Reca-Binding Pile G4 Sequence Essential for Pilin Antigenic Variation Forms Monomeric and 5&Prime

Structure Article

RecA-Binding pilE G4 Sequence Essential for Pilin Antigenic Variation Forms Monomeric and 50 End-Stacked Dimeric Parallel G-Quadruplexes

Vitaly Kuryavyi,1,* Laty A. Cahoon,2 H. Steven Seifert,2 and Dinshaw J. Patel1,* 1Structural Biology Program, Memorial Sloan-Kettering Cancer Center, New York, NY, 10065, USA 2Department of Microbiology-Immunology, Northwestern University Feinberg School of Medicine, Chicago, IL 60611, USA *Correspondence: [email protected] (V.K.), [email protected] (D.J.P.) http://dx.doi.org/10.1016/j.str.2012.09.013

SUMMARY varies at a high frequency by a gene conversion process (Criss et al., 2005). Pilin antigenic variation results from nonreciprocal Neisseria gonorrhoeae is an obligate human DNA recombination between one of many silent pilin loci termed pathogen that can escape immune surveillance pilS and the pilin expression locus, pilE (Hagblom et al., 1985). through antigenic variation of surface structures Genetic, pharmacological, and biophysical experiments show that pilin antigenic variation requires a cis-acting DNA element such as pili. A G-quadruplex-forming (G4) se- 0 0 situated upstream of pilE with the sequence 5 -G3TG3TTG3TG3 quence (5 -G3TG3TTG3TG3) located upstream of the N. gonorrhoeae pilin expression locus (pilE)is that forms a G-quadruplex structure in the bacterial cell and in- vitro (Cahoon and Seifert, 2009). necessary for initiation of pilin antigenic variation, G-quadruplexes are four-stranded structures formed by a recombination-based, high-frequency, diversity- G-rich sequences through formation of stacked G,G,G,G generation system. We have determined NMR-based tetrads (G-tetrads) in monovalent cation-containing solutions structures of the all parallel-stranded monomeric (Burge et al., 2006; Neidle, 2009; Patel et al., 2007). The length 0 and 5 end-stacked dimeric pilE G-quadruplexes in and number of individual G-tracts and the length and sequence monovalent cation-containing solutions. We demon- context of linker residues define the diverse topologies adopted strate that the three-layered all parallel-stranded by G-quadruplexes. Bearing homology rather than complemen- monomeric pilE G-quadruplex containing single- tarity of four interacting bases, and thus devoid of coding func- residue double-chain reversal loops, which can be tions, G-quadruplexes have long been considered as genomic modeled without steric clashes into the 3 nt DNA- outliers (Davis, 2004). However, recently, G-quadruplexes have binding site of RecA, binds and promotes E. coli drawn significant interest because of an ever-expanding reper- toire of experiments indicative of their potential biological RecA-mediated strand exchange in vitro. We discuss roles impacting oncogenic promoter regions (Balasubramanian how interactions between RecA and monomeric pilE et al., 2011; Qin and Hurley, 2008), telomeric DNA (Paeschke G-quadruplex could facilitate the specialized recom- et al., 2005; Smith et al., 2011), regions of genomic instability bination reactions leading to pilin diversification. (Ribeyre et al., 2009), and the discovery of enzymes acting on G-quadruplex topology (Huber et al., 2002; Wu et al., 2008; reviewed in Lipps and Rhodes, 2009; Maizels, 2006; Sissi INTRODUCTION et al., 2011). It was also earlier hypothesized that intermolecular parallel-stranded G-quadruplexes may be involved in the align- DNA recombination is common to all organisms and is utilized to ment and recombination of sister chromatids during meiosis repair DNA and generate genetic diversity. In normal cells, most (Liu and Gilbert, 1994; Sen and Gilbert, 1988). DNA recombination reactions occur at low frequency, but in We have recently initiated a systematic analysis of possible many genetic diversity-generating systems, such as immunoglo- molecular mechanisms of normal and pathogenic genome rear- bin class switching, yeast mating-type switching, or pathogen- rangements based on various G-quadruplexes as alternative esis-associated antigenic variation, programmed recombination precursor structures for recombination or gene conversion. reactions occur at a relatively high frequency (Criss et al., 2005; Thus, we have identified a structural motif that exhibits similarity Haber, 1998; Maizels, 2006). DNA recombination is generally between the human intronic G-quadruplex and the active site of beneficial to the organism but can also be detrimental; such is group I intron (Kuryavyi and Patel, 2010), as well as invoked the case for chromosomal translocations, which are the molec- parallel-stranded crossing-over mediated by a novel dimeric ular signature for many types of cancer (Zhang et al., 2010). G-quadruplex formed by the c-kit2 intergenic region (Kuryavyi The obligate human pathogen Neisseria gonorrhoeae is et al., 2010). a Gram-negative bacterium, which relies on type IV pili for twitch- Here, we present solution structures of monomeric and 50 end- ing motility, adherence to host cells, and natural DNA transfor- stacked dimeric G-quadruplexes formed by the N. gonorrhoeae 0 mation (Rudel et al., 1992; Sparling, 1966; Wolfgang et al., pilE G4 sequence (5 -G3TG3TTG3TG3) in monovalent cation 1998). Pili are mainly composed of pilin, which antigenically solution. We show that the pilE G-quadruplex binds to

A Figure 1. pilE G4 DNA Oligomer Sequences Used in the Current Study, Their Imino GGGTGGGTTGGGTGGG NG16 NG19 TAGGGTGGGTTGGGTGGGG Proton NMR Spectra, and Their Mobilities GGGTGGGTTGGGTGGGG NG17 GGGTGGGTTGGGTGGG NG22 TA GAAT in Native PAGE GGGTGGGTTGGGTGGGGAAT NG20 AGAATAGGGTGGGTTGGGTGGGGAATTTT NG29 (A) pilE G4 DNA oligomers starting with 50G (NG16, NG17, and NG20) (left) and those containing a 50-preﬁx (NG19, NG22, NG29) (right). The core B NG16 NG19 C segment is highlighted in bold lettering. (B) Imino proton NMR spectra of aforementioned- 93del h-telo NG16 NG17 NG20 NG19 NG22 NG29 listed DNA oligomers in 100 mM KCl, 5 mM K-phosphate buffer (pH 6.8) at 25C. (C) Electrophoretic mobilities of dimeric NG16, NG17 NG22 NG17, and NG20 and monomeric NG19, NG22, and NG29 pilE G4 sequences are compared with } dimeric 93 del (93del) (Phan et al., 2005b) and

Dimers monomeric human telomere (h-telo) (Phan et al., 2007b) G-quadruplex markers in 25% acrylamide } gel with 25 mM KCl at 4C. NG20 NG29 Monomers

0 11.5 11 10.5 1 12 11.5 11 10.5 NG20 all begin with the 5 -G residue of H ppm 0 5 -G3TG3TTG3TG3), whereas the remaining sequences tested have extra bases Escherichia coli RecA. RecA binds to the pilE G4 with affinity at their 50 end (50-TA in NG19 and NG22, and 50-AGAATA in similar to single-stranded DNA (ssDNA) but does not bind other NG29) (Figure 1A). Contrary to the profound impact of residues G-quadruplex structures. We also show that the pilE G-quadru- at the 50 end, addition of residues at the 30 end has no impact plex can promote E. coli RecA-mediated strand exchange. In on either of the two distinct guanine imino proton spectral addition, we hypothesize how interactions between RecA and patterns (Figure 1B). monomeric pilE G-quadruplex could facilitate the specialized We have monitored the native gel electrophoresis migration recombination reactions leading to pilin diversification. patterns of various pilE G4-containing DNA oligomers (NG16– NG29) so as to elucidate the oligomeric state of these sequences RESULTS that exhibit the two distinct guanine imino proton spectral patterns (Figure 1B). The DNA oligomers that begin with a 50-G Impact of the 50-Prefix Sequence on G-Quadruplex residue (NG16, NG17, and NG20) and exhibited a narrower imino Formation proton spectral pattern (Figure 1B, left panel) migrated as dimers The N. gonorrhoeae pilE G4 sequence is essential for pilin anti- in both K+ (Figure 1C) and Na+ (data not shown) cation-contain- genic variation and is located upstream of pilE in the gonococcal ing solution, whereas those that had extra bases at the 50 end genome (Cahoon and Seifert, 2009). The pilE G4 16-mer (NG19, NG22, and NG29) exhibited a more dispersed imino 0 sequence (5 -G3TG3TTG3TG3) exhibits perfect mirror symmetry proton spectral pattern (Figure 1B, right panel) and migrated centered between two central T residues (NG16, Figure 1A). predominantly as monomers (Figure 1C). Molecularity based One extra G nucleotide at the 30 end (NG17), as well as AT- on gel migration has its limitations because molecular shape, containing segments at one or both ends (NG19, NG20, NG22, net charge, and dangling ends can impact on mobility of and NG29), breaks this symmetry (Figure 1A). We have system- G-quadruplexes. The gel migration of dimeric NG29 G-quadru- atically recorded NMR spectra of sequences NG16–NG29 (Fig- plex is at the border of monomeric and dimeric G-quadruplexes, ure 1A) in monovalent cation solution so as to investigate the and this could reflect the contributions of 6 nt dangling 50 and 30 effect of flanking sequences on G-quadruplex formation by the ends on either side of the G-quadruplex fold for this sequence. 0 5 -G3TG3TTG3TG3 pilE G4 sequence. We observed two distinct imino proton spectral patterns as a function of flanking sequence Proton NMR Assignments of NG19 and NG22 pilE G4 between 10.5 and 12.0 ppm (Figure 1B), a region characteristic Sequences of guanine imino protons involved in G-tetrad formation. The numbering systems for the NG19 and NG22 sequences are 0 0 0 One spectral pattern, adopted by NG19, NG22, and NG29 as follows: 5 -T1AGGG5TGGGT10TGGGT15GGGG-3 and 5 - 0 sequences (Figure 1A), exhibits guanine imino proton reso- T1AGGG5TGGGT10TGGGT15GGGGA20AT-3 , with the 16-mer nances dispersed over a wide spectral range between 10.7 core segment highlighted in underlining. We observe 12 imino and 12.2 ppm (Figure 1B, right panel). The other spectral pattern, protons between 11.0 and 12.1 ppm in the exchangeable proton adopted by NG16, NG17, and NG20 sequences (Figure 1A), NMR spectrum of the NG19 sequence in 100 mM KCl, 5 mM exhibits partially overlapping guanine imino proton resonances K-phosphate, H2O (pH 6.8), at 25 C, in a spectral range charac- dispersed over a narrower spectral range between 10.6 and teristic of guanine imino protons involved in N-H,,O hydrogen 11.5 ppm (Figure 1B, left panel). The essential defining charac- bond formation, a feature characteristic of G-tetrad formation teristic between the two imino proton spectral patterns for the (Figure 2A) (reviewed in Patel et al., 2007). We observe doubling listed DNA sequences is their 50 end prefix (NG16, NG17, and of intensity for proton chemical shifts of four peaks (at 11.78,

