The nucleic acids molecular mechanics program DUPLEX is specifically designed to compute atomic resolution structures of DNA that is damaged by linkage with carcinogenic environmental chemicals. These chemicals, polycyclic aromatic hydrocarbons and amines, are present in automobile exhaust, tobacco smoke, dyes, and barbecued meat and fish.
According to our current understanding, cancer is initiated by these substances, once they gain entry into the cell, by their biochemical activation to reactive moieties which then attack the DNA bases (A, G, C, T). Once bound to the bases, they can alter the DNA shape (conformation). This change can profoundly impact on the normal functioning of DNA during replication. In this process the replication machinery unwinds the double helix and synthesizes a daughter strand with bases that are the complement of the parent template-A and T are paired, and G and C are paired. Thus, for example, the machinery places a G in the daughter where the parent has a C. When the DNA is damaged by a carcinogen, this process can fail in accuracy such that a mismatch occurs at the modified base, or the modified base may be skipped altogether, or an extra base may be added at the lesion site. These errors constitute mutations, which can cause the synthesis of aberrant proteins that lead to cancer.
Obtaining atomic resolution views of carcinogen modified DNA has long been deemed essential for understanding the initiation of cancer by these substances. However, the task is formidable. Studies of recent years have revealed that a given substance may be mutagenic or not depending on what the bases adjacent to the modified one are (a sequence that is mutagenic is termed a ``hot-spot'' while a non-mutagenic one is a ``cold-spot''). Both hot-spot sequences, and cold-spot sequences as controls, must be examined. Furthermore, it is now understood that the damaged DNA must be examined in both normal double helices (the shape of the DNA when it is not replicating) and at single strand-double strand junctions that more accurately mimic the DNA shape as it is unwinding during the replication process. Double strands must be studied because lesion repair processes may succeed or fail, depending on the damaged DNA's shape; single strand-double strand replication intermediates must be studied because they model the situation during replication when the mutagenic event actually takes place. Furthermore, a given carcinogen may damage the DNA at more than one locus, and all the damaged loci must be examined. Moreover, certain polycyclic aromatic hydrocarbons are carcinogenic, while chemically very similar ones, even mirror images, are harmless; it is necessary to study both harmful substances and harmless ones, as controls.
Thus, it is necessary to obtain molecular views of DNAs modified by many different chemicals, linked to the DNA bases at different sites, in many different base sequence neighbor contexts, and in double strands and single strand-double strand junctions. With a large atomic resolution data base, structural hallmarks that are related to known mutagenic outcome from biological studies will emerge. Indeed, our investigations have already uncovered some such hallmarks. Once the structural data base is achieved, it will be possible to screen chemicals by computing their conformational impact on the DNA. If mutagenic hallmark structural damage is computed for a given substance, it would likely be harmful. This computational prediction of the mutagenic and carcinogenic potential of environmental chemicals will be extremely valuable. Current screening procedures require laborious, expensive and controversial animal testing to be relevant to the human condition with each test costing upwards of half a million dollars. Computational modeling would both lower the cost and controversy of these mandated screening procedures.
However, obtaining atomic resolution views experimentally is a daunting undertaking. For unambiguous structural information at atomic resolution, x-ray crystallography is the tool. However, to date, there is not a single crystal structure of any DNA modified by a polycyclic aromatic carcinogen; it has, so far, been impossible to obtain crystals of suitable diffracting quality. Average structures in solution are obtained by high resolution nuclear magnetic resonance (NMR) methods. However, fewer than a dozen structures have been obtained by this method to date. Synthesis has been a great problem. In many cases, the synthetic problems have not yet been overcome. In others, very laborious synthetic procedures (one preparation can take several man-years) are required. Furthermore, even when synthesis succeeds, the preparation may fail to yield interpretable NMR data if the structure is very mobile. Moreover, the NMR data, in the form of inter-proton distances, requires molecular mechanics for translation of the distances into a structure.
Our group has been involved in the molecular mechanics aspect of the NMR studies in most of the structures published so far. Thus, the experimental approach cannot possibly examine the large number of compounds, in the multiple base sequence contexts, and in both double strands and single-double strand junctions, that must be investigated to identify structural hallmarks of mutagenicity and carcinogenicity which will permit a predictive computational approach. Molecular mechanics calculations, on the other hand, have the capacity to model them, including especially the many whose synthesis is extremely difficult and/or for which high resolution NMR data is uninterpretable. In addition, for full understanding, both static and dynamic views are needed, while the experimental methods offer only static ones. Molecular dynamics simulations can, however, offer the animation that will yield the full picture.
DUPLEX has been tailored with special features that permit reliable computation of carcinogen-modified DNA structures, by surveying the potential energy surface with energy minimization. These features include (1) employing the reduced variable domain of internal coordinates (torsion angle space), (2) studying small modified subunits extensively and employing special building strategies to generate larger structures, and (3) including experimental data in the form of penalty functions (restraints) when these are available. Currently, DUPLEX can compute up to 24 residues of double stranded DNA in principle. However, the practical limit is about a dozen, since obtaining a single such structure can take 50 - 100 hours on a Cray YMP-C90 (such as the a-machine at NERSC), due to the computational intensity of this work. More powerful computing resources would permit analysis of the huge array of structures that must be examined in a much more reasonable time frame. Furthermore, much longer DNA strands could be studied.
The 140 residues of the DNA nucleosome core that is the first order building block of mammalian chromosomes could be treated instead of the small model duplexes that are now being studied of necessity. Even larger biologically important structures should become treatable; one key example is supercoiled DNA comprised of hundreds of residues. Supercoiling, the higher order winding of the double helix in topologically constrained situations like closed circles, to form figure-8 and other inter-wound forms, plays a leading role in DNA replication; furthermore, carcinogens are known to alter this essential topology, yet nothing is known about it at atomic resolution. DUPLEX will be able to investigate this morphology at atomic resolution with more powerful resources. The only way to accomplish such calculations is from advances in massively parallel computer architectures in addition to algorithm improvements.
Recently, we have carried out an extensive study of a substance, present in barbecued meat, known as PHIP (2-amino-1-methyl-6-phenylimidazo [4,5-b]pyridine). PHIP (see Fig. 32) is mutagenic and carcinogenic in rodents. When biochemically activated it binds to DNA, primarily to the base guanine at a position known as carbon-8. PHIP causes the damaged guanine to pair with adenine instead of the normal partner cytosine during DNA replication, causing a mutation that is probably an initiating event in the carcinogenic process.
We have carried out extensive computations to achieve atomic
resolution views of the PHIP-bound DNA in a double helix in which the
modified guanine is mispaired with adenine. In this work we used the
nucleic acids molecular mechanics program DUPLEX. The computations
revealed that binding of PHIP caused the damaged guanine to rotate 180
degrees from its normal ``anti'' position in the double helix to an
abnormal position termed ``syn''. In this orientation it can readily
accommodate an adenine in opposition, placing the carcinogenic
PHIP-moiety so that it is comfortably snuggled into the helix ``minor
groove.'' Thus, the computed structure offers a molecular view of an
initiating mutagenic event in PHIP induced carcinogenesis: the
mismatching of the modified guanine with adenine (see Figure
32). This is the only available atomic resolution
structure of DNA bound by PHIP, as experimental studies have not yet
been successful.
