53wt-phCMV plasmid and 100 ng of pGL3Mdm2-Luc or pGL3-p21-Luc plasmids (pGL3-Basic containing firefly Photinus pyralis cDNA from Promega (Madison, WI), linked to sequences derived from the human Mdm2 or

CE

p21 promoters) [19]. The total DNA amount was adjusted to that of the vector DNA and was equal in all wells (500 ng). Luciferase activity was measured in triplicate and normalized by

AC

determining the protein concentrations using the Coomassie G-250 protein-dye assay (BioRad). Proteins from cellular lysates of transfected cells were resolved on SDS/15% polyacrylamide gels and transferred onto PVDF membranes (Bio-Rad). Endogenous HMGB1 and the plasmid-encoded HMGB1 or the mutants were detected using a 1:1000 dilution of the affinity-purified rabbit polyclonal anti-HMGB1 antibody (ab18256-ChIP grade). The -actin levels were measured to demonstrate the equal loading of all samples.

6

ACCEPTED MANUSCRIPT

3. Results 3.1. DNA bending by the isolated HMGB1 A and B domains cannot satisfactorily explain the increase in the sequence-specific binding of p53 to DNA in vitro Based on published HMG box-DNA structures (reviewed by [1] and refs. therein), it clearly

PT

appears that the binding of HMGB1 to DNA results in the intercalation of the bulky hydrophobic amino acid residues in HMG-boxes (referred to as HMGB1 domains) between

RI

successive base-pairs within the DNA minor groove. This is accompanied by a partial unwinding and widening of the minor groove, and a bending towards the major groove.

SC

HMGB1 has three intercalating residues: the “primary” intercalating phenylalanine F102 and the “secondary” intercalating isoleucine I111 in the B domain, and a “secondary” intercalating

NU

F37 residue in the A domain (Fig. 1). Apart from mutations in intercalating amino acid residues, we have recently shown that the oxidation of cysteine residues C22 and C44 within

MA

the HMGB1 A domain may also significantly affect the ability of the protein to bend DNA [12, 13].

Previous studies have indicated that HMGB1 and its isolated domains could facilitate the

D

specific binding of p53 to DNA duplexes derived from the GADD45 promoter in vitro [17,

PT E

21, 22]. The fact that HMGB1 could enhance the sequence-specific binding of p53 to linear, but not to circular, DNA was proposed as indicative of a requirement for the HMGB1mediated DNA bending to increase the binding of p53 to DNA [21]. In the present study, we

CE

have reinvestigated the importance of the HMGB1-induced DNA bending in promoting the

bend DNA.

AC

specific binding of p53 to DNA in vitro using HMGB1 mutants that differ in their abilities to

We have performed site-directed mutagenesis of intercalating residues in HMGB1, individual HMGB1 domains (A and B) or in HMGB1 lacking the acidic C-terminal tail (HMGB1C) (Fig. 1). These HMGB1 proteins/peptides have been expressed in E. coli (see Materials and Methods). To study the role of the redox state of HMGB1 in increasing p53 binding to DNA in vitro, HMGB1 and the peptides were subjected to a reversible mild oxidation, using dialysis against low concentrations of Cu2+ (see Materials and Methods). The formation of an intramolecular disulfide bond between the opposing C22 and C44 residues was confirmed

7

ACCEPTED MANUSCRIPT

through MALDI-TOF mass spectrometry of trypsin-digested samples, as previously described [12, 13]. A number of HMGB1 mutants have been used to address the importance of DNA bending for increasing p53-specific DNA binding in vitro using EMSA. HMGB1 or its isolated A or B domains were pre-incubated with p53, followed by the addition of a short

32

P-labeled DNA

PT

duplex containing the p53-binding site. Protein-DNA complexes were finally subjected to non-denaturing gel electrophoresis. The amount of p53 was kept very low so that only traces

RI

of p53 bound to DNA in the absence of HMGB1 or its domains (Fig. 2). In agreement with previous reports [14, 15, 18, 19], the addition of HMGB1 could enhance p53 binding to DNA,

SC

as evidenced by the formation of one or two discrete bands of lower mobility (Fig. 2A). Moreover, p53 was detected within these supershifted complexes using an anti-p53 antibody

NU

(Supplementary Fig. 1). HMGB1 was excluded from these complexes (designated as bands 1 and 2 in Supplementary Fig. 1), as no supershift was observed upon incubation with an anti-

MA

HMGB1 antibody. HMGB1C, rather than the full-length HMGB1, was more prone to forming two p53-DNA complexes of different mobility when a short DNA duplex derived from the Mdm2 promoter was used in EMSA (Supplementary Fig. 1). Previous reports have

D

demonstrated that p53 could interact with free HMGB1 in solution ([20, 22, 26] and refs.

PT E

therein). However, upon “delivery” of p53 to its DNA binding site, HMGB1 protein most likely dissociates from the “ternary complex” which is most likely of transient nature. The only report demonstrating the existence of the stable “ternary complex” p53-HMGB1-DNA

CE

originated from experiments using hemicatenated DNA lacking p53-binding sites ([20], see also the Discussion).

AC

We have also observed that HMGB1C was ~4-fold more efficient than the full-length HMGB1 in enhancing the specific binding of p53 to short DNA duplexes containing p53binding sites (Supplementary Fig. 1). Thus, the acidic C-terminal tail of HMGB1 could significantly reduce the ability of the protein to promote the specific binding of p53 to DNA in vitro (see also [21]). Furthermore, we have shown that the isolated A domain of HMGB1 was more efficient in enhancing the specific binding of p53 to DNA than the full-length HMGB1 or the isolated B domain (Fig. 2A, see also [21]). In addition, neither the oxidation nor the mutation of phenylalanine 37 (F37) to alanine in domain A could reduce the formation of p53-DNA 8

ACCEPTED MANUSCRIPT

complexes (Fig. 2B, top panels), even though both these modifications abolished the ability of the A domain to bend DNA (Fig. 2B, low panels). Interestingly, both the isolated B domain lacking either the N/C-terminal flanking sequences (designated as domain Bmin, Fig. 1) or the N-terminal flanking sequence (designated as domain Bʹ, Fig. 1), could promote p53 binding to DNA (Fig. 2A), despite the fact that these

