Extrahelical cytosine bases in DNA duplexes containing d[GCC](n).d[GCC](n) repeats: detection by a mechlorethamine crosslinking reaction.

The cytosine-cytosine (C-C) pair is one of the least stable DNA mismatch pairs. The bases of the C-C mismatch are only weakly hydrogen bonded, and previous work has shown that, in certain sequence contexts, they can become unstacked from the core helix, and adopt an 'extrahelical' location. Here, using DNA duplexes with d[GCC](n).d[GCC](n) fragments containing C-C mismatches in a 1,4 bp relationship, we show that cytosine bases of different formal mismatch pairs can be crosslinked by mechlorethamine. For example, in the duplex d[CTCTCGCCGCCGCCGTATC].d[GATACGCCGCCGCCGAGAG], where underlined cytosine bases are present as the formal C-C mismatch pairs C(7)-C(32), C(10)-C(29) and C(13)-C(26), we show that two mechlorethamine crosslinks form between C(13) and C(29) and between C(10) and C(32), in addition to crosslinks at C(7)-C(32), C(10)-C(29) and C(13)-C(26) (we have reported previously the crosslinking of formal C-C pairs by mechlorethamine). We interpret the formation of the C(13)-C(29) and C(10)-C(32) crosslinks as evidence of an extrahelical location of the crosslinkable cytosines. Such extrahelical cytosine bases have been observed previously for a single C-C mismatch pair (in the so-called E-motif conformation). In the E-motif, the extrahelical cytosines are folded back towards the 5'-end of the duplex, consistent with our crosslinking data, and also consistent with the absence of C(7)-C(29) and C(10)-C(26) crosslinks in the current work. Hence, our data provide evidence for an extended E-motif DNA (eE-DNA) conformation in short d[GCC](n).d[GCC](n) repeat fragments, and raise the possibility that such structures might occur in much longer d[GCC](n).d[GCC](n) repeat tracts.

DNA interstrand crosslink formation by mechlorethamine at a cytosine-cytosine mismatch pair: kinetics and sequence dependence.

Expansion of the triplet repeat DNA sequence d[CGG]n.d[CCG]n is a characteristic of Fragile X syndrome, a human neurodegenerative disease. Stable intrastrand conformations formed by both d[CGG]n and d[CCG]n, and involving G-G and C-C mismatch pairs, respectively, are believed to be of importance in the development of the disease. We have shown previously that C-C mismatch pairs can be crosslinked covalently by mechlorethamine, a nitrogen mustard alkylating agent, and hence this reaction may be of value as a probe for conformers of d[CCG]n. To characterize the mechlorethamine C-C crosslink reaction further, here we report the kinetics and sequence dependence of formation of the crosslink species, using a series of model duplexes. The rate of reaction depends on the base sequence proximal to the C-C mismatch pair. Hence, in 19mer duplexes containing a central d[M4M3M2M1Cn1n2n3n4].d[N4N3N2N1Cm1m2m3m4] sequence, where M-m and N-n are complementary base pairs, the amount of crosslink increased with increasing G-C content of the eight base pairs neighboring the C-C mismatch and with the proximity of the G-C pairs to the C-C mismatch. Molecular dynamics simulations of the solvated duplexes provided an explanation of these data. Hence, for a C-C pair flanked by G-C base pairs the mismatched cytosine bases remain stacked within the duplex, but for a C-C pair flanked by A-T base pairs, the simulations suggested local opening of the duplex around the C-C pair, making it a less effective target for mechlorethamine.

Water dynamics at the binding interface of four different HLA-A2-peptide complexes.

Because only a limited number of MHC molecules are available for presentation of a large number of peptides, each of these MHC molecules must be able to bind promiscuously many different peptides at an affinity sufficient for stable presentation. Here we show, for the MHC molecule HLA-A2, that this ability may be facilitated by a flexible water network that forms an interface between the MHC molecule and the peptide. Using the SURFNET program we have computed the 'gaps' present in the peptide-binding groove in the X-ray structures of complexes of HLA-A2 with four different bound peptides. The volume of these gaps increases with increasing peptide hydrophilicity. Using molecular dynamics simulations, we show that the water molecules in the binding groove of complexes of HLA-A2 with the more hydrophilic peptides are largely disordered, but a number of defined water-binding sites are also discernable. Conversely, for complexes of HLA-A2 with the more hydrophobic peptides, the water molecules are more rigidly bound at the MHC-peptide interface and a number of well-defined water-binding sites exist. However, even these well-defined sites may not be permanently occupied by the same water molecule and in the dynamics calculations we observed exchange of water molecules between such sites.

