A structural analysis of G-quadruplex/ligand interactions
Shozeb M. Haider a,*, Stephen Neidle b, Gary N. Parkinson b,c,*
a Centre for Cancer Research and Cell Biology, Queen’s University Belfast, Belfast BT9 7BL, Northern Ireland, UK
b Cancer Research UK Biomolecular Structure Group, The School of Pharmacy, University of London, London WC1N 1AX, UK
c Department of Pharmaceutical and Biological Chemistry, The School of Pharmacy, University of London, London WC1N 1AX, UK
a r t i c l e i n f o
Article history:
Received 24 March 2011
Accepted 16 May 2011
Available online 26 May 2011
Keywords:
G-quadruplex Structural analysis Ligand binding
a b s t r a c t
This focused review article discusses in detail, all available high-resolution small molecule ligand/G- quadruplex structural data derived from crystallographic and NMR based techniques, in an attempt to understand key factors in ligand binding and to highlight the biological importance of these complexes. In contrast to duplex DNA, G-quadruplexes are four-stranded nucleic acid structures folded from guanine rich repeat sequences stabilized by the stacking of guanine G-quartets and extensive WatsoneCrick/ Hoogsteen hydrogen bonding. Thermally stable, these topologies can play a role in telomere regulation and gene expression. The core structures of G-quadruplexes form stable scaffolds while the loops have been shown, by the addition of small molecule ligands, to be sufficiently adaptable to generate new and extended binding platforms for ligands to associate, either by extending G-quartet surfaces or by forming additional planar dinucleotide pairings. Many of these structurally characterised loop rearrangements were totally unexpected opening up new opportunities for the design of selective ligands. However these rearrangements do significantly complicate attempts to rationally design ligands against well defined but unbound topologies, as seen for the series of napthalene diimides complexes. Drawing together previous findings and with the introduction of two new crystallographic quadruplex/ligand structures we aim to expand the understanding of possible structural adaptations available to quadruplexes in the presence of ligands, thereby aiding in the design of new selective entities.
© 2011 Elsevier Masson SAS. All rights reserved.
⦁ Introduction
Organized around double-stranded B-form DNA, genomic DNA is both resilient and flexible enough to store genetic information and pass that information on. Helical B-form DNA is the norm of several secondary structural motifs that utilizes an anti-parallel arrangement of two complimentary strands, stabilized by Wat- soneCrick base pairing, base stacking, all within a suitable hydrated environment. Once freed from the associations of an extended complimentary sequence, single stranded DNA and RNA can adopt a vast array of other stable secondary structure motifs, such as stem- loop, pseudoknots, and tetraloops, ideal for its involvement in other biological settings other than as a store of genetic information. In particular, G-rich sequences can utilize both the WatsoneCrick and Hoogsteen faces of a guanine base to self-associate in a hydrated environment containing cations to form extended four stranded stacked structures termed G-quadruplex or tetraplexes [1] The stability of extended quadruplexes has fuelled research beyond
* Corresponding authors.
E-mail addresses: [email protected] (S.M. Haider), gary.parkinson@pharmacy. ac.uk (G.N. Parkinson).
biology into the development of nanowires and sensors. As poten- tial therapeutic targets, G-quadruplexes have been associated with telomeres located at the ends of chromosomes by interfering with the enzyme telomerase and other associated protein. The ability of small molecules to stabilise G-quadruplex DNA and interference with telomere extension in cancer cells by inhibiting the enzyme telomerase have highlighted the potential role of quadruplexes as anti-cancer drug targets [2e4]. The working hypothesis is that the ligand stabilises telomeric DNA into quadruplex structures that inhibit the primer from interacting as a single stranded species with the RNA template of telomerase. The approach to target quadruplexes overrides the problem of telomerase resistance along with inhibiting telomere-independent mechanisms of telomere maintenance in cancer cells, making them more universally applicable than conventional telomerase inhibitors. More recently, this strategy has been extended to the promoter sequences of oncogenes where the stabilisation of quadruplexes leads to tran- scriptional regulation of oncogenes [5,6]. The recently identified transcription of telomeric DNA into G-rich RNA (telRNA or TERRA) composed of r(UUAGGG) repeats of 100e1000 bp with an average length of 200 bp [7] has propelled interest towards structural descriptors for RNA G-quadruplexes. Both NMR and x-ray structural
0300-9084/$ e see front matter © 2011 Elsevier Masson SAS. All rights reserved. doi:10.1016/j.biochi.2011.05.012
1240 S.M. Haider et al. / Biochimie 93 (2011) 1239e1251
determinations of telomeric RNA are now available revealing a stable parallel alignments and a crystallographically determined x-ray structure of a ligand/RNA complex shows rearrangement of the loops.
A wide range of ligands has been reported and a number of structure-activity relationships derived [2] for small molecule binding to quadruplexes. The ligands possess several common features, notably planar aromatic chromophores and positively charged side chains. Fortunately several high-resolution structures are now available (both from X-ray crystallography and NMR studies) from human and protozoan ciliates complexed with ligands that provide detailed insights into the quadruplex-ligand interac- tions. Currently topological data of G-quadruplexes in complex to ligands is limited to two main structural arrangements: bimolecular (intermolecular), (and intramolecular parallel stranded arrange- ments with connecting propeller (double chain reversals loops); and a bimolecular antiparallel topology with diagonal loops d(G4T4G4). Even with this limited starting set of folded topologies, the simple introduction of a few ligands has generated an immensely diverse set of structural adaptations, revealing significant structural rearrangements of connecting loops.
Here we describe in detail, the overall structural details of G-quadruplexes in complex with a variety of small molecule ligands, explore the types of interaction that each ligands make with each type of topology and explain how the native structures adapt to the addition of the ligands to create new sites of interaction. We also discuss these associations with a view to how we might improve future ligand design based on high resolution structural data.
⦁ Building blocks of G-quadrupexes
Central to the G-quadruplexes sit a series of stacked planar G-quartets comprised of four guanine nucleotides stabilized by eight hydrogen bonds (N1eO6)(N2eN7) exploiting both Hoogsteen and WatsoneCrick faces (Fig. 1). In a similar arrangement to duplex DNA, the stacked G-quartets have a regular rise (3.13e3.3 Å) and a twist (30◦) that develops a groove into which a network of ordered solvent molecules are organised around the guanine N3, ribose O40 and phosphate OP1/2 atoms. Central to the G-quartets are the carboxyl O6
Fig. 1. Guanine can hydrogen bond on both its WatsoneCrick and Hoogsteen faces in a co-planar array to form a G-quartet.
atoms that when stacked creates a series of cages formed from eight oxygen running central to the quadruplex ideally oriented to create a negatively charged channel for the bipyramidal antiprismatic coordination of metal cations. The G-quartets form an ideal platform for the pep stacking of polyaromatic small molecule ligands. The guanines are attached to the ribose sugars through a glysosidic bond c (N9eC10) restrained in the G-quartets to either anti (—120 to 180◦) or syn (0 to 90◦) dihedral angle (O40eC10eN9eC4). The deoxyribose in the case of DNA can be seen to adopt both C30 or C20 -endo confor- mations, while the RNA ribose tends to adopt the C30 -endo confor- mation due to anomeric and gauche [primarily O40eC10eC20eO20] stereoelectronic effects [8]. The flexible phosphodiester backbone links the ribose sugars together in either a 50 / 30 or 30 / 50 orientations generating an external negatively charged backbone. The grooves formed between the backbones provide channels that selective ligands can be designed to exploit for specificity. The back- bone strand arrangements for quadruplexes can run both parallel and antiparallel to each other generating various groove widths, narrow, wide, or medium arrangements (Fig. 2) [9]. Three different loop connections can be made in a standard intramolecular arrangement, lateral, diagonal and propeller, each with a nominal minimum number of bases. The connecting nucleotides can take on many structural arrangements, normally involving base stacking over the G-quartets. The association of ligands containing polycyclic aromatic ring systems can result in significant rearrangements of these loops, including the expansion of the G-quartets to hexads and octads.
⦁ Quadruplex ligand complexes
⦁ Structure of oxytricha quadruplex with disubstituted acridines (BSU6039, 6042, 6066, 6038, 6045, 6048, 6054,
bis 3-fluropyrrolidone)
The crystal structure of quadruplex DNA formed from the telo- meric sequence of Oxytricha, d(G4T4G4), was the first to be solved in complex with a disubstituted acridine, BSU6039 [10]. Disubstituted acridines belong to a class of ligands that have shown to be quad- ruplex stabilising agents and can inhibit telomerase in the low mM range [11e13]. The core polyaromatic ring of acridines has high affinity to duplex DNA utilizing base stacking and intercalation, but little selectivity. Substitutions at the 3 and 6 positions usually have little influence on quadruplex to duplex DNA selectivity, however some disubstituted triazole-linked acridine compounds have shown selectivity to human telomeric sequences combined with limited binding to duplex DNA [14].
The structure is bimolecular with two strands of the sequence d(G4T4G4) coming together to form an intermolecular antiparallel quadruplex. The thymine bases that form the loops diagonally connect one edge of the terminal G-quartets to one another and are present across the top and the bottom of the stack of G-quartets. The sugar puckering is C30-endo and the guanine glycosidic bond angles are in an alternating syneanti arrangement along the strand and synesyneantieanti arrangement within the quartet.