Figure 2. Exchangeable Proton NMR Spectra and Proton Assignments of Mono- meric NG19 and NG22 pilE G4 Sequences (A) Imino proton NMR spectrum of NG19 sequence (0.7 mM strands) in 100 mM KCl, 5 mM K-phosphate buffer (pH 6.8) at 25C. Imino proton assignments are listed over the spectrum. (B) Imino protons of NG19 sequence were assigned in 15N-ﬁltered NMR spectra of samples containing 2% uniformly 15N,13C-labeled guanines at the indicated positions. (C) Imino proton NMR spectrum of NG22 sequence (0.8 mM strands) in 100 mM KCl, 5 mM K-phosphate buffer (pH 6.8) at 25C. Imino proton assignments were established by selective labeling, as well as comparison with assignments of the imino proton spectrum of the NG19 sequence. (D) Guanine H8 proton assignments of NG22 sequence (6 mM strands) based on through-bond correlations between assigned imino and to-be assigned H8 protons via 13C5 at natural abundance, using long-range J couplings shown in (F). (E) Guanine H8 proton assignments of NG19 sequence (1.7 mM strands) using assignment protocol as outlined in (D). (F) A schematic drawing indicating long-range J couplings used to correlate imino and H8 protons within the guanine base. See also Table S1.

11.42, 11.38, and 11.05 ppm), reflective of spectral overlap at Monomeric G-Quadruplex Fold of NG22 pilE G4 these positions. We have made unambiguous site-specific Sequence guanine imino proton assignments by recording 15N-filtered The observation of 12 guanine imino protons between 11.0 and imino proton NMR spectra on samples containing 2% uniformly 12.2 ppm in the NG22 sequence spectrum (Figure 2C) is 15N,13C-labeled guanines at the indicated positions (Figure 2B). consistent with G-quadruplex formation stabilized by three The resulting imino proton assignments are listed above the stacked G-tetrads. We have assigned guanines within each of control NMR spectrum of the NG19 sequence in Figure 2A, the three G-tetrads by monitoring NOEs between assigned with chemical shift overlap observed for four pairs (G3/G7, G4/ guanine imino and H8 protons around individual G-tetrad planes G9, G13/G17, and G5/G18) of imino protons. (shown schematically in Figure 3B) in the expanded NOESY The corresponding imino proton spectrum of the NG22 spectrum (mixing time 250 ms) of the NG22 quadruplex in 2 sequence is shown in Figure 2C, with assignments based on a H2O buffer solution (Figure 3C). The observed imino-H8 subset of labeling experiments, listed over the spectrum. This NOE connectivities (Figure 3C) identify formation of G3,G7, spectrum exhibits resolved imino proton chemical shifts for G3 G12,G16 (assignments in red), G4,G8,G13,G17 (assignments and G7, as well as for G5 and G18 (Figure 2C), thereby resolving in blue), and G5,G9,G14,G18 (assignments in orange) part of the assignment ambiguity due to spectral overlap. The G-tetrads (Figure 3D). imino protons were next correlated with their assigned H8 We also observe four slowly exchanging guanine imino protons within individual guanines based on through-bond protons, assignable to G4, G8, G13, and G17, in the imino proton correlations via 13C at natural abundance for both the NG22 (Fig- spectrum of NG19 recorded 40 min following transfer from ure 2D) and NG19 (Figure 2E) sequences using long-range J 100 mM KCl, 5 mM phosphate, H2O solution (after lyophilization) 2 couplings shown schematically in Figure 2F(Phan, 2000; Phan to its H2O counterpart (Figure 3E). Because the imino protons of and Patel, 2002). Such an approach, combining data for both the G4,G8,G13,G17 G-tetrad exhibit the slowest exchange NG19 (Figures 2A and 2E) and NG22 (Figures 2C and 2D) rates (Figure 3E), this G-tetrad must constitute the middle layer sequences, resolved overlap ambiguities and provided guanine of the monomeric G-quadruplex, thereby defining the fold shown imino proton and H8 assignments for both sequences. in Figure 3F. All guanines adopt anti glycosidic torsion angles, We next traced NOEs between base protons and their own and all three loops are of the double-chain reversal type (Wang 50-flanking sugar H10 protons in an expanded NOESY contour and Patel, 1994) for the deciphered fold of the monomeric 2 plot (250 ms mixing time) for the NG22 sequence in H2O buffer NG22 G-quadruplex (Figure 3F). solution (Figure 3A), thereby yielding base and sugar H10 proton We did note that the NOE walk (Figure 3A) could be monitored assignments. The remaining sugar protons were next assigned for 50 end TA and 30 end GAAT segments, suggestive of a by through-bond COSY and TOCSY experiments, with base helical-like arrangement at both ends. By contrast, the NOE and sugar proton chemical shifts listed in Table S1, which is walk was interrupted at G5-T6-G7, G9-T10, and G14-T15-G16 available online. We did not observe strong NOEs between steps (Figure 3A), suggestive that these thymines, connecting base and sugar H10 protons at short (50 ms) mixing times, indic- GGG tracts, that are part of the double-chain reversal loops, ative of anti-torsion angles (Patel et al., 1982) for all guanines of were most likely extruded out of the monomeric G-quadruplex the NG22 sequence. fold (Figure 3F).

