Distinct Activities of Exonuclease 1 and Flap Endonuclease 1 at Telomeric G4 DNA

Background Exonuclease 1 (EXO1) and Flap endonuclease 1 (FEN1) are members of the RAD2 family of structure-specific nucleases. Genetic analysis has identified roles for EXO1 and FEN1 in replication, recombination, DNA repair and maintenance of telomeres. Telomeres are composed of G-rich repeats that readily form G4 DNA. We recently showed that human EXO1 and FEN1 exhibit distinct activities on G4 DNA substrates representative of intermediates in immunoglobulin class switch recombination. Methodology/Principal Findings We have now compared activities of these enzymes on telomeric substrates bearing G4 DNA, identifying non-overlapping functions that provide mechanistic insight into the distinct telomeric phenotypes caused by their deficiencies. We show that hFEN1 but not hEXO1 cleaves substrates bearing telomeric G4 DNA 5′-flaps, consistent with the requirement for FEN1 in telomeric lagging strand replication. Both hEXO1 and hFEN1 are active on substrates bearing telomeric G4 DNA tails, resembling uncapped telomeres. Notably, hEXO1 but not hFEN1 is active on transcribed telomeric G-loops. Conclusion/Significance Our results suggest that EXO1 may act at transcription-induced telomeric structures to promote telomere recombination while FEN1 has a dominant role in lagging strand replication at telomeres. Both enzymes can create ssDNA at uncapped telomere ends thereby contributing to recombination.

Like the telomeres, the immunoglobulin (Ig) heavy chain switch regions are composed of G-rich repeats with considerable potential to form G4 DNA. The Ig switch regions are targets for class switch recombination, a process of DNA deletion that enables a B cell to juxtapose a new constant region to the expressed variable region [39,40]. EXO1 is necessary for efficient class switch recombination [41], and we recently showed that human EXO1 (hEXO1) and human FEN1 (hFEN1) exhibit distinct activities on substrates representing switch recombination intermediates [42]. Ig class switch recombination is induced by transcription; and hEXO1 but not hFEN1 excised characteristic G-loop structures formed by transcribed switch region substrates [43]. Those experiments identified a property of EXO1 that could contribute to its function not only at the Ig switch regions, but also at other transcribed genomic domains with potential to form G4 DNA.
Those results prompted us to examine the biochemical activities of EXO1 and FEN1 that might contribute to telomere maintenance or instability, using substrates that recapitulate structures predicted to form upon telomere replication, uncapping, recombination or transcription. We show that hFEN1 but not hEXO1 cleaves 59 flaps containing telomeric G4 DNA, or other G4 DNA. This is consistent with the importance of FEN1 and not EXO1 in telomere lagging strand replication, but provides a surprising exception to the documented inactivity of FEN1 at other structured 59 flaps in vitro [44,45]. We demonstrate that hEXO1 and hFEN1 both exhibit robust exonucleolytic activity on both strands of substrates with Grich 39 telomeric tails, enabling either enzyme to expose a singlestranded region for recombination, or to participate in removal or formation of a G-rich tail. We show that hEXO1 but not hFEN1 excises G-loops formed at transcribed telomeric repeats, suggesting a mechanism by which telomeric transcription could promote recombination to maintain telomere length in the absence of telomerase. These results define new activities for these enzymes, and provide mechanistic understanding of how deficiencies in these enzymes cause distinct telomeric phenotypes.

hFEN1, but Not hEXO1, Cleaves Telomeric 59 G4 Flaps, Possible Intermediates in Lagging Strand Replication
The canonical activity of FEN1 is endonucleolytic excision of a 59 flap from a duplex substrate, which represents the intermediate in processing Okazaki fragments during lagging strand replication [5]. During telomeric lagging strand replication, a 59 flap may become structured as G4 DNA. A variety of 59 flap structures have been shown to inhibit FEN1 activity in vitro [45]. This raised the question of whether the presence of G4 DNA at a 59 flap would affect cleavage by EXO1 or FEN1. To address this, we assayed cleavage of two duplex substrates bearing 59 flaps, one carrying telomeric G4 DNA, the other unstructured (Fig. 1A). hFEN1 displayed robust flap endonuclease activity on the substrate bearing a 59 telomeric G4 DNA flap (70% cleavage at 4.5 nM enzyme), while hEXO1 was inactive on this substrate ( Fig. 1B, C, left). Both enzymes were active on the substrate bearing an unstructured 59 flap, although hFEN1 was somewhat more active than hEXO1. hFEN1 activity at a 59 G4 DNA flap was comparable to its activity on an unstructured 59 flap (Fig. 1B,

