X-ray characterization of mesophases of human telomeric G-quadruplexes and other DNA analogues

Observed in the folds of guanine-rich oligonucleotides, non-canonical G-quadruplex structures are based on G-quartets formed by hydrogen bonding and cation-coordination of guanosines. In dilute 5′-guanosine monophosphate (GMP) solutions, G-quartets form by the self-assembly of four GMP nucleotides. We use x-ray diffraction to characterize the columnar liquid-crystalline mesophases in concentrated solutions of various model G-quadruplexes. We then probe the transitions between mesophases by varying the PEG solution osmotic pressure, thus mimicking in vivo molecular crowding conditions. Using the GMP-quadruplex, built by the stacking of G-quartets with no covalent linking between them, as the baseline, we report the liquid-crystalline phase behaviors of two other related G-quadruplexes: (i) the intramolecular parallel-stranded G-quadruplex formed by the 22-mer four-repeat human telomeric sequence AG3(TTAG3)3 and (ii) the intermolecular parallel-stranded G-quadruplex formed by the TG4T oligonucleotides. Finally, we compare the mesophases of the G-quadruplexes, under PEG-induced crowding conditions, with the corresponding mesophases of the canonical duplex and triplex DNA analogues.

bathing PEG solutions and combining it with the x-ray structural probe in an osmotic stress experiment 11 , we examine the liquid-crystalline phase behavior of the 22-mer HT-quadruplex in order to assess its likely state under similar conditions in the cellular environment.
The G-quartet is a planar structure formed by the cyclic arrangement of four hydrogen-bonded guanine residues. We first examine the liquid-crystalline phases of the G-quartets that are made of four GMP monomers 12 (Fig. 1c). The columnar structure built by the stacking of such G-quartets (abbreviated GMP-quadruplex) is a simple model of biologically relevant G-quadruplexes. Helical stacking of G-quartets (Fig. 1d) in the crystalline state was proposed based on fiber diffraction 13 . Formation of columnar structures was detected also in the non-crystalline GMP solutions in the presence of K + ions 14 . By varying the osmotic pressure of the solution set by the concentration of PEG, we investigate the first-order transition from a loosely linked and disordered G-quartet column formed in K + solution to the highly-ordered GMP-quadruplex driven by the osmotic pressure changes in the regime mimicking the biologically relevant molecular crowding conditions. This osmotic-pressure-induced change in order along the columnar axis consequently permits tighter packing of the GMP-quadruplex array between the columns, with progressively longer-ranged hexagonal order in the plane perpendicular to the columnar axis.
The columnar liquid-crystalline mesophases of planar disc-shaped structures are known and understood 15 . In these phases, disc-shaped structures are stacked on top of each other and form columns, which in turn self-assemble into arrays. As with other columnar assemblies 16 , the nature of intra-and intercolumnar ordering can vary, depending on solution conditions. The disordered columnar phases (Φ dc ) exhibit fluid-like positional intracolumnar order, while in the ordered columnar phases (Φ oc ) there is long-range positional order within each column. This leads to column-column positional and orientational correlations and consequently to long-range intercolumnar order 16,17 . In this respect, the mesophase transition of GMP-quadruplex is similar to the Φ dc → Φ oc transitions observed in the columnar aggregates of other disc-shaped structures built from the molecules with aromatic rings 15,17,18 .
Recent experiments demonstrated that the formation of higher-order G-quadruplex motifs in the human telomere is sensitive to the phase in the cell-cycle 9 , which, on the other hand, is related to macromolecular crowding, that can fine-tune the gene circuit response 19 . Thus, the Φ dc → Φ oc transition of the GMP-quadruplex induced by changing molecular-crowding conditions is relevant for the higher-order G-quadruplex structure formation built from the 22-mer HT-quadruplex repeats 5,8 . In this view, we also examined the columnar assemblies and the mesophase behaviors of (i) the 22-mer HT-quadruplex and (ii) the intermolecular parallel-stranded G-quadruplex formed by four TG 4 T oligonucleotides 20 (PDB:244D). The latter (abbreviated TG 4 T-quadruplex) is shown in Fig. 2a. This quadruplex was selected because of its resemblance to the 22-mer HT-quadruplex. In both structures, parallel sugar-phosphate backbones interconnect the stacked G-quartets. Contrary to the 22-mer HT-quadruplex, the TG 4 T-quadruplex is also missing the flexible side loops.
