p53 Binds Preferentially to Non-B DNA Structures Formed by the Pyrimidine-Rich Strands of GAA·TTC Trinucleotide Repeats Associated with Friedreich’s Ataxia

Expansions of trinucleotide repeats (TNRs) are associated with genetic disorders such as Friedreich’s ataxia. The tumor suppressor p53 is a central regulator of cell fate in response to different types of insults. Sequence and structure-selective modes of DNA recognition are among the main attributes of p53 protein. The focus of this work was analysis of the p53 structure-selective recognition of TNRs associated with human neurodegenerative diseases. Here, we studied binding of full length p53 and several deletion variants to TNRs folded into DNA hairpins or loops. We demonstrate that p53 binds to all studied non-B DNA structures, with a preference for non-B DNA structures formed by pyrimidine (Py) rich strands. Using deletion mutants, we determined the C-terminal DNA binding domain of p53 to be crucial for recognition of such non-B DNA structures. We also observed that p53 in vitro prefers binding to the Py-rich strand over the purine (Pu) rich strand in non-B DNA substrates formed by sequence derived from the first intron of the frataxin gene. The binding of p53 to this region was confirmed using chromatin immunoprecipitation in human Friedreich’s ataxia fibroblast and adenocarcinoma cells. Altogether these observations provide further evidence that p53 binds to TNRs’ non-B DNA structures.


Introduction
Wild type p53 (p53) is an important transcription factor that plays a significant role in the prevention of tumor cells development and is widely known as "the guardian of the genome and epigenome" [1,2]. As a response to potentially oncogenic conditions, the tumor suppressor p53 protein Formation of non-B DNA structures inside TNR sequences is correlated to TNR expansion, that is in turn associated with the progression of a range of human diseases [26]. Previous studies of CTG·CAG tracts (11 repeats) showed that p53 exhibits preferential recognition of hairpin DNA over their linear forms, canonical (CTG·CAG) duplexes or mismatched (CTG·CTG; CAG·CAG) duplexes [18]. In our study, we focused on GAA·TTC TNRs associated with Friedreich's ataxia, on differences in p53 recognition of the pyrimidine and purine strands of the GAA·TTC repeat and also on comparison to other repeats. We examined TNR sequences containing 11 repetitions, which is in the normal, non-disease range for the Friedreich's ataxia region [16]. TNR sequences were locked into a non-B DNA form (NB), as previously described for the CTG·CAG repeat [18] and schematically portrayed in Figure 1 and Table S1. This non-B DNA form represents a slipped strand intermediate and may contain several non-B DNA structural motifs: three-way junctions, hairpins, and mismatched duplexes formed by self-annealed TNR strands or loop structures. The binding of p53 protein was examined for each strand of GAA·TTC and CTG·CAG repeats locked within non-B DNA structures, by EMSA ( Figure 1A, lanes [4][5][6][7][8][9][10][11][12][13][14][15] and ELISA ( Figure 1C). To compare the amount of p53 bound to Py-strand (TTC NB , Figure 1A, lanes [4][5][6] with protein bound to Pu-strand (GAA NB , Figure 1A, lanes 7-9), unbound free DNA (usually forming sharper bands, the intensities of which are easier to evaluate, compared to often more smeary bands of the bound DNA fraction) was monitored. The p53 preference for TTC NB was evident from the amount of free DNA after p53 binding ( Figure 1A, compare lane 6 to lane 9). Similarly, as in the case of CAG·CTG repeats and in agreement with a previous study [18], CTG hairpin (CTG NB ) was preferred substrate for p53 protein over CAG hairpin form ( Figure 1A, compare lanes 12 and 15). A relatively strong difference in p53 interaction with Pu-and Py-strands of repeat was also observed with the T NB ( Figure 1B, lanes [10][11][12] and A NB ( Figure 1B, lanes [13][14][15], which were "hairpin-like" structures formed by (T/A) 33 tracks locked in a loop in the same way as the TNR sequences. Within all of this data, it is important to note that p53 bound the most strongly to the p53 consensus sequence (CON) ( Figure 1A lanes 1-3). To better characterize differences in affinities of p53 to different non-B DNA structures and CON, we used ELISA ( Figure 1C) followed by quantitation using a specific DO1 antibody as was recently described for p53-telomeric quadruplex DNA-binding [6]. With this system ( Figure 1C), we confirmed the EMSA data that p53 binds to TTC NB and T NB with higher affinity than to an equivalent GAA NB and A NB .
