BER is critical for dealing with a variety of single strand lesions. Many enzymes in this pathway are conserved from microorganisms to humans and serve as antimutators, especially in terms of tumor suppression and preventing hereditary neurodegenerative disease 
. Aberrant BER processes might result in the eventual appearance of DSBs, which are a major source of genome instability. MMS-induced lesions are considered a source of DSBs as a result of collapsed replication forks at the lesions or processed intermediates. Based on genetic evidence, these replication-associated DSBs have been considered to be repaired by HR mechanisms.
Previously, we showed that MMS-induced single-strand damage in G1 arrested cells had the potential for generating derived DSBs and highlighted the role that Rad27 and Pol32 play in preventing such breaks 
. We had concluded that closely-spaced opposing lesions could be a source of the derived DSBs and that well-coordinated BER assures prevention of these downstream DSBs. The present study using G2/M cells is the first to characterize directly the generation, processing and repair of derived DSBs following treatment by an alkylating agent. While two closely-opposed SSBs with 5′-blocked termini could be “moved closer” to form a DSB during repair-associated DNA synthesis and strand displacement 
, this is not expected to be the case for SSBs with 3′-blocked termini since they would not support DNA synthesis directly. The present results are consistent with derived DSBs resulting from generation of SSBs at closely opposed lesions. It is also possible that derived DSBs could arise through processing of more distant SSBs with 3′-blocked ends in cells lacking AP endonucleases, in essence the breaks are “moved closer,” as discussed below. Furthermore, we have described a novel repair intermediate, SMD, which can be generated if abasic sites are nicked by AP lyases.
While the APE1
gene, which codes for the major mammalian AP endonuclease (the APE1 homologue APE2 has only weak endonuclease activity, and its role in human BER is not clear), is essential for human cell survival and results in embryonic lethality when knocked out in mouse 
, yeast can survive the deletion of both AP endonucleases with almost no growth defect. It is, therefore, possible to study alternative mechanisms/pathways that deal with AP sites and 3′-blocked SSBs in vivo
and their role in generating DSBs. Earlier studies had demonstrated that MMS does not cause DSBs directly 
. With an assay that can specifically monitor the processing of closely-opposed single strand lesions, our previous study showed that PFGE-detected DSBs were accumulated in G1 apn1/2
haploid yeast after MMS damage. However, since closely-opposed SSBs might lead to chromosome DNA breakage during in vitro
handling, the extent to which the PFGE-detectable DSBs were actually formed in vivo
remained a question. Here, we confirm that DSBs do appear after MMS treatment of G2 cells lacking AP endonucleases, as demonstrated by i) resection, ii) a requirement for HR components to reconstitute chromosomes, and by iii) the formation of Chr III dimers. While resection is generally considered essential for DSB repair mechanisms 
, we have demonstrated that it also occurs at the MMS-derived DSBs and like radiation-induced DSBs they are subject to MRX control. As in the case of randomly generated radiation-induced DSBs 
, we aimed to determine other factors affecting resection at the MMS-derived DSBs, especially factors that may lead to increased resection. We have recently shown that UV as well as MMS damage to single-strand DNA formed at site-specific DSBs cause high level of mutagenesis 
. Increasing resection at MMS derived breaks could further enhance its mutagenic potential.
