Spatial separation of replisome arrest sites influences homologous recombination quality at a Tus/Ter-mediated replication fork barrier

ABSTRACT

The Escherichia coli replication fork arrest complex Tus/Ter mediates site-specific replication fork arrest and homologous recombination (HR) on a mammalian chromosome, inducing both conservative “short tract” gene conversion (STGC) and error-prone “long tract” gene conversion (LTGC) products. We showed previously that bidirectional fork arrest is required for the generation of STGC products at Tus/Ter-stalled replication forks and that the HR mediators BRCA1, BRCA2 and Rad51 mediate STGC but suppress LTGC at Tus/Ter-arrested forks. Here, we report the impact of Ter array length on Tus/Ter-induced HR, comparing HR reporters containing arrays of 6, 9, 15 or 21 Ter sites—each targeted to the ROSA26 locus of mouse embryonic stem (ES) cells. Increasing Ter copy number within the array beyond 6 did not affect the magnitude of Tus/Ter-induced HR but biased HR in favor of LTGC. A “lock”-defective Tus mutant, F140A, known to exhibit higher affinity than wild type (wt)Tus for duplex Ter, reproduced these effects. In contrast, increasing Ter copy number within the array reduced HR induced by the I-SceI homing endonuclease, but produced no consistent bias toward LTGC. Thus, the mechanisms governing HR at Tus/Ter-arrested replication forks are distinct from those governing HR at an enzyme-induced chromosomal double strand break (DSB). We propose that increased spatial separation of the 2 arrested forks encountering an extended Tus/Ter barrier impairs the coordination of DNA ends generated by the processing of the stalled forks, thereby favoring aberrant LTGC over conservative STGC.

Introduction

The stalling of a replication fork at sites of abnormal DNA structure is a recognized cause of genomic instability.^1-4 Indeed, a number of human diseases, ranging from developmental disorders to cancer predisposition syndromes, have been associated with defective processing of stalled replication forks.^5,6 Stalled replication forks can be processed by a variety of different mechanisms to generate double strand breaks (DSBs) local to the site of fork arrest. These DSBs, in turn, may be repaired by sister chromatid recombination (SCR)—a potentially error-free form of homologous recombination (HR).⁷ The hereditary breast/ovarian cancer predisposition genes, BRCA1 and BRCA2, in concert with other Fanconi anemia genes and other HR genes, have been implicated in regulating HR/SCR in response to replication fork stalling, underscoring the importance of this process for cancer predisposition.^8-10

We recently adapted the Escherichia coli Tus/Ter replication fork arrest complex ^11-14 to provoke site-specific replication fork stalling and chromosomal HR in mammalian cells.¹⁵ In E. coli, binding of the monomeric Tus protein to the asymmetrical ∼23 bp Ter site provokes polar replication fork arrest, depending on whether the replicative helicase DnaB arrives at the non-permissive or permissive end of Ter. We found that an array of 6 Ter sites (6xTer), when bound by wild type (wt)Tus, provokes site-specific replication fork arrest within a replicating episomal plasmid, driven by the Epstein-Barr Nuclear Antigen-1 (EBNA1). In this system, the replication origin is specified by binding of EBNA1 to the OriP sequence, but replication itself is performed by the mammalian host replication machinery.¹⁶ The Tus/6xTer replication fork barrier was ∼70% effective in blocking the mammalian replisome in this setting.¹⁵ Unexpectedly, encounter of the mammalian replisome at the permissive end of the Ter array also causes replisome arrest, indicating that fork arrest at a Tus/6xTer replication fork barrier in mammalian cells is at least in part non-polar. By placing the 6xTer array within a reporter of mammalian HR, targeted to a specific locus of mouse embryonic stem (ES) cells, we found that transient expression of wild type Tus triggers HR at the 6xTer array, while expression of a mutant form of Tus (H144A) that has a low affinity for Ter failed to induce HR in the same cells.^14,15