The diagonal loop topology results in one wide, two medium and one narrow groove. Though the diagonal loops are symmetric, the ligand molecule is bound in only one of the two T4 loops. The acridine chromophore intercalates in this loop and stacks on half of the terminal G-quartet, thus maximising the pep stacking interactions. Thymine2 in the loop is positioned in the plane of the acridine chromophore and is directly involved in ligand binding. There are two hydrogen bonds formed between the O2 atom of thymine2 and the central ring nitrogen atom in the ligand and also between the N3 atom from the same thymine and the amide oxygen atom of one of the two side chains. Thymine3 in the loop is stacked asymmetrically on top of the acridine chromophore.
S.M. Haider et al. / Biochimie 93 (2011) 1239e1251 1241
Fig. 2. Schematics of tetrad and quadruplex structural arrangements. a) Guanine bases shown as rectangular blocks, with glycosidic torsion angle c syn (grey) and anti (blue). Groove widths (w) wide, (m) medium, (n) narrow. Strand orientation (þ) or (—). Central metal cation drawn as a pale blue sphere. b) Linking the G-runs together the phosphate backbone connections can take the form of lateral, propeller and diagonal loops. The relationship of glycosidic torsion angle c, syn (grey) and anti (blue) orientations of the guanine bases to the backbone orientations are shown (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.).
The direct hydrogen bonds between thymine2 and the ligand and the sandwich between thymine3 and the guanine bases in the top G-quartet ensures effective stacking and locks the acridine chromophore in the binding site. Both the amide groups in the side- chains are in a cis conformation. The aminoalkylamido side chains are in an extended conformation that positions them close to the grooves, although neither of them penetrates the groove. One of the
protonated nitrogen atoms in the side chain terminal pyrrolidino group is positioned directly above the 50 end of one strand and forms hydrogen bonds with the O4’ atom of a terminal guanine deoxyribose. The second pyrrolidino ring nitrogen atom interacts with an exocyclic N2 atom of a guanine in the first base via water- mediated interactions. G-quartet (Fig. 3).
By comparison of the structure of the ligand-boundloop with the native loop [10,15], shows that the thymine2 is swung out from the loop and hangs on the mouth of the wide groove. Thymine2 is in the position of native thymine3 and stacked on the terminal G-quartet. It is thymine that makes direct hydrogen bonds with the ligand. Thymine3 sandwiches the acridine chromophore on to the terminal G-quartet and thymine4 is oriented away from the quadruplex making no interactions other than those in the crystal lattice. The conformation of the ligand bound loop suggests that only the backbone of the residues is required to maintain loop topology whereas the bases are not involved. This confirms that the loops are highly flexible and can adopt multiple conformations depending upon the interactions required to take place with a ligand. There are networks of water molecules that lie along the grooves making hydrogen bonds with the base edges and sugar-phosphate backbone. There is a preference for solvent to interact with the N2 exocyclic amino group other than with the N3 group. A considerable amount of hydration is present in the loops and contributes towards the maintenance of loop conformation, especially in the absence of ligand [15].
Neidle and co-workers have also crystallised an additional six disubstituted acridines (having differing substituent sizes) with the Oxytricha telomeric quadruplex and examined how the thymine loop accommodates differences in substituent size in a range of closely related compounds [16]. All these acridine compounds (BSU6038, 6042, 6045, 6048, 6054, 6066) bind in a similar manner as BSU6039 [10]. The formation of a hydrogen bond between the thymine O2 and central ring nitrogen atom confirms that this nitrogen atom is
protonated. None of these structures have any direct interaction between the amine nitrogen atom in the side chain and the phos- phate backbone, thus the charged groups would be exerting their effects on ligand binding via long-range electrostatic interactions. The conformations of the side chains of the ligands mostly cluster together, although the methyl groups attached to the piperidino rings are oriented in a variety of directions suggesting the available space for substituent extension on the side chain ring system e.g. the para methyl substituted to the piperidino ring is flipped by w60◦ into a region with a larger volume that can accommodate the para substitution. Furthermore, the ortho ethyl group is relatively close to the backbone and sugar oxygen atoms suggesting that a suitable ortho substituent such as a methylene hydroxyl group could form
additional hydrogen bonds to O30 or phosphate oxygen atoms.
The bimolecular diagonal loop topology of the Oxytricha quadruplex is able to accommodate larger side-chain groups, such as a seven- membered ring, without any distortions thus allowing considerable latitude in the length and size of substituent. A similar diagonal loop is also present in the c-myc quadruplex DNA with bound TMPyP4 [17]. The porphyin substituent groups are also exposed in a similar way to the side chains as observed in ligands interacting with the Oxytricha quadruplex.
An effect on the hydrogen bonding properties of the side chains was examined using bis-3-fluoropyrrolidine enantiomers (R,R) and (S,S) [18]. It is known that when a CeF bond is positioned close to a positively charged quaternary amine, a charge dipole interaction is generated between the CeF bond and the ammonium NþeH. The acridine chromophore binds in identical positions to other struc- tures [10,16] whereas the pyrrolidinium NþeH in both enantiomers is oriented in opposite direction to that in the non-fluorinated structures suggesting that the hydrogen bonding interactions are very different following the introduction of the CeF bond. The fluorine adopts a pseudo axial orientation in all four pyrrolidinium ring moieties, forcing greater puckering of the pyrrolidinium ring
than that observed in the non-fluorinated ligands. The rings are rotated by around 180◦ relative to BSU6039 and all interact with phosphate atoms. Steric factors may perhaps play a role in change in ring conformation, however the pKa of the protonated pyrrolidines will reduce from w10.7 to 9.0 due to b-fluorination, thus the NþeH hydrogens should contribute towards forming stronger hydrogen bonding interactions [18]. The replacement of CeH to CeF is
1242 S.M. Haider et al. / Biochimie 93 (2011) 1239e1251
Fig. 3. Structures of quadruplexes in complex with ligands coloured by atom types with VDW surface representation added for the ligands (a) BSU6039; 1L1H (b) RHPS4; 1NZM (c) Daunomycin; 1O0 K (d) Distamycin-A; 2JT7 (e) Distamycin-A analogue; 2KVY.
sterically neutral. Fluorines that cause puckering and change the orientation of pyrrolidine ring do not contribute towards stabili- sation other than by stronger electrostatic interactions with the phosphate backbone. Thus F.NþeH charge dipole interaction can be used to influence the molecular conformation of ligands binding to macromolecules [18].
⦁ Structure of d(TTAGGGT)4 with RHPS4
The NMR structure of the fluorinated pentacyclic quino[4,3,2-kl] acridinium cation (RHPS4) in complex with the human telomeric repeat d(TTAGGGT)4 has been reported by Searle and co-workers [19]. Four strands of the sequence come together to forms a parallel stranded intermolecular quadruplex. The four-fold symmetry is maintained in both the bound and unbound structures. The pattern of nuclear Overhauser effects (nOes) exhibit characteristics of a right-handed twist backbone conformation.
Detailed analysis of the inter-residue nOes reveals that the adenine bases adopt anti glycosidic torsional angles and stack on top of the G-quartets as A-quartets. They are partially stabilised by well-defined intermolecular hydrogen bonding interactions between 6-NH2 and N1. Only weak nOe connectivity’s are observed for the TTA sequence, highlighting the dynamic interplay between these nucleotides, with no evidence of T-quartet formation for the terminal 50 and 30 residues.
The ligand binds to the quadruplex with a 2:1 stoichiometry, which is confirmed by the sharper resonances when plotted as a function of the bound ratio of ligand. The effects of temperature on line widths further suggest the occurrence of multiple orientations
at the two non-overlapping binding sites. The primary site for ligand binding is between the ApG step i.e. sandwiched between an A-quartet and a G-quartet. There are some structural distortions observed in the A-quartet that are a result of the intercalation of the RHPS4 at the ApG step. The second ligand-binding site is at the GpT step. The binding modes of RHPS4 in both site is very similar. In order to maximise the p-stacking interactions, the ligand is slightly offset only stacking on two bases of each G-quartet and does not adopt a centro-symmetric position. This also allows the partial positive charge on ligand atom N13 to be positioned directly above the electronegative channel running through the central axis of the quadruplex. The partial positive charge behaves like a pseudo monovalent cation and imparts stability to the structure. The electron-withdrawing nature of the two fluorine atoms increase the positive charge density of N13 and thereby enhancing its electro- static contribution. The other CH3 substituent groups in the ligand are oriented towards the grooves. The simple end stacking mode of
RHPS4 is able to stabilise core G-quartets (DTmw20 ◦C) and there is
no evidence of the ligand intercalating between the G-quartets.
⦁ Structure of d(TGGGGT)4 with daunomycin
There is evidence that the duplex DNA targeting anticancer drugs such as the anthracyclines doxorubicin and daunomycin can interact with telomeric DNA [20]. Although these drugs lack selectivity, they can readily bind to quadruplex DNA via stabilisation of the anthraquinone chromophore. The asymmetric unit in the crystal lattice of the daunomycin complex contains four parallel strands of the sequence d(TGGGGT) coming together to form an
S.M. Haider et al. / Biochimie 93 (2011) 1239e1251 1243
intermolecular quadruplex, along with three daunomycin mole- cules [21]. The four G-quartets are stacked on top of one another with the thymine bases oriented away from the quadruplexes into the solvent. Three molecules of daunomycin are stacked on the 50 end, making p-stacking interactions with the terminal G-quartet. The platform for ligand binding is large enough to accommodate three daunomycin molecules simultaneously and there is no evidence of the ligand intercalating between the layers of the quadruplex. The asymmetric units in the crystal lattice are related to one another by a crystallographic two-fold axis that stacks two quadruplex units on one another in a 50e50 arrangement. Two layers of daunomycin ligands are sandwiched between the quadruplexes, one from each unit. The degree of overlap between the two layers of ligands is greater than that observed between the ligands and
guanines in the G-quartets. The hydrophobic van der Waals inter- actions hold each trio of daunomycin in one layer. In addition to the stacking interactions, the daunosamine sugar moieties makes interactions in the three of the four grooves. The positively-charged amine in the daunosamine sugar of the first daunomycin makes hydrogen-bonding interactions with phosphate oxygen groups on both sides of the groove. For the second daunomycin, hydrogen bonds are formed between cationic amine and exocyclic hydroxyl groups from the ligand and phosphate oxygens on only one side of the groove. There are no direct interactions between the third daunosamine sugar and the quadruplex groove; however, there are some water-mediated interactions. The insertion of daunosamine sugars in the groove does not have any effect on quadruplex groove widths. The backbone conformation adopted by the complex is similar to that of the native quadruplex [22,23]. Recent NMR studies
on the d[(TGGGGT)4]sequence have identified the additional pres- ence of T-quartet formation (15% to 30%) on the 50 end, this is in addition to the four expected G-quartets [24].
