Novel Insights on the DNA Interaction of Calicheamicin c1I Claudia Sissi,1 Stefano Moro,1 Donald M. Crothers2 1
Department of Pharmaceutical and Pharmacological Sciences, University of Padova, Via Marzolo 5, 35131 Padova, Italy
2
Chemistry Department, Yale University, 225 Prospect Street, New Haven, CT 06511
Received 30 September 2014; accepted 13 November 2014 Published online 20 November 2014 in Wiley Online Library (wileyonlinelibrary.com). DOI 10.1002/bip.22591
INTRODUCTION ABSTRACT: Calicheamicin g1 I (Cal) is a unique molecule in which a DNA binding motif (aryl-tetrasaccharide) is linked to a DNA cleaving moiety (calicheamicinone). The hallmark of this natural product rests in the impressive optimization of these two mechanisms leading to a drug that is extremely efficient in cleaving DNA at well-defined sites. However, the relative contributions of these two structurally distinct domains to the overall process have not been fully elucidated yet. Here, we used different experimental approaches to better dissect the role of the aryltetrasaccharide and the enediyne moieties in the DNA sequence selective binding step as well as the in the cleavage reaction. Our results highlight the remarkable cooperation of the two components in producing an amazing molecular machine. The herein presented molecular details of this concerted mechanism of action can be further applied to rationally design more druggable comC 2014 Wiley Periodicals, Inc. Biopolymers 103: pounds. V
449–459, 2015. Keywords: calicheamicin g1I; DNA binding; DNA cleavage
This article was originally published online as an accepted preprint. The “Published Online” date corresponds to the preprint version. You can request a copy of any preprints from the past two calendar years by emailing the Biopolymers editorial office at
[email protected]. Correspondence to: Claudia Sissi; e-mail:
[email protected] Contract grant sponsor: AIRC C 2014 Wiley Periodicals, Inc. V
Biopolymers Volume 103 / Number 8
fter its discovery in 1987, Calicheamicin c1 I (Cal) attracted the interest of a wide part of the scientific community with alternating periods of enthusiasm and frustration (Figure 1). This compound was initially explored for its remarkable cytotoxic activity.1,2 However, this property turned out to be hardly tunable, thus making the drug apparently not suitable for clinical applications. Subsequently, its fusion with an antibody represented the proof-of-concept of the pharmacological potential of bioconjugation.3,4 Indeed, a humanized antibody directed toward CD33 antigen covalently linked to a N-acetyl dimethyl hydrazide derivative of Cal, succeeded in directing the toxic agent to the malignant cells and in promoting its uptake.5 The transfer occurs through lysosomes, where the drug is released to finally reach the nucleus.6,7 Consistently, in 2000, gemtuzumab ozogamicin (Mylotarg) was the first approved bioconjugate product: in particular, according to the early clinical studies, it received accelerated marketing approval for the treatment of acute myeloid leukemia (AML).8 Unfortunately, the product was voluntarily withdrawn from U. S. and European markets due to limited efficiency that emerged in subsequent Phase III clinical trials. Nevertheless, novel studies seem to suggest that the history of this product is not over.9 Behind its clinical fate, the chemistry of Cal represents a sort of amusement park for organic chemists. This molecule is composed by two main moieties: an aryl-oligosaccharide tail composed of four deoxysugars and a iodo-thiobenzoate and a bicyclic aglycone (calicheamicinone) containing an enediyne group.10 Although these domains cooperate to the overall mechanism of action, their different chemical identities reflect different functional roles. The biological activity of the whole drug is exerted at DNA level where the carbohydrate tail of Cal recognizes well defined sequences (preferentially homopyrimidine strands) through its insertion into the minor groove.11 Permanent DNA damage is then produced by the enediyne.12 Indeed, the bicyclic ring is functionalized with a methyl trisulfide moiety that, upon bioreductive activation,
A
449
450
Sissi et al.
FIGURE 1 Chemical structures of Calicheamicin c1 I and related derivatives used in this work.
