Crystal structure of Z-DNA in complex with the polyamine putrescine and potassium cations at ultra-high resolution

An ultra-high-resolution crystal structure of Z-DNA with the sequence d(CGCGCG)2 reveals the pattern of putrescinium(2+) and K+ binding. The complete polyamine dication is visible in the electron density and assumes a position that allows for hydrogen-bond interactions with three Z-DNA duplexes. The K+ cation occupies a mixed K+/H2O site and forms an irregular coordination sphere interacting with water molecules and the O atoms of two phosphate groups of two Z-DNA molecules, with some of the ligands present at fractional occupancy.


Introduction
Biogenic polyamine cations are essential for cell growth and differentiation, and their biochemical significance in a wide spectrum of physiological functions has been repeatedly reviewed (Gugliucci, 2004;Moinard et al., 2005;Larqué et al., 2007;Igarashi & Kashiwagi, 2010;Jastrzab et al., 2016). Common compounds in this group include the diamines 1,3diaminopropane(2+) (Dap 2+ ), putrescine(2+) (Put 2+ ) and cadaverine(2+) (Cad 2+ ), the triamines spermidine(3+) (Spd 3+ ) and norspermidine(3+), and the tetramine spermine(4+) (Spm 4+ ). Put 2+ and Cad 2+ are the two most common biogenic diamines in bacteria. For example, at a concentration of 10-30 mM, putrescine is the predominant polyamine in Escherichia coli (Cohen, 1997). Polyamines(n+) are involved in maintaining the cell-envelope integrity of bacteria. They are also crucial to the virulence phenotype of many bacterial pathogens (Shah & Swiatlo, 2008;Di Martino et al., 2013). Moreover, polyamines(n+) are important resources for viruses, which evolved mechanisms to maintain, enhance or manipulate polyamine(n+) metabolism to support viral infection (Firpo & Mounce, 2020). For instance, Mounce et al. (2017) showed the enhancement of viral polymerase activity ISSN 2052-5206 by polyamines(n+). Polyamines(n+) are important factors stabilizing nucleic acids and stimulating their replication; only $7-10% of the total cell content exists as free polyamines(n+), the vast majority remaining in association with nucleic acids (Gugliucci, 2004). It was also shown that both polyamines(n+) and metal polycations (even at low concentration) can stimulate B to Z conversion of DNA and promote DNA condensation (Ohishi et al., 2008;Vijayanathan et al., 2001). Generally, both polyamines(n+) and metal cations can interact with Z-DNA molecules in one of two different modes. Polyamine(n+) cations are found to bind to the O atoms of different phosphate groups either directly or through watermediated bridges. In the second mode, the cations are coordinated simultaneously by the guanine bases from different helices and also by the O atoms of dual-conformation phosphate groups (Tomita et al., 1989;Gao et al., 1993).
While the structure of the Put 2+ cation has been thoroughly studied (Woo et al., 1979;Bratek-Wiewiorowska et al., 1986;Jaskó lski, 1987;Pospieszna-Markiewicz et al., 2007), including the structures of phosphate salts (Jaskó lski et al., 1986;Bartoszak & Jaskó lski, 1990) and a study at helium temperature (Jaskó lski & Olovsson, 1989), it is rather poorly characterized as a component of nucleic acid structures. The Put 2+ dication was specified as a nucleic acid ligand only in nine structures deposited in the PDB, all of them corresponding to E. coli ribosome.
In this article, we describe a very-high-resolution crystal structure of Z-DNA with the sequence d(CGCGCG) 2 , which for the first time provides insight into the interaction of Z-DNA with Put 2+ and K + cations. We also discuss the questions of competition and effectiveness of polyamine and metal cations in interactions with Z-DNA.

Oligonucleotide synthesis, purification and crystallization
The methods of synthesis, deprotection and purification of the oligodeoxynucleotide have been described previously (Xia et al., 1998;Drozdzal et al., 2013). A 1.5 mM water solution of the DNA oligonucleotide with the self-complementary sequence d(CGCGCG) was heated at 338 K for 10 min and then annealed slowly to room temperature overnight. Single crystals of the d(CGCGCG) 2 /Put 2+ /K + complex were grown at 292 K by the hanging-drop vapour-diffusion method by mixing a 2 ml nucleic acid solution and a 2 ml precipitating solution consisting of 10%(v/v) (+/À)-2-methyl-2,4-pentanediol (MPD), 40 mM sodium cacodylate pH 6.0, 80 mM KCl, 12 mM NaCl and 14 mM putrescinium dichloride. The drops were equilibrated against 0.5 ml 80%(v/v) MPD. Crystals appeared within one week and grew to dimensions of 0.3 mm Â 0.1 mm Â 0.1 mm.