To roughly check the proper formation of designed structures, i.e., TNR hairpin or unfolded region within a double-helical DNA, we performed enzymatic footprint of the structures ( Figure 1D,E). T7 endonuclease I, specific for mismatched base-pairs and junctions, gave a strong signal at the junction between the TNR part and the locked terminal B-DNA part for both TTC NB and GAA NB . Single-strand specific P1 nuclease cleaved at optimal concentration only the TNR part, indicating proper formation of locked B-DNA, but without significant site preference, indicating unfolded TNR part, for both structures.
In summary, our results show p53 binding to slipped strand mimicking intermediates formed by TNR repeats with preference for TTC repeats over their purine-rich counterparts. Although the p53 binding to some TNRs has been studied previously, here the effects of TTC non-B DNA structure were studied for the first time. We compare p53 binding to TTC with other TNR non-B DNA (GAA, CTG and CAG), with ELISA and EMSA results showing that affinity for the TTC repeat was higher than for the previously published CTG and GAA non-B DNA structures [18,19].  Table S1. In panels (A) and (B), radiolabeled oligonucleotides (1 pmol) were bound by p53 in binding buffer in the presence of non-specific DNA competitor (75 ng, BSK/SmaI) and analyzed on a 6% 0.5× TBE polyacrylamide gel. The p53/DNA ratios are given per p53 tetramer. (A) DNA molecules are slipped strand mimicking substrates formed by trinucleotide repeat sequences (11 repeats of TTC, GAA, CTG or CAG) and a duplex p53 consensus sequence (CON). (B) DNA molecules formed by TANB, TNB, ANB, and DNA sequence nonspecific for p53 (dsLOCK, ssLOCK). (C) p53 binding to biotinylated oligonucleotides by ELISA, portrayed are mean values and SD. The dissociation constants (Kd) calculated per p53 tetramer are indicated. (D,E) Polyacrylamide gels showing resolved products of TTCNB (D) and GAANB (E) enzymatic or chemical cleavage with increasing concentration of T7 endonuclease I (left), P1 nuclease (middle) and potassium permanganate followed by piperidine treatment (right). Right panels show schematic drawings of formed structures. Small arrows indicate preferential cleavage by T7 endonuclease I.

Role of p53 Core and C-Terminal DNA Binding Domains in TNR Non-B DNA Recognition
To examine the roles of both p53 DNA binding domains (Figure 2A) in p53 TNR slipped-strand mimicking substrate recognition, we analyzed the DNA interactions of isolated p53 core domain (p53CD, aa 94-312, Figure 2B) and p53 C-terminal segment (p53CT aa 320-393; containing p53CTDBD and tetramerization domains, Figure 2C,D). Binding of p53CD and p53CT to CTGNB hairpin, TTCNB and TNB was compared with their binding to the CON ( Figure 2B,C, lanes 4-6) and a non-specific DNA sequence (NON, Figure 2B, C lanes 1-3). Importantly, although p53CD was able to form a specific complex with the consensus sequence under the experimental conditions ( Figure  2B, lanes 4-6), it was unable to form a stable complex with any of the structure-specific substrates ( Figure 2B, lanes 7-15). In contrast to p53CD, binding of p53CT to non-B DNA substrates (CTG hairpin, TTCNB and T-loops) was clearly selective over binding to duplex CON or NON substrates  Table S1. In panels (A,B), radiolabeled oligonucleotides (1 pmol) were bound by p53 in binding buffer in the presence of non-specific DNA competitor (75 ng, BSK/SmaI) and analyzed on a 6% 0.5× TBE polyacrylamide gel. The p53/DNA ratios are given per p53 tetramer. (A) DNA molecules are slipped strand mimicking substrates formed by trinucleotide repeat sequences (11 repeats of TTC, GAA, CTG or CAG) and a duplex p53 consensus sequence (CON). (B) DNA molecules formed by TA NB , T NB , A NB , and DNA sequence nonspecific for p53 (dsLOCK, ssLOCK). (C) p53 binding to biotinylated oligonucleotides by ELISA, portrayed are mean values and SD. The dissociation constants (Kd) calculated per p53 tetramer are indicated. (D,E) Polyacrylamide gels showing resolved products of TTC NB (D) and GAA NB (E) enzymatic or chemical cleavage with increasing concentration of T7 endonuclease I (left), P1 nuclease (middle) and potassium permanganate followed by piperidine treatment (right). Right panels show schematic drawings of formed structures. Small arrows indicate preferential cleavage by T7 endonuclease I.