DNA structure control by polycationic species: polyamine, cobalt ammines, and di-metallo transition metal chelates.

Many polycationic species bind to DNA and induce structural changes. The work reported here is the first phase of a program whose long-term aim is to create a class of simple and inexpensive sequence-selective compounds that will enable enhanced DNA structure control for a wide range of applications. Three classes of molecule have been included in this work: the polyamine spermine (charge: 4(+)) and spermidine (charge: 3(+)) (which are known to induce a wide range of DNA conformational changes but whose binding modes are still not well understood); cobalt (III) amine transition metal complexes as potential polyamine mimics and [Fe(H(2)O)(6)](3+); and the first member of a new class of di-metallo tris-chelated cylinders of helical structure (charge 4(+)). Temperature-dependent absorption, circular dichroism, linear dichroism, gel electrophoresis, and molecular modeling data are presented. The cobalt amines prove to be effective polyamine mimics, although their binding appears to be restricted to backbone and major groove. All the ligands stabilize the DNA, but the 4(+) di-iron tris-chelate does so comparatively weakly and seems to have a preference for single-stranded DNA. All the molecules studied bend the DNA, with the di-iron tris-chelate having a particularly dramatic effect even at very low drug load.

Biopharmaceutics of transmucosal peptide and protein drug administration: role of transport mechanisms with a focus on the involvement of PepT1.

Non-invasive delivery of peptide and protein drugs will soon become a reality. This is due partly to a better understanding of the endogenous transport mechanisms, including paracellular transport, endocytosis, and carrier-mediated transport of mucosal routes of peptide and protein drug administration. This paper focuses on work related to the elucidation of structure-function, intracellular trafficking, and regulation of the intestinal dipeptide transporter, PepT1.

Spermine inhibition of the 2,5-diaziridinyl-1,4-benzoquinone (DZQ) crosslinking reaction with DNA duplexes containing poly(purine). poly(pyrimidine) tracts.

Upon reduction, 2,5-diaziridinyl-1,4-benzoquinone (DZQ) can form an interstrand guanine to guanine crosslink with DNA duplexes containing a d(GC).d(GC) dinucleotide step. The reaction is enhanced by a thymine positioned 5[prime] to each guanine [i.e. in a d(TGCA). d(TGCA) duplex fragment]. Here we show that spermine can inhibit DZQ crosslink formation in duplexes of sequence d[C(N6)TGCA(M6)C]. d[G(M[prime]6)TG-CA(N[prime]6)G]. For N6= M6= GGGGGG, N6= M6= a 'random' sequence and N6= GGGGGG and M6= a 'random' sequence, spermine concentrations of 20, 1 and 3 microM, respectively, were required for 50% inhibition of the DZQ crosslink. This suggests that spermine is more strongly bound to the polyguanosine tract than the random sequence, making it less available for crosslink inhibition. When the polyguanosine tract is interrupted by N 7-deazaguanine (D) located three bases, d(CGGGDGGTGCAGGDGGGC), and four bases, d(CG-GDGGGTGCAGGGDGGC), from the d(TGCA).d(TGCA) site, 30 and 3 microM spermine, respectively, were required for 50% crosslink inhibition. We suggest that this difference is due to the relative proximity of the three-guanosine tract to the d(TGCA).d(TGCA) site. We were able to confirm these conclusions with further experiments using duplexes containing three-guanosine and two-guanosine tracts and from computer simulations of the spermine-DNA complexes.

Anomalous cross-linking by mechlorethamine of DNA duplexes containing C-C mismatch pairs.