X-ray data collection and processing
X-ray diffraction data for the Z-DNA/Put 2+ /K + complex were collected to a resolution of 0.60 Å on the EMBL beamline P13 (Cianci et al., 2017) of the PETRA III storage ring at DESY, Hamburg. The crystal was vitrified in a stream of cold nitrogen gas at 100 K. The mother liquor served as the cryoprotectant solution. The diffraction data were collected in two passes using two wavelengths and the following crystal-todetector distances, oscillation ranges and numbers of images: 0.6880 and 0.7293 Å , 134 and 134 mm, 0.5 and 0.1 and 360 and 1800, respectively. The diffraction data were indexed, integrated and scaled using the XDS package (Kabsch, 2010). The X-ray data statistics are summarized in Table 1.  Notes: (a) values in parentheses correspond to the last resolution shell; (b) Bijvoet pairs separated; (c) correlation between intensities from random half-sets as defined by Karplus & Diederichs (2012) where F o and F c are the observed and calculated structure factors, respectively. R free was calculated analogously for the test reflections, which were randomly selected and excluded from the refinement.

Structure solution and refinement
The structure was solved by molecular replacement using PHASER (McCoy et al., 2007). The DNA part of the PDB structure 4hig, corresponding to our earlier model of the d(CGCGCG) 2 /Spm 4+ /Mn 2+ complex (Drozdzal et al., 2013), served as a molecular probe. At the initial stages of the refinement, the model was refined using REFMAC5  from the CCP4 program suite . The final anisotropic refinement was carried out with SHELXL (Sheldrick, 2015) using the full resolution of the diffraction data. The details of the SHELXL refinement were the same as described for our previous Z-DNA structures (Drozdzal et al., 2013(Drozdzal et al., , 2015, except for the use of the DISP instruction, which allows the definition of the dispersion and absorption coefficients of a particular element without having to use the full format of the SFAC instruction. The instruction OMIT was used to exclude ten of the most disagreeable reflections (error/e.s.d. > 10) from the refinement. The model was validated using the NuCheck program (Feng et al., 1998) and the free R test (Brü nger, 1992), with 1876 reflections selected at random and set aside for R free calculations.
It should be noted that the atomic scattering factor for the K + site was declared on the SFAC instruction by specifying a neutral K atom. This small inaccuracy means assigning one excess electron per 18 actual electrons. This should have very little, if any, effect on the refinement, possibly underestimating the refinable occupancy of the K + cation by a factor of 18/19.
At the wavelengths used in the diffraction experiments (0.6880, 0.7293 Å ), the imaginary component of the anom-alous scattering (f 00 ) of K and P atoms are, respectively, 0.236, 0.286 and 0.095, 0.104 electron units (Cromer, 1983). The anomalous signal is significant in the diffraction data set up to $0.9 Å resolution, as illustrated by clear peaks at the K + ion and P atoms in the anomalous electron-density map ( Fig. 1).
At this resolution, no stereochemical restraints are necessary to supplement the experimental observations (Jaskolski, 2017). However, restraints may still be needed for some disordered or highly mobile fragments. In the present structure, restraints were applied only to the putrescinium dication and to bonds and angles (33 for the sugar and eight for the phosphodiester moieties) of dual-conformation Z-DNA fragments. The ideal geometry targets for Put 2+ were taken from Pospieszna-Markiewicz et al. (2007). Conformation-dependent geometrical restraints on bond lengths (DFIX) and bond angles (DANG) for the polynucleotide chains were generated using the RestraintLib server (http://achesym.ibch.poznan.pl/ restraintlib/) as described by Kowiel et al. (2016Kowiel et al. ( , 2020 and Gilski et al. (2019). The CSD-derived conformation-dependent RestraintLib dictionary supersedes the classic nucleic acid restraints compiled by Parkinson et al. (1996). The final cycles of CGLS (conjugate-gradient least-squares) refinement converged with R/R free values of 8.88/9.50%. The very last round of refinement, calculated with the test reflections included in the working set, converged with R = 8.77%. In order to provide estimations of standard uncertainties in all individual refined parameters and of all derived geometrical parameters, at the final stage of the refinement, one cycle of full-matrix least-squares minimization was calculated. Model placement in the unit cell was standardized with the Achesym server (Kowiel et al., 2014). Stereoview of the d(CGCGCG) 2 /Put 2+ /K + complex with anomalous difference map (gray) peaks for the K + ion (purple sphere) and P atoms. The map is contoured at 3. Note the alternative conformations (I is green and II is orange) along the DNA chain.
The final model includes all the H atoms of the oligonucleotides and Put 2+ added as riding contributions to Fc. The ammonium -NH 3 + groups of Put 2+ were refined using the instruction AFIX 33. The rotatable O3 0 /5 0 -H groups of the terminal sugars were refined using the instruction AFIX 87.
Figures presenting the model and electron density were prepared with PyMOL (DeLano, 2002). The Coot (Emsley et al., 2010) program was used for visualization of the electrondensity maps and manual rebuilding of the atomic model. The program 3DNA (Lu & Olson, 2003) was used to calculate the Z-DNA helical parameters. The pseudorotation parameters were calculated by the method of Jaskó lski (1984) using the PseudoRotation server (http://www.cryst.ump.edu.pl/pseudorotation.php).