PT

two peptides are incapable of DNA bending [24]. In addition, the F102A/I111A mutations in the B domain could not reduce the increase in p53 binding (Fig. 2C, top panel), although this

RI

B domain mutant had lost the ability to bend DNA (Fig. 2C, low panel). Although our original EMSA experiments have been performed using short DNA duplexes derived from the

SC

GADD45 promoter (Fig. 2), a similar conclusion was reached using short DNA duplexes derived from the Mdm2 or p21 promoters (data not shown). Collectively, the above data

NU

demonstrate that the A or B HMGB1 domains could promote the specific binding of p53 to DNA, even when their abilities to bend DNA were severely compromised. These results

MA

suggest that HMGB1 could modulate p53 binding to DNA independent of its DNA bending

D

ability.

PT E

3.2. The HMGB1-mediated increase in p53 DNA binding in vitro and the HMGB1 DNA bending ability are not correlated EMSA and DNA circularization experiments (Fig. 2) have demonstrated that the isolated A or B HMGB1 domains could efficiently promote the specific binding of p53 to short DNA

CE

duplexes, even when their DNA bending ability was abolished. HMGB1 contains two covalently interlinked domains, A and B (Fig. 1), that can cooperate or act independently in

AC

DNA bending, which is further modulated by the acidic C-terminal tail (reviewed in [1]). Thus, it may be possible that the results obtained with isolated HMGB1 domains or mutants (Fig. 2) could differ from those obtained using full-length HMGB1. From this reason, our next set of EMSA experiments were performed with the full-length HMGB1 and its mutants. We first analyzed the importance of the A domain in the context of the full-length HMGB1, regarding its ability to promote p53 binding to DNA, by studying the impact of the F37A mutation. In our EMSA experiments, negligible amounts of p53 bound to DNA in the absence of HMGB1 (Fig. 3A). The addition of HMGB1 could stimulate the specific binding of p53 to DNA and the binding was dependent on the amount of HMGB1 added. Mutating the F37 9

ACCEPTED MANUSCRIPT

residue decreased the ability of HMGB1 to enhance p53 DNA binding by only ~25% (Fig. 3A, right panel), while the same mutation in a HMGB1 construct lacking the acidic Cterminal tail (HMGB1C) did not affect the ability of HMGB1 to promote p53 binding (Fig. 3B). Moreover, we observed that the oxidation of C22 and C44 within the A domain of the full-

PT

length HMGB1 had the strongest negative impact on the ability of HMGB1 to promote p53 binding to DNA in vitro (a ~40% decrease; Fig. 3A, right panel). Furthermore, the removal of

RI

the HMGB1 acidic C-terminal tail fully protected the truncated protein from the negative impact of oxidation (the oxidized protein was even more efficient in enhancing p53 binding;

SC

Fig. 3B, right panel). We have also found that replacing these two cysteine residues with alanines did not affect the ability of HMGB1 to promote p53 binding to DNA (data not

NU

shown, see also the Discussion).

Next, we investigated the combined effect of oxidation and that of the F37 mutation on the

MA

ability of HMGB1 or that of HMGB1 lacking the C-terminal tail (HMGB1C) to increase the specific p53 binding. As shown in Fig. 3A (right panel), the oxidized HMGB1(F37A) mutant

D

could promote p53 binding to the same extent as the wild-type, while the oxidized HMGBC(F37A) mutant was even more efficient than the wild-type HMGB1 in promoting

PT E

this binding (Fig. 3B, right panel). The above data provide evidence that the acidic C-terminal tail of HMGB1 renders the protein more sensitive to structural changes induced by oxidation

CE

[27] or the F37 mutation [22] (see also ref. [13] and the Discussion). To investigate whether the observed differences regarding the impact of oxidation or that of

AC

the F37 mutation on the ability of HMGB1 or HMGB1C to promote the specific binding of p53 to DNA (Figs. 2-3) could be due to differences in HMGB1-mediated DNA bending, we analyzed the DNA bending ability of these proteins using DNA circularization assays. This assay measures the efficiency with which DNA minicircles form during the ligation of DNA fragments below the persistence length (< 150-bp). Since the stiffness of the short DNA fragments prevents the intramolecular alignment of their ends (in the absence of an internal curvature), DNA minicircles can only be detected in the presence of proteins that bend DNA ([24, 28-30], reviewed in [1]). In agreement with previous reports (reviewed in [1]), HMGB1C increased the formation of DNA circles (~2.5-fold, Fig. 4C) when compared to the full-length HMGB1, suggesting that 10

ACCEPTED MANUSCRIPT

the acidic C-terminal tail of HMGB1 negatively modulates the DNA bending ability of the A and B HMGB1 domains. Oxidation or the F37 mutation markedly diminished the DNA bending ability of the full-length HMGB1 (~10-fold). On the other hand, the percentage of DNA circles induced using oxidized HMGB1C was similar to that obtained using the wildtype (un-oxidized) HMGB1, while the ability of the oxidized HMGB1C (F37A) mutant to

PT

promote the formation of DNA circles was only ~2-fold lower than that of the wild-type (Fig. 4C, see also [12, 13]). The above data imply that the acidic C-terminal tail of HMGB1 could

RI

enhance the negative impact of oxidation and that of the F37 mutation within the A domain on the DNA bending ability of HMGB1 (Fig. 4C). Thus, our EMSA (Figs. 2 and 3) and DNA

SC

circularization experiments (Fig. 4) demonstrate that the A domain-induced DNA bending in the context of the full-length HMGB1 cannot explain the HMGB1-mediated increase in the

NU

specific binding of p53 to DNA in vitro. This idea was further supported by EMSA using HMGB1 harboring mutations in both domains (Fig. 5). As shown in Fig. 5A, only a very

MA

small reduction in the formation of p53-DNA complexes was observed using HMGB1 mutated at all three intercalating amino acids (simultaneous mutations within the A domain (F37A) and the B domain (F102A/I111A)), compared with wild-type HMGB1. The latter

D

finding is intriguing given that the HMGB1 (F37A/F102A/I111) triple mutant completely

PT E

lost the ability to bend DNA in our DNA circularization assay (Fig. 5B/C, see also the Discussion).