The Φ dc → Φ oc transition of the GMP-quadruplex is also analogous to the mesophase transitions in duplex (Fig. 2b) and triplex (Fig. 2c) DNA arrays under similar crowding conditions. For duplexes, this transition is described as the cholesteric-columnar hexagonal transition for short DNA 21 and as the cholesteric-hexatic transition for long DNA 22,23 . While the columnar hexagonal phase of the DNA triplexes has been observed before 24 , the triplex mesophase transition was not measured. The changes observed in the x-ray diffraction patterns at the mesophase transitions of the duplex, triplex, and G-quadruplex are essentially similar. At these transitions, both the intra-and the intercolumnar order change significantly and abruptly. We discuss the mesophase transitions of the analogue DNA structures with or without sugar-phosphate backbones and with different numbers of bases contributing to the stacking units (i.e., base-pair for duplex, base-triplet for triplex, and G-quartet for G-quadruplex). We then assess the relevance of these transitions for the observed mesophase behaviors.
Finally, 22-mer HT-quadruplex has a propeller-like shape 5 though not as pronounced as when 22-mer HT-quadruplex blocks are stacked in a column with TTA linkers between them 1,5 . In this particular model for the columnar stacking of G-quadruplex blocks (that are formed by four-repeat telomeric sequences), TTA segments connect the consecutive blocks that in turn would form an ordered G-quadruplex column with quasi-continuous helical characteristics. Here, we investigate the columnar mesophase behavior of 22-mer HT-quadruplex blocks when they are unlinked. These are similar to the unlinked very-short duplex DNA fragments 25 , except that the strong stacking interactions between the exposed hydrophobic cores of the duplex DNA fragments 25 are missing in the 22-mer HT-quadruplex case. If anything, because of the relatively weak end-to-end stacking interactions, the 22-mer HT-quadruplex blocks could be more prone to repel each other then to attract. This would force the columns of 22-mer HT-quadruplex blocks to be highly disordered.

Results
Under sufficient molecular crowding, as mimicked by the PEG-induced osmotic pressure (see Methods), duplex and quadruplex (GMP-quadruplex, TG 4 T-quadruplex, and 22-mer HT-quadruplex) DNA structures self-assemble into stable aggregates in the presence of K + ions without any other ions being added. These aggregates are transferred into PEG solutions of various concentrations for equilibration against known external osmotic pressures. PEG (molecular weight 8000 Daltons) is excluded from the DNA arrays during equilibration. Temperature-dependent osmotic pressures produced by the solutions of PEG at various concentrations are from ref. 11. Unless otherwise stated, all the measurements are at [KCl] = 0.3 M.
The osmotic pressures required for inducing DNA-analogue mesophase transitions strongly depend on solution ionic conditions. At [K + ] = 0.3 M, in the absence of any multivalent salts, the transition osmotic pressures for duplex and GMP-quadruplex are about the same, varying nearly from 6 to 8 atm (corresponding to from ≈19 to ≈22 wt% PEG 8000 concentration at 20 °C). Increasing K + concentration (under fixed external pressure) results in compression in the arrays of duplexes and G-quadruplexes. This observation can be explained as the screening of the electrostatic interactions between intercolumnar phosphate charges. In this section, we describe and present the experimental data systematically, by giving the priority to G-quadruplex DNA structures (GMP-quadruplex, TG 4 T-quadruplex, and 22-mer HT-quadruplex) and by using the duplex and triplex DNA data as additional information. The osmotic-pressure induced mesophase transitions for duplex DNA in the presence of monovalent NaCl has been observed recently 22 , while in this paper we additionally present the duplex DNA mesophase transition data in the presence of monovalent KCl. Furthermore, the triplex mesophase transition was not measured elsewhere. Although the main focus of this work is to present and discuss the mesophase behavior of the G-quadruplex structures, we also present duplex DNA data (in the presence of KCl) and the triplex mesophase transition that was measured the for the first time. We then compare the observed G-quadruplex DNA mesophases with duplex and triplex DNA mesophases. We used "long" wild-type DNA (~1 micron long) in the duplex DNA measurements. See below for details on the triplexes.