Several lines of evidence indicate that Tus/Ter-induced HR in mammalian cells entails the processing of 2 converging forks arrested at the Tus/Ter fork block. Tus/Ter-induced HR predominantly generates “short tract” gene conversions (STGC)—a conservative HR outcome. Crucially, Southern blot analysis of Tus/Ter-induced STGC products revealed that STGC termination entailed interaction with a second homologous DNA end—presumably derived from the second arrested fork.¹⁵ An alternative, aberrant HR outcome termed “long tract” gene conversion (LTGC) entails the copying of several kilobases from the donor sister chromatid and is considered an error-prone outcome.^17-20 One cause of LTGC can be a failure of termination of conventional HR, when the second (non-invading) DNA end is either missing—as might occur in a “one-ended” break—or imperfectly coordinated with the invading strand.²⁰

Analysis of Tus-Ter interactions in vitro showed that partial strand separation at the non-permissive end of Ter—such as might occur when the replicative helicase stalls within Ter—allows a cytosine at position 6 (C6) to flip into a pocket of Tus, thereby “locking” Tus onto Ter.¹⁴ This lock mechanism may contribute to the polarity of replication fork arrest observed in E. coli. Importantly, Mulcair et al. described a Tus mutant, F140A, which, although exhibiting higher affinity for duplex Ter than wtTus, cannot accommodate the base-flipped C6 and is defective for the lock mechanism.¹⁴ Thus, the performance of the Tus F140A mutant is a useful probe for testing the mechanism of action of Tus/Ter in the context of eukaryotic replication blocks. We found previously that Tus F140A induces higher levels of HR than wtTus in mammalian cells carrying a 6xTer array HR reporter.¹⁵ This suggests that the C6 base-flipping lock mechanism is not required for Tus/Ter-induced HR in mammalian cells and that the affinity of Tus for duplex Ter is the critical determinant of HR induction in mammalian cells. In contrast, Larsen et al. found that Tus/Ter mediates polar replication fork arrest in Saccharomyces cerevisiae.^21,22 Indeed, Tus F140A is defective for replication fork arrest at a Ter array in S. cerevisiae. Thus, current evidence suggests that Tus/Ter mediates replication fork arrest through distinct mechanisms in yeast and mammalian cells. The reasons for this species-specific difference are unclear at present.

BRCA1, BRCA2 and Rad51 have known functions at stalled replication forks additional to their roles in HR.^8,9 Together with the central Fanconi anemia protein FANCD2, BRCA1, BRCA2 and Rad51 protect hydroxyurea-stalled replication forks from processing by the MRE11 nuclease.^23-25 BRCA1 also promotes the processing of replication forks stalled by UV irradiation.²⁶ We found that BRCA1, BRCA2 and Rad51 are required for efficient STGC at Tus/Ter stalled replication forks, whereas loss of BRCA1 or BRCA2 led to a paradoxical increase in aberrant “long tract” gene conversions (LTGCs) at Tus/Ter stalled forks.¹⁵ LTGC is an error-prone, non-crossover outcome of HR, entailing copying of several kilobases from the donor sister chromatid prior to HR termination, thereby generating a tandem duplication at the site of repair.^17,19,20 LTGC shares several properties with break-induced replication (BIR), a known mediator of genomic instability in yeast.^27-31 Indeed in BRCA1 mutant cells, LTGC is the most abundant HR product observed at Tus/Ter-stalled forks. Thus, determining the mechanisms underlying Tus/Ter-induced LTGC is crucial to understanding how the BRCA genes, and the HR machinery in general, suppress genomic instability at stalled replication forks. One such mechanism could entail the stabilization of an arrested replication fork until the arrival of the converging second fork.³² To determine whether increased spatial separation of the 2 forks arrested at Tus/Ter affects HR, we studied the impact of extended Ter arrays on the induction of HR at Tus/Ter-stalled forks, using defined reporters targeted, in parallel, to the ROSA26 locus of mouse embryonic stem (ES) cells. Our results suggest that coordination between the 2 arrested converging forks influences the quality of HR at the Tus/Ter replication fork barrier.