undergoes an intramolecular Michael addition to the adjacent a,b-unsatured ketone. At this point, a Bergman reaction generates a highly reactive benzenoid biradical which removes two hydrogens from the nucleic acid chain. The abstraction occurs preferentially at the C-5’ of the pyrimidine strand and at the C4’ position of the complementary one leading to a double strand cleavage.13,14 As the reactive group is located in the minor groove, the two DNA breaks are shifted one to the other by three base pairs. The elucidation of this mechanism required extensive work which was exploited mainly in the 90s. A great contribution to fulfill this task came from a huge synthetic work.15 Indeed, for organic chemists, the total synthesis of Cal represented a fascinating challenge and two groups succeeded in performing it: Nicolaou’s and Danishefsky’s groups.16,17 Behind the great significance that their work represented for the whole chemistry community, an additional value of their efforts derived from the synthesis of defined Cal fragments (in particular the carbohydrate tail and the enediyne containing moiety) as well as of several derivatives. It was the critical use of these compounds in controlled experimental conditions that provided the actual available knowledge of the mechanism that this molecular machine uses to bind and cleave DNA in a such concerted and efficient fashion.18 In
the hand of expert molecular biologists, chemists and biophysics, these synthetic products allowed to exactly map the interaction site19–21 and to confirm that the carbohydrate tail of Cal is able to recognize the same sequences as the whole drug.22 At the same time, Cal was found to exhibit a higher DNA binding affinity when compared to its carbohydrate derivative thus supporting an active interaction of the aglycone with the nucleic acid.23 Finally, DNA has been confirmed to be actively involved in the binding process. Indeed, not only its sequence and shape are recognized by the drug, but its intrinsic flexibility is a critical requirement to allow efficient binding at selected sites.24,25 All these pieces of information are now taken into consideration in order to produce novel bioconjugates with more proficient pharmacological profiles. An example is provided by studies concerning the properties of novel antibodybioconjugates in which the use of synthetic Cal derivatives, more versatile for conjugation purposes and in terms of activation in biological environment, is explored. Among them, Cal h (Figure 1) is currently used in novel bioconjugates designed for targeted therapy of neoroblastoma and leukemia.26,27 In this work, we provide our ultimate data acquired on Cal in order to highlight the contribution of each single structural domain to the DNA binding and damage processes. In Biopolymers
Calicheamicin cI1-DNA Interaction
particular we addressed the energetic contributions of the saccharide tail and the relevance of its localization with reference to the reactive domain. To fulfill this task we used the derivatives reported in Figure 1: they represent a panel of compounds in which selected combinations of conserved and modified fragments of the natural product were realized.
MATERIALS AND METHODS Calicheamicin c1 I , the aryl-tetrasaccharide as well as Cal h, NAT a and UNN a were prepared as previously described.28 Stock solutions (1 mg/ml) were prepared in ethyl acetate and stored at 220 C. Immediately before use, the required amount was dried under vacuum and dissolved in 20% ethanol (calicheamicin) or 10% tetrahydrofuran (carbohydrate domain).
451
in 50 mM Tris, pH 8.0, 2% EtOH. The resulting solutions were divided into two aliquots: one was added of DTT to a 2 mM final concentration, on the second one the same volume of water was added. After variable time (0–30 min) at the required temperature, reaction volumes corresponding to 0.1 mg of plasmid were collected, cooled down on ice, added with 1.5 lL of gel loading buffer (30% glycerol, 10 mM Tris, 1 mM EDTA, 0.1% xylene cyanol, 0.1% bromphenol blue), and loaded on 1% agarose gel in 1X TAE (4 mM Tris, 4 mM acetic acid, 0.1 mM EDTA). Gels were run at 8 V cm21 and the reaction products were visualized by ethidium bromide staining and quantified. To assess the presence of apurinic sites, before loading, the reaction mixtures resulting after 5 min incubation, were added of putrescin (100 mM final concentration). The samples were further incubated for at 37 C for 1.5 h. Reaction products were then resolved as above described.
Isothermal Titration Calorimetry (ITC) DNA Fragments The required DNA sequences were synthesized on an Applied Biosystem Synthesizer and purified by denaturating polyacrylamide gel (10– 15% according to the oligonucleotide size) in 1X TBE (89 mM Tris, 89 mM boric acid, 2 mM Na2EDTA). After purification, DNA concentration was spectroscopically determined. Prior to use, hairpins were annealed in 50 mM Tris, pH 8.0 by heating the solutions for 5 min at 90 C and then rapidly cooling them in ice. For ITC titrations, they were further dialyzed against the working buffer. The double stranded sequences were used as uniquely 3’ or 5’end-labeled DNA. In order to obtain 5’-labeled fragments, the purified top strand was incubated for 30 min at 37 C with [c-32P]ATP in the presence of T4-Kinase (NEB in the required buffer). The reaction product was then annealed o.n. to the complementary one in 50 mM Tris, pH 8.0. For 3’-labeling, an equimolar solution of the two complementary strands was annealed o. n. in 50 mM Tris, pH 8.0. Then, DNA was incubated with [a-32P]ATP and Klenow DNA polymerase fragment (NEB). The double stranded fragments were purified on an 8% native PAGE in 1X TBE prior to use.
ITC titrations were performed on a Omega ITC instrument (MicroCal, Inc) previously equilibrated for 12 h at the working temperature. Working solutions were prepared in 50 mM Tris, pH 8.0, 2% THF and degassed for 15 min prior to use. The measurement cell was filled with 1.234 ml solution of aryl-tetrasaccharide (40–50 mM). After thermic equilibration under stirring conditions, 25 injections (9 ml each) of previously folded DNA hairpin (0.4–0.8 mM, strand concentration) were performed. The injection time was 15 s and the spacing between each injection was 5 min. At least three titrations were performed for each experimental condition. Before data analysis, raw data were corrected for the heat of dilution of the complex. Heats were integrated and binding parameters were calculated according to a “one binding site” model using Origin Software. Data analysis provides DH (reaction enthalpy change, kcal mol21), Ka (binding constant, M21), and n (number of bound ligands) whereas the Gibbs energy and the entropic contribution were calculated using the relationships DG 5 2RT ln Ka and DG 5 DH 2 TDS, respectively.
