Linear transgene constructs lacking vector backbone sequences generate low-copy-number transgenic plants with simple integration patterns.

Whole plasmids are used in both Agrobacterium-mediated transformation and direct DNA transfer, generally leading to the integration of vector backbone sequences into the host genome along with the transgene(s). This is undesirable, as vector backbone sequences often have negative effects on transgene or endogenous gene expression, and can promote transgene rearrangements. We, therefore, bombarded rice tissue with two constructs: a plasmid containing the bar gene, and a linear DNA fragment isolated from the same plasmid, corresponding to the minimal bar gene expression cassette (promoter, open reading frame and terminator). We recovered phosphinothricin-resistant plants from both experiments, showing that the selectable marker was efficiently expressed. Transformation with such constructs resulted in predominantly 'simple' integration events (one or two bands on Southern blots), producing low-copy-number transgenic plants with a low frequency of transgene rearrangements. Conversely, transformation with supercoiled or linearized whole plasmids generated plants with 'complex' integration patterns, that is, higher copy numbers and frequent transgene rearrangements. We monitored transgenic lines through to the R4 generation and observed no silencing in plants carrying minimal constructs. We also carried out experiments in which rice tissue was simultaneously bombarded with minimal linear hpt and gusA cassettes. We observed robust GUS activity in hygromycin-resistant plants, confirming co-expression of the selectable and nonselectable markers. Furthermore, the efficiency of cotransformation using minimal constructs was the same as that using supercoiled plasmid cointegrate vectors.