This information was most recently updated October 31, 2006.

Repair of Strand Breaks

Double-strand breaks (ds breaks) are repaired by two different types of mechanism. One type takes advantage of proteins that promote homologous recombination (HR) to obtain instructions from the sister or homologous chromosome for proper repair of breaks. The other type permits joining of ends even if there is no sequence similarity between them. The latter process is called non-homologous end joining (NHEJ).

The process by which complex single-strand breaks (those that cannot be directly religated) are repaired (SSBR) in some ways resembles NHEJ. Here we shall discuss all three mechanisms: HR, NHEJ and SSBR.

Ds break repair by HR

When ds break repair is by HR, there are at least two categories of repair mechanism. The first and most important category, "synthesis-dependent strand annealing" (SDSA), requires that a single DNA strand find its complementary sequence within a double-stranded DNA molecule. The second category, called "single-strand annealing" (SSA), requires that two complementary single strands find each other. All of the proteins encoded by genes in the RAD52 epistasis group of S. cerevisiae (or their homologs in other eukaryotic organisms) are important for SDSA. These proteins include RAD51, RAD52, RAD54, RAD55, RAD57 and RAD59. However, only RAD52 and RAD59 are needed for SSA.

The RAD51 protein is a key member of the group of proteins needed for SDSA. It contains a central core that is rather similar to the RecA protein of E. coli. RecA forms a structured nucleoprotein complex with single-stranded DNA. When RecA-ssDNA complexes interact with homologous dsDNA, strand exchange can take place: the RecA-bound single strand can base-pair with its complementary strand in the dsDNA, displacing the other strand of the dsDNA.

RecA action

Similar reactions are catalyzed in vitro by eukaryotic RAD51 proteins, but the eukaryotic reactions are not as efficient as those carried out by RecA. Several types of evidence suggest that both of the proteins that are frequently defective in hereditary breast cancer, BRCA1 and BRCA2, can interact with human RAD51 and thus may play a role in ds break repair, either directly or indirectly in a regulatory capacity.

The following diagram illustrates that SDSA requires invasion of dsDNA by ssDNA, whereas SSA does not. That is why RAD51 and additional proteins (RAD54, RAD55, RAD57) are needed for SDSA but not for SSA.


It now appears likely that most or all mechanisms that utilize information from a sister or homologous chromosome to repair a ds break with HR use some variation of the SDSA pathway. One of the simplest versions is shown above. In more complicated versions, the new synthesis primed by the invading 3' strand from the broken chromosome can lead to the development of a full-fledged replication fork, which in some cases can continue synthesis all the way to the end of the chromosome serving as information donor. This is called "break-induced replication." What distinguishes all versions of SDSA from earlier models of break repair by HR is the fact that the DNA synthesis is "conservative." That is, within the newly synthesized stretch of DNA in the final repaired chromosome, both strands are newly synthesized, and both template strands are restored to the template chromosome. SDSA models are based on the considerable genetic evidence (mostly from yeast) indicating that new DNA synthesis during ds break repair is ultimately conservative and thus differs from the semi-conservative replication that takes place during normal chromosome duplication.

In the simple version of SDSA shown above, repair by conservative DNA synthesis is accomplished in the following way. A ds break is introduced into one of two homologous chromosomes (green). The 5' ends are then resected by an exonuclease to expose the 3' ends (shown with arrowheads) in single-stranded form. With the help of RecA-like proteins, the 3' ends locate complementary regions in the sister (or homologous) chromosome (blue). This is sometimes called "strand invasion." Then the 3' ends are used as primers for new DNA synthesis (red), using the donor chromosome strands (blue) as template. After sufficient synthesis to permit the new strands to anneal with each other, the new red strands are unwound from the blue template and allowed to anneal with each other. Any overhangs are removed by a flap endonuclease, and any gaps are filled in by a polymerase. Remaining nicks are sealed by a ligase. The green chromosome is now repaired, but it contains information from the blue chromosome in the newly synthesized region.