Molecular Modeling Study Double Stranded DNA Cleavage and Sequencing Reactions Uniquely end-labeled DNA (10,000 cpm) was dissolved in 50 mM TRIS, pH 5 8.0, 2% EtOH, in the presence of variable concentrations of Cal (0–40 mM). Proper concentrations of DNA hairpins or aryltetrasaccharide were included, when required. After 10 min of incubation at 25 C, the reaction was started by adding 2 mM DTT. The mixture was further incubated for 15 min at 25 C and the reaction was quenched by addition of an excess of yeast tRNA. The reaction products were finally ethanol precipitated, dried, suspended in sequencing loading buffer (80% formamide, 10 mM NaOH, 1 mM EDTA, 0.1% xylene cyanol, 0.1% bromophenol blue), heated for 2 min at 90 C, chilled in ice, and loaded onto a 10–12% denaturing polyacrylamide gels (19 : 1) in 1X TBE. Gels were transferred to Whatman 3MM paper, dried and autoradiographed.
DNA Cleavage on Supercoiled DNA Plasmid A solution of supercoiled phage plasmid (UX174 replicative form I) was prepared to contain increasing concentrations of Cal derivatives
Biopolymers
The NMR structure of Calicheamicin c1 I -DNA complex was retrieved from the Protein Data Bank (PDB code: 2PIK).19 Hydrogen atoms and ionization states were carried out using standard properties to the DNA structure with the Molecular Operation Environment (MOE, version 2013.08) program.29 The structure of the isomeric analog of Cal (NAT a) was reconstructed starting from its NMR structure using the “Builder” tool implemented in MOE. NAT a was docked into the DNA minor groove using flexible MOE-Dock methodology.29 The best scored NAT a-DNA complex pose was subjected to Amber99 force field minimization until the rms of conjugate gradient was < 0.1 kcal mol21 A˚21.
RESULTS DNA Sequence Selectivity of Calicheamicin aryl-tetrasaccharide One of the fascinating properties of Calicheamicin c1 I is represented by its impressive DNA-sequence selectivity. The
452
Sissi et al.
FIGURE 2
DNA sequences used in this work.
preferred binding site is the homopyrimidine TCCT sequence. Some base mutations are tolerated by the drug, although they generally lead to a decrease in the cleavage efficiency. This reduction is more prominent when alternate purine2pyrimidine sequences are considered. This point can be experimentally summarized in cleavage experiments in which the DNA cleavage sites introduced upon reductive drug activation are sequenced and eventually quantified in order to obtain indication on the affinity of the drug for the tested binding sites. It is well accepted that, in the Cal-DNA binding process, the sequence selectivity is largely determined by the aryltetrasaccharide domain that is able to recognize the same sequences cleaved by the whole drug. However, a key point, not fully elucidated yet, is whether the modulation of Cal affinity determined for different DNA sequences is actually shared by its carbohydrate domain. This question derives from the higher DNA binding affinity showed by Cal when compared with the saccharide domain: this evidence suggests that all Cal structural domains actively contribute to the binding process.19,30 Here we decided to investigate in details the drug recognition process toward DNA sequences which are recognized by Cal to a significant different extent. In order to properly identify them, the preferential homopyrimidine TCCT site and an alternating ACAT sequence have been inserted within a 45 base pair long double stranded sequence (ds_45mer, Figure 2). As reported in Figure 3, by comparing the cleavage extent at the TCCT and ACAT sites, a significant difference clearly emerged with the first site being cleaved with almost 10 folds higher efficiency. Available data on Cal mechanism of action suggest this
effect to be due to a reduced affinity for the ACAT site by the drug. To highlight the contribution of the full drug in this process, we performed competition experiments. According to this protocol we used the extent of DNA cleavage upon reductive activation as a signal of binding; thus we monitored it when the reaction was performed in the presence of the two competing sequences under investigation. Since the DNA structural features are critical factors in allowing proper drug-DNA complex formation, the two selected sequences TCCT and ACAT were inserted in the paired regions of hairpin (h_TCCT and h_ACAT, Figure 2). This approach provides short double stranded DNA stems that are stable enough to avoid melting equilibria which could disrupt the oligonucleotide structure. Indeed their melting temperatures, determined by monitoring the absorption signal at 260 nm, were 58 C and 52 C for h_TCCT and h_ACAT, respectively. Moreover, this sequence design allows to limit the length of the oligos in order to avoid the presence of multiple binding sites on each DNA molecule. The results obtained are summarized in Figure 4. It should be noted that, by including the cold hairpins as competing sequences, a homogeneously reduced cleavage pattern of the labelled sequence was recorded, thus corresponding to Cal displacement from TCCT, GCCT, and TCCG sites. In particular, from a quantification of the cleavage products it emerged that about a 10-fold incremented concentration of h_ACAT in comparison to h_TCCT were required to impair to the same extent the cleavage of the labelled sequence by Cal, thus, likely, to displace the whole drug from the binding site. These results were confirmed both at 25 C and 10 C as working temperature. Biopolymers