Role of p53 Core and C-Terminal DNA Binding Domains in TNR Non-B DNA Recognition
To examine the roles of both p53 DNA binding domains ( Figure 2A) in p53 TNR slipped-strand mimicking substrate recognition, we analyzed the DNA interactions of isolated p53 core domain (p53CD, aa 94-312, Figure 2B) and p53 C-terminal segment (p53CT aa 320-393; containing p53CTDBD and tetramerization domains, Figure 2C,D). Binding of p53CD and p53CT to CTG NB hairpin, TTC NB and T NB was compared with their binding to the CON ( Figure 2B,C, lanes 4-6) and a non-specific DNA sequence (NON, Figure 2B, C lanes 1-3). Importantly, although p53CD was able to form a specific complex with the consensus sequence under the experimental conditions ( Figure 2B, lanes 4-6), it was unable to form a stable complex with any of the structure-specific substrates ( Figure 2B, lanes 7-15).
In contrast to p53CD, binding of p53CT to non-B DNA substrates (CTG hairpin, TTC NB and T-loops) was clearly selective over binding to duplex CON or NON substrates ( Figure 2C). The ELISA system ( Figure 2D) confirmed p53CT binding to TNR with often higher affinity than to p53CON, it also demonstrated p53CT preference for TTC NB over GAA NB . Overall, the C-terminal region of p53, containing CTDBD and tetramerization domain, seems to be chiefly responsible for p53 TTC NB recognition. This corresponds to data acquired by Walter et al. [18], who observed decreased p53 binding to CTG or CAG hairpins upon obstructing the p53 C-terminus by PAb421, an antibody mapping to the C-terminal region.
( Figure 2C). The ELISA system ( Figure 2D) confirmed p53CT binding to TNR with often higher affinity than to p53CON, it also demonstrated p53CT preference for TTCNB over GAANB. Overall, the C-terminal region of p53, containing CTDBD and tetramerization domain, seems to be chiefly responsible for p53 TTCNB recognition. This corresponds to data acquired by Walter et al. [18], who observed decreased p53 binding to CTG or CAG hairpins upon obstructing the p53 C-terminus by PAb421, an antibody mapping to the C-terminal region.
Our results showing that the C-terminal DNA binding domain with the tetramerization domain is crucial for TNR non-B DNA high affinity binding relate well with our previous studies on other non-B DNA structures, e.g., G-quadruplex [6,9] and T.A.T DNA triplex [8].

Binding of Wild-Type p53 to GAA·TTC Sequence in scDNA
To confirm that p53 can bind to TTC repeat sequence also in scDNA an EMSA in agarose gel was performed. Binding of p53 to scDNAs with (pPGM1) or without p53CON (pBSK) was compared with derived plasmids containing (GAA)18·(TTC)18 sequence (pPGM1/TTC and pBSK/TTC). As expected, binding of p53 to pPGM1 ( Figure 3, with p53CON, lanes 11-15) was much stronger than binding to pBSK ( Figure 3, lanes 1-5). Introducing TTC sequence did enhance p53 affinity for both plasmids. The used TTC sequence is a relatively weak binding site since it could not by itself (pBSK/TTC) provide the same affinity as an established p53CON (pPGM1), although the affinity will possibly be affected by TNR length.