G4 Telomeric Tails Inhibit Flap Cleavage by hEXO1 but Not hFEN1
The novel activity of FEN1 at a 59 G4 DNA flap prompted us to test activities of hEXO1 or hFEN1 at unstructured 59 flaps on substrates bearing either a 39 telomeric G4 DNA tail or an unstructured tail ( Fig. 2A). hEXO1 exhibited essentially no flap endonucleolytic activity and very modest exonucleolytic activity on the G4 DNA substrate (Fig. 2B, C, left). This was somewhat surprising, because hEXO1 typically exhibits both endonuclease and exonuclease activities at 59 flaps [42,46]. Thus, a 39 tail may inhibit activity of hEXO1 at a 59 flap. In contrast, hFEN1 exhibited very robust flap endonuclease activity on the substrate with a G4 DNA tail (more than 80% of this substrate cleaved at 4.5 nM hFEN1; Fig. 2B, C, left). hFEN1 was somewhat less active on the substrate with an unstructured 39 tail (60% cleavage at 4.5 nM hFEN1; Fig. 2B, C, right), but was comparably active on this substrate as on a substrate carrying no tail (Figs. 1C and 2C). These results show that a 39 tail can inhibit hEXO1, but not hFEN1; and suggest that a 39 G4 DNA tail may even stimulate hFEN1.  is specific for telomeric sequences, we assayed substrates of similar structures but bearing G4 DNA of identical length but formed from the synthetic, nontelomeric sequence, CCTGGGCTAGG-GATCGGGACCGGG, within the 59-flap or 39 tail (Fig. 3A). hFEN1 was very active on both substrates, and hEXO1 was inactive (Fig. 3B, C). Thus, the effect of G4 structures on the flap endonuclease activities of these enzymes is general, and not specific to telomeric G4 DNA. hEXO1 and hFEN1 Excise from a Nick on a Strand Bearing a 39-G4 DNA Telomeric Tail Both EXO1 and FEN1 can excise from nicks in blunt-ended duplex substrates [42,47], but whether a G4 DNA tail affects excision has not been tested. To do so, we assayed activities of these two enzymes on a nicked DNA duplex bearing a 24 nt 39 (TTAGGG) 4 tail (Fig. 4A), mimicking a structure that may form in vivo at an uncapped telomeric end. The 39 tail spontaneously forms intramolecular G4 DNA in our standard assay conditions (Fig. 1A). Both a duplex bearing a 39-polyA tail and a blunt duplex were included as controls, in order to discriminate between the effects of the G4 structure and a disordered tail. hEXO1 was active on a substrate bearing a G4 tail, but comparably active on a substrate bearing a 39-polyA tail and slightly more active on a control bluntended duplex (Fig. 4B, C). hFEN1 was equally active on all three substrates (Fig. 4D, E). Thus, both enzymes can initiate at a nick to degrade the G-rich strand of DNA carrying a telomeric overhang. This activity could expose the C-rich strand for recombination, but also has the potential to destabilize the telomere by removing the G-rich tail.  