Calicheamicin cI1-DNA Interaction
453
about the potential occurrence of even modest activation events during the titration course which would interfere with the measurement. Conversely, we took advantage of the availability of the aryl-tetrasaccharide moiety which has been previously found to compete with Cal for DNA binding.23 Selected examples of the recorded ITC profiles are reported in Figure 5. It can be readily appreciated that, irrespectively of the tested DNA sequence, the process was exothermic. Data points acquired by titrating the aryl-tetrasaccharide with the two hairpins at different temperatures were well fitted according to a “one binding site” model and provided the results summarized in Table I. According to our DNA design, a single molecule of binder was generally accommodated by the tested hairpins, thus supporting lack of unspecific binding. Concerning binding affinity, higher values were recorded when the hairpin h_TCCT sequence was used. In this context, it is worth to recall that ITC was previously applied to dissect the binding of Cal to a short duplex DNA containing the TCCT sequence, making this data set useful for comparison.31 The binding affinity of the whole drug to this site was Ka 1 * 106 M 21 , in good agreement with the Ka obtained by CD titrations using the same system, thus supporting the reliability of
FIGURE 3 Sequencing gel of ds45-mer (pyrimidine strand labeled at 5’) after treatment with increasing concentration of Cal (0–40 mM) for 15 min at 25 C in 50 mM Tris, pH 5 8.0, 2% EtOH. Cleavage reaction was initiated by addition of 2 mM DTT: C and M refer to untreated control DNA and Maxam and Gilbert purine marker, respectively. Arrows indicate the cleavage products (solid) and the corresponding base involved in hydrogen extraction along the DNA sequence (dashed).
Interestingly, this difference in the sequences recognition by Cal well correlates to the one determined when the two sequences where inserted in a single double stranded sequence, thus further supporting that the hairpin-mediated response is not affected by structural effects.
Thermodynamic Characterization of the Binding Process of aryl-tetrasaccharide to DNA The above described data support the idea that the designed hairpins may properly work as simplified models for the characterization of the binding mechanism of Cal to DNA. Thus, in order to dissect the thermodynamic features of such process, we used them to perform microcalorimetic titrations. As DNA binder we did not use the reactive drug since we were worried Biopolymers
FIGURE 4 Sequencing gel of ds53-mer (purine strand labeled at 3’) after treatment with 5 mM Cal for 15 min at 25 C in 50 mM Tris, pH 5 8.0, 2% EtOH in the presence of increasing concentration of h_TCCT or h_ACAT (0–24 mM). Cleavage reaction was initiated by addition of 2 mM DTT: C and M refer to untreated control DNA and Maxam and Gilbert purine marker, respectively. Arrows indicate the cleavage products (solid) and the corresponding base involved in hydrogen extraction along the DNA sequence (dashed).
454
Sissi et al.
FIGURE 5 ITC profiles corresponding to the titration of 40 mM aryl-tetrasaccharide with h_TCCT (Panel A) or h_ACAT (Panel B) performed at 10 C in 50 mM Tris, pH 8.0, 2% THF. Raw ITC data (top panels) and integrated heat exchanges (bottom panels) are reported.
this result in equilibrium conditions.32 Conversely, available binding affinity derived from Cal-mediated DNA cleavage protocols generally provided higher values.22,23 Interestingly, the ITC-derived Ka for the Cal-TCCT system is higher than the
one herein obtained with the aryl-tetrasaccharide tail thus actually supporting an active interaction of the aglycone with the nucleic acid. More surprising is the remarkable lower enthalpic variation (DH 5 26.47 kcal mol21). Nevertheless,
Table I Thermodynamic Parameters Derived from ITC Titrations Describing the Interaction of Aryltetrasaccharide with Hairpin h_TCCT or h_ACAT in 50 mM Tris, pH 5 8.0, 2% THF at Variable Temperature Sequence
T K
N
Ka * 105 (M21)
DH (kcal/mol)
h_TCCT
283 298 310 283 298 310
1.0 6 0.1 1.0 6 0.1 1.0 6 0.1 1.0 6 0.1 0.9 6 0.1 1.2 6 0.1
9.6 6 0.3 3.5 6 0.4 1.5 6 0.1 2.8 6 0.2 0.9 6 0.1 0.8 6 0.1
213.4 6 1.0 217.9 6 0.5 214.8 6 0.4 212.0 6 0.2 216.0 6 0.5 211.4 6 0.8
h_ACAT
DG (kcal/mol) 27.7 27.6 27.3 27.2 26.7 26.9
2TDS (kcal/mol) 5.6 10.3 7.5 5.4 9.3 4.4
Biopolymers
Calicheamicin cI1-DNA Interaction