Ds break repair by single-strand annealing (SSA) begins in similar fashion. After a break is introduced, the 5' ends are resected. However, this resection exposes complementary regions within the 3' strands (due to repeat sequences (green boxes)) flanking the ds break. In mammalian DNA, with its abundance of short repeat sequences, it is not unlikely that repeat sequences of sufficient length (several hundred base pairs) should be found flanking the ds break. After flap removal (by a FEN1-like nuclease) and ligation, the ds break is repaired—but at the price of deletion of the stretch of DNA between the two repeated sequences. This is a small price to pay, however, when one considers that the alternative would be loss of the entire chromosome fragment lacking a centromere if the ds break were not repaired.

Ds break repair by non-homologous end joining (NHEJ)

Although one might think that non-homologous end-joining would lead to the random joining of any two ends, this does not appear to be the case. Mouse cells lacking Ligase IV (one of the proteins required for NHEJ—see below) undergo numerous chromosome transclocations after DNA damage by ionizing radiation, but wild-type cells do not (Ferguson et al. PNAS 97:6630-6633, 2001), implying that functional NHEJ in wild type cells must preferentially join correct ends of ds breaks. The reason for the ability of NHEJ to join correct ends is not yet known. Several mechanisms may contribute. These will be discussed below.

Many of the proteins important for NHEJ have been identified. They were identified in several ways:

Genes and proteins important for NHEJ

 Yeast gene name  Mammalian gene name  Protein
LIG4 LIG4 DNA ligase IV
LIF1 XRCC4 XRCC4; in collaboration with KU, targets DNA ligase IV to DNA ends
This protein is not present in yeast XRCC7 DNA-PKcs
This protein is not present in yeast ARTEMIS Artemis; nuclease regulated by DNA-PKcs; important for preparing DNA ends to make them ligatable

The core set of NHEJ proteins is conserved in all eukaryotic cells (blue in the above table). Two additional proteins (green in the above table) are present only in vertebrates. These proteins presumably evolved to accomplish special features of the special NHEJ that is required for V(D)J recombination, a process found only in vertebrates. Despite their importance for V(D)J recombination, these proteins (DNA-PKcs and Artemis) are also important for general repair of ds breaks, because vertebrate cells lacking these proteins are hypersensitive to ionizing radiation (which produces ds breaks; note that it also produces other types of damage, especially oxidative damage and single-strand breaks).

For some of the mammalian genetic analyses, rodent cells sensitive to X irradiation (a form of ionizing radiation; X-rays produce ds breaks and additional types of damage) were separated into different complementation groups, human DNA was transfected into mutant rodent cells from each complementation group, and then human genes were isolated from those cells rendered X-irradiation-resistant by the human DNA. Several X-ray cross-complementing (XRCC) human genes important for NHEJ, XRCC4 - XRCC7, were isolated in this way.

The Ku heterodimer

The Ku heterodimer is an abundant DNA-binding protein with ATPase and possible helicase activity. It binds strongly to DNA ends, to stem-loop and bubble structures, and to transitions between ds DNA and two single strands. Once Ku binds to a DNA molecule at its ends, it can translocate along the DNA. There is weak evidence that Ku heterodimers bound to separate nearby ends may interact with each other and thus help bring the ends together. Ku recruits other proteins, including XRCC4, ligase IV, and DNA-PKcs to DNA ends.

DNA-PK and Artemis

The 465-kDal protein encoded by XRCC7 is the catalytic subunit (cs) of a DNA-dependent protein kinase (DNA-PK) activity. The Ku heterodimer interacts with and regulates DNA-PKcs. Thus DNA-PK is frequently considered to be a trimeric protein consisting of DNA-PKcs, Ku70 and Ku80. In the presence of DNA ends, DNA-PK phosphorylates serine or threonine preceding glutamine in a wide variety of substrates in vitro. Until recently none of its in vivo substrates were known.

Now we know that the Artemis protein is a substrate of DNA-PK. Artemis binds to DNA-PKcs and is thus recruited to Ku-bound DNA ends along with DNA-PKcs. Once bound to a DNA end, DNA-PKcs becomes active as a kinase, and it phosphorylates Artemis. It also phosphorylates itself (auto-phosphorylation). Recent experiments in which the DNA-PK-targeted amino acids in Artemis were changed to non-phosphorylatable amino acids demonstrated that phosphorylation of Artemis by DNA-PK has no functional significance. In contrast, auto-phosphorylation of DNA-PK produces a change in DNA-PK conformation, which, apparently, makes the DNA ends to which DNA-PK is bound accessible by Artemis, so that Artemis can use its nuclease activities to "clean up" the ends and make them good substrates for ligation.

