An In-Depth Look at DNA Crystals through the Prism of Molecular Dynamics Simulations

Abstract

•Stable simulations of Drew-Dickerson dodecamer crystals (B-DNA) on the μs timescale•Analysis of all the space groups in which Drew-Dickerson dodecamer was resolved•Molecular description of the role of spermine in the stability of crystals•Detailed analysis of crystals’ biophysical properties in comparison with solution Almost 90% of biomolecular structures deposited in public databases to date have been determined by X-ray crystallography. The method owes its success partly to the fact that many materials can form crystals; however, it is difficult to know beforehand which specific conditions could facilitate their formation. Thus, optimal crystallization conditions are determined by gradually changing the buffer composition, temperature, and pressure in a trial-and-error manner. Considering how time consuming and costly such an approach is, the field would greatly benefit from knowing how external conditions determine the stability and specific symmetry of given biomolecular crystals, thereby enhancing the efficiency of the crystallization process. Computational methods—such as molecular dynamics simulations—show great potential in this respect, and by studying the behavior of DNA in crystals and the effects crystallization additives have on crystal stability, we lay the foundation for such studies. X-ray crystallography is the primary tool for biomolecular structural determination. However, contacts formed through the crystal lattice are known to affect structures, especially for small and flexible molecules such as DNA oligomers, by introducing significant structural changes in comparison to solution. Furthermore, why molecules crystallize in certain symmetry groups, which role crystallization additives play, and whether they are just innocuous and unspecific crystallization catalysts remain unclear. By using one of the currently best-performing DNA force fields and applying significant computational effort, we described the nature of intermolecular forces that stabilize B-DNA crystals in various symmetry groups and solvent environments with an unprecedented level of detail. We showed a tight coupling between the lattice stability and the type of crystallization additives and that certain symmetry groups are stable only in the presence of a specific additive. Additives and crystal contacts induce small but non-negligible changes in the physical properties of DNA. X-ray crystallography is the primary tool for biomolecular structural determination. However, contacts formed through the crystal lattice are known to affect structures, especially for small and flexible molecules such as DNA oligomers, by introducing significant structural changes in comparison to solution. Furthermore, why molecules crystallize in certain symmetry groups, which role crystallization additives play, and whether they are just innocuous and unspecific crystallization catalysts remain unclear. By using one of the currently best-performing DNA force fields and applying significant computational effort, we described the nature of intermolecular forces that stabilize B-DNA crystals in various symmetry groups and solvent environments with an unprecedented level of detail. We showed a tight coupling between the lattice stability and the type of crystallization additives and that certain symmetry groups are stable only in the presence of a specific additive. Additives and crystal contacts induce small but non-negligible changes in the physical properties of DNA. Ever since Linus Pauling laid the foundations of structural biology, X-ray crystallography has been the cornerstone method for solving biomolecular 3D structures. Over 130,000 models derived from X-ray data and deposited in the Protein Data Bank (PDB) (out of ∼147,000 models) are the best witnesses of the power of this technique, which has also become the gold standard for validating other structural methods. However, just as any other method, X-ray crystallography has its own shortcomings, and obtaining diffractable crystals is not always an easy task.1Zheng J. Birktoft J.J. Chen Y. Wang T. Sha R. Constantinou P.E. Ginell S.L. Mao C. Seeman N.C. From molecular to macroscopic via the rational design of a self-assembled 3D DNA crystal.Nature. 