FANCM limits ALT activity by restricting telomeric replication stress induced by deregulated BLM and R-loops

Telomerase negative immortal cancer cells elongate telomeres through the Alternative Lengthening of Telomeres (ALT) pathway. While sustained telomeric replicative stress is required to maintain ALT, it might also lead to cell death when excessive. Here, we show that the ATPase/translocase activity of FANCM keeps telomeric replicative stress in check specifically in ALT cells. When FANCM is depleted in ALT cells, telomeres become dysfunctional, and cells stop proliferating and die. FANCM depletion also increases ALT-associated marks and de novo synthesis of telomeric DNA. Depletion of the BLM helicase reduces the telomeric replication stress and cell proliferation defects induced by FANCM inactivation. Finally, FANCM unwinds telomeric R-loops in vitro and suppresses their accumulation in cells. Overexpression of RNaseH1 completely abolishes the replication stress remaining in cells codepleted for FANCM and BLM. Thus, FANCM allows controlled ALT activity and ALT cell proliferation by limiting the toxicity of uncontrolled BLM and telomeric R-loops.

upon stalling of replication forks by physical impediments including DNA crosslinks 31 . Independently of the FA complex, FANCM remodels replication forks, recruits DNA repair factors at damage sites, suppresses meiotic crossovers and facilitates ATR checkpoint activation [32][33][34][35] . Moreover, the ATPase/translocase activity of FANCM resolves RNA:DNA hybrids in vitro and that R-loops accumulate genomewide in FANCM-deficient cells 36 . We hypothesized that FANCM suppresses replication stress at ALT telomeres, and while this work was in progress, a report from Pan et al. 17 confirmed our hypothesis. The authors showed that, in ALT cells, FANCM allows efficient progression of the replication fork through the telomeric tract, and depletion of FANCM induces telomeric replication stress 17 . The same study also reported that FANCM depletion leads to accumulation of BLM and Brca1 at ALT telomeres and that codepletion of FANCM with Brca1 or BLM is lethal 17 .
Here we show that FANCM depletion in ALT cells causes robust telomere replication stress and damage, activation of ATR signaling, nearly complete abrogation of proliferation, and cell death. In FANCM-depleted ALT cells telomeric ssDNA, ECTRs and mitotic DNA threads accumulate. Moreover, features of ALT activity including APBs and DNA synthesis in G2/M augment when FANCM is depleted. An ATPase/translocase inactive variant of FANCM fails to revert telomeric replication stress and APB accumulation in cells depleted for endogenous FANCM. Finally, FANCM resolves telR-loops in vitro and restricts them in cells, and the replicative stress induced by FANCM depletion is completely averted by simultaneous codepletion of BLM and overexpression of RNaseH1. We propose that FANCM keeps replicative stress and ALT in check by assuring regulated BLM activity and resolving telR-loops.

Results
FANCM supports viability of ALT cells. We depleted FANCM in several ALT (U2OS, HuO9, Saos2 and WI-38 VA13) and telomerase-positive (Tel+; HeLa, HOS, HT1080 and SKNAS) cells using short interference RNAs (siRNAs) against two sequences from FANCM coding region (siFa and siFb). Nontargeting siRNAs were used as controls (siCt). Two days after transfection, nearly complete depletion of FANCM protein was detected by western blot in Fa-and Fb-transfected cells, with the exception of siFb-transfected SKNAS cells, where about 10% of the protein remained (Fig. 1a). Fluorescence-activated cell sorting (FACS) of ethanol-fixed, propidium iodide (PI)-stained cells revealed that FANCM-depleted ALT cells, but not Tel+, accumulated in G2/M phase (Fig. 1b, c; Supplementary Fig. 1A). The clonogenic potential of ALT cells was largely abolished upon transfection of FANCM siRNAs, while the one of Tel+ cells remained essentially unaffected (Fig. 1d, e). For colony formation experiments, cells were transfected only once with siRNAs before seeding and colonies were counted at least 8 days later. Hence, the antiproliferative effects exerted by FANCM depletion on ALT cells are fast and irreversible. Cell growth analysis upon prolonged siRNA treatment showed that FANCM-depleted U2OS cells were quickly eliminated from the population, while HeLa cells continued to grow although at lower rates (Fig. 1f). Finally, FANCM-depleted U2OS cells, but not HeLa cells, started to be permeable to PI already after 3 days of siRNA treatment denoting cell death (Supplementary Fig. 2A). No major changes in PARP1 cleavage were detected in the same cells (Supplementary Fig. 2B). Thus, FANCM depletion causes aberrant accumulation of ALT cells in G2/M phase, followed by PARP1independent cell death.
Our data indicate that FANCM is essential for cell-cycle progression and viability in ALT cells. This is different to what was observed previously 17 . It is possible that less efficient protein depletion obtained by Pan and colleagues or retained expression of crucial FANCM splice variants (possibly including cell-type specific ones that are not reported in public databases) left residual amounts of FANCM protein sufficient to sustain cell proliferation. Less sensitive cell viability assays might also have underestimated the effects of FANCM depletion in the previous study.
