Iodo-CCGG: A Single Crystal Structure of A-DNA Citation Conner, Benjamin Norman (1982) Iodo-CCGG: A Single Crystal Structure of A-DNA. Dissertation (Ph.D.), California Institute of Technology. doi:10.7907/3t68-dc11. https://resolver.caltech.edu/CaltechTHESIS:05142018-150830492 Abstract Iodo-CCGG was synthesized by the triester method and crystallized in space group P4 3 2 1 2. Electron density maps were phased using the anomalous scattering from the iodines and refined with the Jack-Levitt constrained refinement program. The final crystallographic R factor was 16.5% for 2 σ data and 19.9% for all data. Solution of the structure revealed the first fragment of A-DNA to be seen in a single crystal. The helix has a 2.2Å rise per residue, 10.6 base pairs per turn, and an average propeller twist of 17°. A total of 86 water molecules were located in the crystal. Of these, 47 were within 3.5Å of the I CCGG molecule. A second layer of hydration is formed by water molecules which are within 3.5Å of two or more first layer water molecules. The hydration sites on I CCGG are ordered based on the relative strengths of DNA to water bonds. as judged by their bond lengths. I CCGG packs together with the bottom of one helix in the minor groove of another. The packing scheme indicates that the minor groove of A-DNA is a relatively hydrophobic site. This fact may be important in the recognition of A-DNA by proteins and drugs. A list of refined coordinates of I CCGG is given in the appendix. Item Type: Thesis (Dissertation (Ph.D.)) Subject Keywords: (Chemistry) Degree Grantor: California Institute of Technology Division: Chemistry and Chemical Engineering Major Option: Chemistry Thesis Availability: Public (worldwide access) Research Advisor(s): Marsh, Richard Edward Thesis Committee: Dickerson, Richard E. (chair) Campbell, Judith L. Dervan, Peter B. Marsh, Richard Edward Defense Date: 18 May 1982 Funders: Funding Agency Grant Number NIH UNSPECIFIED Caltech UNSPECIFIED Record Number: CaltechTHESIS:05142018-150830492 Persistent URL: https://resolver.caltech.edu/CaltechTHESIS:05142018-150830492 DOI: 10.7907/3t68-dc11 Default Usage Policy: No commercial reproduction, distribution, display or performance rights in this work are provided. ID Code: 10880 Collection: CaltechTHESIS Deposited By: Mel Ray Deposited On: 15 May 2018 18:11 Last Modified: 26 Apr 2025 00:21 Thesis Files Preview PDF - Final Version See Usage Policy. 37MB Repository Staff Only: item control page

Iodo-CCGG A Single Crystal Structure of A-DNA Thesis by Benj am.in N. Conner In Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy California Institute of Technology Pasadena, California· 1982 (Submitted May 18, 1982) -iiAcknowledgements I wo.u ld like to thank my thesis advisor, Professor Richard E. Dickerson, for his advice and support throughout my graduate years. I would also like to thank Dr. Keiichi Itakura for his part in our productive collaboration, and Dr. Shoji Tanaka for his unconventional and informative introduction into the art of nucleic acid syn- • thesis. I am grateful to Dr. Tsunehiro Takano for his patience, and his willingness to share his complete knowledge of all things crystallographic. My lab partners, Horace Drew and Mary Kopka deserve credit for their help in the realities of experimental science. Finally, I would like to thank Steve Schichman and Perry Norton for their friendship during my years at Caltech. My graduate program was made possible through the financial support of the National Institute of Health and the California Institute of Technology. -iii- Abstract Iodo-CCGG was synthesized by the triester method and crystallized in space group P~2 1 2 . Electron density maps were phased using the anomalous scattering from the iodines and refined with the Jack-Levitt constrained refinement program. The final crystallographic R factor was 16.5% for 2 a data and 19.9% for all data. Solution of the structure revealed the first fragment of A-DNA to be seen in a single crystal. The helix .has a 2.2A rise per residue, 10.6 base pairs per turn. and an average propeller twist of 17°. A total of 86 water molecules were located in the crystal. Of these, 47 were within 3.5A of the 1CCGG molecule. A second layer of hydration is formed by water molecules which are within 3 .5A of two or more first layer water molecules . The hydration sites on IccGG are ordered based on the relative strengths of DNA to water bonds. as judged by their bond lengths . IccGG packs together with the bottom of one helix in the minor groove of another. The packing scheme indicates that the minor groove of A-DNA is a relatively hydrophobic site.This fact may be important in the recognition of A-DNA by proteins and drugs . A list of refined coordinates of IccGG is given in the appendix. -ivTable of Contents INTRODUCTION 1 1 .1 Historical Background 2 1.2 Definition of the Problem- 4 1 .3 Results 5 CHAPTER 1. CHAPTER 2. SYNTHESIS OF THE TETRANUCLEOTIDES:.. 7 2.1 General Considerations 8 2.2 Choice of Heavy Atom Derivative 9 2.3 Overview of the Triester Method 10 2.4 Materials Supplied by Collaborators · 13 2 .5 Synthetic Reactions Performed 16 2.5.1 Protection of Ic 16 2.5.2 Phosphorylation of Ic 19 2 .5.3 Preparation of Coupling Reagent 21 2 .5 .4 Coupling Reaction 21 2 .6 Purification During Synthesis 22 2 .7 Tests of 5-iodo-2'-deoxycytidine Stability 22 2.8 Deblocking 23 2.9 Purification after· deblocking 23 2.10 Enzymatic Digestions 25 ANOMALOUS PHASING 27 3 .1 The Path from CCGG to IccGG 28 3 .2 Data Collection for Anomalous Phasing 30 3 .3 Normal vs. Anomalous X-ray Scattering 33 3.4 Summary of Anomalous Scattering Cause and Effect 37 CHAPTER 3. -v3.5 Location of the Anomalous Scatterers 38 3.6 Use of Anomalous Differences in Phasing 44 3.7 Solution of the Phase Problem 46 3 .8 The Final Choice 49 CHAPTER 4. INTERMEDIATE RESULTS 52 CHAPTER 5 . SUBSEQUENT REFINEMENT AND HYDRATION 61 5 .1 Results of Subsequent Refinement 62 5 .2 The Structure: of IccGG 63 5 .3 Water Around IccGG 67 CHAPTER 6 CONCLUSIONS 6 .1 Structural Analysis 81 82 6 .1.1 Comparison of IccGG with fiber -studies 82 6 .1.2 Other single crystaLA-DNA structures 83 6.1.3 Major groove opening at purine pyrimidine sequence 83 6 .1.4 Packing of A-DNA 84 6 .1.5 Regularity of IccGG 85 6. 1.6 Variety in sugar pucker 87 6 .1. 7 Solution work with CCGG 89 6 .1.8 Other structural studies of A-DNA 90 6 .2 Water structure around IccGG 91 6 .2 .1 Review of solution studies 91 6 .2 .2 Correlations with IccGG 92 6.2.3 New ordering of hydration sites on DNA 94 6 .3 Spermine 96 6 .4 Summary of Advances from IcGG 97 APPENDIX REFINED IccGG COORDINATES 6.1 Coordinate Transformation 98 99 - 1 - Chapter 1. Introduction - 2CHAPI"ER 1. INTRODUCTION 1.1 Historical Background When this project began, crystallographic structural information concerning DNA came from the two extreme limits of the diffraction method. The gross structure of nucleic acids, the double helix with anti-parallel strands, was known from the earliest fiber diffraction studies. 1 2 The fine structure of single bases was known from high resolution single crystal studies. 3 Fibers of DNA from natural and artificial sources had been studied under various conditions of humidity and salt content. These studies resulted in the classification of DNA fibers into the A, 4 B, 5 C6 and D7 8 families . The limit of this fiber diffraction work was reached with least-squares linked-atom refinement of the A and B structures .9 However, fiber diffraction studies were not completely satisfactory because the molecules within these fibers were neither rotationally nor translationally aligned. This inherent disorder meant that fine details of the structure, including any structural effects arising from base sequence, were averaged out. The lack of specificity in these fiber diffraction photos became apparent with the disagreement over the structure of D-DNA. The initial D DNA structure 10 constituted what was later termed 11 "a bizarre left-handed double-helix with unusual furanose 1. J.D. Watson and F. H. C. Crick {1953). Nature (London} 171, 737-738 2. R. E. Franklin and R. G. Gosling {1953). Nature (London) 171, 740-741 3. J . Donohue {1968). Arch. Biochem. Biophys. 128, 591-594 4 . R. Langridge, W. E . Seeds, H. R. Wilson, C. W. Hooper, M. H. F . Wilkins and L. D. Hamilton (1957). J. Biophys. Biochem. Cytol. 3, 767-778 5. R. Langridge, D. A. Marvin, W. E. Seeds, H. R. Wilson, C. W. Hooper, M. H. F. Wilkins and L. D. Hamilton {1960). J. Mol . Biol . 2, 38-64 6. D. A. Marvin, M. Spencer, M. H. F. Wilkins and L. D. Hamilton {1961). J. Mol . Biol. 3, 547565 7. Y. Mitsui, R. Langridge, B. E. Shortie, C. R. Cantor, R. C. Grant, M. Kodama and R. D. Wells {1970) . Nature (London) 228, 1166-1169 8. S . Arnott, R. Chandrasekaran, D. W. L. Hukins, P . J. C. Smith and L. Watts {1974). J. Mol. Biol . 88, 523-533 9. S . Arnott and D. W. L. Hukins {1972). Biochem. Biophys. Res. Com. 47, 1504-1509 10. Y. Mitsui et al. {1970) . ibid -3- ring shapes", while the alternative structure 12 was praised as "a conformationally unexceptional right-handed double-helix with standard furanose rings". Left-handed structures eventually were proposed as alter natives for both A and B families of DNA. 13 14 These cracks in the supposedly infallible crystallographic method of structure determination highlighted the need for further crystallographic studies on DNA crystals, which had no fiber disorder. What were needed were single-crystal studies of DNA oligomers. Considerable single-crystal x-ray diffraction work had already been done with nucleosides and nucleotides of RNA as well as with nucleotides that paired to form single steps of a double helix. 15 This body of work provided a detailed look not only at the geometry of the nucleic acid components, but also at alternate base-pairing mechanisms, 16 water structure near the DNA, 17 and interactions of both intercalating 18 and non-intercalating 1 ~ drugs with RNA. These structures showed, in detail, what could be expected from small fragments of RNA. However, only one DNA dimer, pTpT, had been studied 20 when the work described in this thesis began, and it was not double helical. Furthermore, studies of short pieces of RNA were not a sufficient base for extrapolation to longer pieces of DNA. A frequently voiced criticism of studies with nucleic acid dimers was that the crystallographer was only "seeing two end effects" in his crystals . The missing informati on 11. S . Arnott et al. (1974) . 12 . S . Arnott et al . (1974) . 13. G. Gupta, M. Bansal. V. Sasisekharan (1980). Proc. Nat. Acact. Sci. USA 77, 6486-6490 14. V. Sasisekharan, M. Bansal a n d G. Gupta (1981) . Biochem. Biophys. Res. Com. 102, 10871095 15. J. Rosenberg, N.C . Seeman, R. 0 . Day, and A. Rich (1976) . J. MoL. BioL. 104, 109-144 16. K. Hoogstee n ( 1959). Acta, Oryst. 12, 822-823 17 . B. Hi n gerty, E. Subramanian, S . D. Stellman, T. Sato, S. B. Broyde and R. Langridge {1976) . Acta, Oryst. B 32, 2998-3 013 18. H . M. Sobell, C. Tsai, S . C. Jain and S. G. Gilbert {1977). J. MoL . BioL. 114, 333-365 19. D. M. L. Goodgame, I. Jeeves, F. L. Phillips and A. C. Skapski {1975). Biochem. Biophys. · Acta 378 , 153-157 20. N. Came rman, J. K. Fawcett and A. Camerman (1976) . J . MoL. BiaL . 107, 601-621 -4- would be found in structures of intermediat'e chain length. These would have sufficient resolution to see the details of DNA structure and yet would also be long enough to show some of the structural constraints on DNA of biological size. 1.2 Definition of the Problem As a first step in this direction, work was begun on a four-base segment, a tetramer, of DNA. This length was chosen because it was thought that a smaller piece would not provide enough information, and that a larger piece would be harder to crystallize. The latter point subsequently proved to be incorrect. In fact, the tetramer length is so short that, in terms of packing within the crystal, its behavior is as much that of a sphere as that of a long cylinder. In at least one case, the tetramer proved to be disordered 21 while a hexamer of the same sequence crystallized well. 22 The sequence CpCpGpG was chosen for several reasons. The sequence is self complementary so that only one strand need be synthesized but in solution two molecules will pair to form a double helix. The sequence is CG rich, so that even for this short piece the melting temperature, T m• of the double helix is not too low. (T m has been measured by NMR as 42°C at 20 mM [CCGG], in 0.1 M phosphate, 0.01 M ethylenediamine-tetracetic acid (EDTA) and neutral pH.) 23 CCGG was also known to be the target site for the restriction endonuclease Hpa II from Hemophilus parainfluenza and its associated methylase. For Hpa II, methylation takes place on the 5 position of the ring of the second cytosine 24 and cutting takes place on the 3' side of the first cytosine. 25 An important reason for choosing this particular sequence of self complementary 21. J. L. Crawford, F. J. Kolpack, A. H.-J. Wang, G. J. Quigley, J. H. van Boom, G. van der Marel and A. Rich (1980). Proc. Nat. Acad. Sci. USA 77, 4016-4020 22. A. H.-J. Wang, G. J. Quigley, F. J. Kolpack, J. L. Crawford, J. H. van Boom, G. van der Marel and A. Rich (1979). Nature (London) 282, 680-686 23 . D. J. Patel (1977) . Biopolymers 16, 1635-1656 24. M. 8 . Mann and H. 0. Smith (1977). Nuc. Acid Res. 4, 4211-4221 25 . D. E. Garfin and H. M. Goodman (1974). Biochem. Biophys . Res. Com. 59, 108-116 -5- CG-rich DNA was that it contained the GpG sequence. This sequence was of interest because it was known to be the most active site for binding of the most commonly used inorganic cancer chemotheraputic agent cis-dichlorodiaminoplatinum(II) (cisplatin). 26 Since an alternating sequence, CGCG, was already being used in the Dickerson group on a separate project, it could also be used in the study of cisplatin. One part of the project was to be a comparison of cisplatin + CCGG and cisplatin + CGCG to study the binding mechanism of this drug. Along with the cisplatin + DNA experiments we also planned to try crystallizing the sequence without any drugs. Not only would the structure of CCGG itself be interesting in its own right for reasons detailed previously, but it would also be useful for comparisons with the cisplatin studies. 1.3 Results As often happens, the aspect of the experiment which was planned to be the major part turned out to be the minor part, and vice versa. Co-crystallization of DNA with cisplatin proved to be very difficult. Although a very large number of conditions were tried, suitable crystals were never obtained. On the other hand, the native crystallizations were successful and provided the first detailed look at a member of the A-DNA family. This single-crystal IccGG structure gives a more detailed picture of A-DNA than the fiber structures, and points out sequence specific changes in the helix geometry. It also allows us to see the water structure around the DNA and to compare it with the recent study of water structure in B-DNA. 27 Subsequent single-crystal x-ray analyses of GGCCGGCC 28 and GGTATACC 29 have allowed us to study A-DNA structure as a function of both length and sequence. This body of 26. P. J . Stone, A. D. Kelman, F. M. Sinex, M. M. Bhargava and H. 0. Halvorson (1976). J. Mot. Biol. 104, 793-801 27. H. R. Drew and R. E. Dickerson (1981). J. Mol. Biol. 151, 535-556 28. A. H.-J. Wang, S. Fujii, J. van Boom and A. Rich (1982). submitted- Proc. Nat. Acad. Sci. USA 29. z. Shakked, D. Rabinovich, W. B. T. Cruse, E. Egert, 0. Kennard, G. Sala, S. A. Salisbury and M.A. Viswamitra (1981). Proc. R. Soc. Lond. B 213, 479-487 - 6- information represents a large advance in our understanding of DNA st ructure. -7- Chapter 2. Synthesis of the Tetran ucleotides - B- CHAPrER 2. SYNTHESIS OF THE TETRANUCLEOTIDES 2.1 General Considerations Tetranucleotides were not commercially available when this project was begun, so compounds for study had to either be isolated from biological sources or synthesized by biochemical or organic means. Isolation of a given sequence from biological sources in sufficient quantity for x-ray crystallization involves enormous ° problems in purification. 3 For a piece of DNA as small as a tetramer, these problems are probably insurmountable. By manipulating specific DNA polymerases, biochemical synthesis had in the past provided homopolymers, alternating sequences, and occasionally even repeating triplet sequences 31 of DNA suitable for xray fiber work. However, the choice of sequence is very limited, and does not include CCGG. Therefore, the remaining method was organic synthesis . The native tetramer, CCGG, was generously provided by Dr. Keiichi Itakura's laboratory at the City of Hope Medical Research Center. In order to solve the crystallographic problems, one or more heavy-atom derivatives of the native compound are required. One method, that of isomorphous replacement, involves soaking a heavy metal into the crystal and using its position and the resulting change in intensity as a basis for solving the structure. This approach had been tried with CGCG by a fellow graduate student, Horace Drew, without success. He found that soaking heavy metals into a polyanion of regular repeating structure resulted in multiple substitution, and uninterpretable results. Another possibility is to modify the DNA after synthesis but before crystallization. Since DNA has so many chemically reactive groups, I felt that this approach would lead to modification of varying degree at multiple sites. For example, it would be very difficult to prevent modification of both cytosines in CCGG instead of just one. 30. 0. B. Kallai, J. M. Rosenberg, M. L. Kopka, T. Takano, R. E. Dickerson, J. Kan and A. D. Riggs (1980) . Bioch. Bioph. Acta 601, 113-124 31. E. Selsing, S. Arnott and R. L. Ratliff {1975). J. Mol. Biol. 98, 243-248 -9Such multiple modification increases the crystallographic problems tremendously. Therefore, it seemed best to synthesize the required derivatives starting from commercially available modified nucleosides, following the procedures of the triester method of synthesis. 2.2 Choice of Heavy Atom Derivative For the sequence CCGG, the two heavy atoms, iodine and bromine, were chosen for both synthetic and crystallographic reasons. These two halogens were available as 5-iodo-2'-deoxycytid.ine from P. L. Biochemicals and 5-bromo-2'-deoxycytidine from Sigma Chemical Company. Bromine was chosen because its atomic number of 35 makes it a heavy but acceptable choice for the isomorphous phasing technique . In order for the isomerphous phasing method to be effective the change in diffraction intensity must be significantly above errors in measurement. 