Fanconi anemia and Bloom syndrome are genomic instability syndromes caused by mutations in proteins that participate in overlapping DNA repair and replication pathways. Here, we show that the monoubiquitinated form of the Fanconi Anemia protein FANCD2 acts in opposition to the BLM DNA helicase to restrain telomere replication and recombination in human cells that utilize the Alternative Lengthening of Telomeres (ALT) pathway. ALT relies on exchanges of telomeric DNA to maintain telomeres, a process that we show FANCD2 suppresses. Depletion of FANCD2 results in a hyper-ALT phenotype, including an increase in extrachromosomal telomeric repeat DNAs, putative recombinational byproducts that we show exist as intertwined complexes forming the nucleic acid component of ALT-associated PML bodies. Increases in telomeric DNA are suppressed by loss of BLM but not RAD51, occur without parallel upregulation of shelterin proteins TRF1 and TRF2, and are associated with increased frequencies of deprotected and fragile telomeres. Inactivation of the FA pathway does not trigger ALT, as FANCD2 depleted telomerase positive cells do not acquire ALT-like phenotypes. We observe frequent fragile telomeres in ALT cells, suggesting that telomere sequences are prone to replication problems. We propose that, in ALT cells, FANCD2 promotes intramolecular resolution of stalled replication forks in telomeric DNA while BLM facilitates their resection and subsequent involvement in the intermolecular exchanges that drive ALT.
Homologous recombination (HR) promotes genome stability in mitotic cells through its involvement in DNA repair, DNA replication, and telomere maintenance. Mutations in HR genes result in increased genomic instability and have been implicated in human syndromes associated with tumor susceptibility. Two such syndromes are Fanconi anemia (FA), caused by mutations in any of 19 or more genes (FANCA-T) (1–8), and Bloom syndrome (BS), caused by mutations in the BLM gene (9). The clinical phenotypes of FA and BS share several common features: growth retardation, skin hyperpigmentation, reduced fertility and cancer predisposition (10,11). However the lack of bone marrow failure in BS patients, the distinctive FA phenotype of malformations, and the broader cancer spectrum in BS indicates that even though the pathways are intertwined, important functional differences exist (11–13). Cellular studies support these differences as the sensitivity to interstrand cross-link (ICL) inducing agents is more modest in BS then FA cells (14,15), and only the FANCM and FANCJ complementation groups show the increase in spontaneous genomic sister chromatid exchanges (SCEs) that characterize BS (16–18).
Despite clear differences that exist between FA and BS, at the molecular level the pathways are tightly intertwined. FANCD2 monoubiquitination during replication and after DNA damage is regulated by a core complex of FA and related proteins (19) that also promotes the phosphorylation of BLM after interstrand DNA crosslinks (ICLs) (15). A subcomplex of FANCM and associated proteins recruits both the FA core complex and the BLM/topoisomerase IIIα/RMI1/RMI2 complex to chromatin in response to ICL exposure (17), and FANCD2 and BLM act in the same epistatic pathway to suppress ICL induced quadriradials (20). FANCD2 also plays a role in BLM protein stability and recruitment to chromatin during replication (21) and FANCD2 accumulates at the base of fragile sites connected by BLM coated ultrafine bridges present during mitosis (22,23). Together, these observations illustrate the still enigmatic relationships between FA proteins and BLM.
At a mechanistic level both FANCD2 and BLM bind single stand DNA (ssDNA) and HR intermediate structures. BLM facilitates HR repair by unwinding double strand DNA (dsDNA) and promoting end resection (24,25), single strand DNA annealing (26), strand exchange (27) and DNA synthesis (28). The early HR activities of BLM may contribute to its observed role in the restart of stalled replication forks (29). In the later phases of HR repair, BLM exhibits anti-recombinase activities, by promoting the dissolution of Holliday junctions (30,31), melting of displacement loops (32), and displacing RAD51 from ssDNA filaments (28).
Unlike BLM, FANCD2 does not appear to be a canonical repair factor, but functions in a context dependent manner (33,34). E.g., FANCD2 regulates essential nuclease reactions during ICL repair by restraining DNA2 activity and recruiting the FAN1, SLX4/FANCP, and XPF/FANCQ-ERCC1 nucleases to stalled/collapsed forks (35–40). In a different context, FANCD2 plays a critical role in restraining MRE11 activity at replication forks stalled by exposure to HU or aphidicolin (41). Together, this suggests a central role for FANCD2 in the regulation and coordination of nucleases during HR.
