TRF1 prevents permissive DNA damage response, recombination and Break Induced Replication at telomeres

Telomeres are a significant challenge to DNA replication and are prone to replication stress and telomere fragility. The shelterin component TRF1 facilitates telomere replication but the molecular mechanism remains uncertain. By interrogating the proteomic composition of telomeres, we show that telomeres lacking TRF1 undergo protein composition reorganisation associated with a DNA damage response and chromatin remodelers. Surprisingly, TRF1 suppresses the accumulation of promyelocytic leukemia (PML) protein, BRCA1 and the SMC5/6 complex at telomeres, which is associated with increased Homologous Recombination (HR) and TERRA transcription. We uncovered a previously unappreciated role for TRF1 in the suppression of telomere recombination, dependent on SMC5 and also POLD3 dependent Break Induced Replication at telomeres. We propose that TRF1 facilitates S-phase telomeric DNA synthesis to prevent illegitimate mitotic DNA recombination and chromatin rearrangement.


Introduction 32
Telomeres are specialised nucleoprotein structures at the ends of chromosomes, composed of 33 repetitive sequences (TTAGGG repeats in mammals) (Moyzis et al., 1988), long non-coding RNA 34 called TERRA and six associated proteins, TRF1, TRF2, POT1a/b, RAP1 and TIN2, that form the 35 shelterin complex (de Lange, 2005). These capping structures have the crucial function of maintaining 36 genome stability by protecting the chromosome end from being recognised as DNA double strand 37 breaks (DSBs) (Palm & de Lange, 2008). They also represent challenging structures for the 38 replication machinery, which is associated to telomere fragile sites (Martinez et  During tumorigenesis, cancer cells can achieve replicative immortality by activation of telomere 47 maintenance mechanisms. The majority of cancer cells reactivate telomerase, while a minority (10-48 15%) uses an alternative mechanism named ALT for alternative lengthening of telomeres (Bryan,49 Englezou, Dalla-Pozza, Dunham, & Reddel, 1997; Kim et al., 1994). Intriguingly, ALT is 50 characterised by the appearance of ALT-associated PML bodies (APBs), specialised sites where a 51 subset of telomeres co-localises with PML protein and several DNA repair and homologous 52 recombination (HR) proteins (Draskovic et al., 2009; G. Wu, Lee, & Chen, 2000;Yeager et al., 1999). 53 ALT telomeres can be maintained by more than one mechanism of recombination. Indeed, in yeast, 54 two different ALT-like pathways have been described: Type I, requires Rad51 to mediate the invasion 55 of a homologous sequence, while Type II is Rad51 independent and rely on Rad52 dependent 56 elongation mechanism, which consists in the annealing of ssDNA regions. Both Type I and II 57 mechanisms require the DNA polymerase Pol32, which initiates DNA synthesis for several kilobases, 58 in a process known as Break Induced Replication (BIR) (Ira & Haber, 2002). Recently, multiple 59 groups have revisited this Rad51 independent DNA synthesis repair pathway at mammalian ALT 1B). Cells were fixed and isolation of telomeres was performed using a probe complementary to 93 TTAGGG repeats or a scrambled probe as a negative control. Finally, telomeric chromatin was 94 isolated from both control cells (wt) and TRF1 deleted cells before mass spectrometry identification 95 ( Figure 1C). We identified a list of 1306 proteins that was subjected to refinement in order to remove 96 unspecific bound proteins or contaminants found with the scrambled probe (see experimental 97 procedure for detailed description). Based on the analysis of label free quantification (LFQ 98 intensities), we found 119 proteins presenting a gain of abundance at TRF1 depleted telomeres 99 (Log2>-2) and 206 factors were displaced from these telomeres (Log2>2), considering that a cut-off 100 for differential expression is set to log2 fold change (TRF1deletion/wt)> |2| and -Log (p-value) >1 101 ( Figure 1D). Amongst these 206 proteins, we found TRF1, as expected due to the knock-out of its 102 gene, but also one component of the CST complex (CTC1), important player in the efficient restart 103 of stalled replication forks at telomeres (Gu et al., 2012) and recruited through POT1b interaction (P. 2015; Potts & Yu, 2007). To validate the specific association of some of these factors with TRF1 118 depleted telomeres in telomerase positive MEFs, we carried out chromatin immunoprecipitation 119 (ChIP) experiments using ChIP-grade specific antibodies followed by telomeric dot-blot. TRF1 120 antibody was used as a negative control for our experiment, while the recruitment of BRCA1, BAZ1b, 121 and some subunits of the nucleosome remodeling and deacetylase (NurD) complex (p66a, MTA1, 122 ChD4, zinc-finger protein ZNF827) was assessed. For all these factors, with the exception of p66a 123 for which no statistical significance was achieved, we observed a specific enrichment at telomeres 124 upon TRF1 deletion ( Figure 2B; Figure S1A-B). In addition, to confirm the presence of PML at 125 replication stress induced telomeres, as suggested by our PICh data (  suppressing recombination events as well as many other phenotypic features related to ALT. Hence, 132 to test this hypothesis, we revisited the incidence of telomeric sister chromatid exchanges (T-SCE) 133 using chromosome orientation FISH (CO-FISH) in TRF1 deficient cells ( Figure 2D). We identified 134 an increase in T-SCE in TRF1 -/-MEFs (2.8%) compared to control cells (0.4%) ( Figure 2D). This 135 