CE

3.3. Distinct impact of DNA bending induced by HMGB1 or the acidic C-terminal tail of HMGB1 on the transactivation of the p53-responsive Mdm2 and p21 promoters

AC

Previous reports have demonstrated that HMGB1 can transactivate p53-responsive gene promoters during transient transfections [17, 19, 31]. To investigate whether HMGB1mediated DNA bending is required for the p53-mediated transactivation of p53-responsive gene promoters, we performed luciferase gene-reporter assays. Plasmid constructs bearing two p53-regulated genes promoters (Mdm2 or p21) were cotransfected into SKOV-3 (p53−/−) human ovarian cancer cells using plasmids encoding p53, and either the wild-type HMGB1 or HMGB1 mutants with distinct DNA bending abilities. Western blotting using cellular lysates from transfected cells demonstrated that the plasmid-encoded HMGB1 and its mutants were properly expressed (Fig. 6, low panels). As shown in Fig. 6, HMGB1 was ~2-fold less efficient in the p53-dependent transactivation of the Mdm2 promoter compared to that of the 11

ACCEPTED MANUSCRIPT

p21 promoter. Mutations in either the A (F37A) or B domains (F102A/I111A) could only decrease the transactivation of both gene promoters by ~30-40% (Fig. 6). The latter data are in opposition to a previous report, suggesting that only the A domain of HMGB1 is important for the transactivation of p53-responsive promoters [31]. Our data are supported by the finding that only the simultaneous mutation of all three intercalating amino acids within both the A and B domains (F37A/F102A/I111A), which resulted in a HMGB1 protein incapable of

PT

DNA bending in vitro (Fig. 5B/C), could either abrogate the ability of the protein to transactivate the Mdm2 promoter or markedly reduce (by ~ 90%) the transactivation of the

RI

p21 promoter (Fig. 6). The above findings indicate that the efficient transactivation of the

SC

p53-responsive Mdm2 or p21 promoters by HMGB1 strongly depends on the DNA bending ability of both HMGB1 domains (Fig. 6).

NU

Previous luciferase assays have revealed that the co-transfection of plasmids encoding p53 and pG13-luc (a luciferase reporter containing 13 copies of the p53-binding consensus

MA

sequence) with plasmids encoding HMGB1 lacking the C-terminal tail (HMGB1C) into p53-deficient H1299 cells could reduce (but not abolish) promoter transactivation by ~4-fold when compared with that in wild-type HMGB1 [31]. The authors concluded that the acidic C-

D

terminal tail of HMGB1 was necessary for the activation of p53 in vivo [31]. To determine

PT E

any potential promoter specificity regarding the requirement of the acidic C-terminal tail in the transactivation of p53-responsive gene promoters, we assessed the impact of the HMGB1C or the HMGB1C(F37A) mutants on the transactivation of Mdm2 and p21

CE

promoters (Fig. 6). We observed that HMGB1C could not transactivate the p53-responsive Mdm2 promoter, whereas the transactivation of the p21 promoter by HMGB1C was reduced

AC

by ~2-3-fold (relative to that induced by the full-length HMGB1). Moreover, the F37A mutation could further reduce the transactivation potential of HMGB1C on the p53responsive Mdm2 and p21 promoters (Fig. 6).

4. Discussion The functional link between HMGB1 and p53 is emerging from a number of reports demonstrating the impact of a deregulated HMGB1 expression on p53-dependent events such as cell proliferation, apoptosis, or autophagy ([17, 19, 26, 32-34] and refs. therein). One of the roles of HMGB1 in the latter processes likely depends on the ability of the protein to interact 12

ACCEPTED MANUSCRIPT

with p53 (and other specific proteins involved in transcription) and to act as a DNA chaperone by promoting the binding of p53 to DNA by bending/twisting or unwinding specific sequences ([22] and refs. therein). The binding of p53 to its cognate sites and its role in transactivation are the key properties that control the biological functions of the p53 tumor suppressor protein. Human p53 can bind its cognate DNA sites with high cooperativity and bend the DNA towards the major groove (bending angle 25°-52°), with p53 binding on the

PT

inside of the bend [14]. The binding of HMGB1 in the minor groove, on the outside of a bend towards the major groove (DNA bending angle of ~90° [35] and refs. therein), can support a

RI

favorable DNA conformation for the binding of p53 on the inside of the bend [14]. The

SC

HMGB1-mediated DNA bending requires the intercalation of residues F37 (in the A domain) and F102/I111 (B domain) into the minor groove of DNA (reviewed in [1]). The

NU

HMGB1(F37A) mutant and to a lesser degree oxidized HMGB1 may still promote p53 binding to DNA, despite their very limited DNA bending potential ([13] and this paper). The more apparent inhibition of DNA bending and the promotion of DNA binding by p53 in vitro

MA

in the presence of oxidized HMGB1 rather than in that of the HMGB1(F37A) mutant (this paper and [13]) may be explained by a structural change in the A domain [27]. This is due to

D

the formation of a disulfide bridge between C22 and C44, and the flipped ring orientation of F37, restricting the intercalation of F37 into DNA [27]. Although the structures of the

PT E

individual domains A or B of HMGB1 have been previously reported [27, 36-39], no structure of the full-length HMGB1 protein has been published so far. Thus, we can only speculate whether the reported subtle conformational change of the oxidized A domain [27]

CE

might affect the global structure of the full-length HMGB1 protein.

