Crystal structure of a DNA aptamer bound to PvLDH elucidates novel single-stranded DNA structural elements for folding and recognition

Structural elements are key elements for understanding single-stranded nucleic acid folding. Although various RNA structural elements have been documented, structural elements of single-stranded DNA (ssDNA) have rarely been reported. Herein, we determined a crystal structure of PvLDH in complex with a DNA aptamer called pL1. This aptamer folds into a hairpin-bulge contact by adopting three novel structural elements, viz, DNA T-loop-like motif, base–phosphate zipper, and DNA G·G metal ion zipper. Moreover, the pL1:PvLDH complex shows unique properties compared with other protein:nucleic acid complexes. Generally, extensive intermolecular hydrogen bonds occur between unpaired nucleotides and proteins for specific recognitions. Although most protein-interacting nucleotides of pL1 are unpaired nucleotides, pL1 recognizes PvLDH by predominant shape complementarity with many bridging water molecules owing to the combination of three novel structural elements making protein-binding unpaired nucleotides stable. Moreover, the additional set of Plasmodium LDH residues which were shown to form extensive hydrogen bonds with unpaired nucleotides of 2008s does not participate in the recognition of pL1. Superimposition of the pL1:PvLDH complex with hLDH reveals steric clashes between pL1 and hLDH in contrast with no steric clashes between 2008s and hLDH. Therefore, specific protein recognition mode of pL1 is totally different from that of 2008s.


Results and Discussion
Dimeric PvLDH binds to one DNA aptamer, folding into a unique hairpin-bulge contact. The pL1:PvLDH complex was crystallised under two different conditions. The crystals belonged to P2 1 2 1 2 and P4 1 2 1 2 at 1.9 Å and 1.71 Å resolution, respectively (Table S1). Two aptamers were bound to tetrameric PvLDH in an asymmetric unit of P2 1 2 1 2 (Fig. 1A). One aptamer was bound to dimeric PvLDH in an asymmetric unit of P4 1 2 1 2 ( Fig. 1B). Despite different space groups, the crystal packing and interface of the pL1:PvLDH complexes were similar (Fig. 1A,B). Therefore, all descriptions provided below are based on the pL1:PvLDH complex structure belonging to P4 1 2 1 2.
Electron density of pL1 was clearly observed (Fig. S1). pL1 folds into a unique hairpin-bulge contact in complex with PvLDH. The main elements of pL1 include two stems, an intervening bulge and a hairpin loop (Fig. 1C). The two stem elements are stabilised by stacking base pairs. Four Watson-Crick-type base pairs (T3-A34, C4-G33, G5-C32 and A6-T31) are stacked as a B-form helical DNA duplex at Stem I. Stem II comprises four Watson-Crick-type base pairs (G16-C30, C17-G29, C18-G28 and G20-C26) and one wobble base pair (G19·T27). In contrast to intervening bulges or internal loops that result in distorted helices, coaxial stacking occurs at the interface between the terminal base pairs of Stem I and Stem II. Owing to coaxial stacking, unpaired nucleotides of the bulge are pushed out of the stems. However, the hairpin and intervening bulge are stabilised by the hairpin-bulge contact. Two Watson-Crick-type base pairs are created (A22-T12 and A21-T13) between the hairpin loop and bulge. Similar to RNA pseudoknots, coaxial stacking is formed at the interface between Stem II and the hairpin-bulge contact. Thus, a pseudo-continuous helix is built from Stem I to the hairpin-bulge contact by coaxial stacking (Fig. S2).