Telomeric replication stress sensitizes ALT cells to FANCM depletion. Several features of ALT cells could explain their sensitivity to FANCM depletion: absence of telomerase activity, very long telomeres, ATRX inactivation, and sustained telomeric replication stress. We ectopically expressed the catalytic (hTERT) and RNA (hTR) subunits of telomerase in U2OS and HeLa cells to generate supertelomerase cells 37 . Overexpression of hTERT and hTR was confirmed by quantitative RT-PCR ( Supplementary   Fig. 3A). As expected, HeLa supertelomerase cells had much longer telomeres than HeLa control cells; U2OS supertelomerase cells had reduced TFE frequencies, while the incidence of underreplicated, fragile telomeres (TFs) remained unchanged (Supplementary Fig. 3B) 11,37 . FANCM depletion inhibited cell proliferation and led to G2/M accumulation in U2OS supertelomerase cells, but not in HeLa supertelomerase cells (Fig. 1b-e). Moreover, HeLa cells treated with the telomerase inhibitor BIBR 1532 38 did not accumulate in G2/M when depleted for FANCM ( Supplementary Fig. 4A). We then codepleted FANCM and ATRX in HeLa cells and did not observe the accumulation of G2/M cells ( Supplementary Fig. 4B). Thus, the presence of ultra-long telomeres or the absence of active telomerase or ATRX alone do not explain the sensitivity of ALT cells to FANCM depletion.
We then overexpressed the shelterin factor TRF1 in U2OS cells by retroviral infection, as this treatment halves the incidence of  FTs 39 . FANCM depletion in TRF1 overexpressing cells still led to G2/M accumulation, yet less severely than in cells infected with empty vector (ev) retroviruses (Fig. 1g, h). However, HeLa cells depleted for TRF1 using an siRNA previously shown to induce telomere fragility 39 did not accumulate in G2/M when codepleted for FANCM ( Supplementary Fig. 4C). Similarly, FANCMdepleted HeLa cells did not show an altered cell-cycle distribution when treated with the replication stress inducer hydroxyurea (HU) followed by block release (Supplementary Fig. 4D). Hence, telomeric replication stress contributes to the sensitivity of ALT cells to FANCM depletion; nevertheless, telomeric or generalized replication stress alone are not sufficient to sensitize non-ALT cells to FANCM depletion.

FANCM suppresses telomeric replication stress in ALT cells.
To test the involvement of FANCM in telomere stability, we performed indirect immunofluorescence (IF) using antibodies against TRF2 combined with antibodies against RPA32 phosphorylated at Serine 33 (pS33) or p53 binding protein 1 (53BP1). RPA32 is phosphorylated at serine 33 during S phase by ATR upon replication fork stalling 40 ; 53BP1 forms foci at dysfunctional telomeres that have activated either ATR, or the other DNA damage signaling kinase ataxia-telangiectasia mutated (ATM), or both 41,42 . Within 48 h of transfection, pS33 and 53BP1 accumulated at telomeres in FANCM-depleted ALT cells ( Fig. 2a, b) forming the so-called telomere dysfunction-induced foci (TIFs) 41 . FANCM depletion did not induce TIF formation in Tel+ cells (Fig. 2a, b). Accumulation of pSer33 and 53BP1 outside of telomeres was negligible in all FANCM-depleted cell lines (Fig. 2a). FANCM-mediated suppression of telomere instability is likely to be direct, because the protein associated with telomeric DNA in chromatin immunoprecipitation (ChIP) experiments (Fig. 2c, d). FANCM also immunoprecipitated with the abundant, genomewide-spread Alu repeat DNA (Fig. 2c, d), indicating that the protein is not exclusively associated with telomeres. This is consistent with the reported localization of FANCM to cellular chromatin fractions 43 . Western blot analysis confirmed that FANCM depletion causes pS33 accumulation and revealed phosphorylation of the other ATR target checkpoint kinase 1 (CHK1) in U2OS but not HeLa cells (Fig. 2e). The ATM target KRAB domain-associated protein 1 (KAP1) was not phosphorylated in any of the tested cell lines (Fig. 2e). Moreover, pS33 accumulation was weakened in FANCM-depleted U2OS cells overexpressing TRF1 (Fig. 1g), while telomerase inhibition, ATRX or TRF1 depletion and HU treatment did not promote pS33 accumulation in FANCMdepleted HeLa cells ( Supplementary Fig. 4A-D). Actually, pS33 failed to accumulate efficiently in HeLa cells depleted for FANCM and treated with HU, consistent with a role for FANCM in supporting activation of the canonical ATR-dependent intra Sphase checkpoint 32 . We propose that FANCM deficiency in ALT cells activates a specific ATR-dependent signaling cascade, which is not fully identical to the one triggered by generalized replication stress and stems at least partly from excessive telomeric replication stress. Such ATR response likely provokes the observed G2/M arrest and cell death.