A pairwise rmsd values for the stacked G-tetrad core in the 0.44 A˚ range (Table S2). A20 5.6 The first and the third single-residue loops are of the double- A2 A21 G19 G18 T1 5.8 chain reversal type, with T6 and T15 nucleotides directed out G17 G16 T22 of the groove. However, T6 is positioned closer to the wall of G3 6.0 the groove and inclined toward residue T10 from the neighboring G4 6.2 G12 G7 G8 loop (Figures 4A, S1B, and S1D). The second loop is also of the G9 T6 G13 T10 G5 6.4 double-chain reversal type but is composed of 2 nt: T10 and T11. T15 G14 Residue T10 turns toward the neighboring groove (Figures 4A 6.6 T11 and S1A), whereas residue T11 is fixed by hydrogen bond 1 8.2 8.0 7.8 7.6 7.4 7.2 H ppm formation between atom O2 of T11 and amino proton of G8 B D E (Figure S1A) inside the groove formed by the G7-G8-G9 and Gδ H8 H8 G4 G12-G13-G14 columns. 0 Gα 3 7 12 16 G8 G13 Surprisingly, the 5 -TA overhang caps the top G3,G7, H1 G17 4 8 13 17 , H1 H1 G12 G16 tetrad layer, forming a T-A platform (Cate et al., H1 5 9 14 18 1996), with the residue A2 inclined out of plane (Figures 4A and Gγ 4D). We also observe an unanticipated topology for the single- 11.8 11.4 1H ppm H8 H8 0 Gβ stranded overhang at the 3 end of the all parallel-stranded C F monomeric pilE G-quadruplex fold. The G19-A20-A21-T22 G3 G16 G7 segment makes a sharp turn at the junction between A20 and 14/18 7.2 G12 8/13 G4 A21, bringing T22 in proximity with the bottom G5,G9,G14,G18 17/4 7.6 G17 G8 12/16 3/7 9/14 18/5 tetrad (Figures 4A and 4E). We observe NOEs between the 7/12 13/17 8.0 G13 4/8 16/3 5/9 G5 8.4 methyl protons of T22 and imino protons of G5 and G9, between 1 G18 12.0 11.8 11.6 11.4 11.2 11.0 H ppm G9 G19 G14 H6 of T22 and imino proton of G5, between the methyl proton of T22 and H10 sugar proton of G9, and between H10 sugar proton Figure 3. NOESY Contour Plots, Hydrogen-Deuterium Exchange, of T22 and imino proton of G5. These NOEs, along with intra- and and Schematic Representation of NG22 and NG19 pilE G4 inter-residue NOEs observed for other residues within the G19- Sequences Adopting Monomeric G-Quadruplex Fold T22 segment, define a unique ‘‘30 hook’’ configuration, with no (A) Expanded NOESY contour plot (250 ms mixing time) correlating base and 0 hydrogen bond interactions between its constitutive nucleotides sugar H1 nonexchangeable protons of NG22 sequence (1.7 mM strands) in (Figures 4A, 4B, and 4E). 100 mM KCl, 5 mM K-phosphate buffer (pH 6.8) at 25C. The lines trace NOEs between base protons (H8 or H6) and their own and 50-flanking sugar H10 protons. Intra-residue base-to-sugar H10 NOEs are labeled with residue Proton NMR Assignments of NG16 pilE G4 Sequence numbers. The N. gonorrhoeae pilE G4 symmetric 16-mer core sequence 0 0 (B) Schematic of guanine imino-guanine H8 connectivities around a G-tetrad. 5 -G1GGTG5GGTTG10GGTGG15G-3 , labeled NG16, migrates (C) Expanded NOESY contour plot (mixing time, 250 ms) showing imino-to-H8 in native gel electrophoresis as a dimer, both in K+ (Figure 1C) connectivities of NG22 sequence (1.7 mM strands) in 100 mM KCl, 5 mM and Na+-containing solution. The imino proton NMR spectra of K-phosphate buffer (pH 6.8) at 25C. Cross-peaks identifying connectivities NG16 have been recorded in both K+- and Na+-containing solu- within individual G-tetrads are color coded, with each cross-peak framed and labeled with the residue number of imino proton in the first position and that of tion. Because single inosine for guanine substitutions has been the H8 proton in the second position. shown to enhance the resolution of both imino and aromatic (D) Guanine imino-guanine H8 connectivities observed for G3,G7,G12,G16 protons, we substituted I15 for G15 in NG16 with resultant (red), G4,G8,G13,G17 (blue), and G5,G9,G14,G18 (orange) G-tetrads. improvement in spectral quality, for spectra recorded in Na+ (E) Imino proton NMR spectrum of monomeric NG19 sequence form recorded cation-containing solution (Figure 5A). We observe 11 imino 1 hr following transfer from H O (after lyophilization) to 2H O solution. 2 2 protons between 10.5 and 12.0 ppm in Figure 5A (the down- Assignments of slowly exchanging imino protons are listed over the spectrum. (F) Schematic representation of the all parallel-stranded monomeric NG19 pilE field-shifted I15 imino proton is outside this range), requiring G-quadruplex fold. Anti bases are indicated in cyan and loop connectivities that the dimer be composed of two symmetry-related mono- shown by red lines. mers. We assigned the guanine imino protons following site- See also Figures S2 and S11 and Table S1. specific incorporation of 2% 15N-labeled guanines one at a time into the I15-modified NG16 sequence and recorded Solution Structure of the Monomeric NG22 pilE 15N-filtered imino proton NMR spectra (Figure 5B) (Phan and G-Quadruplex Patel, 2002), which yielded the assignments listed over the Initial distance-restrained and subsequent intensity-restrained NMR spectrum in Figure 5A. The guanine H8 protons were molecular dynamics calculations of the solution structure of next linked through long-range through-bond J-coupling exper- monomeric pilE NG22 G-quadruplex were guided by exchange- iments to the assigned guanine imino protons (Phan, 2000)as able and nonexchangeable proton restraints (numbers listed by shown in Figure 5C, with the exception of I15, whose H8 proton category in Table S2). The ensemble of 12 refined superposi- stands out due to its significantly downfield-shifted position. tioned structures is shown in stereo (Figure 4A), with a represen- The expanded NOESY contour plot (300 ms mixing time) tative refined structure in the same orientation shown in correlating base and sugar H10 protons of I15-modifed NG16 is ribbon (Figure 4B) and surface (Figure 4C) representations. The plotted in Figure 5D, together with tracing of NOE connectivities ensemble of refined structures is well converged, exhibiting between the base protons and their own and 50-flanking sugar

A A2 T1 A2 T1 Figure 4. Solution Structure of the Mono- meric NG22 pilE G-Quadruplex in K+ Solution G16 7G G16 7G G12 3G G12 3G 6T 6T (A) Stereo view of 12 superpositioned refined structures of the monomeric NG22 pilE G-quad- G13 G13 ruplex. Guanine bases with anti glycosidic torsion T11 8G T11 8G angles in the G-tetrad core are colored in cyan. G19 G19 Bases in connecting loops and in 50 and 30 T15 G14 T22 T15 G14 T22 T10 T10 elements are shown in brown, with the backbones A20 A20 in gray, sugar ring oxygens in red, and phosphorus atoms in yellow. 12A 12A (B) Ribbon view of a representative refined structure of the monomeric NG22 pilE G-quadruplex. B C T1 (C) Surface view of a representative refined struc- A2 G16 G7 G12 G3 ture of the monomeric NG22 pilE G-quadruplex. (D) Platform stacking of the 50-TA prefix dinucleo- 0 G17 G8 tide over the 5 tetrad. G13 (E) Base alignment within the four-residue G19- G18 T11 G4 0 T15 T6 A20-A21-T22 3 hook element. See also Figure S1. G9 T10 G14 G5 G19 T22

A20 A21 protons (Figure 6D, bottom) over a period D E of 6 hr, with protons fully substituting for G3 deuterons after 30 hr (Figure S2B). G14 G5 The resulting dimeric NG16 pilE G7 G18 G9 G-quadruplex fold, which integrates elec- A2 T22 trophoretic mobility (Figure 1C), proton T1 exchange (Figure 6D), and NMR spectral A21 G19 analysis (Figure 6A), is shown schemati- G16 G12 A20 cally in Figure 6E. Within this dimeric G-quadruplex, four imino protons of the outer G3,G7,G12,G16 G-tetrad (Fig- ure 6E) exchange rapidly, whereas eight H10 protons. The remaining sugar protons of I15-modified NG16 imino protons from the internal G2,G6,G11,G15 and intersubu- were assigned by through-bond COSY and TOCSY experi- nit-stacked G1,G5,G10,G14 G-tetrads (Figure 6E) are pro- ments, with chemical shifts listed in Table S3. tected from direct exchange with protons from the solvent. All residues of I15-modified NG16 show comparable intensi- ties for their base to H10 NOEs at short 50 ms mixing time Solution Structure of the 50 End-Stacked Dimeric (data not shown), implying that all glycosidic torsions are anti. NG16 pilE G-Quadruplex All T residues, except at the T9-G10 step, do not show sequential Initial distance-restrained and subsequent intensity-restrained connectivity to 50-neighboring residues (Figure 5D). molecular dynamics calculations of the solution structure of the dimeric I15-modified NG16 pilE G-quadruplex were guided Dimeric G-Quadruplex Fold of NG16 pilE G4 Sequence by exchangeable and nonexchangeable proton restraints listed We have assigned guanines to specific G-tetrads by monitoring by category in Table S4. We observed good convergence within NOEs between guanine imino and H8 protons around individual the 11 refined superpositioned structures (Figure 7A), with the G-tetrad planes (Figures 6A–6C) in the NOESY spectrum (mixing ensemble exhibiting pairwise rmsd values in the 0.40 A˚ range time 250 ms) of the dimeric I15-modified NG16 pilE G4 for the stacked G-tetrad dimeric core component (Table S4). A sequence. These studies allow identification of a G-quadruplex representative refined structure of the dimeric G-quadruplex fold composed of G3,G7,G12,G16, G2,G6,G11,I15, and viewed in the same orientation is shown in Figure 7B, and its G1,G5,G10,G14 G-tetrads for each monomer within the dimer. surface representation is shown in Figure 7C. We observed a very distinct proton-to-deuteron exchange We observe significant stacking overlap between guanines of pattern for the dimeric NG16 and I15-modified NG16 pilE the 50 end-stacked interfacial G1,G5,G10,G14 G-tetrads in the G-quadruplexes. Four imino protons assigned to G3, G7, G12, structure of the dimeric NG16 pilE G-quadruplex (Figure 7D). In and G16 exchanged rapidly upon dissolving a lyophilized addition, T4 (Figures 7A and 7E) and T13 (Figure 7A) single-linker 2 NG16 sample into H2O, whereas the other eight protons re- residues are positioned in partially inward-pointing configura- mained inaccessible (Figure S2A) for several hours (data not tions and directed toward their grooves, with this orientation shown). In a reverse experiment, after transfer of the deuterated supported by observed NOEs between loop residues (T4 or

I15-modiﬁed NG16 sample into H2O, we observed the rapid T13) and their neighboring residues. appearance of G3, G7, G12, and G16 imino proton signals (Fig- The all parallel-stranded stacked G-tetrad cores of in- ure 6D, top), with delayed gradual appearance of the remaining dividual monomeric units of the dimeric NG16 pilE G-quadruplex