this may be a consequence of the different experimental conditions (buffers and DNA substrates) used in the two sets of experiments. A novel result emerging from our ITC titrations concerns the selectivity. On average, the aryl-tetrasaccharide difference in Ka between h_TCCT and h_ACAT is of about 3 folds. This ranking order is in agreement with the one above observed with Cal. However, the selectivity of the carbohydrate domain appears to be remarkably lower when compared with the whole drug. Indeed, cleavage experiments pointed to a 10-fold preference for TCCT vs. ACAT by Cal. Clearly, it is not fully safe to compare data describing an equilibrium conditions (like ITC titrations) with those deriving from cleavage assay, where the involved species are changing along the time. However, this is the first evidence that the whole drug may be more efficient in specifically select one over the two tested sequences. A final comment should be dedicated to the enthalpic contribution. Indeed, at all the applied working temperatures, the preferential binding of the aryl-tetrasaccharide to h_TCCT in comparison to h_ACAT was related to a difference of about 2 kcal mol21 in the binding enthalpy. This is relevant since the binding reaction turned out to be clearly a highly enthalpy driven process. If we compared this result to the binding of other minor groove binders it is peculiar. Indeed, it is reported that the thermodynamic signature for this DNA binding mode is a generally favorable entropic contribution.33 The widely accepted explanation rests in an entropic gain due to solvent reorganization, in particular water release due to the insertion of molecules in the minor groove of the highly hydrated DNA. Nevertheless, several exceptions to this assumption have been described so far, which include DAPI and netropsin. Indeed, for these ligands, osmotic stress measurements supported a relevant uptake of water upon minor groove binding.34,35 This can be the case in the presence of our aryl-tetrasaccharide also. It is worth to remind that, from cleavage assay, the hydrophobic interaction between Cal and DNA was actually found to shift from an entropically to an enthalpically driven process upon temperature increments (T 30 C).36 Moreover, it is worth to underline that although NMR and computational studies support the potential formation of one ionic interaction between the nitrogen present on the aminosugar and the phosphate backbone, its contribution to the binding energetics is considered extremely reduced.30,37 Thus, no favorable entropic contribution connected to ion displacement from DNA spine should be expected. As a consequence, the lack of ionic pairs between the binder and the target requires a precise mutual orientation in order to optimize the nonionic interactions. The presence of the hydrophobic aglycone implements the complexity of the tridimensional Biopolymers
455
target recognition, thus increasing to potential to discriminate among different potential binding sites.
Role of the Relative Orientation of Saccharide vs. Enediyne in the DNA Cleavage So far, our experimental approach was designed to dissect the contributions of the carbohydrate tail and of calicheamicinone in the DNA binding process of Cal. Clearly, to convert an efficient DNA binder into a highly efficient DNA damaging agent, a proper localization of the reactive moiety in close proximity of the target site is required. This is primarily defined by the proper insertion of the aryl-tetrasaccharide into the DNA minor groove and by the relative orientation of this domain with reference to the reactive one: if we alter this parameter, we can expect that the DNA hydrogen extraction process might involve different positions or might occur with altered efficiency. To test how stereochemistry affects the behavior of the drug, we analyzed the DNA cutting properties of three analogues of Cal (Figure 1). The first one was Cal h, in which the trisulfide mojety of the natural compound is substituted by a thioacetate group. This modification allows the compound to undergo a diyl formation in mild basic conditions and without requirement of exogenous reducing agents.38 Structurally, these two derivatives are very similar and, consistently, they share a common DNA recognition mode: as it can be appreciated in sequencing gels, this structural homology results in a conserved distribution of cleavage sites along a double stranded DNA sequence between the two compounds (Figure 6). In Cal and Cal h, the reactive moiety is connected at the equatorial position of sugar A (b anomer). However, derivatives in which the a anomer of the sugar is functionalized (axial position) have been synthetically prepared although the functional effects of such modification have not been fully elucidated yet.28 Here, we used two derivatives in which the aglycone containing the S-acetate substitution was introduced to the a anomeric form of the carbohydrate tail. Indeed, since the natural (2) calicheamicinone39 was found to be less reactive toward DNA than its (1) enantiomer,40 both enantiomers were introduced providing derivatives NAT a and UNN a that contain the (2) and (1) enantiomer of calicheamicinone, respectively (Figure 1). Synthetic problems did not allow us to investigate the properties of the corresponding derivative containing the unnatural aglycone connected to the b anomer. These compounds represent good models to assess the effect of the relative position of the enediyne with reference to the carbohydrate moiety. For these derivatives we monitored the DNA cleavage efficiency focusing on the quantification of the ratio of double-
456
Sissi et al.