result is at odds with previous publications where T-SCE events detected at TRF1 depleted telomeres 136 were not significantly enriched, with only 1% of T-SCEs detected compared to 0.1% in wt cells 137 (Martinez et al., 2009;Sfeir et al., 2009). In fact, this discrepancy might be explained by the difference 138 in timing for the analysis of T-SCEs in TRF1 deficient cells. Both publications report the lack of 139 recombination effect by T-SCEs at 3 or 4 days after TRF1 loss, while we generally carry our 140 investigations at day 7. Therefore, we repeated the experiments in TRF1 -/cells at different time points 141 post infection: day 4 and day 7, finding respectively 1.6% and 2.8% of T-SCEs per chromosome end 142 ( Figure S2, left graph), indicating a lower % of T-SCE events happening at earlier time point. A 143 second distinct difference with previous reports is the type of telomere signal exchanges that we 144 analysed. As in Sfeir et al., 2009, all types of telomere signal exchanges (e.g. the exchanges appearing 145 at single chromatids and the reciprocal exchanges at both chromatids) were considered. However, 146 Martinez et al., 2009 only refers to reciprocal exchanges at both chromatids. Thus, we next classified 147 T-SCEs detected in TRF1 deficient MEFs into these two different types (single and double) and found 148 that 4 days post infection only T-SCEs at single chromatids were significantly increased ( Figure S2, 149 right graph), while the reciprocal exchanges were not enhanced at TRF1 depleted telomeres ( Figure  150 S2, middle graph). Therefore, our detailed analysis of the nature and timing of T-SCEs in TRF1 151  Figure 3A-B) but also 167 at earlier time point (day 4) and in primary MEFs ( Figure S3A-B-C). Collectively, we identify an 168 increase in TERRA molecules upon TRF1 removal from telomeres, confirming transcriptional and 169 telomeric chromatin changes in TRF1 depleted cells. Particularly, the TERRAs molecules increasing 170 upon TRF1 deletion have high molecular weight and can only be detected when an alkaline treatment 171 is performed during Northern-blotting ( Figure 3B; S3D). In addition, we carried out TERRA-FISH 172 ( Figure 3C), confirming a significant increase in numbers and intensity of TERRA foci per nucleus 173 deficient for TRF1 ( Figure 3C). Taken together, these results suggest that TRF1 dependent replication 174 stress at telomeres changes the telomeric chromatin composition by recruiting specific chromatin 175 remodelers, which directly or indirectly affect telomere transcription and contribute to the formation 176 of APBs, platform of recombination. The presence of these ALT-hallmarks suggests that TRF1 177 depleted telomeres present some similarities with ALT telomeres. However, the absence of telomere 178 heterogeneity, c-circle formation and still presence of telomerase activity ( Figure S4A-B-C) also 179 suggest that this ALT-like phenotype is not complete.  deoxyuridine (EdU) and colcemid for 1-hour, mitotic cells were collected to analyse EdU 194 incorporation on metaphase chromosomes ( Figure 4A). We scored for telomeric and non-telomeric 195 EdU foci (mitotic DNA synthesis) and found that CRE induced cells had a significant increase in 196 telomeric mitotic DNA synthesis compared to the GFP control cells ( Figure 4D). This result confirms 197 that TRF1 depleted telomeres present an increased level of non-S-phase DNA synthesis, similar to 198 what is observed in ALT cells. In addition, analysis of EdU incorporation in metaphase spreads 199 allowed us to distinguish between conservative BIR associated DNA synthesis and HR semi- single chromatid. In fact, 72% of the mitotic DNA synthesis at non-telomeric sites localised to a 206 single chromatid, while the remaining 28% of the signal was present at both chromatids ( Figure 4F, 207 upper panel). This result is even more striking when EdU signal was restricted to telomeres, with 208 almost all the co-localisation being present at single chromatids (95%). These observations suggest 209 that TRF1 is crucial for the suppression of mitotic DNA synthesis mediated by BIR at telomeres.   (Figure 2A), we further investigated the role of POLD3 and SMC5 in BIR DNA synthesis 219 observed in TRF1 -/-MEFs. We generated TRF1 F/F cells deficient in SMC5 or POLD3 using specific 220 shRNAs. Upon infection with GFP or CRE adenovirus, we produced respectively single or double 221 deletion TRF1-SMC5 or TRF1-POLD3 cell lines. Loss of SMC5 and TRF1 expression were 222 confirmed by immunoblotting ( Figure 5A-B), while mRNA levels of POLD3 were analysed by RT-223 QPCR ( Figure 5C). We first confirmed that these deletions did not elicit a cell cycle arrest. We only 224 noticed a slight decrease in population doublings in the double mutants, while all cell lines were still 225 able to properly divide and incorporate EdU ( Figure S5A-B). Thus, we carried out EdU-FISH in these 226 cells to check for the presence of BIR ( Figure 5D). We found that the enrichment of DNA synthesis 227 at telomeres in TRF1 deleted cells was suppressed in the double mutant TRF1-POLD3, while the 228 double mutant TRF1-SMC5 revealed similar telomeric DNA synthesis when compared to the single 229 TRF1 mutant ( Figure 5E). First, these results confirm that BIR is the molecular mechanism taking 230 place at TRF1 depleted telomeres. Second, SMC5 appears to be dispensable for BIR dependent DNA 231 synthesis at these replication-stressed chromosome ends. 232 233 SMC5 and POLD3 are required for APBs formation and recombination at TRF1 deficient 234

