Mechanism of Suppression of Chromosomal Instability by DNA Polymerase POLQ

Although a defect in the DNA polymerase POLQ leads to ionizing radiation sensitivity in mammalian cells, the relevant enzymatic pathway has not been identified. Here we define the specific mechanism by which POLQ restricts harmful DNA instability. Our experiments show that Polq-null murine cells are selectively hypersensitive to DNA strand breaking agents, and that damage resistance requires the DNA polymerase activity of POLQ. Using a DNA break end joining assay in cells, we monitored repair of DNA ends with long 3′ single-stranded overhangs. End joining events retaining much of the overhang were dependent on POLQ, and independent of Ku70. To analyze the repair function in more detail, we examined immunoglobulin class switch joining between DNA segments in antibody genes. POLQ participates in end joining of a DNA break during immunoglobulin class-switching, producing insertions of base pairs at the joins with homology to IgH switch-region sequences. Biochemical experiments with purified human POLQ protein revealed the mechanism generating the insertions during DNA end joining, relying on the unique ability of POLQ to extend DNA from minimally paired primers. DNA breaks at the IgH locus can sometimes join with breaks in Myc, creating a chromosome translocation. We found a marked increase in Myc/IgH translocations in Polq-defective mice, showing that POLQ suppresses genomic instability and genome rearrangements originating at DNA double-strand breaks. This work clearly defines a role and mechanism for mammalian POLQ in an alternative end joining pathway that suppresses the formation of chromosomal translocations. Our findings depart from the prevailing view that alternative end joining processes are generically translocation-prone.


Introduction
A diverse group of at least 16 DNA polymerases carry out DNA replication, repair, and damage tolerance in the mammalian genome [1,2]. One of these is DNA polymerase theta (POLQ). POLQ homologs are found in multicellular eukaryotes including plants, but an equivalent enzyme is absent from yeast [3]. The large 290 kDa human POLQ protein has an unusual bipartite structure with an N-terminal helicase-like domain and a Cterminal DNA polymerase domain [4]. This domain arrangement and the POLQ protein sequence is highly conserved in vertebrates [3].
Several functions have been suggested for POLQ [3] including bypass of template DNA lesions such as abasic sites and thymine glycols [5,6], a backup role in DNA base excision repair [7,8], and influencing the timing of DNA replication origin firing [9]. Loss of POLQ homologs in Drosophila and C. elegans causes hypersensitivity to DNA interstrand crosslink (ICL)-forming agents [10,11] such as nitrogen mustards or cisplatin. A consistent picture of hypersensitivity to DNA damage in mammalian cells lacking POLQ has not emerged from studies reported so far [3]. Mice devoid of or carrying mutant alleles of Polq display an elevated level of micronuclei (indicating chromosome breaks) in their peripheral erythrocytes [12][13][14]. A further increased frequency of micronuclei is observed after ionizing radiation exposure, and is much elevated in Polq mutant animals [12,14]. The majority (,90%) of mice with double homozygous deficiencies in Polq and Atm die during the neonatal period, with surviving double mutant mice showing severe growth retardation [13]. From this observation it was suggested that POLQ has a unique role in maintaining genomic stability that is distinct from the major homologous DNA recombination pathway regulated by ATM [13].
DNA double-strand breaks (DSBs) can be formed in cellular genomes by environmental agents such as ionizing radiation. DSBs also arise during normal cellular duplication cycles, when DNA replication stalls at naturally occurring structures or at sites of internally-generated DNA damage. In diversification steps of the mammalian immune system, DSBs are deliberately formed by regulated enzymatic action, to initiate rearrangement of antibody and receptor segments, and as a means to introduce local variation. Because DSBs are toxic and/or mutagenic if not repaired, organisms have multiple mechanisms for DSB repair [15,16]. The primary strategies are end-joining mechanisms, which process and rejoin the ends of a DSB, and homologous recombination (HR) pathways which employ an undamaged copy of the DNA [17]. End-joining pathways appear to be the first line of defense again DSBs. The most studied pathway is ''classical'' non-homologous end-joining (cNHEJ), which relies on the DNAbinding Ku70 (XRCC6) and Ku80 (XRCC5) gene products, and the DNA protein kinase (DNA-PK, PRKDC). One or more ''alternative'' end-joining pathways (altEJ) also exist, which are independent of these factors [18,19]. During immunoglobulin diversification, the regional end-joining process of class switch recombination (CSR) replaces one constant region coding exon for another. This CSR process is known to occur through both cNHEJ and alternative end joining pathways [20]. In mammalian cells, an alternative end-joining repair pathway repair of DSBs is thought to play a role in driving the formation of chromosomal translocations, although the specific enzymology is unclear [21,22].
Here, we report experiments that define a specific function and mechanism of action for POLQ in a pathway for alternative endjoining of DNA double-strand breaks in mammalian cells.