G5 Figure 5. Exchangeable Proton NMR A G6 G7G16 Spectra and Proton Assignments of Dimeric G2 G12 I15-Modified NG16 pilE G4 Sequence G11 G14 G3 C 13 G10 C ppm (A) Imino proton spectrum of I15-modified NG16 G1 117.8 sequence (0.3 mM strands) in 100 mM NaCl, 5 mM G5 G14 G14 G5 G10 G1 Na-phosphate buffer (pH 6.8) at 25C. Unambig- 118.0 G1 G16 G3 G10 G3 G16 uous imino proton assignments are listed over the G6 G6 G11 G12 G12 G11 118.2 spectrum. 12.0 11.5 11.0 1H ppm G7 G7 (B) Imino protons of I15-modified NG16 sequence G2 G2 15 B 118.4 were assigned in N-filtered NMR spectra of G1 11.6 11.2 10.8 1H ppm 7.8 7.6 samples containing 2% uniformly 15N,13C-labeled G2 guanines at the indicated positions. 300 ms (C) Guanine H8 proton assignments of I15-modi- G3 D fied NG16 sequence (3.5 mM strands) based G5 G2 G1 G6 6.0 on through-bond correlations between assigned G6 imino and to-be assigned H8 protons via 13C5 at G10 natural abundance, using long-range J couplings G7 G11 G16 6.2 G14 G5 G7 shown in Figure 2F. G10 G12 (D) Expanded NOESY contour plot (300 ms G11 I15 G3 6.4 mixing time) correlating base and sugar H10 non- T13 G12 T4 T8 exchangeable proton assignments of I15-modified NG16 sequence (1.7 mM strands) in 100 mM NaCl, G14 T9 6.6 5 mM Na-phosphate buffer (pH 6.8) at 25C. The 8.2 8.0 7.8 7.6 1H ppm 12.0 11.5 11.0 1H ppm line connectivities trace NOEs between base protons (H8 or H6) and their own and 50-flanking sugar H10 protons. Intra-residue base-to-sugar H10 NOEs are labeled with residue numbers. See also Figures S2 and S11 and Table S3.

(Figures 7A and 7B) are similar to their counterpart within the alignment shown in rotamer I at the dimeric interface monomeric pilE NG22 G-quadruplex (Figures 4A and 4B). (Figure 8A). For the I10-substituted NG16 dimeric pilE G-quadruplex, the Stacking Alignment of 50-G-Tetrads in the Dimeric H2 proton of I10 is predicted to show a NOE to the H8 proton NG16 pilE G-Quadruplex of G14 in rotamer I (Figure 8C, labeled a), to the H8 proton of We observe a set of NOE cross-peaks between stacked G1 in rotamer II (Figure S3D, labeled d), to the H8 proton of G5 interfacial G1,G5,G10,G14 G-tetrads that supports the 50 in rotamer III (Figure S3E, labeled e), and the H8 proton of I10 end-stacked arrangement of the two monomeric units that in rotamer IV (Figure S3F, labeled f). Experimentally, we observe constitute the dimeric pilE G-quadruplex (Figure 7D). These a strong NOE between the H2 proton of I10 and the H8 proton of include NOEs involving proton pairs G10(H10)-G14(H10), G14 (peak a, Figure 8D), but not the H8 protons of G1, G5, and G1(H10)-G5(H10), G1(H10)-G5(H8), and G5(H10)-G1(H8) in the I10, strongly supporting the stacking alignment shown in rotamer NOESY spectra of dimeric I15-modified NG16 G-quadruplex I at the dimeric interface (Figure 8C). (Figure S2C). The remaining cross-peaks in the NOESY spectra of the The unique 50-stacking alignment of the dimeric NG16 pilE I1-substituted (Figure 8B) and I10-substituted (Figure 8D) G-quadruplex (Figure 7D) was further experimentally validated NG16 pilE G-quadruplexes have also been assigned and following analysis of NOE patterns for the dimeric I1- and analyzed to ensure that they are consistent with the stacking I10-substituted NG16 pilE G-quadruplexes. Substitution of arrangement corresponding to rotamer I and the exclusion of inosine (with its H2 proton) for guanine at the interfacial rotamers II, III, and IV. G1,G5,G10,G14 G-tetrad of NG16 results in characteristic NOE patterns that allow definitive differentiation between alter- Alignment of Central TT Double-Chain Reversal Loop nate stacking arrangements (Phan et al., 2005a). Because of Loops containing G-tracts play critical roles in defining G-quad- the 4-fold symmetry of the all-anti guanine tetrad, there are theo- ruplex topology (Gue´ din et al., 2010). Similar to what was retically four rotational isomers involving stacked guanines observed in the monomeric pilE G-quadruplex fold (Figure 4A), possible at the dimeric interface. the T8-T9 double-chain reversal loop in the dimeric pilE G- For the I1-substituted NG16 dimeric pilE G-quadruplex, the quadruplex has the T9 residue positioned inside of the groove H2 proton of I1 is predicted to show a NOE to the H8 proton (Figure 7A), stabilized by hydrogen bonding between the O2 of G5 in rotamer I (Figure 8A, labeled a), to the H8 proton of atom of T9 and the amino proton of G6. This alignment of T9 in G10 in rotamer II (Figure S3A, labeled a), to the H8 proton of the groove is supported by the following observed NOEs: G14 in rotamer III (Figure S3B, labeled b), and the H8 proton G10(H30)-T9(H10), in I10-substituted spectrum; and T9(H10)- of I1 in rotamer IV (Figure S3C, labeled c). Experimentally, G10(H8, H20,H10) and T9(H40)-G10(H50), in I15-substituted spec- we observe a strong NOE between the H2 proton of I1 and trum. The T8 residue is directed toward the 30-neighboring the H8 proton of G5 (peak a, Figure 8B), but not the H8 groove (Figure 7A), similar to its counterpart in the monomeric protons of G10, G14, and I1, strongly supporting the stacking pilE G-quadruplex structure.

A D G16G12 Figure 6. NOESY Contour Plots, Hydrogen- Deuterium Exchange, and Schematic pilE 5 min Representation of I15-Modiﬁed NG16 G3G7 D O 0 12/16 2 G4 Sequence Adopting 5 End-Stacked 16/3 5/10 Dimeric G-Quadruplex Fold 10/14 7.5 7.0 1/5

2/6 (A) Expanded NOESY contour plot (mixing time: 250 ms) showing imino-to-H8 connectivities of 8.0

3/7 30 hr I15-modiﬁed NG16 sequence (3.5 mM strands) in 7/12

14/1 H O 11/15 6/11 2 100 mM NaCl, 5 mM Na-phosphate buffer (pH 6.8) 8.5 at 25 C. Cross-peaks identifying connectivities 11.8 11.6 11.4 11.2 11.0 10.0 1 1 H ppm 14.0 13.0 12.0 11.0 H ppm within individual G-tetrads are color coded, with each cross-peak framed and labeled with the residue number of imino proton in the ﬁrst position B Gδ H8 H8 E G3 and that of the H8 proton in the second position. G7 (B) Guanine imino-guanine H8 connectivities Gα H1 G16 around a G-tetrad. G12 (C) Guanine imino-guanine H8 connectivities H1 H1 T4 G2 G6 observed for G1,G5,G10,G14 (red), G2,G6, H1 Gγ G15 G11,G15 (green), and G3,G7,G12,G16 (orange) G11 G1 G-tetrads. H8 H8 G5 (D) Imino proton spectrum of I15-modiﬁed NG16 Gβ G14 G10 sequence dimeric form recorded 5 min and 30 hr G5 2 following transfer from H2O (after lyophilization G1 2 C G10 from H2O solution) to H2O solution. Assignments 1 5 10 14 G14 G6 of fast exchanging imino protons are listed over G2 the spectrum. G11 (E) Schematic representation of the 50 end- 2 6 11 15 G15 G7 T4 stacked all parallel-stranded NG16 dimeric pilE G3 G-quadruplex fold. Anti bases are indicated in 3 7 12 16 G12 cyan and connectivities shown by red and black G16 lines for two monomers related by symmetry. See also Figures S2 and S9 and Table S3.

Susceptibility of pilE G-Quadruplexes to S1 Nuclease and double-stranded DNA (dsDNA) in filaments of 3-nt-unit Current models of recombination mechanisms imply initial increments (Figures 9A and S5A) (Chen et al., 2008). Because single- or double-strand breaks by nucleases, and hence, we the monomeric G-quadruplex unit is composed of three stacked were interested in the sensitivity to nuclease cleavage of con- G-tetrads (Figure 3F), we postulated that the all parallel-stranded necting loops within G-quadruplex folds. Digestion of the mono- monomeric pilE G-quadruplex might fit into the 3-nt-binding site meric NG22 pilE G-quadruplex (in K+ cation-containing solution) of a monomeric RecA unit (Chen et al., 2008). The amino acid by S1 nuclease leads to formation of a ladder on a denaturing sequences of RecA proteins from N. gonorrhoeae (RecANg) gel, with product lengths ranging between 16 and 22 nt (Fig- and E. coli (RecAEc) exhibit 81% similarity and 65% identity, ure S4A). After 30 min of reaction, the ladder clears out toward and the E. coli RecA can mediate pilin antigenic variation in the product migrating as 16 nt, without observation of shorter N. gonorrhoeae (Stohl et al., 2002, 2011). The RecA IGVMFGNP products in the range of 8–12 nt. These results indicate that S1 motif, which constitutes the loop that separates DNA into 3 nt nuclease digests single-stranded overhangs of the monomeric segments in RecAEc-DNA complex, is identical in both proteins, pilE G-quadruplex but does not cut into its core composed of except for a single substitution of Asn (in RecAEc) by Ser (in three stacked G-tetrads. RecANg). Therefore, the two proteins are expected to be rather Surprisingly, digestion of the NG16 dimeric pilE G-quadruplex similar in their structure and mechanism of action. results in two oligonucleotide bands migrating approximately as Fluorescence anisotropy was used to measure the affinity of 7-mers and 3-mers (Figure S4B). The intensity of the original E. coli RecA for different G-quadruplex-forming G-rich DNA 16-mer gradually decreases as a function of the reaction time substrates. RecA binds NG19-like monomeric G-quadruplexes at two tested (0.2 and 0.4 mM) DNA concentrations. Thus, the in monovalent cation solution with an apparent KD = 0.88 ± dimeric pilE G-quadruplex serves as a substrate of S1 nuclease 0.31 mM(Figure 9B), with this binding affinity comparable to and unexpectedly is digested within the G-tetrad core. that for RecA binding to oligo dT16 (KD = 0.64 ± 0.26 mM) (Fig- ure S6A). Interestingly, a three-layered G-quadruplex containing Binding of the Monomeric pilE G-Quadruplex to RecA a single-residue (minimal size) double-chain reversal loop (Fig- In N. gonorrhoeae, recombination events that occur at pilE and ure 9B, right panel) is a prerequisite for binding in K+ solution result in a new pilin antigenic variant are dependent on RecA because three-layered G-quadruplexes containing two- or (Koomey et al., 1987). Because pilin antigenic variation initiates three-residue double-chain reversal loops, formed by two other at the pilE G4 sequence, we investigated whether RecA could G-quadruplex sequences within the gonococcal genome, do bind to monomeric pilE G-quadruplex, thereby effectively cata- not bind RecA in K+ solution (Figure S6B; data not shown). These lyzing pilin antigenic variation in N. gonorrhoeae (Stohl et al., latter G-quadruplex-forming sequences were not able to substi- 2002). RecA has been shown to bind and assemble on ssDNA tute for the pilE G4 sequence to restore pilin antigenic variation