AC

HMGB1-mediated DNA bending, as a mechanism of increasing the specific binding of p53 to DNA could support the observed HMGB1-mediated increase in the specific binding of p53 to linear [17, 19, 21, 22], but not to circular (bent) DNA [21]. However, based on our results, we believe that DNA bending by HMGB1 may not represent a key mechanism by which HMGB1 promotes the specific binding of p53 to short DNA duplexes in vitro. Alternatively, the “residual” abilities of HMGB1 mutants to bend DNA, which cannot be detected using the DNA circularization assay, may suffice to promote p53 binding to DNA in vitro. The latter explanation seems unlikely, as FRET measurements revealed that the DNA bending angle induced by the F37A/F102A/I111A HMGB1 triple mutant was less than 20° (MohanaBorges, R., personal communication), corresponding to a ~4-5-fold decrease relative to the 13

ACCEPTED MANUSCRIPT

DNA bending angle obtained for the wild-type HMGB1 [35]. As p53 consensus sequences are usually bent by ~25°-52° upon p53 binding [14], the “residual” DNA bending potential of the HMGB1 triple mutant may not be enough to efficiently promote p53 binding to short DNA duplexes in vitro. Finally, the previously reported slightly diminished DNA binding affinities of the F37A, F102A/I111A and/or F37A/F102A/I111A HMGB1 mutants relative to that of the wild-type HMGB1 protein [8, 40] cannot explain the markedly impaired capacity

PT

of the HMGB1 mutants to bend DNA (this study).

RI

HMGB1 is the most mobile chromatin-associated protein and only 1 to 2 seconds are required for it to cross the nucleus ([41] and refs. therein). The interaction of HMGB1 with other

SC

proteins could efficiently decrease the mobility of HMGB1, which could affect the ability of the protein to bend DNA and to exercise its role as a DNA chaperone at the chromatin level.

NU

p53-HMGB1 interactions are well documented both in a free solution or within the cell [16, 20, 22, 26]. Although both the A and B domains of HMGB1 can interact with p53 [20, 22],

MA

there are discrepancies regarding the region of p53 involved in HMGB1 binding. While initial reports suggested that the C-terminus of p53 was not involved in mutual interactions [17, 21], other reports revealed that the C-terminal region of p53 (residues 363–376) participated in the

D

interaction with HMGB1 [16]. Irrespective of the binding region through which p53 binds to

PT E

HMGB1, the interaction of HMGB1 with p53 could represent a prevailing mechanism by which HMGB1 enhances the specific binding of p53 to short DNA duplexes in vitro, while the DNA bending capacity of HMGB1 may be dispensable for this function (this paper). The

CE

interaction of HMGB1 with p53 is unlikely to be affected by mutating intercalating amino acids in HMGB1. This may also explain the almost negligible impact of these mutations on

AC

the ability of HMGB1 to promote p53 binding to DNA. On the other hand, structural changes in HMGB1 due to its mild oxidation [27] may significantly affect HMGB1 binding to p53, thus explaining why oxidation, unlike the mutation of intercalating residues, could significantly diminish the HMGB1-mediated increase in p53 binding to DNA in vitro (this report). The function of HMGB1 in the transactivation of p53-responsive gene promoters in vivo may require, in addition to the pre-bending of p53-binding sites and the promotion of p53 binding via direct p53-HMGB1 interactions, other architectural functions of the protein, such as interactions with a plethora of specific proteins (reviewed in [1]). This can lead to the recruitment of other factors (perhaps in a promoter- or cell-specific manner) and facilitate the 14

ACCEPTED MANUSCRIPT

formation of specific DNA-protein complexes, by bringing together distant regulatory sequences and their binding partners through DNA looping via HMG-boxes [10, 42]. However, the possible impact of the HMGB1-induced DNA-looping on p53 binding can only be observed in the context of longer DNA (such as in the context of chromatin within the cell, or in plasmid DNAs with p53-responsive promoters in luciferase assays). Therefore, this cannot represent a mechanism contributing to p53 binding in EMSA experiments using short

PT

DNA fragments. DNA unwinding by HMGB1 can also contribute to the observed HMGB1mediated p53-dependent transactivation of responsive gene promoters ([42, 43] reviewed in

RI

[1]).

SC

The A domain of HMGB1 is less tightly sequestered by the acidic C-terminal tail than the HMGB1 B domain at ‘‘physiological ionic strength’’ [44]. Thus, based on its greater

NU

availability for recruiting p53 [22], and the ability of the HMGB1 B domain to bend DNA (therefore providing a favorable DNA structure to which p53 can bind), a dual chaperone role

MA

was proposed for HMGB1: a protein chaperone role for the A domain, and a DNA chaperone role for the B domain [22]. However, the apparent dual chaperone role of HMGB1 is not supported by our findings. First, both HMGB1 domains can bend DNA (reviewed in [1]) and

D

interact with p53, albeit with a slight preference of the A domain for binding p53 [26] and the

PT E

B domain being slightly more efficient in bending DNA [20, 22, 26]. Second, the oxidation or simultaneous mutation of intercalating residues in the A and B HMGB1 domains, could abrogate the ability of the protein to bend DNA, without significantly affecting its function in

CE

promoting p53 binding to DNA in vitro (this paper). Although our gene reporter assays using HMGB1 lacking the A domain (i.e., the HMGB1 B domain attached to the acidic C-terminal

AC

tail, HMGB1A) indicated that the A domain was critical for the transactivation of p53responsive promoters by HMGB1 in vivo [31], these results may be inconclusive due to the fact that the HMGB1A peptide was found to be unstable in solution due to the loss of secondary and tertiary structure, as revealed by NMR [45]. In addition, the mutation of the intercalating F37 residue in the A domain could only reduce the transactivation of p53responsive promoters by ~30-40%. Further mutations in intercalating residues within the B domain (F102A/I111A) were necessary to (nearly) abrogate the HMGB1-mediated transactivation, possibly via the inhibition of the DNA bending potential of the mutant. Thus, the above data question the conclusions of a previous report, suggesting that only the A domain of HMGB1 is important for the transactivation of p53-responsive promoters [31]. 15

ACCEPTED MANUSCRIPT

Considering the above data, we favor the hypothesis that both the A and B domains of HMGB1 could play simultaneous dual chaperone roles in the facilitation of p53 binding to its cognate sites. This may include the binding of p53 and the delivery of the p53-HMGB1 complex to DNA (likely pre-bent by both HMGB1 domains), as well as the transactivation of p53-responsive promoters, which may require the selective binding of the HMG-boxes and/or the acidic C-terminal tail to other proteins involved in this process. The absence of HMGB1

PT

from the p53-DNA complex in EMSA assays ([17, 21, 22] and this report) may suggest that the p53-HMGB1 complexes are likely disrupted upon encountering specific p53 DNA

RI

binding sites. Alternatively, the p53-HMGB1-DNA complexes may be short-lived.