We then compared the ability of hEXO1 or hFEN1 to create single-stranded regions on DNA duplex substrates on the strand opposite a telomeric tail. Substrates were internally labeled by 39filling to enable assays of 59-39 exonucleolytic digestion initiating opposite the structured end (Fig. 5A). hEXO1 proved to be considerably more active on the substrate bearing a 39 telomeric tail than on the substrates bearing a polyA tail or a blunt end (Fig. 5B, upper), while hFEN1 displayed robust exonucleolytic activity on the substrate bearing a 39 telomeric tail, but little activity on the other substrates (Fig. 5B, lower). The activities of hEXO1 and hFEN1 on the substrate bearing a 39 telomeric tail were comparable: 50% of the substrate was digested by approximately 3 nM enzyme. Thus, a G4 DNA telomeric tail, but not a polyA tail or blunt end, can stimulate 59-39 excision by either hEXO1 or hFEN1 on the opposite strand.
We also compared digestion of these same substrates initiated by hEXO1 and hFEN1 at an internal nick (Fig. 5C). hEXO1 was most active on the blunt duplex substrate, but less active on the substrates bearing polyA tail or telomeric tail (Fig. 5D, upper). A telomeric tail appeared to inhibit excision from a nick on the opposite strand on duplex DNA, even though separated from the nick by 26 base pairs. hFEN1 was less active than hEXO1 on these substrates, but was approximately twice as active on the substrate bearing the telomeric tail than on the blunt-ended or polyA-tailed substrate (Fig. 5D, lower).
Transcribed Telomere Repeats Are Excised by hEXO1 but Not hFEN1 Telomeric repeats are transcribed in vivo, and the G-rich strand is the non-template strand [48,49]. Transcription of telomeres and other G-rich sequences produces characteristic structures, called Gloops, which contain a stable RNA/DNA hybrid on the C-rich template strand, and G4 DNA interspersed with single-stranded regions on the G-rich nontemplate strand [43,50]. We tested activity of hEXO1 and hFEN1 on G-loops formed by transcribed telomeric sequences formed in the pTELN plasmid, which bears 800 bp of TTAGGG human telomeric repeat downstream of a T7 promoter (Fig. 6A). pTELN was transcribed, free transcript removed by RNaseA treatment, and DNA nicked. We then compared digestion by hEXO1 and hFEN1 of substrates that had been transcribed and nicked substrates, transcribed but not nicked, nicked but not transcribed, and the supercoiled template DNA. Digestion was quantitated by monitoring loss of one of the PvuII sites that occurs as a consequence of excision from the nick, and which results in production of a 3.8 kb PvuII fragment. hEXO1 proved to be quite active on the transcribed, nicked telomeric substrates: approximately 80% of the substrate was excised by 2.4 nM hEXO1 (Fig. 6B,  upper; Fig. 6C), comparable to hEXO1 activity at transcribed Ig switch regions [42]. hEXO1 was also somewhat active on the nicked substrate, consistent with its documented activity on nicked DNA in mismatch repair. hEXO1 was not active either on the supercoiled template, or on transcribed, unnicked DNA. In contrast, hFEN1 showed no activity on any of the substrates (Fig. 6B lower; Fig. 6C). These results identify EXO1 as a candidate activity for resection of DNA at transcribed telomeres.

