Duplication or amplification of chromosomal segments is important for evolution, phenotypic variation, human genetic disorders, and cancer 
. Many of these duplications or amplifications are arranged in direct tandem repeat and have homologous sequence elements at their boundary, suggesting they were formed through recombination between non-allelic homologous sequences. Evidence for such non-allelic homologous recombination (NAHR) events is found in the genomes of nearly all species, including humans, where almost half of the human genome is comprised of low or high copy number repeat sequences 
The mechanisms responsible for these duplications or amplifications have been difficult to discern because these events are usually too rare to characterize their molecular intermediates. Nonetheless, studies primarily in microorganisms have led to a number of models for how these duplications/amplifications might arise. The most established model assumes that these NAHR events occur through the same fundamental mechanism as allelic homologous recombination 
. In this model NAHR is initiated by a simple DNA double-strand break (DSB) in a repeat sequence, which normally provokes a homology search for the intact allelic counterpart as a repair template. An imperfect search arising from misalignment of sister chromatids or homologs, however, would lead to establishment of a double Holliday junction structure between non-allelic homologous sequences that can resolve into an unequal crossover. Evidence of the reciprocal copy number expansions and contractions expected to arise from such unequal crossing over is limited, having only been observed in the context of large tandem arrays of rDNA 
, at subtelomeric repeats 
, and in some human genetic disorders 
. A recent variation on this model suggests that NAHR-mediated tandem duplications/amplifications may be generated by break-induced replication (BIR) 
. In this model, a broken chromosomal end initiates strand invasion and replication fork assembly at a non-allelic homologous sequence. The fork then duplicates the chromosomal segment between the homologous sequences before proceeding to the end of the chromosome. Again, direct support for this model is minimal.
Recently, we demonstrated that re-replication of a chromosomal segment due to dysregulation of replication controls can efficiently induce NAHR-medicated tandem duplication/amplification of that segment in the budding yeast Saccharomyces cerevisiae
. Importantly, introduction of simple DSBs failed to induce duplication/amplification with similar efficiency. Our findings raise the possibility that an alternative mechanism initiated by the loss of replication control might be responsible for some NAHR-mediated tandem duplications/amplifications. We have thus been eager to elucidate the mechanism of re-replication induced gene amplification (RRIGA).
The initiation of eukaryotic DNA replication is controlled by a battery of overlapping mechanisms that prevent re-initiation of DNA replication from the hundreds to thousands of replication origins in eukaryotic genomes. Replication initiation normally occurs at these origins in a two-stage process 
. In G1 phase, origins are licensed for initiation by loading the Mcm2-7 core replicative helicase onto them, a process that requires the origin recognition complex (ORC), Cdc6, Cdt1, and Mcm2-7. During S-phase, licensed origins are triggered to initiate DNA replication. To ensure that none of these origins re-initiate, multiple mechanisms inhibit ORC, Cdc6, Cdt1, and Mcm2-7 to minimize the chance that origins will be relicensed after they have initiated 
. Consistent with the non-redundant nature of these controls, experimentally inactivating increasing numbers of these mechanisms leads to progressively increasing amounts of re-initiation and re-replication in budding yeast 
These replication controls are critical for cell viability and genome stability. When sufficient controls are disrupted to cause overt re-replication (i.e. an increase in genomic DNA content detectable by flow cytometry), extensive DNA damage and a major DNA damage response is observed 
. While the source of damage is not well understood, the amount of damage apparently overwhelms the DNA damage response and leads to massive cell death. In budding yeast, we developed the ability to induce and detect much lower levels of re-replication compatible with cell viability 
. Retention of viability allowed us to examine the effect of re-replication on genome stability 
. We found that limited, transient re-replication of a chromosomal segment induced tandem duplication and occasionally higher order amplification of that segment at a rate approximating 10−2
per cell division, about five orders of magnitude higher than spontaneous duplication rates 
. The tandem duplications were bounded by Ty retrotransposon elements, a class of repetitive elements scattered throughout the yeast genome 
Here we uncover the mechanism of this re-replication induced gene amplification. These studies show that re-replication induces DNA damage because re-replication forks are highly susceptible to breakage. Our data support a model for RRIGA in which the two forks of a re-replication bubble both proceed beyond repetitive sequence elements flanking the re-initiating origin and then break. Should these breaks occur in trans with respect to the chromosome axis, normal 5′ to 3′ strand resection at each break will expose complementary strands of the non-allelic repetitive elements, providing a ready substrate for recombination by a single strand annealing (SSA) mechanism. Such repair of these broken re-replication bubbles will result in a tandem duplication arranged in direct repeat. In this model both the susceptibility of re-replication forks to breakage and the special structural context provided by the re-replication bubble contribute to the extraordinary efficiency of RRIGA. Importantly, the critical event triggering the formation of these tandem direct duplications is re-initiation of DNA replication within the duplicated segment. The remarkably efficient channeling of these re-initiation events into tandem direct duplications raises the possibility that even rare spontaneous re-initiation events may be a potent source of copy number variation in evolution and disease.