T-loop-like motif is a novel ssDNA secondary structural element contributing as a nucleation site. The T-loop motif, a RNA hairpin motif, consists of a UA trans-Watson-Crick/Hoogsteen closing base pair and a U-turn sub-motif following three specific nucleotides, 5ʹ -(G/U)NR-3ʹ (N: any nucleotide; R: purine). The U-turn sub-motif is stabilised by H-bonds between the first (G/U) and third (R) residues. The imino proton of the G/U interacts with the phosphate of the R residue via a BPh interaction belonging to the 5BPh interaction. Moreover, one hydrogen bond (H-bond) is formed between the 2ʹ hydroxyl group of R and the imino proton of the G/U residue 39 . PvLDH subunits A, B, C, and D are shown in orange, yellow, limon, and green, respectively. pL1 A and B are colored in blue and marine, respectively. (B) Dimeric PvLDH interacts with one DNA aptamer. PvLDH subunits A and B are shown in orange and yellow, respectively. PvLDH-contacting nucleotides and non-contacting nucleotides of pL1 are shown in cyan and blue, respectively. (C) Structure and schematic of a DNA aptamer called pL1. Ribbon representation of pL1 and the secondary structure derived from the tertiary structure. Strands forming the bulge, hairpin loop, and stems are shown in cyan, purple, and black, respectively. Adenines, guanines, thymines, and cytosines are shown in red, green, yellow, and blue respectively. Nucleotides at the bulge and hairpin loop are boxed with cyan and purple dashed lines, respectively. To date, no study has reported a similarity between ssDNA structural elements and the RNA T-loop motif. Interestingly, the apical bulge of pL1 is stabilised by a novel structural element similar to the RNA T-loop motif. Of the apical unpaired nucleotides of the bulge, including 5ʹ -A11, T12, T13, G14 and T15-3ʹ , 5ʹ -A11 interacts with T15 via a reversed Hoogsteen base pair. In other words, the TA trans-Watson-Crick/Hoogsteen closing base pair occurs within the bulge instead of the TU trans-Watson-Crick/Hoogsteen closing base pair of the RNA T-loop. T12, T13 and G14 form sharp bend-like U-turn sub-motifs. However, the second (T13) and third (G14) nucleotides are stabilised by H-bonds, unlike RNA U-turn motifs where H-bonds occur between the first and third nucleotides. At the sharp turn containing 5ʹ -T12, T13 and G14-3ʹ , there are no H-bonds between the sugars and bases because of the absence of 2ʹ -hydroxyl DNA groups. However, the 1-endocylic amine group and 2-exocyclic amino group of G14 interact with the phosphate oxygen group of T13 through a 4BPh interaction. Thus, the third guanine nucleotide (G14) plays a critical role in the sharp bend formation ( Fig. 2A-C). Moreover, the third guanine is conserved between pL1 and pL2, through which pLDH is specifically recognised 37 . Therefore, the DNA turn motif is defined as a G-turn motif (Fig. 2B).
In addition to the formation of thermodynamically stable structures, RNA T-loop motifs contribute as nucleation sites, inducing long-range tertiary interactions such as T-loop:D-loop and T-loop:receptor complexes. The DNA T-loop-like motif also induces a T-loop-like motif:hairpin loop complex via a hairpin-bulge contact. The T-loop:D-loop and T-loop:receptor complexes contain one Watson-Crick-type base pair with other non-canonical interactions. However, the hairpin-bulge contact occurs between the T-loop-like motif and hairpin loop via two Watson-Crick base pairs because of the absence of H-bonds between the sugars and bases, which confers more accessible T12 and T13 surfaces for interacting with complementary sequences ( Fig. 2A,D).

BPh zipper as a novel tertiary interaction element that stabilises loops adjacent to stems.