⦁ Structures of d(TGGGGT)4 with distamycin
Distamycin-A is a antibiotic-like small molecule that has been shown to bind to (1:1 stoichiometry) and expand (2:1 stoichiom- etry) the minor groove of adenine and thymine rich tracts of duplex B-DNA in drug/DNA complexes [25] and [26]. Two molecules of Distamycin-A bind in an antiparallel orientation in the 2:1 stoichiometry. There has been no structural consensus on how Distamycin-A interacts with quadruplex DNA. Several models had been proposed which included distamycin-A binding as dimers in two opposite grooves, two Distamycin-A molecules stacking on terminal G-quartets and a mixed groove/G-quartet stacking binding mode [27,28]. A recent NMR structure by Randazzo and co-workers has unequivocally shed light on the structural binding mode of Distamycin-A [29]. The quadruplex-ligand complex exhibits four-fold symmetry forming an intermolecular parallel stranded topology with a right handed B-form helical rotation, the backbone conformation resembling that of the native d(TGGGGT)4
structure and all sugar puckers adopt an anti glycosidic confor- mation. The deoxyribose rings are predominantly in the C20 endo conformation, except for the second guanine that is in the C20 endo conformation. Distamycin-A binds to d(TGGGGT)4 with a 4:1 stoichiometry, with two distamycin-A dimers binding simulta- neously in two opposite grooves of the quadruplex [29].
The distamycin-A molecule makes simultaneous interactions with the guanine bases in the quartets and the sugar-phosphate backbone. The four ligands make eighteen head to tail drug-drug contacts and the ligand is positioned closer to the backbone and to the base when compared with the similar binding mode in the minor groove of duplex DNA [25,26]. The changing NMR signals during titration suggests that the binding of two Distamycin-A molecules are side-by-side in a highly cooperative mode. The NOESY spectra
obtained during titration also confirm the presence of same head to tail drug-drug interactions observed for the 4:1 complex. The NOE data also confirms the binding of Distamycin-A in two opposite grooves of the quadruplex. Binding of each Distamycin-A molecule expands the width of the groove by w2 Å to 17.8 Å and a subsequent reduction in the width of the adjacent groove by w1 Å to 15 Å. This is also similar to the increase in groove width observed in the 2:1 duplex DNA complex, upon binding of Distamycin-A [25,26]. The ligand dimers span almost the entire length of the grooves, however,
they are slightly shifted towards the 50 end. The N-methylpyrrole
rings and the peptide bonds are planar while the links connecting these planar rings are flexible. The peptide-pyrrole and pyrrole- peptide hinges adopt twist angles between 2◦ and 26◦. The two staggered antiparallel Distamycin-A molecules overlap by w90%. The N-methylpyrrole ring of one ligand faces the peptide bond of another. This results in the crescent shape of the ligand that fits the curvature of the grooves in quadruplex DNA. The dipole moments of the pyrrole ring and carbonyl groups are in opposite direction resulting in favourable dipoleedipole interactions. This along with p-stacking interactions favours the antiparallel orientation of Distamycin-A binding to quadruplex DNA. Furthermore, the van der Waals complementarity between the Distamycin-Aand the grooves maximises favourable ligand-DNA interactions. The terminal amidi- nium moiety in the ligand interacts with the phosphate backbone
while there are another four hydrogen bonds formed between NH2 groups of Distamycin-A and N2/N3 of guanine bases from the G-quartet. This is similar to the hydrogen-bonding pattern observed in Distamycin-A-Duplex DNA binding where NH group from the drug interact with N2 (purine) or O2 (pyrimidine) base atoms [25,26]. The fact that there are no observed NH atom interactions in ligandequadruplex DNA makes the duplex and quadruplex grooves chemically different.
In order to further explain the details of groove binding of Distamycin-A to d(TGGGGT)4 quadruplex, Randazzo and co-workers modified the end groups of Distamycin-A, where the amidinium groups are replaced by uncharged N-methylamide moiety [30]. The structure adopted by the analogue-complex is similar to that of Distamycin-A-complex [29], binding in a similar staggered, anti- parallel overlapping manner fitting the curvature of the quadruplex groove. The N-methylpyrrole ring of one molecule makes interac- tions with the terminal amide bond of the other. One of the ligand molecules is slightly more solvent exposed and the other sand- wiched between the first and the quadruplex structure. This is due to the absence of the positively charged groups in the analogue that help in anchoring the ligand in the groove. Unlike with Distamycin-
A, the NOE contacts highlight that the Distamycin-A analogues do not stack on the 50 edges but are able to span the entire groove. The ligands make both intramolecular and ligandeligand and ligandequadruplex intermolecular hydrogen bonds. The flexible terminal methylcarboxamide group makes intramolecular hydrogen bonds between NH group and carbonyl oxygen attached to the third pyrrole ring and by an intermolecular bond between the terminal carbonyl oxygen and the terminal NH2 group of the adjacent ligand. The 50 terminal guanine base makes two hydrogen bonding inter- actions with the analogues via its O40 and N3 atoms. While polar interactions are predominant in the centre of the quadruplex groove, the hydrophobic interactions are stronger towards the 30 end where the terminal pyrrole ring of one molecule is surrounded by an aromatic cage formed by the two thymine bases of the groove and the third pyrrole ring of the other ligand molecule.
ITC measurements confirm that interactions between Distamycin-A (and its analogue) and quadruplex grooves is entro- pically driven [31]. While in the case of Distamycin-A, a small favourable enthalpy change was observed, the change in case of Distamycin-A analogue is slightly unfavourable. This is possibly due
1244 S.M. Haider et al. / Biochimie 93 (2011) 1239e1251
to the differences in the chemical structures of the ligands. In case of Distamycin-A, the charged amidinium moiety of the ligand interacts with the phosphate groups of the quadruplex, generating a favour- able enthalpy change, which is not observed in the analogue, due to the presence of the uncharged N-methylamide moiety.
⦁ Porphyrin binding to quadruplexes
Several tetrasubstituted porphyrins typified by TMPyP4 (tetra- N-methyl-pyridyl-porphin) bind with high affinity to quadruplexes [32], although it does not possess high selectivity for quadruplex DNA compared to duplex DNA [33,34]. TMPyP4 itself is an effective telomerase inhibitor, with an EC50 of 6.5 mM in a cell free assay [32,35] and at subcytotoxic concentrations inhibit cell growth in a time-dependent manner leading to decrease in telomerase activity over time [32,36]. Porphyrins induce chromosomal insta- bilities in sea urchin embryos, which have been attributed to direct quadruplex stabilisation [37]. It also downregulates the expression of c-myc [6] and k-ras [38] oncogenes by binding to quadruplexes formed from the promoter sequences. The structure and topology of quadruplexes formed from the human telomeric repeats d(TTAGGG) have been studied by both x-ray and NMR [17] and [39].
⦁ Crystal structure of G-quadruplex-TMPyP4
þ
The biomolecular form of the human telomeric G-quadruplex in complex with TMPyP4 has been resolved to 2 Å [40]. Two strands of the human telomeric repeat d(TAGGGTTAGGG) associate to form an asymmetric parallel stranded quadruplex. The core structure consists of three planar stacked G-quartets with the connecting TTA loops forming the external propeller loop topology. Three monovalent K ions are positioned between the quartets in an antiprismatic bypyrimidal arrangement. Two TMPyP4 molecules interact with quadruplex DNA. The first TMPyP4 molecule is posi- tioned on a 2-fold axis where it is sandwiched between the quad- ruplex it is directly stacked on and a symmetry-related quadruplex. The second TMPyP4 molecule is located perpendicular to the quadruplex stacks and interacts with the TTA loops. The N-methyl- 4pyridyl groups in the TMPyP4 are perpendicular to the porphyrin plane in both orientations. An interesting structural feature is the conformation of one of the loops. Adenine-8 from one strand flips out and makes reverse WatsoneCrick A-T base pairing interactions
with the terminal 50 Thymine-1 from the same strand. A direct
result of this reduces one of the loops to a di-nucleotide TeT loop. In the crystal structure, there is no observed stacking of TMPyP4 on or between G-quartets. It is on the AeT base pair that TmPyP4 stacks. The AeT base pair then positioned over A2eT12eA13 base stepped triad platform that is contributed by residues from both strands. This then is stacked on the G-quartet. As a result of the two-fold symmetry axis, through the porphyrin plane, the TMPyP4 molecule is sandwiched between two AeT base pairs. The two Thymines: T6 and T7, in the di-nucleotide loop are stacked on one another and the second TMPyP4 stacks on Thymine-6. Thymine-6 also forms a symmetry-related base pair with Thymine-18 from the other loop. The TMPyP4 interacting with the loops is positioned in the space between the stacked quadruplexes in the crystal lattice such that all four of its pyrrole rings are sandwiched between the two TeT base pairs.