We and others previously reported a role for BLM in regulating telomeric DNA production in cells that maintain telomeres via the Alternative Lengthening of Telomeres (ALT) mechanism (42,43). In this paper we utilize ALT as a means to study the roles of FANCD2 and BLM in spontaneous recombination events involving telomeres, events which may favor different pathways than those triggered by exogenous damage or replication fork stalling agents (44). Through a comparison of ALT and telomerase immortalized cells we also further explore the role of FANCD2 within non-ALT telomere maintenance, as telomere shortening and other abnormalities have been reported in some FA cells (45–49).
The ALT pathway is an HR-dependent process that is active in the 10-15% of human cancers that lack telomerase (50–52). Common ALT phenotypes include frequent exchanges of telomeric DNA, variable telomere lengths, extrachromosomal telomeric repeat (ECTR) DNA, telomeric DNA that has activated a damage response, and the presence of ALT-associated PML bodies (APBs) (50,53–57). The precise role(s) of APBs in ALT are unclear, with hypotheses ranging from APBs as structures that facilitate active recombination between telomeres to passive storage depots for ECTR DNA. In this study, we use electrospectroscopic imaging (ESI) to provide the first look at the ultrastructure of APBs in wildtype and FANCD2 depleted ALT cells, and together with our Halo-FISH technique, demonstrate that the primary component of APBs is not telomeres, but intertwined molecules of ECTR DNA. We further report that monoubiquitinated FANCD2 plays a critical role in restraining BLM within ALT, suggesting that these two HR proteins have antagonistic roles in ALT. Lack of FANCD2 results in the activation of RAD51-independent pathways in ALT cells that drive replication of and recombination between telomeric DNA sequences through a BLM-dependent mechanism, but does not result in telomerase positive cells acquiring ALT-like phenotypes.
Monoubiquitinated FANCD2 limits telomeric DNA synthesis in ALT, but not telomerase positive, human cells
In agreement with Fan et al (58), we found that FANCD2 formed frequent spontaneous nuclear foci that colocalized with telomeric DNA and telomere binding proteins (TRF1, TRF2) in cells that utilize ALT telomere maintenance, but not in cells that utilize telomerase (
To investigate the role of monoubiquitinated FANCD2 at telomeres, we examined the effect of reduced FANCD2 expression on telomeric DNA in ALT and telomerase positive human cells. We found that depleting FANCD2 using two different siRNAs targeting FANCD2 mRNA resulted in significant increases in total telomeric DNA content in ALT cells (Fig. 1). Increased telomeric DNA content in FANCD2-depleted ALT cells was observable by both microscopic examination and flow cytometric measurement of FISH stained TTAGGG telomere repeat sequences in 3 different ALT lines (Fig. 1, data not shown). Telomerase positive cells depleted of FANCD2 did not show detectable increases in telomeric DNA content through microscopic imaging or flow cytometric analysis of telomeric DNA content (data not shown, Fig. 1C), demonstrating that this is an ALT-specific phenomenon. Similar to FANCD2 depletion, knockdown of FANCA in ALT cells resulted in a change in the population distribution detectable with flow-FISH, with a higher proportion of cells containing increased amounts of telomeric DNA (Fig. 1C). The similar phenotypes of FANCA and FANCD2 deficient cells confirm that the monoubiquitinated form of FANCD2 regulates the amount of ALT telomeric DNA. Of note, the most extreme change in telomeric DNA content occurred following FANCD2 depletion in GM847, followed by VA13, then U2OS cells, correlating with the fraction of total telomeric DNA that is ECTR DNA in each cell line (59).