32 The change in average intensity of a CCGG tetramer because of a single bromine atom can be estimated 33 as 130%, far above the required minimum. Iodine was chosen for use in heavy-atom phasing. As a rule of thumb for heavyatom derivatives, 34 2: ZHght 2 cell ~ l.O 'cell 2: Zheavy2 (2.1) where Zught and Zheavy refer to the atomic numbers of all light and heavy atoms in the cell. For iodine attached to a CCGG tetramer, this ratio is 0.66. Thus iodine's 53 electrons make it a light but acceptable choice for heavy-atom phasing. Four different tetramers, IccGG, cicGG, BrccGG, and c 8 rcGG, were synthesized by the 32 . T. L. Blundell and L. N. Johnson {1976). Protein Crystallography, Academic Press, New York, p161 33. F. H. C. Crick and B.S. Magdoff {1956). Acta Oryst. 9, 901-908 34. G. H. Stout and L. H. Jensen (1968). X-ray Structure Determination, The Macmillan Co., New York, p278 - 10triester method, in the hope that at least two of the five sequences (including the original CCGG) would crystallize as an isomorphous pair. 2.3 Overview of the Triester Method In the hands of an expert, the triester method of DNA synthesis allows organic synthesis of DNA segments up to around 25 base pairs in length. For smaller pieces, the yields at each coupling step are often greater than 90%. However, in order to crystallize DNA extremely high purity is required. Stringent purification at each step produces yields that are lower, but are still above 50%. The procedure is relalively rapid, and bases can usually be added at a rate of one per day. As originally proposed, 35 36 the triester method was used only to make poly dT DNA. However, development since then, including that done by Dr. Itakura in the laboratory of S. A. Narang, 37 allows any sequence to be synthesized. An overview of the triester synthesis is shown in Figure 1. Basically, the method is the following: 38 1. Block all reactive groups on the DNA monomers 2. Deblock only the reactive sites (ie. the 5' end of one sugar and the 3' end of another) 3. Couple the free 3' end to the free 5' end 4. Purify the reaction mixture with silica gel chromatography 5. Start again with step 2 and repeat cycle until desired chain length is reached 6. After desired chain length is reached, remove all blocking groups from the DNA 35. A. M. Michelson and A. R. Todd ( 1955). J. Chern. Soc. 2632-2638 36. R. L. Letsinger and K. K. Ogilvie (1967). J. Am. Chern. Soc. 89,4801-4803 37. K Itakura, N. Katagiri, C. P. Ball, R. H. Wroghtman and S. A. Narang (1975). J. Am. Chern. Soc. 97, 7327-7332 38. K. Itakura and A. D. Riggs (19?0). Science 209, 1401-1405 - 11 - 7. Purify the deblocked DNA by chromatography. - 12- • J8 X 8 DMT 2 Stepa OH 2 B OMT B 3 B B J- OJ-OJ-0 II II II ~ ~ ~ -OCE 0 0 0 R ~---3 ----t~ OMT B J- ~J-~J0 R R ··j B 4b 4a t B B B . ·J-OJOJu 5 + HO R R R OMT R R 1) 0 OBz R TPSTt ~ 3) Et3N B B B B B P P P P- P 1 I I I I I 0 0 0 0 0 0 R R R R R R B HO 1) TPSTe 2) H• 3) Silica qfl B B B B B P. 1- P P r B I 1 "'-J 0 0 ·R R P I R P P I I o- P I o- P I o- P I o- Silico qel 3) NH 4 0H 4) AcOH B B P I P I I 0 R 9 2) o- B 0 R B B B B B J ~~ ~u ?u ?J ~~ ?u ?U- ~U- ~U- ~J- ?,Uo- B B J- ~J- ~ I c~J-~J-~j TPSTt 8 HO ~ 7 J- ?.J-~J-~J-~J-~j~ oP ~ R R 21 Silica qel B II 0 6 ~ B oez 0 R P I o- P , o- P 1 o- P I o- P OH t o- 10 R= @- Cl X= Bases B =Protected bases 0 CE = /3- Cyanaethyl Bz = C-@ DMT : d imethoxyh1tyl Figure 1. Overview of DNA synthesis by triester method9 OBz - 132.4 Materials Supplied by Collaborators The syntheses were run in collaboration with Dr. Itakura's laboratory at the City of Hope and under the guidance of Dr. Shoji Tanaka from that lab . They supplied not only the necessary experience but also some of the materials used. In particular, the fully protected monomers G. Gp and Cp were generously donated. Horace Drew was kind enough to donate fully protected 5-bromo-2'-deoxycytidine. For the CCGG derivatives, the only nucleotides involved are guanine, cytosine and the halogenated cytosines . The 06 site on guanine and the 02 site on cytosine are not reactive enough under subsequent conditions to require protection. However, the base amino groups, N2 for guanine and N4 for cytosine, must be b locked. The N2 of guanine is protected by condensation with isobutyric acid. The N4 of cytosine is blocked by condensation with benzoic anhydride. (see section 2.5.1) The sugar phosphate backbone contains several different blocking groups because the coupling sequence requires that a protecting group on a specific reactive site of the DNA be removed or replaced without affecting the other protecting groups . The 5' hydroxy of the deoxyribose is blocked with a 4,4'-dimethoxytrityl group (DMT in Figure 1). For all residues except the 3' terminal residue, the 3' hydroxy of the deoxyribose is phosphorylated in preparation for the coupling reaction . As the name of the synthesis suggests. the phosphate is protected as a triestar. It is esterified to 3' hydroxy of the sugar, to a p-chlorophenol. and to a (3cyanoethyl group. If the deoxyribose 3' hydroxy is not to be phosphorylated, it is blocked by either anisole or benzoyl. Anisole was used for the two brominated tetramars. However, when the iodinated tetramers were made, benzoyl was recommended because of problems encountered by Dr. Itakura's group in deblocking anisole derivatives . All of the protecting groups are left in place throughout the synthesis with the exception of the (3-cyanoethyl on the phosphate and the 4,4'dimethoxytrityl on the 5' hydroxy of the deoxyribose. These two reactions are - 14- discussed in greater detail in the section 2.5. The fully protected forms of these nucleotides received from Dr. Itakura's lab are shown in t he following figure . - 15- Fully protected Fully protected terminal Cytosine Fully protected Figure 2. Fully protected groups not synthesized - 162.5 Synthetic Reactions Performed The reactions I performed dealt with the protection and phosphorylation of 5iodo-2'-deoxycytidine and the actual chemical linking of the monomers together. This latter part involved synthesis of the coupling agent, removal of protection groups in preparation for coupling and the coupling react ion itself . 2. 5.1 Protection of Ic: For cytosine and its iodine and bromine derivatives, the amine is the only reactive group on the base itself. It is protected by refiuxing 5iodo-2'-deoxycytidine with a 2 molar excess of benzoic anhydride in dry ethanol for 3 hours while constantly stirring. 39 Progress of the reaction was checked hourly by chromatography on silica gel 60 F254 plates run in 5/ 1 CHCl3 /CH 3 0H. Disappearance of the low mobility iodocytidine marked the end of the reaction. After the reaction was completed, the mixture was put in an ice bath and the product precipitated in the form of light yellow crystals. These were filtered out and washed with diethyl ether. A sample of this N-benzoylated 5-iodo-2'-deoxycytidi.il.e was submitted to the Caltech analytical lab for elemental analysis of H, C, N, and Iodine. The reported value of the iodine was within 2.% of the calculated value, indicating that the iodine was not destroyed during the protection reaction. The 5' hydroxy of the deoxyribose was protected by condensation with a 50.% molar excess of 4,4'-dirn.ethoxytrityl chloride in 5 ml dry pyridine/ mM N-benzoyl cytidine. Progress of the reaction was checked every 30 minutes by chromatography on silica gel 60 F-254 plates in 9 I 1 CHC1 3 /CH 3 0H. After chromatography, these plates were sprayed with perchloric acid and heated on a hot plate. Under these conditions the dimethoxytrityl group turns bright orange and the deoxyribose becomes dark brown. Thus the tritylated sugar can be identified as an orange spot 39. Synthetic Procedures in Nucleic Acid Chemistry Vol. I pg 285 et seq. Zorbach and Tipson editors ( 1968) Interscience Publishers. - 17- which turned brown upon heating. The reaction was stopped with 5 ml/mM dimethoxytrityl of NaHC0 3 . and the product extracted in CHC1 3 . The CHC13 was washed twice with water and evaporated to dryness . Purification was done on a silica gel column of the type described in section 2.6. - 18- Figtire 3. Protection of·lc - 19- 2.5.2 Phosphorylation of 1c: The iodocytidine is now protected and ready for phos- phorylation of the 3' hydroxy end. This reaction is run in two steps as shown in Figure 4. First, to prepare the activated reagent, the phosphorylating reagent (50% molar excess) is added to 1,2,4-triazole (twofold molar excess) in dry pyridine (5ml/mM phosphorylating reagent). Triethylamine (3 .3 molar excess) in dry dioxane (twofold molar excess over triethylamine) is added dropwise to the mixture while it is stirring in an ice bath. After addition, the mixture is stirred 15 minutes in its ice bath and then left at room temperature for 1 hour. The sticky white triethylamine salt is filtered out with a glass frit funnel and the liquid reduced to a volume of 6 ml. This reduced volume is added to the protected monomer and allowed to react at room temperature for 4 hours. Every hour, aliquots of the reaction mixture are treated with (CH 3 CH 2 )sN/pyridine/water and chromatographed on silica gel plates in 15/1 CHC13 /CH 3 0H. In this system the charged product shows low mobility with respect to the uncharged starting material. The reaction is stopped by addition of 1 M triethylamine bicarbonate pH=7.5, evaporated to a small volume and the product extracted with chloroform. - 20- nJ Cl I I o=P-CI TEA 7>, . ~\ -or=N:::/ 'f ' N-ij b 'N I o Cl Figure 4. Phosphorylation of rc Cl -21- 2 .5.3 Preparatian of Coupling Reagent: The actual coupling of protected monomers is performed by TPS-tetrazole. The synthesis of this reagent is shown in Figure 5. Triisopropylphenylsulfonyl chloride, TPS-Cl, is mixed with an equimolar amount of tetrazole in a dry dioxane (1.5 ml/mM TPS-Cl). An equimolar amount of triethylamine is added dropwise while the mixture was stirring rapidly in an ice bath. After two hours, the triethylamine chloride salt is filtered out and the reaction mixture reduced to a small volume. The product is crystallized overnight at 4°C from benzene/n-pentane. H-a Figure 5. TPS-tetrazole synthesis 2.5.4 Coupling Reactian: As shown in Figure 1., the actual linking of two residues. X-p and Y, together to form the 3' to 5' phosphate linkage X-p-Y is done in four steps: 1. Remove p-cyanoethyl protecting group from the phosphate group of X-p. This is done by treatment of the monomer with 3/1 I 1 pyridine/water/ (CHsCH2)sN at - 22room temperature for 20 minutes. 2. Remove dimethoxytrityl group from 5' hydroxy of deoxyribose of Y. This is done in 2% benzenesulfonic acid for 20 minutes at 0°C. 3. Add a 50% molar excess of X toY and dry thoroughly. In a small volume of pyridine add the two monomers and the coupling reagent, triisopropylsulfonyl tetrazole. The reaction runs about 1 hour at room temperature and is checked every 30 minutes by chromatography on a silica gel plate in 9 I 1 CHC13 / CHsOH. The reaction is stopped by addition of water and the mixture evaporated to dryness. The final traces of pyridine are removed by evaporation of toluene and the product is purified by chromatography on a silica gel column as described in section 2.6. 2.6 Purification During Synthesis Because of the high purity required for crystallization, extra care must be taken during the synthesis to purify intermediate products. This is usually done by chromatography on a silica gel column. A typical column is 10 em. in diameter and consists of 5.5 em. bed of Silica Gel 60 H (MCB Manufacturing Chemists Inc.) in chloroform. The material to be purified is loaded on the column in chloroform and then chromatographed with a 0% to 5% gradient of methanol in chloroform. Fractions are collected and spotted on Silica Gel F-254 thin layer plates from E . M. Laboratories Inc. Good fractions are combined and concentrated. Careful chromatography during synthesis produces much cleaner reactions later and a purer final product. This purification is a critical step for DNA intended for crystallization. 2. 7 Tests of 5-iodo-2"-deoxycytidine Stability Three experiments were performed in order to check the stability of iododeoxycytidine under deblocking conditions. - 231. 10% isopropylamine in methanol for 23 hours at room temperature 2. 2/1 cone . NH4 0H/pyridine for 12 hours at room temperature 3. 2 I 1 cone. NH 4 0H/pyridine for 4 hours at 50°C When judged by chromatography on thin layer plates, all reactions showed deblocking with no major side products . I used the last method for deblocking the tetramers because it was the quickest and most commonly used. 2.8 Deblocking After synthesis is completed, the compounds must have all of their blocking agents removed. The blocked DNA is dissolved in a small amount of pyridine and added to 80 ml of 2/1 cone. NH4 0H/pyridine in a round bottom flask. The flask is then sealed with a glass stopper, wrapped with tape for protection is case of explosion, and heated to 50°C in a water bath for 12 hours. This procedure removes all base-labile protecting groups, ie. aU protecting groups except the dimethoxytrityl on the 5' end of the DNA. After cooling to room temperature and evaporating the mixture to dryness, 100 ml of galcial acetic acid are added for 30 minutes at room temperature. The short reaction time and moderate temperature prevent depurination of the DNA. The mixture is evaporated to dryness again, the blocking agents are extracted with ether and the solution then is made ready for chromatography on DEAE cellulose. The sequence of deblocking given here, base before acid, is important in order to prevent the amination reaction shown in Figure 6. Therefore, the dimethoxytrityl group is left on until after the basic deblocking step is completed. 2.9 Purification Mter Deblocking After deblocking, each tetramer had to be purified to remove blocking agents, partially blocked tetramers, unreacted materials and products of side reactions. Two separate portions of the native compound, CCGG, were received from Dr. - 24- c;.NH 3 Base ofP b ~ase 0 ko ~Base •I q I Figure 6. Amination side reaction of deblocking step. Itakura's laboratory. Purification was done on DEAE cellulose in a triethylammonium bicarbonate pH=7 .5, 15% ethanol buffer system. A gradient from 0.1M to 0 . 7M TEAB was used to chromatograph the tetramer. Of these two preparations of the same compound, only one would crystallize, although they had identical UV absorption and appeared identical by high performance liquid chromatography, HPLC. This situation prompted an increase in concern about purity so a new method of purification was adopted for the halogenated tetramers. The new method of DNA purification was one used for longer pieces of DNA. 40 It involved chromatography on DEAE cellulose in 7 M urea, 20 mM tris pH=7.5 with a NaCl gradient from 30 mM to 60 mM. DNA from the desired peak was trapped on another DEAE column, washed free of urea, and taken off with TEAB buffer with 15% ethanol. This chromatography system was far from ideal and usually resulted in the 40. E . Ohtsuka, A. Kumar and H. G. Khorana (1972). J .Mol. Biol. 72, 309-327 -25- product peak appearing as a relatively broad peak close to impurities. A decision about which fractions to pool was made on the basis of HPLC results from selected fractions. This allowed separation of impurities which could not be differentiated by their UV absorbance. Samples were run on a Spectra Physics HPLC equipped with AAX. ion exchange column and using a gradient from 0.005 M to 0.05M KH 2 P04 , pH 4.4 as a buffer system. This system served very well in distinguishing compounds which could not be cleanly separated by DEAE chromatography. The final products of these purification efforts were always single peaks with less than 1% contamination as judged by HPLC. All four of halogenated tetramers were purified in this manner and three of them were subsequently crystallized. The best that can be said for this expensive, time consuming and messy purification procedure is that it had the desired result. This fact is mainly due to the ability of the HPLC to discriminate between product and impurity cleanly when the DEAE column could not. 2.10 Enzymatic Digestions A final test of chemical composition was performed using snake venom phosphodiesterase which cleaves the tetramers on the 3' side of the phosphate. The monomers were then chromatographed on thin layer plates of Cellulose F. In order to prevent distortion of chromatography by the phosphodiesterase, the material was applied to a cellulose F plate and chromatographed in water for a short distance. The migrating monomers were scraped off the plate, eluted into water, concentrated by lyophilization, applied to another cellulose F plate and chromatographed against . standards in 6 I 3 I 1 iospropanol/water I cone. NH 4 0H This procedure eliminated .the distortion of the faster migrating cytosine residues. After chromatography, the separated components were scraped off the plate, eluted with water and their UV spectra taken. The spectra identified the compound and served to determine the - 26relative amounts of each monomer present. From these spectra the compositions of the tetramers were confirmed. In the case of 1CCGG , this technique also verified the fact that the iodine was still intact after purification, at a time when it could not be located crystallographically . - 27- Chapter 3. Anomalous Phasing -28 ... CHAPTER 3. ANOII.ALOUS PHASING 3.1 The Path from. CCGG to 1CCGG The project began as a study of the compound CpCpGpG. To solve the crystallographic phasing problem, we intended to use the technique of isomorphous replacement. This involves either making a heavy atom derivative of the molecule while it is in the crystal by soaking in heavy ions, or modifying the molecule before crystallization and then crystallizing it in the same space group with the same unit cell packing as the unmodified form. With this goal in mind, the two bromine derivatives described in Chapter 2 were synthesized and crystallization trials were begun. Only one of these compounds, BrccGG, could be crystallized, and it was useless as an isomorphous derivative because it was not in the same space group as the native tetramer. Since BrccGG could not be used as an isomorphous derivative, could the crystal structure be solved by itself, using bromine as a heavy atom? The rule of thumb for heavy atom derivatives expressed in equation (2.1) gives a value of 0.29 for bromine on the CCGG tetramer. For a good derivative this ratio should be ;:::j 1 .0. Therefore, bromine is not heavy enough by itself to phase these crystals. At this point, the phasing problem was reevaluated and two iodine derivatives of CCGG were synthesized as described in Chapter 2, for use in heavy atom phasing. Both of these iodine derivatives were crystallized successfully. One, C1CGG, formed disordered crystals. The other, 1CCGG, formed crystals suitable for x-ray diffraction study. A survey of all crystals grown is given in Tables 1 and 2. Unit cell dimensions and density measurements of 1CCGG indicated two molecules in the asymmetric unit. This meant that two iodines must be located before phasing could begin. The fact that this 1CCGG crystal is in a different space group from both CCGG and BrccGG meant that isomorphous phasing could still not be used. There- -29- DNA CCGG BreeGG e 8 reaG IcCGG c rcaa Conditions 1.49 mM CCGG Z. 94 mM MgCl2 isopropanol, zoe 2 .Z3 mM BrceGG 4.33 mM Mgel 2 isopropanol, zoe no xtal 5. ZmM IeeGG 0 .4 mM spermine e1 30 m M Na Cacodylate pH=7 .5 isopropanol. zoe 4 .4 mM c ieGG 7 .3 mM Mgel 2 ethanol. zoe Space Group PZ 1 Notes thin laths l .OxO.Zx0.03mm PZ 1 Z 1 Z thin plates 0 .5x0.2x0. 