2009; 461: 74-77Crossref PubMed Scopus (711) Google Scholar, 2Zhang W. Szostak J.W. Huang Z. Nucleic acid crystallization and X-ray crystallography facilitated by single selenium atom.Front. Chem. Sci. Eng. 2016; 10: 196-202Crossref Scopus (17) Google Scholar, 3Saenger W. Methods: X-ray crystallography, potential energy calculations, and spectroscopy.in: Saenger W. Principles of Nucleic Acid Structure. Springer-Verlag, 1984: 29-50Crossref Google Scholar Through robotic equipment, numerous variants of crystallization buffers are tested until a suitable one is found.4Yeung H. Squire C.J. Yosaatmadja Y. Panjikar S. López G. Molina A. Baker E.N. Harris P.W.R. Brimble M.A. Radiation damage and racemic protein crystallography reveal the unique structure of the GASA/Snakin protein superfamily.Angew. Chem. Int. Ed. 2016; 55: 7930-7933Crossref PubMed Scopus (26) Google Scholar However, why a given buffer promotes crystallization, what its influence is on the symmetry of the unit cell, as well as the overall structure and physical properties of the biomolecule, is typically unknown. Thousands of X-ray-derived DNA structures have been crucial in understanding the fine details of isolated and protein-bound DNAs.5Mooers B.H.M. Crystallographic studies of DNA and RNA.Methods. 2009; 47: 168-176Crossref PubMed Scopus (26) Google Scholar, 6Westhof E. Perspectives and pitfalls.in: Ennifar E. Nucleic Acids Crystallography. Humana Press, 2016: 3-8Crossref Scopus (1) Google Scholar, 7Egli M. Nucleic acid crystallography: Current progress.Curr. Opin. Chem. Biol. 2004; 8: 580-591Crossref PubMed Scopus (55) Google Scholar Unfortunately, artifacts in DNA crystals can lead to conformations that are otherwise undetected in solution. Even in ideal cases, such as the Drew-Dickerson dodecamer (DDD),8Drew H.R. Wing R.M. Takano T. Broka C. Tanaka S. Itakura K. Dickerson R.E. Structure of a B-DNA dodecamer: Conformation and dynamics.Proc. Natl. Acad. Sci. USA. 1981; 78: 2179-2183Crossref PubMed Scopus (1187) Google Scholar, 9Drew H.R. Samson S. Dickerson R.E. Structure of a B-DNA dodecamer at 16 K.Proc. Natl. Acad. Sci. USA. 1982; 79: 4040-4044Crossref PubMed Scopus (134) Google Scholar where crystal structures resemble the solution ones reasonably well, it is unclear how crystallization conditions modulate the symmetry space in which the molecule crystallizes, the fine details of the structure, and the physical properties of the DNA.1Zheng J. Birktoft J.J. Chen Y. Wang T. Sha R. Constantinou P.E. Ginell S.L. Mao C. Seeman N.C. From molecular to macroscopic via the rational design of a self-assembled 3D DNA crystal.Nature. 2009; 461: 74-77Crossref PubMed Scopus (711) Google Scholar, 2Zhang W. Szostak J.W. Huang Z. Nucleic acid crystallization and X-ray crystallography facilitated by single selenium atom.Front. Chem. Sci. Eng. 2016; 10: 196-202Crossref Scopus (17) Google Scholar Knowing the molecular interactions that stabilize the crystals would help to understand the actual properties of DNA in such systems and the mechanisms through which highly charged molecules are packed in severely crowded conditions, e.g., the cellular nucleus. Here, we used a new generation of supercomputers and a recently developed DNA force field10Ivani I. Dans P.D. Noy A. Pérez A. Faustino I. Hospital A. Walther J. Andrio P. Goñi R. Balaceanu A. et al.Parmbsc1: A refined force field for DNA simulations.Nat. Methods. 2016; 13: 55-58Crossref PubMed Scopus (555) Google Scholar, 11Dans P.D. Danilāne L. Ivani I. Dršata T. Lankaš F. Hospital A. Walther J. Pujagut R.I. Battistini F. Gelpí J.L. et al.Long-timescale dynamics of the Drew-Dickerson dodecamer.Nucleic Acids Res. 2016; 44: 4052-4066Crossref PubMed Scopus (70) Google Scholar, 12Dans P.D. Ivani I. Hospital A. Portella G. González C. Orozco M. How accurate are accurate force-fields for B-DNA?.Nucleic Acids Res. 2017; 45: 4217-4230PubMed Google Scholar to analyze the nature of DDD in three known crystal lattices and to explore the effects of crystallization additives on the stability of the crystals and properties of the DNA.13Sines C.C. McFail-Isom L. Howerton S.B. VanDerveer D. Williams L.D. Cations mediate