The structure of the one of the loops that does not interact with the porphyrin is similar to that observed in the crystal structure of the native human telomeric sequence [40]. The pattern of hydration is also similar to that from the native crystal structure where N2 and N3 of guanines are coordinated to solvent atoms. The preference of TmPyP4 to stabilise bimolecular quadruplexes over unimolecular has been reported by Kim et al. and may be relevant to induce
formation of anaphase bridges at the ends of telomeres [41]. The crystal structure of the TMPyP4 with human telomeric sequence also provides a basis for a model for bridge formation and TMPyP4 stablisation, where two separate chromosomes are linked with the formation and stabilisation for a strand from each single-stranded overhang to form a parallel bimolecular quadruplex that is held together by porphyrin binding [41]. However, the crystal structure does not support the suggestion of a direct stacking of the porphyrin on G-quartets. A direct stack of the porphyrin on G-quartets results in steric hinderance by the pyridyl group and an effective loss of pep stacking. As a result the optimal distance of interaction is extended from 3.4A to 4.2A. This has been observed in the NMR structure of the parallel stranded G-quadruplex from the c-myc promoter sequence in complex with TMPyP4 [17]. The steric restraints imposed by the structure of TMPyP4 are more suited to stacking on base pairs than G-quartets, as observed in this crystal structure but also consistent with the experimental findings of the lack of speci- ficity of TMPyP4 for G-quadruplexes over duplex DNA. The structural reorganisation of one TTA loop to form a di-nucleotide loop and a platform to interact with a second TMPyP4 is in accord with the observation where association with loops in free solution can be expected, which is suggestive of an external binding mode for TMyP4 [42]. The lack of ability of TMPyP4 to exploit the difference in binding surface area between G-quartets and duplex DNA makes it a suboptimal platform for further ligand design.
⦁ NMR structure of cMYC-TMPyP4
Stabilisation of a G-quadruplex in the NHE by TMPyP4 decreases the transcriptional level of MYC [6]. The solution structure of the 24-nucleotide guanine-rich strand of the MYC NHE III1 was the first five-guanine-tract sequence to be reported [17]. In order to improve the spectrum a single guanine to inosine substitution was made at position 10. (Pu24I). The sequence adopts a parallel stranded unimolecular (intra-) topological fold with three propeller loops connecting adjacent parallel strands. Of these, two are single nucleotide loops that connect three G-quartet layers and the third tri-nucleotide loop bridges two G-quartet layers. Another four nucleotide diagonal loop spans across and connects the edges of
the bottom G-quartet. A striking feature of this diagonal loop is the formation of a GeAeG triad. The terminal 30 guanine is inserted back into the G-quartet core while displacing the inosine residue. This is directly reflected in the stronger GeGeGeG G-quartet interactions over GeIeGeG quartet. The three G-quartets are sandwiched by the GeAeG triad on the 30 end and by an AeA pair on the 50 end. The first two residues positioned above the AeA base pair on the 50 end are less defined where as the 30 triad is capped by a stacked adenine.
A comparison of the differences in the chemical shifts between protons of the free and TMPyP4 bound sequences suggest that the ligand stacks over the top G-quartet, while slightly shifted to one corner. The AeA base pair formed in the native structure is not formed. One of the adenines from this base pairing is located between two guanines that contribute to the G-quartet formation; where as the other adenine is positioned between two pyridyl rings. The positive charges in the ligand make interactions with the sugar-phosphate backbone from the top G-quartet and adjoining adenine residues. The two 50 end residues are located above TMPyP4 and thereby forming a sandwich.
⦁ Crystal structure of G-quadruplex-BRACO-19 complex
The 3,6,9-trisubstituted acridine ligand BRACO-19 was the first ligand that was rationally designed, based on the assumption that the substituent in the side chains would occupy each groove in a quadruplex [12]. It exhibited extensive in vitro anticancer activity in
S.M. Haider et al. / Biochimie 93 (2011) 1239e1251 1245
tumour xenografts, which is associated with telomere uncapping [12]. The biological unit in the crystal structure consists of two propeller-loop topology parallel stranded quadruplexes stacked end- to-end in a 50 / 30 orientation. The ligand is sandwiched between the two quadruplexes. The topology is similar to that observed in the native uni- and bimolecular crystal structure of human telomeric
sequence [43] and also to that complexed with TMPyP4 [40], indi- cating that it is the preferred topology for parallel stranded structures. Each quadruplex is bimolecular consisting of three stacked
þ
G-quartets. The interspersing TTA bases form the extended propeller-type loops that join the ends of the G-quartets. The ligand, BRACO-19 is sandwiched between a G-quartet at the 30 end of one quadruplex and TATA terad at the 50 end of the other quadruplex. The stacking is asymmetrical with only two bases from the quartet forming pep stacking interactions with the ligand. The positively charged nitrogen atom in the acridine ring is positioned directly on the central axis of the electronegative channel in one quadruplex. It seems as if a monovalent cation has been displaced and then compensated for by the positively charged nitrogen. The TATA G-quartet at the 50 end is formed from reverse WatsoneCrick base pairing from 50-TA ends of two strands of one quadruplex and the 30-Thymine from a third strand on the other. The G- and TATA G-quartets are inclined by 30◦ and offset with respect to each other. A 30-thymine base is also flipped into the binding site such that it interacts with the side chain amide nitrogen in BRACO-19. The parallel loop has to undergo a considerable conformational change in order to place the thymine in-plane. This interaction has also been observed in the bimolecular antiparallel crystal structures of di- substituted acridine ligands complexed with Oxytricha nova. These conformational changes results in the formation of the binding pocket that are bounded by phosphate groups from the loops and accommodate the pyrrolidino rings at the end of the BRACO-19 side chains. There are two other interactions between the solvent and ligand moiety, with carbonyl oxygen in the side chain of BRACO-19 and flipped thymine and also with the central ring nitrogen of the acridine. Another row of monovalent cations (K ) runs through the central axis of the second quadruplex, and this time terminating at the positively charged nitrogen present in BRACO-19 at the 9- position on the 50 end. The two cationic side chain at 3- and 6- position extends into grooves, located on each side of the G-quartet face. However, there are no direct interactions observed between the ligand and backbone of DNA. The only interactions are solvent-mediated. The phenyl ring at the 9- positions itself into a small hydrophobic pocket formed at the dimeric interface between the ligand and the quadruplex. The grooves in both the quadruplexes are filled with water molecules that resemble DNA duplex minor groove. Thus of the overall eight hydrogen bond donors and accep-
tors, seven form hydrogen bonding interactions, six of which are with solvent molecules and not directly with BRACO-19. Any extension of side chains would extend the charged substituent from the groove regions and thus reduce binding by being unable to participate in water-mediated hydrogen bonding network in the grooves. The crystal structure provides a detailed explanation of the role of anilino group in increasing quadruplex affinity by 10-fold [44]. This group is positioned in a tight narrow pocket that is able to accommodate a planar aromatic ring. Addition of substituent beyond the ring system would lose the tight fit in the binding pocket.
⦁ Crystal structures of G-quadruplexes with napthalene diimide
⦁ Structural data for naphthalene diimide with two diethylamines and two hydroxyl groups
In an effort to rationalise biophysical data on a focused library of naphthalene diimides designed to interact with human telomeric sequences a series of structural investigations were recently
undertaken and partially reported on. In these investigation two human telomeric sequences were successful in providing high resolution crystallographic structural data, crystallising in two forms: an intra- (uni); and inter- (bi) molecular arrangements [39] in complex with a napthalene diimide (L1) ligand containing four side-chains with two dimethylamines and two hydroxyl groups (compound No. 24 [45]). In both crystal forms, the parallel stranded propeller (double chain reversal) loop topology is observed, similar to the native crystal structures [43]. The deoxyribose sugar puck-
ering for G-quartet guanosine residues are C20-endo and their anti-
glycosidic angles lie in 112◦ 9◦ (c), unmodified by ligand binding. The 23-nucleotide intramolecular quadruplex is formed from four human telomeric repeats of the sequence d[TAGGG(TTAGGG)3]. Within the asymmetric unit two independent intramolecular quadruplexes are observed interacting with six naphthalene diimide ligands. The ligands are stacked on both 50 and 30 ends in each quadruplex. The two quadruplexes are then stacked 50e50, end-to-end, such that the arrangement forms a double sandwich (30eligandequadruplexeligande50e50ligandequadruplexeligand- e30). Two other ligands interact each with a TTA loop. The inter- molecular quadruplex is formed from two crystallographically related strands containing two telomeric repeats within the 12-nucleotide repeat sequence d(TAGGGTTAGGGT). In the biologi- cally relevant unit, two ligands directly interact with the intermo- lecular form of quadruplex; one on the 30 G-quartet surface and the other one perpendicular to the quadruplex axis. A crystallographic 2-fold axis sits between the two quadruplexes generating a stack 50e50 interaction in the crystal lattice, however, this time without any ligands sandwiched between them. The third ligand is stacked against the second ligand lying on two 2-fold crystallographic axes. The two quadruplex interacting ligands (L1) are both bisected by 2-fold crystallographic axes, resulting in disordered 2-fold averaged electron density and modelled ligand compounds.