03mm --- can't win them all 0 .5x0.5x0.3mm P4aZlZ tetragonal 0 .3xO.ZxO. Zmm blocks TABLE 1. Summary of crystals grown of CCGG and its derivatives. fore the heavy atom approach was begun . Diffraction data were collected from 1eeGG crystals, and Patterson maps were studied in an attempt to locate the heavy atoms. Because of the inherent symmetry of the DNA molecule, many large peaks appeared on the Patterson maps which were not caused by the iodines, making interpretation of the maps difficult. Several solutions to the Patterson maps were tried but none produced interpretable electron density maps . Harker sections of this map are shown in Figure 1. - 30- DNA CCGG Space Group P2 1 BrccGG P2 1212 c 8_rcGG ICCGG no xtal P4 3 2 1 2 ere aa tetragonal Unit Cell Dimensions b=25.6A c=30.4A a=39.7A b=31.BA. c=13.7A Asymmetric Unit two single strands two single strands --- --- a=41.1A b=41.1A c=26.7A a=56.A b=56.A. c=26 .A two single strands four single strands a=22.2~ TABLE 2. Crystallographic parameters of CCGG and its derivatives. At this point the method of anomalous dispersion was considered for phasing. It had not been tried previously because the anomalous effect is usually too small to be considered when dealing with molecules of this size containing only H, C, N, 0, and P. However, investigation of the anomalous method showed that the iodine attached to CCGG had twice as much phasing power as the chlorine used to calculate phases for ephedrine hydrochloride.'H Therefore data collection of Friedel pairs began in preparation for anomalous phasing. 3.2 Data Collection for Anomalous Phasing Since the anomalous difference to be measured is small, great care must be taken in data collection. The Friedel pairs, Fhkl and Fm. must be collected within a short time of each other in order to prevent crystal decay from affecting the anomalous difference. For IccGG, Friedel pairs were collected within 30 minutes of each other. The decay in intensity during this period of time was less than 0.1%. The Friedel pairs must also be collected in such a manner that absorption differences due to crystal orientation are negligible. Two methods suitable for our 41. G. N. Ramachandran and S. Raman (1956). Ourr. Sci. 11, 348-351 - 31- o.s D·f# •l (1.~ "·' D -0 U• V Figure 1. Harker sections from 4/mmm averaged data -32- four circle di.tfractometer were tried. The first method is to collect Ihkl at 29 c.>, rp, and x and Im at 29, c.>, rp + 180°, and I -x. This is the method supplied in the software for our Syntex PI diffractometer. The advantage of this method is that the reflections are collected 180° apart in q;. The absorption correction1 applied to these diffraction data assumes that x-ray absorption is equal when the crystal is at rp and rp + 180°. Experimentally, there is only a small difference so this assumption is usually good. Thus, the same absorption correction is applied to both reflections and the anomalous difference between Ihkl and Irn is not obscured by errors in absorption. The other method was suggested by Dr. Tsunehiro Takano. IccGG crystals were usually mounted with b • along the goniometer head axis. Since in P4 3 2 1 2, Ihkl = Irn, one can collect Ihkl at 29, c.>, q;, and x and Ihl<l a t 29 c.>, q;, and I -x. The advantage of this method is that both reflections are collected at exactly the same rp value so the absorption correction is exactly the same for both reflections. The Pl collects data with the l index incremented most rapidly. Therefore, Friedel pairs could be collected close together in time if the crystal were reindexed with the old b • axis as the new c • axis. The new indexing was done as follows. hnew =hold knew= lold In order for this method to collect data efficiently, the diffractometer must be stopped and reset every time the l index reaches its limit. This method of data collection requires constant operator intervention. 1. A. C. T. North, D. C. Phillips and F. S. Matthews (1968). Act«Z Oryst. A24, 351-359 - 33- Three angstrom data were collected by both methods on a single crystal. In order to compare the two methods, an R factor for each data set was calculated from the two measurements of each centric reflection. Both data sets had the same centric R factor of 5% in F, indicating equal quality data by both methods. Hence, the rest of the data collected (to 2.1 Aresolution) used the method of rp and rp + 180°, because it required less operator intervention. 3.3 Normal vs. Anomalous X-ray Scattering To understand the cause of anomalous scattering of x-rays by iodine, a brief review of normal x-ray scattering is in order. In a single crystal x-ray diffraction experiment, x-rays are scattered by the electrons of the molecule of the crystal. The three-dimensional periodicity of the crystal lattice causes interference between the scattered x-rays and results in the formation of a three dimensional lattice of diffraction spots of varying intensity. From the value and position of these intensities the electron density throughout the crystal, and thus the structure of the molecule, is deduced. In the classical derivation of x-ray scattering by an atom, the electron is considered to be weakly coupled to its nucleus to form a simple harmonic oscillator. The normal oscillation frequency of this electron, CJ0 , is the frequency with which this electron oscillates when it is displaced from equilibrium. In the diffraction experiment, this atom is placed in an incident beam of x-rays whose frequency. CJ, is large compared to CJ 0 • Under such conditions, the electron is driven by the sinusoidal electric field of the incoming x-ray. The solution to the problem of a damped simple harmonic oscillator driven by a sinusoidal force is well known. 2 An oscillator driven at frequency CJ reaches a steady state in which its frequency is equal to the driving frequency, but its amplitude of oscillation and the phase difference between the driving force and the particle are both functions of CJ and CJo· 2. J. B. Marion {1970). Classical Dynamics of Particles and Systems, Academic Press, p118 -34- The amplitude is given by (3.1) where • F is the force on the oscillator • (3 is the viscous damping force of the oscillator and • o, the phase difference between the driving force and the oscillating particle, is defined by o = tan -l{ c.>P } 2 c.>o 2 _ CJ2 (3.2) It is this variation of amplitude and phase difference with c.> and c.>0 that is at the heart of anomalous scattering. Both of these functions change rapidly when c.> ~ c.> 0 but are relatively constant otherwise. In the case of normal scattering, c.>>> c.>0 and scattering occurs with the scattered wave 180° out of phase with the incident wave. 3 However, if the scattering atom happens to have an x-ray absorption edge near the frequency of the incident radiation, then c.> ~ c.> 0 and both A and o undergo drastic changes. For example, t5 changes from 180° when c.>>> c.>0 to 90° when c.> = c.>0 • It is important to remember that different atoms usually have absorption edges at different frequencies. Since the amount of anomalous scattering depends on how close c.> is to c.>0 , when one atom is scattering with a large anomalous component, the other atoms in the structure probably will not be. In the specific case of 1CCGG, only iodine has a significant anomalous scattering component for the CuKa whose wavelength is 1.5418A Therefore, the phase and amplitude of a wave scattered from an iodine would be different from those scattered from a carbon in the same position. These differences in phase and amplitude must be taken into account when 3. R. W. James ( 1967). The OpticaJ Principles of the Diffraction of X-rays G. Bell and Sons Ltd. (London) p135 - 35- the electron density of the cell is calculated. This correction to the electron density map is done by placing two mathematically di.fl'erent types of atoms at the site of the anomalous scatter. (It is important to note here that in order to make this correction the position of all of the anomalous scatters in the unit cell must be known.) One of these atoms scatters as would a normal atom, such as a carbon, but the other atom scatters goo out of phase from a normal atom. The mathematics of this correction involves changing the scattering factor of the atom from a real number, f, to a complex number, f + f' + ~f", having both real and imaginary parts. Now, instead of an atom's contribution to a scattered wave being f [ cos2rr(hx + ky + lz) + isin2rr(hx + ky + lz)] (3.3) it becomes (f + ~f') [ cos2rr(hx + ky + lz) + isin2rr(hx + ky + lz)] (3.4) + ~f" [ - sin2rr(hx + ky + lz) + icos2rr(hx + ky + lz)] Now the real term has a sine component (which is goo out of phase from its normal cosine component) and the imaginary term has a cosine component (which is goo out of phase with its normal sine component). That is to say, at the position (x,y,z) there appear to be two atoms. One scatters as would a normal atom while the other scatters goo out of phase from a normal atom. By varying the number of electrons on these two atoms, the scattered wave from the anomalous scatter can be reproduced and the electron density of the crystal calculated. Usually the number of electrons on the 'imaginary" atom is much less than that on the other normal atom. In the case of iodine in CuKa radiation the "imaginary" atom has 7.2 electrons while the normal has 52 electrons. Values of f, ~f', and ~f" for all atoms in the structure are given in Table 3. - 36- f ~f' ~f" H . c N 1.0 0.0 0.0 6.0 0.0 0 .0 7.0 0.0 0.0 0 8.0 0.0 0.1 p 15.0 0.2 0.5 I 53.0 -1.1 7.2 TABLE 3. Corrections for anomalous scattering for all atoms in 1CCGG at sin ® I A. = 0 with CuKa radiation One of the manifestations of anomalous scattering is that Friedel's law, which says that the intensity of the hld and hkl reflections must be equal, is no longer valid. The reason for this breakdown can be seen by examining the structure factor equation. For normal x-ray scattering, Fhkl = ~fj[ cos21r(hxj + kyJ + lzj) + isin21r(hxJ + kyJ + lzj)] (3.5) j = Ahkl + iBhkl where the sum runs over the j atoms in the unit cell which have positions (xj,yj,zj), and scattering factor fj. For the Friedel mate of the above reflection, Frn = 2:fj[cos21r( -hxj -kyj -lzj) +isin2rr( -hxJ -kyj -lzj)] (3.6) j = 2:fj[cos21r(hxj + kyj + lzj) - isin2rr(hxj + kyj + lzj)] j Since the intensity of a given reflection, Ihkl, is proportional to (3.7) -37- the change in sign of the imaginary part of the structure factor equation cancels However, we have seen that in the case of anomalous scattering, the scattering factor, fj, becomes the complex function fj + ~fj' + i8f(. The imaginary component of the scattering factor adds cross terms of the type ~f(sin2rr(hxj + kyj + lzj) to the structure factor equation. These cross terms are not the same for IFhkll 2 and I Flikl 1 2 so Friedel's law breaks down. It is this difference in Ihkl and Irn , the Bijvoet difference, which is used to measure the anomalous effect and provides the basis for anomalous phasing techniques. Although this anomalous effect usually is negligible, in the case of IccGG the average difference between IFhkll and IFrnl was 12.%. This was easily measured over a noise level of 5.%. Reflections from an anomalous dispersion data set can be divided into two groups, centric and acentric reflections. In P4 3 2 1 2, the structure factor equation for centric reflections can be reduced to: Fhkl = ~fj cos2rr(hxj + kyj + lzj) or Fhkl = ~fj isin2rr(hxj + kyj + lzj) (3.8) j j The first form contains only a cosine term which is invariant when changing hkl to hkL The second form contains only a sine term which does becomes negative when moving from hkl to hkl. However, this change is not seen in the intensity because it is masked by the conversion from structure factor to intensity in the same manner as equation (3.7). Therefore, centric reflections do not show Bijvoet differences. 3.4 Summary of Anomalous Scattering Cause and Effect The causes and effects of anomalous dispersion are summarized here. -38- • Occurs when the x-ray wavelength is near an absorption edge of the anomalous scatter • Caused by shift in phase and amplitude of the anomalously scattered wave relative to that of a normally scattered wave. • Usually a negligible effect but may be significant (ie. > 10%) • Affects the intensities of only acentric, not centric reflections • May be used in locating anomalous scattering atom in Patterson map. • May be used to calculate the phases necessary to create an electron density map. 3.5 Location of the Anomalous Scatterers As pointed out earlier, the location of all anomalous scatterers in the cell must be known in order to use the Bijvoet differences to phase a reflection. In the case of isomorphous replacement techniques, this is done using difference Patterson maps or by using phases calculated from one derivative to phase a second derivative's Fourier map. In the case of IccGG however, there is only one data set so no previous phases are available. Therefore, the iodines must be located by Patterson techniques. A Patterson map is made using the measured intensities of a crystal without the need for phase information. Peaks in this map correspond to vectors between atoms in the crystal. The heavier the peak, the greater electron density of the two atoms which make up that vector. The advantage of this map is that the phase of a given reflection need not be known. However, interpretation of such a map in terms of coordinates for individual atoms is often not straightforward. Peaks in a Patterson map are broader than those in the corresponcling electron density map. For this reason the Patterson map becomes more cluttered as the number of atoms - 39- increases, because the peaks begin to overlap one another. This overlap problem may become particularly misleading in cases where the molecule itself possesses high symmetry, as with DNA. Such symmetry increases the odds that several peaks will overlap one another and create the impression of a single peak of higher weight. For example, ma.ny small carbon-to-carbon peaks may pile up to simulate one iodine-to-iodine peak. Once again, anomalous differences can be of some help. Rossmann has demonstrated that a Patterson summation with (I Fhkll - IFm I) 2 coefficients gives peaks at the end of vectors relating the anomalous scatters in the structure.4 Theoretically, the major peaks in the anomalous Patterson map should correspond to iodine to iodine vectors since the other atoms do not scatter anomalously to an appreciable extent. These anomalous Patterson maps were made for 1CCGG. Most useful were the Harker sections. For P4 3 2 1 2, the relevant Harker sections are the sections u = v, u = .L w = ..L and w = ..L These sections are shown in Figure 2. 2' 4. 4' 2' M.G. Rossmann (1961). Acta. Cryst. 14, 383-388 -40- Figure 2. Anomalous Harker sections Since there are two iod.ines in the asymmetric unit, each one of these sections - 41- should have at least four peaks on it. Possible solutions for iodines A and B were tested by .two methods. To test a single iodine position, all of the self vectors, eg. A to A vectors, should appear on the Harker sections. When two possibilities were found, they were tested with each other by looking in the three dimensional Patterson map for appropriate cross vectors, eg. A to B vectors. The positions of the correct I-I peaks are marked on the Patterson as A or B. The fact that these positions are only rarely in the center of the large peaks resulted in a good deal of uncertainty in their assignment. Supposedly, the anomalous Patterson helps peak assignment by enhancing those peaks that result from anomalously scattering atoms. That is to say, if a peak is stronger in the anomalous Patterson than in the normal Patterson, it should be due to the anomalous scatterer, iodine. Conversly, if the peak is weaker in the anomalous Patterson than in the regular Patterson, it must be due to a pileup of vectors from atoms that scatter normally. These guidelines turned out to be false. The problem is that when the (I Fhkll - IFrnl )2 coefficients are formed for an anomalous Patterson, the centric data are effectively omitted, since they have no anomalous differences contributions . In a space group with many centric zones, low resolution data may contain a large percentage of centric reflections. In P4s212, the 0 3 A data set contains 36% centric reflections, and severe distortions occur in the anomalous Patterson map as a result of this omission. This loss of centric data may not be a problem unless the anomalous scatters are located near crystallographic symmetry elements of the cell . However, in the case of IccGG , the two iodines are positioned close to twofold axes in such a way that A, B, and X vectors (vectors between positions A and B) often point along cell axes . In a normal Patterson map the centric reflections are important in defining these positions. Since the centric reflections are missing from an anomalous Patterson, these peaks are not as well defined as they should be. Th e refore, there were some cases -42- where the I-I peaks was actually diminished in the anomalous Patterson map rather than enhanced. To distinguish such peaks, a total of three Patterson maps were used. • A normal Patterson map (ie.l Fhkll 2 as coefficients) • A normal Patterson map containing only acentric data (ie. only acentric IFhk11 2 as coefficients) • An anomalous Patterson map (ie. (I Fhkll - IFlil(] I)2 as coefficients) These Patterson maps are shown in Figures 1, 2, and 3. -43- CJ ~~----. OD w __CL.. - __ o_ - Figure 3. Harker sections from 4/mmm averaged acentric data The following algorithm was used to sort peaks from the Patterson maps: -44- 1. If a peak is higher in the anomalous map than in the normal map, it is considered a possible I-I vector. 