Interactions between RNA helical and unpaired nucleotides occur via A-minor motifs, base triplets and ribose zippers. DNA unpaired nucleotides adjacent to stems are solely stabilised by base triplets owing to the formation of B-form helices and the absence of 2ʹ -hydroxyl groups. To date, it has not been reported that nucleic acids adopt consecutive BPh interactions. Unlike previously reported motifs, consecutive BPh interactions mediate stabilisation between helical and unpaired nucleotides from the region between the bulge to Stem II of pL1 (Fig. 3). Of the nine unpaired nucleotides (T7, T8, G9, G10, A11, T12, T13, G14 and T15) of the bulge, five unpaired nucleotides (A11, T12, T13, G14 and T15) are stabilised by the T-loop-like motif ( Fig. 2A). Of the remaining four unpaired nucleotides (T7, T8, G9 and G10), two nucleotides (T7 and T8) are looped-out. However, G9 and G10 are adjacent to Stem II, with unpaired G10 being stacked above unpaired G9 but below T15 within the bulge (Fig. 1C). Interestingly, the two unpaired nucleotides G9 and G10 interact with the 1-phosphate and 2-phosphate oxygens of C17, respectively. Both the 1-endocylic amine group and 2-exocyclic amino group of G9 and G10, respectively, contribute as H-bond donors to the phosphate oxygens via 4BPh interactions (Fig. 3B). Therefore, sequentially consecutive BPh interactions between stems and unpaired nucleotides form the BPh zipper. Interestingly, the BPh zipper is a unique ssDNA structural element that differs from RNA structural elements because RNA does not use consecutive BPh interactions. Instead, RNA molecules favour the use of 2ʹ -hydroxyl groups. In addition to DNA T-loop-like motifs, 4BPh interactions are observed in BPh zippers. The 4BPh interaction is powerful because the 1-phosphate oxygen forms two H-bonds with the 1-endocylic amine group and 2-exocyclic amino group of guanine 40 . Therefore, exclusive ssDNA structural elements appear to favour 4BPh interactions to compensate for the lack of 2ʹ -hydroxyl groups.

Stabilisation of the bulge and BPh zipper by a novel DNA G·G metal ion zipper.
The metal-binding site was observed in pL1; validation of this pL1 magnesium-binding site was performed with the CheckMyMetal web server (http://csgid.org/csgid/metal_sites). All parameters were acceptable for magnesium ions (Fig. S3, Table S2). The metal-binding site of pL1 was found to differ from G-quadruplexes, well-known metal-binding ssDNA structures. The phosphate backbone of the bulge region (T7, T8, G9 and G10) is sharply bent at approximately 90°. To stabilise the bent bulge, the magnesium ion holds the highly bent conformation of the backbone (Fig. 4A,B). The phosphate oxygen of T8, the 7-imino group of G9 and four waters are present in the inner sphere of the magnesium ion, in the octahedral geometry. Of the classified magnesium ion-binding sites in the large ribosomal subunit of RNA, Type IIa shows an orthogonal geometric arrangement of two inner-sphere atoms of RNA with four waters 41 . The magnesium ion-binding site of pL1 belongs to the same geometric arrangement of Type IIa (Fig. 4C). To reduce repulsion, nucleotide atoms at the bulge and Stem II are directly or indirectly coordinated with the magnesium ion. The magnesium ion is directly linked with one 1-phosphate oxygen of T8. Out-sphere atoms are three 2-phosphate oxygens of T7, T8 and G16, although the phosphate of G16 at Stem II is distant from the phosphates of T7 and T8 at the bulge. A magnesium ion is bound through the outer sphere by two distant phosphates in the RNA metal ion zipper. Interestingly, this motif was only observed in rRNA (The term 'distant' means that two phosphates come from different chains or are separated by at least seven nucleotides in the same chain) 9 . This outer-sphere-coordinating pattern is similar to that of pL1 (Fig. 4D). Moreover, the oxygen atoms of 6-oxo groups at G9 and G10 are outer-sphere-coordinating atoms. This is the same outer-sphere-coordinating pattern of the RNA G·G metal-binding site in which two sequentially consecutive guanine moieties are connected to the magnesium ion 9 (Fig. 4E). Considering all coordination patterns, it can be said the magnesium coordination of pL1 in the outer sphere is similar to that in the RNA metal ion zipper and RNA Strands forming the bulge, hairpin loop, and stems are shown in cyan, purple, and black, respectively. (A) Basephosphate zipper and pseudo-continuous helix are in the secondary structure, which is derived from the tertiary structure. Nucleotides at the base-phosphate zipper and pseudo-continuous helix are boxed in green and blue dashed lines, respectively. Strands forming the bulge, hairpin loop, and stems are shown in cyan, purple, and black, respectively. Guanines, adenines, thymines, and cytosines are shown in green, red, yellow, and blue, respectively. (B) Two sequentially consecutive guanine bases interact with phosphate oxygens of the quasi-continuous helix. The pseudo-continuous helix is represented as a transparent sphere. Stick representation of the hairpin-bulge contact indicates the pseudo-continuous helix by coaxial stacking. Strands forming the bulge, hairpin, and stems are shown in cyan, purple, and black, respectively. Guanines, adenines, thymines, and cytosines are shown in green, yellow, pink, and blue, respectively.