FANCM suppresses ALT features. In our IF images, TRF2 foci in FANCM-depleted ALT cells are both larger and brighter than in control cells (Fig. 2a). To confirm that this was not simply due to increased TRF2 at telomeres, we subjected siRNA-transfected U2OS interphase cells to DNA fluorescence in situ hybridization (FISH) using telomeric probes, and measured the number and area of telomeric foci. We controlled for possible secondary effects related to cell-cycle stage by arresting siCt-transfected cells at the G2/M border with the cyclin-dependent kinase 1 (CDK1) inhibitor RO-3306 44 (Fig. 3a, b; Supplementary Fig. 1B). The overall number of telomeric foci decreased upon FANCM depletion (Fig. 3c, d), while their area distribution was broader, with slightly increased frequencies of very small foci (S in Fig. 3c-e) and substantially increased frequencies of very large foci (L in Fig. 3c-e). RO-3306-treated cells also had less telomeric foci than control cells (Fig. 3d), likely due to clustering of ALT telomeres in G2 19,45 . However, the increase in very small and very large foci was more pronounced upon FANCM depletion than RO-3306 treatment (Fig. 3c-e). Approximately 60% of FANCM-depleted cells had at least five large foci, vs. approximately 10 and 15% of untreated or RO-3306-treated siCttransfected cells, respectively (Fig. 3f).
We then analyzed the localization of PML, RAD51 and POLD3 at telomeres by combining PML and RAD51 IF with telomere FISH, and double IF for POLD3 and RAP1. We observed increased telomeric localization of all three factors in FANCMdepleted U2OS cells, without obvious increase in PML, POLD3 and RAD51 total protein levels (Figs. 3a, 4a, b). RO-3306 treatment did not substantially affect the number of telomeric PML and RAD51 foci, while it increased the one of telomeric POLD3 foci yet less importantly than FANCM depletion (Fig. 4a, b). Moreover, we incubated cells treated as above with the thymidine analog 5-Ethynyl-2′-deoxyuridine (EdU) for 2.5 h, and performed telomere FISH combined with EdU detection to visualize newly synthesized telomeric DNA (Fig. 4a). To exclude S phase cells, we only scored cells showing a punctuate EdU staining and with not more than 25 EdU foci. FANCM depletion increased the incidence of telomeric EdU foci, as it did RO-3306 treatment albeit to lower extents (Fig. 4a, b).
We conclude that FANCM depletion exacerbates ALT activity as shown by robust telomere clustering within large APBs containing PML, RAD51 and POLD3, and increased synthesis of telomeric DNA outside of S phase. FANCM depletion also generates short telomeric species, possibly representing ECTRs (see below). G2/M arrest alone cannot explain the aberrantly elevated ALT features observed in FANCM-depleted cells.

FANCM suppresses telomeric ssDNA and ECTRs in ALT cells.
We performed in-gel telomere restriction fragment (TRF) analysis of genomic DNA from ALT (U2OS and WI-38 VA13) and Tel+ (HOS and HeLa) cells harvested 48 h after siRNA transfection. Blots were hybridized with telomeric oligonucleotides of either 5′-TTAGGG-3′ or 5′-CCCTAA-3′ repeats. When hybridization was performed under native (nondenatured) conditions, we observed increased C-rich telomeric ssDNA of very diverse lengths in FANCM-depleted ALT cells (Fig. 5a, upper panel). Conversely, a decrease of G-rich ssDNA was observed in correspondence of the bulk of telomeres, likely due to shortening of the G-overhang (Fig. 5a, lower panels). For both probes, a fraction of the signal was in the gel wells, possibly corresponding to ssDNA exposed from molecules with significant secondary structures (Fig. 5a). We did not observe alteration of telomeric ssDNA in Tel + cells (Fig. 5a). Hybridization of the same gels in denatured conditions using a long telomeric probe (Telo2 probe) revealed no appreciable alteration of telomere length in FANCM-depleted cells (Fig. 5a).
We then dot-blotted genomic DNA from cells as above and hybridized it under native conditions to telomeric oligonucleotides, followed by denaturation and hybridization with an Alu repeat, as a control for total DNA loaded. This experiment confirmed that FANCM-depleted ALT cells contain more telomeric C-rich ssDNA than siCt-transfected cells (Fig. 5b). As previously reported 11 , depletion of RNaseH1 in U2OS cells also  increased telomeric C-rich ssDNA, albeit at lower levels than FANCM depletion (Fig. 5b). No major difference in total telomeric DNA was detected using dot-blot hybridization of denatured DNA for U2OS, HOS and HeLa cells (Fig. 5b). An increase in total C-rich telomeric DNA was observed in FANCMdepleted WI-38 VA13 cells (Fig. 5b).