2. If a peak is lower in the anomalous map than in the normal map, its height must be checked in the acentric map. A. If the peak is relatively the same height in normal and acentric maps, then its diminution in going from normal to anomalous maps indicates it is not an I-I vector. B. If the peak decreases in height in going from normal to acentric maps, then the diminution in going from normal to anomalous maps may only be caused by the loss of centric data. Therefore, this peak must also be considered a possible I-I vector. Using the vectors added by this procedure, a self consistent set of positions for the two iodines was found . One method used to check these two iodine positions was to use one position as a heavy atom in phasing the electron density map, and look for the other iodine position to appear . This was not a good test. Even when the correct positions were used, maps phased with only the first iodine did not show the second iodine well above the no ise level. The positions of the iodines was uncertain until the very end, when the correct anomalously phased electron density map was calculated. 3.6 Use of Anomalous Differences in Phasing The difference in intensity between Ihkl and Im. the Bijvoet difference, can help determine the phase of the hkl reflection. In macromolecular crystallography, several different methods are used . These techniques may use the anomalous differences by themselves or combine them with other phase information. As will be discussed later, a single anomalously scattering atom can reduce the phase determination problem for an acentric reflection to a choice between two pos sibilities. To -45- make this choice, several other sources of information can be used. In one example, 5 data were collected on a single crystal using two different x-ray wavelengths near the absorption edge of the anomalous scatterer. The way in which amplitude and phase of the anomalously scattered wave change with input wavelength allows phases to be calculated for the reflections. This method is particularly effective when a tunable x-ray source, such as a synchrotron, is used, so that any desired wavelength can be chosen. 6 Anomalous phasing is often combined with the method of isomorphous replacement to provide phasing information. Isomorphous replacement involves making a heavy atom derivative of a given crystal, usually by soaking the crystal in a solution containing the heavy atom. By studying the change in intensity between the native crystal and the derivative, phases for the reflections can be found. In practice, suitable derivative crystals are difficult to make and it is best to extract the most infermation from each one. When only a single isomorphous derivative exists, the phase problem can be reduced to choice between two possibilities. This choice is usually resolved by making a second isomorphous derivative. However, since these derivatives usually contain heavy atoms such as samarium, uranium or iodine, they often have significant anomalous scattering components. Bijvoet differences collected from these derivatives can resolve the twofold choice in single isomorphous phasing. 7 In effect, each isomorphous derivative becomes twice as useful. 8 If multiple isomorphous derivatives are used, the phase problem can be solved using only isomorphous data, without considering anomalous scattering. However, 5. W. Hoppe and U. Jakubowski (1975). in Anoma.lou.s Sca.ttering, edited by S. Ramaseshan and S. C. Abrahams, Copenhagen:Munksgaard, 437-461 6. J. C. Phillips, A. Wlodawer, J. M. Gooodfellow, K. D. Watenpaugh, L. C. Seeker, L. H . Jensen and K. 0 . Hodgson (1977). Acta. Cryst. A 33, 445-455 7. A. C. T. North ( 1965). Acta. Oryst. 18, 212-216 8. J. R. Knox, J. A. Kelly, P. C. Moews and N. S. Murthy (1976). J. Mol. Biol. 104, 865-875 -46- addition of anomalous data allows a more precise determination of the phase angle by adding more information, which narrows the range of errors. This additional information improves the quality of the electron density map and allows interpretation to proceed with greater ease and confidence. 9 3.7 Solution of the Phase ProbleJD. IccGG could not be solved by the isomorphous method because no two of the available crystals belonged to the same space group. A multiple wavelength analysis similar to that of Hoppe and Jakubowski 10 was briefly considered. However, the expense of a new x-ray tube, the difficulty in aligning the diffractometer, and the need to collect twice as much data all argued against this technique. However, the iodine is almost heavy enough to phase the reflections itself using standard heavy atom techniques. It was believed that the combination in the proper weighting scheme of heavy atom information and anomalous dispersion information would be sufficient to solve the structure. In fact, this method worked very well. The initial 3 AFourier map showed the DNA molecules clearly separated from the solvent background. Within the molecule itself, the base planes were resolved and the sugar phosphate backbone could be easily traced. The clarity of the initial map allowed the elimination of various left handed 11 12 and other 13 models which have proposed in recent years. A method to phase reflections based on Bijvoet pairs from a single crystal has already been worked out. 14 Once the positions of all the anomalous scatterers in the cell are known , the phase of a particular reflection can be calculated by the equa9. J. R. Herriot, L. C. Sieker, L. H. Jensen and W. Levenberg (1970). J. Mol. Biol. 50, 391-406 10. W. Hoppe and U. Jakubowski (1975). ibid 11. V. Sasisekharan, M. Bansal and G. Gupta (1981). Biochem. Biophys. Res. Com. 102. 10871095 12. G. Gupta, M. Bansal, V. Sasisekharan (1980). Proc. Na,t. Aca,d. Sci. USA 77. 6486-6490 13. R. C. Hopkins (1981). Science 211. 289-291 14. G. N. Ramachandran and S. Raman (1956). ibid -47- tion: (Please refer to Figure 4 for identification of components.) (3.9) where and • CXJuci is the phase for the scattering from all atoms in the cell and includes the real scattering component from iodine. • a:A is the phase of the scattering due to the normal dispersion from the iodine. • F"hkl is the magnitude of the anomalous scattering. Once a:hkl is known, it can be used in a Fourier synthesis with Fhkl• the amplitude of scattering due only to normal dispersion, to calculate the electron density map for the crystal. An example of phasing from 1CCGG, F 3 , 1 , 1 is shown in Figure 4. The two possible choices for phase occur where the two circles intersect. In order to minimize the mean square error in the electron density map, each of the two phases is included in the map with a weight proportional to the probability of its being the correct choice. 15 Different weighting schemes were used for the two types of data, centric and acentric. Since the centric data contained no anomalous differences, they were weighted by the heavy atom procedure of G. A. Sim 16 which assigns to the phase calculated by the heavy atom method a weight which depends on the probable error between the heavy atom phase and the true phase. The acentric data are phased by the Bijvoet differences, so they are weighted by G. A. Sim.'s method for anomalous phasing. 17 This scheme weights each of the two phase angles 15. M. M. Woolfson (1956). Acta Oryst. 9, 804-810 16. G. A. Sim (1969). Acta, Cryst. 13,511-512 17. G. A. Sim (1964). Acta Cryst. 17, 1072-1073 -48- Figure 4. Reflection 3,1, 1 phased by iodines at ( x, y, z ) in P4 3 2 1 2 based on the following factors. • The phase angle closest to the heavy atom phase angle receives the most weight. • Reflections with larger heavy atom contributions receive heavier weight. • Reflections with larger total intensity receive more weight. Figure 5 shows an example of phasing from IccGG where the two phase choices are close together. In this situation, Sim's weighting method assigns 50% probability to -49- Figure 5. Phasing diagram with 9 approx 0 °; F843 each solution and includes them both. Figure 6 shows an example of phasing from IccGG where the two phases are relatively far apart and the ambiguity of the phase is at its greatest. In this situation, Sim's weighting method assigns the phase closest to the heavy atom phase the greatest weight, 82%. 3.8 The rmal Choice There is one final choice which must be made when using a single crystal for anomalous phasing, Although a self consistent set of iodine positions was found in -50- Figure 6. Phasing diagram with® approx 90 a; F 211 the Patterson maps, there actually are two possible solutions. An atom at (x,y,z) in space group P4 3 2 1 2 produces a Patterson map identical to an atom at (x, y, z) in P4 1 2 1 2. When anomalous phasing methods are used, this becomes an important problem. Figure 4 illustrates the calculation of the· phase with what later turned out to be the correct choice, space group P4 3 2 1 2. Figure 7 shows the same reflection phased by iodines at ( x, y, z) in P4 1 2 1 2. The phase is considerably different . The only way to make this choice is to actually construct separate electron density maps for each of the possible solutions. The correct solution, P4 3 2 1 2. gave a very clear -51- ~II Figure 7. Reflection 3,1,1 phased by iodines at (X, y, z) in P4 1 2 1 2 picture of the tetramer in which even the base pairs could be resolved from one another. The other solution, space group P4 1 2 1 2, produced a map with no intelligible features. The decision for P43 2 1 2 was unambigious. -52- Chapter 4. Intermediate Results -53CHAPTER 4. INtERMEDIATE RESULTS Although only a few water molecules had been located after 28 cycles of refinement, most of the DNA had settled into place. At this point a structural analysis was performed and the results published in Natura. 18 1CCGG was compared with the fiber structure of A-DNA and some preliminary information on the hydration structure was presented. The remainder of this chapter is that paper. One set of useful illustrations which were not included in the paper are the stereo drawings in Figure 1. These show the sequence CCGG as determined by fiber diffraction. 19 This figure was not included in the paper because of space limitations. However, it is very useful for comparison with IccGG because it is drawn from the same viewpoint as Figure 1 of the Nature paper. The change in propeller twist can be clearly seen by comparing the side views of the two tetramers. 18. B. N. Conner, T. Takano, S. Tanaka, K. Itakura, and R. E. Dickerson (1982). Na;ture 295, 294-299 19. S. Arnott and D. W. L. Hukins (19?2). Biochem. Biophys. Res. Com.. 47, 1504-1509 -54- Figure 1. Fiber A-DNA from Arnott and Hukins (1972). Biochem. Biaphys. Res. Com.. 47, 1504-1509 -55294 Nature Vol. 295 28 January 1982 The molecular structure of d( 1CpCpGpG), a fragment of right-handed double helical A-DNA Benjamin N. Conner·, Tsunehlro Takano·, Shoji Tanaka·, Keiichi Itakurat & Richard E. Dickerson• Divaion of Chemistry and Chemical Engineering, California Institute of Technology, Pasadena, California 91125, USA t City of Hope National Medical Center, Duarte, California 91010, USA The DNA tetramer deCpCpGpG) or 1CCGG crystallizes as a double-stranded 4-base pair (bp) segment of an A helix. Two such tetramer helices are packed together in the crystal with local helix axes nearly coincident, simulating an 8-bp helix, and four such octamers make up the tetragonal unit cell. Restrained energy and reciprocal space refinement has led to an R factor of20.5% at2.1 A resolution. The 1CCGG helix has a twistco"esponding to 10.7 bppertum, a 19°base tilt and a 2.3 A rise per base pair along the helix axis. The mean propeller twist of 18° is comparable with, and has the same rotational sense as that observed in the B-DNA dodecamer CGCGAA TTCGCG at similarly high alcohol concentration. Backbone phosphate groups in A-DNA are extensively hydrated, including a network across the opening of the major groove, whereas base edge N and 0 groups in major and minor grooves are less hydrated than in B-DNA. The minor groove spine of hydration observed in B-DNA is totally absent. These observations of relative hydration confirm and extend the model for the B-to-A helix transition proposed earlier on the basis of the B helix structure. THE first double-helical DNA structures were proposed on the evidence of X-ray diffraction data from drawn fibres obtained from natural sources, supplemented by spectroscopic and chemical information, and model building from the known stereochemistry of component parts 1- 7 • Two major families of helix were proposed, A and B, differentiated by sugar puckering, base tilt relative to the helix axis, rise per residue along the axis, distance of base pairs from the centre of the helix and relative dimensions of major and minor grooves (Table 1). A-DNA is characterized by a deep and narrow major groove, a shallow minor groove and base pairs tilted by roughly 19" to the helix axis. In various members of the B family, the bases are more nearly perpendicular to the helix axis and the grooves are of comparable depth, although the major groove is wider. Fibre diffraction patterns can give the overall average helix parameters for DNA of known composition or repeating sequence, but cannot reveal local variations in helix structure that might be produced by that sequence. This kind of information can only come from single-crystal structure analyses of discrete, homogeneous DNA oligomers. The development of triester and related methods has made the synthesis of such oligomers possible'. Structure analyses have been carried out for both a 12-base pair (bp) B helix 9 • 10 and a totally unexpected new family, the left-handed Z helix 11 - 13 • We describe here the first single-crystal structure analysis of a member of the A-DNA family, the self-complementary tetramer d(iodo-CpCpGpG) or 1 CCGG. It is also the first example of a nucleic acid structure analysis carried out without isomorphous heavy atom derivatives, using initial phasing derived from anomalous scattering as was done for the proteins haemerythrin and cram bin (ref. 14 and W. A . Hendrickson and J. Smith, personal communication). (In another structure analysis of an RNA dimer, r(ApU), complexed with 9-aminoacridine, anomalous scattering was used to locate positions of phosphorus atoms, which then became the starting point for non-anomalous phase analysis 15 .) • Present addresses: Molecuiar Biology Institute, University of California at Los Angeles, Los Angeles, California 90024, USA (B.N.C., R.E.D.); Institute of Protein Research, Osaka University, Osaka, Japan (T.T.); Laboratory of Microbiology, Suntory Institute for Biomedical Research, Mishima-aun. Osaka 618, Japan (S.T.). 0028-0836/ 12/ 040294--06SO 1.00 Structure analysis To find an isomorphous pair of crystal forms, the native CCGG tetramer and four variants with 5-bromocytosine or 5-idocytosine in the first or second position along the chain were synthesized independently by the triester method. Four of these compounds were crystallized, but no two adopted. the sam.e space group. Having discussed with W. Hendnck.son h1s experiences with haemerythrin and crambin, we decided to try to solve the better crystal form of the two iodo derivatives, 1 CCGG, using anomalous scattering from iodine atoms. Crystals of the 1-iodo derivative were grown at 2 oc by vapour diffusion against isopropanol, from concentrated aqueous solution of 1CCGG containing spermine hydrochloride buffered at pH 7.5 with sodium cacodylate. The cacodylate buffer was later found to be unnecessary for crystal growth. Crystals began appearing at around 40% isopropanol , and crystal growth was continued to 80% alcohol . The space grout? of 1CCGG crystals is P4 3 2 12, with cell dimensions a= b = 41.1 A, c = 26.7 ~an~ two tetramer strands (one double helix) per asymmetric umt. A measured density of 1.50 g cm- 3 and a calculated molecular weight of 1,298 per strand indicate that the crystals are 51 .0% DNA by weight. Data were collected on a Syntex PT automatic diffractometer with graphite monochromator, and corrected for X-ray decay and absorption 16 • The pattern was quite strong to 3.5 A resolution, but declined rapidly beyond that point. One .anomal<:>us data set to 2.1 A resolution contained 1,486 Fnedel pa1rs. Multiple data sets were collected on different crystals for accuracy, so that each reflection within the 3.0 A sphere was measured at least three times. Mean R factors between independent data sets were - ~% in F. 11 Patterson and anomalous d1fference Patterson maps were employed to locate the two iodine positions per asymmetric unit (16 per cell), using the 472 pairs of reflections a.bove the 2u confidence level within the 3.0 A sphere. Approximately 36% of these reflections were centric and hence gave no anomalous contribution, but the iodine anomalous scattering factor of = 7.2 electrons for CuKa radiation produced an avera~e difference between Friedel pairs of 11% in F among acentnc reflections. The location of iodine atoms was complicated by two factors : the iodine of cytosine Cl was accidentally situated 11r <0 198 2 Macm1 llan Jo urnal• l td -56Nature Vol. 295 28 January 1982 295 Table 1 Structural parameten in double-helical DNA Fibre diffraction results Bfamily 1 Helix sense Supr pucker Glycosyl angle Twist per base pair Bases per tum Rise per bue pair (A) Base tilt Groove width (A) Minor Major Groove depth (A) Minor Major Refs Zfamily CCGG crystal structure Right A Right C-3 ' -~ndo C-3'-~ndo Anti 33.6" 10.7 2.3 19" Anti 32.7• 11 2.6 19" 36° 10 3.4 -6· -a• 45° 8 3.0 -16° 8.8 2.0 11.0 2.7 S.1 11.7 4.8 10.5 1.3 8.9 3.7 13.8 This work 2.8 13.5 4 7.5 8.5 4 7.9 7.5 4,5 6.7 5.8 4,5 B c C-2'-~ndo Anti 38.6° 9! 