Nitrogen mustards such as mechlorethamine have previously been shown to covalently cross-link DNA through the N7 position of the two guanine bases of a d[GXC].d[GYC] duplex sequence, a so-called 1,3 G-G-cross-link, when X-Y = C-G or T-A. Here, we report the formation of a new mechlorethamine cross-link with the d[GXC].d[GYC] fragment when X-Y is a C-C mismatch pair. Mechlorethamine cross-links this fragment preferentially between the two mismatched cytosine bases, rather than between the guanine bases. The cross-link also forms when one or both of the guanine bases of the d[GCC].d[GCC] fragment are replaced by N7-deazaguanine, and, more generally, forms with any C-C mismatch, regardless of the flanking base pairs. Piperidine cleavage of the cross-link species containing the d[GCC].d[GCC] sequence gives DNA fragments consistent with alkylation at the mismatched cytosine bases. We also provide evidence that the cross-link reaction occurs between the N3 atoms of the two cytosine bases by showing that the formation of the C-C cross-link is pH dependent for both mechlorethamine and chlorambucil. Dimethyl sulfate (DMS) probing of the cross-linked d[GCC].d[GCC] fragment showed that the major groove of the guanine adjacent to the C-C mismatch is still accessible to DMS. In contrast, the known minor groove binder Hoechst 33258 inhibits the cross-link formation with a C-C mismatch pair flanked by A-T base pairs. These results suggest that the C-C mismatch is cross-linked by mechlorethamine in the minor groove. Since C-C pairs may be involved in unusual secondary structures formed by the trinucleotide repeat sequence d[CCG]n, and associated with triplet repeat expansion diseases, mechlorethamine may serve as a useful probe for these structures.

Proton-driven dipeptide uptake in primary cultured rabbit conjunctival epithelial cells.

PURPOSE: To characterize proton-driven carrier-mediated dipeptide uptake in primary cultured conjunctival epithelial cells of the pigmented rabbit using beta-alanyl-L-histidine (L-carnosine) as a model dipeptide substrate. METHODS: Uptake of tritiated L-carnosine was monitored using conjunctival epithelial cells on days 6 through 8 in culture on a filter support. The structural features of dileucine stereoisomers and cephalexin contributing to interaction with the dipeptide transporter were evaluated by computer modeling and inhibition of tritiated L-carnosine uptake. RESULTS: Uptake of L-carnosine by primary cultured conjunctival epithelial cells in the presence of an inwardly directed proton gradient showed directional asymmetry (favoring apical uptake by a factor of five), temperature dependence, and saturability correlated with substrate concentration, with a Michaelis-Menten constant (Km) of 0.3 +/- 0.03 mM and a maximum uptake rate (Vmax) of 22.0 +/- 1.0 picomoles per milligram protein per minute. L-Carnosine uptake was optimal at pH 6.0 and was reduced by 60% and 35%, respectively, by 50 microM p-trifluoromethoxyphenylhydrazone (a proton ionophore) and by acid preloading with 50 mM NH4Cl. The constituent amino acids did not inhibit L-carnosine uptake. L-Carnosine uptake was inhibited, however, from 50% to 80% by other dipeptides and structurally similar drugs such as bestatin, beta-lactam antibiotics, and angiotensin-converting enzyme inhibitors. The LL, LD, or DL forms of the dipeptide Leu-Leu inhibited tritiated L-carnosine uptake by approximately 60%, 40%, and 70%, respectively. By contrast, the DD form did not inhibit uptake. Results from computer modeling suggest that an appropriate dipeptide N-terminal to C-terminal distance and a favorable orientation of the side chains may be important for substrate interaction with the conjunctival dipeptide transporter. CONCLUSIONS: Uptake of the dipeptide L-carnosine in primary cultured pigmented rabbit conjunctival epithelial cells is probably mediated by a proton-driven dipeptide transporter. This transporter may be used for optimizing the uptake of structurally similar peptidomimetic drugs.

Structure, function, and molecular modeling approaches to the study of the intestinal dipeptide transporter PepT1.

The proton-coupled intestinal dipeptide transporter, PepT1, has 707 amino acids, 12 putative transmembrane domains (TMD), and is of importance in the transport of nutritional di- and tripeptides and structurally related drugs, such as penicillins and cephalosporins. By using a combination of molecular modeling and site-directed mutagenesis, we have identified several key amino acid residues that effect catalytic transport properties of PepT1. Our molecular model of the transporter was examined by dividing it into four sections, parallel to the membrane, starting from the extracellular side. The molecular model revealed a putative transport channel and the approximate locations of several aromatic and charged amino acid residues that were selected as targets for mutagenesis. Wild type or mutagenized human PepT1 cDNA was transfected into human embryonic kidney (HEK293) cells, and the uptake of tritiated glycylsarcosine [3H]-(Gly-Sar) was measured. Michaelis-Menton analysis of the wild-type and mutated transporters revealed the following results for site-directed mutagenesis. Mutation of Tyr-12 or Arg-282 into alanine has only a very modest effect on Gly-Sar uptake. By contrast, mutation of Trp-294 or Glu-595 into alanine reduced Gly-Sar uptake by 80 and 95%, respectively, and mutation of Tyr-167 reduced Gly-Sar uptake to the level of mock-transfected cells. In addition, preliminary data from fluorescence microscopy following the expression of N-terminal-GFP-labeled PepT1Y167A in HEK cells indicates that the Y167A mutation was properly inserted into the plasma membrane but has a greatly reduced Vmax.