Scientific RepoRts | 6:34998 | DOI: 10.1038/srep34998 G·G metal-binding motif. Therefore, this novel metal-binding element is defined as a DNA G·G metal ion zipper. Moreover, sequentially consecutive G9 and G10, contributing to the BPh zipper, are stabilised by the magnesium ion. In addition to contributing to ssDNA folding, the magnesium ion interacts with the ammonium group of Lys84(B) through one salt bridge (Fig. 4B).
Nature of pL1-mediated PvLDH recognition. The dimeric PvLDH places one pL1 in a single cluster of DNA-contacting residues that forms a concave region determined by the Cx algorithm (Fig. S4) 42 . The interface between the dimeric PvLDH and pL1 buries the total accessible surface area (ASA) of 1145.25 Å 2 . The interface between pL1 and monomeric PvLDH buries a solvent-accessible surface of 613.8 Å 2 and 530.0 Å 2 for PvLDH subunits A and B, respectively. Although the same five amino acids (Met14, Val36, Val37, Gly81, Phe82 and Ile105) of both subunits participate in DNA contact, another eight amino acids at each subunit are involved in DNA interaction (Table S3). Therefore, the DNA-contacting surfaces of each subunit are asymmetric in the dimeric PvLDH (Fig. 5A).
The PvLDH-contacting region of pL1 is composed of two paired nucleotides at Stem II and nine unpaired nucleotides at the bulge and hairpin loop. For interactions between Stem II and pL1, Leu232(A) is solely engaged in van der Waals contacts with the sugar phosphate backbone of C17 and C18. However, the main interactions between PvLDH and pL1 occur with unpaired nucleotides. Unpaired nucleotides at the bulge and hairpin loop continuously stretch away into the concave protein region because the two separated DNA regions of the bulge and hairpin loop are connected by the hairpin-bulge contact. The bulge contains PvLDH-contacting nucleotides (T7, T8, G9, G10, A11 and T12), while the hairpin loop involves PvLDH-contacting nucleotides (G23, T24 and G25) (Fig. 5A,B).
The PvLDH-contacting unpaired nucleotides contain four unpaired and flipped nucleotides (T7, T8, T24 and G25). T8 and T24 extend into the NADH-binding clefts of subunits B and A, respectively. Unpaired and flipped nucleotides do not participate in hydrogen bonding for complex formation, with the exception of T8. Only one H-bond occurs between the 3ʹ oxygen of T8 and the backbone amide NH group of Gly81(B). In contrast to using H-bonds, the two bases of G23 and T8 interact with the two aromatic rings of Phe82(A, B) through hydrophobic interactions (Fig. 5B).
The unpaired G9, G10, A11, T12 and G23 nucleotides are not flipped out of their structure because they are stabilised by the three novel structural elements described above. As a result, the sugar phosphates of G9, G10, A11 and T12 are primarily exposed to the concave protein surface instead of the bases. The unpaired and not flipped nucleotides are stabilised by four H-bonds. The 1-endocyclic amine group and 2-exocyclic amino group of G23 form two H-bonds with the carbonyl backbone of Thr83(A). The 1-phosphate oxygen of A11 interacts with the phenolic hydroxyl group of Tyr236(A) by one H-bond. The 1-phosphate oxygen of G10 interacts with  the ammonium group of Lys44(B). In addition to H-bonds, one salt bridge occurs between the phosphate of G10 and ammonium group of Lys44(B) to stabilise the negatively charged phosphate backbone. Therefore, a total of five intermolecular H-bonds and one salt bridge are observed at the interface between pL1 and dimeric PvLDH (Fig. 5A,B, Table S3). Moreover, the magnesium ion interacts with the ammonium group of Lys84(B) through one salt bridge (Fig. 4B).