Data deposition in public repositories
Atomic coordinates and anisotropic ADPs, as well as the processed structure factors corresponding to the final model presented in this work, were deposited in the PDB with accession code 7atg. Raw X-ray diffraction images were deposited in the Integrated Resource for Reproducibility in Macromolecular Crystallography repository (proteindiffraction.org; Grabowski et al., 2016) with DOI https:// dx.doi.org/10.18430/m37atg.

Quality of the results
The estimated standard uncertainties (e.s.u.) of the fully occupied DNA atomic positions in the structure are in the range 0.003-0.01 Å for C atoms, 0.003-0.007 Å for N atoms, 0.003-0.009 Å for O atoms and 0.001-0.002 Å for P atoms. The e.s.u. values for the full-occupancy covalent bonds are $0.005, $0.004, $0.004 and $0.003 Å for C-C, C-O, C-N and P-O, respectively. The agreement with stereochemical standards (r.m.s.d.) is 0.010 Å for bond lengths and 1.56 for bond angles. These results are comparable in terms of accuracy and precision with the record-setting model of Z-DNA (0.55 Å , R = 7.77%, PDB ID 3p4j) described by Brzezinski et al. (2011).

Overall structure and helical parameters
The overall structural parameters of the DNA moieties of the d(CGCGCG) 2 /Put 2+ /K + complex classify them within the Z-DNA family. The average helical twist h (i.e. the angle of rotation about the helical axis that brings successive base pairs into coincidence) per CG/GC tandem of base pairs is À60 . Other average base-pair and base-pair-step parameters are as follows: helical rise 4.37 Å , inclination () 14.55 , tip 2.6 , tilt 0.44 , roll À3.32 , shift À0.02 Å , slide 2.77 Å and rise 3.45 Å . Comparison of these helical parameters and base-pair geometries for the present and previously described Z-DNA complexes shows that they are within the range typical for Z-DNA duplexes (see supplementary Table S3 in Drozdzal et al., 2015).
In the present complex, the sugars at the 3 0 -termini do not have the alternating C2 0 -endo/C3 0 -endo pucker of the pyrimidine/purine nucleotides, as is typical for Z-DNA, but all assume the C2 0 -endo conformation. The ZII conformation of the phosphate group can be assigned only to G4(I) with = 64.6 [ is a backbone torsion angle defined as: C3 0 -O3 0 -P(i + 1)-O5 0 (i + 1)]. This ZII conformation is stabilized by a water-mediated OP2(I)_G4Á Á ÁWat81Á Á ÁN2_G4 hydrogen bond.