Hypersensitivity to DNA strand-breaking agents in the absence of Polq
To clarify the cellular role of POLQ in response to DNA damage, we measured the sensitivities of Polq-null and Polqproficient bone marrow stromal cell (BMSC) lines to various DNA damaging agents. Cells lacking POLQ exhibit hypersensitivity to ionizing radiation ( Figure 1) [12,23], and to the double-strand break-inducing chemical bleomycin, as previously reported [12].
We found that Polq 2/2 cells were also hypersensitive to other agents which directly cause DNA breaks, including the topoisomerase II inhibitors etoposide and ICRF-193 [24] and camptothecin, a topoisomerase I inhibitor. In contrast, loss of Polq did not cause hypersensitivity to agents that largely form DNA replicationblocking adducts on one DNA strand including ultraviolet radiation and the methylating agent temozolomide. The Polq 2/2 cells were also not more sensitive than control cells to mitomycin C, cisplatin, and UVA-photoactivated psoralen plus UVA, all of which induce some interstrand DNA crosslinks (ICLs) (Figure 1).
These data indicate that POLQ is most important in a process conferring resistance to direct DNA strand-breaks, particularly double-strand breaks. Cells lacking Polq were not hypersensitive to the PARP inhibitor olaparib ( Figure 1) while control RAD51Ddefective cells were hypersensitive ( Figure S1A). This suggests that POLQ does not function in the BRCA/homologous recombination pathway [25]. POLQ-proficient cells treated with both olaparib and camptothecin were significantly sensitized compared to camptothecin alone. However, addition of olaparib to Polq-null cells only modestly increased the sensitivity to camptothecin ( Figure 1). Consequently, PARP and POLQ may operate within the same subpathway of DNA repair.

Loss of Polq enhances chromosomal instability in somatic cells
It is important to know whether the elevated level of micronuclei in Polq-defective cells extends to cell types other than peripheral erythrocytes. To answer this question, matched wildtype and Polq 2/2 BMSC lines were exposed to etoposide or xrays, and the number of cells with micronuclei after 48 h were enumerated (Figure 2A and B). Polq-null cells exhibited a ,3 fold increase in frequency of spontaneous micronuclei formation ( Figure 2C). Upon exposure to DNA damaging agents, the percentage of cells with micronuclei increased about 1.5-fold more per unit of damage for Polq 2/2 cells in comparison to Polq +/ + cells (Figure 2A and B). This shows that the susceptibility to micronucleus formation in Polq 2/2 cells is not confined to cells of the hematopoietic lineage, but occurs also in cultured cells, including fibroblast-like BMSCs.
Cells lacking Polq were analyzed for their ability to proliferate in culture. Two independent BMSC lines devoid of Polq expression proliferated at a rate comparable to a pair of wild-type control cells, the Polq BMSCs showing only a 5% increase in population doubling times ( Figure 2D and E). We extended this analysis to isogenic immortalized mouse embryonic fibroblast (MEF) cell lines ( Figure 2F and G). Polq 2/2 cells divided at a rate comparable to Polq-proficient cells. These findings fit with our previous observations that hematopoietic cell counts in irradiated Polq-null mice recovered at rates comparable to wild-type mice [12]. We have observed no major alterations in growth or development in unchallenged Polq null or mutant mice, consistent with previous reports [13,14,26]. These observations indicate that despite some increased chromosomal instability, POLQ-defective cells originating from a variety of tissues can proliferate at near-normal rates.