FIGURE 6 Sequencing gel of ds45-mer (pyrimidine strand labelled at 5’) after treatment with Cal (5 mM) Cal h (100 mM or NAT a for 15 min at 25 C in 50 mM Tris, pH 5 8.0, 2% EtOH, in the presence/absence of DTT: C and M refer to untreated control DNA and Maxam and Gilbert purine marker, respectively. Arrows indicate the cleavage products (solid) and the corresponding base involved in hydrogen extraction along the DNA sequence (dashed).
vs. single-stranded DNA breaks. This task was fulfilled by monitoring the kinetic of DNA cleavage on a supercoiled plasmid. To properly compare our data with those relative to Cal, experiments were performed both in the absence and in the presence of DTT. Cal h showed a very similar behavior to the natural compound. Indeed, it was able to produce double strand breaks (Figure 7). The difference between the two rests in the requirement of higher concentration of the synthetic derivative which, on turn, was able to produce DNA single and double strand brakes even in the absence of any reductive species in the reaction mixture. Assuming a Poisson distribution of the DNA lesions, it was possible to calculate the average number of single and double strand breaks.41 Data from experiments performed at different drug concentration, revealed that both Cal h produce a ratio double vs. single strand breaks 1.7:1 (Figure 8). This value is higher than the one reported from the
same lab with the leading compound Cal (ratio double vs. single strand breaks 1 : 2). However, by addition of the reducing agent an overall implemented DNA cleavage by Cal h occurred and, in this experimental condition, the double vs. single strand break ratio was actually reduced to 1.2 : 1. Moreover, we took into account that the DNA damage deriving from the hydrogen abstraction at 4’ position of the deoxisugar at the bottom strand (pyrimidine strand) may not always result in a cleavage but can produce abasic site.42 Thus, we also assessed the potential production of this kind of DNA lesion by our derivatives by treating the supercoiled plasmid reaction products with putrescine that is known to introduce a cut at apurinic site level (Figure 9). This experimental step converted the single strand brakes, produced by Cal as well as Cal h, into double strand lesions thus supporting that these compounds always generate double strand DNA damage. When we tested the two compounds connected at the a anomeric form of the sugar tail, a very different picture emerged. With NAT a, even working up to 5 mM drug concentration, no lesions were observed either in the presence or in the absence of DTT in the reaction mixture. A very peculiar result was obtained with the UNN a compound. In the whole range of tested concentrations (0– 3.5 mM), only single strand brakes were detected and their occurrence was linearly related to the incubation time. This result refers to the presence of DTT since, in a nonreducing environment, the drug-induced DNA damage was too limited to be properly quantified.
FIGURE 7 Time-dependent induction of cleavage on supercoiled phage plasmid UX174 induced by fixed concentration of tested derivatives upon incubation at 20 C s in 50 mM Tris, pH 5 8.0, 2% EtOH. When reported, 2 mM DTT was used included. I, II, and III refer to supercoiled, open circular and linearized plasmid, respectively.
Biopolymers
Calicheamicin cI1-DNA Interaction
457
FIGURE 8 Correlation plot of single (n1) vs. double (n2) strand cleavage recorded upon treatment of supercoiled phage plasmid UX174 with variable concentrations of Cal h in the absence (Panel A) or presence (Panel B) of DTT in 50 mM Tris, pH 5 8.0, 2% EtOH, 20 C.
Distinctly form b anomers, all single stranded cuts produced by UNN a, were completely insensitive to the action of putrescine. Thus, in this case, no apurinic sites are likely produced by this compound on the opposite strand, in close proximity of the cleavage site (Figure 9). The changes in the relative spatial organization of the binding and reactive domains should be used to rationalize the results provided by the a-anomers. The carbohydrate moiety of Cal is a rigid pre-organized structure that well fits the shape of the DNA minor grove. If we assume that this overall arrangement is conserved between the two anomeric forms, we could expect that the tested analogs bind DNA comparably. In this case, our data may advise that NAT a exposes the reactive radical towards the solvent so that no breaks are observed. Distinctly, using the non-natural enantiomer of calicheamicinone (UNN a), the change in the geometry of the reactive center can be sufficient to locate it in close proximity to just one DNA strand. These data would agree with the observation that single strand breaks for UNN a are not related to apurinic sites on the complementary strand. However, we cannot exclude that epimerization of the glycosidic bond may directly affect the insertion of the rigid carbohydrate structure into the DNA groove to a large extent. In fact, calicheamicinone is a quite bulky group that, in addition, actively contributes to the drug-DNA complex formation. A reduced overall binding affinity would thus justify why very high UNN a concentrations (2 orders of magnitude higher compared with NAT b) are required to nick supercoiled plasmid. To discriminate between these two models we performed a molecular modeling study of the a anomer of Cal h starting from the available NMR structure of Cal bound at a TCCT Biopolymers
site. Minimization of the docked structure revealed that a large part of the carbohydrate tail fits in an overlapping position of the minor groove irrespectively of the position of calicheamicinone insertion on the sugar. However, in the case of the a
FIGURE 9 Detection of apurinic sites by putrescine treatment after 30-min reaction of supercoiled plasmid with Cal h or UNN a at the reported concentration in 50 mM Tris, pH 5 8.0, 2% EtOH. I, II, and III refer to supercoiled, open circular, and linearized plasmid, respectively.