SC

Another mechanism by which HMGB1 could modulate p53 binding to its cognate sites, and the p53-dependent transactivation in vivo, involves the acidic C-terminal tail of HMGB1 ([31]

NU

and this paper). The acidic C-terminal tail occludes the DNA- and p53-binding faces of the HMGB1 domains and likely regulates their binding to p53, DNA, and other proteins involved

MA

in the transactivation of p53-responsive promoters ([22] and refs. therein). The C-terminal tail of HMGB1 can downregulate the HMGB1-mediated DNA bending and reinforce the negative impact of oxidation and/or mutations in intercalating residues on the ability of the protein to

D

bend DNA and enhance p53 binding to DNA in vitro (this paper). Our finding that the

PT E

transactivation of p53-responsive gene promoters by HMGB1C was significantly decreased (p21) or non-existent (Mdm2) upon the removal of the acidic C-terminal tail from HMGB1 further indicates promoter specificity involving the C-terminal tail for efficient transactivation

CE

(this report and [31]).

In summary, we report that HMGB1 can enhance p53 binding to short DNA duplexes derived

AC

from different p53-responsive gene promoters, even when the ability of HMGB1 to bend DNA is significantly impaired by oxidation or the mutation of intercalating residues. Conversely, the transactivation of p53-responsive gene promoters (Mdm2 or p21) strongly depends on HMGB1-mediated DNA bending (and/or DNA looping) and is regulated by the acidic C-terminal tail of the protein in a promoter-specific manner.

Acknowledgements 16

ACCEPTED MANUSCRIPT

RI

PT

Financial support for this research was provided through the Grant Agency of the Czech Republic (Grant P305-15-01354S awarded to M.Š.). We are grateful to Professor Stephen J. Lippard (MIT, Cambridge) for providing us with plasmids encoding rat HMGB1(F37A) and HMGB1(F102A/I111A) mutants. We would like to acknowledge a kind gift of the monoclonal anti-p53 antibody (ICA-9) and purified human p53 protein from Dr. Marie Brázdová (Institute of Biophysics, Brno). We also thank Professor Ronaldo Mohana-Borges (UFRJ, Rio de Janeiro) and his coworkers for providing us with preliminary data from FRET measurements of HMGB1-induced DNA bending of the mutants used in this study.

SC

References

AC

CE

PT E

D

MA

NU

[1] M. Štros, HMGB proteins: interactions with DNA and chromatin, Biochim Biophys Acta, 1799 (2010) 101-113. [2] G. Gerlitz, R. Hock, T. Ueda, M. Bustin, The dynamics of HMG protein-chromatin interactions in living cells, Biochemistry and cell biology = Biochimie et biologie cellulaire, 87 (2009) 127-137. [3] J.O. Thomas, HMG1 and 2: architectural DNA-binding proteins, Biochem Soc Trans, 29 (2001) 395-401. [4] M.T. Lotze, K.J. Tracey, High-mobility group box 1 protein (HMGB1): nuclear weapon in the immune arsenal, Nat Rev Immunol, 5 (2005) 331-342. [5] R. Kang, Q. Zhang, H.J. Zeh, 3rd, M.T. Lotze, D. Tang, HMGB1 in cancer: good, bad, or both?, Clinical cancer research : an official journal of the American Association for Cancer Research, 19 (2013) 4046-4057. [6] S. Calogero, F. Grassi, A. Aguzzi, T. Voigtlander, P. Ferrier, S. Ferrari, M.E. Bianchi, The lack of chromosomal protein Hmg1 does not disrupt cell growth but causes lethal hypoglycaemia in newborn mice, Nat Genet, 22 (1999) 276-280. [7] M. Štros, D. Launholt, K.D. Grasser, The HMG-box: a versatile protein domain occurring in a wide variety of DNA-binding proteins, Cell Mol Life Sci, 64 (2007) 2590-2606. [8] S.C. Roemer, J. Adelman, M.E. Churchill, D.P. Edwards, Mechanism of high-mobility group protein B enhancement of progesterone receptor sequence-specific DNA binding, Nucleic Acids Res, 36 (2008) 3655-3666. [9] P.C. Billings, R.J. Davis, B.N. Engelsberg, K.A. Skov, E.N. Hughes, Characterization of high mobility group protein binding to cisplatin-damaged DNA, Biochem Biophys Res Commun, 188 (1992) 1286-1294. [10] M. Štros, J. Reich, A. Kolíbalová, Calcium binding to HMG1 protein induces DNA looping by the HMG-box domains, FEBS Lett, 344 (1994) 201-206. [11] S. Park, S.J. Lippard, Redox state-dependent interaction of HMGB1 and cisplatinmodified DNA, Biochemistry, 50 (2011) 2567-2574. [12] E. Polanská, S. Pospíšilová, M. Štros, Binding of histone H1 to DNA is differentially modulated by redox state of HMGB1, PloS one, 9 (2014) e89070. [13] M. Štros, E. Polanská, M. Kučírek, S. Pospíšilová, Histone H1 Differentially Inhibits DNA Bending by Reduced and Oxidized HMGB1 Protein, PloS one, 10 (2015) e0138774.