Author Summary
The reason for the hypersensitivity of POLQ-defective mammalian cells to ionizing radiation has been elusive. Here we show that POLQ-defective mammalian cells are selectively susceptible to double-strand breaks in DNA. We present experiments in mammalian cells showing that a specific double-strand break repair pathway is POLQdependent. To analyze the repair function in more detail, we examined class switch joining between DNA segments in antibody genes. Insertions of DNA bases are sometimes found at the joins between such segments, but the origin of these insertions has been mysterious. We show that this class of insertion joins during immunoglobulin classswitching is entirely POLQ-dependent. In experiments with purified human POLQ protein, we found a novel biochemical mechanism explaining the formation of the insertions. POLQ has a unique biochemical ability to extend DNA with minimal base pairing. Finally, we examined the biological consequences for chromosome stability. Unexpectedly, the Burkitt lymphoma translocation (a major cancer-associated genome instability) is enhanced in the absence of POLQ. This alters the current view about the action of DNA end joining in mammalian cells, revealing that a POLQ-dependent DNA repair pathway combats potentially damaging chromosome translocations.
The DNA polymerase activity of POLQ is required to confer resistance to DNA damaging agents We sought next to investigate which catalytic activities of POLQ are necessary to confer resistance to DNA damaging agents. Lentiviral-delivered expression vectors were constructed to express wild-type or mutant versions of POLQ in immortalized MEFs, in order to test for functional complementation ( Figure 3A). A tandem (D2330A,Y2331A) mutation was introduced into the DNA polymerase domain (POL); mutation of the corresponding residues in other DNA polymerases completely inactivates polymerase activity [27]. In a separate construct, a mutation was introduced into the conserved ATP-binding site of the Walker A motif (K121M) in the helicase-like domain (HLD). An equivalent mutation eliminates DNA helicase activity in related enzymes, including HELQ [28]. A third construct (DM) was made harboring mutations in both domains. These vectors expressed full-length recombinant POLQ as tested in transfected 293T cells ( Figure 3B and C).
The mutant cDNAs were tested for their ability to genetically complement the bleomycin sensitivity of Polq-null MEFs. Stable clones with each of the constructs were generated and analyzed for expression of POLQ ( Figure 3D). Independent clones of knockout MEFs expressing wild-type recombinant POLQ (WT4 and WT6) were able to rescue bleomycin hypersensitivity ( Figure 3E) as an antibody that recognizes endogenous POLQ does not yet exist. Neither the polymerase domain mutant (POL) nor the polymerase-helicase double mutant (DM) restored bleomycin sensitivity ( Figure 3E, Figure S1B). Expression of a construct with a mutation only in the helicase-like domain (HLD) was, however, still able to restore resistance to bleomycin. These data indicate that POLQ polymerase activity is essential for conferring resistance to DNA damage, while the ATPase activity of the helicase-like domain is not necessary. Similarly reintroduction of polymerase activity of POLQ into Polq-deficient MEFs was able to rescue chromosomal instability (micronuclei and DNA DSBs, as measured by 53BP1 and cH2AX colocalization ( Figure 3F and 3G, Figure S2).
Mice with an S1932P mutation in Polq (the ''chaos1'' allele) have an increased spontaneous frequency of micronuclei [13]. We generated a human POLQ cDNA mimicking the chaos1 mutation (S1977P), but attempted expression of POLQ with this mutation in 293T cells did not yield detectable protein ( Figure S3). This suggests that the chaos1-encoded mutant protein is unstable, consistent with the finding that chaos1 mice have a phenotype essentially indistinguishable from Polq knockout mice [13].
POLQ operates in a pathway of altEJ during mouse Ig class-switching Immunoglobulin class-switch recombination (CSR) uses DNA end joining to exchange one constant region of an antibody gene for another constant region. CSR can occur by both Kudependent classical non-homologous end joining and Ku-independent altEJ [20]. The overall frequencies of CSR are similar in Polq-defective mice [29] and cultured B cells [30]. To determine whether POLQ is involved in a mechanistically distinct subset of Figure 1. Hypersensitivity of Polq 2/2 bone marrow stromal cells to DNA strand-breaking agents. BMSCs were exposed to x-rays or UVC at the indicated doses, and to etoposide, ICRF-193, camptothecin, olaparib, temozolomide, mitomycin c, cisplatin, and HMT psoralen+UVA at the indicated concentrations and plated in triplicate. Two isogenic bone marrow stromal cell lines were used of each genotype, Polq +/+ or Polq 2/2 . Colonies were crystal violet stained and counted seven to ten days later. Experiments were repeated three times. Circles, Polq +/+ clone 1; Squares, Polq +/+ clone 1; Triangles, Polq 2/2 clone 1; inverted triangles, Polq 2/2 clone 3. doi:10.1371/journal.pgen.1004654.g001 CSR joins, we isolated and analyzed DNA sequences at such joins. Naïve B cells were isolated from the spleens of wild-type and Polqnull mice and stimulated for IgM to IgG class switching, and then the fraction of IgG1-positive B cells was measured by flow cytometry. Parallel B-cell cultures were incubated with NU7026, a DNA-PKcs inhibitor that suppresses cNHEJ [31]. It has been shown that B cells incubated with NU7026 have an increased proportion of CSR junctions with .1 bp insertion at the junction [31]. This suggests that when a pathway of altEJ is used during CSR, it more frequently results in insertion of nucleotides.
We found that B cells from Polq-proficient and deficient mice had similar overall frequencies of CSR ( Figure 4A), and inhibition of DNA-PKcs increased the frequency of CSR in both genotypes by 1.5 to 2 fold ( Figure 4B). The Sm-Sc1 junction was then sequenced from 100 clones of each group of IgG1-positive B cells. These data revealed that in wild-type B cells, insertions of .1 bp at Sm-Sc1 junctions, that are thought to be altEJ-dependent, comprised about 9% of total events, and that this increased to ,21% in cells incubated with NU7026 ( Figure 4C, Table 1).
Strikingly, in cells lacking Polq, this class of insertions at CSR junctions was absent, even in the presence of NU7026 ( Figure 4D, only one insertion of 2 bp observed). Insertion of .1 bp therefore requires POLQ. This class of Polq-dependent joining events included insertions of between 2 and 35 bp. For longer insertions (greater than ,10 bp) homologous sequences were unambiguously detected up to 2-5 kbp away from the junction site (Table 1), as has been reported for long insertions at Sm-Sc1 junctions in ATMdefective B cells [31]. This suggests that most or all of such insertions are formed in a templated manner during altEJ by POLQ.