The structures adopted by the loops in both inter- and intra- molecular forms show significant conformational variability and thus highlighting the flexible nature of the TTA connecting loops. Unlike the loops in the native structures where the backbone torsion angles are A-DNA type, the g and z torsion angles in the intra- molecular structure resemble those identified in Z-DNA. As a result of this variation, the bases in the loops are swung out and act as stacking platforms for ligands. In one loop, a thymine is displaced from its native conformation to sandwich a naphthalene diimide ligand between adjacent adenine. A second ligand molecule is stacked between a thymine from one quadruplex loop and an adenine from another. All six loops in the intramolecular structure exhibit some involvement of inter- or intra-stacking interactions. Four of these loops form stacking interactions with two ligands external to the quadruplex core. Similar conformational flexibility of thymine bases in the loops is also observed in the intramolecular structure. Both thymine bases are utilized as stacking platforms for naphthalene diimide ligands. While the N6 and N7 atoms in adenine makes hydrogen bonding interactions with N2 and N3 atoms
of guanine via a GeA mismatch at the terminal 50 end and thus
presenting a A-(GeGeGeG)-A hexad. This hexad arrangement has been observed in the NMR studies of the sequence d(G2AG2AG) [46]. In both structures, the ligand promotes the formation of planar binding surfaces, by the formation of AeT and TeT base pair inter- actions from the TTA loops and 50-TA terminal nucleotides, to generate stable interaction platforms, formed by the conformational
remodelling of the loops [39] (Table 1).
⦁ New structural data for naphthalene diimide dimethylamines and diethylamines ligand complexes
Further structural studies are reported here with two similar ligands, a tetra-substituted naphthalene diimide with four
1246
Table 1
List of structures summarising sequences and ligands.
S.M. Haider et al. / Biochimie 93 (2011) 1239e1251
and 30 surfaces positioned almost centrally to the electro negatively charged central core. The quadruplexes stack 50e50 with two
PDB id Sequence Ligand
symmetry related ligands held between the G-quartets. The 30
1L1H, 3NZ7, 3NYP, 3EUI,
3EUM, 3ET8, 3ERU, 3ES0,
3EQW, 3EM2, 3CE5,
d(GGGGTTTTGGGG) Disubstituted Acridine
G-quartet bound ligand is also interacting with T19*eA2* a dinu- cleotide pair formed from thymine 19 of propeller loop 3 and the terminal 50 residue A2 of adjacent symmetry related molecules.
2A5R d(TGAGGGTGGIGA
GGGTGGGGAAGG)
TMPyP4
The third ligand is almost perpendicular to the helical axis, stacked onto a symmetry related ligand and sandwiched between two
1O0K d(TGGGGT) Daunomycin
2KVY d(TGGGGT) Distamycin-A analogue
2JT7 d(TGGGGT) Distamycin-A
1NZM d(TTAGGGT) RHPS4
2HRI d(TAGGGTTAGGG) TMPyP4
3CE5 d(TAGGGTTAGGGT) Braco19
3CDM/3CCO d(TAGGGTTAGGGTT Napthalene Diimide AGGGTTAGGG)
3MIJ r(UAGGGUUAGGGU) Disubstituted Acridines d(TAGGGTTAGGGTT Napthalene Diimide AGGGTTAGGG)
d(TAGGGTTAGGGT) Napthalene Diimide
dimethylamines (L2) (compound No. 3 [45]) complexed with the intramolecular sequence d[TAGGG(TTAGGG)3] in a new space group and packing arrangement. Second, a tetra-substituted naphthalene diimide with four diethylamines (L3) (compound No. 7) [45] complexed to d(TAGGGTTAGGGT) in a similar space group, but with a reduced cell volume. The 23-nucleotide intra(uni) molecular telomeric sequence folds in complex with three ligands (L2) as a parallel stranded G-quadruplex with three propeller loops (Fig. 5aed). Similar to the previously determined structure, the
naphthalene diimide ligands (L2) (Fig. 4d) are bound on both the 50
WatsoneCrick base paired nucleotides T13*-A8, where again the binding surfaces are generated by crystallographic symmetry (Fig. 6a). Here the normal propeller loop geometry is disrupted for loop 1 separating the nucleotides allowing T7 to rotate into a groove of the quadruplex and hydrogen bond with G4-N2 T7-O4 (2.79 Å) to form an unusual pentad arrangement (Fig. 5d). In all other aspects this quadruplex mimics the previously determined 23 nucleotide naphthalene diimide complex and the unbound native 22mer. The addition of two further positive charges over ligand L1 to a total of four for ligand L2, appears to have little influence in the type of associations of the ligands have and with their abilities to stack onto the quadruplex (Figs. 4d and 5d) G-quartet surfaces. The flexibilities previously observed for the loops containing TTA linking nucleotides appear to be exploited to generate a further set of binding surfaces to accommodated these positively charged ligands.
In contrast the 11-nucleotide inter- bimolecular quadruplex bound to ligand L3 packs into the same space group but a slightly smaller cell volume (Table 2) than the previously determined L1 complex [39]. The reduced unit cell volume reflects the loss of one ligand, previously observed sandwiched between two other ligands
Fig. 4. Structures of quadruplexes in complex with ligands coloured by atom types with VDW surface representation added for the ligands (a) TMPyP4; 2A5R (b) TMPyP4; 2HRI
(c) BRACO19; 3CE5 (d) Napthalene Diimide; 3CDM (e) Acridine; 3MIJ.
S.M. Haider et al. / Biochimie 93 (2011) 1239e1251 1247
Fig. 5. Structure of the telomeric d(TA(GGGTTA)3GGG) DNA/ligand (L2) complex. Surfaces coloured by atom type a) Ligand L2 stacked over the G-tetrad (yellow). b) Space filling representation showing the four charged substituents pointing towards the grooves. c) Opening up of loop 1 to accommodate the ligand binding. The 30◦ twist of the stacked tetrads is highlighted along with the groove. Ligand have been removed for clarity. d) Overall structural arrangement for all components.
positioned central to two intersecting 2-fold crystallographic axes. All other structural aspects of the DNA arrangement in the crystal lattice are almost identical to the previously reported structure [39], forming an intermolecular G-quadruplex structure generated by a 2-fold crystallographic symmetry element on an 11-nucleotide repeat sequence d(TAGGGTTAGGG) (Fig. 6b). The removal of the 30
terminal thymine and shortening the sequence as compared to the previous crystal form does not influence the packing arrangement or change any ligand interactions. The biologically relevant unit stacks 50 to 50 directly onto an adjacent A2eG-quartet surface while on the 30 G-quartet surface has one bound ligand central to the
G-quartet. The change in packing volume and ligand arrangement may be a consequence of having four positive charges on the ligands, thereby limiting self-associations restricting the formation of extended stacked ligand structures. We observe that the tetra- substituted naphthalene diimide diethylamine ligand (L3) is again stacked on a TeT* base pair with all four diethylamine substituents directed towards backbone phosphates. Unfortunately, the majority of the atoms of the ligand four substituents cannot be adequately visualized in the residual electron density maps, indicating signif- icant disorder, far greater than that observed for the L1 bound inter molecular complex. The additional charges appear to limit the
Fig. 6. Cartoon representations of DNA quadruplex/ligand complexes. a) The telomeric d(TA(GGGTTA)3GGG) DNA/ligand (L2) complex interacting with three ligands (purple).
b) The telomeric d(TAGGGTTAGGGT) ligand (L3) complex draw as the biological unit. The position of the crystallographic 2-fold axis to generate the second strand and G-quadruplex unit is shown.
1248 S.M. Haider et al. / Biochimie 93 (2011) 1239e1251
Table 2
Crystallographic data.
Complex Complex
Structure info Sequence d(TAGGGTTAGGGTTAGGGTTAGGG) d(TAGGGTTAGGG)
Ligand Tetra-substituted naphthalene Tetra-substituted naphthalene
diimide dimethylamine (L2) diimide diethylamine (L3)
Solution method Mol. replacement Mol. replacement
Data collection Space group
Unit cell
a, b, c (Å) P 3121
59.00, 59.00, 57.74 P 6422
58.33, 58.33, 40.244
Resolution (Å) a, b, g (◦) 90.0, 90.0, 120.0
51e3.00 90.0, 90.0, 120.0
29e2.70
Rint (%) overall 9.02 (56.0) 6.9 (36.3)
I/s 12.8 (2.10) 10.6 (2.7)
Completeness (%) 99.8 (100.0) 97.5 (100.0)
Redundancy 10.0 (10.6) 4.7 (5.0)
Refinement Resolution (Å) 19.3e3.1 20.0e2.95
No. reflections 2390 920
Rwork/Rfree (%) 25.0/34.0 26.70/39.0
Overall B-factor (Å2) 14 23
RMS deviations Bond-lengths (Å) 0.019 0.016
Bond-angles (◦) 2.64 3.08
Numbers in brackets represent values for the highest resolution shell 3.11e3.00 (23mer) and 2.80 to 2.70 (11mer) Å.
ligands ability to self-association through extensive self-stacking, and so changing the overall packing arrangement.