Unique properties of the pL1:PvLDH complex interface. To understand the properties of the pL1:Pv-LDH interface, several parameters of the pL1:PvLDH complex were compared with those of protein:RNA, protein:DNA and protein:DNA aptamer complexes (Table 1, S4). The buried surface area within 5 Å upon complex formation is defined as Δ ASA (Å 2 ). H-bonds (/100 Å 2 Δ ASA) are defined as the number of H-bonds per 100 Å 2 Δ ASA. Bridging waters (/1000 Å 2 Δ ASA) are defined as the number of water molecules forming H-bonds with both proteins and nucleic acids per 1000 Å 2 Δ ASA. All parameters were calculated by the COCOMAP web server, PISA, PyMol and CONTACT in the protein:DNA aptamer complexes. The data for the interface areas, H-bonds and bridging waters of the protein:RNA, protein:DNA and protein:protein complexes were obtained from a previous interface analysis 43,44 . Although most protein-binding nucleotides are unpaired nucleotides of pL1, the number of H-bonds (/100 Å 2 Δ ASA) of the pL1:PvLDH complex was 0.4, which was the lowest among the mean number of the other complexes. On the other hand, extensive H-bonds are observed in ssDNA:thrombin, 2008s:PfDLH and ARC1172:VWF complexes because most protein-interacting nucleotides are unpaired nucleotides. Although the main protein-binding nucleotides of anti-ATX and anti-HIV-1 RT DNA aptamers are paired nucleotides, the number of H-bonds (/100 Å 2 Δ ASA) of the ATX:RB011 and HIV-1 RT:apt-DNA complexes is higher than that of the pL1:PvLDH complex. Moreover, the bridging waters (/1000 Å 2 Δ ASA) of the pL1:PvLDH complex far outnumbered those of the other complexes (Table 1, S4). Fourteen bridging waters form water-mediated H-bonds between pL1 and PvLDH (Fig. 5D, Table S5). To understand the contribution of bridging waters to shape complementarity, Sc values for the pL1:PvLDHW and pL1:PvLDH complexes were calculated. PvLDHW is used to designate PvLDH with bridging waters; the Sc value for pL1:PvLDHW (0.720) is higher than that for pL1:PvLDH (0.675) (Table S6). Therefore, fourteen bridging waters maximise shape complementarity. The electrostatic surface of the pL1-binding region is partially negatively and positively charged in PvLDH, whereas that of the PvLDH-binding region is mostly positively charged in pL1 (Fig. 5C). Bridging waters serve to buffer unfavourable electrostatic interactions (Fig. 5D).
However, unique properties in their interactions with pLDH are shown because of different DNA folding. In contrast to the metal-independent 2008s, pL1 is a metallo-DNA aptamer and the magnesium ion participates in the interaction between pL1 and PvLDH. This magnesium ion, coordinated with pL1, interacts with the ammonium group of Lys84 by one salt bridge (Fig. 4B). Instead of metal ions, the ammonium group of Lys84 interacts with the phosphate of A12 in the 2008s:PfLDH complex via one salt bridge. Moreover, pL1 exclusively interacts with Gly11, Lys38, Thr79, Thr83, Asp97, Leu98 and Tyr236 of PvLDH, while 2008s solely interacts with Phe34, Tyr67, Ala85, Pro86, Gly87, Lys88, Ser89, Asp90, Lys91, Glu108, Val229 and Ala233 of PfLDH (Table S3). In addition to different DNA-contacting residues, one cluster of DNA-contacting residues in the pL1:PvLDH complex is different from the double cluster of DNA-contacting residues in which the dimeric PfLDH places one 2008s (Fig. S5).
Furthermore, the main interaction mode of the pL1:PvLDH complex is different from that of the 2008s: PfLDH complex. Hereafter, the loop containing ASPR is referred to as FT. Although disordered in apo-PfLDH, FT is well ordered in the pL1:PfLDH complex because FL plays a key role in 2008s recognition via extensive H-bonds. To understand the conformational change between apo-PvLDH and pL1-bound PvLDH, we determined the crystal structure of apo-PvLDH belonging to I222 at 1.6 Å resolution (Table S1). Conformational changes are subtle between the apo-PvLDH dimer and pL1-bound PvLDH dimer, with an overall root mean square deviation of 0.94 Å (Fig. S6). Moreover, FL is disordered in both apo-PvLDH and pL1-bound PvLDH. Therefore, shape complementarity between pL1 and the inflexible regions of apo-PvLDH is critical for interactions between pL1

Structural insight into discriminatory pLDH recognition by pL1.