The observed binding improvement to p53CON when in combination with TTC (pPGM1/TTC) encourages further exploration into non-B DNA structures as modulators of p53 regulated transcription near p53 consensus sites. In the case of the T.A.T type of triplex, we have previously shown that triplex forming sequence in supercoiled plasmid reporter can enhance transcription from CTDBD-C-terminal DNA binding domain) and its deletion variants p53CD (aa 94-312) and p53CT (aa 320-393) were used in this study. (B,C) EMSA analysis of p53CD (B) and p53CT (C) binding to CTG NB , TTC NB , T NB , nonspecific (NON) and p53 specific sequence (CON). Radiolabeled oligonucleotides (1 pmol) were bound by p53CD (B) or p53CT (C) in binding buffer in the presence of non-specific DNA competitor (50 ng BSK/SmaI) and analyzed on polyacrylamide gels (for more details see Figure 1). The p53/DNA ratios are given per p53 tetramer. (D) ELISA analysis of p53CT binding to biotinylated oligonucleotides: non-B DNA substrates CTG NB , TTC NB , GAA NB , A NB , T NB and p53 specific sequence Our results showing that the C-terminal DNA binding domain with the tetramerization domain is crucial for TNR non-B DNA high affinity binding relate well with our previous studies on other non-B DNA structures, e.g., G-quadruplex [6,9] and T.A.T DNA triplex [8].

Binding of Wild-Type p53 to GAA·TTC Sequence in scDNA
To confirm that p53 can bind to TTC repeat sequence also in scDNA an EMSA in agarose gel was performed. Binding of p53 to scDNAs with (pPGM1) or without p53CON (pBSK) was compared with derived plasmids containing (GAA) 18 ·(TTC) 18 sequence (pPGM1/TTC and pBSK/TTC). As expected, binding of p53 to pPGM1 (Figure 3, with p53CON, lanes 11-15) was much stronger than binding to pBSK ( Figure 3, lanes 1-5). Introducing TTC sequence did enhance p53 affinity for both plasmids. The used TTC sequence is a relatively weak binding site since it could not by itself (pBSK/TTC) provide the same affinity as an established p53CON (pPGM1), although the affinity will possibly be affected by TNR length. a neighbor p53 consensus sequence [8]. In our current work, we detect p53 selective binding to scDNA with TTC repeats indicating that this type of slipped DNA can serve as a non-B DNA binding motif for p53 in complex molecules.

Analysis of p53 Interaction with TNR Region from the First Frataxin Intron in Cells
Expansion of GAA·TTC repeat in the first intron of the frataxin gene is associated with Friedreich's ataxia [21]. Using the strategy used to prepare other clamped TNRs, each strand from this region (Table S1)  To determine if endogenous p53 binds to GAA·TTC expansion region in the first intron of FXN we chose this GAA·TTC rich sequence for ChIP analysis in adenocarcinoma cell line MCF7 and Friedreich's ataxia fibroblast cell line FXN 4654. To detect the p53-DNA binding in cells, we used semiquantitative PCR analysis ( Figure 4C) and qPCR ( Figure 4D), both employing primers that map the first intron of FXN and controls. Detection of p53 binding to P21 promoter was used as a positive control and binding to P21 exon was used as negative control ( Figure 4C). The GAA·TTC rich frataxin region was bound in the investigated MCF7 cells and we observed increase of p53 binding after p53 stabilization in response to nutlin-3 (NUT) and doxorubicin (DOX) treatment ( Figure 4C,D). In case of p53-DNA analysis in Friedreich's ataxia fibroblast cell line FXN 4654 we were able to detect binding to p21 promoter region and FXN GAA·TTC rich sequence by ChIP-qPCR analysis. Data confirmed the increased region occupancy of P21 and FXN genes after p53 stabilization by NUT and DOX in both tested cell lines ( Figure 4D). Taken together, ChIP analysis has confirmed the EMSA results, connecting p53 and TNR associated with Friedreich's ataxia disease. Interestingly, when probing other genomic GAA·TTC rich regions, p53 bound to the intragenic region on chromosome 2q ( Figure 4C, 2q) but not the last intron of RNF150 gene ( Figure 4C, 4q). Bioinformatics blast analysis revealed ( Figure S1) that both p53 bound regions (FXN and 2q) contain DNase I hypersensitivity sites, usually indicating transcriptionally active chromatin. In contrast, 4q GAA·TTC rich region did not contain such sites and was unbound by p53 (Fig, 4C). The nucleosome occupancy, methylation status and other factors are probably responsible for the differences in p53 recognition of the three investigated TTC rich segments in human genome.