17

ACCEPTED MANUSCRIPT

AC

CE

PT E

D

MA

NU

SC

RI

PT

[14] A.K. Nagaich, V.B. Zhurkin, S.R. Durell, R.L. Jernigan, E. Appella, R.E. Harrington, p53-induced DNA bending and twisting: p53 tetramer binds on the outer side of a DNA loop and increases DNA twisting, Proceedings of the National Academy of Sciences of the United States of America, 96 (1999) 1875-1880. [15] W.A. Freed-Pastor, H. Mizuno, X. Zhao, A. Langerod, S.H. Moon, R. RodriguezBarrueco, A. Barsotti, A. Chicas, W. Li, A. Polotskaia, M.J. Bissell, T.F. Osborne, B. Tian, S.W. Lowe, J.M. Silva, A.L. Borresen-Dale, A.J. Levine, J. Bargonetti, C. Prives, Mutant p53 disrupts mammary tissue architecture via the mevalonate pathway, Cell, 148 (2012) 244-258. [16] T. Imamura, H. Izumi, G. Nagatani, T. Ise, M. Nomoto, Y. Iwamoto, K. Kohno, Interaction with p53 enhances binding of cisplatin-modified DNA by high mobility group 1 protein, J Biol Chem, 276 (2001) 7534-7540. [17] L. Jayaraman, N.C. Moorthy, K.G. Murthy, J.L. Manley, M. Bustin, C. Prives, High mobility group protein-1 (HMG-1) is a unique activator of p53, Genes Dev, 12 (1998) 462472. [18] M. Brazdova, J. Palecek, D.I. Cherny, S. Billova, M. Fojta, P. Pecinka, B. Vojtesek, T.M. Jovin, E. Palecek, Role of tumor suppressor p53 domains in selective binding to supercoiled DNA, Nucleic Acids Res, 30 (2002) 4966-4974. [19] M. Štros, T. Ozaki, A. Bačíková, H. Kageyama, A. Nakagawara, HMGB1 and HMGB2 cell-specifically down-regulate the p53- and p73-dependent sequence-specific transactivation from the human Bax gene promoter, J Biol Chem, 277 (2002) 7157-7164. [20] M. Štros, E. Muselíková-Polanská, S. Pospíšilová, F. Strauss, High-affinity binding of tumor-suppressor protein p53 and HMGB1 to hemicatenated DNA loops, Biochemistry, 43 (2004) 7215-7225. [21] K. McKinney, C. Prives, Efficient specific DNA binding by p53 requires both its central and C-terminal domains as revealed by studies with high-mobility group 1 protein, Mol Cell Biol, 22 (2002) 6797-6808. [22] J.P. Rowell, K.L. Simpson, K. Stott, M. Watson, J.O. Thomas, HMGB1-facilitated p53 DNA binding occurs via HMG-Box/p53 transactivation domain interaction, regulated by the acidic tail, Structure, 20 (2012) 2014-2024. [23] S. Kaku, Y. Iwahashi, A. Kuraishi, A. Albor, T. Yamagishi, S. Nakaike, M. KuleszMartin, Binding to the naturally occurring double p53 binding site of the Mdm2 promoter alleviates the requirement for p53 C-terminal activation, Nucleic Acids Res, 29 (2001) 19891993. [24] M. Štros, DNA bending by the chromosomal protein HMG1 and its high mobility group box domains. Effect of flanking sequences, J Biol Chem, 273 (1998) 10355-10361. [25] S. Jaouen, L. de Koning, C. Gaillard, E. Muselíková-Polanská, M. Štros, F. Strauss, Determinants of specific binding of HMGB1 protein to hemicatenated DNA loops, J Mol Biol, 353 (2005) 822-837. [26] K.M. Livesey, R. Kang, P. Vernon, W. Buchser, P. Loughran, S.C. Watkins, L. Zhang, J.J. Manfredi, H.J. Zeh, 3rd, L. Li, M.T. Lotze, D. Tang, p53/HMGB1 complexes regulate autophagy and apoptosis, Cancer research, 72 (2012) 1996-2005. [27] J. Wang, N. Tochio, A. Takeuchi, J.I. Uewaki, N. Kobayashi, S.I. Tate, Redox-sensitive structural change in the A-domain of HMGB1 and its implication for the binding to cisplatin modified DNA, Biochem Biophys Res Commun, (2013). [28] T.T. Paull, M.J. Haykinson, R.C. Johnson, The nonspecific DNA-binding and -bending proteins HMG1 and HMG2 promote the assembly of complex nucleoprotein structures, Genes Dev, 7 (1993) 1521-1534.