B-DNA conformational heterogeneity.J. Am. Chem. Soc. 2000; 122: 11048-11056Crossref Scopus (120) Google Scholar, 14Johansson E. Parkinson G. Neidle S. A new crystal form for the dodecamer C-G-C-G-A-A-T-T-C-G-C-G: Symmetry effects on sequence-dependent DNA structure.J. Mol. Biol. 2000; 300: 551-561Crossref PubMed Scopus (44) Google Scholar, 15Liu J. Subirana J.A. Structure of d(CGCGAATTCGCG) in the presence of Ca(2+) ions.J. Biol. Chem. 1999; 274: 24749-24752Crossref PubMed Scopus (43) Google Scholar Through extensive unbiased molecular dynamics (MD) simulations on the multi-microsecond timescale, we demonstrated how the stability of DNA crystals depends on subtle interactions between the packed DNA molecules and the components of the crystallization buffer. We obtained stable atomistic simulations of DNA crystals in biologically relevant timescales16York D.M. Yang W. Lee H. Darden T. Pedersen L.G. Toward the accurate modeling of DNA: The importance of long-range electrostatics.J. Am. Chem. Soc. 1995; 117: 5001-5002Crossref Scopus (172) Google Scholar, 17Liu C. Janowski P.A. Case D.A. All-atom crystal simulations of DNA and RNA duplexes.Biochim. Biophys. Acta. 2015; 1850: 1059-1071Crossref PubMed Scopus (16) Google Scholar, 18Babin V. Baucom J. Darden T.A. Sagui C. Molecular Dynamics simulations of polarizable DNA in crystal environment.Int. J. Quantum Chem. 2006; 106: 3260-3269Crossref Scopus (11) Google Scholar, 19Korolev N. Lyubartsev A.P. Nordenskiöld L. Laaksonen A. Spermine: An “invisible” component in the crystals of B-DNA. A grand canonical Monte Carlo and molecular dynamics simulation study.J. Mol. Biol. 2001; 308: 907-917Crossref PubMed Scopus (69) Google Scholar, 20Korolev N. Lyubartsev A.P. Laaksonen A. Nordenskiöld L. On the competition between water, sodium ions, and spermine in binding to DNA: A molecular dynamics computer simulation study.Biophys. J. 2002; 82: 2860-2875Abstract Full Text Full Text PDF PubMed Scopus (110) Google Scholar, 21Korolev N. Lyubartsev A.P. Laaksonen A. Nordenskiöld L. A molecular dynamics simulation study of oriented DNA with polyamine and sodium counterions: diffusion and averaged binding of water and cations.Nucleic Acids Res. 2003; 31: 5971-5981Crossref PubMed Scopus (74) Google Scholar in various symmetry groups (in which DDD has been crystallized) and in different solvent environments, which allowed us to understand with a high level of detail the nature of intermolecular interactions that stabilize the crystals. Through the analysis of numerous simulations, we characterized the physical properties of DNA in a highly crowded environment and how the presence of additives affects them. The Drew-Dickerson dodecamer is a prototypical B-DNA molecule whose structure has been determined by X-ray crystallography in various space groups and under diverse conditions13Sines C.C. McFail-Isom L. Howerton S.B. VanDerveer D. Williams L.D. Cations mediate B-DNA conformational heterogeneity.J. Am. Chem. Soc. 2000; 122: 11048-11056Crossref Scopus (120) Google Scholar, 14Johansson E. Parkinson G. Neidle S. A new crystal form for the dodecamer C-G-C-G-A-A-T-T-C-G-C-G: Symmetry effects on sequence-dependent DNA structure.J. Mol. Biol. 2000; 300: 551-561Crossref PubMed Scopus (44) Google Scholar, 15Liu J. Subirana J.A. Structure of d(CGCGAATTCGCG) in the presence of Ca(2+) ions.J. Biol. Chem. 1999; 274: 24749-24752Crossref PubMed Scopus (43) Google Scholar as well as through high-quality NMR data in solution. We chose to perform our simulations on structures crystallized in three different space groups: P3212, P212121, and H3 (PDB: 1EHV, 1FQ2, and 463D, respectively). We followed the multistep protocol developed by the Case group,17Liu C. Janowski P.A. Case D.A. All-atom crystal simulations of DNA and RNA duplexes.Biochim. Biophys. Acta. 2015; 1850: 1059-1071Crossref PubMed Scopus (16) Google Scholar which includes a careful stabilization of the internal pressure by calibrating the number of water molecules in the system (see the Supplemental Experimental Procedures for details; Table S2). Once the internal pressure was adjusted, we ran microsecond long simulations of each crystal, which contained 27, 24, and 36 copies of DDD (Table 1). To our surprise, two