As for frataxin, a p53-responsive element was recently found in the proximal FXN promoter [28,29] and FXN was identified as a p53 target gene. The dysfunction of p53 by mutation in cancerous cells was found to decrease the expression of frataxin and deregulate mitochondrial iron homeostasis, similarly to Friedreich's ataxia disease [29]. However, the previously investigated area did not cover the TNR region, which as we show may add a new regulatory feature to the p53-frataxin interplay. The observed binding improvement to p53CON when in combination with TTC (pPGM1/TTC) encourages further exploration into non-B DNA structures as modulators of p53 regulated transcription near p53 consensus sites. In the case of the T.A.T type of triplex, we have previously shown that triplex forming sequence in supercoiled plasmid reporter can enhance transcription from a neighbor p53 consensus sequence [8]. In our current work, we detect p53 selective binding to scDNA with TTC repeats indicating that this type of slipped DNA can serve as a non-B DNA binding motif for p53 in complex molecules.

Analysis of p53 Interaction with TNR Region from the First Frataxin Intron in Cells
Expansion of GAA·TTC repeat in the first intron of the frataxin gene is associated with Friedreich's ataxia [21]. Using the strategy used to prepare other clamped TNRs, each strand from this region (Table S1)  To determine if endogenous p53 binds to GAA·TTC expansion region in the first intron of FXN we chose this GAA·TTC rich sequence for ChIP analysis in adenocarcinoma cell line MCF7 and Friedreich's ataxia fibroblast cell line FXN 4654. To detect the p53-DNA binding in cells, we used semiquantitative PCR analysis ( Figure 4C) and qPCR ( Figure 4D), both employing primers that map the first intron of FXN and controls. Detection of p53 binding to P21 promoter was used as a positive control and binding to P21 exon was used as negative control ( Figure 4C). The GAA·TTC rich frataxin region was bound in the investigated MCF7 cells and we observed increase of p53 binding after p53 stabilization in response to nutlin-3 (NUT) and doxorubicin (DOX) treatment ( Figure 4C,D). In case of p53-DNA analysis in Friedreich's ataxia fibroblast cell line FXN 4654 we were able to detect binding to p21 promoter region and FXN GAA·TTC rich sequence by ChIP-qPCR analysis. Data confirmed the increased region occupancy of P21 and FXN genes after p53 stabilization by NUT and DOX in both tested cell lines ( Figure 4D). Taken together, ChIP analysis has confirmed the EMSA results, connecting p53 and TNR associated with Friedreich's ataxia disease. Interestingly, when probing other genomic GAA·TTC rich regions, p53 bound to the intragenic region on chromosome 2q ( Figure 4C, 2q) but not the last intron of RNF150 gene ( Figure 4C, 4q). Bioinformatics blast analysis revealed ( Figure S1) that both p53 bound regions (FXN and 2q) contain DNase I hypersensitivity sites, usually indicating transcriptionally active chromatin. In contrast, 4q GAA·TTC rich region did not contain such sites and was unbound by p53 ( Figure 4C). The nucleosome occupancy, methylation status and other factors are probably responsible for the differences in p53 recognition of the three investigated TTC rich segments in human genome. To examine whether the human FXN expression is regulated by p53 in tumor adenocarcinoma cell line (MCF7) and non-tumor cells (fibroblast BJ and Friedreich's ataxia fibroblast cell line FXN 4654), the cells were treated by nutlin and doxorubicin and RT-qPCR was carried out. As shown in Figure S2A,B in MCF7 cells, the level of FXN mRNA and p21 mRNA markedly increased in dependence on time of treatment (4 hours or 16 hours) and drugs; doxorubicin strongly increasing oxidative stress response or nutlin primarily stabilizing p53 protein level. The strongest activation of FXN expression was observed after 16 hours treatment by nutlin. RT-qPCR analysis in non-tumor fibroblast cell lines BJ and FXN 4654 after 4 hour treatment showed FXN repression in FXN 4654 cells ( Figure S2C) and no influence on FXN expression in contrast to p21 after nutlin or doxorubicin treatment in BJ cells ( Figure S2D). Our expression analysis of FXN in different cell lines showed that As for frataxin, a p53-responsive element was recently found in the proximal FXN promoter [28,29] and FXN was identified as a p53 target gene. The dysfunction of p53 by mutation in cancerous cells was found to decrease the expression of frataxin and deregulate mitochondrial iron homeostasis, similarly to Friedreich's ataxia disease [29]. However, the previously investigated area did not cover the TNR region, which as we show may add a new regulatory feature to the p53-frataxin interplay.