Molecular identification of a role for tyrosine 167 in the function of the human intestinal proton- coupled dipeptide transporter (hPepT1).

hPepT1 is a proton-coupled peptide transporter that mediates the absorption of di- and tripeptides. Here we show that tyrosine 167 (Y167) in transmembrane domain 5 (TMD5) of this 12-transmembrane spanning protein contributes to its transport function. We identified this particular amino acid by a computer model of the arrangement of the TMDs of hPepT1 and investigated its role by site-directed mutagenesis and dipeptide uptake studies. [3H]Gly-sar uptake in cells transiently transfected with Y167A-hPepT1 was abolished completely, even though the level of Y167A-hPepT1 expression by Western blot analysis and cell surface expression by immunofluorescence microscopy was similar to those of the wild type. Therefore, mutation affected transport function, but apparently not the steady-state protein level or trafficking of the transporter to the plasma membrane. Moreover, mutation of Y167 into phenylalanine, serine, or histidine all abolished gly-sar uptake in transfected HEK 293 cells. Taken together, these findings suggest that Y167 plays an essential role in hPepT1 function, perhaps due to the unique chemistry of its phenolic side chain.

A model of water structure inside the HLA-A2 peptide binding groove.

Based on molecular dynamics simulations, it is proposed that water within the binding groove of the human MHC class I molecule HLA-A2 plays a role in the formation of its complex with the influenza matrix protein (residues 58-66; GILGFVFTL) peptide. In these simulations, a loosely structured network of water molecules is present in the binding groove between the peptide and the MHC molecule, and may be important in completing the peptide-MHC interface. In two independent 400 ps simulations where groove-based water molecules were included, the peptide remained essentially in the conformation observed in the crystal structure. In contrast, in a 400 ps simulation in which no water molecules were placed between the peptide and the MHC molecule, the crystal structure conformation was rapidly lost. The basis for this behavior appears to be that the groove-based water molecules help to maintain the appropriate orientation of the Arg-97 side chain of HLA-A2 and, in turn, the conformation of the central part of the peptide.

At physiological pH, d(CCG)15 forms a hairpin containing protonated cytosines and a distorted helix.

To investigate potential structures of d(CGG/CCG)n that might relate to their biological function and association with triplet repeat expansion diseases (TREDs), the structure of a single-stranded (ss) oligonucleotide containing d(CCG)15 [ss(CCG)15] was examined by studies of the pH and temperature dependence of electrophoretic mobility, UV absorbance, circular dichroism, chemical modification, and P1 nuclease digestion. ss(CCG)15 had an unusually high pKa (7.7 +/- 0.2). At pH 8.5, ss(CCG)15 formed a relatively unstable (Tm = 30 degrees C in 1 mM Na+) hairpin containing CpG base-pair steps. At pH 7.5, the hairpin contained protonated cytosines but no detectable C x +C base pairs, increased thermal stability (Tm = 37 degrees C), increased stacking of the CpG base-pair steps, and a single cytosine that was flipped away from the central portion of the helix. Examination of ss(CCG)18 and ss(CCG)20, which were designed to adopt hairpins containing alternative GpC base-pair steps, revealed hairpins containing CpG base-pair steps, pKas of approximately 8.2 and approximately 8.4, respectively, and distorted helices. The results suggest that DNA sequences containing (CCG)(n > or = 15) adopt hairpin conformations that contain CpG rather than GpC base-pair steps; the mismatched cytosines are protonated at physiological pH but are not H-bonded. We propose that protonation arises from the stacking of two cytosines in the minor groove of a distorted helix.

Spermine binding to GC-rich DNA: experimental and theoretical studies.

The DNA binding of spermine has been studied using experimental and computational approaches. Spermine blocks 5'-GC interstrand crosslinking by 2,5-diaziridinylbenzoquinone in the oligonucleotide duplex 5'-CTTCCAAGATGCATCAGATG 5'-CATCTGATGCATCTTGGAAG (where the underlined nucleotide bases represent the crosslinking site). Molecular dynamics simulations suggest that this is a result of preferential spermine binding at the 5'-GC major groove site of the oligonucleotide. A further simulation with a GC-alternating sequence shows a similar preference for the 5'-GC step. In a simulation including multiple spermine molecules, occupation of alternate 5'-GC steps occurred. From this, we deduced a mechanism for the experimentally observed cooperativity of spermine binding to poly(dGdC)2.