Results

The HR reporter system used here enables simultaneous flow cytometric quantitation of both STGC and LTGC within a chromosomally integrated HR reporter, in response to either Tus/Ter-induced fork arrest or a DSB induced by the rare-cutting homing endonuclease I-SceI.^15,20 Briefly, 2 mutant copies of the cDNA encoding enhanced green fluorescent protein (“GFP”) are placed in tandem within the HR reporter. One GFP copy is full length but is interrupted by a Ter array followed by an I-SceI target site (Fig. 1). The neighboring GFP copy is truncated at the 5′ end but is otherwise wild type in sequence. Either Tus/Ter-induced replication fork stalling or I-SceI-induced DNA breakage stimulates recombination between the 2 GFP copies, generating wild type (wt)GFP by gene conversion and converting the uncolored cell to GFP⁺. LTGC produces a tandem duplication within the reporter, generating 3 copies of GFP, the central one of which is converted to wtGFP. The reporter cassette contains 2 artificial exons (A and B, Fig. 1) of the gene encoding red fluorescent protein 1.3 (“RFP”), placed in an inverted orientation. In the context of STGC, the exons remain misplaced with respect to one another and the cell remains RFP⁻. In contrast, following LTGC, the RFP cassette undergoes tandem duplication, allowing expression of wtRFP by splicing (Fig. 1). Thus, STGC products are GFP⁺RFP⁻ while LTGC products are GFP⁺RFP⁺. Note that crossing over can also triplicate the GFP copies on the repaired sister chromatid and produce GFP⁺RFP⁺ products. However, crossing over is suppressed in somatic HR and, where analyzed, both I-SceI-induced and Tus/Ter-induced “long tract”-type HR products have been found to be the product of LTGC (i.e., extensive copying from the donor sister chromatid), rather than crossing over.^15,17,19,20

Figure 1.

Homologous recombination products induced at a mammalian chromosomal Tus/Ter replication fork barrier. Model shows a hypothetical double strand break (DSB) intermediate during Tus/Ter-induced HR. Gray boxes: mutant GFP copies. Green box: wtGFP. Ovals A and B: artificial RFP exons. Additional elements including promoter, introns and polyadenylation signal sequences not shown. Red triangle: 6xTer array adjacent to an I-SceI target site (vertical blue line). Processing of the bidirectionally stalled fork or I-SceI endonuclease-induced DSB triggers sister chromatid recombination (SCR). Green arrows indicate extent of gene conversion with the neighboring sister chromatid. SCR products: STGC, short tract gene conversion; LTGC, long tract gene conversion. LTGC is distinguished from STGC repair events by expression of wtRFP via RNA splicing.

We found previously that increasing the number of Ter sites between 1 and 6 within the HR reporter produces progressively higher levels of Tus/Ter-induced HR, the largest array studied previously containing an array of 6 Ter elements.¹⁵ To determine the impact on Tus/Ter-induced HR of increasing the Ter array size beyond 6, we generated reporters containing arrays of 9, 15 or 21 Ter repeats, each with an adjacent I-SceI site (Fig. 2A). We targeted each reporter, in parallel, to the ROSA26 locus of mouse ES cells carrying one conditional allele of BRCA1 and one mutant BRCA1 allele that inactivates the tandem BRCT repeat of the gene product to generate Brca1^fl/BRCT Ter/HR cells.³³ (These cells behave as if wild type with regard to BRCA1 ^15,20) We used Southern blotting and PCR to identify clones of each type containing only one intact, ROSA26-targeted HR reporter (Fig. S1 and Materials and Methods). We analyzed Tus/Ter-induced HR and, in parallel, I-SceI-induced HR in 3 clones of each type, comparing them to a previously characterized clone that contains a 6xTer array HR reporter ¹⁵ (see Materials and Methods). We found that increasing the Ter array size beyond 6 had no significant impact on overall Tus/Ter-induced HR (total GFP⁺) or on Tus/Ter-induced STGC (GFP⁺RFP⁻) (Fig. 2B and Fig. S2). In contrast, with increasing size of the Ter array, we noted a significant trend of increased LTGC (GFP⁺RFP⁺) and increased probability of LTGC (ratio of LTGC:Total HR) in reporters containing 15x and 21xTer arrays. Thus, increasing the size of the Ter array beyond 9 progressively biases Tus/Ter-induced HR in favor of LTGC.

Figure 2.