We then performed phi-29-mediated C-circle assays 46 using DNA from ALT and Tel+ cells and found a remarkable increase in C-circles in ALT cells depleted for FANCM (Fig. 5c). Accumulation of ECTRs, likely to correspond partly but not exclusively to C-circles, was also detected in FANCM-depleted U2OS cells using two-dimensional gel electrophoresis (Fig. 5d). Metaphase chromosome FISH of FANCM-depleted U2OS cells showed abundant extrachromosomal telomeric signals, probably corresponding to ECTRs, and DNA threads extending from the termini of single chromosomes, or bridging two independent . 3D images were sum projected and areas of individual nuclear FISH signals were measured using DAPI staining to identify nuclei (not shown). A total of at least 300 nuclei from three independent experiments were analyzed for each sample. Areas of telomeric foci (in pixels) are binned into 25 intervals of 5-pixel width (x axis; numbers indicate bin centers) and plotted against frequencies (y axis; %). S small foci (0−2.5 pixels), N normal foci (2.5−57.5 pixels), L large foci (57.6−125.5 pixels). The distribution of large foci is represented in the right graph using a smaller y axis scale to facilitate visualization. f Quantification of cells with at least five Large (L) foci in experiments as in (c). P values were calculated with a twotailed Student's t test. *P < 0.05, **P < 0.005, ***P < 0.001, ****P < 0.0001. Source data are provided as a Source Data file chromosome ends (Supplementary Fig. 5A and B). C-rich ssDNA-containing ECTRs and DNA threads may explain the well-retained DNA molecules observed in our TRF analysis and the increased telomeric ssDNA observed in our dot-blot analysis ( Fig. 5a, b). We did not observe an increase in the incidence of TFEs in FANCM-depleted U2OS cells (Supplementary Fig. 5A and B).
FANCM regulates BLM in ALT cells. FANCM and BLM were reported to collaborate in maintaining ALT telomeres 17 . We depleted FANCM in U2OS and HeLa cells and performed indirect IF using anti-BLM and anti-TRF2 antibodies. Because FANCM is necessary for BLM recruitment to damage sites induced by stalled replication 34 Fig. 6B and C). CPT treatment marginally affected TIF frequencies in all samples (Supplementary Fig. 6B and C). BLM nuclear relocalization occurred without major changes in total protein levels ( Supplementary  Fig. 6A). Hence, we confirm that FANCM depletion causes BLM accumulation at ALT telomeres 17 , while it prevents it at nontelomeric sites of damage both in ALT and Tel+ cells 34 . The telomeric accumulation of BLM upon FANCM depletion might involve reported interactions with TRF1 and TRF2 47 . We then depleted FANCM and BLM simultaneously in U2OS cells (Fig. 6a). BLM depletion alone did not alter cell-cycle FANCM suppresses TERRA and telR-loops in ALT cells. To test whether FANCM suppresses telomere replication stress in ALT cells by regulating TERRA and/or telR-loops, we first performed TERRA northern blot and found that the levels of this lncRNA were 3.5 and 2.5 folds higher in siFa-and siFbtransfected cells, respectively, than in siCt-transfected ones. TERRA species up to~2 kb in length were the most affected (Fig. 7a). We then performed in vitro R-loop resolution assays using telR-loop-containing plasmids generated by T7 transcription of a telomeric tract of approximately 1 kb 11 . We used two  . After signal acquisition, gels were denatured and rehybridized to a long radiolabeled telomeric probe (Telo2). The position of the wells and the sizes in kb of a molecular weight marker are indicated on the left of the gels. b Dot-blot hybridizations of digested genomic DNA from cells as in (a). Control transfections with siRNAs against RNaseH1 (siRH) were also included. Native or denatured DNA was first hybridized to radiolabeled telomeric oligonucleotides. After signal acquisition, membranes were denatured and rehybridized to radiolabeled Alu repeat oligonucleotides (loading). For quantifications (table below), telomeric signals were normalized through the corresponding Alu signal and expressed relative to siCt-transfected samples. Means and SDs from three technical replicates are indicated. Note the accumulation of C-rich ssDNA in FANCM-depleted ALT cells (thick borders). c C-circle assay analysis of genomic DNA from the indicated siRNA-transfected cells harvested 48 h after transfection. Products were dot-blotted and hybridized to a radiolabeled Telo2 probe. Control reactions were performed without phi29 polymerase (Φ29). Note that Tel+ cells had no detectable signals. The graph at the bottom shows quantifications of C-circle signals relative to siCt-samples. Bars and error bars are means and SDs from three independent experiments. P values were calculated with a two-tailed Student's t test. **P < 0.005, ***P < 0.001. d 2D gel electrophoresis of genomic DNA from siRNA-transfected U2OS cells as in a. DNA was denatured and hybridized to a radiolabeled Telo2 probe. Arrowheads point to arches corresponding to circular DNA. Source data are provided as a Source Data file plasmids with different insert orientations as to produce transcripts containing TERRA-like, G-rich RNA repeats, or complementary C-rich transcripts (Fig. 7b). As expected 11 , G-rich transcripts were less efficiently produced than C-rich ones (Fig. 7b). TelR-loop plasmids were incubated with recombinant FANCM in heterodimer with its stabilization partner FAAP24, with or without ATP and then resolved in agarose gels. FANCM promoted complete release of both G-rich and C-rich transcripts from R-loop-plasmids without RNA degradation and in an ATPdependent manner (Fig. 7b). Thus, FANCM efficiently unwinds the RNA moiety of telR-loops in vitro.