Loss of Polq impairs an altEJ pathway but not cNHEJ in cultured cells
The most important factor in determining which double-strand break repair pathway is used is whether or not the 59 termini of broken ends are resected [32]. Ends with little or no single stranded overhang are typically rejoined by Ku-dependent cNHEJ. In contrast, CtIP and MRN-dependent resection of 59 termini generates ends with extended single stranded 39 overhangs; resection is thought to block cNHEJ [33] and enable repair by altEJ [34,35].
To analyze differing requirements for end joining, with or without end resection, we generated two linear DNA substrates with 39 single stranded overhangs; one with a short overhang (6 nt), and one a long overhang (45 nt, a ''pre-resected end'') ( Figure 5A). Both can be aligned with the same 4 nt of terminal complementary sequence. These substrates were then introduced into wild-type mouse fibroblasts or fibroblasts harboring deficiencies in Ku70 or Polq. Repaired products were recovered from cells and quantified. Repair of the short overhang substrate was, as anticipated, over 10-fold less efficient in cells without Ku70 ( Figure 5B) when compared to Ku70-complemented controls. The absence of Polq 2/2 had no consequence for repair of this substrate.
End joining with the 45 nt overhang substrate was assessed using qPCR primers located to ensure that at least 10 nt of overhang was included in joined products ( Figure 5A). Recovery of these products was no longer dependent on Ku; instead, it was increased 2.8-fold in Ku70-deficient cells ( Figure 5C). This is consistent with previous studies arguing Ku suppresses repair by altEJ. Strikingly, joining of the long overhang substrate in Polq 2/2 cells was reduced 10-fold, near background levels of signal observed using this assay. Complementation of the knockout cells with POLQ returned joining to wild-type levels ( Figure 5C). These data demonstrate that POLQ participates in some form of alEJ, but cells lacking POLQ maintain proficiency for cNHEJ.

POLQ extends 39 DNA ends in a template-dependent manner
Our results demonstrate that POLQ is necessary to form the insertions found in CSR junctions in a process of altEJ. We next sought to determine the mechanism. Like other DNA polymerases, an active polymerase fragment of POLQ [36] can catalyze template-dependent DNA synthesis from an annealed primer ( Figure 6A). As is common for family-A DNA polymerases, only a single nucleotide is added to the end of duplex DNA [5]. Unusually, however, POLQ can catalyze extension of singlestranded oligonucleotides [37]. It was unclear whether this reflects a robust terminal deoxynucleotide transferase activity of POLQ on single-stranded DNA, or some form of template-dependent synthesis. For example, POLQ can extend a single-stranded 16mer oligonucleotide provided without a complementary template (products up to 35 nt long), while E. coli pol I Klenow fragment has no activity on this substrate ( Figure 6B). The major 22 nt extension product produced by POLQ on the 16-mer used in Figure 6B may be accounted for by inter-or intra-oligonucleotide pairing ( Figure S4C). Neither POLQ nor Klenow fragment could extend an oligonucleotide that was incapable of annealing to itself ( Figure S4) [37].
To identify the mechanism of 39 single-stranded DNA extension by POLQ, we used a different single-stranded oligonucleotide designed to be unable to form self-complementary base pairs longer than a single nucleotide [37], and sequenced the products of POLQ-mediated extension. Individual extension products of 1 to 30 nt were found (Table S1). Most of the sequenced extension products feature AAC or AAAC sequences that could arise from copying GTTT sequences in the template via inter-or intramolecular priming and re-priming ( Figure 6C) following minimal base pairing at the 39-primer end. These data reveal that POLQ uniquely extends 39 DNA tails through template-dependent DNA synthesis from a primer with minimal base pairing and that the polymerase lacks true TdT-like activity. POLQ indeed has unique biochemical properties compatible with these observations. Unlike other DNA polymerases, POLQ can efficiently extend a DNA chain with a nucleotide incorporated opposite an abasic site [5], or from a mismatched primer-terminus [38]. Further, there is evidence that primers slip on DNA templates with an increased frequency during POLQ-mediated synthesis, as shown by the high frequency of single base pair frameshift mutations generated by purified POLQ [39].

POLQ suppresses chromosomal translocations in B cells
Double-strand breaks initiated by AID activity in the immunoglobulin heavy chain (IgH) locus of B cells are necessary to generate immunological diversity, but breaks are sometimes generated at other chromosomal sites, providing an opportunity for dangerous chromosome translocations [21,22,40,41]. For instance the oncogenic Myc/IgH translocation that causes Burkitt lymphoma is AID-dependent and requires breaks at both loci, with breaks in the Myc gene rate-limiting [42]. An altEJ process is implicated in the formation of oncogenic translocations in lymphoid tissues, including the Myc/IgH translocation in murine B cells [21,43,44]. cNHEJ suppresses the formation of such chromosomal translocations [45]. To determine the role of POLQ in chromosomal translocations, Polq +/+ and Polq 2/2 naïve splenic B cells were stimulated in culture and assayed for the frequency of Myc/IgH translocations ( Figure 7A). Notably, in the absence of  Polq there was a 4-fold increase in translocation frequency ( Figure 7B and C). This indicates that mammalian POLQ acts in a subset of altEJ events to suppress chromosomal translocations.
Additionally, an increase in intramolecular IgH rearrangements was found in B cells lacking Polq ( Figure 7B). Therefore, although POLQ is involved in an altEJ pathway, it prevents rather than promotes chromosomal instability, rearrangements and the formation of Myc/IgH translocations.