In conclusion, all three tetra substituted naphthalene diimide ligands (L1, L2, L3) can successfully stack externally on both the 30 and 5 G-quadruplex surfaces without disrupting the propeller geometry, and parallel stranded G-stacked arrangement. These same parallel stranded, propeller loop arrangements for the human
¼
¼
¼ ¼
telomeric sequences have now been observed in three different space groups, four different packing arrangements, with three different sequence bound to naphthalene diimides. However, when stacked on the G-quartets all three naphthalene diimide ligands fail to fully exploit their full hydrogen bonding potential expected from for the various substitutents. In fact, they present very few direct contacts with the quadruplex backbone or groove hydration structure, and display high mobility or disorder for the atoms in the extended side chains of the substituents. For those few observed interactions the ligands do make we observe indirect water- mediated interactions with the sugar-phosphate backbone and the bases. This might imply that the side-chains are also not of optimal length to penetrate the grooves and make specific hydrogen bonding interactions. Modelling of extended side-chains with n 3 to n 5 carbon atoms terminating with bulky N-methylpiperazine groups was suggestive of enhanced affinity to human telomeric DNA with direct interactions with the grooves [47]. Currently reported experimental data show that extending lengths of the naphthalene diimide’s side-chains from n 3 to n 5 has not led to an improvement in affinity for human telomeric sequences although it does alter selectivity [47]. In addition to side-chain length the size of the chromophore is not quite large enough to fully overlap the G- quartets. The situation is different when stacked onto the dinucle- otide base-pairs observed in these crystal structures, as the ligand’s substituents can reach around and make electrostatic interactions with otherwise unavailable backbone phosphate groups.
⦁ Crystal structure of a RNA quadruplexeacridine ligand complex
It has recently been established that the C-rich strand of telo- meric DNA can be transcribed into telomeric RNA termed as TERRA
⦁ and [49]. The 100e1000 base pair r(UUAGGG) repeats may be involved in a number of cellular processes [50] and [51]. In addition, the TERRA sequences are negative regulators of telomerase func- tion [52]. Several proteins have been identified that interact with
TERRA sequences [53e55]. The emerging role of TERRA in telomere biology suggests that they can be used as therapeutic targets and ligands with high affinity and selectivity can be designed that can differentiate between DNA and RNA quadruplexes [2,56,57]. Both crystallographic and NMR analysis have shown that short TERRA sequences can form stable parallel stranded bimolecular (inter-) G-quadruplex structures with propeller loop topology, similar to that observed for telomeric DNA [58] and [59]. This is primarily
due to the presence of 20-OH groups together with the preference of ribose sugars to adopt a C30-endo pucker plays a significant role in stabilising the parallel arrangement [9].
The crystal structure of a TERRA RNA quadruplex in complex with 3,6 di-substituted acridine with triazole-phenyl-diethylamine side chains has been reported by Neidle and co-workers [60]. The ligand was originally designed to bind DNA quadruplexes and shows selectivity for telomeric quadruplex over those found in the promoter sequence of the c-kit gene [14]. The biological unit contains two strands of r(UAGGGUUAGGGU) complexed with two molecules of triazole acridine formed around a two-fold crystal- lographic axis. The G-quartets are stacked with the chain reversal UUA propeller-like loops linking the top to the bottom.
Despite having a high degree of internal conformational flexibility, the two triazole acridine molecules are stacked in a flat coplanar arrangement on the terminal 50 G-quartet, with only a slight deviation from planarity observed for the triazole-phenyl link where the side chain curves up towards the 50 face of a symmetry related biological unit. Two complete intermolecular quadruplexes are stacked 50e50 sandwiching two layers of ligands between them. The central nitrogen atoms in the acridine chromophores are pointing away from the centre of the G-quartet. The two ligand molecules are laterally displaced across the crys- tallographic two-fold axis and around the central electronegative channel. There are no direct contacts between the two ligands and the predominant means of stabilisation is via pep stacking inter- actions. There is a substantial reorganisation of loop conformation where the adenines from the UUA propeller-loop and the 50 UA sequence flip and align within the plane of the 50 G-quartet forming a novel pseudo four-fold G4A4 octet. The core of the octet is the 50 G-quartet. This further extends the surface area for p-stacking of triazole rings of the ligand. The formation of the G4A4 octet is a direct consequence of the hydrogen bonding in the loops, some of which is RNA specific. In addition, cross-loop hydrogen bonds help in maintaining the conformation of the loop. The N1 atom from
S.M. Haider et al. / Biochimie 93 (2011) 1239e1251 1249
adenine makes interactions with the 20OH group of the adjacent guanosine. A water-mediated interaction is formed between N2 and N3 atoms of this guanine. This pattern is repeated as a result of the two-fold symmetry.
The aromatic groups in the ligands are stacked on purine bases except for one phenyl ring that is not positioned above the purine octet platform. One end of the ligand is pseudo-intercalated between an adenine and the terminal uracil base while at the other end this overlap is minimal. Instead there are several inter- actions including a weak hydrogen bond between the amide carbonyl group from the ligand side chain and the N3 nitrogen
atom of the 50 terminal uracil residue. Thus it is evident that the
structural reorganisation observed in the loop is driven by the need to maximise interactions with the ligands and is stabilised by both stacking interactions with the planar ligands and also by hydrogen bonds made by 20OH groups. Although there are no direct inter- actions made between the ligand and the quadruplex, selectivity is achieved in a more indirect manner.
⦁ Conclusions
It is telling that very little structural data is available for G-quadruplex/ligand complexes. Many high affinity ligands have been synthesised and studied but it is proving elusive to structurally determine ligand interactions in a routine manner. Currently the application of crystallographic techniques is providing us with the most comprehensive picture of ligand binding interactions, however the validity of the model is in some ways compromised by crystal packing forces within a crystalline lattice. Even within these constraints, it is clear that the stacked G-quadruplex core remains intact with little scope for the intercalation of planar ligands. Planar
aromatic p stacking external to the G-quartet quartets dominates ligand binding on either the 30 or 50 interfaces. Positively charged
groups, as seen with the protonated nitrogen of the acridine ring, can exploit the central electronegative channel to enhance binding. While the diagonal and propeller loops are sufficiently adaptable to create many new surfaces for ligand binding. The propeller loops can use both base stacking to form stable planar platforms and hydrogen bonding to form extended surfaces. These form from AeT base pairing and TeT base paring, while further extended platforms come from adenine interactions with the G-quartets. For example the G4A4 octet pairings in the RNA acridine complex, stabilized in part
by the ribose 20OH of the guanosine. The NMR derived structures
reveal distamycin bound within the groove confirming the potential to selectively target the grooves, however it would also be inter- esting to understand the role of different connecting loops on groove binding. The hydration structure around the quadruplexes has yet to be fully explored, although ordered spines of hydration are clearly important for stabilising groove and loop arrangements, the role of counter ions such as potassium may yet play a role in loop stability.
⦁ Methods
The HPLC-purified DNA sequence d[TAGGGTTAGGG] (11mer) was purchased from Eurogentec (Belgium) while the DNA sequence d[TAGGG(TTAGGG)3] (23mer) was prepared on an Applied Biosystems DNA synthesizer using solid-phase b-cyanoethylphos- phoramidite chemistry, described elsewhere. Both DNA prepara- tions were first dissolved at 3 mM into solutions containing 20 mM potassium cacodylate buffer at pH 6.5, and 50 mM potassium chloride heated to 363 K before annealing by slow cooling to room temperature. The analytically-pure tetra-substituted naphthalene diimide ligand compounds1 No. 3 with four dimethylamines (L2) and with four diethylamines No. 7 (L3 [45]) were prepared to a final stock solution concentration of 10 mM in 100% DMSO and 40 mM
HCl. Crystals of the 23-mer/ligand complex were grown by vapor diffusion where 0.6 mL of the premixed drop solution containing
× ×
0.7 M ammonium sulphate, 50 mM magnesium chloride, 100 mM sodium chloride, 50 mM potassium chloride, 50 mM potassium cacodylate (pH 6.5) was added to 0.6 mL of a preformed complex of quadruplex DNA at a concentration of 1.6 mM and the ligand L1 at 1.6 mM. The drop solution was equilibrated against a 0.9 M ammonium sulphate well solution at 283◦ K. Crystals grew as blue rectangular flat plates with dimensions of 0.01 0.1 0.05 mm3 and were cryo-protected using glycerol (25% w/v water) in addition to the initial in-drop crystallization conditions. Crystals of the 12- mer/ligand complex were also grown by vapor diffusion methods at 283◦ K where 1 mL of a preformed complex of quadruplex DNA and ligand L2 at 2 mM in a 1:1 molar ratio was added to 1 mL of well solution to generate a 50% gradient. The well solution contained 300 mM sodium chloride, 50 mM sodium cacodylate (pH 6.5), and 30% v/v MPD and gave blue crystals after 1 month with dimensions of 0.1 × 0.2 × 0.05 mm3.
⦁ Data collection and structure determination
Data from flash-frozen crystals for the 23mer were collected at Diamond Light Source on beam line IO4, at wavelength 0.9785 Å, and for the 11mer data were collected at the Diamond Light Source (DLS-0117-05) on beam line IO4 at wavelength 0.9702 Å 100 K. Processing and data reduction were carried out using the D*TREK part of the CrystalClear software package [61]. The structure of the intramolecular quadruplex containing four human telomeric repeats was solved by molecular replacement using the program PHASER from the CCP4 package [62], utilizing the parallel-stranded intramolecular ligand quadruplex, PDB ID 3CDM with the ligands and waters removed, as the search model. The structure of the bimolecular quadruplex containing two human telomeric repeats in each strand was also solved by molecular replacement using the program PHASER [63], with a search model PDB ID 3CCO, stripped of waters and ligands. Model building and refinement cycles were performed using REFMAC5 [64] and COOT [65] for both complexes.