DNA ligase IV and XRCC4

In addition to recruiting DNA Ligase IV, XRCC4 also recruits polynucleotide kinase, and probably also recruits DNA polymerases of the polymerase beta family, to ds breaks. These are end-processing enzymes that can assist in making DNA ends ligatable. Their recruitment to ends is independent of DNA-PKcs and Artemis and probably accounts for the ability of yeast cells to carry out efficient NHEJ in the absence of DNA-PKcs and Artemis.

The XRCC4 protein is an essential collaborator with DNA ligase IV in the ligation of double-stranded DNA molecules. Before this ligation can take place, both the 5' and 3' ends of each of the DNA strands to be joined must be processed to become proper ligation substrates (simple nick with 3'-OH and 5'-phosphate). The stoichiometry of XRCC4 and ligase IV molecules at ds breaks is not yet entirely clear, but it is likely that two molecules of ligase IV are involved, since there are two nicks to be sealed.

The basic steps of NHEJ

Our current view of the basic steps of of NHEJ in mammalian cells is summarized in this diagram:


DNA breaks are produced in numerous ways. In most cases, the ends of the break will not be suitable for direct ligation. In the case shown here, some nucleotides have been lost from the 5' end of the bottom strand of the left-hand fragment and from the 3' end of the bottom strand of the right-hand fragment.

As indicated above, in most cases NHEJ rejoins the correct ends, thus preventing chromosome rearrangements. To accomplish this, the two ends must be held together until they can be ligated. How this is accomplished is not known, but holding the ends together, which is called "synapsis", is likely to be accomplished by a combination of the following:

It has been calculated that Ku heterodimers (yellow spheres in above diagram) are so abundant in mammalian nuclei that any ds break would on average be only about 5 molecular diameters away from the nearest Ku dimer. Thus binding of Ku to broken DNA ends is unlikely to be the rate-limiting step in repair by NHEJ.

Although DNA-PKcs (magenta oval in above diagram) binds DNA ends in the absence of Ku, its affinity for ends is increased about 100-fold when Ku is already bound to those ends. Artemis (small blue oval) binds to DNA-PKcs even in the absence of DNA ends. Thus when DNA-PKcs is recruited to Ku-bound DNA ends, Artemis is probably brought to those ends along with DNA-PKcs.

Binding of DNA-PKcs to Ku at DNA ends activates the protein kinase activity of DNA-PKcs, permitting it to phosphorylate Artemis and itself. As indicated above, DNA-PKcs auto-phosphorylation appears to make DNA ends available to Artemis' endonuclease activity, permitting that endonuclease activity to open hairpin loops (if present) and to cut away single-stranded overhangs at DNA ends. Additional nucleases and polymerases may be necessary to process both ends in such a way that they are separated by ligatable nicks (a 3'-OH end adjacent to a 5'-phosphate end). In budding yeast the relevant polymerase is Pol4, which is in the same family as mammalian polymerase beta. The corresponding polymerase(s) in mammalian cells has/have not yet been identified but may prove to be polymerase lambda and/or mu, which are other members of the same family.

Polynucleotide kinase also contributes importantly to end processing. This protein has two relevant enzyme activities. It can transfer phosphate from ATP to 5'-OH groups (which is why it is called a "kinase"), and it can remove phosphate from 3'-phosphate groups. Thus it is also a phosphatase. The consequence of these actions is the conversion of nicks that were nonligatable (due to 5'-OH or 3'-phosphate groups) to ligatable 5'-phosphate/3'-OH nicks.

XRCC4 (small orange oval) and DNA ligase IV (small green spheres) are also recruited by Ku. XRCC4 recruits polynucleotide kinase and probably polymerase-beta-like polymerases to assist in end processing. Then ligase IV catalyzes the ligation of the two strands. The resulting "healed" DNA molecule is likely to have altered DNA sequence, with the extent of alteration depending on the amount of damage at the break and the extent of processing that was required to make the break ligatable.