The trinucleotide repeat sequence d(CGG)15 forms a heat-stable hairpin containing Gsyn. Ganti base pairs.

To investigate potential structures of d(CGG/CCG)n that might relate to their biological function and association with triplet repeat expansion diseases (TREDs), electrophoretic mobility, chemical modification, and P1 nuclease studies were performed with a single-stranded (ss) oligonucleotide containing (CGG)15 [ss(CGG)15]. The results suggest that ss(CGG)15 forms a hairpin with the following features: (i) a stem containing Gsyn. Ganti base pairs; (ii) at > or = 200 mM K+, CGG repeats on the 5' portion of the stem base-paired to GCG repeats on the 3' side (referred to as the (b) alignment); and (iii) heat stability (Tm = 75 degrees C in low ionic strength). At < or = 100 mM K+, dimethyl sulfate reactions indicated that the hairpin in the (b) alignment was in equilibrium with another structure, presumably a hairpin in the alternative (a) alignment (CGG repeats on the 5' portion of the stem base-paired to CGG repeats on the 3' portion of the stem). Molecular dynamics simulations suggested that the loop region of the (a) alignment contained two guanines stacked on top of one another. The same guanines in the (b) alignment were base-paired in a syn-anti arrangement. We propose that the stability of the loop partially determines the stem alignment.

Molecular cloning, characterization, and mapping of a full-length cDNA encoding human UDP-galactose 4'-epimerase.

Galactose metabolism in all organisms is catalyzed by three enzymatic steps: the galactokinase, galactose-1-phosphate uridyltransferase, and UDP galactose 4'-epimerase reactions. We report here the molecular cloning, characterization, and mapping of a full-length cDNA encoding human UDP-galactose 4'-epimerase (GALE). Our cDNA is 1488 bp long and matches the mRNA size of 1.5 kg detected in fibroblasts and lymphoblasts. The human GALE cDNA encodes a predicted protein of 348 amino acids with a molecular mass of 38,266. The human GALE enzyme is 87% identical to the rat protein, 53% identical to the homologous GAL10 protein from the yeast Kluyveromyces lactis, and 51% identical to the galE protein from the prokaryote Escherichia coli. This extraordinary degree of sequence identity has allowed us to build a homology model of the human protein based on the bacterial crystal structure. This predicted human structure is very similar to the E. coli galE enzyme, suggesting that both enzymes use similar mechanisms. The human gene encoding GALE maps, as expected, to a single locus on chromosome 1 and appears to be compact. The human GALE gene is structurally intact in 19 patients with epimerase-deficiency galactosemia, an inborn error of metabolism secondary to GALE deficiency. Therefore, we propose that this disorder is due to small mutations within the gene.

The trinucleotide repeat sequence d(GTC)15 adopts a hairpin conformation.

The structure of a single-stranded (ss) oligonucleotide containing (GTC)15 [ss(GTC)15] was examined. As a control, parallel studies were performed with ss(CTG)15, an oligonucleotide that forms a hairpin. Electrophoretic mobility, KMnO4 oxidation and P1 nuclease studies demonstrate that, similar to ss(CTG)15, ss(GTC)15 forms a hairpin containing base paired and/or stacked thymines in the stem. Electrophoretic mobility melting profiles performed in approximately 1 mM Na+ revealed that the melting temperature of ss(GTC)15 and ss(CTG)15 were 38 and 48 degrees C respectively. The loop regions of ss(GTC)15 and ss(CTG)15 were cleaved by single-strand-specific P1 nuclease at the T25-C29 and G26-C27 phosphodiester bonds respectively (where the loop apex of the DNAs is T28). Molecular dynamics simulations suggested that in ss(GTC)15 the loop was bent towards the major groove of the stem, apparently causing an increased exposure of the T25-C29 region to solvent. In ss(CTG)15 guanine--guanine stacking caused a separation of the G26 and C27 bases, resulting in exposure of the intervening phosphodiester to solvent. The results suggest that ss(GTC)15 and ss(CTG)15 form similar, but distinguishable, hairpin structures.

Multiple DNA binding modes of anthracene-9-carbonyl-N1-spermine.