POLQ suppresses hypersensitivity to direct DNA doublestrand breaks
We show that in mammalian cells, POLQ has a specific role in defense against DNA damaging agents that cause direct DNA double-strand breaks, including ionizing radiation, bleomycin, and topoisomerase inhibitors. Our findings indicate that POLQ participates in a novel pathway of alternative-end joining of DSBs, a process that can occur throughout the cell cycle in mammalian cells [17]. The minimal additional sensitization to camptothecin by olaparib in Polq-defective cells suggests that one function of PARP is to participate in a Polq-dependent altEJ pathway. Our experiments indicate that POLQ is an important factor in DNA DSB repair in all cells, not just cells of the hematopoietic lineage. Indeed, Polq is broadly expressed in murine tissues ( Figure S5).
Mutants of POLQ homologs in Arabidopsis (TEBICHI), C. elegans (polq-1), and Drosophila (Mus308) are hypersensitive to ICL-inducing agents [3], whereas Polq-defective mammalian cells are not appreciably hypersensitive to such agents (Figure 1). This difference may arise because of differences between organisms in the priority of DNA repair pathway engagement. In proliferating mammalian cells, ICLs are usually dealt with through the Fanconi anemia pathway, which produces enzymatically induced doublestrand breaks that are channeled into homologous recombination repair [46]. In Drosophila and some other organisms, an altEJdependent pathway may be more important for resolving ICLassociated double-strand breaks. Although Drosophila Mus308 mutants are not hypersensitive to IR, pronounced IR sensitivity occurs in a double mutant when HR is also inactivated [47]. The phenotypic consequences of POLQ-dependent altEJ of doublestrand breaks may thus depend on the relative dominance of HR which varies between organisms.
We show here that the DNA polymerase activity of POLQ is necessary to prevent cell death and chromosome breaks (micronuclei) caused by a double-strand break-inducing agent. Disruption of the ATPase activity in the helicase-like domain of POLQ did not, however, alter the correcting function of POLQ addition to knockout cells. A previous study with mouse cell lines suggested that disruption of the polymerase domain of the murine Polq gene is less severe than complete disruption of Polq [30], but the result is difficult to evaluate in the absence of quantitative measurements of expression of the partially deleted form. No activity has yet been shown for the helicase-like domain, other than DNA-dependent ATPase function [4]. It is likely that an additional role remains to be discovered that is dependent on the ATPase function of POLQ.
POLQ aids DNA double-strand break repair through alternative end joining and nucleotide insertions When double-strand breaks form in mammalian cells, a majority will be repaired through cNHEJ. However, a subset of these breaks will be handled by alternative end-joining pathways in situations where the DNA end is not compatible with processing by Ku-dependent cNHEJ, or if core components of the cNHEJ machinery are absent or unavailable ( Figure 7D). In general, altEJ is defined as a means for repair of chromosome breaks that is exclusive of Ku-dependent, classically defined NHEJ [48], and dependent on factors (CtIP, MRN) that resect double-strand breaks to generate extended 39 ssDNA tails [34,35] (Figure 5A). Accordingly, we showed joining of a ''pre-resected'' extrachromosomal substrate (substrate with 45 nucleotide 39-ssDNA tails) was stimulated in Ku-deficient cells, similar to results using chromosomal substrates [35]. Joining of this substrate was also dependent on Polq ( Figure 5C). Our experiments thus define an altEJ subpathway in mammalian cells that involves POLQ (termed synthesis-dependent end joining, SD-EJ, in Figure 7D), Additional Polq-independent altEJ subpathways may also be operational ( Figure 7D). To some extent, different end-joining pathways can be been distinguished from one another by the ligase employed in the pathway, with DNA ligase IV (LIG4) suggested as essential for cNHEJ, and DNA ligase III (LIG3) for altEJ in mammalian cells [21,43,49]. There are caveats, however. For example, some functional redundancy is apparent between LIG1 and LIG3 in altEJ [44,[50][51][52]. Ligase deficiencies may thus not be the best marker for distinguishing different end-joining pathways. For the  altEJ subpathway under consideration here, dependence on POLQ is the best available definition. The biochemical properties of POLQ provide a mechanistic explanation for its contribution to altEJ. POLQ has a unique ability to add nucleotides to the 39 ends of single-stranded DNA [37], primed by minimal pairing with other available DNA molecules ( Figure 6 and Figure S6). Synthesis by POLQ in this context is consistent with the unusually efficient ability of the polymerase to extend from mismatched DNA termini [5,38], and its tendency towards primer-template slippage [39]. In further biochemical experiments it will be of interest to examine the action of POLQ and DNA ligases at double strand breaks with 39-singlestranded overhangs that closely mimic the resected ends of a DNA double-strand break. In vivo studies with such substrates, including those that can form hairpins in the single-stranded region, would give insight as to the preferred structures for POLQ-catalyzed extension.
Unique to the POLQ-dependent altEJ process are frequent joins displaying templated DNA insertions. Some form of altEJ has been implicated in resolution of a subset of double-strand break intermediates in CSR, producing templated insertions [20]. Our data support a role for POLQ in generating the CSR products with these templated insertions. These events are consistent with the templated insertions that occur during Mus308-dependent repair of directed double-strand breaks in Drosophila [47,53] and in C. elegans [54]. In the absence of POLQ, the lack of insertioncontaining joins is observed, but the global CSR frequency is relatively unchanged (Figure 4). These insertions are best explained by repeated initiation of synthesis by POLQ ( Figure 6C) on template sites, ultimately leading to a joined product.