GMP-quadruplex.
Increasing order continuously with increasing osmotic compression in the disordered columnar mesophase (Φ dc ) is followed by a sudden collapse into the ordered columnar mesophase (Φ oc ) with remarkable changes in the intercolumnar distance (d int ) and the radial order ( Fig. 3b). At the Φ dc → Φ oc transition of the GMP-quadruplex, in particular, the change in the intercolumnar distance (Δd int ) is about 6.5-7 Å (Fig. 4a). This change in the packing density corresponds to about 0.3-0.4 nm 3 volume change per GMP nucleotide. It occurs concurrently with a significant lowering of the packing disorder. The radial disorder in the Φ dc phase, due to lateral displacements of loosely stacked G-quartets around the mean columnar axis, is evident in the Gaussian-shaped broad diffraction radial intensity peaks. By comparison, the Lorentzian peak shape in the Φ oc phase attests to the long-range nature of positional order. The correlation length in the ordered phase (equal to the inverse of the full-width-at-half-maximum of a Lorentzian function fitted to the x-ray diffraction radial intensity peak) is 5-to-6 neighbor separations for duplex DNA and 9-to-10 neighbor separations for GMP-quadruplex.
The slight temperature sensitivity of the Φ dc → Φ oc transition is shown for GMP-quadruplex (open circles in Fig. 4a). Temperature has no detectable effect on the packing density in the Φ oc phase. The transition osmotic In the course of the Φ dc → Φ oc transition, the shape of the diffraction peak changes dramatically while the peak center is shifted to a higher q r value. Two distinct types of peaks in the intensity profile coexist over a small range of Π, i.e., a phase-coexistence region. The procedure for x-ray diffraction peak fits is described in SI Appendix. Insets: X-ray images of the GMP-quadruplex arrays in the Φ dc (left) and Φ oc (right) phases. Higher order diffraction rings in the x-ray images of the Φ oc phase confirm hexagonal packing (see SI Appendix). (c) 1D intensity profiles from 22-mer HT-quadruplex (Fig. 1a, pressure (Π tr ) increases by about 1 atm upon increasing temperature from 20 to 40 °C. The effects of temperature on Π tr (as well as the effect of temperature on the DNA density at a fixed osmotic pressure Π) are appreciably smaller for other DNA structures and not shown. The transition for the GMP-quadruplex system is fairly sharp, and the integrated diffraction intensity (area under the diffraction peak) does not change through the transition. The integrated diffraction intensities from the same sample in the Φ dc and Φ oc phases are nearly the same (Fig. 3b). Phase-coexistence is pronounced over a significantly narrower range of osmotic pressures in the transitions of the duplex and GMP-quadruplex than in the transitions of the other structures. In particular, for GMP-quadruplex, the width of the phase-coexistence region is about 0.5 atm osmotic pressure. TG 4 T-quadruplex. When we followed the same procedure with the intermolecular TG 4 T-quadruplex, we observed first-order transitions similar to the Φ dc → Φ oc transition of the GMP-quadruplex. This suggests an abrupt change in the columnar organization of the short intermolecular TG 4 T-quadruplex, similar to the change in the columnar organization of the G-quartets at the Φ dc → Φ oc transition of the GMP-quadruplex. When this change in order is tuned by varying the osmotic pressure, we observe a much broader coexistence-region in the Φ dc → Φ oc transition for TG 4 T-quadruplex than for the GMP-quadruplex. For clarity, in Fig. 4a we show the intercolumnar spacings for the TG 4 T-quadruplex only at the upper and lower boundaries of the coexistence-region (inverted triangles). Intercolumnar spacings are about the same in the Φ oc phases of the TG 4 T-quadruplex and GMP-quadruplex arrays under the same pressure.