Extensive H-bonds determine the specific recognition between 2008s and ASPR 35 . Selectivity tests using cationic surfactant-based colorimetric detection show that pL1 also discriminates pLDH from LDH-A 38 . Moreover, electrophoretic mobility shift assays (EMSAs) reveal that pL1 discriminates pLDH from hLDH-A and hLDH-B (Fig. 6A). However, ASPR does not contribute to pL1 recognition in the pL1:PvLDH complex. Compared with four salt bridges and 15 H-bonds in the 2008s:PfLDH complex, only one salt bridge and five H-bonds occur in the pL1:PvLDH complex. Therefore, pL1 applies different strategies for specific pLDH recognition, instead of extensive H-bonds and electrostatic complementarity. Superimposition of the 2008s:PfLDH complex with hLDH reveals shape complementarity between 2008s and hLDH (Fig. S7). However, superimposition of the pL1:PvLDH complex with hLDH shows steric clashes between pL1 and hLDH. Tight packing of the pL1:PvLDH interface is evident because of the hairpin-bulge contact that confers a close approach to the concave PvLDH region on the unpaired nucleotides A11, T12 and T13. Superimposition of pL1-bound PvLDH with hLDH reveals that the concave region of hLDH is shallower than that of PvLDH. Moreover, protruding regions of hLDH aid in the steric clashes between pL1 and PvLDH. The protruding Arg99, Gln100, Gln101 and Y247 of hLDH-A overlap with A11, T12, G23 and A24, respectively. The protruding Arg99, Gln101, Arg112 and Tyr247 of hLDH-B overlap with A11, T12 and G23 of pL1, respectively (Fig. 7). As a result, the shape of pL1 is not complementary to the shallowly concave and protruding region of hLDH, clashing with pL1 binding to hLDH. Therefore, shape complementarity plays a critical role in the specific discrimination of pLDH from hLDH by pL1. In conclusion, the specific protein recognition mode of pL1 appears to be completely different from that of 2008s.

Structural analysis of deleterious mutations on pL1.
Mutations in the set of conserved sequences between pL1 and pL2 dramatically decrease the affinity between pL1 and PvLDH 37 (Table S7). We designed base substitution mutations to understand the critical roles of the three novel structural elements for PvLDH recognition. In the bulge mutant, the DNA sequence was changed from 5ʹ -GG-3ʹ to 5ʹ -AA-3ʹ at the bulge to understand the importance of the BPh and G·G metal ion zippers. Furthermore, the hairpin mutant was designed by changing A22 to G22 at the hairpin loop because the Watson-Crick A22-T12 pair plays a key role in the DNA T-loop-like motif:hairpin loop complex. In addition to mutations, the presence of magnesium ions is critical for proper pL1 folding and PvLDH recognition. Therefore, EMSA was performed under various conditions (Fig. S8, Table S7). As a result, the disturbance in the hairpin-bulge contact, BPh zipper and DNA G·G metal ion zipper has deleterious effects on PvLDH recognition (Fig. 6). As described above, the hairpin-bulge contact, BPh zipper and DNA G·G metal ion zipper play critical roles in the stabilisation of PvLDH-binding unpaired nucleotides at the bulge and hairpin loop. Therefore, pL1 does not recognise PvLDH without the three novel structural elements because the combination of the three novel structural elements confers the stable shape to the unpaired nucleotides.