Impact of extended Ter arrays on Tus/Ter-induced and I-SceI-induced HR. A. Structure of the 6x, 9x, 15x and 21xTer/HR reporter Ter-GFP copies. 5′ and 3′ GFP sequences are 387 bp and 328 bp respectively. Length of the Ter array is as shown. Additional elements including restriction enzyme cloning sites within the Ter arrays are not shown. B. Frequencies of Tus/Ter-induced and I-SceI-induced HR in Brca1^fl/BRCT Ter/HR reporter mouse ES cells containing ROSA26-targeted HR reporters with 6x, 9x, 15x or 21xTer arrays. HR frequencies were measured as described in Materials and Methods, following transient transfection of Tus or I-SceI expression vectors. Columns shown for the 9x, 15x and 21xTer/HR reporters represent the aggregate mean frequency observed for 3 independently derived clones for each reporter cell line. Data for the 6xTer/HR reporter was generated using a single clone. For the 6x, 9x, 15x and 21xTer/HR reporters, each column represents the mean of triplicate samples from 8, 5, 3, and 4 independent experiments respectively. Error bars represent the standard error of the mean. Tus/Ter-induced HR, one-way ANOVA: Total: P = 0.7771; STGC: P = 0.2941; LTGC: P < 0.0001; LTGC/total HR: P < 0.0001; I-SceI-induced HR, one-way ANOVA: Total: P = 0.0006; STGC, P = 0.0005; LTGC, P = 0.1920; LTGC/total HR: P = 0.0619. Tus-induced HR: t-test: 6xTer vs. 9xTer: Total: P = 0.3726; STGC: P = 0.3630; LTGC: P = 0.8772; LTGC/total HR: P = 0.3789; 6xTer vs. 15xTer: Total: P = 0.1725; STGC: P = 0.1311; LTGC: P = 0.0073; LTGC/total HR: P = 0.0007; 6xTer vs. 21xTer: Total: P = 0.3035; STGC: P = 0.1586; LTGC: P< 0.0001; LTGC/total HR: P< 0.0001; 9xTer vs. 15xTer: Total: P = 0.5286; STGC: P = 0.3794; LTGC: P = 0.0182; LTGC/total HR: P = 0.0005; 9xTer vs. 21xTer: Total: P = 0.8927; STGC: P = 0.4764; LTGC: P< 0.0001; LTGC/total HR: P< 0.0001; 15xTer vs. 21xTer: Total: P = 0.5737; STGC: P = 0.8627; LTGC: P = 0.0028; LTGC/total HR: P = 0.0013. I-SceI-induced HR: t-test: 6xTer vs. 9xTer: Total: P = 0.0163; STGC: P = 0.0147; LTGC: P = 0.9233; LTGC/total HR: P = 0.0048; 6xTer vs. 15xTer: Total: P = 0.0052; STGC: P = 0.0049; LTGC: P = 0.5716; LTGC/total HR: P = 0.0178; 6xTer vs. 21xTer: Total: P = 0.0004; STGC: P = 0.0004; LTGC: P = 0.0768; LTGC/total HR: P = 0.0068; 9xTer vs. 15xTer: Total: P = 0.6467; STGC: P = 0.6565; LTGC: P = 0.5066; LTGC/total HR: P = 0.3153; 9xTer vs. 21xTer: Total: P = 0.0526; STGC: P = 0.0545; LTGC: P = 0.0576; LTGC/total HR: P = 0.4704; 15xTer vs. 21xTer: Total: P = 0.0564; STGC: P = 0.0557; LTGC: P = 0.1746; LTGC/total HR: P = 0.6884.

We analyzed I-SceI-induced HR, in parallel, in the same experiments. In contrast to Tus/Ter-induced HR, I-SceI-induced HR revealed a progressive reduction in overall HR (Total GFP⁺) and STGC (GFP⁺RFP⁻) with increasing Ter array size (Fig. 2B and Fig. S2). There was a trend of reduced LTGC (GFP⁺RFP⁺) with increasing Ter array size, which did not reach statistical significance by ANOVA or pair-wise t-test. The ratio LTGC: Total HR was affected by Ter array size greater than 6, however, pair-wise clone comparisons involving clones with longer than 6x Ter arrays showed no statistically significant differences. The LTGC: Total HR ratio bias observed for I-SceI induced HR is due to the reduction in STGC frequencies in reporters with Ter array lengths greater than 6 while for Tus-induced HR, the LTGC: Total HR skewing is directly due to increased incidence of error-prone LTGC repair. Effects of Ter array length on repair events are not due to inherent properties of the Ter array on HR repair itself. Increasing Ter array length did not result in significant increase in background levels of HR in clones transfected with empty vector control plasmid (Fig. S3). Thus, accumulation of background HR events cannot account for differences in the performance of HR reporters carrying different size Ter arrays. In summary, increasing the size of the Ter array has distinct consequences for Tus/Ter-induced HR and I-SceI-induced HR.

