DNA shuffling is a powerful process for directed evolution, which gene
rates diversity by recombination(1,2), combining useful mutations from
individual genes. Libraries of chimaeric genes can be generated by ra
ndom frag;mentation of a pool of related genes, followed by reassembly
of the fragments in a self-priming polymerase reaction. Template swit
ching causes crossovers in areas of sequence homology. Our previous st
udies used single genes and random point mutations as the source of di
versity(3-6). An alternative source of diversity is naturally occurrin
g homologous genes, which provide 'functional diversity.' To evaluate
whether natural diversity could accelerate the evolution process, we c
ompared the efficiency of obtaining moxalactamase activity from four c
ephalosporinase genes evolved separately with that from a mixed pool o
f the four genes. A single cycle of shuffling yielded eightfold improv
ements from the four separately evolved genes, versus a 270- to 540-fo
ld improvement from the four genes shuffled together, a 50-fold increa
se per cycle of shuffling. The best clone contained eight segments fro
m three of the four genes as well as 33 amino-acid point mutations, Mo
lecular breeding by shuffling can efficiently mix sequences from diffe
rent species, unlike traditional breeding techniques, The power of fam
ily shuffling may arise from sparse sampling of a larger portion of se
quence space.