To examine whether the human FXN expression is regulated by p53 in tumor adenocarcinoma cell line (MCF7) and non-tumor cells (fibroblast BJ and Friedreich's ataxia fibroblast cell line FXN 4654), the cells were treated by nutlin and doxorubicin and RT-qPCR was carried out. As shown in Figure S2A,B in MCF7 cells, the level of FXN mRNA and p21 mRNA markedly increased in dependence on time of treatment (4 h or 16 h) and drugs; doxorubicin strongly increasing oxidative stress response or nutlin primarily stabilizing p53 protein level. The strongest activation of FXN expression was observed after 16 h treatment by nutlin. RT-qPCR analysis in non-tumor fibroblast cell lines BJ and FXN 4654 after 4 h treatment showed FXN repression in FXN 4654 cells ( Figure S2C) and no influence on FXN expression in contrast to p21 after nutlin or doxorubicin treatment in BJ cells ( Figure S2D). Our expression analysis of FXN in different cell lines showed that further studies are necessary to understand the p53 role in frataxin biology. Frataxin repression after doxorubicin (oxidative stress inducer) treatment was observed previously in normal cells (cardiomyoblasts) [30].
The role of oxidative stress in Friedrich ataxia's cells [31] and p53 function in stress response have to be considered. The reduction of frataxin expression is associated with increased oxidative stress and cell damage, accumulation of iron in mitochondria [32]. These factors can strongly influence p53 function by oxidative damage of protein. The central domain of p53 protein is sensitive to cysteine oxidation which leads to reduction or abolishment of sequence specific interaction with p53 consensus element [33] but p53 non-B DNA structure recognition is partially maintained for oxidized p53 as was observed for G-quadruplexes [6] or supercoiled DNA [33]. We do show an increase in p53 occupancy on frataxin TNR regions when treated with oxidative stress inducing agent doxorubicin. The C-terminal domain directed binding may represent a potential mechanism by which p53 could regulate frataxin under oxidative stress.

Future Prospects of p53 and TNR Non-B DNA Structures
In cancer genomes, (GAA) n repeats, together with (AT) n , and (GAAA) n , constitute the most frequent repeats at translocation breakpoints, which represent an intrinsic risk factor for genomic rearrangements [34]. Interestingly, unlike CNG repeats, GAA repeats display unique behavior during TNR repeat expansion: a) strict strand orientation dependence of the replication stalling; b) GAA repeats expansion occurs once they reach a certain threshold [35]. As p53 is involved as a transcription factor and also by direct interactions with several relevant proteins in processes that might be important in TNR stability such as homologous recombination, mismatch repair, replication [2,36,37], it is a future prospect to investigate p53 influence on TNR stability.
The GAA triple repeat expansion in Friedreich's ataxia is directly responsible for the transcriptional silencing of frataxin gene [21]. Nowadays two model potential mechanisms are considered to be responsible for this reduction of frataxin expression: a) the formation of non-B DNA structures (including RNA_DNA hybrid non-B structures or GAA triplexes) and b) heterochromatin formation [32]. Inhibition of the frataxin transcription process by non-B DNA conformation was observed in vitro and in vivo [32,38].