Interestingly, the G-C rich stem mutant results in a low affinity between pL1 and PvLDH, although the secondary structure of Stem II is maintained (Fig. S8B). The stability of coaxial stacking depends on the sequence of the helix-helix interface. In pL1, coaxial stacking occurs at the helix-helix interface between the Watson-Crick A6-T31 pair and the Watson-Crick G16-C30 pair. In the G-C rich stem mutant, the Watson-Crick A6-T31 pair stacks below the Watson-Crick A16-T30 pair at the helical interface. Base-stacking interactions are stabilised to a greater extent by the Watson-Crick G-C pair than by the Watson-Crick A-T pair 46,47 . As a result, pL1 has more stable coaxial stacking than the G-C rich stem mutant. Because coaxial stacking is the main scaffold of pL1 for maintaining other structures, the G-C rich stem mutation reduces the affinity between pL1 and PvLDH. Considering all mutations, it is conceivable that mutations in nucleotides involved in structural elements cause a critical malfunction in the interaction between pL1 and PvLDH.

Methods
Purification of recombinant PvLDH and oligo DNA preparation. PvLDH gene sequence was inserted into the pET28a (Novagen) vector and then recombinant PvLDH was transformed into E. coli. strain ER2566. Bacteria were incubated at 37 degrees in Luria-Bertani (LB) broth until OD 600 = 0.5. Then, the recombinant PvLDH was expressed by adding 0.2 mM isopropyl β -d-1-thiogalactopyranoside (IPTG) for 4 hours. Cells were harvested using centrifugation at 5,000 rpm for 10 min. Cells suspended in resuspension buffer (20 mM Tris HCl pH 7.5 and 500 mM NaCl) were sonicated. Then, cell debris was removed by centrifugation at 13,000 rpm for 1 hour. Supernatants were loaded onto a 5-ml Histrap column (GE Healthcare Life Science) and eluted in an elution buffer (20 mM Tris HCl pH 7.5, 500 mM NaCl, and 300 mM imidazole). The elution samples were thereafter passed over Superdex 200 (16/60 GE Healthcare Life Science) in the final buffer containing 20 mM Tris HCl pH 7.5, 50 mM NaCl, 5% (w/v) glycerol, and 5 mM β -mercaptoethanol to remove aggregates. DNA aptamer pL1 with the sequence 5ʹ -GTTCGATTGGATTGTGCCGGAAGTGCTGGCTCGAAC-3ʹ was synthesised using MOPC TM purification at Macrogen (Seoul, Korea). Oligonucleotides were placed in the binding buffer (20 mM Tris HCl pH 7.5, 50 mM NaCl, 5 mM KCl, and 5 mM MgCl 2 ). For proper folding, oligonucleotides were denatured at 90 degrees for 3 minutes and then incubated at 4 degrees for 1 hour.  by mixing 1 μ l of the pL1:PvLDH complex with 1 μ l of the reservoir solution (0.1 M HEPES pH 7.0 and 15% (w/v) PEG-20 K) after 1 month. To get high-resolution data, additive screening was performed. Then, rod-or octahedron-shaped crystals were grown by mixing 2 μ l of the pL1:PvLDH complex with 2 μ l of the crystal solution containing 0.1 M HEPES pH 7.0 and 15% (w/v) PEG-20 K with the addition of 0.3 M NDSB-195 or 5% (w/v) PEG-400, respectively, for 1 month. Crystals of apo-PvLDH were observed by mixing 1 μ l of the pL1:Pv-LDH complex with 1 μ l of the precipitant containing 0.2 M ammonium nitrate and 20% (w/v) PEG-3350 after 2 months.

Complex formation and crystallization. Recombinant
Data collection and processing. Native diffraction data of the pL1:PvLDH complex and apo-PvLDH were collected at the beamline of 5C of Pohang Accelerator Laboratory at a wavelength of 0.9796 Å by using cryo-cooled crystals at 100 K with cryoprotectants involving 30% (w/v) glycerol or 50% (w/v) PEG-20 K. Data processing was achieved with HKL2000 software 48 . The high-resolution cutoff of a dataset was performed according to rules of three for determining resolution limit (1. completeness > 70%, 2. R merge (a.k.a R symm , R linear ) < 0.30 (or 30%), I/σ (I) > 3). In spite of a conservative cut for I/σ (I) and completeness, the resolution limits were determined for appropriate R merge values. Crystallographic data statistics are described in Table S1.