POLQ prevents the formation of Myc/IgH chromosomal translocations
In the absence of POLQ, we found a ,4-fold increase in the formation of the oncogenic translocation Myc/IgH in mice. This increase is comparable to that seen in B cells that have lost Tdrd3, a regulator of R-loop formation during transcription [55] and miRNA-155 which regulates AID and suppresses oncogenic translocations [56]. In the absence of Polq there is also an apparent enhancement of rearrangement events in the IgH locus, consistent with the elevated level of chromosomal instability observed in cells lacking POLQ [57].
altEJ is typically associated with frequent annealing of the DNA ends at existing microhomologies (2-5 bp) and large deletions at repair junctions [19]. Since translocations commonly feature such microhomologies at their breakpoint junctions [58,59] and occur more frequently in cNHEJ defective cells, altEJ is considered the primary mechanism by which translocations occur. Thus, a striking finding of the present work is that the formation of Myc/ IgH translocations is suppressed when the POLQ-dependent altEJ subpathway is operational. It is possible that DNA DSBs persist for a longer time in the absence of POLQ, giving more opportunity for the formation of translocations. Alternatively, the POLQdependent pathway may be the most efficient at repairing a structurally distinct class of translocation-prone DNA breaks.
These studies clearly define a role for POLQ in the repair of DNA strand-breaking agents and provide a mechanism of template-dependent extension of DNA ends necessary to repair breaks in a subpathway of altEJ. This distinct altEJ pathway is necessary to prevent the formation of chromosomal translocations as shown by our in vivo experiments. It has been suggested that suppression of POLQ may be useful in increasing the efficacy of DNA damaging treatments in cancer [3,23,60]. This promising prospect should be tempered with the knowledge that loss of POLQ may also lead surviving cells to be prone to potentially oncogenic chromosome translocations.

Ethics statement
Research mice were handled according to MD Anderson Cancer Center Institutional Animal Care and Use Committee policies and protocol 08-08-08732. Mice were euthanized by CO 2 euthanasia followed by cervical dislocation.

Cellular proliferation assay
Polq +/+ and Polq 2/2 bone marrow stromal cells and mouse embryonic fibroblasts were plated in triplicate (200,000 cells per 10 cm dish) with 15 mL of complete media (Dulbecco's Modified Eagle Medium+Glutamax, 10% FBS, 1% PennStrep). On the indicated days, cells were trypsinized and live cells were counted using trypan blue exclusion (Countess automated cell counter, Life Technologies). Experiments were repeated three times in order to generate standard deviations. Viability was consistently high for all cell lines examined (.95% trypan blue-excluding cells).

Clonogenic assays with bone marrow stromal cells
For X-irradiation 5610 5 cells were plated on a 10 cm plastic culture dish, and exposed the following day at 2 Gy/min, 160 kV peak energy (Rad Source 2000 irradiator, Suwanee, GA). Cells were then trypsinized for replating. For UVC-irradiation (254 nm peak germicidal lamp) cells were irradiated in 500 ml PBSA (10 5 cells/ml) at 5 J m 22 min 21 and then plated. For psoralen-UVA treatment, 5610 5 cells were plated on a 10 cm dish and incubated in medium with the indicated concentration of HMT-psoralen for 1 h, the dish was irradiated for with 0.9 kJ m 22 UVA (365 nm peak, 30 min, 0.5 mJ m 22 sec 21 ), the psoralen-containing medium was removed, and the dish UVA-irradiated in fresh medium for a further 30 min before replating. Chemicals were added at the indicated concentrations to dishes at the beginning of the experiment. Drugs were solubilized in ethanol (mitomycin c), DMSO (ICRF-193, etoposide, camptothecin, HMT-psoralen, temozolomide, olaparib), or 150 mM NaCl (cisplatin). All chemicals were from Sigma (St. Louis, MO) except ICRF-193 (Enzo LifeScience, Farmingdale, NY), olaparib (AZD2281, Selleck Chemicals, Houston, TX), and mitomycin c (Calbiochem, Darmstadt, Germany). Cells were plated in triplicate in 10 cm dishes and grown for 7-10 days before being fixed and stained with crystal violet. Colonies of 50 or more cells were quantified and experiments were repeated three times to generate standard  Table S1. This model depicts a 12 nt extension product in Table S1. The product can be produced by a series of annealing, extension, slippage and repriming events. doi:10.1371/journal.pgen.1004654.g006 deviations. A clonogenic assay was performed with Rad51D +/+ and Rad51D 2/2 Chinese hamster ovary (CHO) cell lines exposed to varying concentrations of olaparib.