Discussion
We have shown that hEXO1 and hFEN1 exhibit both shared and distinct biochemical activities on telomeric substrates. These activities correlate with the roles of EXO1 and FEN1 in telomere maintenance, as determined by genetic analysis, and thus provide mechanistic insight into the distinct telomeric phenotypes caused by deficiencies in these enzymes. Notably, distinct activities of each enzyme depended upon a G4 structure, supporting the importance of G4 DNA recognition in telomere maintenance.

FEN1 May Process G4 Structures Formed during Lagging Strand Replication
hFEN1 but not hEXO1 displayed robust endonuclease activity on substrates bearing telomeric or nontelomeric 59 G4 DNA flaps, representing intermediates in lagging strand replication. This contrasts with the comparable activities of these enzymes at unstructured 59 flaps [42]. hFEN1 activity on these structures constitutes an exception to the documented inhibition of FEN1 in vitro by other 59-structures, such as ds DNA and hairpins formed by triplet repeats [44,45]. Moreover, while either a G4 DNA tail or unstructured tail inhibited endonucleolytic cleavage by hEXO1, the 59 flap endonuclease activity of hFEN1 was not impaired by proximity of a 39 tail.
G4 DNA structures may create blocks to replication at telomeres [31]. The ability of FEN1 but not EXO1 to cleave a flap containing G4 DNA would then explain both the instability of telomeric lagging strands that occurs in the absence of FEN1, and the inability of Exo1 overexpression to rescue this instability in S. cerevisiae [6][7][8]. In human cells, telomere lagging strand instability is observed in the absence of FEN1 [8], and also in the absence of the RecQ family helicase, WRN [32]. WRN has robust G4 DNA unwinding activity [23], and interacts with and stimulates FEN1 [34,51,52]. Like most RecQ family helicases, WRN contains a high affinity G4 DNA binding domain, the RQC domain [53], which could promote FEN1 recognition of G4 DNA in vivo. Alternatively, FEN1 may itself specifically recognize G4 structures, a possibility supported both by the robust activity of hFEN1 on 59 flaps containing G4 DNA, and by the ability of a G4 39 tail, but not a polyA tail or blunt end, to stimulate hFEN1 exonucleolytic activity. FEN1 and EXO1 Activities at Telomeric G4 DNA Tails Both hEXO1 and hFEN1 were able to excise substrates bearing 39 telomeric tails, which resemble uncapped telomeres. This activity could expose the C-rich strand for recombination, or remove the G-rich tail and thereby destabilize the telomere. Stabilization of telomeres that occurs upon ablation of Exo1 in telomerase-deficient mice [54] may be one manifestation of this activity.
hFEN1 and hEXO1 were also very active at the 59 end of the strand opposite a telomeric tail, where hFEN1 in particular was stimulated by G4 DNA. In vivo, excision from this end could lead to an increase in the length of a 39 telomeric tail, at the expense of duplex DNA. The roles of FEN1 and EXO1 in telomerasedeficient cells have been ascribed to functions in recombination, including exposing single-stranded DNA or resolving structures produced by branch migration. Our results identify another possible function, in exposing G-rich regions for 39 tail formation.

Telomere Transcription May Enable EXO1 to Create Substrates for Recombination
In vivo, telomeres are transcribed from promoters in subtelomeric regions [49,55]. Transcription of G-rich telomeric sequences results in formation of characteristic G-loop structures, which contain a stable RNA/DNA hybrid on the C-rich template strand and G4 DNA interspersed with single-stranded regions on the Grich strand [43]. hEXO1 (but not hFEN1) excised telomeric Gloops, in a reaction that parallels excision of G-loops formed within transcribed Ig switch regions [42]. While the function of telomeric transcription is not yet understood, it has been shown to be more active in cells that lack telomerase and depend upon recombination for telomere maintenance [56]. Our results suggest that one function of telomere transcription may be to promote formation of recombinogenic structures that are substrates for EXO1. Because telomere transcription initiates in subtelomeric regions, promoter-proximal sequences will be enriched among these substrates, conferring the potential to transfer long regions of sequence to the recipient telomere.

G4 DNA as a Target of FEN1 and EXO1
The results documented here show that both FEN1 and EXO1 must be included on the growing list of critical DNA maintenance and repair factors that are active at G4 DNA. This list also includes BLM, WRN, Sgs1 and related RecQ family helicases [21,22,24,32,57,58]; and XPD-family helicases such as FANCJ, RTEL1, DOG-1 and Srs2 [26][27][28][29]. hFEN1 and hEXO1 activities were stimulated by both telomeric and nontelomeric sequences, so both enzymes could act upon G4 DNA that formed outside the telomeric regions. This may explain why G-rich ribosomal DNA repeats [59] and the G-rich minisatellites CEB1 and MS32 [60,61] are unstable in S. cerevisiae rad27 strains which lack the FEN1 homolog. The evidence that robust activity of hFEN1 was evident on structures on which hEXO1 exhibited limited or no activity suggests that EXO1 might not be able to compensate for FEN1 to maintain stability of genomic regions with high potential for G4 DNA formation.
The G4 DNA signature motif, and consequently potential for G4 DNA formation, is unevenly distributed and also selected in the human genome, characterizing not just specialized domains like the telomeres, ribosomal DNA and Ig switch regions, but also promoters and specific functional classes of genes [62][63][64]. The ability of G4 structures to stimulate FEN1 and EXO1 not only identifies these enzymes as key to stability of regions bearing the G4 motif, but also raises the further possibility that the G4 motif may confer special localized properties in replication or recombination.