22-mer HT-quadruplex. With the aim of obtaining ordered columnar phases of the 22-mer HT-quadruplex
arrays (similar to the Φ oc phase of simpler model GMP-quadruplex), we increased the osmotic pressure compressing the disordered arrays of 22-mer HT-quadruplex and measured their x-ray diffraction patterns. The radial packing order increases with increasing osmotic compression. However, the very sharp changes observed in the radial intensity profiles at the first-order Φ dc → Φ oc transition of the GMP-quadruplex are not seen in the radial intensity profiles of 22-mer HT-quadruplex arrays (Fig. 3c).
When equilibrated under osmotic pressures less than the Φ dc → Φ oc transition osmotic pressure of the GMP-quadruplex, the measured positional disorder in 22-mer HT-quadruplex array is smaller than the disorder in GMP-quadruplex array at the same pressure (Fig. 5), which can be argued to be a consequence of the constrained lateral motion of the G-quartets, connected by the sugar-phosphate backbone. This connectivity increases the stability of 22-mer HT-quadruplex relative to the GMP-quadruplex and decreases its positional as well as stacking disorder. This is obviously true despite the TTA loops of the 22-mer HT-quadruplex that extend laterally from the core G-quartets 5 , and could be conceived as promoting and not suppressing the disorder of the columns. However, when equilibrated under osmotic pressures greater than that of the Φ dc → Φ oc transition of GMP-quadruplex, d int is 2-3 Å bigger for 22-mer HT-quadruplex than for GMP-quadruplex at the same pressure (Fig. 4a). Under these conditions, the radial disorder in the 22-mer HT-quadruplex array remains almost the same as in the disordered phase, while that of the GMP-quadruplex array drops significantly (Fig. 5, blue circles), possibly signaling the reverse action of laterally extended TTA loops in this case, amounting to a simple increase of the effective diameter of the columns.  (Fig. 4a). The changes were reversible for d int less than ≈38 Å. At larger spacings, once the Poly(AT*T)-triplexes disassociate, the triplex formation cannot be reestablished by simply increasing the osmotic pressure. Thus, the d int values for the Poly(AT*T)-triplex that we report in Table 1 are near the biggest distances where the transition can be observed at [K + ] = 0.3 M.

Discussion
Fiber diffraction data provide substantive evidence of helical stacking of the G-quartets 13 in GMP-quadruplex, as well as helical stacking of the base-pairs 26 in duplex DNA, at relative humidities corresponding to the osmotic pressures produced by the bathing PEG solutions that induced Φ oc phases of these structures. These helical details are expected to be pronounced only in the presence of strong correlations between repeating units (base-pair for duplex and G-quartet for GMP-quadruplex) along the columnar axes. Based on this, one can argue that the helical nature of the base-stacking in DNA structures plays a key role in the Φ dc → Φ oc transitions of the DNA arrays. Transitions occur when the intercolumnar spacings (d int ) are comparable to the helical pitch length (P), i.e., the axial distance per helical turn along the columnar axis (see Table 1). This implies that the formation of long-range translational and helical order along the columnar axis leads to long-range intercolumnar order in Figure 5. FWHM of the x-ray diffraction radial intensity peaks from the arrays of DNA structures: Symbols and structures are the same as in Fig. 4a,b. In the less-ordered phases (Φ dc ), packing disorder decreases with increasing osmotic pressure for all structures. Further compression leads to Φ dc → Φ oc transitions with discontinuous changes in the packing order in duplex and GMP-quadruplex arrays, which occur concurrently with discontinuous changes in the packing densities (shown in Fig. 4a,b). The latter is not observed in the 22mer HT-quadruplex arrays (discussed in the text). Phase-coexistence is shown only for the TG 4 T-quadruplex (shaded area). Phase-coexistence is observed over significantly narrower ranges of osmotic pressures for the duplex and GMP-quadruplex compared with the TG 4 T-quadruplex. The error in the determination of FWHM is as big as ≈0.1 nm −1 in the case of broad Gaussian peaks (in the disordered phases). The errors in FWHM are smaller for the sharp Lorentzian peaks in the ordered phases (see SI Appendix). the Φ oc phase. Additionally, the strength of the attraction, as seen in the volume change per nucleotide (ΔV pn ) in Table 1, increases with increasing number of strands in the structure. Counterintuitively, the attraction induced at the transition increases with increasing linear or surface charge density (λ and σ, respectively). This scenario has been elaborated 27,28 theoretically but is difficult to corroborate experimentally. We examined the Φ dc → Φ oc mesophase transitions of duplex, triplex, and quadruplex DNA structures. In particular, the spontaneous formations of highly-ordered G-quadruplex columns (GMP-quadruplex and TG 4 T-quadruplex) under biologically relevant molecular crowding conditions are significant for their analogies with the stacking organization of G-quadruplex structures in the human telomere. A quasi-continuous helix that runs along the columnar axis by the stacking of G-quartets within the TG 4 T-quadruplex columns 20 (similar to the helical stacking of the G-quartets in GMP-quadruplex) is possible with azimuthal rotations and arrangements of the TG 4 T-quadruplex blocks relative to the adjacent blocks.