Micronucleus assay
BMSCs were plated at 1.5610 4 cells per well in chambered slides and treated with the indicated amount of x-rays or etoposide the following day. 48 hr later, cells were fixed with 2% para-formaldehyde, stained with DAPI and coverslipped. Micronuclei were scored by immunofluorescence for 300 cells per group. Experiments were repeated three times to generate standard deviations.

Generation and complementation of Polq MEFs
Polq-null (Polq 2/2 ) mice [13] were obtained from Jackson Laboratories and maintained on a C57BL/6J background. Isogenic primary MEFs were generated from 13.5 day pregnant females and cultured in a 2% O 2 atmosphere. MEFs were then transfected with 1 mg of pSV-Tag [61,62] and grown in atmospheric oxygen for six population doublings to allow for immortalization. To generate lentivirus used for transduction, 293T cells were cotransfected with psPAX2 (6 mg), pMD2G (6 mg), and pCDH (12 mg) expression vector (See Text S1 for construction of expression vectors) using Lipofectamine 2000. One day prior to transduction Polq 2/2 MEFs were seeded into a 10 cm dish at 1.5610 5 cells with 12 mL complete media. 48 hr posttransfection virus-containing media was harvested, filtered through a 0.45 mm syringe filter and used to replace the media on the plated MEFs. MEFs were incubated in the virus-containing media for 24 hr before being split into T-75 flasks and allowed to grow to 80% confluence before undergoing three weeks of puromycin selection (2.5 mg/mL). Following selection, pure clones were isolated and cultured with complete media containing puromycin (1 mg/mL).

Quantitative real-time PCR analysis of complemented MEFs
RNA isolated from the complemented MEF lines were analyzed for quality and purity using RNA 6000 Nano kit (Agilent Technologies, Santa Clara, CA). 1 mg of total RNA was used to generate cDNA using the High Capacity cDNA RT kit (Life Technologies). qPCR analysis was performed in triplicate using the ABI Prism 7900 HT thermocycler and the following Taqman

Bleomycin sensitivity
Complemented MEF lines were plated in triplicate into white 96 well plates at 1250 cells per well and grown overnight using complete media containing puromycin (1 mg/mL). The following day, cells are cultured with complete media containing the indicated amounts of bleomycin (dissolved in 150 mM NaCl) for 24 hr before the media was replaced. Cells were allowed to recover for 72 hr before cellular viability was measured using the ATPlite 1Step kit (Perkin Elmer, Waltham, MA) using a Biotek plate reader. Experiments were repeated three times.

Immunofluorescence
Complemented MEF lines were plated at a density of 1.5610 4 cells per well in 4-well chamber slides and the following day were irradiated with either 0 or 6 Gy of x-rays. Media was changed and cells were allowed to recover for 48 hr after damage before fixation with 2% para-formaldehyde and permeabilized with Triton X-100. Samples were blocked with donkey serum for 30 minutes before being incubated overnight with primary antibodies against 53BP1 (Bethyl, Montgomery, TX, A300-272A, 1:500) and cH2AX (EMD Millipore 05-636, 1:400). Cells were later incubated with AlexaFluor-488 goat-anti-mouse or AlexaFluor-594 goat-anti-rabbit secondary (Life Technologies, 1:1000) and then stained with DAPI before being coverslipped. Cells were scored for DSBs by enumerating the percentage of cells with .2 53BP1 foci and .2 cH2AX foci [61,63]. The majority of cells that contained .2 foci for each of the DSB markers, exhibited colocalization of the foci. Cells with pan-staining of cH2AX were not included in the analysis as they are proposed to represent pre-apoptotic cells [64]. Many of the cells with 53BP1 foci, exhibited enlarged foci that are associated with nuclear OPT (Oct-1, PTF, transcription) domains that sequester damaged DNA in G1 [65,66]. Thus, most of the MEFs that were foci positive contained DSBs [65]. DAPI-stained micronuclei were also scored. For each experiment 250 cells were scored for three independent experiments for a total of 750 cells.

B cell culture and CSR analysis
B cells were isolated from mouse spleens (n = 6 per genotype) and stimulated for class-switching in culture for 72 hr. Where indicated, cultures were incubated with DNA-PKcs inhibitor 20 mM NU7026 (Tocris, Bristol, UK) dissolved in DMSO, or mock-treated. The stimulation procedure and flow-sorting for CSR analysis was as described [31,67]. Prior to this analysis, cells were counted; numbers and viability were similar for all groups. Sm-Sc1 CSR junctions were amplified by PCR using the following conditions for 25 cycles at 95uC (30 s), 55uC (30 s), 68uC (180 s) using the primers (FWD 59-AATGGATACCTCAGTGGTTTT-TAATGGTGGGTTTA-39; REV 59 CAATTAGCTCCTGCT-CTTCTGTGG-39) and Pfu Turbo (Stratagene, La Jolla, CA). To the PCR reaction, 5 U of Taq polymerase (Promega, Madison, WI) was added and incubated at 72uC for 10 min. The resulting product was TOPO TA cloned and transformed into Top10 E. coli cells (Life Technologies, Carlsbad, CA) and plasmids were purified and sent for sequencing using M13 FWD and REV primers in addition to the amplification primers for sequencing. 100 clones for each group were analyzed for mutations, deletions, insertions, and sequence overlaps at the junction and both 30 nt upstream and downstream of the junction. p-values were determined by using two-tailed Fisher's exact test.