The MRN/MRX protein group

The MRN/MRX proteins play important, but apparently varying, roles in ds break repair in eukaryotic organisms. This group consists of Mre11, Rad50 and Nbs1 (mammals and fission yeast) or Xrs2 (budding yeast).

Orthologues of Mre11 and Rad50 exist in all kingdoms of life and perform a wide variety of DNA-related functions. The third members of this group, Nbs1 and Xrs2, are much less well conserved and less is known about their roles. The MRN/MRX groups are essential during meiosis for both the creation and the repair of the programmed ds breaks that initiate meiotic recombination. In budding yeast, the MRX proteins stimulate Rad51-dependent HR and are required for the bulk of NHEJ. In contrast, the MRN proteins play no essential role in NHEJ but are required for ds break repair by HR in fission yeast and mammalian cells. Available evidence does not rule out the possibility that MRN may contribute to NHEJ in mammals and fission yeast, but if MRN does contribute to NHEJ in these organisms, then another protein(s) must be able to substitute for the MRN proteins when any of these proteins is mutated or depleted.

The Mre11 proteins are also important for checkpoint signaling. Certain mutations in the human NBS1 gene cause "Nijmegen breakage syndrome" (NBS), a rare autosomal recessive disease characterized by microcephaly, immunodeficiency and increased frequency of hematopoietic cancers. Cells from NBS patients suffer from frequent chromosome breakage. In normal cells, the RAD50, MRE11, and NBS1 proteins all co-localize in numerous spots (foci) within nuclei after induction of ds breaks by ionizing radiation. Ds-break-induced co-localization in nuclear foci does not occur in cells from NBS patients. Many of these symptoms are identical to those of the checkpoint disease, ataxia telangiectasia (AT), in which the key checkpoint signaling protein, ATM, is mutated. Similar symptoms are also present in "Ataxia telangiectasia-like disease" (ATLD), which has recently been found to be due to mutations in the MRE11 gene. In fact, the only major difference between AT, NBS and ATLD is that p53 responses are defective in AT but normal in NBS and ATLD. These observations suggest that NBS1 and Mre11 (and probably Rad50 as well) may mediate all checkpoint responses downstream of ATM with the exception of the p53 responses.

The mechanisms by which the MRE11 complex affects checkpoint pathways are currently being elucidated. Evidence at this time suggests that mammals respond to ds breaks by activating the ATM protein kinase, which then phosphorylates several different substrates. Among these are Rad50, Nbs1 and Chk2. Chk2 is also a protein kinase. Phosphorylation of Chk2 by ATM is required to activate Chk2 kinase activity, and Chk2 kinase activity is required for signal transmission downstream of ATM. However, the phosphorylated form of Nbs1 (and possibly also of Rad50) is also required for ATM-dependent activation of Chk2. This explains why the symptoms of AT, NBS and ATLD are so similar: in each case, Chk2 protein is not activated in response to ds breaks. The pathways downstream of Chk2 that enhance cell survival after ionizing radiation are under current study. A similar pathway exists in budding yeast cells, where the Tel1 kinase (similar to ATM) is required to phosphorylate Xrs2 (which seems to function similarly to Nbs1), Mre11 and Rad53 (similar to Chk2), and phosphorylation of Rad53 requires functional Xrs2 and Mre11 proteins.

Recent studies are beginning to suggest biochemical mechanisms by which the MRN/MRX proteins carry out their multiple functions. Mre11 is a small protein with a 3' to 5' exonuclease activity. In addition, genetic studies in yeast suggest that Mre11 regulates the 5' to 3' exonuclease that is responsible for resection of the 5' ends of ds breaks prior to SDSA or SSA (see diagram above). In contrast, Rad50 is a large protein with a long coiled coil domain separating two half-ATPase domains. Rad50 and Mre11 interact to form heterotetramers composed of two Mre11 molecules and two Rad50 molecules. I shall refer to this as an MR dimer. A zinc-hook protein-protein interaction motif in the middle of the coiled-coil domain permits dimerization of the MR dimer to form MR tetramers, as diagrammed here. This diagram is based on structural studies by Hopfner et al., Nature 418:562-566, 2002.