3.3 z D Right z· Left Left Alternating Syn(G )/anti{C) -60" / 2 -60" /2 12 12 3.7 3.8 -7° -9· 11 13 Groove width is the perpendicular separation of ideal helix strands drawn through the phosphate groups, decreased by 5.8 A to represent van der Waals radii of two phosphate groups. Groove depth is defined precisely in ref. 4, but again is based on van der Waals radii, as though measured from a space-filling model. directly above a 2-fold axis, with coordinates close to (l, t i>. and that on cytosine C5 of the other strand of the double helix was nearly just one-quarter of the cell away from it along the a axis. Each of these two coincidences introduced false centrosymmetry that decreased the iodine anomalous contribution to many of the acentric reflections. Thus the simple (.1F".no) 2 = 2 (Fhl<l- Fm ) anomalous difference Patterson map was not interpretable. Iodine positions were ultimately found by comparing F 2 Patterson maps using both centric and acentric reflections, with F 2 Patterson maps using only acentric data. As we suspected that our difficulties arose because the iodines sat on or near symmetry elements, those Patterson peaks which were deleted by the removal of centric data were reinforced in the simple anomalous difference Patterson map, and the composite map was then interpretable in terms of a self-consistent set of iodine-iodine vectors. Once the iodines were located, electron densitr maps were calculated using Sim-weighted heavy atom phases 1 for centric reflections, and Sim-weighted anomalous phases 19 · 20 for acentric reflections. Of the two possible enantiomorphic maps, one was uninterpretable, while the other gave a clear and clean image of the tetramer double helix. The helix obtained from this map was subjected to 28 crcles of simultaneous energy and reciprocal space refinemene , using the same energy constraints as for the B-ONA dodecamer 10 • Refinement progress was monitored periodically by Fourier and difference Fourier electron density maps. During this process, the R factor fell from 39% at 3.0 A resolution (2u data) to the current 20.5% at 2.1 A. The model contains 29 solvent molecules at present and refinement is continuing. Tetramer structure The complete unit cell contains eight identical double-helical tetramers, one of which is shown in Fig. 1. The tetramer is a fragment of an A-DNA double helix. A helix-generating program provided by J . Rosenberg enabled us to use the relationships between base pairs within one tetramer to define the parameters of an ideal continuous helix (see Fig. 1). The helix twist angle is 33.6° per base, corresponding to 10.7 bp per turn, and the rise per base pair along the the axis is 2.3 A. Bases are tilted by 19" from perpendicularity to the helix axis. Sugar rings are all in the C-3' -endo conformation with the exception of the 3' -terminal sugar of guanine G4, which is C-2' -endo. (See upper left of Fig. 1b, or lower left of Fig. 1c.) However, as the G4 sugar has no chain continuation to force it into a regular geometry, it is not surprising that it adopts an anomalous conformation. Table 1 compares observed CCGG helix parameters with idealized A-DNA values derived from fibre data and with those of other types of helix. The base pairs have large propeller twists: 19", 18°, 18" and 23° for Cl-G8, C2-G7, G3-C6 and G4-C5 respectively. The large propeller twist was visible early in the analysis of the anomalously phased electron density map, and hence is not a consequence of the refinement process. The last value is perturbed by the same intermolecular crystal packing that produces the tilted C5 plane visible in Fig. 1, but the other three values are quite consistent. They are larger than the 13° average propeller twist observed for B-ONA in the low-alcohol c!6stals (35% methylpentanediol, MPD) of CGCGAATICGCG 1 , and more like the 19° observed in the higher alcohol (60% MPD) crystals of CGCGAATI8 'CGCG (M. L. Kopka, A . V. Fratini, H. R. Drew and R.E.D., in preparation). The sense of the propeller rotation in A-helical CCGG is the same as that in these B helices: clockwise rotation of the nearer base plane when the base pair is viewed along its long axis, defined here as the positive rotational sense for propeller twist. In contrast, the published coordinates for a model A-DNA structure derived from fibre diffraction data 22 yield an opposite rotation, with a propeller twist angle of -12". (A recent unpublished set of new fibre-derived A-DNA coordinates supplied by S. Amott has positive propeller twist angles of + 7.5°.) Two RNA dimers are known which are fragments of an A helix : r(ApU) and r(GpC) 23 ·24 • In these molecules the propeller twist is smaller and of variable sign: + 12.5° (average ) for r(ApU) and -7.2° for r(GpC) . It seems likely that a substantial propeller twist is a consequence of efficient base stacking at successive steps along one strand of a continuous double helix, and hence reasonable that the full twist typical of an infinite helix would not be generated in helices shorter than 3 or 4 bp. Indeed, one unexpected feature apparent from the single-crystal analyses of oligomers of A- and B-ONA has been the large propeller twist angles, which were not anticipated from fibre and dimer studies. A-helical octamers in the crystal The most apparent distortion of the tetramer in Fig. 1 is the pushing of the iodine on cytosine C5 towards the centre of the molecule, with a resultant bending of the C5 plane away from that of G4. This is simply explained in terms of the way in which tetramers are packed within the crystal. Two tetramers are stacked with their Cl-G8 base planes in contact, their helix axes nearly coincident and a rotation between tetramers that closely approximates a continuous 8-bp A helix (Fig. 2). Four of these octamers are then packed around the 4-fold screw axes of the -57196 Natur~ Vol. 295 28 January 1982 f'la. 1 Stereo views of the 1CCGG double-helical A-DNA molecule . a, View into the major groove with helix axis vertical. Bases are numbered: 5'-Cl-C2-G3-G4-3' in one strand and S'-CS-C6 -G1-G8-3' along the other. The two largest atoms are iodines on cytosines Cl (lower right) and CS (upper left), and other atoms in order of decreasing size are P, 0, N and C. Hydrogen bonds between base pairs are shown as thin lines. Base pair Cl-G8 is at the bottom and G4-CS at the top. b, View 90" to the left of a, showing the shallow minor groove in profile at left, and the large propeller twist of base planes. c, Top view looking directly down the helix axis onto base pair G4-CS . The iodocytosine base planes are tilted away from the planes of their hydrogen-bonded partners, CS more so than Cl, because of intermolecular contacts within the crystal. f'la. 2 Two 1CCGG tetramers stacked around a 2-fold axis (normal to the page through the centre of the drawing) as they are observed in the crystal. Their Cl-G8 base pairs are in contact, and have very nearly the correct displacement to continue the A helix for 8 bp. Local helix axes, as determined independently for the two tetramers, are shown by vertical lines, and are almost coincident. This view is directly into the major groove, lined by phosphates 6, 7 and 8 (numbered on one strand). After the missing phosphate between tetramers, a continuous helix would use phosphates 2, 3 and 4 around the back of the drawing on the other tetramer. Seven water molecules and their symmetry-related p:u-tners are shown spanning the major groove (9). Distances of .. 3.4 A are shown by thin solid lines, and distances up to 3.8 A by dotted lines. As discussed in the text, in a continuous A-DNA helix the major groove would be narrower than shown bere, and the bridging network of hydration would be different in detail although similar in character. -58Nature Vol. 295 28 January 1982 P4)2.2 cell. The iodocytosine ring CS at each end of an octamer (top and bottom of Fig. 2) is packed against the minor groove of a neighbouring octamer, between bases C2 and G3. Although the CS iodine atom is pushed back towards the centre of the octamer, it still is in tight van der Waals contact with the C-2' sugar atom of cytosine C2 and the C-5' atom of guanine G3 on a neighbouring octamer. At the centre of each octamer (deep inside the major groove in Fig. 2), the iodine of cytosine Cl on one tetramer is displaced slightly because of a close contact with the five-membered ring of G8 on the other tetramer (the next base step along the helix), but the effect on its cytosine orientation is small. The two local helix axes as viewed in Fig. 2 are displaced sideways by 1.5 A, which serves to lessen the overlap between each iodine and its ~acing G8 ring. The extreme top and bottom of the two axes in Fig. 2 are tipped down into the plane of the page so the axes make an angle of 8° to one another, which also lessens the iodine/guanine clash. The absence of phosphate groups to connect the tetramers allows a degree of adjustment that might not be present in a true double-stranded octamer. Four octamers as shown are packed into the tetragonal unit cell of the crystal in helical arrays around water channels with elliptical cross-sections, parallel to the z axis. Each octamer uses its relatively flat minor groove side to define a long side of one elliptical channel, and its more concave major groove to make the sharp bend at the short side of a different elliptical channel. The individual octamers are hydrated as described below, and Fla· 3 PLUTO space-filling drawings of three 1CCGG tetramers stacked atop one another along their helix axes to form a complete tum of A-DNA helix. In the actual octamer found in the crystal (Fig. 2), one tetramer is inverted before being stacked on the other, with equivalent base pairs in contact. In contrast, in this simulation of a complete turn of helix, all three tetramers are stacked upright in the sense of Fig. 1, with base pair G4-CS at the top and ClG8 at the bottom. No attempt has been made to idealize or regularize individual molecules, or to fill in the missing phosphate groups that would connect tetramers in a continuous helix ; each tetramer is exactly as observed in the crystal. a, Side view looking into the cavernous major groove at the top, with the shallow minor groove in profile at right. b, Top view down the helix axis, showing the ·spiral staircase' arrangement of base pairs and the hollow centre. 297 networks of loealized solvent molecules extend across the channels. The idealized extended A helix The octamer in Fig. 2 does not provide a true impression of a continuous A-DNA .helix with this helix geometry . Figure 3 shows two views of a full turn of helix, constructed by stacking three identical tetramers on top of one another with their local helix axes coincident. The very deep major groove is particularly striking in stereo (Fig. 3a ). The groove is narrower in the continuous helix (Fig. 3) than in the crystal structure octamer (Fig. 2) because of the 8° angle between local helix axes in the octamer. If this angle were straightened out to 0°, then the major groove in the octamer would appear as in Fig. 3a, with phosphorus-phosphorus separations across the groove of 7 .8 A rather than 10.9 A, the P6-P6 separation in Fig. 2. This factor will become important in the following discussion of hydration. The shallow minor groove is hardly a groove at all in comparison with its structure in B-ONA. Figure 3b shows the hollow channel down the centre of the helix, a feature peculiar to A-DNA. In a sense the A helix is more a twisted, double-edged ribbon than a cylinder of stacked base pairs . The major groove is only one-quarter as wide as the minor one , but is four times as deep (Table 1). Hydration and contrast with B-DNA The A-DNA structure itself has been completely refined ; -59298 remaining refinement will be concerned principally with adding more solvent peaks until the difference electron density map has no significant residual features. The 29 solvent peaks added so far give a good impression of the hydration state of the A helix, s.i nce most of the peaks that have yet to be incorporated lie in the open solvent channels between molecul~. A detailed analysis of intermolecular water networks, and comparison with other results (refs 25, 26 and M. L. Kopka, A. V . Fratini, H . R . Drew and R.E.D., in preparation) will be described elsewhere (B.N.C. and R.E.D., in preparation). The helical octamers are long enough to establish true major and minor groove solvent environments, even though the chemically unique unit is only four bases long. The hydration state of A-DNA in CCGG is significantly different from that of B-ONA in CGCGAATICGCG. In the 35% MPD crystals of B-DNA 26 , all theN and 0 groups on the edges of base pairs in the major groove are hydrated by a monolayer of solvent molecules, but the bulk water within the major groove appears disordered. Similarly, except for three examples where a water molecule is trapped in a clathrate-like manner between a phosphate group and a thymine methyl, hydration of B-ONA along the phosphate backbone also is unlocalized, without clearly defined peaks. But taking the BONA dodecamer to the higher alcohol concentration of 60% MPD produces a more extensive second- and third-shell hydration within the major groove, and a more extensive and ordered hydration along the phosphate backbone (M. L. Kopka, A. V. Fratini, H. R. Drew and R.E.D., in preparation). Almost every free phosphate oxygen has a solvent ligand, and another solvent molecule frequently bridges the two oxygens. The most striking and unvarying aspect of B-ONA hydration, however, is a zigzag spine of water molecules running down the narrow minor groove, bridging purine N-3 and pyrimidine 0-2 atoms on adjacent base pair steps, where not interrupted by N-2 amino groups on guanines. This feature is identical in low and high MPD crystals, though slightly more extensive in the latter. lt has been proposed that the breakup of this minor groove spine of hydration is the crucial first step in the cooperative transition from the B to the A form 26 •2'. With respect to the phosphate backbone, CCGG in 80% isopropanol resembles CGCGAA1T 8 'CGCG in 60% MPD: a chain of localized water molecules follows each tetramer backbone strand, bonding to the two unesterified oxygens on each phosphorus. Eighteen well defined solvent peaks are associated with the phosphate backbone, and many of these first hydration shell ligands are themselves bridged by other solvent molecules (see Fig. 4 ). In the view of the octamer in Fig. 2, a network of six solvent molecules (and their six symmetry-related partners) arches across the opening of the major groove, connecting phosphates P-6 and P-7 on opposite sides. This octamer is so short that the two edges of the major groove are in opposition only at P-6 and P-7 . In a continuous A helix, such as in Fig. 3, one would expect the opening of the major groove to be crossed by a continuous network of solvent molecules, especially as the actual A-DNA groove would be 3 A narrower than that in the octamer of the crystal structure. This network of solvent molecules stitching together the two edges of the major groove may be a distinctive feature of the A helix. In contrast, fewer ordered water molecules seem to be located deep in the major groove of A-DNA, or in the very shallow minor groove, even though the major groove is more accessible in this crystal structure analysis than it would be in continuous A-DNA. Only one-third of theN and 0 atoms on base edges have clearly defined full-occupancy ligands, five in the major groove and three in the minor one. Another one-quarter has small nearby peaks that may represent partial occupancy sites (three each in major and minor), but hydration within the grooves is less impressive than in either of the grooves of B-ONA, even at comparable alcohol.concentrations. Water structure and the B-to-A transition In summary, in A-DNA there is no trace of the minor groove Nature Vol. 295 28 January 1982 f'la. 4 Schematic drawing of hydration in the immediate vicinity of the CCGG double helix in the crystal. The four central ovals represent purinepyrimidine base pain identified by base numben. as 'G4 / C5 ' at the top. Base rina nitrogen and oxyaen atoms are shown by N3 . N~ . 02 and so on around the perimeter of each oval. The upper edge of each oval faces the minor groove, and the lower edge faces the major. The two sugar- phosphate chains run vertically at either side of the base pair ovals, with C -3'. C-4' and C-5 ' atoms indicated merely by 3'. 4' and 5'. (Atoms C-1'. C-2' and 0-1 ' are omined for clarity.) Phosphates are numbered P2-P4 and P6-P8 accordina to the base to which they arc linked via C-5' atoms. Sohd black oxygen atoms on phosphates are those facing the major groove . Large numbers ide ntify the most prominent solvent molecules in the immediate vicrn it)' of the double helix. x Indicates that a peak was specifically looked for in the electron density map but not found . y,-y. Are lesser peaks not included in the original hydration search, but found in the map when specifically examined for base edae coordination. - Indicates that base edge atoms are blocked by neiahbourina DNA substituents, and henao unavailable for hydratoon . The reaion of the major &roove opening depicted rn Fia. 2 appean at the upper riaht edae . A chain of five solvent molecules (24- 10-34 -11-36o runs diagonally downwards to the right from phosphate P-6 in Fig. 2. with the parallel symmetry-related chain (36- 11 -34-10-24) just belo"'· it. Solvent molecules 19 and 20 in two symmetry-related pain lie just below and above these two strands in Fig. 2. The solvent network shown at the left of the diagram is necessarily incomplete and dostoncd because a three -dimensional network cannot easily be flattened into a planar doagram . Solvent atom 32 at the lower left, for example. is pan of a network connecting phosphate P-2 with phosphate P-• on a different double helix (upper left ). It also is coordinated directly to major groove atoms N-7 and 0-6 on base 03 of the same helix. However, the figure does provide an accurate depiction of the tint hydration shell of the hehx. spine of hydration that is so prominent in B-ONA, nor does there appear to be as uniform a monolayer of ordered solvent lining the major groove. The extensive solvation of phosphate backbone observed in B-ONA at 60 % MPD reappears in A-DNA, where it seems to lace together the two edges of the major groove . These observations permit the ideas proposed in ref. 26 on the 8-to-A helix transition to be carried one step further. The backbone phosphate groups are more polar, and have a greater attraction for water molecules, than do the N and 0 groups on the edges of base pairs within the grooves. In conditions where the activity coefficient of water is high (conditions favouring B-ONA), all these polar groups should be solvated ; but when the water activity is decreased, either by dehydration or by addition of alcohol, then the groove Nand 0 groups should lose their solvation shells before the phosphates do. When the spine of hydration in the minor groove is disrupted, the integrity of the 8 structure is destroyed and the helix slips into the A form , in which the still hydrated phosphate groups participate in a network of solvation that laces together the two edges of the now quite narrow major groove. The phosphate-associated lacing across the major groove in A-DNA may have a stabilizing role comparable with that of the base-associated spine within the minor groove in B-ONA. When the water activity around an A helix is increased again, rehydration of base edge N and 0 groups, including re-establishment of the minor groove spine , -60Natur~ 299 Vol. 295 28 January 1982 shifts the equilibrium in favour of B-ONA, because the number of molecules solvating the phosphate backbone is not altered, but only their disposition. Since submission of this manuscript, 0. Kennard and coworkers at Cambridge have solved the structure of another A-DNA helix, of sequence: GGTATACC (personal communication). We thank A1 Fratini, Mary Kopka and Horace Drew for permission to quote their dodecamer results in 60% MPD before publication, and Lillian Casler for assistance with the figures. This work was supported by NIH grant GM-12121 and NSF grant PCM79-13959. B.N.C. was the recipient of an NIH predoctoral traineeship. This is contribution no. 6482 from the Division of Chemistry and Chemical Engineering. Received 27 July; O<X>e1>ted 27 October 1981. I. Watoon. J D . cl< Criclr., F. H. C. N11r•r< 171, 737- 738 (1 9S3 1. 