Figure 7. Solution Structure of the NG16 + T8 T8 Dimeric pilE G-Quadruplex in Na Solution A T4 T4 G12 G12 (A) Stereo view of 11 superpositioned refined G16 G7 G16 G7 structures of the 50 end-stacked dimeric NG16 G3 G3 pilE G-quadruplex. Guanine bases in the G-tetrad 51G 51G core are colored cyan (anti). Bases in connecting T13 G2 T13 G2 loops are shown in brown, with the backbones of two individual monomers in gray and orange, G1 G5* G5 G5* G5 G14 G10* G14 G10* G1 sugar ring oxygens in red, and phosphorus atoms in yellow. G6* G6* (B) Ribbon view of a representative refined *31T *31T structure of the 50 end-stacked dimeric pilE G-quadruplex. G12* G12* (C) Surface view of a representative refined *9T *9T T4* T4* structure of the 50 end-stacked dimeric pilE T8* T8* G-quadruplex. 0 B C (D) Stacking of two interfacial 5 -G-tetrads T8 (G1,G5,G10,G14) between monomers in the G12 G16 G7 dimeric pilE G-quadruplex. G3 (E) Side projection of the dimeric NG16 pilE T9 T4 G-quadruplex, showing mutual alignment and twist of the T4-T4* residue-connected grooves. T13 G2 G14 G10 See also Figure S8 and Tables S2 and S4.

G5 G10* G1 G1* G14* G5*

T13* G11* G2* facing the enzyme-binding pocket, T9* G6* T4* whereas the double-residue loop points G12* G16* G3* outward (Figures 9C and S5B). By con- T8* G7* trast, the same three-layered all parallel- stranded G-quadruplex with grooves connected by two- or three-residue D E G3 T4 G7 double-chain reversal loops (topology shown in Figure S6B, right panel) would G14* clash with the RecA-binding pocket, G14 G10 as would a four-layered G-quadruplex G2 G6 (Smith et al., 1995; Wang and Patel, 1995) (topology shown in Figure S6C, G10* right panel). These data show that some G5* G5 G1 G1* but not all G-quadruplex structures can G1* bind to RecA. G6* G2* G1 Strand Exchange Reaction G5 G5* G7* Mediated by Monomeric pilE G3* G-Quadruplex and RecA T4* The N. gonorrhoeae monomeric pilE G-quadruplex fits well in the space of the two DNA-binding sites (Figures 9C and (Cahoon and Seifert, 2009). In addition, the number of stacked S5B). Once formed as a bead on a ssDNA, the pilE G-quadruplex G-tetrads involved in G-quadruplex formation is also critical for could perhaps produce a seeding point for RecA binding and binding because a four-layered G-quadruplex (Smith et al., increase the efficiency of RecA-mediated strand exchange. To 1995; Wang and Patel, 1995) does not show affinity for RecA in test this hypothesis, a series of oligonucleotides were synthe- K+ solution and cannot substitute for the pilE G4 sequence to sized (listed in Table S5) either having the G-quadruplex (simpli- restore pilin antigenic variation (Figure S6C; data not shown). In fied designation as G4 in Figures 10 and S7) sequence or addition, the dimeric pilE G-quadruplex composed of six stacked a mutated G4 sequence on either the 30 end (Figure 10A) or 50 G-tetrad layers (Figure 6E) also does not bind RecA (Figure S6D). end (Figure S7A). These oligomers were first determined to Modeling studies based on the crystal structure of the RecA- form ssDNA, dsDNA, or G-quadruplex DNA by CD spectroscopy dsDNA complex (Figures 9A and S5A) (Chen et al., 2008) estab- (Figures 10B and S7B) and were then tested for RecA-mediated lish that the three-layered all parallel-stranded G-quadruplex strand exchange (Figures 10C and S7C). Oligomers having a containing single-residue double-chain reversal loops (Figure 4A) G-quadruplex structure showed more efficient strand exchange can be positioned in a unique alignment within the RecA mono- than a version of the oligomer where that G-rich sequence meric unit, with the no-residue groove and single-residue loops was mutated and could not form the G-quadruplex structure

Figure 8. Analysis of Interface through A B D2O NOEs of Reporter Point Substitution of I1

5H1´ for G1 and I10 for G10 for Rotamer I in the I1H8 h Dimeric NG16 pilE G-Quadruplex 6.2 1H1´ (A) Signature cross-peak of rotamer I in I1-sub- G14 G14 g stituted dimeric NG16 G-quadruplex is labeled ‘‘a.’’ 5H8 14H8 H O 10H8 (B) Observed and expected NOEs for four theo- 2 6.4 G10 G10 retically possible rotamers I–IV of I1-substituted a NG16 dimeric G-quadruplex with NOE assign- I1H1 10H1 I1H2

5H8 ments listed in the contour plot. 7.8 I 5H8 d 14H8 c (C) Signature cross-peak of rotamer I in I10-sub- 10H8 b f stituted dimeric NG16 G-quadruplex is labeled ‘‘a.’’ I1 I1 (D) Observed and expected NOEs for four theo- 8.0 I1H2 I1H2 I1H1 I1H8 retically possible rotamers I–IV of I10-substituted a 1 dimeric NG16 G-quadruplex with NOE assign- G5 G5 8.0 7.9 7.8 H ppm ments listed in the contour plot. Positions of nonexchangeable H2 proton of I1 in I1H1 e 8.3 (A) and nonexchangeable H2 proton of I10 in (C) 10H1 are colored in blue. 11.1 See also Figure S3.

14.2

C D D2O assembly of higher-order G-quadru-

6.0 plexes and the principles underlying their recognition. G14 G14 6.2 a 0 0

I10H8 1H8 Unique 5 and 3 End Features of the 5H8 I10 I10 H2O 14H8 Monomeric pilE G-Quadruplex I10H1´ 6.4 We have observed a two-residue over- 1H1 I10H2 0 I I10H1

7.8 hang at the 5 end that forms a T-A

d 14H8 5H8 c 14H8 platform (Figures 4A and 4D), which f a 0 G1 G1 caps the 5 G-tetrad of the monomeric 8.0 I10H2 I10H2 I10H8 I10H1 b NG22 pilE G-quadruplex. Previous struc- G5 G5 8.0 7.9 7.8 1H ppm tural studies have identiﬁed capped ends of G-quadruplexes, whereby T-A, I10H1 8.3 C-A, and A-A platforms form triads (Kur-

1H1 e yavyi and Jovin, 1995) by a hydrogen 11.6 bonding interaction with a third base 14.1 (Kettani et al., 1997, 2000a; Kuryavyi et al., 2000). In the present work, the TA platform has no interacting hydrogen (Figures 10D and S7D). These results suggest that the mono- bonds to a third base, thus establishing that stacking between meric pilE G-quadruplex structure may recruit RecA and facilitate a platform and a G-tetrad constitutes an energetically stabiliz- strand exchange during pilin antigenic variation. ing interaction. We postulate that the O2 atom of T1 could be involved in monovalent cation coordination in the center of DISCUSSION the top G-tetrad, which could enhance the stability of un- paired residues T1 and A2 in a platform configuration. Further- All Parallel-Stranded G-Quadruplexes more, formation of this platform prevents dimerization of two The propensity of the pilE G4 sequence to adopt all parallel- monomeric G-quadruplexes by stacking of their 50-terminal stranded monomeric and dimeric G-quadruplexes highlights G-tetrads, thereby explaining the role of the 50 overhang in the importance of this all parallel-stranded topology. This fold the observed switching between monomeric and dimeric pilE was first identified for the human telomere by crystal structure G-quadruplexes. determination (Parkinson et al., 2002) and subsequently ob- An unexpected snapback topology was also observed at the served for the c-myc sequence (Ambrus et al., 2005; Phan 30 end of the structure of the monomeric NG22 pilE G-quadru- et al., 2004) in solution, as well as other sequences containing plex (Figure 4B), where the four-base single-stranded overhang short linkers spanning G-tract segments (reviewed in Burge (G19-A20-A21-T22) forms a reverse turn, such that its terminal et al., 2006; Patel et al., 2007). In addition, the identification T22 residue is positioned for stacking over the 30 end G-tetrad of a 50 end-stacked all parallel-stranded dimeric pilE G-quadru- (Figure 4E). We anticipate that potential involvement of T22 in plex (Figure 6E) provides insights not only into the principles cation coordination together with the propensity for hydrogen of G-tetrad stacking at dimeric interfaces but also into the bond formation between its O2 atom and amino proton of G19