Rad50 structure and interactions with Mre11

The diagram of Rad50 shows the two half ATPase domains (A and B; green), the two Mre11-interaction domains (M; blue), the coiled-coil regions (CC; orange), and the zink-hook interaction motif (H; red). In Rad50 molecules, the two ATPase domains come together to create a "head" with a long coiled-coil tail. Mre11 binds to Rad50 near its head. Both MR dimers and MR tetramers can bind to DNA. The diagram shows how two DNA molecules might be tethered at their ends by an MR tetramer.

These properties of the MRN/MRX proteins suggest that they may play two roles in ds break repair (whether by NHEJ or HR): (i) helping to hold two DNA ends close to each other and (ii) end processing (with Mre11's nuclease activity). Indeed, it is tempting to speculate that the MRX complex in budding yeast cells may contribute so much to end processing that it compensates for budding yeast's lack of DNA-PKcs and Artemis. Additional roles for MRN/MRX are also possible. There's clearly need for more research about the roles and mechanisms of this interesting protein complex.

Chromatin modifications at sites of ds break repair

In mammalian cells, a minor form of histone H2A is called H2AX. This form has a serine at position 139 that becomes phosphorylated (by the ATM and/or ATR checkpoint kinases) in response to local DNA damage. The analogous form of H2A is the major form in budding and fission yeasts, and it also becomes phosphorylated in response to local damage. The phosphorylated form is called gamma-H2AX. In mammalian cells, regions of gamma-H2AX can extend for millions of base pairs from a ds break. However, in yeast cells the region of gamma-H2AX is relatively short, only approximately 25 kb on either side of the break, as was recently demonstrated for a break that could be induced with high efficiency at a unique site in S. cerevisiae chromosome III:

Distribution of Gamma-H2AX and Mre11 at a defined DSB

Note that there was a decrease in the amount of gamma-H2AX right at the break. In the same region, Mre11 (one of the proteins of the MRX complex [see above]) could be detected, consistent with the possibility that the MRX complex might be holding the two DNA ends together at the break. In addition, Rad51 and other proteins required for ds break repair by HR were found in the same region as Mre11, suggesting that the basic break repair process is confined to the region close to the break.

What, then, is the role of gamma-H2AX? Other laboratories have found that gamma-H2AX helps to recruit cohesins (proteins that hold sister chromatids together) to large regions surrounding ds breaks. This may help to promote error-free homologous repair (using the sister chromatid as a template) as opposed to error-prone single-strand annealing or NHEJ. Additional laboratories have found that gamma-H2AX helps to bring a chromatin remodeling complex, INO80, to ds breaks, and that the resulting chromatin remodeling facilitates progression of the exonuclease that resects 5'-terminated strands at ds breaks (see above). I suspect that in the near future we'll learn a great deal more about how both cohesins and chromatin-remodeling complexes facilitate repair of ds breaks and other types of DNA damage.

Repair of complex single-strand breaks

This diagram shows our current view of the basic steps of single-strand break repair (SSBR). The pathway shown here operates in mammalian cells but not in yeasts. Yeasts lack homologs to PARP, XRCC1 and DNA ligase III. How yeasts repair complex single-strand breaks is not yet clear.

Single-strand break repair diagram

The first step is detection of the break by poly (ADP-ribose) polymerase (PARP; yellow oval in diagram), an abundant protein that builds up chains of poly (ADP_ribose) on itself or other target proteins after it binds to single-strand breaks. Although it seems clear that PARP performs a signalling function that ultimately leads to the recruitment of other proteins to the break, it is not known if PARP recruits those other proteins directly. PARP may perform additional roles by modulating chromatin structure (some of its targets are chromatin proteins) and/or inhibiting recombination.

XRCC1 (orange oblong) and DNA Ligase III (green circle) are among the next proteins to be recruited to the break. XRCC1 resembles XRCC4 in the sense that both recruit DNA ligase and processing enzymes, such as polynucleotide kinase, to break sites. The processing enzymes usually needed for SSBR include polynucleotide kinase and polymerase beta.

Once the strand ends have been converted to 5'-P and 3'-OH, and once gaps have been filled in so that only a nick remains, the nick is sealed by DNA ligase III, and the SSBR proteins dissociate.

Link to next page