2. Lan,nd~ . R .. Wilson, H. R., Hooper. C. W., Wilkins. M. H. F. It Hamilton, L. D./. mol«. Brol. l. 19-37 (1 9601. 3 Fulle r, W .. W illd ""' M. H . F .. Wilson . H . R.. . Hamil ton, L. D . 4c. Amot1 . S . J. lft.Ol« . s ;ol. 11. 60-40 11 96Sl. 4. Amott, S. in Qrrt~n iz arion 11nd Eip<nsion of Q,omosom<~ Ieels Allfre y, V. G ., Bautz, E . K. F .. McCarthy. B. J., Schimke. R. T. cl< Tissiereo. A. I 209-222 !Dahlem Konlerenzcn , Berlin, 1976). S. Amott, S. It Sclli na. E. / . mol.c. Brol. ,._ 26S- 269 11 97SI. 6. M..,.,n, D . A ., Spencer, M.. Wilkins, M. H. F. A Hamilton, L. D./. mol«. Biol. l. S47- S6S (1 96 11. 7 . Davia, D. R. It Baldwin. R. L. / . m oloc. Bioi. 6, 2S I-2SS 11 9631. 8. ltak ura, K. It Rrus. A. D . Sc~•u lOt, I40 1-1 40S 11 980). 9 . Wina. R. M. ., a/. Natwr< %11, 7SS-7S8 tl 980). 10 . Drew, H. R . ., a/. Proc. ""'" · A cad. Scr. U.S.A . 11, 2179-2 183 11 981 1. II. Wana- A. H -J . ., a/. Nor•,.11l.680-686 (1 979 ). 12. Amott, S.. Cbandruckara.n. R., Birdsal ~ D . L.. Lalie. A. G . W. It RaUill, R. L. NatuiYl&l, 743-74S !1 9801. 13. Drew, H., Takano, T ., Tanaka. 5., llllkura. K. It Dickenon. R. E . .'laru,. :116, ~ 67 -S73 (1 980). 14. Hendrickaon. W. A. It Teeter. M. M. Na ru,.lM. 107- 11311 98 1l. IS . Seeman, J'ol. C .. Day, R. 0 . I< lllch, A. Nor.,. 153, 324-326 (1 97SI. 16. Nonh, A C . T .. Phi1 1ips, 0 . C. 4t ,_,bthe.,.., F. S Artd' cn·s111IM, A U 3.JJ-JJP ' ' 068 ) 17. Rcmman. M. G . Aero cr,·srallorr. 14, 38 3-388 t 196 11. 18. Sirn, G . A . Acta crysu.tlorr. 13.5 11 - 512 (19601 19. Sim. G. A . A c<D cr.·srallorr. 11, 1072-1 073 (19641. 20. Ramachandran, G . N cl< Raman, S. Cu,., Sc1. 11. 348-3S111 9S61. 21. Jack, A. It Levitt, M. Acra crystal/oar. AJ-4, 931-93S (1 976). 22. Amott, S. 4t Hukins, D . W. L. Broc~''"· biopil ys. Rl!. Commun 47. IS04-1~09 • 19 72 1. 23 . Seeman, N. C .. Rooenbcra. J. M.• Suddath, F. L., Kim. J. J. P. cl< Rtch. A. / . mokc. B rol. 104. 109- 144 11 976). 24. Rooenbcra. J. M.. Seeman. N. C .. Day, R. 0 . I< Rich, A . / . motu . Bioi. 104. I4S-1 67 119761. 2S . NeiCiel, S.. Berman, H . M. cl< Shieh. H. S. Narur. 211. 129- 133 (1 9801. 26. Drew. H. R. I< Dic.k cnon , R. E. / . motu. Brol. 151. SJS-S~6 11 98 11. 27. Dickenon. R . E .. Drew, H . R . &. Conner, 8 . N. 1n B1omoftcu/ar Sl,tod wtam •n Vol. 1 1ed. Sarma, R. H.l l - 34 (Ad<:ninc, New York ,_1 981 1. - 61- Chapter 5. Subsequent Refinement and Hydration -62. CH.APrER 5. SUBSEQUENT REFINEMENT AND HYDRATION 5.1 Results of Subsequent Refinement The helix obtained from the anomalously phased electron density map was subjected to 48 cycles of simultaneous energy and reciprocal space refinement, 20 using energy constraints similar to those employed for the B-DNA dodecamer. 21 For the first 45 cycles of refinement, an averaged structure factors defined as , F average = ~ (Fhkl+ Frn) were used for refinement. During the last three cycles, F normal• the sum of the normal scattering from all atoms in the cell, 22 was used for refinement. This quantity may be calculated as (5.1) This change in data sets resulted in only a 0.3% change in R factor since the R factor between the F average and the F normal was less than 1.%. The path of refinement can be followed on the R factor curves in Figure 1. During the refinement process the R factor for 2cr data fell from 39.% at 3.0 A resolution to the current 16.5.% for reflections between 8A and 2.1 Aresolution. The R factor for all data between 8A and 2 .1 Aresolution is 19.9.%. The asymmetric unit contains two tetramer strands (one double helix) with 86 water molecules positioned in the final map. The RMS error in atomic position, as estimated by the method of Luzzati, 23 is 0 .25 Aas shown in Figure 2. Structure factors and atomic coordinates are available from the Brookhaven Protein Data Bank. 24 The atomic coordinates from the final cycle of refinement may be found in the Appendix of this thesis. 20 . A. Jack and M. Levitt (1976). Acta Cryst. A34, 931-935 21. H. R. Drew, R. M. Wing, T. Takano, C. Broka, S. Tanaka, K. Itakura and R. E. Dickerson (1981). Proc. Nat. Acad. Sci. U.S.A. ?8, 2179-2183 22. G. N. Ramachandran and S. Raman (1956). Ourr. Sci. 11, 348-351 23. V. Luzzati (1952). Acta, Cryst. 5, 802-810 24. F. C. Berstein, Chemistry Department, Brookhaven National Laboratory, Upton, New York 11973 - 63- 1 CCGG Refinement Figure 1. Crystallographic R factor as a function of resolution and refinement cycle. 5.2 The Structure of IccGG The helix parameters of 1CCGG shown in Table 1 were calculated with a program provided by John Rosenberg. In a short piece of DNA there is some ambiguity as to what constitutes the 'best" helix. Rosenberg's program calculates helical parameters by determining the best helix to move a given vector from one position to its helix related position. The values in Table 1 were calculated using a vector that pointed from the sugar C1' to the base nitrogen, N9 or Nl. The parameters may vary slightly depending on the set of atoms used to define the vector. The helix parameters of 1CCGG show a regular helix of A-DNA with a rise /residue of 2.2 Aand 10.6 base pairs/turn. All sugars, except the terminal G4 sugar, adopt the C3'-endo conformation. Torsion angles of the refined molecule are shown in Table 2 . Although these torsion angle values are clumped together in relatively tight groups, they do differ significantly from the fiber A-DNA values 25 shown at the -64- Figure 2. Luzzati estimate of RMS error in atomic position for IccGG. bottom of the table. As shown in the cell packing diagram of Figure 3, , the IccGG tetramer fits into the P4 3 2 1 2 unit cell with two single strands (one double helix) as the asymmetric unit. The crystallographic 4 3 axis passes near the G4 sugar and causes the bottom of G4/ C5 base pair to fit into the fiat minor groove of the adjacent helix in a manner similar to that described by Wang et al. for GGCCGGCC. 26 A better view of the close contacts caused by packing the bottom and minor groove side of residues G4, C5, and C6 into the minor groove side of C2, G3, G4, G7, and GB on the 4 3 related molecule is shown in Figure 4. However, none of these contacts seem to be ionic or hydrogen bonding interactions. In this area, the molecules must be held together by van der Waals interactions. 25. S . Arnott and D. W. L. Hukins (1972). Biochem. Biophys. Res. Com. 47, 1504-1509 26 . A. H.-J. Wang, S. Fujii, J. van Boom and A. Rich ( 1982). submitted - Proc. Nat. Acad. Sci. USA -65Helix Parameters of 1CCGG Helix Step bp/turn dee;/turn Single Step 2 to 1 3 to 2 4 to 3 Ave. Sini!le Steo 9.9 rise/res 10.1 10.6 36.2° 30.6° 35.7° 34.0° 3.2A 2.9A 3.3A. 2.2A. Double Steps 3 to 1 4 to 2 Ave. Double Step 10.8 11.0 10 .6 33.3° 32.7° 34.1° 3.oA. 3.1A. 2.5A. First to Last 10.8 33.4° 3.2A. Octamer Parameters Ave. Sine:le Step Step Btwn Helices 10.4 9.9 34.6° 36.2° 2.8A 2.7A. 11.8 TABLE 1. Helix parameters from IccGG Near the 4 3 axis, the crystal packing is very tight and some distortions of the helix occur. The x-o plot in Figure 5 shows that while seven of the sugars have the 3'endo conformation, residue G4, near the 4 3 axis, has the 2'-endo conformation. This is not significant, however, since G4 has no continuation of the helix to keep it in the regular C3'-endo conformation. Also, residue C5 has been shoved underneath residue C6. Figure 6 shows that the helix axis is displaced toward the cytosines . -66- Residue 1 1c 2C 3G 4G 5 1c 6C 7G BG Ap BF a -70 -73 -72 -68 -80 -75 -90 -41 Torsion An' les of 1CCGG (3 0 E "Y 60 76 -157 174 57 77 -151 166 70 73 -170 -174 79 150 143 90 -168 -170 51 80 -162 -178 72 87 -160 -176 80 55 -149 47 83 -175 136 38 139 -133 t ){ -69 -65 -72 -169 -162 -171 -127 -150 -159 -58 -68 -67 -164 .. 45 -157 -152 -154 -102 TABLE 2. Torsion angles of iodo-CCGG This is caused by the G4/C5 base pair being pushed underneath the G3/C6 base pair. The movement of C5 underneath C6 causes some shift in the phosphate and thus an increase in the propeller twist of the G3/C6 pair. The base pair propeller twists are 15°,12°,22°, and 17° for C1/G8, C2/G7, G3/C6 andG4/C5 respectively. At the other end of this double helix is a crystallographic twofold axis which is positioned perpendicular to the helix axis in such a way as to stack two tetramers on top of each other to make a continuous octamer of A-DNA. This interaction can be seen in the cell packing diagram of Figure 3, but a clearer view is shown in Figure 7. The twist between base pair C 1 I GB and its twofold related base pair is 36. zo with a rise/residue of 2.7 A This is comparable with the helix rotations for other steps shown in Table 1. The smoothness of the interface between the two tetramers in the octamer may a lso be seen by looking at the torsion angles in Table 2. The torsion angles -y of residue C 1 and 6 of residue G8 near the twofold axis are comparable to those belonging to base pairs C2/G7 and G3/C6. However, the torsion angles 6 of G4 and-y of C5 at the other end of the molecule, near the tightly packed 4-s axis, take values different from the angles of the other base pairs in the tetramer. Of course, this situation emphasizes the distortion of the helix around the 4 3 end of IccGG but also shows that the interface at the 2 1 end of the helix is very smooth. In fact, these -67- t t t+ 00 c9+ i+o0 0+ . ' -. ' !11' I!_ "\ !y f I z- t+05 l+c9 oO+ ' 4' / t t I ' l- 00+ ·--' IJ/ 4 f t /" t '+ ' ' /• ' ~ --I /~ • I I •"' I . ' ' -I ?4' J i/ ~ !/ I A• I I " Figure 3. Crystal packing in rccGG. two stacked tetramers may be considered to be one octamer with single phosphate atom per chain missing between them. 5.3 lfater Around lcCGG \ The electron. density maps show IccGG surrounded by water molecules. Figure 8 shows the 47 solvent molecules within 3.5 A of the tetramer. - 68- Figure 4. Interactions about the 4 3 axis of IccGG. Distances closer than 5A are marked with enlarged atoms. Three of these molecules, Wl9, W31, and WB6, appear in the figure in more than one place because of the symmetry of the unit cell. Such waters are marked with an apostrophe the second time they appear, e.g. W19 becomes W19'. First layer water contacts and distances are shown in Tables 3 and 4. The distance of 3 .5A was chosen for hydration analysis because it represents the outer bound of hydrogen bonding distance. lt is probably fair to say that water at distances greater than 3.5 A represents an interaction but not a true hydrogen bond. Water interactions have been studied in some of the earlier high resolution studies: the RNA sequences ApApA, 27 and GpC, 28 and the DNA sequence pTpT .29 Since these structures all have resolution limits near lA and crystallographic R factors 27. D. Suck, P. C. Manor and W. Saenger (1976). Acta Cryst. B 32, 1727-1737 28. B. Hingerty, E. Subramanian, S. D. Stellman, T. Sato, S. B. Broyde and R. Landgridge Acta Cryst. B 32, 2998-3013 29. N. Camerman, J. K. Fawcett and A. Camerman (1976). J. Mol. Biol. 107, 601-621 - 69- -80° -90° -100o -110° -120° Xo -130° -140° -150° -160° (T) -170° -180° 60° 80° 1000 120° 140° 160° 8 Figure 5. Plot of backbone torsional angles x vs. o showing the clustering of sugar values about the C3'-endo conformation. near or below 10%, their solvent to DNA distances are very accurate. They have water to water and water to phosphate oxygen distances near 2 .8A, but water to 03' or 05' distances of 3 .0A. Taking the number of 3 .0A as the longest average bond distance and adding twice the Luzzati estimated RMS error in atomic distance, gives the 3.5A limit used in this study. When comparing the major and minor groove views of the hydrated tetramer in Figure 8, one immediately sees that the major groove and sugar phosphate backbone are heavily hydrated while the minor groove is almost dry. One reason for the dryness of the minor groove is the crystal packing about the 4 3 axis that was described earlier. The bottom base pair of one molecule effectively fills most of the minor groove of its 4 3 related molecule. - 70- Even so, that part of the minor groove which is not covered by the symmetry related molecule still has only a small number of waters in it. All of the waters in the minor groove are attached to guanine 4 or to oxygens in the sugar phosphate backbone. In IccGG, cytosines 1 and 6 are not blocked by symmetry related helices, and yet still have no hydration. The minor groove G7 and GB is blocked by the 4 5 related molecule, but G3 and G4 are free. One water in the minor groove near G3 is pulled away from the base by an interaction with the 4s related molecule. Around G4 are several water molecules which attach to N2 and N3 of the base. Several waters also are attached to the sugar phosphate backbone in the minor groove. However, these waters are usually stabilized by attachments to other waters or to a crystallographically related double helix. In general. those parts of the sugar phosphate backbone accessible from the minor groove are not heavily hydrated. In contrast to the minor groove, the 12 waters in the major groove occupy every possible hydrogen bonding site. This major groove hydration pattern has approximate twofold symmetry reflecting the approximate twofold symmetry of the DNA itself. The overall impression is that of a meshwork of water molecules stretched over the form of the A-DNA to cover the surface of the major groove entirely with water. The use of 5-iodo-cytosine in place of cytosine in this crystal resulted in two slight alterations of the hydration pattern. First, each iodine has one water attached to it about 3.4A directly along a line drawn from the base through the iodine. This distance is close to the normal Van der Waals iodine to oxygen distance of 3.55A. However, the water,s alignment with the iodine indicates that some interaction is taking place. This may be a charge transfer bond involving donation of the lone pair electrons of the oxygen to the empty 5d orbitals on the iodine. Such an interaction has been reported in the high resolution crystal structure of 5-iodo-2,-deoxyuridine. 30 30. N. Camerman and J. Trotter (1965) . Acta Cryst. 19, 203-211 - 71- Second, the iodine also serves to prevent water from binding to the cytosine 4-amino group. Those cytosines without iodines, cytosine 2 and 6, have waters attached to their 4-amino groups. Both the N7 and the 06 of ·guanine are hydrated, although the positioning of the waters is not the same in all cases. One particularly interesting case is that of guanine 4, where both water 31 and water 33 bridge to other moieties in the helix. Water 31 connects 06 of guanine 4 with 06 of guanine 3, above it. Water 33 links together the guanine 06 with one of the phosphate oxygens. It should be emphasized that this guanine is not in the typical A-DNA conformation, but more in the B form. Thus, the positioning of the backbone which allows a single water to link a base with a phosphate oxygen is not found on any of the other guanines. The backbone is the most heavily hydrated portion of the structure. It contains 30 first layer waters of hydration, in contrast to the 12 waters found in the major groove and the 7 waters of the minor groove. Since the 01' oxygens in the deoxyribose rings are only accessible from the minor groove, they have no major groove waters. However, there are minor groove waters attached to two of them. The 03' and 05' oxygens are accessible from the major groove or from the outside of the helix. Several waters are close to these oxygens, but in most cases this proximity is probably due to simple backbone geometry rather than to hydrogen bonding. One example of this may be seen in water 37, which bridges between the charged phosphate oxygens on cytosine 2 and those on guanine 3. Here the backbone geometry pushes the sugar 05' toward water 37, which is firmly held in place by three hydrogen bonds . Two of these hydrogen bonds are to charged phosphate oxygens and the third is to a guanine bound water. In the middle of the chain, 05' contains no hydrogen bonds except those characterized above as being due to backbone geometry. - 72On the other hand, there are some waters which do appear to be bound primarily to 03' or 05'. All of the free ends, for example, contain at least one and usually two bound waters. The 03' oxygen often has attached waters which are usually toward the minor groove side of the helix. Along the major groove side of the backbone of residues 1 through 4 is a chain of waters containing 105'-W14-20M-W37-30M-W69-W33-40M. (Phosphate oxygen OM is on the major groove side of the backbone while oxygen Om is on the minor groove side .) Such a chain is not found on the other side of the helix. It is possible, though probably somewhat meaningless, to continue analysis of the water structure for many levels. However, it is instructive to consider the second layer of hydration. I have defined a "second layer water" as water which touches at least two other first layer waters. This distinguishes them from those which touch only a single first layer water and are therefore probably less tightly bound. The twenty waters involved in this layer are shown in Figure 9 (along with the first layer, of course). - 73- The contacts and distances of these waters are listed in Table 5. Comparison of Figures 8 and 9, which show first and second layers of hydration respectively, illustrates how the second layer begins to knit together some of the more isolated sections of the first layer. A good example is W45 which links together W60 on Guanine 7, W58 on Guanine 8 and W20 attached to a phosphate oxygen of residue 7. A similar case is found in W4B which links together waters from cytosine 5, cytosine 6 and phosphate oxygens from residue 6. These two examples show how the water layers onto the DNA: first by attaching directly to the charged and polar atoms of the DNA and then by filling in between appropriately placed first layer waters. In summary it can be said that no major water structure, such as the "spine" in the minor groove of AT rich B-DNA, 31 is found in this structure of A-DNA. Instead, water binds in a more haphazard fashion which appears to cover the molecule rather than shape it. All three types of water bridges appear: base to base, phosphate to phosphate, and base to phosphate. However, none of these forms a systematic network. The overall impression of hydration for this fragment of A-DNA is fortuitous hydration, heavy along the backbone and in the major groove, but light in the minor groove. 