A Secondary Primary C Secondary Primary Figure 9. Ribbon Representations of X-Ray binding site binding site binding site binding site Structure of RecA in Complex with DNA, and Binding Studies and Model of RecA Bound by a Monomeric pilE G-Quadruplex (A) Ribbon view of dsDNA bound after strand K237 L1 K237 exchange in the primary binding site of RecA L1 filament (PDB 3CMX). Three bases of dsDNA R235 R235 are separated between two loops, L1 and L2, colored red. (B) Left panel shows fluorescence anisotropy (FA) L2 measurements of RecA binding to N. gonorrhoeae L2 L2 monomeric pilE G-quadruplex sequence in K+ 0 solution with apparent KD = 0.88 ± 0.31 mM. 5 - FAM-labeled pilE G-quadruplex sequence (top) Structure (pdb: 3CMX) Model with loop elements highlighted in red. Middle panel B illustrates circular dichroism (CD) spectrum of pilE + fam-TAGGGTGGGTTGGGTGGGG G-quadruplex in K solution. Right panel is a schematic representation of the three-layered all 2.0 parallel-stranded monomeric pilE G-quadruplex with single-residue loops. mdeg, millidegree. G3 (C) Ribbon view of model of monomeric pilE 1.0 G16 G7 G-quadruplex bound to primary and secondary G12 G4 binding sites of RecA filament. G17 G8 See also Figures S4–S6 and S10 and Table S5. 0.0 G13 G5 CD (mdeg) G18 G14 G9 -1.0

220 240 260 280 300 320 linker spanning a third groove, and Wavelength (nm) RecA:DNA Ratio no residue positioned within the remaining 50–30 groove (Figure 7B). In this and the hydrodynamic advantage of a compact 30 end most symmetry-governed configuration of the dimeric pilE G-quadru- likely together contribute to define the structure of the ‘‘30 plex, two-residue grooves continue into no-residue-containing hook’’ configuration of the G19-A20-A21-T22 overhang segment 50–30 grooves, whereas single-residue grooves continue into (Figure 4E). This 30 end topology contrasts with the snapback single-residue grooves with similar T residues from the two feature observed for c-myc and c-kit1 monomeric G-quadru- monomers (Figure 7E, residues T4 and T4* are oriented in plexes, whereby 30 end G residue(s) participates in completion opposite directions). This alignment appears to be favored for of 30-terminal G-tetrad(s) formation (Phan et al., 2005a, 2007a). positioning of the mass center and likely reflects favorable solu- The structure of the monomeric pilE G-quadruplex illustrates tion hydrodynamic properties of the dimeric pilE G-quadruplex how a G-quadruplex core could serve as a long-range coordina- with looped-out residues. tion center for optimal positioning of a single-stranded overhang at the 30 end. Comparison of Stacking Patterns among Dimeric G-Quadruplexes A50 End-Stacked Dimeric pilE G-Quadruplex There have been several examples of end-on-end association to A notable feature of the structure of the dimeric NG16 pilE form dimeric G-quadruplexes that exhibit distinct characteristics G-quadruplex (Figures 7A and 7B) involves end-on-end from that observed in the current study on the 50 end-stacked stacking of two monomeric all parallel-stranded three-layered dimeric pilE G-quadruplex. The earliest examples involved G-quadruplexes through their 50 end G-tetrads. The purine six- dimeric G-quadruplex formation including two-layered mono- membered rings overlap with each other, whereas the purine mers that are mediated by 50 end-stacked A,(G,G,G,G),A hex- five-membered rings overlap with amino groups of guanine ads (Figure S8A) (Kettani et al., 2000b) and heptads (Matsugami bases, thereby resulting in a very large stacking area at the inter- et al., 2001, 2003). Additional examples of 50 end-stacked face (Figure 7D). For this interface overlap geometry, sugar resi- G-tetrads involve formation of an interlocked 93 del dimeric dues are shifted by one step of helical twist, such as that G-quadruplex targeted to HIV integrase, where the interfacial observed in a side projection (Figure 7E), whereby the first G-tetrads involve one G(syn) residue from one monomer aligned G-tetrad tier residues G1/G1* become vertically aligned with antiparallel to three G(anti) residues from the other monomer residues G6*/G6 from the second G-tetrad tier of the (Figure S8D) (Phan et al., 2005b), as well as an interlocked V- symmetry-related monomer, with the same assignments also shaped dimeric G-quadruplex containing A,(G,G,G,G) pentads observed for the G5/G2* and G5*/G2 pairs. (Figure S8G) (Zhang et al., 2001). The two monomeric components are positioned so as to The observed features of the overlap geometry of the interfa- achieve the overall second-order rotational symmetry of the re- cial G-tetrads in the dimeric pilE G-quadruplex include stacking sulting dimeric pilE G-quadruplex. There are four types of between guanine six-membered rings, as well as between grooves associated with each monomeric G-quadruplex, with guanine amino groups and five-membered rings (Figure 7D). single-reside linkers spanning two of the grooves, a two-residue These overlap geometries are distinct from their counterparts

0 A B D Figure 10. A 3 pilE G4 Structure Promotes More Efﬁcient RecA-Mediated Strand No overhang 3´G4 Exchange (A) RecA strand exchange reaction schematic. A 50FAM-labeled 69-mer containing a 30 pilE G4 0 No overhang structure (3 G4) or the same 69-mer with a mutated pilE G4 sequence that can no longer form the structure (30G4mt) was combined with dsDNA of the same length and sequence with or without Overhang an overhang containing the pilE G4 sequence. (B) CD spectra of the 50FAM-labeled G4 (black) or G4 mutant sequence (red), and unlabeled dsDNA (green). (C) RecA strand exchange. Shown is a represen- C Overhang tative 12% polyacrylamide gel of RecA strand 3´G4mt exchange reactions performed with the 50FAM- labeled 30G4 or 30G4mt sequence. (D) The percent total of RecA-mediated strand exchange is plotted versus time. The 30G4 (black) and 30G4mt (red) combined with dsDNA without or with an overhang. See also Figure S7 and Table S5.

in dimeric G-quadruplexes involving interfacial A,(G,G,G,G),A canonic A,A base pair, with a resulting parallel orientation of hexads (Figures S8A and S8B), interlocked 93 del (Figures S8D incoming and outgoing DNA strands (Figure S9B). and S8E), and V-shaped (Figures S8G and S8H) topologies. For these three families of dimeric G-quadruplexes, the overlap RecA Is Targeted by the Monomeric pilE G-Quadruplex pattern between junctional G-tetrads involves stacking between The RecA family of ATPases catalyzes ATP-dependent strand guanine five-membered rings and between guanine amino exchange between ssDNA and heterologous duplex DNA (Rad- groups (Figures S8B, S8E, and S8H). ding, 1981). The enzyme has two binding sites: a primary site, In addition, sugar residues of the junctional G-tetrads do not which binds ssDNA and forms a RecA-ssDNA filament (presyn- stack over each other but rather are twisted by one helical aptic complex); and a secondary site, which binds a dsDNA step, such that sugar of G1 projects vertically on the sugar of target (site locations in Figure 9A). RecA binds ssDNA and G6 rather than G5, in the dimeric pilE G-quadruplex (Figures dsDNA with differing affinities in these two sites (Mazin and 7D and 7E). By contrast, sugar residues of junctional G-tetrads Kowalczykowski, 1998). In the postsynaptic complex, the either stack directly over each other as observed for the dimeric primary site is bound to dsDNA, which is the product of recom- G-quadruplex involving interfacial A,(G,G,G,G),A hexads bination (Figure 9A) (Chen et al., 2008). Thus, although three (Figures S8B and S8C) or are twisted to a lesser extent in the in- strands are involved in strand exchange, RecA has enough terlocked 93 del (Figures S8E and S8F) and V-shaped (Figures space to simultaneously accommodate two DNA duplexes in S8H and S8I) dimeric G-quadruplexes. the primary and secondary binding sites. Two critical reaction residues in the secondary site, Arg235 and Comparison of Dimeric pilE and c-kit2 G-Quadruplexes Lys237 (Kurumizaka et al., 1999), are separated from the primary For the dimeric pilE G-quadruplex elucidated in this work, two site by a distance exceeding 24 A˚ , which is more than the diam- monomeric units associate through 50 end stacking (Figure S9A). eter of a DNA double helix (Figure 9A) (Chen et al., 2008). Our fluo- Recently, we have demonstrated that the G-rich sequence from rescence anisotropy measurements establish that RecA binds intergenic region upstream of the c-kit oncogene promoter, a monomeric three-layered all parallel-stranded G-quadruplex termed c-kit2, forms a different and unique dimeric G-quadru- containing single-residue loops (Figure 9B), but not its counter- plex in solution. In the ckit-2 dimeric G-quadruplex structure, part containing three-residue loops (Figure S6B), nor does it two 50-guanine tracts and two 30-guanine tracts self-assemble bind to four-layered G-quadruplexes (Smith et al., 1995; Wang in register to form an all parallel-stranded dimeric G-quadruplex and Patel, 1995)(Figure S6C) or to the dimeric pilE G-quadruplex scaffold (Figure S9B). Each of these two dimeric G-quadruplexes (Figure S6D). Our studies have identified a higher-order DNA (Figures S9A and S9B) is composed of two three-layered substrate targeted to this important bacterial enzyme. subunits, but the pattern of their association exhibits distinct topologies. In the case of pilE G-quadruplex, two monomeric Potential Role for Monomeric pilE G-Quadruplex units self-associate through 50 end stacking resulting in an anti- in Recombination parallel orientation of outgoing DNA strands (Figure S9A), The absolute requirement of RecA for pilin antigenic variation whereas in the ckit-2 G-quadruplex, the two three-layered (Koomey et al., 1987) coupled with its reliance on the RecF-like monomeric units are covalently linked, and separated by a non- recombination pathway (Mehr and Seifert, 1998) has implied