Increased telomeric DNA content and exchanges in FANCD2-depleted cells occur through a BLM dependent, RAD51 independent mechanism
The ALT-specific increase in telomeric DNA observed when FANCD2 is depleted is similar to observations made after transient over-expression of helicase active BLM (42). We performed western blot analysis of BLM expression in siRNA treated cells and found that knockdown of FANCD2 does not increase BLM expression (
RAD51 was recently identified as playing a key role in bringing ALT telomeric DNA together and promoting recombination (52). As both FANCD2 and BLM can affect RAD51 mediated processes, we performed siRNA co-depletion experiments to further investigate the roles of BLM and RAD51 in the FANCD2-depletion phenotype (Fig. 2A). Using flow-FISH to monitor total telomeric DNA content per cell we found that co-depletion of BLM suppresses the increase in telomeric DNA caused by FANCD2 depletion, shifting the median telomere fluorescence and population distribution back to control levels (Fig. 2B). Strikingly, we see that, for FANCD2-depleted GM847 cells, co-depletion of BLM reduces the fraction of cells with telomeric DNA content at or exceeding the limit of detection from 38% to 0%. Co-depletion of BLM with FANCD2 in VA13 ALT cells also reduces the frequency of cells with high amounts of telomeric DNA from 14% to 0%. siRNA depletion of BLM alone also reduces the telomeric DNA content per cell relative to control cells (43).
In contrast to BLM, co-depletion of RAD51 slightly increases the median telomeric DNA content in FANCD2-depleted GM847 and VA13 cells (Fig. 2B). To determine if the reduction in RAD51 protein was biologically significant, we scored metaphases from GM847 cells treated with either siRAD51 or siControl for endoreduplication, a phenomenon associated with reduced RAD51 loading (60). A random assessment of 1235 metaphases showed that 42/600 siRAD51 but 0/635 siControl treated cells had undergone endoreduplication, confirming biologically significant reductions in RAD51 protein levels.
To assess the stability of telomeres in FANCD2 depleted ALT cells we measured the frequency of recombination events involving telomeres in the first cell cycle following FANCD2 depletion using Chromosome Orientation-FISH (CO-FISH). With the CO-FISH assay, it is not possible to determine whether exchanges occur between telomeres on sister chromatids, other chromosomes, or with ECTR DNA; therefore, these events are referred to simply as telomeric DNA exchanges. In GM847 ALT cells the mean frequency of chromosome ends with evidence of a telomeric DNA exchange increases from 5.0 ± 0.7% in siControl cells to 14.4 ± 1.9% in siFANCD2 treated cells (Fig. 2C). This increase is ALT specific, as depletion of FANCD2 in HT1080 telomerase positive cells does not cause an increase in the frequency of exchanges (1.5 ± 0.09% in siControl and 1.5% ± 0.2% siFANCD2 treated cells). Similar to changes in total telomeric DNA content, co-depletion of BLM with FANCD2 reduces the frequency of chromosome ends with a telomeric DNA exchange to 6.0 ± 0.9% in GM847 ALT cells, while exchanges remain elevated in FANCD2 and RAD51 depleted cells at 14.1 ± 6.2% (Fig. 2C).
In the CO-FISH assay, there normally will be a single signal at each chromosome end and these two telomeric signals will be in trans. If, however, a genomic sister chromatid exchange (SCE) has caused a cross over to occur upstream of the telomere, then that chromosome’s telomeric signals will be in a cis configuration. When we examined the frequency of telomeres in cis vs trans, we found that co-depletion of BLM with FANCD2 resulted in a 3.1 fold increase, a phenotype indicative of frequent genomic SCEs. This observation is consistent with the well documented role of BLM in limiting genomic SCEs, and indicates that distinctly different mechanisms underlie telomeric and genomic SCEs.
In addition to telomeric exchanges measurable by CO-FISH, we also found evidence of interactions between interphase telomeric foci in the form of linear telomeric DNA fibers often connecting large telomeric foci to smaller distal telomeric foci (Fig. 2D). Knockdown of FANCD2 resulted in a 3-fold increase in the frequency of nuclei with linear fibers, with 15.0 ± 7.2% of control vs 44.5 ± 7.3% of FANCD2- depleted nuclei having at least one detectable fiber (Fig. 2D). Again, co-depletion of BLM reduces the frequency of cells with a telomere fiber to 21.7 ± 8.5%, while co-depletion of RAD51 (50.3 ± 6.7%) does not (Fig. 2D).