To examine telR-loops in FANCM-depleted U2OS cells, we performed DNA:RNA immunoprecipitations (DRIP) using the monoclonal antibody S9.6 48 . Dot-blot hybridization detected telomeric DNA in immunoprecipitated material from all samples, with a~3-fold increase in siFa and siFb samples (Fig. 7c). Treatment of nucleic acids with recombinant RNaseH prior to antibody incubation largely abolished hybridization signals, confirming that they emanated from DNA:RNA hybrids (Fig. 7c). We also performed native DNA FISH using G-rich telomeric probes on interphase nuclei treated or not with RNaseH 11 . A punctate staining corresponding to C-rich telomeric DNA was already visible in untreated siCt-transfected cells, and its intensity was higher in RNaseH-treated cells likely due to degradation of TERRA transcripts within telR-loops and consequent increased binding sites for the probe (Fig. 7d, e). In untreated siFatransfected cells, the C-rich ssDNA signal was more prominent than in control cells and it was further augmented by RNaseH treatment (Fig. 7d, e). The total number of foci per cell was higher in FANCM-depleted cells but was not affected by RNaseH treatment (Fig. 7d, e). We conclude that FANCM suppresses TERRA and TERRA-containing telR-loops in ALT cells. Considering the ability of FANCM to resolve telR-loops in vitro (Fig. 7b) and the localization of FANCM to telomeres (Fig. 2c, d), we propose that FANCM directly resolves telR-loops on telomeric chromatin. The more prominent C-rich ssDNA signal already present in FANCM-depleted cells not treated with RNaseH (Fig. 7d, e) might originate from gaps in DNA replication or cellular degradation of the RNA moiety of telR-loops. Also, although we refer to the telomeric RNA:DNA hybrid structures arising upon FANCM depletion as telR-loops, our experiments do not distinguish between conventional R-loops, three-stranded nucleic acids comprising an RNA:DNA hybrid and a displaced ssDNA, and ds RNA:DNA hybrids devoid of a displacement loop.

Discussion
We demonstrate here that, in absence of FANCM, ALT cells experience severe telomeric replication stress and activate an ATR-mediated DNA damage signaling, which is likely the trigger of the observed G2/M arrest and cell death 50 . The fact that overexpression of TRF1 renders ALT cells less sensitive to FANCM depletion (Fig. 1g, h) and the lack of accumulation of nontelomeric pSer33 and 53BP1 foci in FANCM-depleted ALT cells (Fig. 2a) confirm the centrality of the signal emanating from damaged telomeres in promoting G2/M arrest. However, FANCM may serve essential functions in ALT cells also outside telomeres, as suggested by its physical interaction with Alu repeat DNA. 3D images were sum projected and integrated intensities of FISH signal were measured within individual nuclei identified by DAPI staining (not shown) and background subtracted. Each dot represents an individual nucleus. A total of 100-120 nuclei were analyzed for each sample. One representative experiment is shown. Bars are means. P values were calculated with a Mann−Whitney U test. *P < 0.05, **P < 0.005, ***P < 0.001, ****P < 0.0001. Source data are provided as a Source Data file We also confirm that FANCM is not essential in all cells, as proliferation and viability of Tel+ cells was not majorly affected by FANCM depletion. Consistently, adult humans carrying biallelic loss of function FANCM mutations were reported 51,52 . Moreover, Tel+ human colorectal carcinoma cells, mouse embryonic fibroblasts and chicken lymphoblasts knocked-out for FANCM were successfully generated and proliferated normally unless challenged with DNA damage [53][54][55] . As such, FANCM represents an attractive target in ALT cancer therapy. While it is true that FANCM deficiency is associated with higher risk of breast and liver cancer 51,55 , the irreversible lesions rapidly inflicted by FANCM depletion on ALT cells indicate that shortterm inhibition of FANCM may efficiently eradicate ALT tumors in absence of secondary effects. On the other side, the previously proposed therapy based on coinhibition of FANCM and BLM should be avoided 17 .
We also show that FANCM suppresses ALT-associated features, including clustering of telomeres in PML-, POLD3-and RAD51-containing APBs, and production of ECTRs which include C-circles but possibly other forms (Figs. 4a, b, 5c, d). Because the same features were not evident when we depleted FANCM in Tel+ cells, FANCM deficiency alone is not sufficient to initiate ALT de novo. Further in line with FANCM limiting ALT, its depletion increased synthesis of telomeric DNA outside of S phase (Fig. 4a, b) and led to appearance of DNA threads (Supplementary Fig. 5A and B), which may represent intermediates of intermolecular recombination events. We propose that FANCM suppresses POLD3-dependent telomeric BIR in G2 and possibly MiDAS [24][25][26] . Consistently, deletion of the yeast helicase Mph1, the Saccharomyces cerevisiae FANCM ortholog 56 , directed repair of an HO endonuclease-induced DSB towards BIR 57 . Moreover, Mph1 overexpression inhibited BIR at intrachromosomal DSBs 58 , while it did not prevent insurgence of telomerase-deficient type II survivors, which are ALT yeasts that maintain their telomeres though BIR [58][59][60][61][62] . Finally, MphI localizes to short telomeres in an R-loop-dependent manner 63 . FANCM proteins appear to play specific roles at ALT and uncapped telomeres that are different from the ones that they exert at intrachromosomal damage sites, and that are mediated by R-loops. TelR-loops may directly promote recruitment and/or stabilization and in turn activation of FANCM at telomeres in human ALT cells, thus regulating POLD3-dependent telomeric BIR. The relevance of RAD51 accumulation in APBs when FANCM is depleted remains unclear (Fig. 4a, b)  Comparisons between D145A and ev and WT are not indicated. *P < 0.05, ***P < 0.001, ****P < 0.0001. Source data are provided as a Source Data file NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-10179-z ARTICLE telomere clustering observed in FANCM-depleted cells or other molecular events which we did not investigate, such as sister telomere exchanges. Despite increasing ALT activity, FANCM depletion did not elicit major gain of total (G-rich plus C-rich) telomeric DNA (Fig. 5b). It is possible that in our experiments the amount of newly produced telomeric DNA was below the detection limit. Additionally, de novo synthesis of telomeric DNA in FANCMdepleted cells is likely counterbalanced by incomplete semiconservative replication of telomeric DNA in S phase 17 and by elimination of ECTRs from cells. The exact mechanism by which circular ECTRs are generated remains to be elucidated; nevertheless, they are associated with features that increase in FANCM-depleted ALT cells, including replication stress, activated ATR, C-rich telomeric ssDNA and telR-loops 11,[64][65][66] . One observation of our study is that FANCM depletion does not alter TFE frequencies (Supplementary Fig. 5A and B). This suggests that the observed ECTRs do not derive from excision of entire telomeric tracts.