Translocation assay
Naïve B cells from three pairs of Polq +/+ and Polq 2/2 mice were harvested as above, cultured for 72 hr, and DNA was isolated. 32 separate PCR reactions, each containing the genome from 1610 5 cells, was performed with primers to amplify Myc/IgH translocations and amplified translocations were verified by Southern blotting using internal probes to the Myc and IgH loci as described [68,69]. Three independent experiments were performed and the p-value was determined using two-tailed Fisher's exact test. %IgG1 was also measured as an internal control to ensure the B cells from each genotype were switching at a comparable level. Figure S1 Cell sensitivity assays. (A) Clonogenic assay of Rad51D +/+ and Rad51D 2/2 CHO cells treated with the indicated doses of olaparib, as a positive control [70]. Colonies were crystal violet stained and counted eleven days later. (B) Polq +/+ , Polq 2/2 , Polq 2/2 Empty, and multiple clones of Polq 2/2 POL and Polq 2/ 2 DM MEF lines were treated with 1 mM bleomycin for 24 hr and cellular ATP levels were measured 72 hr later. (EPS) Figure S2 Analysis of DNA double-strand breaks and micronuclei in complemented Polq MEFs. Representative immunofluorescence images of Polq 2/2 Empty and Polq 2/2 WT4 MEF lines stained with DAPI and antibodies against 53BP1 and cH2AX. Scale bar represents 25 mm. Arrows note micronuclei. Pan staining of cH2AX (P), OPT domain staining by 53BP1 (OPT), and examples of colocalization (Co) are noted. (EPS) Figure S3 Full-length chaosI mutant protein is poorly expressed. 293T cells were transfected with pCDH constructs that contained either the human POLQ cDNA, ChaosI mutant (S1977P, corresponding to the S1932P mutated residue in mice), or empty control. (A) 48 hr post transfection lysates were made and immunoblotted with antibodies against FLAG and alpha-Tubulin. (B) Total RNA was isolated from cells and qPCR analysis of POLQ transcript levels were performed using the DDCt method. * denotes non-specific band. (EPS) Figure S4 POLQ does not extend a single-stranded oligonucleotide that cannot self-anneal. Increasing amounts of exonucleasedefective E. coli polI Klenow fragment denoted as Kf exo-and POLQ were incubated with (A) 59-32 P-labeled 20-mer dC oligonucleotide or (B) a 20-mer dC oligonucleotide with a synthetic abasic site on the 39 end. All reactions include all four deoxynucleotides and were incubated at 37uC for 20 min. The first lane contained no enzyme. The percentage (%) of the product extension from the primer is shown below each lane. * indicates nonspecific band. (C) Model of major 22 nt product formation produced by extension from the primer in Fig. 6B, which has a limited ability to self-anneal. (EPS) Figure S5 Absolute quantification of transcript number shows that Polq is broadly expressed in tissues. Total RNA was isolated from necropsied C57BL/6 mice by Trizol (Life Technologies). Transcript abundance for Polq and DNA polymerase delta catalytic subunit Pold1 was determined by absolute quantification method as described in the Experimental Procedures. Gapdh was used as an internal control to normalize samples. (EPS) Figure S6 Mechanism of insertion formation by POLQ during double-strand break repair. After a double-strand break is formed (i), the broken ends are frequently resected enzymatically to form 39 single-stranded tails (ii). POLQ can extend a 39 end by templated synthesis from another available DNA strand (iii). This may be the 39-tailed end near the break, or a DNA strand available at a more distant location, or through possible ''snapback synthesis'' whereby the tail serves as its own intramolecular template. POLQ has the unique ability to prime DNA synthesis (blue) with minimal base pairing, sometimes slipping in the process (main text Fig. 6). The newly synthesized DNA end then anneals via microhomology to the other 39 tail at the break (iv), and repair is completed (v) after further DNA synthesis (gray). This process results in an insertion of DNA sequences (blue).

(EPS)
Table S1 POLQ-dependent extension of 39 DNA ends relies upon homology. Primer extension assays were carried out with a defined substrate (top) incubated in the presence of an active polymerase fragment of POLQ. Products were then incubated with terminal deoxynucleotidyl transferase (TdT) in the presence of only dA, dC, dG, or dTTP to create extension products that terminate with homopolymeric runs. The reaction products were then cloned and sequenced for analysis of extension products. Extension products are listed above for each homopolymeric run. (XLSX) Text S1 Additional materials and methods. Expanded methods, including generation of POLQ antibody and POLQ expression constructs, PCR primers, oligonucleotide substrates, and sequencing of DNA extended by POLQ. (DOCX)