18

ACCEPTED MANUSCRIPT

AC

CE

PT E

D

MA

NU

SC

RI

PT

[29] P.M. Pil, C.S. Chow, S.J. Lippard, High-mobility-group 1 protein mediates DNA bending as determined by ring closures, Proceedings of the National Academy of Sciences of the United States of America, 90 (1993) 9465-9469. [30] S.H. Teo, K.D. Grasser, J.O. Thomas, Differences in the DNA-binding properties of the HMG-box domains of HMG1 and the sex-determining factor SRY, Eur J Biochem, 230 (1995) 943-950. [31] S. Banerjee, T.K. Kundu, The acidic C-terminal domain and A-box of HMGB-1 regulates p53-mediated transcription, Nucleic Acids Res, 31 (2003) 3236-3247. [32] A.R. Davalos, M. Kawahara, G.K. Malhotra, N. Schaum, J. Huang, U. Ved, C.M. Beausejour, J.P. Coppe, F. Rodier, J. Campisi, p53-dependent release of Alarmin HMGB1 is a central mediator of senescent phenotypes, J Cell Biol, 201 (2013) 613-629. [33] Q. Li, J. Li, T. Wen, W. Zeng, C. Peng, S. Yan, J. Tan, K. Yang, S. Liu, A. Guo, C. Zhang, J. Su, M. Jiang, Z. Liu, H. Zhou, X. Chen, Overexpression of HMGB1 in melanoma predicts patient survival and suppression of HMGB1 induces cell cycle arrest and senescence in association with p21 (Waf1/Cip1) up-regulation via a p53-independent, Sp1-dependent pathway, Oncotarget, 5 (2014) 6387-6403. [34] A.J. Bagherpoor, D. Dolezalova, T. Barta, M. Kučírek, S.A. Sani, M. Esner, M. Kunová Bosaková, V. Vinarsky, L. Peskova, A. Hampl, M. Štros, Properties of Human Embryonic Stem Cells and Their Differentiated Derivatives Depend on Nonhistone DNA-Binding HMGB1 and HMGB2 Proteins, Stem cells and development, 26 (2017) 328-340. [35] F.S. Belgrano, I.C. de Abreu da Silva, F.M. Bastos de Oliveira, M.R. Fantappie, R. Mohana-Borges, Role of the acidic tail of high mobility group protein B1 (HMGB1) in protein stability and DNA bending, PloS one, 8 (2013) e79572. [36] H.M. Weir, P.J. Kraulis, C.S. Hill, A.R. Raine, E.D. Laue, J.O. Thomas, Structure of the HMG box motif in the B-domain of HMG1, Embo J, 12 (1993) 1311-1319. [37] R.W. Broadhurst, C.H. Hardman, J.O. Thomas, E.D. Laue, Backbone dynamics of the Adomain of HMG1 as studied by 15N NMR spectroscopy, Biochemistry, 34 (1995) 1660816617. [38] R. Sanchez-Giraldo, F.J. Acosta-Reyes, C.S. Malarkey, N. Saperas, M.E. Churchill, J.L. Campos, Two high-mobility group box domains act together to underwind and kink DNA, Acta crystallographica. Section D, Biological crystallography, 71 (2015) 1423-1432. [39] C.H. Hardman, R.W. Broadhurst, A.R. Raine, K.D. Grasser, J.O. Thomas, E.D. Laue, Structure of the A-domain of HMG1 and its interaction with DNA as studied by heteronuclear three- and four-dimensional NMR spectroscopy, Biochemistry, 34 (1995) 16596-16607. [40] Y. Jung, S.J. Lippard, RNA polymerase II blockage by cisplatin-damaged DNA. Stability and polyubiquitylation of stalled polymerase, J Biol Chem, 281 (2006) 1361-1370. [41] A. Agresti, P. Scaffidi, A. Riva, V.R. Caiolfa, M.E. Bianchi, GR and HMGB1 interact only within chromatin and influence each other's residence time, Mol Cell, 18 (2005) 109121. [42] M. Štros, J. Štokrová, J.O. Thomas, DNA looping by the HMG-box domains of HMG1 and modulation of DNA binding by the acidic C-terminal domain, Nucleic Acids Res, 22 (1994) 1044-1051. [43] L.G. Sheflin, S.W. Spaulding, High mobility group protein 1 preferentially conserves torsion in negatively supercoiled DNA, Biochemistry, 28 (1989) 5658-5664. [44] K. Stott, M. Watson, F.S. Howe, J.G. Grossmann, J.O. Thomas, Tail-mediated collapse of HMGB1 is dynamic and occurs via differential binding of the acidic tail to the A and B domains, J Mol Biol, 403 (2010) 706-722. [45] M. Carballo, P. Puigdomenech, T. Tancredi, J. Palau, Interaction between domains in chromosomal protein HMG-1, EMBO J, 3 (1984) 1255-1261. 19

ACCEPTED MANUSCRIPT

Legends to Figures

MA

NU

SC

RI

PT

Fig. 1. HMGB1 protein, mutants, and domains. HMGB1 organization (amino acid residues 1214, the numbering does not include the N-terminal methionine; A, A domain; B, B domain; the acidic C-terminal tail is depicted as a gray oval). Wild-type HMGB1 and mutants. HMGB1-lacking the acidic C-terminal tail (HMGB1ΔC, amino acid residues 1-184) and mutants. Reduced A domain (amino acids 1-83), oxidized A domain and mutant A domain. Different peptides corresponding to the isolated B domain: the Bmin domain (amino acids 91161), the Bʹ domain (amino acids 91-184), B domain (amino acids 86-184), and the F102A/I111A B domain mutant. Positions of mutations and/or oxidation are indicated. F, phenylalanine; A, alanine; I, isoleucine; C, cysteine. F/F/I, residues F37, F102 and I111 mutated to alanine. Experimental data obtained with individual HMGB1, mutants or isolated domains are summarized on the right. ++++, very strong; +++, strong; ++, intermediate; +, low; +/-, very weak (if any); −, no effect; n.d., not determined.

AC

CE

PT E

D

Fig. 2. Enhancement of p53-specific binding to DNA in vitro by HMGB1 and its domains. (A) Increase in p53 binding to DNA, mediated by HMGB1, HMGB1 lacking the acidic Cterminal tail (HMGB1C) and individual HMGB1 domains. HMGB1 proteins (0-30 M, as detailed in the figure) were preincubated with human p53 (20 ng), followed by the addition of a 32P-labeled DNA duplex (~5 nM) containing a p53-binding site derived from the GADD45 promoter (EMSA). (B) Enhancement of p53 binding to DNA, mediated by the A domain of HMGB1 (EMSA, top panel). WT, A domain (reduced); oxi, oxidized A domain; F37, mutated A domain (phenylalanine 37 to alanine). Different HMGB1 A domain peptides (1, 2, and 6 μM, left to right) were preincubated with p53 (20 ng), followed by the addition of a 32Plabeled DNA duplex (5~ nM) containing a p53-binding site derived from the GADD45 promoter. (C) Enhancement of p53 binding to DNA, mediated by the B domain of HMGB1 (EMSA, top panel). WT, B domain; F/I, B domain mutated at F102 and I111 to alanine. HMGB1 B domain peptides (1, 2, and 6 μM, left to right) were preincubated with p53, as indicated in panel B. Protein-DNA complexes from EMSA assays were resolved on 5% polyacrylamide gels. (B) and (C), low panels: DNA circularization assay using the A and B HMGB1 domains. A 32P-labeled 123-bp DNA duplex (~1 nM) was pre-incubated with 1, 2, and 6 μM HMGB1 domain peptides (left to right), followed by ligation using T4 DNA ligase. Deproteinized DNA samples were separated by electrophoresis on 5% non-denaturing polyacrylamide gels as detailed in Materials and Methods.