that RecA could play a role in promoting strand exchange during Balasubramanian, S., Hurley, L.H., and Neidle, S. (2011). Targeting G-quadru- recombination (Stohl et al., 2011). However, the finding that plexes in gene promoters: a novel anticancer strategy? Nat. Rev. Drug Discov. RecA binds the monomeric pilE G-quadruplex suggests that 10, 261–275. RecA may play additional roles in promoting pilin antigenic Burge, S., Parkinson, G.N., Hazel, P., Todd, A.K., and Neidle, S. (2006). Quadruplex DNA: sequence, topology and structure. Nucleic Acids Res. 34, variation. 5402–5415. The G-quadruplex-forming pilE G4 DNA sequence is only Cahoon, L.A., and Seifert, H.S. (2009). An alternative DNA structure is necessary present at one location in the gonococcal genome, upstream for pilin antigenic variation in Neisseria gonorrhoeae. Science 325, 764–767. of pilE (Cahoon and Seifert, 2009). These data suggest that the Cate, J.H., Gooding, A.R., Podell, E., Zhou, K., Golden, B.L., Szewczak, A.A., monomeric pilE G-quadruplex could recruit RecA to the pilE Kundrot, C.E., Cech, T.R., and Doudna, J.A. (1996). RNA tertiary structure locus to promote recombination between pilE and a pilS copy. mediation by adenosine platforms. Science 273, 1696–1699. Such a model should take into account the role of RecF-like Chen, Z., Yang, H., and Pavletich, N.P. (2008). Mechanism of homologous pathway factors, RecQ, RecJ, RecO, and RecR, which are all recombination from the RecA-ssDNA/dsDNA structures. Nature 453, 489–494. involved in pilin antigenic variation, and would have to act before Criss, A.K., Kline, K.A., and Seifert, H.S. (2005). The frequency and rate of pilin or along with RecA (Mehr and Seifert, 1998; Sechman et al., antigenic variation in Neisseria gonorrhoeae. Mol. Microbiol. 58, 510–519. 2005; Skaar et al., 2002). Davis, J.T. (2004). G-quartets 40 years later: from 50-GMP to molecular biology We have outlined a proposed recombination mechanism and supramolecular chemistry. Angew. Chem. Int. Ed. Engl. 43, 668–698. based on a monomeric pilE G-quadruplex binding to RecA (Fig- Gue´ din, A., Gros, J., Alberti, P., and Mergny, J.L. (2010). How long is too long? ure S10) in the Supplemental Experimental Procedures. This Effects of loop size on G-quadruplex stability. Nucleic Acids Res. 38, 7858–7868. proposed mechanism needs further validation but has been Haber, J.E. (1998). Mating-type gene switching in Saccharomyces cerevisiae. put forward in the interest of stimulating further experiments. Annu. Rev. Genet. 32, 561–599. Hagblom, P., Segal, E., Billyard, E., and So, M. (1985). Intragenic recombination EXPERIMENTAL PROCEDURES leads to pilus antigenic variation in Neisseria gonorrhoeae. Nature 315, 156–158. Huber, M.D., Lee, D.C., and Maizels, N. (2002). G4 DNA unwinding by BLM and Protocols for sample preparation, NMR spectroscopy, gel electrophoresis, Sgs1p: substrate specificity and substrate-specific inhibition. Nucleic Acids circular dichroism spectroscopy, fluorescence anisotropy, structure calcula- Res. 30, 3954–3961. tions, modeling of monomeric pile G-quadruplex into RecA, experiments Kettani, A., Bouaziz, S., Wang, W., Jones, R.A., and Patel, D.J. (1997). Bombyx outlining RecA-mediated strand exchange, and S1 nuclease digestion are mori single repeat telomeric DNA sequence forms a G-quadruplex capped by presented in detail under Supplemental Experimental Procedures. base triads. Nat. Struct. Biol. 4, 382–389. Kettani, A., Basu, G., Gorin, A., Majumdar, A., Skripkin, E., and Patel, D.J. ACCESSION NUMBERS (2000a). A two-stranded template-based approach to G.(C-A) triad formation: designing novel structural elements into an existing DNA framework. J. Mol. The coordinates of the NG22 monomeric (accession code 2LXQ) and NG16 Biol. 301, 129–146. dimeric (accession code 2LXV) G-quadruplexes have been deposited in the Kettani, A., Gorin, A., Majumdar, A., Hermann, T., Skripkin, E., Zhao, H., Jones, Protein Data Bank. R., and Patel, D.J. (2000b). A dimeric DNA interface stabilized by stacked A.(G.G.G.G).A hexads and coordinated monovalent cations. J. Mol. Biol. SUPPLEMENTAL INFORMATION 297, 627–644. Koomey, M., Gotschlich, E.C., Robbins, K., Bergstro¨ m, S., and Swanson, J. Supplemental Information includes eleven figures, five tables, Supplemental (1987). Effects of recA mutations on pilus antigenic variation and phase transi- Experimental Procedures, and Supplemental Discussion and can be found tions in Neisseria gonorrhoeae. Genetics 117, 391–398. with this article online at http://dx.doi.org/10.1016/j.str.2012.09.013. Kurumizaka, H., Ikawa, S., Sarai, A., and Shibata, T. (1999). The mutant RecA proteins, RecAR243Q and RecAK245N, exhibit defective DNA binding in ACKNOWLEDGMENTS homologous pairing. Arch. Biochem. Biophys. 365, 83–91. Kuryavyi, V., and Patel, D.J. (2010). Solution structure of a unique G-quadru- V.K. was responsible for the NMR-based structure determinations of the plex scaffold adopted by a guanosine-rich human intronic sequence. monomeric and dimeric pilE G-quadruplexes and proposed the RecA-binding Structure 18, 73–82. hypothesis, under the supervision of D.J.P., whereas L.A.C. was responsible for the RecA-mediated binding and strand exchange experiments under the Kuryavyi, V., Kettani, A., Wang, W., Jones, R., and Patel, D.J. (2000). A supervision of H.S.S. The research was supported by NIH Grant GM34504 diamond-shaped zipper-like DNA architecture containing triads sandwiched to D.J.P. who is a member of the NY Structural Biology Center supported by between mismatches and tetrads. J. Mol. Biol. 295, 455–469. NIH Grant GM66354. This work was also supported by NIH Grants RO1 Kuryavyi, V., Phan, A.T., and Patel, D.J. (2010). Solution structures of all AI044239 and R37 AI033493 to H.S.S. In addition, L.A.C. was partially sup- parallel-stranded monomeric and dimeric G-quadruplex scaffolds of the ported by NIH Grant T32GM08061. human c-kit2 promoter. Nucleic Acids Res. 38, 6757–6773. Kuryavyi, V.V., and Jovin, T.M. (1995). Triad-DNA: a model for trinucleotide Received: July 18, 2011 repeats. Nat. Genet. 9, 339–341. Revised: September 7, 2012 Lipps, H.J., and Rhodes, D. (2009). G-quadruplex structures: in vivo evidence Accepted: September 24, 2012 and function. Trends Cell Biol. 19, 414–422. Published online: October 18, 2012 Liu, Z., and Gilbert, W. (1994). The yeast KEM1 gene encodes a nuclease specific for G4 tetraplex DNA: implication of in vivo functions for this novel REFERENCES DNA structure. Cell 77, 1083–1092. Ambrus, A., Chen, D., Dai, J., Jones, R.A., and Yang, D. (2005). Solution Maizels, N. (2006). Dynamic roles for G4 DNA in the biology of eukaryotic cells. structure of the biologically relevant G-quadruplex element in the human Nat. Struct. Mol. Biol. 13, 1055–1059. c-MYC promoter. Implications for G-quadruplex stabilization. Biochemistry Matsugami, A., Ouhashi, K., Kanagawa, M., Liu, H., Kanagawa, S., Uesugi, S., 44, 2048–2058. and Katahira, M. (2001). An intramolecular quadruplex of (GGA)(4) triplet