The replication stress arising at ALT telomeres upon FANCM depletion mainly originates from two sources, deregulated BLM and telR-loops. Given the increased BLM recruitment to ALT telomeres when FANCM is depleted ( Supplementary Fig. 6A−C), FANCM could directly displace BLM from telomeres. Alternatively, FANCM could suppress the triggers provoking BLM recruitment such as arrested telomeric replication forks or R-and D-loop intermediates. BLM activity promotes ALT by supporting telomeric recombination and BIR-based synthesis of telomeric DNA possibly by resolving recombination intermediates formed during BIR-associated strand invasion as part of the BLM-TOP3A-RMI (BTR) dissolvase complex 15,67,68 . Consistently, FANCM depletion increases telomere synthesis outside of Sphase (Fig. 4a, b). Moreover, because BLM mediates long-range resection of DNA ends 69,70 , hyperactive BLM is likely to directly contribute to the production of telomeric ssDNA when FANCM is depleted. This is consistent with the low levels of pS33 TIFs detected in cells double depleted for FANCM and BLM (Fig. 8f). As FANCD2 has also been shown to suppress BLM toxicity in ALT cells 68 , one could speculate that this may be a general role for the FA pathway. Nevertheless, the ATPase/ translocase activity of FANCM is dispensable for FANCD2 monoubiquitination 31 , and overexpression in U2OS cells of a variant of FANCM unable to recruit the FA complex to chromatin did not suppress any ALT-associated feature (accompanying manuscript, "The FANCM-BLM-TOP3A-RMI complex suppresses alternative lengthening of telomeres (ALT)", Lu et al.). It seems therefore unlikely that the entire FA complex functions to maintain ALT.
As for the nature of telR-loops in FANCM-depleted cells, our data suggest that they accumulate because they are not properly dismantled at telomeric chromatin by the ATPase/translocase activity of FANCM (Fig. 7b-e). TelR-loops could be generated cotranscriptionally, as telomeric DNA is a difficult substrate for RNA polymerases 11,49 . The increased short TERRA species observed in FANCM-depleted ALT cells (Fig. 7a) might indeed derive from premature termination of telomere transcription due to improper telR-loop resolution. Also, the accumulation of pS33 indicates that FANCM most likely resolves telR-loops during Sphase. Thus, improper telR-loop resolution can at least in part explain the diminished efficiency of replication fork progression through the telomeric tract in FANCM-depleted cells 17 . Moreover, some FANCM-depletion-associated features of replicationstress, accumulation of C-rich telomeric ssDNA and C-circles (Fig. 5a-e), are more evident in U2OS cells than in other ALTcells. This might be explained by the fact that TERRA and telRloops are particularly abundant in U2OS cells 11 .
Due to the fast and dramatic response of ALT cells to FANCM inactivation (Fig. 1b-f), our studies had to be performed using siRNAs rather than CRISPR/Cas9-based gene inactivation, obfuscating the analysis of genetic interactions. Nevertheless, the synergism between BLM depletion and RNaseH1 overexpression in suppressing replicative stress in FANCM-depleted cells suggests that BLM activity and telR-loops may be functionally related. Although BLM suppresses R-loops genome-wide 71 , it could promote telR-loop formation specifically in ALT cells, for example by generating C-rich ssDNA followed by TERRA annealing. Conversely, telR-loops could recruit BLM to telomeres by stalling telomeric replication forks or by forming D-loop mimicking structures. Future studies, possibly utilizing conditional knockout cells for FANCM and BLM, should refine this intriguing cellular scenario and open the way to novel avenues for curing ALT cancers.