The current results further confirm that DSBs can be derived from AP sites arising during BER, since the appearance of DSBs could be blocked either at the step in which methylated bases are removed or if cleavage of AP sites is prevented by MX (). It is clear that DSBs were generated by the bifunctional glycosylases because deletion of NTG1
along with APN1
blocked the formation of DSBs as well as resection. The targets of these enzymes are limited to AP sites instead of methylated bases based on efficient DSB inhibition following deletion of MAG1
(). Though OGG1
is known to deal primarily with oxidative damage 
, we have shown that this bifunctional glycosylase provides a backup for cleavage at AP sites following induction of MMS damage since derived DSBs that appeared in the apn1/2 ntg1/2
mutant were prevented by a further ogg1
mutation (). This is the first direct demonstration for the Ogg1 glycosylase dealing with lesions other than oxidative damage in vivo
, suggesting a potentially more general role for this gene in repair. Considering that the predominant lesions induced by MMS are N7-methylguanine and N3-methyladenine 
, the function of Ogg1 in the development of DSBs and SMD is likely due to its action on AP sites derived from N7-methylguanine. There was still a small amount of DSBs after removal of all the bifunctional glycosylases/lyases () which might be due to NER or some other enzymes. It was shown that DNA Topoisomerase I (Top1) forms DNA-protein adducts with nicked and gapped DNA structures 
. Possibly the AP sites could also be processed by yeast topoisomerases to generate DSBs. This might explain the small amount of SMD presented in apn1/2 ntg1/2 ogg1
As summarized in , the generation of derived DSBs would require that opposed AP sites either be sufficiently close (left side of figure) so that DSBs are created directly in vivo
or there is a nick-processing mechanism that “moves” the relatively distant opposing-nicks closer (central part of figure) to form a DSB. Considering that MMS is an SN2 type of alkylating agent that methylates DNA bases in a random manner with a limited ability to produce closely-spaced lesions under the conditions used in this study (in contrast to ionizing radiation 
), many of the MMS-derived DSBs might be generated from distant single-strand breaks during processing/repair of the end-blocking groups as also suggested from our previous study with rad27
. Since AP lyases generate blocked 3′-ends (3′-dRP) while repair of either SSBs or DSBs requires an unblocked 3′-OH end for repair synthesis or ligation, we suggest that both the formation of DSBs and SMD are related to the processing of 3′-blocking groups. Either or both might be generated through development of 3′-flaps, possibly by helicases or nucleases. For example, opposing SSBs could be “moved” together to form a DSB if 3′-flaps are generated toward each other. Possibly it is the generation of multiple flaps that leads to the reduced mobility of large DNAs on PFGE, and the SMD molecules; however, the reduction in mobility is not as great as observed with replicating chromosomes, which remain in the well during PFGE. Although exonuclease 1 generated gaps at UV-damage sites can lead to reduced mobility, they are unlikely to be the source of SMD in the present experiments, While we have shown that the DSBs and SMD arise from a common BER intermediate, their subsequent appearance and disappearance are genetically separable. Importantly, we have established that SMD does not involve the HR pathway. The derived DSBs are subject to processing by MRX and the subsequent DSB repair as well as the appearance of dimer recombination products requires HR.
Scheme for the appearance of DSBs and SMD.
Regardless, there are limitations on the appearance of SMD. The loss of chromosomal DNA along with the appearance of the wide band of SMD following MMS treatment of apn1/2
cells is dependent on the size of the chromosome. SMD was substantially greater for larger chromosomes than smaller ones (). This is clearly shown in a Southern blot comparison of linearized Chr III with Chr II and in comparisons of the 230 kb Chr I and 270 kb VI with the larger chromosomes where there was little if any loss of the smaller chromosome bands (). Based on our previous results 
, we anticipate ~0.4 SSBs/kb which would lead to considerable damage in even the smallest chromosomes (~100 SSBs/Chr I). Thus, while SMD requires the generation of SSBs, other factors determine its appearance. Possibly, the appearance of SMD depends simply on the likelihood of producing some minimum amount of lesions or certain types of structures (i.e.
, sensitive to T7-endonuclease) that are stable in vitro
The requirement for generation of a 3′-flap to remove 3′-blocked termini had been proposed previously to explain the synthetic lethality between apn1/apn2
or rad10 
. Although in vitro
studies demonstrated that a 3′-flap can be removed by Rad1/Rad10 proteins 
, direct evidence for flap removal in vivo
has been lacking. The observation of SMD in our current study fits well with this hypothesis though the actual mechanism for its formation and release might be more complex than previously proposed. It is interesting that while we have eliminated SMD as a recombination product, it is sensitive to the T7 endonuclease I which can cleave structures that might arise during recombination as well as branched molecules containing single strand regions (possibly as a result of 3′-flap formation as proposed in ).
In conclusion, our study identifies new mechanisms for processing abasic sites and provides the first direct demonstration in nonreplicating G2/M cells of MMS-derived DSBs and that the DSBs are subject to recombinational repair. In addition, we identify and characterize the generation of SMD. While not previously described, possibly because of the techniques used to assess DNA damage and repair, SMD might be a general repair intermediate for various types of DNA damage, a view that we are currently pursuing. Interestingly, there has been an indication, though not directly addressed, of SMD-like material in exonuclease 1 defective yeast cells during excision repair of UV damage 
. The combination of genetics and systems for detection of novel structures has provided a unique opportunity to address processed events at intermediates in repair of DNA lesions. While the derivation and repair of derived DSBs has been addressed as well as the generation of SMD, it will be interesting to determine the specific nature of the actual DNA changes that lead to SMD and the eventual resolution including the genetic controls. To our knowledge, this is the first report of a novel branched repair intermediate being generated during the processing of 3′-blocked termini. These findings are expected to expand our understanding of mechanism for repair of 3′-blocked ends as well as their impact on genome stability.