Initial calculated electron density maps revealed strong density on both the 30 and 50 faces for the 23mer complex, along with strong disc like density in a TTA loop. The shapes of the residual electron density matched the core tetra-substituted naphthalene diimide ligands and so were introduced into the final model. The ligand modified TTA loop was re-fitted into Fo-Fc omit maps
and refined, the ligand was then modeled to sit above a T*:A base pair partly formed from a symmetry related molecule. The final model contains one intramolecular parallel stranded G-quadruplex, three ligands (L2), and two potassium ions between the three stacked G-quartets. The relatively low resolution crystallographic data available did not allow for the fitting of the hydration struc- ture. The relatively high R-factor and R-Free is due to weak diffraction and the disorder of the ligands.
The 11mer bimolecular quadruplex sits adjacent to a crystallo- graphic 2-fold axis, generating a biologically relevant parallel stranded G-quadruplex with potassium ions in the central channel on the 2-fold axis. The TTA loops nucleotides were removed from the molecular replacement solution model and rebuilt into the omit map electron density. After placement and model building and refinement all three loop nucleotides are observed to be directly involved in packing interactions in a similar arrangement to the
search model used PDB ID 3CCO. Disc like residual density on the 30
guanine G-quartet surface was modelled with a core naphthalene diimide fragment and refined. The ligand is disordered as it sits on the 2-fold axis making the placement of the four substituent’s side chains problematic. Overall, the crystal lattice has contracted in volume from the initial search model dimensions that results in the
1250 S.M. Haider et al. / Biochimie 93 (2011) 1239e1251
exclusion of one of the ligands that was positioned on two inter- secting 2-folds. The second ligand in the crystal lattice interacts with the TTA loops by stacking on a T*-T base pair partly formed from a symmetry related molecule. The final biological model contains a bimolecular parallel stranded G-quadruplex, two ligands (L3), and two potassium ions between the three stacked G-quartets. The high R-factor and R-free can be attributed to the disordered ligands positioned on crystallographic symmetry elements. The scattering from the ligands contributes strongly to the overall total scattering within the asymmetric unit.
Acknowledgements
Deeksha Munnur for assistance in crystallizations of the 23mer DNA complexed with the tetra-substituted naphthalene diimide dimethylamine (L2), and Tony Reszka for solid phase DNA synthesis; Francisco Cuenca for the preparation of the ligands used in these studies; Gavin Collie, Nora Cronin and the staff at ID04 beam line for assisting in data collection at Diamond Light Source facilities. This work was supported by Cancer Research UK (Programme Grant No. C129/A4489).
References
⦁ M. Gellert, M.N. Lipsett, D.R. Davies, Helix formation by guanylic acid, Proc. Natl. Acad. Sci. U.S.A. 48 (1962) 2013e2018.
⦁ A. De Cian, J. Gros, A. Guedin, M. Haddi, S. Lyonnais, L. Guittat, J.F. Riou,
C. Trentesaux, B. Sacca, L. Lacroix, P. Alberti, J.L. Mergny, DNA and RNA quadruplex ligands, Nucleic Acids Symp. Ser. (Oxf.) (2008) 7e8.
⦁ A. De Cian, L. Lacroix, C. Douarre, N. Temime-Smaali, C. Trentesaux, J.F. Riou,
J.L. Mergny, Targeting telomeres and telomerase, Biochimie 90 (2008) 131e155.
⦁ T.M. Ou, Y.J. Lu, J.H. Tan, Z.S. Huang, K.Y. Wong, L.Q. Gu, G-quadruplexes: targets in anticancer drug design, ChemMedChem 3 (2008) 690e713.
⦁ L.H. Hurley, D.D. Von Hoff, A. Siddiqui-Jain, D. Yang, Drug targeting of the c-MYC promoter to repress gene expression via a G-quadruplex silencer element, Semin. Oncol. 33 (2006) 498e512.
⦁ A. Siddiqui-Jain, C.L. Grand, D.J. Bearss, L.H. Hurley, Direct evidence for a G-quadruplex in a promoter region and its targeting with a small molecule to repress c-MYC transcription, Proc. Natl. Acad. Sci. U.S.A. 99 (2002) 11593e11598.
⦁ A. Porro, S. Feuerhahn, P. Reichenbach, J. Lingner, Molecular dissection of telomeric repeat-containing RNA biogenesis unveils the presence of distinct and multiple regulatory pathways, Mol. Cell Biol. 20 (2010) 4808e4817.
⦁ J. Plavec, W. Tong, J. Chattopadhyaya, How do the gauche and anomeric effects drive the pseudorotational equilibrium of the pentofuranose moiety of nucleosides? J. Am. Chem. Soc. 115 (1993) 9734e9746.
⦁ M. Webba da Silva, M. Trajkovski, Y. Sannohe, N. Ma’ani Hessari, H. Sugiyama,
J. Plavec, Design of a G-quadruplex topology through glycosidic bond angles, Angew. Chem. Int. Ed. Engl. 48 (2009) 9167e9170.
⦁ S.M. Haider, G.N. Parkinson, S. Neidle, Structure of a G-quadruplexeligand complex, J. Mol. Biol. 326 (2003) 117e125.
⦁ R.J. Harrison, S.M. Gowan, L.R. Kelland, S. Neidle, Human telomerase inhibition by substituted acridine derivatives, Bioorg. Med. Chem. Lett. 9 (1999) 2463e2468.
⦁ M. Read, R.J. Harrison, B. Romagnoli, F.A. Tanious, S.H. Gowan, A.P. Reszka,
W.D. Wilson, L.R. Kelland, S. Neidle, Structure-based design of selective and potent G quadruplex-mediated telomerase inhibitors, Proc. Natl. Acad. Sci. U.S.A. 98 (2001) 4844e4849.
⦁ M.A. Read, A.A. Wood, J.R. Harrison, S.M. Gowan, L.R. Kelland, H.S. Dosanjh,
S. Neidle, Molecular modeling studies on G-quadruplex complexes of telomerase inhibitors: structure-activity relationships, J. Med. Chem. 42 (1999) 4538e4546.
⦁ S. Sparapani, S.M. Haider, F. Doria, M. Gunaratnam, S. Neidle, Rational design of acridine-based ligands with selectivity for human telomeric quadruplexes, J. Am. Chem. Soc. 132 (2010) 12263e12272.
⦁ S. Haider, G.N. Parkinson, S. Neidle, Crystal structure of the potassium form of an Oxytricha nova G-quadruplex, J. Mol. Biol. 320 (2002) 189e200.
⦁ N.H. Campbell, M. Patel, A.B. Tofa, R. Ghosh, G.N. Parkinson, S. Neidle, Selec- tivity in ligand recognition of G-quadruplex loops, Biochemistry 48 (2009) 1675e1680.
⦁ A.T. Phan, V. Kuryavyi, H.Y. Gaw, D.J. Patel, Small-molecule interaction with a five-guanine-tract G-quadruplex structure from the human MYC promoter, Nat. Chem. Biol. 1 (2005) 167e173.
⦁ N.H. Campbell, D.L. Smith, A.P. Reszka, S. Neidle, D. O’Hagan, Fluorine in medicinal chemistry: beta-fluorination of peripheral pyrrolidines attached to acridine ligands affects their interactions with G-quadruplex DNA, Org. Biomol. Chem. (2011).
⦁ E. Gavathiotis, R.A. Heald, M.F. Stevens, M.S. Searle, Drug recognition and stabilisation of the parallel-stranded DNA quadruplex d(TTAGGGT)4 con- taining the human telomeric repeat, J. Mol. Biol. 334 (2003) 25e36.
⦁ L.W. Elmore, C.W. Rehder, X. Di, P.A. McChesney, C.K. Jackson-Cook, D.A. Gewirtz,
S.E. Holt, Adriamycin-induced senescence in breast tumor cells involves func- tional p53 and telomere dysfunction, J. Biol. Chem. 277 (2002) 35509e35515.
⦁ G.R. Clark, P.D. Pytel, C.J. Squire, S. Neidle, Structure of the first parallel DNA quadruplex-drug complex, J. Am. Chem. Soc. 125 (2003) 4066e4067.
⦁ G. Laughlan, A.I. Murchie, D.G. Norman, M.H. Moore, P.C. Moody, D.M. Lilley,
B. Luisi, The high-resolution crystal structure of a parallel-stranded guanine tetraplex, Science 265 (1994) 520e524.
⦁ K. Phillips, Z. Dauter, A.I. Murchie, D.M. Lilley, B. Luisi, The crystal structure of a parallel-stranded guanine tetraplex at 0.95 A resolution, J. Mol. Biol. 273 (1997) 171e182.
⦁ P. Sket, J. Plavec, Tetramolecular DNA quadruplexes in solution: insights into structural diversity and cation movement, J. Am. Chem. Soc. 132 (2010) 12724e12732.
⦁ X. Chen, B. Ramakrishnan, M. Sundaralingam, Crystal structures of the side-by-side binding of distamycin to AT-containing DNA octamers d(ICITA- CIC) and d(ICATATIC), J. Mol. Biol. 267 (1997) 1157e1170.
⦁ J.G. Pelton, D.E. Wemmer, Structural characterization of a 2:1 distamycin A.d(CGCAAATTGGC) complex by two-dimensional NMR, Proc. Natl. Acad. Sci. U.S.A. 86 (1989) 5723e5727.
⦁ M.J. Cocco, L.A. Hanakahi, M.D. Huber, N. Maizels, Specific interactions of distamycin with G-quadruplex DNA, Nucleic Acids Res. 31 (2003) 2944e2951.
⦁ M.J. Moore, F. Cuenca, M. Searcey, S. Neidle, Synthesis of distamycin A poly- amides targeting G-quadruplex DNA, Org. Biomol. Chem. 4 (2006) 3479e3488.