458
Sissi et al.
prompted us to further investigate the properties of single domains of Cal. The main focus was deserved to the aryl-tetrasaccharide domain for which we analyzed in deep detail the DNA affinity and sequence specificity. Clearly this domain represents the main driving force that promotes the DNA binding of Cal. However, our binding data indicate that the presence of the aglycon is beneficial not only to improve the affinity of the drug for the nucleic acid but also to increase its sequence selectivity. This is likely realized by proper hydrophobic interactions. When we moved further by evaluating the parameters that lead to an efficient cleavage process, a strict correlation between the relative tridimensional organization of all the drug domains emerges as relevant. Although apparently this represents an expected conclusion, our data point to a subtle regulation that involves both the binding affinity and the DNA modification efficacy. Overall our data support a model where the enediyne and the sugar domains are mutually affecting each other, so that the amazing binding affinity, sequence selectivity and cleavage efficiency of the natural drug cannot be considered as a simple algebraic sum of their contribution. This lets still open a great potential for this molecule for diagnostic and therapeutic applications. FIGURE 10 Comparison of DNA minor groove recognition of Calicheamicin c1 I (NMR structure, PDB code 2PIK19, colored in green,) and its isomer NAT a (modeled structure, colored in cyan). Hydrogen atoms are voluntarily omitted. DNA structure is represented using its Connolly’s surface (colored in orange).
anomer, the enediyne is actually pointing towards the solvent, thus resulting too far from the DNA backbone to effectively promote hydrogen abstraction. Interestingly, in this complex, the axial substitution forces the terminal-amino sugar also to be shifted outside the minor groove. Consequently, the interactions supported by this residue are lost. This clearly leads to a reduction of the stability of the drug-DNA complex. Thus, the different anomeric forms cause modifications in the DNA complex formation that cover the localization of both the reactive moiety and the binding domain.
CONCLUSIONS In all the recognition processes between biological systems a proper knowledge of the steric features is not only a valuable source of information to be used in the elucidation of the mechanism of action at molecular level, but is also a reliable starting point in the rational design of new derivatives with improved properties. This was the leading reason that
The authors acknowledge Prof. Samuel Danishefsky for providing them all the compounds used for this work and for his sincere support. The molecular modeling work coordinated by S.M. has been carried out with financial support from the University of Padova, Italy, and the Italian Ministry for University and Research, Rome, Italy. S.M. is also very grateful to Chemical Computing Group for the scientific and technical partnership.
REFERENCES 1. Lee, M. D.; Dunne, T. S.; Chang, C. C.; Ellestad, G. A.; Siegel, M. M.; Morton, G. O.; McGahren, W. J.; Border, D. B. J Am Chem Soc 1987, 109, 3466–3468. 2. Lee, M. D.; Dunne, T. S.; Siegel, M. M.; Chang, D. D.; Morton, G. O.; Borders, D. B. J Am Chem Soc 1987, 109, 3464–3466. 3. Cowan, A. J.; Laszlo, G. S.; Estey, E. H.; Walter, R. B. Front Biosci (Landmark Ed) 2013, 18, 1311–1334. 4. Hinman, L. M.; Hamann, P. R.; Wallace, R.; Menendez, A. T.; Durr, F. E.; Upeslacis, J. Cancer Res 1993, 53, 3336– 3342. 5. Jedema, I.; Barge, R. M.; van der Velden, V. H.; Nijmeijer, B. A.; van Dongen, J. J.; Willemze, R.; Falkenburg, J. H. Leukemia 2004, 18, 316–325. 6. Hamann, P. R.; Hinman, L. M.; Hollander, I.; Beyer, C. F.; Lindh, D.; Holcomb, R.; Hallett, W.; Tsou, H. R.; Upeslacis, J.; Shochat, D.; Mountain, A.; Flowers, D. A.; Bernstein, I. Bioconjug Chem 2002, 13, 47–58.