To investigate this phenomenon further, we compared HR induced by wtTus and by the lock-defective Tus mutant F140A, in clones containing either a 6xTer HR reporter or a 21xTer HR reporter. Consistent with previous findings, Tus F140A induced higher levels of overall HR and STGC than wtTus, which was statistically significant in the 6xTer HR reporter (Fig. 3A).¹⁵ The abundance of transiently expressed wtTus and TusF140A polypeptides were similar (Fig. 3B). STGC and LTGC were proportionately increased in F140A-expressing cells and consequently the ratio LTGC:Total HR within an individual HR reporter clone was similar for Tus F140A and wtTus-expressing samples. Parallel analysis of I-SceI-induced HR in these experiments again revealed reduced frequencies of overall HR and STGC in the 21xTer reporter clone, in comparison to the 6xTer reporter clone (Fig. 3A). However, the LTGC:Total HR ratio was unchanged. These results show that Tus F140A amplifies Tus/Ter-induced HR products of all types, most markedly in 6xTer HR reporter cells. However, in the experiments reported here, Tus F140A does not alter the quality of HR induced by a Tus/Ter block in comparison to wtTus.

Figure 3.

Effect of Tus F140A mutant on Tus/Ter-induced HR. A. Frequencies of HR induced by wtTus, C6 lock defective Tus F140A, or I-SceI in Brca1^fl/BRCT Ter/HR reporter cells containing 6x, 9x, 15x or 21xTer arrays. Each column represents the mean of duplicate samples from 5 independent experiments. Error bars represent the standard error of the mean. Tus-induced HR: t-test: wtTus 6xTer vs. Tus-F140A 6xTer: Total: P = 0.0188; STGC: P = 0.0228; LTGC: P = 0.3333; LTGC/total HR: P = 0.6667; wtTus 21xTer vs. Tus-F140A 21xTer: Total: P = 0.1588; STGC: P = 0.2105; LTGC: P = 0.0119; LTGC/total HR: P = 0.4805. wtTus 6xTer vs. wtTus 21xTer: Total: P = 0.1487; STGC: P = 0.1876; LTGC: P < 0.0001; LTGC/total HR: P = 0.1998; Tus-F140A 6xTer vs. Tus-F140A 21xTer: Total: P = 0.0969; STGC: P = 0.1401; LTGC: P = 0.0034; LTGC/total HR: P = 0.0687; wtTus 6xTer vs. Tus-F140A 21xTer: Total: P = 0.0313; STGC: P = 0.0448; LTGC: P = 0.0021; LTGC/total HR: P = 0.0664; Tus-F140A 6xTer vs. wtTus 21xTer: Total: P = 0.8016; STGC: P = 0.8929; LTGC: P = 0.2006; LTGC/total HR: P = 0.1615; I-SceI-induced HR: t-test 6xTer vs. 21xTer: Total: P = 0.0207; STGC: P = 0.0215; LTGC: P = 0.3604; LTGC/total HR: P = 0.8693. B. Abundance of Myc-tagged Tus or I-SceI proteins in transfected Brca1^fl/BRCT 6xTer and 21xTer/HR reporter cells: EV: empty vector. Additional lanes as indicated.