Coordination of the polyamine cation
The entire putrescinium dication is clearly visible in the electron-density map (Fig. 2) despite its fractional occupancy, which converged on refinement at 0.378 (7). The putrescinium dication has a gauche À -trans-gauche + conformation, with the  The putrescinium(2+) dication in the crystal structure of the d(CGCGCG) 2 /Put 2+ /K + complex. The 2mFo-DFc map is contoured at the 1.6 level. Some water molecules have been omitted for clarity. Hydrogen bonds are marked by dashed lines, with DÁ Á ÁA distances in Å .
The electron-density maps indicate six water molecules in the vicinity of the Put 2+ dication as alternative species populating the polyamine(2+) site at complementary occupancy. It should be noted that there was a $20-fold molar excess of Put 2+ relative to the Z-DNA duplex in the crystallization mixture. The low occupancy of the Put 2+ site is, therefore, not the result of insufficient supply of the ligand, but rather reflects the natural equilibrium of components required for growing those high-quality (from the point of view of diffraction quality) crystals.
There is very little literature information on putrescine(2+) interactions with longer d(CG) n sequences. Putrescine had no effect on the conformation of a plasmid (pDHg16) with a 23base pair d(GC) 23 insert up to a 3 mM concentration (Thomas et al., 1991). Our results indicate that the putrescine(2+) cation has preference for interactions with Z-DNA bases, which may explain why a 3 mM putrescine concentration was not sufficient for the B-to Z-DNA transition in the above plasmid. For longer d(GC) n sequences in vitro, high concentrations of the putrescine(2+) cation may be needed to lower the energetic cost of Z-DNA formation.

Coordination of the K + cation
The electron-density maps clearly revealed one metal coordination site with an occupancy of 0.49 (3), interpreted as potassium. Due to the partial occupation of the potassium cation, a complementary water molecule (Wat202) was also modeled in the 2mFo-DFc map at this site with an occupancy of 0.34 (6). After the refinement of this model, the R factor was reduced from 8.82 to 8.77%. The refined distance between the K + and Wat202 sites is 0.195 (14) Å .
The metal was unambiguously identified as potassium using the following pieces of evidence. The length of the M-O bonds supports the presence of K + rather than, for example, the presence of Na + at higher occupancy. In the anomalous difference map, its peak (6.1) had a height similar to that of a full-occupancy P atom (P_11 at 6.4). The bond-valence method of Brese & O'Keeffe (1991), which correlates bond valences with the identity of the metal atom, is a popular method in coordination chemistry, and is especially reliable at high resolution. The application of this method gives values of the valence (V K ) and bond-valence (R KO ) parameters of 1.18, 1.09 and 2.07, 2.10 (the expected values for K + being V K = 1.00 and R KO = 2.13) for the coordination sphere including Wat110(I)/Wat110(II), respectively (vide infra). Also, the application of the CBVS method of Mü ller et al. (2003) confirmed the identity of the metal site as K + (calcium bondvalence sum CBVS = 0.53; according to the CBVS method, the reference value for K + is 0.64). Finally, the CheckMyMetal server (Zheng et al., 2014) also predicted potassium as the most likely cation at this site.
The K + cation is located between two Z-DNA phosphate groups. There are eight O atoms (four from Z-DNA backbone and four from water molecules) in the immediate vicinity (up The coordination sphere of the hydrated complex of K + (purple) at G6. The 2mFo-DFc map is contoured at 1.0. Coordination bonds are marked as gray dashed lines and hydrogen bonds are marked as orange dashed lines. Bond distances are in Å . Table 2 Coordination geometry (Å , ) around the metal ion in the d(CGCGCG) 2 /Put 2+ /K + structure, with standard uncertainties in parentheses. to a distance of 3.20 Å ) of the K + cation (Table 2). However, one of the water molecules (Wat110) and one phosphate group (OP1_G6) interacting with the potassium cation are disordered and were modelled in two complementary positions. In the presence of the OP1(I) and O5 0 (I) atoms (CN = 8), the coordination sphere can be considered as highly distorted square antiprismatic or dodecahedral. The K + ion is coordinated simultaneously by OP1(I), O5 0 (I) from G6 and OP1, OP2 from C9 iii [symmetry code: (iii) Àx + 1, y À 1 2 , Àz + 3 2 ], as well as by four water sites, one of which (Wat110) has dual occupancy ( Fig. 3 and Table 2). The K + -O bond distances are in the range 2.612 (22)-3.185 (11) Å . The angles within the coordination sphere are irregular (Table 2).