telomeres. 235
We further examined whether POLD3 and SMC5 could be responsible not only for the BIR dependent 236 DNA synthesis but also for the other ALT-like phenotypes observed at TRF1 deficient telomeres. We could not detect any changes in the frequency of telomere fragility in TRF1-POLD3 nor TRF1-242 SMC5 mutants ( Figure S6C) suggesting that neither POLD3 nor SMC5 are involved in the 243 mechanism that gives rise to telomere fragility. As APBs were increased in TRF1 deleted cells ( Figure  244 2C), we investigated the roles of POLD3 and SMC5 in the formation of these specialised bodies. A 245 significant reduction in number of cells having co-localising PML-telomere foci was detected in the 246 double mutant cells TRF1-POLD3 and TRF1-SMC5 ( Figure 6A) suggesting that POLD3 and SMC5 247 are necessary for the formation of these recombination machinery loci. We next explored the 248 involvement of these two factors in HR by scoring for T-SCE ( Figure 6B  formation, but only POLD3 is required to maintain increased TERRA levels and BIR observed in 256 TRF1 deficient cells. This suggests that POLD3 and SMC5 have separate roles or act at different 257 stages of the recombination events happening at TRF1 depleted telomeres, advocating also an 258 intriguing connection between TERRA and BIR. We speculate that TERRA could trigger the 259 homology search by stimulating the initial steps of BIR in which POLD3 is involved (Figure 7). it has been hypothesized that ALT mechanism arises from persistent replication stress, which can be 282 resolved by the initial collapse of the replication fork, subsequently offering substrates for HR repair 283 mechanisms dependent on homology search and telomere synthesis as reported with BIR pathway 284 (Dilley et al., 2016). 285 In this study, we report that replication stress generated at TRF1 depleted telomeres in telomerase 286 positive MEFs is associated with the recruitment of ALT signature factors including PML, subunits 287 of the NuRD complex, BRCA1 and SMC5/6 complex. We suggest that the formation of permissive 288 telomeric chromatin enables transcription of telomeric sequences into TERRAs and increases 289 recombination as measured by T-SCEs, in a POLD3 and SMC5/POLD3 dependent manner, 290 respectively. Moreover, we detect mitotic DNA synthesis at TRF1 depleted telomeres, which is 291 dependent on POLD3 but not SMC5. Collectively, the presence of replication stress, recombination, 292 APBs formation, TERRA increase and recruitment of specific chromatin factors, suggest a strong 293 analogy between MEFs telomeres deleted for TRF1 and ALT telomeres, supporting the hypothesis 294 that replicative stress could be the source of ALT initiation. 295 We suggest that chromatin remodeling factors such as NuRD-ZNF827 are recruited to TRF1 deficient 296 telomeres to counteract the shelterin instability. This may be explained by analogy with ALT 297 telomeres where telomeric DNA sequence is interspersed with variant repeats ( identified through PICh analysis the SMC5/6 complex specifically recruited at TRF1 deficient 307 telomeres. We demonstrate that this complex plays the same role at replication induced telomeres as 308 in ALT cells, targeting telomeres to PML bodies (APBs) and facilitating telomeric HR at these sites 309 MiDAS is decreased in SMC5/6-depleted Saos2 cells. We speculate, this difference is due to an 343 imbalance of factors used for ALT maintenance, compared to the early events observed in our 344 conditional system after only few population doublings. Therefore, we cannot rule out a possible role 345 of SMC5/6 in promoting MiDAS at a later stage, similar to the one observed in ALT maintenance. generating these phenotypes. 351 In conclusion, our analysis of TRF1 function provides a molecular understanding of the 352 level of protection that this shelterin protein offers at telomeres. The role of TRF1 in facilitating 353 DNA replication at telomeres was already described but only until a certain extent. Surprisingly, we 354 establish that TRF1 is essential for the suppression of early ALT-like signature events including