⦁ L. Martino, A. Virno, B. Pagano, A. Virgilio, S. Di Micco, A. Galeone, C. Giancola,
G. Bifulco, L. Mayol, A. Randazzo, Structural and thermodynamic studies of the interaction of distamycin A with the parallel quadruplex structure [d(TGGGGT)]4, J. Am. Chem. Soc. 129 (2007) 16048e16056.
⦁ S. Cosconati, L. Marinelli, R. Trotta, A. Virno, S. De Tito, R. Romagnoli, B. Pagano,
V. Limongelli, C. Giancola, P.G. Baraldi, L. Mayol, E. Novellino, A. Randazzo, Structural and conformational requisites in DNA quadruplex groove binding: another piece to the puzzle, J. Am. Chem. Soc. 132 (2010) 6425e6433.
⦁ J.B. Chaires, A thermodynamic signature for drug-DNA binding mode, Arch. Biochem. Biophys. 453 (2006) 26e31.
⦁ D.F. Shi, R.T. Wheelhouse, D. Sun, L.H. Hurley, Quadruplex-interactive agents as telomerase inhibitors: synthesis of porphyrins and structure-activity rela- tionship for the inhibition of telomerase, J. Med. Chem. 44 (2001) 4509e4523.
⦁ J. Ren, J.B. Chaires, Sequence and structural selectivity of nucleic acid binding ligands, Biochemistry 38 (1999) 16067e16075.
⦁ P. Wang, L. Ren, H. He, F. Liang, X. Zhou, Z. Tan, A phenol quaternary ammonium porphyrin as a potent telomerase inhibitor by selective interac- tion with quadruplex DNA, Chembiochem 7 (2006) 1155e1159.
⦁ L.H. Hurley, R.T. Wheelhouse, D. Sun, S.M. Kerwin, M. Salazar, O.Y. Fedoroff,
F.X. Han, H. Han, E. Izbicka, D.D. Von Hoff, G-quadruplexes as targets for drug design, Pharmacol. Ther. 85 (2000) 141e158.
⦁ S.Y. Rha, E. Izbicka, R. Lawrence, K. Davidson, D. Sun, M.P. Moyer, G.D. Roodman,
L. Hurley, D. Von Hoff, Effect of telomere and telomerase interactive agents on human tumor and normal cell lines, Clin. Cancer Res. 6 (2000) 987e993.
⦁ E. Izbicka, D. Nishioka, V. Marcell, E. Raymond, K.K. Davidson, R.A. Lawrence,
R.T. Wheelhouse, L.H. Hurley, R.S. Wu, D.D. Von Hoff, Telomere-interactive agents affect proliferation rates and induce chromosomal destabilization in sea urchin embryos, Anticancer Drug Des. 14 (1999) 355e365.
⦁ S. Cogoi, L.E. Xodo, G-quadruplex formation within the promoter of the KRAS proto- oncogene and its effect on transcription, Nucleic Acids Res. 34 (2006) 2536e2549.
⦁ G.N. Parkinson, F. Cuenca, S. Neidle, Topology conservation and loop flexibility in quadruplex-drug recognition: crystal structures of inter- and intramolecular telomeric DNA quadruplex-drug complexes, J. Mol. Biol. 381 (2008) 1145e1156.
⦁ G.N. Parkinson, R. Ghosh, S. Neidle, Structural basis for binding of porphyrin to human telomeres, Biochemistry 46 (2007) 2390e2397.
⦁ M.Y. Kim, M. Gleason-Guzman, E. Izbicka, D. Nishioka, L.H. Hurley, The different biological effects of telomestatin and TMPyP4 can be attributed to their selectivity for interaction with intramolecular or intermolecular G-quadruplex structures, Cancer Res. 63 (2003) 3247e3256.
⦁ T. Yamashita, T. Uno, Y. Ishikawa, Stabilization of guanine quadruplex DNA by the binding of porphyrins with cationic side arms, Bioorg. Med. Chem. 13 (2005) 2423e2430.
⦁ G.N. Parkinson, M.P. Lee, S. Neidle, Crystal structure of parallel quadruplexes from human telomeric DNA, Nature 417 (2002) 876e880.
⦁ M.J. Moore, C.M. Schultes, J. Cuesta, F. Cuenca, M. Gunaratnam, F.A. Tanious,
W.D. Wilson, S. Neidle, Trisubstituted acridines as G-quadruplex telomere targeting agents. Effects of extensions of the 3,6- and 9-side chains on quadruplex binding, telomerase activity, and cell proliferation, J. Med. Chem. 49 (2006) 582e599.
⦁ F. Cuenca, O. Greciano, M. Gunaratnam, S. Haider, D. Munnur, R. Nanjunda,
W.D. Wilson, S. Neidle, Tri- and tetra-substituted naphthalene diimides as potent G-quadruplex ligands, Bioorg. Med. Chem. Lett. 18 (2008) 1668e1673.
⦁ A. Kettani, A. Gorin, A. Majumdar, T. Hermann, E. Skripkin, H. Zhao, R. Jones,
D.J. Patel, A dimeric DNA interface stabilized by stacked A. (G.G.G.G).A hexads and coordinated monovalent cations, J. Mol. Biol. 297 (2000) 627e644.
⦁ S. Hampel, A. Sidibe, M. Gunaratnam, J.F. Riou, S. Neidle, Tetrasubstituted naphthalene diimide ligands with selectivity for telomeric G-quadruplexes and cancer cells, Bioorg. Med. Chem. Lett. 20 (2010) 6459e6463.
S.M. Haider et al. / Biochimie 93 (2011) 1239e1251 1251
⦁ C.M. Azzalin, P. Reichenbach, L. Khoriauli, E. Giulotto, J. Lingner, Telomeric repeat containing RNA and RNA surveillance factors at mammalian chromo- some ends, Science 318 (2007) 798e801.
⦁ S. Schoeftner, M.A. Blasco, Developmentally regulated transcription of mammalian telomeres by DNA-dependent RNA polymerase II, Nat. Cell Biol. 10 (2008) 228e236.
⦁ C. Caslini, J.A. Connelly, A. Serna, D. Broccoli, J.L. Hess, MLL associates with telomeres and regulates telomeric repeat-containing RNA transcription, Mol. Cell Biol. 29 (2009) 4519e4526.
⦁ B. Luke, J. Lingner, TERRA: telomeric repeat-containing RNA, EMBO J. 28 (2009) 2503e2510.
⦁ S. Redon, P. Reichenbach, J. Lingner, The non-coding RNA TERRA is a natural ligand and direct inhibitor of human telomerase, Nucleic Acids Res. 38 (2010) 5797e5806.
⦁ Z. Deng, A.E. Campbell, P.M. Lieberman, TERRA, CpG methylation and telomere heterochromatin: lessons from ICF syndrome cells, Cell Cycle 9 (2010) 69e74.
⦁ Z. Deng, J. Norseen, A. Wiedmer, H. Riethman, P.M. Lieberman, TERRA RNA binding to TRF2 facilitates heterochromatin formation and ORC recruitment at telomeres, Mol. Cell 35 (2009) 403e413.
⦁ S. Feuerhahn, N. Iglesias, A. Panza, A. Porro, J. Lingner, TERRA biogenesis, turnover and implications for function, FEBS Lett. 584 (2010) 3812e3818.
⦁ G. Collie, A.P. Reszka, S.M. Haider, V. Gabelica, G.N. Parkinson, S. Neidle, Selectivity in small molecule binding to human telomeric RNA and DNA quadruplexes, Chem. Commun. (Camb.) (2009) 7482e7484.
⦁ S.G. Rzuczek, D.S. Pilch, A. Liu, L. Liu, E.J. LaVoie, J.E. Rice, Macrocyclic pyridyl polyoxazoles: selective RNA and DNA G-quadruplex ligands as antitumor agents, J. Med. Chem. 53 (2010) 3632e3644.
⦁ G.W. Collie, S.M. Haider, S. Neidle, G.N. Parkinson, A crystallographic and modelling study of a human telomeric RNA (TERRA) quadruplex, Nucleic Acids Res. 38 (2010) 5569e5580.
⦁ þ
⦁ H. Martadinata, A.T. Phan, Structure of propeller-type parallel-stranded RNA G-quadruplexes, formed by human telomeric RNA sequences in K solution, J. Am. Chem. Soc. 131 (2009) 2570e2578.
⦁ G.W. Collie, S. Sparapani, G.N. Parkinson, S. Neidle, Structural basis of telomeric RNA quadruplex-acridine ligand recognition, J. Am. Chem. Soc. 133 (2011) 2721e2728.
⦁ CrystalClearTM, An integrated program for the collection and processing of area detector data, Rigaku Corporation.
⦁ N. Collaborative Computational Project, The CCP4 suite: programs for protein crystallography, Acta Crystallogr. D50 (1994) 760e763.
⦁ A.J. McCoy, R.W. Grosse-Kunstleve, P.D. Adams, M.D. Winn, L.C. Storoni,
R.J. Read, Phaser crystallographic software, J. Appl. Crystallogr. 40 (2007) 658e674.
⦁ G.N. Murshudov, A.A. Vagin, E.J. Dodson, Refinement of macromolecular structures by the maximum-likelihood method, Acta Crystallogr. D Biol. Crystallogr. 53 (1997) 240e255.
⦁ P. Emsley, K. Cowtan, Coot: model-building tools for molecular graphics, Acta Crystallogr. D Biol. Crystallogr. 60 (2004) 2126e2132.RHPS 4