FANCD2 localizes to APBs and depletion of FANCD2 leads to APBs composed of non-chromatinized DNA and chromatin like structures
FANCD2 colocalizes with telomeric foci primarily at APBs, with 79% and 88% of FANCD2 foci that colocalize with TRF1 also colocalizing with PML in GM847 and VA13 ALT cells, respectively (
To increase our understanding of the effects of FANCD2 depletion on APBs in ALT cells, we analyzed APBs and ordinary PML nuclear bodies (PML NBs) by electron spectroscopic imaging (ESI). ESI monitors energy loss of electrons undergoing inelastic scattering events, allowing nitrogen (N) and phosphorous (P) atoms to be identified and element specific maps to be generated. Protein (N rich, P poor) and nucleic acid (P rich) based structures can be subsequently identified from these maps (61).
Consistent with our previous data (62), we found that the N to P ratios (N/P) of PML NBs were 8-25 fold (GM847) higher than surrounding chromatin (Fig. 3A). These high N/P ratios are consistent with a protein-based structure that contains no nucleic acid. In contrast, the ultrastructure of APBs is characterized by a N rich, P poor, protein-based outer ring, surrounding a core that contains both N and P, indicating the presence of nucleic acid (Fig. 3). The presence of nucleic acid within APBs was confirmed in the VA13 cell line (data not shown). While nucleic acid clearly resides within the APB core, the N and P signals and morphology of this material are inconsistent with a protein (histone):DNA ratio of that of chromatin, leading us to conclude that it is primarily non-chromatinized nucleic acid.
In contrast to control APBs that do not appear to contain chromatin, approximately one third of APBs in FANCD2-depleted GM847 cells contain blocks of P rich material that have invaded the APB and have a N/P ratio of 1.1-1.9, making these areas more similar to surrounding chromatin than material in the remainder of the body (Fig. 3, solid arrow heads). Analysis of serial sections confirmed that P rich blocks are within, not merely touching APBs, and may represent telomeres themselves, or ECTR DNA derived directly from the telomere, as they morphologically and chemically resemble chromatin. To confirm that nucleic acid within FANCD2-depleted APBs primarily represents DNA, not RNA, a serial section of an APB from a FANCD2-depleted GM847 cells was treated with RNAse (Fig. 3B), which significantly decreased P intensity in a neighboring nucleolus but not the P signal within the APB. As putative chromatin structures were only observed in FANCD2 depleted cells, these interactions are either unique to FANCD2-depleted cells, or occur more frequently or stably when FANCD2 is depleted, allowing them to be observed.
FANCD2 depletion leads to increases in telomeric DNA length and ECTR number in ALT cells
To determine whether the increased telomeric DNA exchanges and enhanced interactions between APBs and chromatin seen with FANCD2-depletion are associated with changes in telomere lengths and ECTR DNA production we used Quantitative-FISH analysis of metaphase spreads to determine telomere lengths, and our novel Halo-FISH assay (59) to quantify ECTR DNA production on a single cell basis. Depletion of FANCD2 in GM847 cells resulted in a significant increase in mean telomere length, from 10.2 to 18.1 kb (Fig. 4A). The change in mean telomere length in FANCD2 depleted cells is driven by a decrease in the frequency of very short telomeres (<5kb) and a concomitant increase in longer telomeres (Fig. 4A). Halo-FISH analysis shows that depletion of FANCD2 also leads to striking increases in both G and C strand ECTR DNAs (Fig. 4B), with 9-fold (G strand) and 12-fold (C strand) increases in mean ECTR DNA content per cell over siControl cells (Fig. 4C). Notably, this increase in ECTR DNA content in FANCD2 depleted cells was driven by an increase in the number of ECTR molecules, not by a shift in the size distribution (
Importantly, co-depletion of BLM with FANCD2 resulted in 20 and 52-fold reductions in mean G and C ECTR DNA production relative to FANCD2 siRNA treated cells, demonstrating that BLM is required for ECTR DNA production in FANCD2 depleted cells (Fig. 4C). Depletion of BLM alone reduced the mean G ECTR content 3.3-fold relative to controls (
To better understand the nature of interactions between ECTR DNAs, we modified our standard Halo-FISH protocol by omitting NaOH treatment after deproteinization in order to preserve DNA aggregates. Using this modified protocol, the majority of ECTR DNAs in FANCD2-depleted cells remained as large deproteinized nuclear foci rather than dispersing into the agar as individual molecules, indicating that, in siFANCD2-depleted ALT cells, ECTR DNA molecules are held together in large paranemic complexes through hydrogen bonding (
TRF1 and TRF2 expression are not upregulated in FANCD2 depleted ALT cells, correlating with an ALT specific increase in TIFs and fragile telomeres
The binding of shelterin complex proteins to telomeric DNA is required to suppress activation of DNA damage responses at telomeres (63) and promote their successful replication (64). Telomere dysfunction induced foci (TIFs) form at ALT telomeres irrespective of telomere length, and can be partially suppressed through over-expression of the shelterin protein TRF2, suggesting an insufficient molar ratio of TRF2 to telomeric DNA contributes to the ALT TIF formation by impairing telomere protection (65). Decreased expression of the shelterin protein TRF1 also leads to replication fork stalling in telomeric DNA and TIF formation (64). We examined TRF2 and TRF1 protein expression (Fig. 5A) in FANCD2 depleted cells, and found that there is no concomitant increase in TRF2 or TRF1 expression (Fig. 5A) even though total telomeric DNA production is increased 3-4 fold (Fig. 1). The result is a further decrease in the ALT cells’ all ready low molar ratio of TRF1/2: telomeric DNA (65) and a further increase in the risk of deprotected telomeres and stalled replication forks.