Gel & post digestion processing 397
Gels were processed using a variant of the in-gel digestion procedure as described in (Shevchenko,

Mass spectrometry analysis 410
Peptides were separated using an Ultimate 3000 RSLC nano liquid chromatography system (Thermo were captured with Zeiss microscope using Carl Zeiss software. Telomeric signals were quantified 454 using the ImageJ software (Fiji). 455 For CO-FISH, the cells were treated with 10µM BrdU:BrdC (3:1) for 16h, followed by 456 colcemid treatment as above. Prior to hybridisation slides were treated with RNAse A (0.5µg/ml in 457 PBS) for 10 min at 37°C, incubated with Hoechst (1 µg/ml in 2XSSC) for 10 min at RT, exposed to 458 UV light for 1h and treated with ExoIII to degrade the neosynthesised DNA strand containing 459 BrdU/C. Slides were next dehydrated through ethanol series, hybridising solution containing TelG-460 FAM probe (Exiqon) in 50% formamide, 2XSSC, 1% blocking reagent was applied to each slide, 461 followed by denaturation for 3 min at 80°C on heating block and hybridisation for 2 hours in the dark. 462 Slides were washed 2 x 15 min in 50% formamide, 2XSSC and 3  5 min in 50 mM Tris pH 7.4, 150 463 mM NaCl, 0.05% Tween-20. Finally, slides were dehydrated, incubated with TelC-cy3 probe for 2 464 hours, followed by the steps described above in the FISH protocol. 465

RNA dot blot 505
RNA extraction was carried out using RNeasy Mini Kit (Qiagen), according to the manufacturer 506 instructions. 2µg of RNA were denatured in 0.2 M NaOH by heating at 65°C for 10 min, incubated 507 5min on ice and spotted onto a positively charged Biodyne B nylon membrane (Amersham Hybond, 508 GE Healthcare). Membranes were UV-crosslinked (Stratalinker, 2000 kJ) and baked for 45 min at 509 80°C, followed by hybridisation at 42°C with digoxigenin (DIG)-labeled telomeric C-rich 510 oligonucleotide TAA(CCCTAA)4, prepared using 3' end labeled kit (Roche). Signal was revealed 511 using the anti-DIG-alkaline phosphatase antibodies (Roche) and CDP-Star (Roche) following the 512 manufacturer's instructions. Images were captured using the Amersham Imager 680 (GE Healthcare) 513 and analysed using the Image Studio Lite software. 514 18s rRNA probe with sequence: 5'-CCATCCAATCGGTAGTAGCG was used for normalisation. 515

Northern Blot 516
10µg of RNA was denatured for 10 min at 65°C in sample buffer (50% formamide, 2.2M 517 formaldehyde, 1X MOPS) followed by ice incubation for 5 min. 10X Dye buffer (50% Glycerol, 518 0.3% Bromophenol Blue, 4mg/ml Ethidium Bromide) was added to each sample and all of them were 519 run on a formaldehyde agarose gel (0.8% agarose, 1X MOPS, 6.5% formaldehyde) at 5V per cm in 520 1X MOPS buffer (0.2M MOPS, 50mM NaOAc, 10 mM EDTA, RNAse free water). The gel was 521 rinsed twice in water, washed twice with denaturation solution (1.5M NaCL, 0.05M NaOH), followed 522 by additional three washes with 20XSSC before transferring the RNA on a positively charged 523 Biodyne B nylon membrane (Amersham Hybond, GE Healthcare) using a neutral transfer in 20XSSC. 524 The membrane was fixed and detected as described for the RNA dot blot. 525  representing the key players for APBs formation and telomere recombination, particularly BIR-856 mechanism. We propose that increased TERRAs molecules at telomeres could lead to increased R-857 loops, which are bypassed by POLD3 dependent BIR to resolve fork progression hindrance.    %T-SCE per chromosome ends (n >2600)