Cross-Linkage by “Intact” Bizelesin and Bisalkylation by the “Separated Halves” of the Bizelesin Dimer: Contrasting Drug Manipulation of DNA Conformation (5‘-TAATTA-3‘) Directs Alkylation toward Different Adenine Targets

Gel electrophoresis analysis of CPI-I bisalkylation of a 21-mer
duplex containing
5‘-TAA2TTA1-3‘
(the
palindromic preferred cross-linking sequence of the (+)-CC-1065
analog Bizelesin) shows same-strand (strand one)
alkylation of first A1 and then A2 instead of
the anticipated symmetrical A1 alkylation of strands one
and two.
Two-dimensional NMR analyses (NOESY and COSY) confirm the
head-to-tail minor groove orientation of the
same-strand-bound drugs. CPI-I contrasts sharply with Bizelesin
(two CPI-I units linked tail-to-tail by a ureadiyl
“linker”), which symmetrically cross-links this sequence at
A1 (strands one and two), but only by first
rearranging
the duplex structure and consequently removing the duplex distortion
stemming from monoadduct formation. CPI-I
induces no such major DNA rearrangement prior to or during
bisalkylation. Why does CPI-I react with the adenines
of only a single strand? Two possible causes for the unexpected
strand one A2 alkylation are, first, retardation
of
strand two A1 site's reactivity by focusing of monoadduct
conformational distortion on this site and, second,
elevation
of A2 reactivity above other competing adenine sites due to
unusual monoadduct strand one A2T-step
conformational
properties. The relative importance of these two nonmutually
exclusive factors was investigated using gel
electrophoresis experiments:  Time-course CPI-I bisalkylation studies
were conducted on the AT-step sequence 5‘-TAA2TTA1-3‘ and an A-tract sequence,
5‘-TAA2AAA1-3‘, to see if the former
sequence's AT-step flexibility, high
base-pair opening rate, and unwinding capability (traits not shared by
the latter sequence) controlled selection of the
second target site. The observed parallel AT-step and A-tract
sequence A1 and A2 bisalkylation patterns
suggest
that AT-step properties play at best a secondary role (compared to
5‘-end TA-step behavior) in directing the second
alkylation reaction to the AT-step site. rMD (solvated)
simulations of the AT-step and A-tract monoadducts display
distortion that is focused on this 5‘-end TA-step site. While
two-dimensional 1H NMR spectra of the final
bisadduct
reveal no significant TA-step conformational distortion, they
demonstrate that conformational adjustment at the
A2
ligand attachment site diminishes head-to-tail steric clash of the two
drugs. These results suggest that the CPI-I
monoadduct propagates bending distortion (to the 5‘-side) through five
base pairs toward the TA-step junction site.
In the AT-step and A-tract sequences, neither adenine straddling
this TA-step junction site is alkylated by CPI-I,
suggesting that the base pairs forming the junction site are distorted
away from a suitable orientation or are unable
to assume a conformation suitable for alkylation. Consequently,
the second alkylation occurs at a site (AT-step)
that requires a modest displacement of the second ligand away from the
already attached drug. The results and
analysis of the data included in this paper provide important lessons
for the design of inter- and intrastrand DNA−DNA cross-linkers.