31. H . R. Drew and R. E. Dickerson (1981). J. Mol. Biol. 151, 535-556 -74- Figure 6. Stereo drawings of each base pair of IccGG viewed down the local helix axis.The position of the helix axis is shown as a dot in each drawing. - 75- Figure 7. 1CCGG tetramers related by crystallographic twofold. -76- Figure 8. First layer waters within ·3.5A of 1CCGG -77- First Laver Waters of IccGG Water Number Wll W12 W13 W14 W16 W17 W19 W19' W20 W21 W22 W23 W24 W26 W28 W30 W31 W31' W32 W33 W36 W37 W40 W46 DNA Contacts 60L 703' 20R 105' 20R 30L 30L DNA Distances BOR Il 70R 806 105' 303' 603' 403T 4N2 4N3 505' 306 406 403' 3N7 3.1A. 3.4.A z.s.A 3.2.A 3.1.A 3.2.A 3.4.A 3.4.A 3.4.A 3.2A. 2.9A. 2.6A. 3.2.A 2.9A. 4N7 40R 60L 20R 205' 30R BOL 803T 6N4 706 7N7 2.1A. 3.o.A z.s.A 2.s.A 3.3.A 3.2.A 3.4A. 3.5A. 3.4A. 2.3.A 3 .4A. z.a.A Other Contacts W36 Other Distances 2.7.A W78 z.a.A W17 W16 W93 3.2.A 3.2.A W46 3.4.A W70 W73 2.4.A W32 W56 WBO W31 W37 W75 W69 3.5A. 3.s.A 3.2.A 3.5.A s.3A. 2.9A. 2.1A. Wll W32 2.7A. 3.3A. W21 W57 W60 2.7A. 3.5A. 3.4.A 3.o.A 3.2A. z.a.A 3.1.A z.a.A. z.a.A z.9A. TABLE 3. Water within 3.5A of 1CCGG 3.3~ 3.4~ {First of twa parts) -78- First Laver Waters of lccGG (Continued) Water Other DNA DNA Other Distances Number Contacts Distances Contacts W56 306 s .o.A W31 3.3! s .1A. W75 W57 6N4 s.oA. W46 2.7A. W58 8N7 2.5A. W59 60R 2.sA. W60 2.6J 3.4A W84 3.5A. W60 7N7 z.aA. W46 z.aA. W59 W61 401 2.7A. 40R s .o.A W63 3.3A. 301 3.4A. W93 3.4A W65 203' 3.5A. W78 201 3.2A W66 505' s.o.A W69 30R 2.7A. 3.4A. W33 305' s.4A W70 403T W28 s.3A 3.s.A 2.a.A W73 2.4A. W73 4N3 s.4A. W28 2.a.A W70 3.s.A W74 s .sA. 701' s.sA. W73 W74 2.9A. W32 W75 2N4 2.aA. s.1.A W56 2.aA. 201 s .o.A W13 W78 3 .4A. W65 801 2.5A W79 s.zA. 401 s .5A W31 WBO s.z.A 51 WBl s.4A. W59 2.7A. 60R W84 401' 3.4A. W86 s .s.A 4N3 W86' 2.9A. 701 W89 2.a.A W17 z.aA. 301 W93 3.sA. W63 3.2A. 30R TABLE 4. Water within 3.5A of 1CCGG {Second of two parts) -79- Figure 9. Second layer waters within 3.5A of two first layer waters -80- Water Number W9 W9' W38 W41 W42 W42' W44 W45 W48 W50 W51 W51' W52 W52' W64 W64' W72 W83 W94 W94' Second Layer Waters of IccGG First Layer Second Layer Distances Contacts 2.5A. W52 s.s.A s .1 A. W51 2.7A. First Layer Contacts W69 WBO W63 W93 W19 W72 WBl W28 W73 W74 W26 W23 W65 W20 W58 W60 W57 W59 W84 W16 W63 W40 W17 W93 W69 W79 W78 W36 W89 W19 W57 W19 W20 W58 W84 W20 Second Layer Distances 2.9A. 2.7A. z.aA. z.aA. 3.4A. 3.2~ s.oA. s.1A. s .oA. 2.9A. z.aA. s.oA. z.aA. z.aA. W83 z.aA. W38 2.7A. W9 2.9A. W41 2.sA. W45 W94' s.1A. W83 s.1A. 3.4A. s.sA. z.aA. s.1A. 2.5A. 3.5A. 2.7A. s.1A. s .oA. s .1 A. s.1A. z .sA. 2.7A. 2.5A. 3.4A. 3.4A. 2.7A. 3 .4A. s .2A. 3.4A. z.aA. TABLE 5. Second layer waters of 1CCGG and their contacts - 81- Chapter 6. CONCLUSIONS -82Chapter 6 Conclusions 6.1 Structural Analysis 6.1.1 Comparison of 1CCGG with fiber studies: All discussions of DNA structure must consider the foundation laid by x-ray diffraction of DNA fibers. For A-DNA, the pertinent structure is the linked atom least square refinement by Arnott and his coworkers. 1 This structure is a regular righthanded helix with a rise per residue of 2.6 angstroms and a twist of 32.7°. The bases are paired in Watson Crick fashion with a propeller twist of -12°. As mentioned earlier, the regularity of the structure is an inevitable consequence of the fiber diffraction method. Variations in the structure are averaged out by sequence heterogeneity, rotational disorder, and translational disorder of the DNA in the fibers. Nevertheless, the fiber patterns from A-DNA are the most crystalline of all DNA fibers .2 It is not surprising, therefore, to find that the structural parameters of A-DNA derived from the fiber are close to those found by this single crystal study. We find the base plane inclination in 1CCGG to be increased to 23° from the fiber value of 19°, and the average rise per residue decreased from 2.6A to 2.2A. As discussed in Chapter 5, these differences may be due in part to the difficulties associated with trying to extract helix parameters from a short piece of DNA. 1CCGG confirms the 3' endo sugar pucker that was found in the A-DNA fiber structure.• The only sugar not in this conformation is residue 4, whose conformation is not constrained by a following residue. The major difference in A-DNA helix parameters found between fiber and crystal is the propeller twist of the GC base pair. Arnott et al. report a value of -12° while we find an average value of +17°. The fact that a difference was also found between fiber and single crystal propeller twist values for 1. S. Arnott and D. W. L. Hukins (1972). Biochem. Biophys. Res. Com. 47, 1504-1509 2. A. G. W. Leslie, S. Arnott, R. Chandrasekaran and R. L. Ratliff (1980). J. Mol. Biol 143, 49-72 - 83- B-DNA3 indicates that fiber patterns are not very sensitive to this parameter. The recent solution of the A-DNA octamer, GGTATACC, 4 confirms the positive propeller twist discovered in 1CCGG. The octamer structure was solved by rotation methods using fiber A-DNA coordinates as a model, so the initial propeller twist of base pairs was negative. However. subsequent refinement of the structure revealed that the twist is indeed positive. 5 6.1.2 Other single crystal A-DNA structures: Two other single crystal A-DNA structures, GGTATACC 6 and GGCCGGCC, 7 have been solved. GGTATACC packs into a P6 1 unit cell with two octamer chains, or one double helix, as the asymmetric unit. GGCCGGCC packs into a P43 2 1 2 unit cell in the same manner as 1CCGG, and so has only one strand, or half a double helix as the asymmetric unit. These structures are both at 2.2A resolution, comparable to 1CCGG's 2.1A resolution. Currently, only preliminary information is available for the AT containing sequence of Shakked et al., but some information from a refined structure is available for the GGCCGGCC sequence of Wang et al. 6.1. 3 Major groove opening at purine pyrimidine sequence: One of the few structural details reported about the Shakked structure is that the major groove opens up by bending the helix axis at the molecular twofold located at ApT, a purine-pyrimidine sequence. This same effect can be seen about the crystallographic twofold in 1CCGG. as shown in Figure 1. This opening of the major groove at a purine-pyrimidine sequence has also been seen in B-DNA. 8 In 3. 4. R. Wing, H. Drew, T. Takano, C. Broka, S. Tanaka, K. Itakura and R. E. Dickerson (1980). Nature 287, 755-758 Shakked, D. Rabinovich, W. B. T. Cruse, E. Egert, 0. Kennard, G. Sala, S. A. Salisbury and M.A. Viswamitra (1982). Proc. R. Soc. Land. B 213, 479-487 0. Kennard- personal communication z. 5. 6. Z. Shakked et al. (1982).- ibid 7. A. H.-J. Wang, S. Fujii, J. van Boom and A. Rich ( 1982). submitted - Proc. Nat. Acad. Sci. USA 8. R. E. Dickerson and H. R. Drew (1981). Proc. Nat. Acad. Sci. USA 78, 7318-7322 -84- Figure 1. Stereo view of the octamer of IccGG IccGG the crystallographic twofold runs between a G(p )C sequence where the phosphate is missing. This opening is probably due to phosphate - phosphate repulsion across the narrow major groove of these short A-DNA helices. 6.1.4 Packing of A-DNA: It is interesting to note that all three of these A-DNA structures have the same packing interactions about their respective screw axes. In each case, the base pair at the end of one helix fits into the dry minor groove of a screw-related helix. as shown in Figure 4 of Chapter 5. This interaction occurs without stabilization from any hydrogen bonds between the two helices. This situation can be contrasted with that of the B-DNA dodecamer, where an interaction between the minor groove of two helices occurs, but is stabilized by hydrogen binding between the involved bases. This A-DNA packing scheme indicates that the minor groove of A-DNA is relatively hydrophobic. Such hydrophobicity may be an important key for molecules, such as proteins or drugs that recognize A-DNA. - 85- Three different types of intermolecular interactions between DNA have been seen in these crystal structures: 1. end-to-end stacking (A, B, and Z DNA) 2 . minor groove hydrophobic packing (A DNA) 3. minor groove H-bonds (B DNA) 6.1. 5 Regularity of 1CCGG: The general regularity of IccGG can be seen in table of torsion angles in Chapter 5. However, it is better illustrated by Figure 2, a cylindrical polar plot of the tetramer's backbone. z tAl Figure 2. Cylindrical polar plot of IccGG Although there is some small variation along the backbone, as evidenced by changes in phosphate oxygen positions, the overall impression is one of regularity. The distortion due to helix packing is easily seen in this figure. The Cl /G8 base pair has its free 03' and 05' in positions similar to the second and third base pair . However, the - 86G4/C5 base pair has its free hydroxy groups in radically different positions. Wang's GGCCGGCC was analyzed at both -18°C and -8°C. The -8° analysis shows an A-DNA helix with alternating sugar-phosphate backbone. These alterations in sugar geometry are visible in the x- o plot of Figure 3. -80° -90° From: A. H.-J. Wang, et al. submitted Proc. Nat. Acad. Sci. 1982 -100o -110° -120° Xo -130° ® -140° ~ -150° -160° -170° -180° 60° 80° 1000 120° 140° 160° 8 Figure 3. x vs. o torsion angles of GGCCGGCC by Wang et al. For comparison, the x - o plot of IccGG is shown in Figure 4. The sugar puckers of the GGCCGGCC fall into two groups, C3' endo and Cl' exo, while the x and o angles of 1CCGG all lie in the C3'-endo region. This variation in backbone conformation is only about half the variation seen in the B-DNA dodecamer, 9 but the two groups are distinct enough that this must be considered a real effect. Why is this sequence variation visible in the octamer but not in the tetramer, IccGG? The answer, as discussed below, may lie in the lengths of the molecules . 9. H. R. Drew, R. M. Wing, T. Takano, C. Broka, S. Tanaka, K. Itakura and R. E. Dickerson (1981). Proc. Nat. Aca.d. Sci. USA 78,2179-2183 - 87- -80° -90° -100o -110° -120° Xo -130° -140° -150° -160° -170° -180° 60° 80° 1000 120° 140° 160° 8 Figure 4. X vs. o torsion angles of rccGG 6.1. 6 Variety in sugar pucker: A recent bone of contention among nucleic acid crystallographers has been the variability of sugar pucker. Previous small molecule work nearly always showed sugar puckers in either 3'-endo or 2'-endo conformations. In only one instance was another conformation, the 01 '-endo pucker, found. 10 This constancy was interpreted as arising from a "rigid nucleotide" which snapped back and forth between the 2'endo and 3'-endo conformation. 11 However, theoretical calculations of the energy required for the transition from 2'-endo to 3'-endo showed that the energy surface was rather fiat. 12 The activation energy for transition between the two states was calculated as 0.6 kcal./mole. Crystal structures of longer pieces of DNA which were done about the same time showed intermediate sugar puckers. 13 As longer 10. J. Konnert, I. L. Karle and J. Karle (1970). Acta Cryst. B 26, 770-778 11. C. Altone and M. Sundralingam (1972). J. Amer. Cn.em. Soc. 94, 8205-8212 12. M. Levitt and A. Warshal (1978). J. Amer. Chern. Soc. 100, 2607-2613 - 88- structures continue to appear, the intermediate sugar puckers are becoming more common. The latest salvo fired by the rigid nucleotide group 14 finishes by saying "... one not only questions the extent to which the restrained refinement procedures [specifically, the Jack-Levitt procedure] determine the reported sugar conformation but also ponders the significance, if any, of the unusually puckered models." The alternative explanation proposed by this group requires the crystallographer to believe that what he sees in the crystal is not the actual conformation, but a mixture of 2'-endo and 3'-endo sugars which is blurred together to look like an intermediate conformation. If this were true, the electron density map should show density for the sugar which is large enough to accommodate sugars of both puckers. This in not the case. It seems better as a rule to make the theory fit the data, rather than the reverse. Perhaps large and small pieces of DNA have different rules for determining sugar pucker. Such a length-dependent conformation change has a precedent in protein crystallography. Two closely related helical protein structures are the a helix and the 3 10 helix. In the a helix, the amino hydrogen from one peptide bond hydrogen bonds to the carbonyl oxygen of another peptide bond 3 residues away. In the 3to helix the same bonding occurs, but to a carbonyl group 2 residues away. These ideal structures are seen in small peptide structures. However, chain folding stresses built up in large proteins can distort the helix and cause the amino hydrogen to appear midway between the two carboxylic oxygens, forming a bifurcated hydrogen bond. The helix is neither an a helix nor a 3 10 helix, but is an intermediate form. This theoretically untidy but realistic state of affairs may have its nucleic acid counterpart in the length-dependent variation of sugar conformation: C3'-endo vs. C2'endo for mono to tetranucleotides, and intermediate conformations in longer hel13. H. R. Drew, T. Takano, S. Tanaka, K. Itakura and R. E. Dickerson (1980). Nature 286, 567573 14. W. K. Olson and J. L. Sussman (1982). J. Am. Chem. Sac. 104, 270-278 - 89- ices. Returning to Wang's octamer, Figure 3 shows that 4 sugar puckers cluster together near Cl'-exo, and well away from either C2'-endo or C3'-endo, even though the structure was refined by the Konnert-Hendrickson constrained refinement program rather than the Jack-Levitt. It is interesting that Wang himself refuses to admit that he has a sugar pucker which is not C2'-endo or C3'-endo. He refers to his Cl'-exo sugars as "a conformation closer to C2'-endo (or Cl'-exo} than C3'-endo". 15 One more piece of information pertinent to oligomer sugar puckers will be available when the refinement of GGTATACC, which is also being refined by the KonnertHendrickson program, is reported. 6.1. 7 Solution work 'With CCGG: Other physical chemical methods have been used to study A-DNA in general and the sequence CCGG in particular. The earliest of these CCGG studies was NMR work by Patel which showed that the sequence melted about 42°C in 0.1M phosphate buffer. 16 This fact was used in selection of this particular sequence. However, of more interest to the crystallographic analysis, is the observation· of end-to-end aggregation of the tetramer in solution. This aggregation later manifested itself as the end-to-end stacking about the crystallographic twofold which makes the tetrameric 1CCGG into an octamer. Patel assigned the B conformation to CCGG by comparing the H-8 proton shift upon melting with that same shift in ribo-CCGG. In the ribo structure, 17 the shift of guanine H-8 upon melting is 0.68 ppm. This was reported as being "consistent with" A-DNA. Patel's find ing that guanine H-8 shifted less than 0.1 ppm in deoxy-CCGG led him to conclude that the tetramer was in the B form. This is a reasonable result 15. A. H.-J. Wang et al. (1982) . ibid 16. D. J. Patel (1977). Biopolymers 16, 1635-1656 17. D. B. Arter, G. C. Walker, 0. C. Uhlenbeck and P. G. Schmidt (1974). Biachem. Biophys. Res . Com. 61, 1089-1094 - 90since the conditions of measurement are not dehydrating enough to form A-DNA. However, a study of the sequences CGCG, GCGC, GGCC and CCGG by circular dichroism 18 indicated that the CCGG tetramer existed in an "unusual" conformation under conditions virtually identical to those used for Patel's NMR study. Three of the tetramers gave what would be considered normal B-DNA CD spectra with maxima near 290 nm and minima near 260 nm, but CCGG, in contrast, had its maximum near 260 nm in a typical A-DNA pattern. 19 This finding may indicate an error in the NMR studies. The solution structure of CCGG must still be considered unknown. Circular dichroism, which indicates an A form for CCGG, is perhaps the least reliable technique for structural characterization. On the other hand, the method used by Patel with NMR is not sufficient proof that CCGG is in the B form. The fact that 1CCGG, GGCCGGCC, and CCGGCCGG 20 all crystallize as A-DNA may indicate that CCGG sequences in fact have a preference for the A form even when well hydrated. If this conformational preference exists, it is undoubtedly critical to the recognition of CCGG by restriction endonucleases and methylases such as Ha.p II, Hpa II, and Msp 6.1 .8 Other structural studies of A-DNA: Recently, the Crothers group at Yale has applied the techniques of rotational correlation time and electric dichroism to the study of A-DNA. 22 This technique has been shown to give fairly good estimates of rise per residue and propeller twist of base pairs in B-DNA. 23 For A-DNA, the rotational correlation time is used to estimate 18. R. V. Kastrup, M.A. Young and T. R. Krugh (19?8). Biochemistry 17, 4855-4865 19. M.-J. B. Tunis-Schneider and M. F. Maestre (19?0). J. Mol. Biol 52, 521 20. A. H. -J . Wang et al. ( 1982). ibid 21. R. J. Roberts (1980). Nuc. Acid. Res. Com. 8, r63-r80 22. H. M. Wu, N. Dattagupta and D. M. Crothers (1981). Proc. Nat. Acad. Sci. USA 78, 68086811 23. M. Hogan, N. Dattagupta and D. M. Crothers (1978). Proc. Nat. Acad. Sci. USA 75, 195-199 - 91- the rise per residue at 2.BA. which is 0.6A from the 2.2A value found in IccGG. The propeller twist was estimated by electric dichroism as 'between oo and 7°" from perpendicular to the helix axis. For IccGG, the average propeller twist is 15°, agreeing with the outer limit of the dichroism measurement. Crothers interprets his results by saying that the base pairs do not have as large a propeller twist as in EDNA. In fact the propeller twist is equally large in A and B. 24 These disagreements indicate that electric dichroism and rotational correlation cannot be considered very precise measurement techniques. 