a(nm) λ(e/nm) σ(e/nm 2 ) P(nm) Change in d int
The formation of uniaxially ordered columnar liquid crystals has been observed also in the case of the stacking of very-short-fragment (6 base-pairs long) DNA duplexes 25 . This stacking behavior was explained by the end-to-end adhesion of the exposed hydrophobic cores of the short DNA segments. From the observed Φ dc → Φ oc first-order mesophase transition of the TG 4 T-quadruplex arrays, as seen in the discontinuous change in the x-ray diffraction patterns tuned by varying molecular crowding conditions, it is possible to argue that the high-planarity of the G-quartets in the TG 4 T-quadruplex blocks makes the end-to-end stacking favorable. However, one could claim that the presence of the thymine bases at the ends of TG 4 T-quadruplex blocks would likely weaken the stacking interactions and in turn make the uniaxial helical ordering unfavorable. The broad range of osmotic pressures where the phase coexistence is observed in TG 4 T-quadruplex arrays (Fig. 5) might be attributed to the increased disorder due to the thymine bases at the ends.
In order to investigate the formation of ordered G-quadruplex columns in the telomere, the 22-mer HT-quadruplex was specifically chosen for two reasons: (i) the four-repeat sequence AG 3 (TTAG 3 ) 3 is the likely candidate for the G-quadruplex formation in vivo and (ii) G-quadruplex conformation of this sequence under molecular crowding conditions is known. However, to the best of our knowledge, there is no direct evidence that the 22-mer HT-quadruplex blocks would be organized into a columnar assembly in the non-crystalline state. Our osmotic stress experiments were designed to detect any sudden changes in the packing density and order, similar to the observed first-order mesophase transitions of the other model G-quadruplexes (GMP-quadruplex and TG 4 T-quadruplex). However, at all osmotic pressures 22-mer HT-quadruplex blocks (when they are unlinked) were observed to make only disordered columns, characterized by almost unchanged radial disorder (Fig. 5), implying also a pronounced stacking disorder. At large osmotic pressures it appears as if the disordered columns would have an effective diameter augmented by the contribution of the dangling TTA loops, while at small osmotic pressures the phosphate backbone connectivity of the 22-mer HT-quadruplex blocks enhances their stability. The robust disordered columnar assembly would possibly be the outcome of the attenuated stacking interactions between the G-quartet cores of the 22-mer HT-quadruplex blocks, compared with the very pronounced stacking interactions within the highly-ordered columnar phases of the GMP-quadruplex and the TG 4 T-quadruplex. This preserves the fluid-like order in the 22-mer HT-quadruplex columns at all crowding conditions. In fact, this could well be important in the in vivo context.