Discussion

The Tus/Ter system provides a powerful new tool for provoking replication fork arrest at a defined chromosomal locus in eukaryotic cells and for analyzing mechanisms of repair at stalled mammalian replication forks. By studying Ter arrays of different lengths within a set of otherwise identical HR reporters, each targeted in parallel clones to the ROSA26 locus, we show here that increasing the size of the Ter array beyond 6 does not affect the overall abundance of Tus/Ter-induced HR products; however, with increasing size of the Ter array, we noted an increased absolute frequency of LTGC, specifically in Tus/Ter-induced HR. Use of the higher affinity Tus mutant F140A significantly stimulated HR in 6xTer HR reporter cells and, to a lesser extent, in 21xTer HR reporter cells and reproduced the bias in favor of LTGC noted in 21xTer HR reporter cells with use of wtTus.

We interpret these results in terms of the distribution of HR-initiating lesions within the Tus/Ter array. Although the mechanism of Tus/Ter-induced HR in mammalian cells is not yet defined at a molecular level, one model proposes that bidirectionally stalled forks at a Tus/Ter block are processed to DSBs at the site of stalling (Fig. 4A and B)—a mechanism analogous to the initial stages of repair of an inter-strand DNA crosslink (ICL) by components of the Fanconi anemia (FA) pathway.^4,34,35 An alternative hypothesis, in which a DSB intermediate is not invoked, proposes that template switches mediate HR at the Tus/Ter block, as has been suggested from analysis of a site-specific replication fork barrier in Schizosaccharomyces pombe.³⁶ For the purpose of discussion, we assume a DSB model of Tus/Ter-induced HR; however, the mechanisms discussed here could also apply to a template switch model. We propose that the LTGC bias noted in HR reporters containing an extended array of Ter sites is a consequence of increased spatial separation of the sites of stalling of the converging replication forks (Fig. 4A and B). Tus/Ter-mediated replication fork arrest may occur in close proximity to the outer edges of the Ter array, and DNA breaks might mirror the distribution of fork stalling events. The two converging forks that encounter Tus bound to the 6xTer array (253 bp in length) will generate breaks in fairly close spatial proximity to one another. In contrast, the 21xTer array (843 bp in length) might generate breaks that are more widely separated in space (Fig. 4A and B). This, in turn, could result in discoordinated processing of the broken forks. Indeed, in frog egg extracts, optimal FA pathway-mediated processing of replication forks stalled at an ICL occurs only once both converging forks have stalled at the ICL.³² Similarly, discoordinated processing of the 2 ends of a chromosomal DSB, such as can occur in DNA end resection mutants in S. cerevisiae,³⁷ may in part explain the bias toward LTGC in cells lacking BRCA1 or CtIP.^15,20 The extreme case of DNA end discoordination is the “one-ended” break, where the second end of the break is entirely absent. One-ended breaks are recognized triggers of LTGC/BIR in yeast and in mammalian cells.^{20,28,30,31,38}

Figure 4.

Models of the impact of an extended Ter array on Tus/Ter-induced HR and I-SceI-induced HR. We propose that an extended Ter array results in spatial separation of the arrest sites of the 2 converging replication forks. DSB-mediated model of Tus/Ter-induced HR is shown. Green elements: GFP coding sequence. Red triangles: Ter sites. A. Bidirectional fork stalling at Tus bound to 6xTer. DNA ends generated as a consequence of stalled fork processing are positioned in close proximity, favoring conservative STGC. B. Bidirectional fork stalling at Tus bound to 21xTer. DNA ends generated as consequence of stalled fork processing are relatively poorly coordinated, producing a bias in favor of LTGC. C. Impact of extended Ter array on DNA end resection requirements for I-SceI-induced HR. As the number of Ter sites within the array increases, the extent of DNA end resection needed to generate GFP⁺ recombinants increases. Green elements: GFP coding sequence. Red elements: Ter arrays. Approximate extent of resection needed to expose ssDNA GFP sequence is indicated for arrays of zero, 6x or 21x Ter. D. Impact of extended Ter array on 3′flap formation during I-SceI-induced strand invasion. Size of 3′flap is indicated for arrays of zero, 6x or 21x Ter.