As might be expected from a decreased TRF1/2: telomeric DNA molar ratio, siRNA knockdown of FANCD2 in GM847 and VA13 ALT cells resulted in a 2.9 fold average increase in the number of TIFs (53BP1/TRF2 colocalization) per cell. 10.6±0.5% and 17.8±1.9% of all TRF2 foci colocalized with 53BP1 in FANCD2 depleted GM847 and VA13 ALT cells, an increase from the 3.9±1.5% and 5.8±2.2% of telomeric foci in siControl cells. In telomerase positive cells, there was no change in the average number of TIFs per cell (Fig. 5B) or the frequency of TRF2 foci that colocalize with 53BP1 (3.0±1.3% to 3.1±1.5%, and 1.9±0.4% to 1.7±0.4% in siControl vs siFANCD2 depleted GM639 and HT1080 cells respectively).
The TRF1 shelterin protein promotes replication of telomeric DNA, prevents replication fork stalling, and limits the incidence of telomere fragility, a phenotype cytogenetically defined by a discontinuous FISH staining pattern in metaphase (64). Interestingly, telomeres in ALT cells more commonly express a fragile telomere phenotype then telomeres in telomerase positive cells, suggestive of ongoing replicative stress at ALT telomeres (Fig. 5C). Consistent with a model in which FANCD2-depletion impairs telomere protection as a result of a relative scarcity of shelterin molecules, we find a 2-fold increase in the frequency of fragile telomeres in GM847 ALT cells when FANCD2 is depleted, relative to controls (Fig. 5C). In contrast, the frequency of fragile telomeres is not increased in FANCD2 depleted HT1080 telomerase positive cells (Fig. 5C).
In this study, we report that depletion of FANCD2 results in a hyper-ALT state in which multiple aspects of the ALT phenotype are exaggerated: Telomeric DNA exchanges, telomere length variation, fragile telomeres, TIFs, ECTR DNA production, APB body size, and interactions between telomeres and APBs. In contrast, depletion of FANCD2 in telomerase positive cells did not result in detectable expression of ALT phenotypes. Taken together, our results indicate that inactivation of FANCD2 is neither necessary nor sufficient to trigger ALT in transformed cells. However, FANCD2 plays a key role in regulating and restraining ALT.
The frequency of telomeric DNA exchanges in FANCD2-depleted cells increased during the first cell cycle after FANCD2 depletion, suggesting that development of the hyper-ALT behavior in FANCD2-deficient cells as well as the ALT phenotype itself may begin with exchanges between telomeric DNAs. Our results, that depletion of FANCD2 increases telomeric DNA exchanges but not genomic SCEs, and that co-depletion of BLM reduces telomeric exchanges but increases genomic SCEs (Fig. 2C), are consistent with data indicating that the mechanism for ALT telomeric DNA exchanges differs from those involved in damage induced and spontaneous genomic SCEs (51,66). Our findings that FANCD2 limits the amounts of and exchanges between ALT telomeric DNA are at odds with an earlier study by Fan and colleagues (58). Although we performed similar experiments, differences in efficiency of protein knockdowns, technique, and timepoints may underlie our varying results. As well, cell lines may have differed as Fan et al (58) reported that GM847 cells are refractory to siRNAs knockdown, a phenomenon not observed in our study or in other published works using this cell line (67–69).