Cell proliferation and viability assays. For colony-forming assays, cells were transfected with siRNAs and 24 h later 300−500 cells were plated in 3 cm dishes and grown until visible colonies were formed. Cells were stained in 1% Crystal violet, 1% formaldehyde, 1% MeOH (Sigma-Aldrich) for 20 min at room temperature, followed by washes in tap water. Plates were air-dried and photographed with a FluorChem HD2 imaging apparatus (Alpha Innotech). Colonies were counted using ImageJ software. For growth curves, cells were transfected with siRNAs and 24 h later 1 × 10 5 cells were seeded in 6 cm dishes and passaged for 10 days. Cells were retransfected with siRNAs and counted every 3 days. For fluorescence-activated cell sorting, cells were trypsinized and pelleted by centrifugation at 500 × g at 4°C for 5 min. Cell pellets were either left untreated for viability assays or fixed in 70% ethanol at −20°C for 30 min and treated with 25 μg/ml RNaseA (Sigma-Aldrich) in 1× PBS at 37°C for 20 min. Cells were then washed in 1× PBS and stained with 20 μg/ml propidium iodide (Sigma-Aldrich) in 1× PBS at 4°C for 10 min. Flow cytometry was performed on a BD FACSCalibur or a BD Accuri C6 (BD Biosciences). Data were analyzed using FlowJo software.
Western blotting. Cells were trypsinized and pelleted by centrifugation at 500 × g at 4°C for 5 min. Pellets were resuspended in 2× lysis buffer (4% SDS, 20% Glycerol, 120 mM Tris-HCl pH 6.8), boiled at 95°C for 5 min and centrifuged at 1600 × g at 4°C for 10 min. Supernatant were recovered and protein concentrations were determined by Lowry assay using bovine serum albumin (BSA; Sigma-Aldrich) as standard. 20-40 μg of proteins were supplemented with 0.004% Bromophenol blue and 1% β-Mercaptoethanol (Sigma-Aldrich), incubated at 95°C for 5 min, separated in 6 or 10% polyacrylamide gels, and transferred to nitrocellulose membranes (Maine Manufacturing, LLC) using a Trans-Blot SD Semi-Dry Transfer Cell apparatus (Bio-Rad). The following primary antibodies were also used: mouse monoclonal anti-FANCM (CV5.  RNA was separated on 1.2% agarose gels containing 0.7% formaldehyde. RNA was then transferred onto nylon membranes and hybridized to a radiolabeled Telo2 probe as above. Ethidium bromide (Sigma-Aldrich) stained tRNAs were used to control for loading. Radioactive signals were detected using a Typhoon FLA 9000 imager (GE Healthcare) and quantified using ImageJ software.
Chromatin immunoprecipitation (ChIP). 10 6 cells were harvested by scraping and resuspended in 1 ml of 1% formaldehyde at room temperature for 15 min. After quenching with 125 mM glycine, cells were washed three times in 1× PBS by centrifuging at 800 × g for 5 min. Cell pellets were resuspended in 500 μl of lysis buffer (1% SDS, 50 mM Tris-HCl pH 8, 10 mM EDTA pH 8) supplemented with cOmplete Protease Inhibitor Cocktail (Roche) and sonicated twice using a Bioruptor apparatus (Diagenode) at 4°C (settings: 30 s "ON" / 30 s "OFF"; power: "High"; time: 15 min). Cellular debris were pelleted by centrifugation at 1600 × g at 4°C for 10 min and 100 μl of supernatant were mixed with 1.1 ml of IP buffer (1% Triton X-100, 20 mM Tris-HCl pH 8, 2 mM EDTA pH 8, 150 mM NaCl). Diluted extracts were precleared by incubation with 50 μl of protein A/G-sepharose beads (GE Healthcare) blocked with sheared E. coli genomic DNA and BSA at 4°C for 30 min on a rotating wheel, followed by centrifugation at 800 × g at 4°C for 5 min. Cleared extracts were incubated with 1 μg of anti-FANCM mouse monoclonal antibody (CE56.1, 72 ) at 4°C for 4 h on a rotating wheel. Immunocomplexes were isolated by incubation with blocked protein A/G beads at 4°C overnight on a rotating wheel. Beads were washed four times with wash buffer 1 (0.1% SDS, 1% Triton X-100, 2 mM EDTA pH 8, 150 mM NaCl, 20 mM Tris-HCl pH 8) and once with wash buffer 2 (0.1% SDS, 1% Triton X-100, 2 mM EDTA pH 8, 500 mM NaCl, 20 mM Tris-HCl pH 8) by centrifuging at 800 × g at 4°C for 5 min. Beads were then incubated in 100 μl of elution buffer (1% Triton X-100, 20 mM Tris-HCl pH 8, 2 mM EDTA pH 8, 150 mM NaCl) containing 40 μg/ml RNaseA at 37°C for 1 h, followed by incubation at 65°C overnight to reverse crosslinks. DNA was purified using the Wizard SV Gel and PCR Clean-up kit (Promega), dot-blotted onto nylon membranes and hybridized overnight to a radiolabeled Telo2 probe as for TRF analysis. After signal detection membranes were stripped and re-hybridized overnight to radiolabeled Alu-repeat oligonucleotides as above. Radioactive signals were detected using a Typhoon FLA 9000 imager (GE Healthcare) and quantified using ImageJ software.