The significance of the observed p53 preference for Py-rich TNR non-B DNA structures is proposed to be associated with a mechanism of gene silencing in Friedreich's ataxia ( Figure S3). The transcriptional block of Pol II on long GAA·TTC repeats occurs via the TTC strand [32]. So far, we can only speculate if the recently observed p53 regulation of DNA-binding function of Pol II [39], can also influence Pol II transcription of frataxin gene. The transcriptional weak spot represented by the extruded TTC strand is however preferentially recognized by p53, as is shown in the current study ( Figure S3). The triplex DNA formation is another proposed mechanism of Pol II transcriptional block on frataxin TNR. In our previous work, we observed preferential p53 recognition of DNA T.A.T triplex. We used the luciferase reporter assay to detect the effect of a T.A.T triplex on p53 driven transcription [8]. We observed T.A.T triplex dependent enhancement of reporter plasmid activation with p53CON sequence and triplex forming sequence by luciferase reporter assay. We also observed that the same sequence can block transcription of reporter plasmid if it contains only T.A.T triplex motif in supercoiled reporter [8]. Any future reporter assays with TNR have to take into account the influence of DNA topology because DNA supercoiling is a necessary condition for non-B DNA formation [8]. The requirement for a threshold level of DNA supercoiling may be the reason why the study of the influence of CTG·CAG repeat track on a linear luciferase reporter did not show profound effect for p53 [18].
Further studies are needed to understand the precise function of p53 TNR non-B DNA recognition in relation to the development of Friedreich's ataxia or other diseases coupled with TNR expansion.
The ability of p53 to interact with two DNA binding domains makes the protein very flexible in terms of its physiological functions. The C-terminal DNA binding domain is mainly responsible for high affinity non-B DNA recognition by p53 [6,8,9] and, for the first time, our work shows that this domain is crucial for TNR non-B DNA recognition. It is expected that the ability of p53 to recognize TNR non-B DNA is conserved in the majority of hot spot cancer mutant p53 proteins [40]. About 90% of TP53 mutations are localized to the central part of gene, the majority of mutant p53 will still have functional C-terminus, so they will be able to recognize non-B DNA structures formed by TNR, as already shown for hot spot R273H mutant p53 to recognize CTG non-B DNA structures [18], which may be a reason why p53 status was not associated with Friedreich's ataxia. In contrast, C-terminal mutants in lysine residues are expected to not recognize TNR non-B similarly to truncated mutant p53 without 30 last amino acids [6,9,41].
In summary, we show that non-B DNA structures formed by TNR (TTC, GAA, CTG, CAG) and simple T-repeat are recognized by p53. Moreover, p53 prefers non-B DNA structures formed by the pyrimidine-rich strands of the investigated repetitive sequences and that the intact C-terminus is responsible for high p53 affinity to TNR non-B DNA structures.

Oligonucleotides
The sequences of oligonucleotides used in this study are presented in Table S1, and these were purchased from Sigma-Aldrich (Praha, Czech Republic). DNA hairpin or DNA loop structures were prepared as described in [18,19]. Briefly, Lock-F oligonucleotide was mixed with the repeat containing complementary strands in buffer (10 mM Tris-HCl, pH 8; 0.1 mM EDTA and 50 mM KCl) and placed in a heating block at 95 • C for 5 min. DNA substrates were allowed to cool slowly and anneal for a minimum of 3 hours. For EMSA Lock-F was 5 -labeled by T4-DNA polynucleotide kinase (NEB) with 32 P-gamma ATP. For ELISA Lock-F was biotinylated at the 3 end.

EMSA in Polyacrylamide Gels
Electrophoretic mobility shift assays were performed with 32 P-radiolabeled oligonucleotide probes (1 pmol) mixed with p53 proteins and incubated in binding buffer (5 mM Tris-HCl, pH 7.6; 0.5 mM EDTA; 0.5 mM DTT; 0.01% Triton-X100 and 50 mM KCl) in the presence of 50 or 75 ng pBSK/SmaI DNA competitor for 15 min on ice to reach equilibrium. Samples were loaded onto a 6% polyacrylamide gel containing 0.5× Tris-borate-EDTA (TBE) buffer. After 30 min electrophoresis (at 6 V/cm) the gels were dried, and DNA was detected by autoradiography using Typhoon FLA 9000 (GE Healthcare). Representative gels from at least three repetitions are shown.