6.2 Yater structure around IccGG 6. 2.1 Review of solution studies The earliest fiber studies 25 showed that the transition from B to A DNA was caused by dehydration. Since that time considerable effort has gone into the study of the precise nature of these water interactions. Ultracentrifuge experiments by Hearst and Vinograd26 established that the number of waters bound to DNA was a function of water activity. Specifically, they found the following relationship between water activity and DNA bound water: • At a.,.= 0.8, there are 9 waters per nucleotide • At al'r = 0.9, there are 13 waters per nucleotide • At a.,.= 1.0, there are 50 waters per nucleotide Falk and his coworkers 27 were able to distinguish waters bound to the DNA by their altered infrared spectrum. This allowed them to determine the numbers of waters bound to the DNA as a function of relative humidity. At 65% relative humidity, there are 5 to 6 waters per nucleotide. At 80% relative humidity, all sites on the DNA are 24. H. R. Drew et al. (1981). Proc. Nat. Acad. Sci. USA- ibid 25. R. E. Franklin and R. G. Gosling (1953). Nature 171, 740-741 26. J. E. Hearst and J. Vinograd (1961). Proc. Nat. Acad. Sci. USA 47, 825-830 27. M. Falk, K. A. Hartman, Jr. and R. C. Lord (1963). J. Amer. Chem. Soc. 85,391-394 - 92filled. Further hydration results in swelling of the DNA. At 92% relative humidity, the DNA has 20 waters per nucleotide. Falk also assigned a specific order to the binding of water molecules to their positions on the DNA. 1. Below 60% relative humidity: • First two waters go to charged phosphate oxygens. • The next waters go to 03' and 05' • Then water goes to 01' 2. Above 65% relative humidity • Base carbonyls and ring nitrogens are hydrated Wolf and Hanlon have used circular dichroism to monitor the A to B transition as a function of relative humidity. 28 Linking their work with that of Falk, they have found that B-DNA has about 18 waters/nucleotide, while A-DNA may have as few as 4 waters/nucleotide. The A to B transition midpoint occurs about 13 waters /nucleotide. They have divided bound waters into two groups, confusingly called A and B. Group A consists of those 11 waters which attach directly to the DNA and disturb its infrared spectrum. For a single residue, there are 5 on the base, 4 on the sugar and 2 on the phosphate. Group B waters are those waters that do not attach directly to the DNA itself but also do not show the spectrum of bulk water. Presumably they attach to the group A waters. 6 .2. 2 Correlations with IccGG: By consulting the first layer hydration stereo drawings shown in Figures 5 and 6, certain correlations can be made between the preceeding solution studies and the IccGG structure. Falk allocates the first two waters to the charged phosphate 28. B. Wolf and S. Hanlon (1975). Biochemistry 14, 1661-1670 - 93- Figure 5. Major groove view of first layer hydration of IccGG oxygens. This is certainly in agreement with the crystal structure. Phosphate oxygens account for 42% of the bonds between first layer waters and DNA. These bonds have an average bond length of 3.0A, indicating a relatively strong hydrogen bond. The phosphates are heavily hydrated in general, but in the case of the phosphate between cytosine 6 and guanine 7, there is only one water on each oxygen. Falk allocates the next waters to 05' and 03', and then to the sugar oxygen, 0 1'. This disagrees with 1CCGG. In 1CCGG, the next most numerous bonds belong to the bases. Those in the major groove account for 23% and those in the minor groove account for 7% of the first layer bonds between DNA and water. In the major groove, these bonds are strong, with an average bond length of 2. 9A. In the minor groove they are much weaker, having an average length of 3.4 A. After the sugars, Falk proposes that the bases hydrate next. IccGG indicates that the order should be reYersed. In IccGG the sugars have the longe.s t, and hence the -94- Figure 6. Minor groove view of first layer hydration of IccGG weakest, bonds of any of the groups. They account for 25% of the first layer water to DNA bonds. Falk seems to be correct when he ranks 05' and 03' before 01'. Sugar 01' atoms have an average bond length of 3.4A while the 03' atoms are stronger as indicated by a average hydration bond length of 3 .3A. The sugar oxygen 05' in a special case in A-DNA. Examination of Figure 5 shows that 05' is located along the major groove side of the sugar phosphate backbone. If a water is bridging between two pho sphate oxygens, the 05' oxygen often will be close enough because of A-DNA stereochemistry to form a hydrogen bond. In IccGG there are five 05' bonds with an average bond length of 3.2 angstroms. However, such a binding scheme would change with a change in DNA conformation. 6. 2. 3 New ordering of hydration sites on DNA With these considerations, IccGG gives us a new ordering of hydration on DNA. -95- 1. Two waters on the phosphate oxygens. (in agreement with Falk) 2. Next, three waters on the major groove side of the base pairs. At this point we have accounted for the 4 to 5 waters/nucleotide that Falk sees below 60% relative humidity. We have also accounted for the four waters that Wolf and Hanlon see as the minimum for A-DNA. 3. It is not clear which area would hydrate next. • The 05' oxygens are positioned in the major groove in such a fashion that they are sterically hindered and cannot be reached by a water mo~ecule unless that molecule also is bound to a phosphate oxygen. It is possible that one of the waters already on a phosphate would shift its position to a 05' and allow room for another water to bind to the phosphate oxygen. A fact consistent with this idea is that those phosphates with the most hydration are those phosphates with their 05' atoms hydrated. See residues two and three of Figure 5 for example. • The 03' atoms are accessible from the minor groove side of the helix. These are regularly hydrated in IccGG with an average bond length of 3.3A. These probably hydrate about the same time as ... • The minor groove guanine N2 and N3 atoms. These are hydrated in IccGG when they are not sterically hindered by the 4 3 related helix. 4. We can add a water to 03', one water each to guanine N2 and N3, and one more water at a phosphate oxygen, for a total of 9 waters . According to Hearst and Vinograd this corresponds to a., = 0.8. Thus we ~ave reached the 80% relative humidity that Falk observed was sufficient to fully cover the DNA without causing swelling. We now have 3 waters in the major groove, 2 in the minor, 3 on the phosphate, and 1 on the 03' of the sugar. - 96- 5. According to Wolf and Hanlon, there are 11 group A waters. (ie. waters that bind directly to the DNA and disturb its spectrum.) If 1CCGG can serve as a pattern, these two extra waters are probably added around the phosphate oxygens. 6. According to Wolf and Hanlon, there are 8 or 9 more waters (group B waters) to add before bulk solvent is reached. These probably correspond to the second and higher layer waters of 1CCGG. Since the midpoint of A to B transition occurs at 13 waters and we already have 11 by now, by the time these group B waters are all added, we will have B-DNA. Some of these undoubtedly go into higher layers of the spine structure 29 in the AT-rich minor groove of B-DNA. No dir ect comparison can be made with 1CCGG since the tetramer has no AT base pairs. 6.3 Spermine Spermine was necessary for the crystallization of 1CCGG. In fact this molecule is commonly used to crystallize nucleic acids. The three A-DNA structures currently under refinement, 1CCGG, GGCCGGCC, and GGTATACC were all crystallized with spermine. Even the B-DNA dodecamer 30 and t-RNA were crystallized using this molecule. Both the B-DNA dodecamer and 1CCGG were crystallized with low spermine to DNA ratios. In the case of 1CCGG, the ratio moles spermine/moles 1CCGG =0.09 because it was found that this low ratio promoted larger crystal growth. In 1CCGG the spermine cannot be located with confidence. In the initial anomalously phased electron density maps, there appeared a band of density stretching across the major groove from one phosphate chain to the other which was believed to be the spermine. The spermine in the dodecamer has been assigned to a similar position across the major groove. 31 However, most of this band later refined away and the rest was replaced by discrete water molecules. In GGCCGGCC 29. H. R. Drew and R. E. Dickerson (1981). J. Mol. Biol. 151, 535-556 30. R. Wing et al. (1980). -ibid 31. H. R. Drew and R. E. Dickerson (1981).- ibid - 97- no mention is made of the spermine location. 32 In GGTATACC the spermine has been tenatively positioned spanning the major groove and passing through the DNA twofold.33 These structures will have to await further refinement before anything is known with certainty. 6.4 S1UD.ID.ary of Advances f:roDl lccGG Advances in DNA structure due to the solution of IccGG can be classed in three groups: 1. Single crystal information about A-DNA: This includes adjustment of most of the helix paramenters and a major change (-12° to +17 °) in propellor twist. 2. Study of water structure around A-DNA including: • The first crystallographic study of A-DNA water structure. • Location and reordering of the water binding sites in DNA • Identification of the minor groove of CG-rich DNA as a relatively hydrophobic area on the DNA. 3. Identification of a possible recognition mechanism for proteins which recognize the sequence CCGG. 32. A. H.-J. Wang et a,L. (1982). ibid 33. Z. Shakked - personal communication - 98- Appendix Refined lcCGG Coordinates - 99- APPENDIX REFINED 'ccGG COORDINATES 6.1 Coordinate Tran.sform.ation 1CCGG was refined in space group P2 . To convert back to the original fractional 1 P~2 1 2 system use the following set of equations: Coordinates of 1CCGG Atom Na_me 1 05' 1 C5' 1 C4' 1 C6 1 I 1 C5 1 N4 1 C4 1 N3 1 02 1 C2 1 N1 1 C1' 1 C2' 1 01' 1 C3' 2 03' 2P 2 OL 2 OR 2 05' 2 C5' 2 C4' 2 C6 2 C5 2 N4 X P2 1 Y P21 Z P2 1 28.69225 28.00716 27.28270 29 .23810 31.35796 29.92827 30.23647 29.60312 28 .67426 27.15561 28.02016 28.28528 27.54564 26.10956 28.24947 26.25705 25.02866 24.03911 22.91927 24.81969 23 .42960 22 .49431 22 .27689 25 .29520 26 .39516 27.83986 21.57280 21.44655 20.12717 18.36084 19 .14377 17 .94553 16.30988 16.70859 15.92104 15.58567 16 .30556 17 .52457 17.96915 18.44748 19.13747 19.96494 20.68398 21.12987 21.93358 21.78647 19 .76349 18 .99651 17 .72490 17.72942 17 .99886 17.29674 26 .84247 25.57956 25.61525 28.15337 30.13559 29.28505 30.98613 29.84845 29.30568 27.69916 28 .20795 27 .64145 26.41919 26 .69147 25.95752 26.71915 26.45670 27.64156 27.09589 28.71227 28.19466 27 .39885 28.17177 30.21312 31.02411 32.76624 - 100Coordinates of 1CCGG Atom Name 2 C4 2 N3 2 02 2 C2 2 N1 2 C1' 2 C2' 2 01' 2 C3' 3 03' 3P 3 01 3 OR 3 05' 3 C5' 3 C4' 3 C4 3 N3 3 N2 3 C2 3 N1 3 06 3 C6 3 C5 3 N7 3 C8 3 N9 3 C1' 3 C2' 3 01' 3 C3' 4 03' 4P 4 OL 4 OR 4 05' 4 C5' 4 C4' 4 C4 4 N3 4 N2 4 C2 4 N1 4 06 4 C6 4 C5 4 N7 (Continued) Z P21 X P2 1 YP21 26.75803 26.06058 24.35825 25.00031 24.62090 23.46608 22.08907 23.55472 21.78563 20.37787 19.76572 18.33730 20.59134 19.94411 19.12143 19.75101 24.00597 24.09447 25.44263 25.19838 26.12482 26.99878 26.07622 24.89333 24.48891 23.32352 22.97954 21.78003 20.74260 21.14400 19.75330 18.45969 18 .21407 16.78127 18.83313 18.98996 18.50179 19.37990 23.59505 24.29764 26.37357 25.55707 26.0?082 26 .01482 25.41417 24.05269 23.08513 17.05347 15.92442 14.62628 15.66821 16.55954 16.23465 16.51270 17 .1 0706 17.91083 18.23874 19.12598 19.41718 20.35608 18.23288 17.05138 16.25188 16.30518 15.32175 14.46164 15.39832 16.38202 18.21805 17.40488 17.32845 18.12122 17.57869 16.46872 15.60839 15 .79318 16.07207 16 .77562 16.72884 17.45447 . 17.41945 18.78960 16.57605 15.28004 14.79678 17.07462 16.42038 16.33452 16.86191 17.84772 19.39754 18.52025 18.08461 18.53397 31.97635 32.13538 31.47336 31.35999 30.39708 29.50011 30.10254 28.35756 29.58649 29.63866 30.82596 30 .55515 30.91347 32.16106 32.42966 33.55945 35.57001 36.48775 38.19722 37.23824 37.05864 36.08577 36.14095 35.34311 34.32776 33.93729 34.66301 34.48987 35.59572 33.28452 34.98082 35.61133 37.02138 37.38628 36.92830 38.11295 38.53394 39.68837 40.23163 41.20238 42 .31299 41.35280 40.56342 38.94162 39.56081 39 .42586 38 .5 9720 - 101Coordinates of 1CCGG Atom Name 4 C8 4N9 4 C1' 4 C2' 4 01' 4 C3' 4 03T 5 05' 5 C5' 5 C4' 5 C6 5 I 5 C5 5 N4 5 C4 5 N3 5 02 5 C2 5 N1 5 C1' 5 C2' 5 01' 5 C3' 6 03' 6P 6 01 6 OR 6 05' 6 C5' 6 C4' 6 C6 6 C5 6 N4 6 C4 6 N3 6 02 6 C2 6 N1 6 C1' 6 C2' 6 01' 6 C3' 7 03' 7P 7 01 7 OR 7 05' (Continued) Z P2 1 X P2t Y P2t 22.01505 22.26778 21.32088 20.09926 20.75835 19.16602 19.52498 34.78635 35.27895 34.45020 31.73105 31.67685 30.90445 28.71814 29.52802 29.00020 29.29889 29.79163 31.16138 32.05501 32.56702 33.23610 34.03992 34.95102 34.50200 35.33524 34.44452 32.97849 32.55234 31.10315 30.72118 30.78351 29.68010 29.62383 28.46944 27.35480 28.40799 29.52588 29.39911 29.43439 30.48848 30.88866 31.13889 31.37346 31.72272 32.38309 29 .96111 17.78011 16.86240 15.86535 16.54770 15.10870 15.34300 14.39168 20.12749 18.81458 17.85727 19.26041 20.97504 19.85645 20.21132 19.67038 18.94954 17.75771 18.38547 18.51201 17.87569 16.43178 18.64339 16.53682 15.28222 14.10829 12.88520 14.88707 13.75417 13.12596 12.80313 15.33912 16.15477 17.14008 16.34882 15.79190 14.51688 15.02847 14.77855 13.87601 12.39521 14.07591 12.05563 10.63095 9.80809 8.40919 10.51128 9.81350 38.90099 39.88634 40.47156 41.06621 39.38306 41.10596 42.12839 42.98569 42.61200 43.42087 41.34337 38.84328 • 40.38106 39.52414 40.47629 41.47049 43.33614 42.40004 42.33446 43.40401 43.23494 43.57718 42.75937 42.88747 41.89252 41.91089 40.65947 42.25365 43.48189 43.15482 40.36136 39.22829 37.36162 38.4?510 38.84682 40.31023 39.94553 40.68929 41.88620 41.54175 42.82198 41.83347 41.87053 40.50525 40.81284 39.69695 39.74106 - 102Coordinates of 1CCGG Atom Name 7 C5' 7 C4' 7 C4 7 N3 7 N2 7 C2 7 N1 7 06 7 C6 7 C5 7 N7 7 C8 7 N9 7 C1' 7 C2' 7 01' 7 C3' 8 03' 8P B OL 8 OR 8 05' 8 C5' 8 C4' 8 C4 8 N3 8 N2 8 C2 8 N1 8 06 8 C6 8 C5 8 N7 8 C8 8 N9 8 C1' 8 C2' 8 01' 8 C3' 8 03T WAT 9 WAT 10 WAT 11 WAT 12 WAT 13 WAT 14 WAT 15 X P2 1 28.93590 27.72371 27.38431 26 .20630 24.88411 26.02950 26.95070 28.87912 28.16124 28.36948 29.41437 29.03844 27.79099 26 .95395 26.76616 27.53400 27 .78574 27.43430 27.89388 27.47580 29 .31975 27 .09305 25.68741 25.22739 27.21689 26 .42334 26.22287 26.91524 28 .09398 29.95340 28.92255 28.40456 28 .90691 27.98183 26.92023 25.65117 25.60973 25 .51009 25 .98424 25.57268 22.87546 39.03223 35.91507 28 .86957 24 .7 1236 27 .38318 38 .12708 (Continued) YP2 1 Z P21 8 .87078 9.02498 12.57121 12.42556 13.32930 13.32393 14.29975 15.43367 14.47759 13.48883 13.31603 12.28167 11.80051 10.76017 9.44390 10.43892 8.52793 7.12893 6.31362 4 .89591 6.57677 6.97316 6.69514 7.35363 11.21548 11.73470 13.51322 12.85171 13.39447 13.51386 12.90638 11.69939 10.92131 9.95850 10 .10149 9.333 9 8 8.19201 8 .74218 6 .99538 5.74169 22 .29451 15.35583 10.77356 9.58239 24.72469 22 .07094 18.84697 40.14951 39.26656 36.17882 35.52907 33.82657 34.55948 34.30614 34.64487 34.94005 35 .94809 36.78545 37.56090 37.24321 37.90857 37.16843 39.17290 37.82599 37.67673 36.37465 36.44881 36 .15659 35 .15828 34.99144 33 .72791 31.91980 30 .96709 29 .46544 30 .43387 30 .85117 32 .11237 31.83022 32.39354 33.37175 33.51059 32.65511 32 .60010 31.59740 33.89490 32 .45929 3 1.88991 34.44115 41.09807 40.56984 44. 13828 28 .39394 29.72832 4 3 .27396 - 103Coordinates of 1CCGG Atom Name WAT 16 WAT 17 WAT 18 WAT 19 WAT 20 WAT 21 WAT 22 WAT 23 WAT 24 WAT 25 WAT 26 WAT 27 WAT 28 WAT 29 WAT 30 WAT 31 WAT 32 WAT 33 WAT 34 WAT 35 WAT 36 WAT 37 WAT 38 WAT 39 WAT 40 WAT 41 WAT 42 WAT 43 WAT 44 WAT 45 WAT 46 WAT 47 WAT 48 WAT 49 WAT 50 WAT 51 WAT 52 WAT 53 WAT 54 WAT 55 WAT 56 WAT 57 WAT 58 WAT 59 WAT 60 WAT 61 WAT 62 X P21 15.71326 16.70154 18 .44806 31 .54509 33.80396 32 .73351 30.30489 20.04610 38.18828 28 .65294 18.56360 39.36414 37.59586 25.43233 24.65195 16.39568 13 .00669 16.02460 37.13176 28.80469 22.68784 23.65076 15.81119 32 .16869 33.70923 31.86336 25.96515 40.19809 20.60822 35.78171 31.17519 35.20650 33.35721 30 .17421 33 .75751 35.28368 20.67754 35.64240 38.23360 34 .88393 29 .22678 32.36420 34.00751 33 .428 47 37.01839 16.18442 35.95784 (Continued) Z P21 YP21 17.84462 20.03473 24.69693 5.90616 8.74271 13.35038 24.16092 19.56577 25.18158 24.65044 16.41988 23.34517 19.04645 25.77808 15.19827 14.27169 13.91278 12.95030 11.13017 5 .34241 3.88231 21.13669 23.26941 23 .36649 23 .66675 21.78467 12.04826 15.03417 20.37468 15.88939 15.42055 14.97240 17.91434 6 .03266 25 .97881 24.92014 23.09845 23.38826 22.37488 21.52393 20.02292 17.53281 16.31920 14.76941 13.77567 20 .01645 26.69983 30.60759 32.73630 31.50084 34.04625 37 .94617 32.72876 25.95244 26.80925 44.26433 32.84888 44.67209 39 .65976 23 .30791 33 .54160 25.25446 35 .18245 35.80127 39.98839 38 .13202 41.39368 25 .47751 31.19217 30 .06293 29.05861 26 .71416 33 .00409 42 .72205 38 .61783 24 .13387 33 .46552 34 .84662 36 .19202 39 .25293 43.39156 36 .44420 29 .57631 36.18973 38.70226 34.96750 35.40392 35.31171 36.04160 31.29778 38 .26389 32 .17072 37 .46584 35 .95711 - 104Coordinates of 1CCGG Atom Name WAT 63 'WAT 64 WAT 65 WAT 66 WAT 67 WAT 68 WAT 69 WAT 70 WAT 71 WAT 72 WAT 73 WAT 74 WAT 75 WAT 76 WAT 77 WAT 78 WAT 79 WAT 80 WAT 81 WAT 82 WAT 83 WAT 84 WAT 85 WAT 86 WAT 87 WAT 88 WAT 89 WAT 90 WAT 91 WAT 92 WAT 93 WAT 94 (Continued) X P2t YP2 1 Z P2 1 37.78879 26.97086 22.92049 35.61533 37.81128 40.20052 20.90021 40.03119 35.32204 32 .96938 23.40503 21.09926 28.35091 27.05965 22.98029 24.33783 26.44994 14.01204 32.65770 38.94791 33.58118 35.76973 31.51845 18.77907 40.53183 24.72504 39 .90665 30.00676 20.97115 40.43021 17.74277 35 .9 8708 25.31223 24.55333 21.61957 22 .10085 21.10889 21.34871 20.79294 21.17653 5.72781 19.28310 13.24294 17 .08371 19 .85747 26.26587 25.37993 24.15001 3.68093 16.37453 22.63315 19.18996 8 .29796 16.90804 23.95116 25.86765 17.53714 26 .69980 5.59533 25 .13 036 23 .88373 26.69954 21.83675 18.49216 33.65392 24.21744 23.88181 40.86673 40.61458 41.85068 34.24814 22.75374 40.51955 33.25403 41.77881 45.12212 32.31110 28.73302 23.30289 25.73256 38.37366 3 5 .58904 36.23270 35.84273 35.32224 39.39597 32.21445 44 .3 5048 39.12218 24.72260 48 .03 49 3 30.13039 29 .67679 40.42981 31.25867 36 .56 966