20

ACCEPTED MANUSCRIPT

MA

NU

SC

RI

PT

Fig. 3. Enhancement of p53 binding to DNA, induced by HMGB1, HMGB1C, and mutants. (A) Enhancement of p53 binding to DNA by the full-length HMGB1 or the HMGB1 (F37A) mutant. (EMSA, left panel): p53 was preincubated with increasing amounts (1, 2, and 3 M, left to right) of the full-length HMGB1 or the HMGB1 (F37A) mutant (reduced or oxidized, as indicated), followed by the addition of a 32P-labeled DNA duplex (~5 nM) derived from the Mdm2 promoter. Protein-DNA complexes from EMSA assays were resolved on 5% polyacrylamide gels (left panel) and dried gels were quantified using PhosphorImager (right panel). (B) Enhancement of p53 binding to DNA, mediated by HMGB1C or the HMGB1 (F37A) mutant. (EMSA, left panel): p53 was preincubated with increasing amounts (1, 2, and 3 M, left to right) of HMGB1C or the HMGB1C (F37A) mutant (reduced or oxidized, as indicated), followed by the addition of a 32P-labeled DNA duplex (~5 nM) derived from the Mdm2 promoter. Protein-DNA complexes from EMSA assays were resolved on 5% polyacrylamide gels (left panel) and dried gels were quantified using PhosphorImager (right panels). The abundance of p53-DNA complexes, obtained at a 3 M concentration of the wild-type HMGB1 protein (WT) or HMGB1C was regarded as 100% (dashed lines). Data are analyzed using Tukey's multiple comparison test (not significant, ns; P<0.001***) and represent the mean (SD) for three independent experiments. F37, HMGB1 protein or peptide mutated at phenylalanine 37 to alanine; oxiF37, oxidized HMGB1 protein or peptide mutated at phenylalanine 37 to alanine.

AC

CE

PT E

D

Fig. 4. The effect of oxidation and mutation of F37 in HMGB1 or HMGB1 lacking the acidic C-terminal tail (HMGB1C) on DNA bending. The 5ʹ-end 32P-labeled 123-bp DNA fragment (~1 nM) was preincubated with 10, 25, 50, and 100 nM (left to right) of HMGB1 (panel A), HMGB1C (panel B) or mutants, followed by ligation using T4 DNA ligase to form DNA minicircles. Deproteinized DNA samples were separated by electrophoresis on 5% nondenaturing polyacrylamide gels. (C), percentage of DNA circles formed by HMGB1, HMGB1C, or mutants in panels A and B. The percentage of the minicircles formed at 100 nM HMGB1 was arbitrarily set to 100%. The percentage of DNA circles was calculated from panels A and B (each of the curves represents an average of three independent DNA circularization experiments). WT, wild-type HMGB1 or HMGB1C; F37A, HMGB1 or HMGB1C mutated at phenylalanine 37 to alanine; oxiF37A, oxidized HMGB1 or HMGB1C mutated at phenylalanine 37 to alanine.

Fig. 5. Enhancement of the specific binding of p53 to DNA in vitro, mediated by HMGB1 and mutants, and comparison of their DNA bending abilities. (A) Increase in p53 binding to DNA, induced by HMGB1 and mutants. HMGB1 proteins (1, 2, and 3 M, left to right) were preincubated with p53 (20 ng), followed by the addition of a 32P-labeled DNA duplex (10 nM) containing a p53-binding site derived from the Mdm2 promoter. Specific p53-DNA complexes were resolved on polyacrylamide gels (EMSA) and quantified using PhosphorImager. The values are expressed relative to 100%, which corresponds to the abundance of p53-DNA complexes obtained at the highest concentration of wild-type 21

ACCEPTED MANUSCRIPT

PT

HMGB1. (B), DNA circularization assay using HMGB1 and mutants. Experiments were performed as in Fig. 5A/B. WT, wild-type. F37, HMGB1 mutated at phenylalanine 37 to alanine; F/I, HMGB1 mutated at phenylalanine102 and isoleucine 111 to alanine; F37/F/I; HMGB1 mutated at phenylalanines 37/102, and isoleucine 111, to alanine. (C), Percentage of DNA circles in panel B, quantified by PhosphorImager. Each of the curves represents an average of three independent DNA circularization experiments. The percentage of DNA circles observed when using 100 nM of wild-type HMGB1 was set to 100%.

AC

CE

PT E

D

MA

NU

SC

RI

Fig. 6. Transactivation of Mdm2 and p21 promoters by HMGB1 and mutants. Human ovarian SKOV-3 (p53−/−) cancer cells were cotransfected with Mdm2-luciferase (left panel) or p21luciferase (right panel) reporter plasmids, plasmid constructs encoding p53, and either the empty vector or different amounts of plasmids encoding the wild-type HMGB1, HMGB1C, or mutants, as detailed in Materials and Methods. All transfections and luciferase assays were performed in triplicate and represent an average of four independent experiments. The luciferase activity from cells transfected with reporter and p53 plasmids only (in the absence of HMGB1) was arbitrarily set as 100%. The plasmid-derived expression of the HMGB1 proteins (and that of endogenous HMGB1, depicted as HMGB1 in the figure) was determined in cellular lysates by western blotting (low panels). recHMGB1, plasmid-based expression of HMGB1 and mutants; WT, wild-type recHMGB1 or recHMGB1C; F37, recHMGB1 mutated at phenylalanine 37 to alanine; F/I, recHMGB1 mutated at phenylalanine 102 and isoleucine 111 to alanine; F/F/I; recHMGB1 mutated at phenylalanines 37/102 and isoleucine 111 to alanine; C or HMGB1C, recHMGB1 lacking the acidic C-terminal tail (the electrophoretic mobility of the HMGB1 peptide is marked by an asterisk).

22

ACCEPTED MANUSCRIPT

Highlights

Sequence-specific p53 binding to short DNA duplexes in vitro depends very little on HMGB1-mediated DNA bending. Transactivation of p53-responsive gene promoters by HMGB1 in the cell is promoter-specific, and requires acidic C-tail and the ability of both HMGB1 domains, A and B, to bend DNA.

AC

CE

PT E

D

MA

NU

SC

RI

PT

Our data suggest that efficient transactivation of p53-responsive gene promoters by HMGB1 depends on complex events, rather than solely on the promotion of p53 binding to its DNA cognate sites.

23

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6