Another feature of the ordered phases of the DNA structures is revealed by comparison of the characteristic decay lengths in the Π vs. d int curves (Fig. 4a) of duplex, triplex, and GMP-quadruplex DNA. On theoretical grounds one can quantify the intercolumnar distance dependence of the osmotic pressure of the array in terms of a short-range hydration characteristic length and a longer-ranged Debye screening length 29 . This approach describes the hydration and the electrostatic components of the total interaction between the columnar structures in an ordered array. In the case of duplex DNA in monovalent salt solutions, at large separations in the ordered phase, the apparent decay length from the Π vs. d int curve is close to the Debye length, suggesting that the electrostatic interactions dominate at these separations 22 . When the surface-to-surface separation is smaller than about 7-8 Å, the Π vs. d int curves for all ionic concentrations converge to a single curve, suggesting also a universal short-range hydration repulsion that is independent of ionic strength (for details see ref. 22). Similarly, triplex DNA exhibits the same interaction regimes, with comparable Debye lengths at large interaxial separations, except that the hydration decay length in the high density regime is smaller than in the case of duplex DNA. In this respect the triplex DNA lies in between the behavior exhibited by the duplex DNA and the G-quadruplexes. In fact, in the case of G-quadruplexes, neither the Debye length nor the expected characteristic length for the short-range hydration interactions (on the order of the size of a water molecule) are as anomalously small as the apparent lengths reported here (~1 Å). The nature of these extremely short range interactions observed in G-quadruplex arrays thus remains to be elucidated.
As a side note, the radial disorder in the Φ dc phase is more pronounced in the GMP-quadruplex arrays than in the duplex, 22-mer HT-quadruplex, and TG 4 T-quadruplex arrays (Fig. 5). In the disordered phases, the repeating units along the columnar (or molecular) axes are constrained by (i) the relatively weak base-stacking interactions and (ii) the sugar-phosphate links between the adjacent units. The data in Fig. 5 point to the increased molecular stability by the sugar-phosphate backbone in duplex, 22-mer HT-quadruplex, and TG 4 T-quadruplex relative to the GMP-quadruplex. Nonetheless, the base-stacking interactions between the G-quartets are obviously strong enough and lead to the formation of disordered GMP-quadruplex columns at osmotic pressures as low as ≈3-4 atm (corresponding to ≈15 wt% PEG 8000 concentration).
Molecular crowding -as quantified by its proxy, the PEG solution osmotic pressure -certainly plays an important role in the stabilization of different DNA structures and regulation of their respective functions 10 . By explicitly showing and analyzing how the solution PEG osmotic pressure controls the DNA density, the related molecular order, and the phase transitions between differently ordered dense DNA-analogue arrays, we can take an additional step in understanding how the complicated in vivo environment regulates different functionalities of nucleic acids.
Scientific RepoRts | 6:27079 | DOI: 10.1038/srep27079 Methods Sample preparation. DNA oligonucleotide samples were ordered from Integrated DNA Technologies and were stored in a freezer until the time of measurements. Below we describe the methods of preparations of each DNA structure considered in this manuscript.
GMP-quadruplex. We prepared the GMP solutions (1 mg/ml) at [KCl] = 0.3 M with stirring at room temperature for about 2-3 hours. We then mixed 1 ml samples of the prepared GMP-solution with 4 ml 25 wt% PEG 8000 solutions (containing 0.3 M K + ), i.e., the final solution contains ~1 mg of GMP under 20 wt% PEG 8000 and 0.3 M K + . Under these conditions, GMP precipitates and pellets are formed by centrifugation. We transferred the collected pellets into new PEG 8000 solutions (at various wt% concentrations) for the x-ray diffraction experiments. TG 4 T-quadruplex. Quadruplex formation was induced by heating TG 4 T oligonucleotide solution at 80 °C for 5 min and then cooling to room temperature at [KCl] = 0.1 M. The oligonucleotide concentration in the annealing solution was ~0.1 mg/ml. TG 4 T-quadruplex is extensively studied in the literature. Their conformation in K + solutions is well-known 30 . We measured the CD spectra of the TG 4 T solutions before and after the heat incubation to confirm the parallel-stranded intermolecular TG 4 T-quadruplex formation (see SI Appendix).