In response to an I-SceI-induced DSB, arrays containing 9 or more Ter sites resulted in progressive reductions in both STGC and LTGC,. With a larger Ter array, the distance between the I-SceI site and GFP sequences to the left of the Ter array increases (Fig. 4C). In a 6xTer HR reporter, one I-SceI-induced DNA end must undergo resection of > 250 bp in order to expose single stranded (ss)DNA GFP sequence that would then be competent for I-SceI-induced HR. The extent of resection needed to support HR increases with increasing Ter array length, the 21xTer reporter requiring >850 bp of resection to reveal recombination competent ssDNA GFP sequences. If the resection processes were limiting, this could explain the reduction in I-SceI-induced HR efficiency with increasing Ter array size (Fig. 4C). By the same reasoning, following I-SceI-induced strand exchange, the size of the 3′ flap of non-homologous (i.e., non-GFP) sequence on the invading strand would increase with increasing length of the Ter array, being 9 bp for the native I-SceI HR reporter, >260 bp for the 6xTer HR reporter and > 850 bp for the 21xTer HR reporter (Fig. 4D). It is not yet clear which of these processes (DNA end resection or 3′ flap processing) is the limiting step responsible for the observed inverse relationship between Ter array size and I-SceI-induced HR efficiency.

The performance of the Tus F140A mutant highlights differences between the mechanisms of action of Tus/Ter as a replication fork barrier in mammals, yeast, and E. coli. Tus F140A fails to support an efficient “lock” mechanism when the Ter site is partially single stranded at the non-permissive end, because it cannot accommodate the base-flipped Ter-C6 residue.¹⁴ However, Tus F140A exhibits a higher affinity than wtTus for duplex Ter ¹⁴ Larsen et al. found that wtTus mediates polar replication fork arrest at a Ter array in S. cerevisiae.²² The same authors found that Tus F140A is unable to generate an efficient replication fork barrier when bound to a chromosomal Ter array in S. cerevisiae, even when the majority of replication forks arrive at the non-permissive face of Tus/Ter.²¹ In contrast, Tus/Ter-mediated replication fork arrest and ensuing HR in mammalian cells is predominantly non-polar and does not require the C6 lock mechanism.¹⁵ Indeed, Tus F140A induces higher levels of HR than wtTus in 6xTer HR reporter cells, suggesting that the critical determinant of Tus-induced HR proficiency in mammalian cells is its affinity for duplex Ter.¹⁵ These considerations suggest that Tus/Ter arrests the mammalian replisome by virtue of a simple affinity/avidity mechanism, involving Tus bound to duplex Ter sites (i.e., Ter sites in which no strand separation has occurred at the non-permissive end and no C6 lock has been engaged).

The above-noted differences in the mechanism of action of Tus/Ter likely reflect species-specific differences in the way the replicative helicase interacts with Tus/Ter complexes on the chromosome.³⁹ We originally suggested that the polarity of the replicative helicase might determine the accessibility of the C6 lock as an explanation for the lack of polarity observed in mammalian cells. Unlike the prokaryotic DnaB replicative helicase, which translocates on the lagging strand,⁴⁰ the eukaryotic CMG helicase complex translocates on the leading strand.⁴¹ Since the critical Ter C6 residue is on the leading strand of the fork approaching the non-permissive end of Ter, this residue could be masked in the barrel of the MCM helicase, thereby preventing C6 lock formation. However, this hypothesis fails to explain the behavior of Tus/Ter in S. cerevisiae, where the C6 lock mechanism is a significant contributor to replisome arrest,²¹ despite the fact that the replicative helicase translocates along the leading strand. An alternative hypothesis is that, in comparison to its yeast counterpart, the mammalian replisome is more prone to arrest at sites of abnormal DNA structure. This might be the case if the force exerted by the mammalian replicative helicase were simply less than the force exerted by the E. coli or S. cerevisiae replicative helicases. In the context of encounters between the mammalian replisome and Tus/6xTer complexes, Tus-Ter binding of intermediate affinity (i.e., lacking the high affinity C6 lock mechanism) might be sufficient to arrest the mammalian replisome. This hypothesis would explain both the dispensability of the C6 lock mechanism and the bidirectionality of the Tus/Ter fork barrier in mammalian cells. An additional possible explanation, non-exclusive with the above-noted considerations, is that replication fork stalling triggers nucleolytic attack of the stalled fork more avidly in mammalian cells than in E. coli or S. cerevisiae. In this regard, the evolution of a multi-subunit Fanconi anemia (FA) pathway distinguishes vertebrates from yeasts, in which FA homologs are limited to the DNA junction-specific helicases/translocases and Hef/FANCM homologs, Mph1 (S. cerevisiae) and Fml1 (S. pombe).^42-44 A functional FA pathway is required for efficient incision of bidirectionally arrested replication forks at an interstrand DNA crosslink in frog egg extracts.^32,45 It will be informative to study the impact of FA pathway dysfunction on Tus/Ter-induced HR in mammalian cells.