The poly(dAdT)2 complex of anthracene-9-carbonyl-N1-spermine, a spermine derivative terminally substituted with an anthracene moiety, has been studied using fluorescence, linear dichroism, circular dichroism, normal absorption spectroscopy (as a function of temperature) and computer modelling. For comparison, some data are also provided for the same ligand with poly(dGdC)2 and calf thymus DNA. Following detailed fluorescence and CD spectroscopic studies, we propose that anthracene-9-carbonyl-N1-spermine intercalates in at least two different binding orientations with poly(dAdT)2. Based on computer simulation data, we deduce that the ligand can intercalate from both the minor groove and the major groove. In contrast, intercalation with poly(dGdC)2 probably occurs only from the major groove. At high ligand concentrations, the CD spectra suggest anthracene-anthracene interactions, whilst the LD data point towards a groove-bound anthracene. Again from computer simulations, we propose binding modes consistent with these observations. Other data from the LD spectra suggest a sequential nature to the binding of the ligand to calf thymus DNA, with GC-rich sites being occupied first. At low ligand concentrations, anthracene-9-carbonyl-N1-spermine is able to stabilize poly(dAdT)2 against thermal decomposition, but not as effectively as spermine. The reverse is found to be true with calf thymus DNA. Both the anthracene-9-carbonyl-N1-spermine and spermine complexes of poly(dAdT)2 show pre-melt transitions in their melting curves. The anthracene-9-carbonyl-N1-spermine complex with poly(dAdT)2 also shows a post-melt transition.

Hairpin properties of single-stranded DNA containing a GC-rich triplet repeat: (CTG)15.

Although triplet repeat DNA sequences are scattered throughout the human genome, their biological function remains obscure. To aid in correlating potential structures of these nucleic acids with their function, we propose their classification based on the presence or absence of a palindromic dinucleotide within the triplet, the G + C content, and the presence or absence of a homopolymer. Five classes of double-stranded (ds) triplet repeats are distinguished. Class I repeats, which are defined by the presence of a GC or CG palindrome, have the lowest base stacking energies, exhibit the lowest rates of slippage synthesis [Schlötterer and Tautz (1992) Nucleic Acids Res., 20, 211] and are uniquely associated with triplet repeat expansion diseases. The six single-stranded (ss) triplet repeats within Class I also have the potential to form hairpin structures, as determined by energy minimization. To explore the possibility of hairpin formation by ss Class I triplet repeats, studies were performed with a ss oligonucleotide containing 15 prototypic CTG repeats [ss (CTG)15]. Electrophoretic, P1 nuclease and KMnO4 oxidation data demonstrate that ss (CTG)15 forms a hairpin containing base paired and/or stacked thymines in the stem. Potential functions of hairpins containing Class I triplet repeats are discussed with respect to protein translation and mRNA splicing. Further, potential roles of hairpin structures in triplet repeat expansion events are discussed.

Molecular dynamics simulations of chlorambucil/DNA adducts. A structural basis for the 5'-GNC interstrand DNA crosslink formed by nitrogen mustards.

The alkylation of DNA by chlorambucil has been studied using a computational approach. Molecular dynamics simulations were performed on the fully solvated non-covalent complex, two monoadducts and a crosslinked diadduct of chlorambucil with the d(CGG3G2CGC).-d(GCG1CCCG) duplex, in which the N7 atoms of G1, G2 and G3 are potential alkylation sites. The results provide a structural basis for the preference of nitrogen mustards to crosslink DNA duplexes at a 5'-GNC site (a 1,3 crosslink, G1-G3) rather than at a 5'-GC sites (a 1,2 crosslink, G1-G2). In the non-covalent complex simulation the drug reoriented from a non-interstrand crosslinking location to a position favorable for G1-G3 diadduct formation. It proved possible to construct a G1-G3 diadduct from a structure from the non-covalent simulation, and continue the molecular dynamics calculation without further disruption of the DNA structure. A crosslinked diadduct developed with four BII conformations on the 3' side of each alkylated guanine and of their respective complementary cytosine. In the first monoadduct simulation the starting point was the same DNA conformation used in the crosslinked diadduct simulation with alkylation at G1. In this simulation the DNA deformation was reduced, with the helix returning to a more canonical form. A second monoadduct simulation was started from a canonical DNA conformation alkylated at G3. Here, no significant motion towards a potential crosslinking conformation occurred. Collectively, the results suggest that crosslink formation is dependent upon the drug orientation prior to alkylation and the required deformation of the DNA to permit 1,3 crosslinking can largely be achieved in the non-covalent complex.