DNA Substrates
Oligonucleotides were 59 end-labeled with c 32 P-dATP using T4 polynucleotide kinase (NEB) or 39-labeled by filling with Exo 2 Klenow polymerase (NEB); and free nucleotides removed using a G50 spin column (GE). Synthetic oligonucleotide substrates (below) were generated by heating oligonucleotides at equimolar concentrations at 90uC in TE containing 40 mM KCl (which promotes G4 DNA formation), followed by slow cooling and overnight incubation at room temperature. Annealed substrates were resolved on an 8% nondenaturing polyacrylamide gel containing 10 mM KCl, eluted, and stored in TE containing 10 mM KCl. Formation of G4 structures was confirmed by analysis of labeled substrates on native gels containing 10 mM KCl, where more than 90% of the label migrated with retarded mobility, diagnostic of G4 DNA ( Figure S1).
Substrates bearing a 59 telomeric or nontelomeric G4 flap were generated by annealing either 59-labeled 59-(TTAGGG) 4  Plasmid substrate pTELN consists of 800 bp of telomere repeat cloned into the Xba1 and BamH1 sites of pBluescript (KS+), downstream of the T7 promoter [43]. Quikchange mutagenesis (Stratagene) generated an Nb.BbvC1 nickase site 131 bp upstream of the promoter and 101 bp from one of two PvuII sites. PvuII cleavage of intact 3.8 kb pTelN produces two fragments, 2.5 kb and 1.3 kb in length. Exonucleolytic digestion from the 59-nick destroys the promoter-proximal PvuII site, so the fraction of 3.8 kb molecules produced upon PvuII digestion provides a measure of exonucleolytic activity [42].
Enzymes and Enzyme Assays hEXO1 was expressed from a baculovirus vector in sf9 insect cells and purified as described [42], and hFEN1 was obtained commercially (Trevigen). Activity assays were carried out in 20 ml containing indicated amounts of enzyme and 5 nM DNA substrate; in 30 mM HEPES, pH 7.6, 40 mM KCl, 8 mM MgCl 2 , 0.1 mg/ml BSA and 1 mM DTT for hEXO1, in manufacturer's buffer supplemented with 40 mM KCl to maintain stability of G4 structures. Products of digestion of labeled synthetic oligonucleotides were denatured by heating at 95uC for 10 min in 0.5 volume of 95% formamide/20 mM EDTA at pH 8.0, and a 5 ml aliquot resolved by denaturing gel electrophoresis on 8 M urea, 12% or 20% (for exonuclease assays) polyacrylamide gels. Products of digestion of plasmid substrates were resolved by native agarose gel electrophoresis and quantified by indirect end-labeling after transfer to a nylon membrane and hybridization to a labeled probe complementary to the T7 promoter [42]. Gels were scanned with a STORM Phosphorimager (Amersham) and label quantitated with Image Quant software (Amersham). Figure S1 Structure formation by substrates. (A) Substrates diagrammed on left were analyzed by electrophoresis on a native gel containing 10 mM KCl. Substrates were 39 end-labeled either opposite the tail or at the nick; asterisk denotes end-label. Retarded mobility on a native gel containing 10 mM KCl is diagnostic of G4 DNA formation by substrates b and e (arrow). (B) Substrates diagrammed on left (asterisk denotes 59-end-label) were gel purified, and structure formation confirmed by electrophoresis on a native gel containing 10 mM KCl. Substrates bearing an unstructured 59-flap (a) migrated more rapidly than substrates containing telomeric G4 DNA 59 flap (b) or 39 tail (c). These substrates (indicated by arrows) were used for subsequent assays. Found at: doi:10.1371/journal.pone.0008908.s001 (0.12 MB PDF)