The consensus sequence (5′-cGGCCgccg-3′) identified in this work agrees well with that deduced from analysis of preferential modification sites (14
), although the modified site positions varied by one base pair inward from each side, still at G on both strands. A combination of earlier reports (14
) and our current work indicates that the base-pairing between C and modified G, if it still occurs, is probably not as strong as normal G–C pairing, and therefore, the primer extension reactions (14
) used earlier to localize preferential modification sites could have run off one base before meeting the modified/cleaved site(s), in all cases.
It is obviously difficult to predict exactly how the five products of the dnd gene cluster interact with each other, and with the nucleotides of the highly conserved central core or the flanking regions. The absolute requirement of the central core (GGCC), in which modification sites were located at the two central bases on complementary strands, indicates unambiguously that the modifying activity is concentrated on these four nucleotide residues (GGCC) both for preferential and random modification in wild-type 1326 and in the dndB− mutant HXY2. Conceivably, these four nucleotides must be in close and direct contact with one of the major modifying enzymes. The preferential recognition specificity, totally unaffected in 1326 but variably abolished or reduced in HXY2, after mutation of the four nucleotides flanking the right, and two separated nucleotides flanking the left of the central core (GGCC), strikingly implicates the necessity of DndB in determining the recognition specificity. This agreed well with the detection of the significant amino acid sequence similarity of DndB to a DNA gyrase, which suggests that the DndB protein could in someway affect DNA topology, and thus the efficiency and/or specificity of S-modification on certain sites flanked by the sequences with a potential to form secondary structures. This could also explain why the DndB protein is not required at ‘simple’ recognition sites without flanking sequence complexity. We assume that the change from preferential modification sites, as detected in wild-type 1326, to a relatively random distribution of modification sites, as detected in its dndB mutant HXY2, is mediated by the DndB protein encoded by dndB. It is likely that random modification does not need DndB but a close association of DndB with the modifying enzymes could stabilize the contact between GGCC and the modifying complex.
It is not clear why a 5.4-kb circular plasmid molecule could only be modified once, although multiple modification sites are available. The presence of numerous modifiable sites on the chromosome suggests that modification frequency does not depend on the number of molecules, large or small, but more likely, by the length of the DNA flanking both sides of the modified site occupied by modifying enzymes. We do not know whether it is the binding of the modifying enzyme complex to a specific GGCC-containing site, or the result of modification, which prevents further binding of the modifying enzyme complex to a neighboring GGCC-containing site on either chromosome or plasmid, but it is hard to think that the result of modification could have contributed to this phenomenon. The fact that the Dnd phenotype could only be observed when a plasmid carrying the dnd gene cluster was present either in the integrated form or on a low copy number (<10) plasmid, but not at high copy number (Li,A.. unpublished data) implied that the expression of the dnd gene cluster in 1326 could be tightly controlled, and thus the dosage of the modifying enzyme complex could be limited in the cell. We propose that a protein complex formed at one specific site which involved nicking, winding and unwinding of the supercoiled double helix could exclude nearby site(s), and additionally, the number of sets of the Dnd protein complex is restricted to a cell-tolerated number by an unknown mechanism. To test the former hypothesis, it would be interesting to define the approximate length of the sequences flanking the consensus GGCC, within which modifiable sites were insensitive to modification, by analyzing plasmids of variable size. This work is now under consideration.
The proposed biochemical pathway leading to DNA modification by S involves five putative proteins encoded by a well-characterized dnd gene cluster. Of these, DndA was characterized as a PLP-containing homodimer that specifically catalyzes formation of L-alanine and elemental S using L-cysteine as substrate, DndC as an ATP pyrophosphatase catalyzing hydrolysis of ATP to pyrophosphate (PPi), and the function of DndC was found to be mediation of the formation of an [4Fe-4S] iron–sulfur cluster protein, whose reconstitution was activated by DndA (25). Instead, the exact biochemical functions of DndD and DndE, have not been demonstrated. While DndD, a bioinformatically predicted SMC-like ATPase with a distinguishable myosin-tail consisting of a coiled-coil region and a flexible hinge, could function as an ATP-modulated DNA cross-linker and energy generator by ATP hydrolysis, DndE was expected to involve in determining the sulfur-existing status. From the present work, it seems reasonable to assume that DndC, forming a [4Fe–4S] iron–sulfur cluster protein, might provide a platform for the orchestrated assembly of the Dnd protein complex, although some member(s) could be associated or not at specific stages of their required activities. The merit of the present work is thus, in essence, to have provided information for additional experiments aimed at demonstrating the interaction(s) between different proteins or between variable combinations of protein(s) and specific target DNA, including those mediating secondary structure formation and/or binding tightly or loosely under different condition(s) using, e.g. immuno-precipitations and gel shift mobility assays. This work is now in active progress.