Materials and methods

Mouse cell lines and cell culture

Mouse embryonic stem (ES) cells were maintained in ES medium on either MEF feeders or gelatinized plates as described previously.^15,20,46,47 10 μg of each Ter/HR reporter ROSA26 targeting plasmid was linearized by Kpn I digest and introduced by electroporation to 2 × 10⁷ cells. Cells were seeded on plates prepared previously with puromycin-resistant feeders. Plates were supplemented with puromycin (4 μg/ml) 24 hours later and colonies were picked 5-10 d later. ROSA26 targeted lines were screened for by PCR and reporter integration and overall structure verified by Southern blotting. ROSA26 genotyping primers: ROSA26-sense- (cat caa gga aac cct gga cta ctg); terBx6 HR reporter antisense- (cct cgg cta ggt agg gga tc).

Recombination assays

1.6 × 10⁵ cells were transfected in suspension with 0.5 μg pcDNA3β-myc NLS-I-SceI,¹⁸ pcDNA3β-myc NLS-Tus, pcDNA3β-myc NLS-TusF140A,¹⁵ or empty vector using Lipofectamine 2000 (Invitrogen). GFP⁺ and GFP⁺RFP⁺ frequencies were scored 72 hours after transfection by flow cytometry using a Becton Dickinson 5 Laser LSRII in duplicate. Values presented are corrected for background events and for transfection efficiency. Transfection efficiency was assessed by parallel transfection with 0.05 μg wild type GFP expression vector and 0.45μg control vector. 6 × 10⁵ total events were scored for each sample. Data presented represents the mean and error bars represent the standard error of the mean (SEM) of at least 3 independent experiments.

Statistical methods

Each figure legend specifies the sample number in terms of number of replicates within each individual experiment and number of independent experiments that were performed to generate the data presented. For statistical analysis of HR repair frequencies, the arithmetic mean of samples for each independent experiment was calculated (i.e., experiments performed on different days) and these single data points for each experiment were used to determine the mean and standard deviation between experiments. The standard error of the mean (SEM) was calculated as standard deviation/√n, in which n indicates the number of experiments (not number of replicates). Differences between sample pairs were analyzed by Student's 2-tailed unpaired t-test, assuming unequal variance, statistical trends across 3 or more samples were analyzed by one-way ANOVA, both analysis were performed using GraphPad Prism v6.0d software. P-values are shown in each figure legend.

Western blotting

Protein was extracted using RIPA buffer (50 mM Tris-HCl, pH 8.0, 250 mM NaCl, 0.1% sodium dodecyl sulfate, 1% NP-40 containing protease and phosphatase inhibitors PMSF and Roche complete protease inhibitor tablet) and resolved by 4–12 % bis-Tris SDS-PAGE (Invitrogen). Protein expression was analyzed by immunoblotting using the following antibodies; β-tubulin (Abcam ab6046, 1:4,000), Myc (Abcam ab9106, 1:10,000).

Southern blotting

Southern blotting of genomic DNA was performed using GFP or ROSA26 probes using methods described previously.^18,47 For all experiments, mouse ES cell clones harboring an intact copy of the reporter integrated at the ROSA26 locus on chromosome 6 were used. Genomic DNA was extracted from ES cells grown to confluency on gelatinized 6-well plates (∼5–10 × 10⁶ cells) using a Puregene DNA Isolation Kit (Gentra Systems).

Disclosure of potential conflicts of interest

No potential conflicts of interest were disclosed.

Acknowledgments

We thank Drs. Ian Hickson, Nicolai Larsen, Hocine Mankouri and Simon Powell, as well as members of the Scully lab for helpful discussions.

Funding

This work was supported by grants R01CA095175 and R01GM073894 to RS and by an American Cancer Society Postdoctoral Fellowship to NAW.