Structure determination and refinement. Phases of the pL1:PvLDH complex and apo-PvLDH were determined by phaser-MR in PHENIX package 49 . Monomeric PvLDH (PDB code: 2A92) was used as a molecular replace search model. DNA nucleotides were manually and gradually built into the difference density map with coot 50 . Thereafter, model structures were refined with phenix.refine in PHENIX package 49 . XYZ coordinates, real-space, rigid body, individual ADPs, and occupancies were selected for refinement strategies in common. Non-crystallographic symmetry (torsion-angle restraints type) was used only for the refinement of apo-PvLDH. The TLS refinement strategy was selected only for the pL1:PvLDH complex (PDB code: 5HRU). TLS groups were automatically found by TLSMD server. Secondary structure restraints were essentially used for the refinement of DNA owing to distance restraints for hydrogen bonding interactions in Watson-Crick base pairs. Model building and refinement were iteratively performed to get the best model of the pL1:PvLDH complex and apo-PvLDH. Amino acids approximately from 83 to 95 and nucleotides of 1, 2, 35, and 36 were not observed in the final modes of apo-PvLDH and pL1:PvLDH complexes owing to weak electron density. The B-factors for DNA are considerably higher than those for proteins in DNA:protein complexes. In detail, the B-factors for solvent-exposed nucleotides are higher than those for protein-interacting nucleotides 51 . Moreover, the RNA backbone is significantly more flexible than that of protein in RNA:protein complexes 52 . A comparison of the B-factors in protein:DNA aptamer complexes shows that the B-factors for DNA aptamers are significantly higher than those for proteins (Table S8). Therefore, the B-factors for nucleic acids are generally higher than those of protein in the complexes. The pL1:PvLDH complex also follows this general trend (Table S1).
Non-radioactive EMSA. DNA probes (pL1 and pL1 mutants) were synthesised using OPC purification at Bionics (Seoul, Korea). Oligonucleotides were dissolved in the binding buffer or binding buffer without magnesium ions. For proper folding, oligonucleotides were denatured at 90 degrees for 5 minutes and incubated at 4 degrees for 1 hour. Next, 10 pmol of oligonucleotides was mixed with 200 pmol PvLDH or hLDH (Novoprotein) for 1 hour at room temperature in the EMSA buffer (20 μ l) or EMSA buffer without MgCl 2 . The EMSA buffer contains 20 mM HEPES pH 8.0, 5% (w/v) glycerol, 50 mM NaCl, 5 mM KCl, 5 mM MgCl 2 , and 5 mM DTT. Samples were loaded into the wells of 10% polyacrylamide gel. The gel was run at 120 V for 1 hour at 4 degrees in 0.5× TBE buffer pH 8.5 containing 45 mM Tris-borate and 1 mM EDTA. DNA and protein molecules were stained with SYBR Green I and Coomassie Brilliant Blue G-250, respectively. pymol.org), Jmol (http://www.jmol.org) and CCP4MG 53 . Ramachandran plot analysis was performed by RAMPAGE webserver (http://mordred.bioc.cam.ac.uk/~rapper/rampage.php) 54 . Interfaces between DNA aptamers and proteins were analyzed with PyMol, PISA, COCOMAPS web server (https://www.molnac.unisa.it/ BioTools/cocomaps/index.p) and CONTACT program from the CCP4 suite 55 . Protein shapes were determined by Concave Finder employing modified Cx algorithm (http://cib.cf.ocha.ac.jp/bitool/CONCAVE) 42 . The magnesium ion position was validated by CheckMyMetal web server (http://csgid.org/csgid/metal_sites) 56 . All parameters were acceptable for magnesium ions (Fig. S3, Table S2). The secondary structures of pL1 mutants were confirmed using mfold web server (http://www.bioinfo.rpi.edu/applications/mfold) 57 . Sc values were calculated with Sc software 58 from the CCP4 suite of programs 54 . Sc_radii.lib file was modified owing to the absence of the van-der-Waals radii for nucleic acid atoms. Atomic radii for proteins were set according to the default in the sc_radii.lib file. Atomic radii for DNA were set according to those for chemically similar atoms of proteins.
Data availability: Apo-PvLDH and two forms of pL1:PvLDH complexes were deposited in the Protein Data Bank under PDB ID codes: 5HS4, 5HRU and 5HTO, respectively.