DNA:RNA immunoprecipitation (DRIP). Cells were harvested by scraping and lysed in 1 ml of RA1 buffer (Macherey−Nagel) containing 1% v/v βmercaptoethanol and 100 mM NaCl. Nucleic acids were extracted with phenol/ chloroform/isoamyl alcohol (25:24:1 saturated with 10 mM Tris-Cl pH 7.0, 1 mM EDTA) and precipitated with isopropanol followed by centrifugation at 15,000 × g at 4°C for 10 min. Pellets were washed in 70% ethanol, resuspended in 200 μl Tris-EDTA, 100 mM NaCl and sonicated using a Bioruptor apparatus (Diagenode) at 4°C (settings: 30 s "ON" / 30 s "OFF"; power: "High"; time: 5 min). Five micrograms of nucleic acids was incubated with 1 μg of S9.6 antibody (a kind gift from B. Luke, IMB, Mainz, Germany) in IP buffer (0.1% SDS, 1% Triton X-100, 10 mM HEPES pH 7.2, 0.1% sodium deoxycholate, 275 mM NaCl) at 4°C for 5 h on a rotating wheel. For RNaseH control experiments, nucleic acids were incubated with 60 U of RNaseH in 1× RNaseH buffer or only with buffer at 37°C for 3 h prior to incubation with the S9.6 antibody. Immunocomplexes were isolated by incubation with protein G Sepharose beads (GE Healthcare) blocked with sheared E. coli DNA and BSA. Beads were washed four times in IP buffer by centrifuging at 800 × g, and incubated in elution buffer (50 mM Tris-Cl pH 8, 10 mM EDTA, 0.5% SDS) containing 10 μg/ml proteinase K (Sigma-Aldrich) and 40 μg/ml RNase A at 50°C for 30 min. Beads were centrifuged as above and supernatants recovered. Isopropanol-precipitated DNA was dot-blotted onto nylon membranes and hybridized to radiolabeled 5′-(TTAGGG) 5 −3′ oligonucleotides as for TRF analysis. After signal detection membranes were stripped and re-hybridized to radiolabeled Alu-repeat oligonucleotides as for ChIP analysis. Radioactive signals were detected using a Typhoon FLA 9000 imager (GE Healthcare) and quantified using ImageJ software.
In vitro R-loop resolution assays. Flag-FANCM-8HIS:FAAP24 complex was purified using a baculovirus expression system in Sf9 cells. Cells were pelleted at 500 × g and lysed on ice in 0.5 M NaCl, 0.02 M Triethanolamine pH 7.5, 1 mM DTT, 10% glycerol plus mammalian protease inhibitors (Sigma-Aldrich) and sonicated on ice 5 × 10 s bursts. Clarified lysates were incubated with equilibrated Flag M2 resin (Sigma-Aldrich) for 1 h and 4°C. Flag resin was subjected to 5× batch washes and eluted with 100 μg/ml Flag peptide. Pooled FANCM-FAAP24 containing elutions were diluted to a final concentration of 100 mM NaCl, 20 mM TEA pH7.5, 10% glycerol, 1 mM DTT (Buffer B) and bound to 400 μl ssDNA affinity resin (Sigma-Aldrich). The resin was washed with 10 CV of buffer B. FANCM-FAAP24 complexes were eluted with buffer B containing 0.5 M NaCl. Two micrograms of pcDNA6-Telo or pcDNA6-TeloR plasmids containing a~1 kb fragment of human telomeric repeats cloned downstream of a T7 promoter were in vitro transcribed using T7 polymerase (New England Biolabs) in presence of CTP, GTP, ATP (2.25 mM each) 825 nM [α-32 P]UTP (3000 Ci/mmol; Perkin Elmer). Reactions were stopped by heating to 65°C for 20 min followed by RNase A (EpiCentre) treatment in 330 mM NaCl. R-loop-containing plasmids were purified by two phenol:chloroform extractions. Unincorporated nucleotides were removed by passing nucleic acids twice through S-400 columns (GE Healthcare). R-loop unwinding reactions (10 μl final volume) contained 1 nM R-loop plasmids, 1 mM ATP, 2.5 nM FANCM-FAAP24 in R-loop buffer (6.6 mM Tris pH 7.5, 3% glycerol, 0.1 mM EDTA, 1 mM DTT, 0.5 mM MgCl 2 ). Reactions were performed at 37°C for 10 min and then stopped by adding 2 μl of stop buffer (10 mg/ml proteinase K (New England Biolabs), 1% SDS) and incubating at 37°C for 15 min. Samples were run on 0.8% agarose TAE gels in TAE buffer (40 mM Tris, 20 mM acetic acid, 1 mM EDTA) at 100 V for 60-90 min, followed by gel drying and autoradiography.
Statistical analysis. For direct comparison of two groups, we employed a paired two-tailed Student's t test using Microsoft Excel or a nonparametric two-tailed Mann−Whitney U test using GraphPad Prism. For comparison of two or more factors for each group and their interaction, we used a two-way analysis of variance (ANOVA) followed by Tukey's HSD for the pairwise comparisons. The analysis was carried out using the aov and TukeyHSD functions of R version 3.3.2. The significance levels are from the Tukey's HSD adjusted P values. P values are indicated as: *P < 0.05, **P < 0.005, ***P < 0.001, ****P < 0.0001.
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
The data that support the findings of this study are available from the corresponding author upon reasonable request.