of our systems showed a systematic degradation of the crystal lattice, as shown by the center-of-mass (COM) displacements of each duplex with respect to the ideal positions in the crystal (Figure 1). The distortion of the lattice was particularly pronounced along the z axis and got worse with simulation time for 1EHV and 1FQ2, whereas the 463D system was completely stable through the trajectory. Not only was the lattice intact in 463D, but the MD conformations sampled in the helical space on the base-pair step level were in complete agreement with the average X-ray values (Figure 2), thus establishing that our force field is fully capable of representing local and global features of crystal structures.Table 1Studied Systems and Simulated ConditionsSystem NamePDBEnvironmentSpace GroupNo. of Unit CellsNo. of DNA DuplexesSalt Type[Salt] (mM)Ratio of dsDNA to SpmTemperature (K)Equilibration Time (ps)Simulation Time (μs)1EHV sol1EHVsolution––1Na/KCl400–284600115 mM1EHVcrystalP32123 × 3 × 381KCl15–2846001200 mM1EHVcrystalP32123 × 3 × 381Na/KCl200–2846001400 mM1EHVcrystalP32123 × 3 × 381Na/KCl400–2846004600 mM1EHVcrystalP32123 × 3 × 381Na/KCl600–2846001800 mM1EHVcrystalP32123 × 3 × 381Na/KCl800–28460011EHV Na/KCl1EHVcrystalP32123 × 3 × 127Na/KCl400–28460041EHV Mg1EHVcrystalP32123 × 3 × 127MgCl2400–28460041EHV Mg hex1EHVcrystalP32123 × 3 × 127MgCl2∙6H2O400–28460041EHFexperiemntalV 1:61EHVcrystalP32123 × 3 × 127SpmCl4–1:62846,00011EHV 1:91EHVcrystalP32123 × 3 × 127SpmCl4–1:92846,00011EHV 1:121EHVcrystalP32123 × 3 × 127SpmCl4–1:122846,00011FQ2 sol1FQ2solution––1Na/KCl400–29560011FQ2 1:6 sol1FQ2solution––1SpmCl4–1:629560011FQ2 Na/KCl1FQ2crystalP2121213 × 2 × 124Na/KCl400–29560041FQ2 Na/KCl equ1FQ2crystalP2121213 × 2 × 124Na/KCl400–2956,00011FQ2 1:31FQ2crystalP2121213 × 2 × 124SpmCl4–1:32956,00011FQ2 1:61FQ2crystalP2121213 × 2 × 124SpmCl4–1:62956,00011FQ2 1:91FQ2crystalP2121213 × 2 × 124SpmCl4–1:92956,0001463D sol463Dsolution––1Na/KCl400–2936001463D463DcrystalH32 × 2 × 136Na/KCl400–2936,0004 Open table in a new tab Figure 2Rotational Helical Parameters for the 463D SystemShow full captionRotational helical parameters (roll, twist, and tilt) of all the duplexes in the 463D system (given for distinct base-pair steps in DDD after removing the ends). The values calculated from the X-ray structure are shown with purple lines. Points are colored according to the smooth 2D densities, estimated by fitting the observed distributions to a bivariate normal kernel (evaluated on a square grid of 90 × 90 bins). Four isodensity curves are shown in white and are quantified on the bottom right side of each plot.View Large Image Figure ViewerDownload Hi-res image Download (PPT) Rotational helical parameters (roll, twist, and tilt) of all the duplexes in the 463D system (given for distinct base-pair steps in DDD after removing the ends). The values calculated from the X-ray structure are shown with purple lines. Points are colored according to the smooth 2D densities, estimated by fitting the observed distributions to a bivariate normal kernel (evaluated on a square grid of 90 × 90 bins). Four isodensity curves are shown in white and are quantified on the bottom right side of each plot. We thus centered our efforts on stabilizing the other two crystals by starting first with the smallest system: 1EHV. We simulated a larger crystal (81 instead of 27 DDD molecules) with five different Na/KCl concentrations (Table 1), ranging from 15 to 800 mM (i.e., up to five times the assumed physiological salt concentration of the cellular nucleus), without being able to stabilize it (Figure S1). Considering that the increase in crystal’s size had little effect on its stability (Figure S2), we decided to proceed with the smaller system (27 DDD copies). The addition of 400 mM MgCl2 produced some improvement in the lattice integrity, but its degradation was still evident during the 4-μs-long simulations (Figure S3 and Supplemental Experimental Procedures for more details). We thus checked whether the global degradation of the crystals was due to internal distortions of the double helices by analyzing the 2D distributions of the base-pair-step helical parameters from MD simulations. We obtained a good agreement between the values