Biopolymers
Calicheamicin cI1-DNA Interaction 7. Rosen, D. B.; Harrington, K. H.; Cordeiro, J. A.; Leung, L. Y.; Putta, S.; Lacayo, N.; Laszlo, G. S.; Gudgeon, C. J.; Hogge, D. E.; Hawtin, R. E.; Cesano, A.; Walter, R. B. PLoS One 2013, 8, e53518. 8. Bross, P. F.; Beitz, J.; Chen, G.; Chen, X. H.; Duffy, E.; Kieffer, L.; Roy, S.; Sridhara, R.; Rahman, A.; Williams, G.; Pazdur, R. Clin Cancer Res 2001, 7, 1490–1496. 9. O’Hear, C.; Rubnitz, J. E. Expert Rev Hematol 2014, 7, 427–429. 10. Lee, M. D.; Dunne, T. S.; Chang, C. C.; Ellestad, G. A.; Siegel, M. M.; Morton, G. O.; McGahren, W.; Border, D. B. J Am Chem Soc 1992, 114, 985–997. 11. Zein, N.; Poncin, M.; Nilakantan, R.; Ellestad, G. A. Science 1989, 244, 697–699. 12. Zein, N.; Sinha, A. M.; McGahren, W. J.; Ellestad, G. A. Science 1988, 240, 1198–1201. 13. DeVoss, J. J.; Townsend, C. A.; Ding, W.-D.; Morton, G.; Ellestad, G. A.; Zein, N.; Tabor, A. B.; Schreiber, S. L. J Am Chem Soc 1990, 112, 9669–9670. 14. Hangeland, J. J.; DEVoss, J. J.; Heath, J. A.; Townsend, C. A.; Ding, W.-D.; Ashcroft, J. S.; Ellestad, G. A. J Am Chem Soc 1992, 114, 9200–9202. 15. Nicolaou, K. C.; Hale, C. R.; Nilewski, C. Chem Rec 2012, 12, 407–441. 16. Hitchcock, S. A.; Boyer, S. H.; Chu-Moyer, M. Y.; Olson, S. H.; Danishefsky, S. J. Angew Chem Int Ed Engl 1994, 33, 858–862. 17. Nicolaou, K. C.; Hummel, C. W.; Pitsinos, E. N.; Nakada, M.; Smith, A. L.; Shibayama, K.; Saimoto, H. J Am Chem Soc 1992, 114, 10082–10084. 18. Ellestad, G. A. Chirality 2011, 23, 660–671. 19. Kumar, R. A.; Ikemoto, N.; Patel, D. J. J Mol Biol 1997, 265, 187–201. 20. Mah, S. C.; Townsend, C. A.; Tullius, T. D. Biochemistry 1994, 33, 614–621. 21. Paloma, L. G.; Smith, J. A.; Chazin, W. J.; Nicolaou, K. C. J Am Chem Soc 1994, 116, 3697–3708. 22. Drak, J.; Iwasawa, N.; Danishefsky, S.; Crothers, D. M. Proc Natl Acad Sci USA 1991, 88, 7464–7468. 23. Li, T.; Zeng, Z.; Estevez, V. A.; Baldenius, K. U.; Nicolaou, K. C.; Joyce, G. F. J Am Chem Soc 1994, 116, 3709–3715. 24. Salzberg, A. A.; Dedon, P. C. Biochemistry 2000, 39, 7605–7612.
Biopolymers
459
25. Sissi, C.; Aiyar, J.; Boyer, S.; Depew, K.; Danishefsky, S.; Crothers, D. M. Proc Natl Acad Sci USA 1999, 96, 10643–10648. 26. Bernt, K. M.; Prokop, A.; Huebener, N.; Gaedicke, G.; Wrasidlo, W.; Lode, H. N. Bioconjug Chem 2009, 20, 1587–1594. 27. Lode, H. N.; Reisfeld, R. A.; Handgretinger, R.; Nicolaou, K. C.; Gaedicke, G.; Wrasidlo, W. Cancer Res 1998, 58, 2925–2928. 28. Hitchcock, S. A.; Chu-Moyer, M. Y.; Boyer, S. H.; Olson, S. H.; Danishefsky, S. J. J Am Chem Soc 1995, 117, 5750–5756. 29. .MOE (Molecular Operating Environment). Chemical Computing Group Inc: 1010 Sherbooke St West, Suite#910, Montreal, Quebec, Canada, 2012, version 2013.08;. 30. Kraka, E.; Tuttle, T.; Cremer, D. Chemistry 2007, 13, 9256–9269. 31. Chatterjee, M.; Mah, S. C.; Tullius, T. D.; Townsend, C. A. J Am Chem Soc 1995, 117, 8074–8082. 32. Krishnamurthy, G.; Ding, W.-D.; O’Brien, L.; Ellestad, G. A. Tetrahedron 1994, 50, 1341–1349. 33. Chaires, J. B. Arch Biochem Biophys 2006, 453, 26–31. 34. Degtyareva, N. N.; Wallace, B. D.; Bryant, A. R.; Loo, K. M.; Petty, J. T. Biophys J 2007, 92, 959–965. 35. Lewis, E. A.; Munde, M.; Wang, S.; Rettig, M.; Le, V.; Machha, V.; Wilson, W. D. Nucleic Acids Res 2011, 39, 9649–9658. 36. Ding, W.-D.; Ellestad, G. A. J Am Chem Soc 1991, 113, 6617– 6620. 37. Krishnamurthy, G.; Brenowitz, M. D.; Ellestad, G. A. Biochemistry 1995, 34, 1001–1010. 38. Nicolaou, K. C.; Li, T.; Nakada, M.; Hummel, C. W.; Hiatt, A.; Wrasidlo, W. Angew Chem Int Ed Engl 1994, 33, 183–186. 39. Ding, W.; Pitts, K.; Ellestad, G. A.; Krishnamurthy, G. Org Lett 2004, 6, 1801–1804. 40. Aiyar, J.; Hitchcock, S. A.; Denhart, D. J.; Liu, K. K.-C.; Danishefsky, S. J.; Crothers, D. M. Angew Chem Int Ed Engl 1994, 33, 855–858. 41. Povirk, L. F.; Wubter, W.; Kohnlein, W.; Hutchinson, F. Nucleic Acids Res 1977, 4, 3573–3580. 42. Dedon, P. C.; Salzberg, A. A.; Xu, J. Biochemistry 1993, 32, 3617–3622.
Reviewing Editors: Stephen Levene and Jonathan Chaires