Hydration
The asymmetric unit contains 123 water sites. All water molecules were refined anisotropically without positional restraints. There was no attempt to model the H atoms of the water molecules. While for some water molecules it might be possible to try to locate their H atoms from difference Fourier maps, such H atoms have notoriously very poor geometry, even in small-molecule structures, and in macromolecular structures the dubious gain from their inclusion in the model usually does not compensate the burden of their individual handling in the refinement (Jaskolski, 2017). Summation of all the water occupancies in the asymmetric unit gives a total water content of 93.15. The positions of many disordered water molecules are correlated with the alternate I/II conformations of the Z-DNA backbones. With respect to their occupancy parameters, those water molecules for which the occupancies converged on refinement to values close to unity (>0.93) had their occupancy fixed at 1.0 (22 sites). Close pairs of sites for which the sum of their refined occupancies converged close to unity (36 pairs) had their combined occupancy constrained to 1.0. The remaining 65 sites had their occupancies refined freely to fractional values (all !0.20). The common patterns of solvent structure, as noted previously for Z-DNA crystals (Drozdzal et al., 2013(Drozdzal et al., , 2015, such as water molecules between N2_G and phosphate O atoms, the spine of hydration (Chevrier et al., 1986), two water molecules hydrogen bonded to each O6_G group or the absence of water molecules hydrogen bonded to the N3_G atoms, are also observed in the present Z-form structure.

Discussion
In this work, we have presented a new crystal structure of Z-DNA/polyamine(n+), in complex with putrescine(2+) and K + cations. It describes the first example of the interactions of putrescine(2+) with a DNA duplex. It is also the first case of Z-DNA crystallized in a complex with potassium ions. Moreover, the structure has the highest resolution and accuracy of the refined parameters among all DNA complexes with biogenic polyamines and/or metal cations deposited in the PDB.
Although the crystallization systems for all the Z-DNA complexes presented in our previous studies (Drozdzal et al., 2013(Drozdzal et al., , 2015 always contained KCl at a concentration of 40-80 mM in the crystallization drop, the K + cation has been identified in the electron-density maps only in the present Z-DNA structure in complex with putrescine(2+). Comparison of the Z-DNA/Spm 4+ /Mn 2+ and Z-DNA/Spm 4+ /Zn 2+ structures with the present Z-DNA/Put 2+ /K + complex in their common unit cell shows that the N1 + atoms of Spm 4+ coincide almost exactly with the site of the K + cation. This may indicate that the Spm 4+ cation has a higher affinity for Z-DNA binding than the K + cation and is preferentially selected when both are present in the crystallization buffer. It should also be stressed that the Z-DNA/K + interaction has not been described in the literature so far in any crystallographic studies of Z-DNA structures. It is interesting to note that the general location of the N1 + and N2 + atoms of the putrescinium(2+) dication in the unit cell of the Z-DNA/Put 2+ /K + complex is analogous to the positions of certain divalent metal cations, such as Mn 2+ (PDB ID 4hig) or Zn 2+ (1) and Zn 2+ (2) (4hif) (Drozdzal et al., 2013) (Fig. 4). Therefore, Put 2+ , having the same net charge as Mn 2+ or Zn 2+ , can effectively replace these cations in interactions with the Z-DNA duplex. It is also worth mentioning that putrescine(2+) has a completely different interaction pattern than 1,3-diaminopropane(2+) (Dap 2+ ) and 1,5-pentanediamine(2+) [cadaverine(2+)]. The structure of Z-DNA with Dap 2+ (PDB ID 2f8w) indicates that this one-carbon-linkshorter polyamine interacts only with the phosphate groups of the Z-DNA (Narayana et al., 2006). In a similar way, cadaverine(2+) (one carbon longer than Put 2+ ) exhibits a preference for binding with Z-DNA phosphates only (Drozdzal et al., to be published).
Our study provides insight into the effectiveness and competition of polyamine and metal cations for interactions with Z-DNA, and confirms that even a simple diamine can adopt different conformations and consequently enter into a variety of interactions with biomacromolecular partners.
The partial occupancy of the K + ion confirms previous research findings suggesting that most observed monovalent cation sites in DNA crystals are partially occupied positions (Tereshko et al., 2001;Dong, 2003). These cations are mobile and easily exchange sites with water molecules. This result also agrees both with (i) the conclusion from molecular-dynamic simulations which suggested that the structures of monovalent counter-ions in DNA are dynamic (Young et al., 1997) and with (ii) the crystallographic studies on which the hybrid solvent model for the solvent structure around DNA is based and in which the solvent sites are occupied by water-cation hybrids (Shui et al., 1998).