calculated from MD simulations and the X-ray structures (Figure S4), which discards the notion that the lattice distortions were caused by the deterioration of duplex geometries. Even the flexible ends (residue pairs 1–24 and 12–13), which exhibit fast but moderate fraying in solution,11Dans P.D. Danilāne L. Ivani I. Dršata T. Lankaš F. Hospital A. Walther J. Pujagut R.I. Battistini F. Gelpí J.L. et al.Long-timescale dynamics of the Drew-Dickerson dodecamer.Nucleic Acids Res. 2016; 44: 4052-4066Crossref PubMed Scopus (70) Google Scholar had in most cases an average RMSD with respect to the X-ray structures under 4 Å, oscillating between 3D conformations compatible with the experimental electron density (Figure S5). We obtained similar results when trying to stabilize, without success, the P212121 1FQ2 crystal (data not shown), which allowed us to discard the possibility of a force field artifact that is space-group specific. Given that we had discarded any obvious explanation for the degradation of the crystal lattice, we focused our attention on the chemical additives used to obtain the DDD crystals in the studied space groups. We noticed that the 463D system was crystallized without spermine (SPM),15Liu J. Subirana J.A. Structure of d(CGCGAATTCGCG) in the presence of Ca(2+) ions.J. Biol. Chem. 1999; 274: 24749-24752Crossref PubMed Scopus (43) Google Scholar whereas both 1EHV14Johansson E. Parkinson G. Neidle S. A new crystal form for the dodecamer C-G-C-G-A-A-T-T-C-G-C-G: Symmetry effects on sequence-dependent DNA structure.J. Mol. Biol. 2000; 300: 551-561Crossref PubMed Scopus (44) Google Scholar and 1FQ213Sines C.C. McFail-Isom L. Howerton S.B. VanDerveer D. Williams L.D. Cations mediate B-DNA conformational heterogeneity.J. Am. Chem. Soc. 2000; 122: 11048-11056Crossref Scopus (120) Google Scholar crystals were formed in the presence of this polyamine (+4 charge at pH 7). SPM is normally found in all eukaryotic cells and has been used as an “inert molecular glue” for obtaining thousands of DNA crystals in all the major forms (A, B, Z, etc.).19Korolev N. Lyubartsev A.P. Nordenskiöld L. Laaksonen A. Spermine: An “invisible” component in the crystals of B-DNA. A grand canonical Monte Carlo and molecular dynamics simulation study.J. Mol. Biol. 2001; 308: 907-917Crossref PubMed Scopus (69) Google Scholar, 20Korolev N. Lyubartsev A.P. Laaksonen A. Nordenskiöld L. On the competition between water, sodium ions, and spermine in binding to DNA: A molecular dynamics computer simulation study.Biophys. J. 2002; 82: 2860-2875Abstract Full Text Full Text PDF PubMed Scopus (110) Google Scholar, 22Katz A.M. Tolokh I.S. Pabit S.A. Baker N. Onufriev A.V. Pollack L. Spermine condenses DNA, but not RNA duplexes.Biophys. J. 2017; 112: 22-30Abstract Full Text Full Text PDF PubMed Scopus (36) Google Scholar However, the molecular basis of its mechanism of action is mostly unknown, and in fact, SPM electron densities are often absent from the crystal. To check whether the lack of SPM in our simulations was the reason of the degradation of P3212 and P212121 simulations, we repeated the simulations for both crystals at three different SPM concentrations ranging from 1:3 (three SPM molecules per duplex) up to 1:12 (Table 1). At last, the lattice integrity for both space groups was now preserved in a consistent concentration-dependent manner (Figure 3). At the medium SPM concentration tested, 1:9 for 1EHV and 1:6 for 1FQ2, the observed stability is already comparable to the one obtained for 463D with no perceptible drift of the crystal lattice along the y or z axes (Figures S6 and S7). Clearly the “inert molecular glue” has a major role in preserving the integrity of some of the crystal lattices. Similarly to 463D, once the crystals were properly stabilized, an impressive agreement between the X-ray structures (1EHV and 1FQ2) and the MD crystals was obtained. This is visible when comparing the rotational helical space (Figure 4), or the groove dimensions, a property that has always been difficult to reproduce accurately by modern force fields (Figure S8). The dynamics of the end residues also seem to be stable according to their RMSD values, which are typically below 2 Å with respect to the crystal position (Figures S9 and S10). In the end, to illustrate the observed structural agreement, we also compared the average structures obtained from crystal MD simulations with the deposited X-ray structure for the three systems stabilized in the μs regime, as shown in Figure S11. To follow each ion along the trajectories of each duplex, we used the curvilinear helicoidal coordinate (CHC)23Lavery R. Maddocks J.H. Pasi M. Zakrzewska K. Analyzing ion distributions around DNA.Nucleic Acids Res. 2014; 42: 8138-8149Crossref PubMed Scopus (78) Google Scholar method as implemented in the Canion module of Curves+, which allowed us to calculate ion populations and concentrations in the inner and outer areas of major and minor grooves of DNA duplexes (see Supplemental Experimental Procedures).24Dans P.D. Faustino I. Battistini F. Zakrzewska K. Lavery R. Orozco M. Unraveling the sequence-dependent polymorphic behavior of d(CpG) steps in B-DNA.Nucleic Acids Res. 2014; 42: 11304-11320Crossref PubMed Scopus (60) Google Scholar, 25Pasi M. Maddocks J.H. Lavery R. Analyzing ion distributions around DNA: sequence-dependence of potassium ion distributions from microsecond molecular dynamics.Nucleic Acids Res. 2015; 43: 2412-2423Crossref PubMed Scopus (75) Google Scholar We found that the sequence-dependent binding sites observed for K+ and the amino groups of SPM were the same in solution and in crystals, although the sequence dependence was weaker in crystals and the cation distributions seemed to be fuzzier (Figure S12).11Dans P.D. Danilāne L. Ivani I. Dršata T. Lankaš F. Hospital A. Walther J. Pujagut R.I. Battistini F. Gelpí J.L. et al.Long-timescale dynamics of the Drew-Dickerson dodecamer.Nucleic Acids Res. 2016; 44: 4052-4066Crossref PubMed Scopus (70) Google Scholar The comparison of aqueous simulations with and without SPM (left column in Figure S12) showed that SPM mainly localizes outside the grooves, which was also previously observed in crystal simulations.19Korolev N. Lyubartsev A.P. Nordenskiöld L. Laaksonen A. Spermine: An “invisible” component in the crystals of B-DNA. A grand canonical Monte Carlo and molecular dynamics simulation study.J. Mol. Biol. 2001; 308: 907-917Crossref PubMed Scopus (69) Google Scholar, 20Korolev N. Lyubartsev A.P. Laaksonen A. Nordenskiöld L. On the competition between water, sodium ions, and spermine in binding to DNA: A molecular dynamics computer simulation study.Biophys. J. 2002; 82: 2860-2875Abstract Full Text Full Text PDF PubMed Scopus (110) Google Scholar, 22Katz A.M. Tolokh I.S. Pabit S.A. Baker N. Onufriev A.V. Pollack L. Spermine condenses DNA, but not RNA duplexes.Biophys. J. 2017; 112: 22-30Abstract Full Text Full Text PDF PubMed Scopus (36) Google Scholar, 26Yoo J. Aksimentiev A. The structure and intermolecular forces of DNA condensates.Nucleic Acids Res. 2016; 44: 2036-2046Crossref PubMed Scopus (51) Google Scholar Interestingly, the interiors of the grooves are depleted of cations in all crystal simulations compared to the solution phase. We also analyzed the cation environment in all the individual duplexes present in each crystal. First, a clear similarity was found in the cation distribution in all the duplexes of a given crystal, confirming the robustness of the average results discussed here (see Figures S13–S15). Second, SPM distribution was similar around the duplexes in simulations of two different space groups (PDB: 1EHV and 1FQ2), indicating that the ion distribution is independent of the crystal packing (see Figures S14 and S15). Moreover, SPM mostly preferred the exterior of the central part of the major groove; however, upon entering the groove, SPM was located at non-central sites, which we also observed for solution simulations (see Figure S12). SPM could concentrate up to 12 M around the phosphate groups in the central portion of the duplexes, where no DNA intermolecular contacts are present. In contrast, these contacts are particularly pronounced in terminal bases (Figure S16). In summary, SPM has a tendency to move around the exterior of the duplex without any perceptible sequence dependence (as observed in NMR27Wemmer D.E. Srivenugopal K.S. Reid B.R. Morris D.R. Nuclear magnetic resonance studies of polyamine binding to a defined DNA