repeat DNA with a G:G:G:G tetrad and a G(:A):G(:A):G(:A):G heptad, and its Sen, D., and Gilbert, W. (1988). Formation of parallel four-stranded complexes dimeric interaction. J. Mol. Biol. 313, 255–269. by guanine-rich motifs in DNA and its implications for meiosis. Nature 334, Matsugami, A., Okuizumi, T., Uesugi, S., and Katahira, M. (2003). 364–366. Intramolecular higher order packing of parallel quadruplexes comprising Sissi, C., Gatto, B., and Palumbo, M. (2011). The evolving world of protein-G- a G:G:G:G tetrad and a G(:A):G(:A):G(:A):G heptad of GGA triplet repeat quadruplex recognition: a medicinal chemist’s perspective. Biochimie 93, DNA. J. Biol. Chem. 278, 28147–28153. 1219–1230. Mazin, A.V., and Kowalczykowski, S.C. (1998). The function of the secondary Skaar, E.P., Lazio, M.P., and Seifert, H.S. (2002). Roles of the recJ and recN DNA-binding site of RecA protein during DNA strand exchange. EMBO J. 17, genes in homologous recombination and DNA repair pathways of Neisseria 1161–1168. gonorrhoeae. J. Bacteriol. 184, 919–927. Smith, F.W., Schultze, P., and Feigon, J. (1995). Solution structures of unimo- Mehr, I.J., and Seifert, H.S. (1998). Differential roles of homologous recombi- lecular quadruplexes formed by oligonucleotides containing Oxytricha telo- nation pathways in Neisseria gonorrhoeae pilin antigenic variation, DNA trans- mere repeats. Structure 3, 997–1008. formation and DNA repair. Mol. Microbiol. 30, 697–710. Smith, J.S., Chen, Q., Yatsunyk, L.A., Nicoludis, J.M., Garcia, M.S., Kranaster, Neidle, S. (2009). The structures of quadruplex nucleic acids and their drug R., Balasubramanian, S., Monchaud, D., Teulade-Fichou, M.P., Abramowitz, complexes. Curr. Opin. Struct. Biol. 19, 239–250. L., et al. (2011). Rudimentary G-quadruplex-based telomere capping in Paeschke, K., Simonsson, T., Postberg, J., Rhodes, D., and Lipps, H.J. (2005). Saccharomyces cerevisiae. Nat. Struct. Mol. Biol. 18, 478–485. Telomere end-binding proteins control the formation of G-quadruplex DNA Sparling, P.F. (1966). Genetic transformation of Neisseria gonorrhoeae to structures in vivo. Nat. Struct. Mol. Biol. 12, 847–854. streptomycin resistance. J. Bacteriol. 92, 1364–1371. Parkinson, G.N., Lee, M.P., and Neidle, S. (2002). Crystal structure of parallel Stohl, E.A., Blount, L., and Seifert, H.S. (2002). Differential cross-complemen- quadruplexes from human telomeric DNA. Nature 417, 876–880. tation patterns of Escherichia coli and Neisseria gonorrhoeae RecA proteins. Patel, D.J., Kozlowski, S.A., Nordheim, A., and Rich, A. (1982). Right-handed Microbiology 148, 1821–1831. and left-handed DNA: studies of B- and Z-DNA by using proton nuclear Stohl, E.A., Gruenig, M.C., Cox, M.M., and Seifert, H.S. (2011). Purification and Overhauser effect and P NMR. Proc. Natl. Acad. Sci. USA 79, 1413–1417. characterization of the RecA protein from Neisseria gonorrhoeae. PLoS One 6, Patel, D.J., Phan, A.T., and Kuryavyi, V. (2007). Human telomere, oncogenic e17101. promoter and 50-UTR G-quadruplexes: diverse higher order DNA and RNA Wang, Y., and Patel, D.J. (1994). Solution structure of the Tetrahymena telo- targets for cancer therapeutics. Nucleic Acids Res. 35, 7429–7455. meric repeat d(T2G4)4 G-tetraplex. Structure 2, 1141–1156. Phan, A.T. (2000). Long-range imino proton-13C J-couplings and the through- Wang, Y., and Patel, D.J. (1995). Solution structure of the Oxytricha telomeric bond correlation of imino and non-exchangeable protons in unlabeled DNA. repeat d[G4(T4G4)3] G-tetraplex. J. Mol. Biol. 251, 76–94. J. Biomol. NMR 16, 175–178. Wolfgang, M., Lauer, P., Park, H.S., Brossay, L., He´ bert, J., and Koomey, M. Phan, A.T., and Patel, D.J. (2002). A site-specific low-enrichment (15)N,(13)C (1998). PilT mutations lead to simultaneous defects in competence for natural isotope-labeling approach to unambiguous NMR spectral assignments in nu- transformation and twitching motility in piliated Neisseria gonorrhoeae. Mol. cleic acids. J. Am. Chem. Soc. 124, 1160–1161. Microbiol. 29, 321–330. Wu, Y., Shin-ya, K., and Brosh, R.M., Jr. (2008). FANCJ helicase defective in Phan, A.T., Modi, Y.S., and Patel, D.J. (2004). Propeller-type parallel-stranded G- Fanconia anemia and breast cancer unwinds G-quadruplex DNA to defend quadruplexes in the human c-myc promoter. J. Am. Chem. Soc. 126, 8710–8716. genomic stability. Mol. Cell. Biol. 28, 4116–4128. Phan, A.T., Kuryavyi, V., Gaw, H.Y., and Patel, D.J. (2005a). Small-molecule Zhang, N., Gorin, A., Majumdar, A., Kettani, A., Chernichenko, N., Skripkin, E., interaction with a five-guanine-tract G-quadruplex structure from the human and Patel, D.J. (2001). V-shaped scaffold: a new architectural motif identified in MYC promoter. Nat. Chem. Biol. 1, 167–173. an A x (G x G x G x G) pentad-containing dimeric DNA quadruplex involving Phan, A.T., Kuryavyi, V., Ma, J.B., Faure, A., Andre´ ola, M.L., and Patel, D.J. stacked G(anti) x G(anti) x G(anti) x G(syn) tetrads. J. Mol. Biol. 311, 1063– (2005b). An interlocked dimeric parallel-stranded DNA quadruplex: a potent 1079. inhibitor of HIV-1 integrase. Proc. Natl. Acad. Sci. USA 102, 634–639. Zhang, Y., Gostissa, M., Hildebrand, D.G., Becker, M.S., Boboila, C., Chiarle, Phan, A.T., Kuryavyi, V., Burge, S., Neidle, S., and Patel, D.J. (2007a). R., Lewis, S., and Alt, F.W. (2010). The role of mechanistic factors in promoting Structure of an unprecedented G-quadruplex scaffold in the human c-kit chromosomal translocations found in lymphoid and other cancers. Adv. promoter. J. Am. Chem. Soc. 129, 4386–4392. Immunol. 106, 93–133. Phan, A.T., Kuryavyi, V., Luu, K.N., and Patel, D.J. (2007b). Structure of two intramolecular G-quadruplexes formed by natural human telomere sequences in Note Added in Proof K+ solution. Nucleic Acids Res. 35, 6517–6525. While our manuscript was under revision, a paper was published on the J19 G-quadruplex, a potential inhibitor of HIV-integrase (Do, N.Q., Lim, K.W., Teo, Qin, Y., and Hurley, L.H. (2008). Structures, folding patterns, and functions of M.H., Heddi, B. and Phan, A.T. (2011). Stacking of G-quadruplexes: NMR struc- intramolecular DNA G-quadruplexes found in eukaryotic promoter regions. ture of a G-rich oligonucleotide with potential anti-HIV and anticancer activity. Biochimie 90, 1149–1171. Nucleic Acids Res. 39, 9448–9457). Similar to the dimeric pilE quadruplex re- Radding, C.M. (1981). Recombination activities of E. coli recA protein. Cell 25, ported in this study, the J19 G-quadruplex also forms a three-layered stacked 3–4. dimeric G-quadruplex alignment (Figure S9C). Nevertheless, the two dimeric G- Ribeyre, C., Lopes, J., Boule´ , J.B., Piazza, A., Gue´ din, A., Zakian, V.A., quadruplexes show distinct stacking patterns involving interfacial G-tetrads. In 0 Mergny, J.L., and Nicolas, A. (2009). The yeast Pif1 helicase prevents genomic the pilE dimeric G-quadruplex reported in our study, the 5 ends of the consti- 0 instability caused by G-quadruplex-forming CEB1 sequences in vivo. PLoS tutive monomers are positioned on one side (Figure S9A), whereas the 5 Genet. 5, e1000475. ends in J19 dimeric G-quadruplex are positioned diagonally on opposite sides (Figure S9C). The top projections reveal the importance of the mass center for Rudel, T., van Putten, J.P., Gibbs, C.P., Haas, R., and Meyer, T.F. (1992). selection of theoretically possible rotamers in these two dimeric G-quadruplex Interaction of two variable proteins (PilE and PilC) required for pilus-mediated structures. The pilE dimeric G-quadruplex possesses four-fold pseudosymme- adherence of Neisseria gonorrhoeae to human epithelial cells. Mol. Microbiol. try so that the number of looped-out residues stays equal to two on each axis 6, 3439–3450. (Figure S9D). By contrast, the J19 dimeric G-quadruplex contains all-single- Sechman, E.V., Rohrer, M.S., and Seifert, H.S. (2005). A genetic screen iden- residue loops and is characterized by an eight-fold pseudosymmetry (Fig- tifies genes and sites involved in pilin antigenic variation in Neisseria gonor- ure S9E), so that looped-out elements (inclusive of single residues at 30 ends) rhoeae. Mol. Microbiol. 57, 468–483. are distributed evenly along the perimeter of the dimeric scaffold.