sequence.J. Mol. Biol. 1985; 185: 457-459Crossref PubMed Scopus (96) Google Scholar and RAMAN28Deng H. Bloomfield V.A. Benevides J.M. Thomas Jr., G.J. Structural basis of polyamine-DNA recognition: spermidine and spermine interactions with genomic B-DNAs of different GC content probed by Raman spectroscopy.Nucleic Acids Res. 2000; 28: 3379-3385Crossref PubMed Scopus (178) Google Scholar experiments), most likely screening intermolecular phosphate-phosphate repulsion and facilitating crystal packing. Combined with its rare long-term presence within the insides of the grooves, it is not surprising that SPM electron density is seldom captured. Although the temperature for both experiments and MD simulations was in the 284–293 K range for 1EHV, 1FQ2, and 463D (see Supplemental Experimental Procedures), the intermolecular DNA contacts in the crystal froze the duplex structure. DNA’s internal effective temperature was lowered by 200 K (with respect to the thermostat temperature) at the end of the duplexes, which are mostly rigidified, whereas we observed only a ∼50 K decrease in the central portion of the duplex (Figure S17). Water molecules and SPM are also affected by the crowded and highly negatively charged environment produced by the packing of the backbone phosphate groups of neighboring duplexes. The rescaled diffusion coefficient29Hospital A. Candotti M. Gelpí J.L. Orozco M. The multiple roles of waters in protein solvation.J. Phys. Chem. B. 2017; 121: 3636-3643Crossref PubMed Scopus (14) Google Scholar of water molecules (Figure S18 and Supplemental Experimental Procedures) was reduced by one order of magnitude—from 1.70 ⋅ 10−5 cm2 s−1 in solution DNA simulations to 1.67 ⋅ 10−6 cm2 s−1 in crystal simulations. The reduction in water mobility is visible from a 30 K decrease in its effective temperature. Furthermore, water mobility in crystals is anisotropic, with the preference for the z axis in 463D and 1EHV and the y axis in 1FQ2, following the water channels that are created in each crystal due to the orientation of the duplexes (Figure S19). The effect was even more pronounced for SPM when comparing the 1FQ2 1:6 solution simulations with the 1FQ2 crystal with the same SPM concentration: the diffusion coefficient in the crystal was reduced by 3 orders of magnitude in comparison with the solution: from 1.90 ⋅ 10−6 cm2 s−1 to 2.62 ⋅ 10−9 cm2 s−1. Although the mobility of SPM is greatly reduced in the crystal, the lack of specificity in its binding to DNA (extensive and non-specific binding outside the major groove) makes its detection in the experimental electron densities very difficult (see Figure S12). As expected, the diffusion of water and SPM in the crystal was correlated and exhibited an SPM-concentration-dependent effect, as shown in Figure S20. We computed the global RMSD of MD simulations (crystal and solution) with respect to the experimental crystal configuration (Table 2). Our calculation expectedly showed that the average MD structure in solution deviates much more from the X-ray structure (with larger deviations in the backbone) than the average conformation from the crystal MD simulations.17Liu C. Janowski P.A. Case D.A. All-atom crystal simulations of DNA and RNA duplexes.Biochim. Biophys. Acta. 2015; 1850: 1059-1071Crossref PubMed Scopus (16) Google Scholar However, for all three systems, we obtained an excellent agreement for the internal 10 base pairs, with the RMSD in all cases under 0.9 Å (even 0.52 Å for 1FQ2), ratifying that MD simulations are able to capture the effect of the crystal environment on the DNA. Irrespectively of the original conformation present in the crystal, when we transferred the duplexes to solution, all the simulations converged to the same ensemble that was previously described by multi-microsecond simulations of DDD,11Dans P.D. Danilāne L. Ivani I. Dršata T. Lankaš F. Hospital A. Walther J. Pujagut R.I. Battistini F. Gelpí J.L. et al.Long-timescale dynamics of the Drew-Dickerson dodecamer.Nucleic Acids Res. 2016; 44: 4052-4066Crossref PubMed Scopus (70) Google Scholar, 12Dans P.D. Ivani I. Hospital A. Portella G. González C. Orozco M. How accurate are accurate force-fields for B-DNA?.Nucleic Acids Res. 2017; 45: 4217-4230PubMed Googl