Using ESI, we provide the first look at the ultrastructure of wildtype APBs in unaltered interphase ALT cells, and reveal that the primary DNA species within APBs is likely to be ECTR DNA, not telomeres, as the material within wildtype APBs have N/P ratios and morphology that are inconsistent with chromatin and ALT telomeres are chromatinized structures (71). APBs do however, make contact with neighboring chromatin on the periphery of the body, leaving open the possibility that telomeres may contact the periphery of the body, an idea supported by large aggregates of telomeric DNA detected in close proximity to metaphase telomeres (65). The continued association between ECTR DNA and telomeres would enable the stable transmission of ECTR DNA to daughter nuclei, similar to the well described ability of viral episomes to ‘piggyback’ onto cellular chromosomes during mitosis (72). Our examination of metaphase spreads suggests that within a single cell there are multiple fates of ECTR, with some remaining associated with chromosomes and therefore stably inherited, and other ECTR DNA that are disassociated from chromosomes and therefore lost during mitosis.
Approximately one third of APBs in FANCD2-depleted cells have blocks of chromatin-like material that have invaded the APB core, a structure consistent with direct interactions between telomeres and ECTR DNAs. Chromatin-like material was not observed in APBs in control cells, suggesting that this phenomenon is unique to, or more readily detectable, in FANCD2-depleted cells. Double strand breaks (DSBs) occurring within ALT telomeric DNA trigger interactions between telomeric foci and intermolecular exchanges of telomeric DNA (52) raising the possibility that broken telomeres may be targeted to APBs. FANCD2-depletion may result in more frequent DSBs within telomeres, contributing to the observed increase in TIFs (Fig. 5) and increasing the probability of detecting chromatin like material within APBs. Alternately, interactions between telomeres and ECTR DNA may be more stable in FANCD2-depleted cells. However, even in FANCD2-depleted cells the majority of the DNA within APBs is non-chromatinized, reinforcing the idea that these structures should not be thought of as aggregates of replicating or recombining telomeres, but rather as complexes of ECTR DNA.
We propose that ECTR DNAs within APBs represent a rich mixture of recombinational substrates that actively engage in multiple recombination and replication reactions amongst themselves, generating additional ECTR DNA. This mechanism of ECTR DNA production would be mechanistically distinct from ECTR DNA produced during telomere trimming through aberrant resolution of the t-loop (57,70), and would allow for the concomitant increase in telomere lengths and ECTR DNA production observed in FANCD2-depleted ALT cells. The idea, that APBs are not passive storage depots for sequestering ECTR DNA, is supported by observations of newly synthesized DNA within APBs (73,74) and our Halo-FISH experiments which demonstrate a requirement for disruption of intermolecular hydrogen bonding through NaOH treatment for the liberation of ECTR DNA from large foci. The latter suggests that, in vivo, ECTR DNAs exist in higher-order complexes that are held together by multiple paranemic DNA-DNA interactions (e.g., D-loops, Holliday junctions, and annealed ssDNAs/gapped dsDNAs) rather than by protein-DNA interactions or knotted/catenated/plectonomic structures that cannot be disassembled by NaOH treatment.
The underlying stimulus driving exchanges of ALT telomeric DNA and production of ECTR DNA is unclear, but may involve the cellular responses to replication fork problems, as large APBs form in ALT cells after replication fork stalling by nucleotide depletion, and FANCD2 is recruited to these stress induced APBs (58,75). FANCD2 and BLM are also both involved in the response to replication problems arising in common fragile sites (22,23,80), although it is noteworthy that the repetitive nature of telomeric DNA is one factor that distinguishes it from common fragile sites. ALT telomeric DNA is further distinguished by interspersed noncanonical telomeric repeats, C rich overhangs, ssDNA breaks/gaps, and an open chromatin structure (71,76–79), which may increase endogenous replication stress. In addition, ALT cells have insufficient molar ratios of telomere binding proteins to deal with these issues (65), a problem which appears to be further exacerbated in FANCD2-depleted ALT cells, as indicated by their elevated frequencies of fragile telomeres and TIFs (Fig. 5) as well as their 3-4 fold lower molar ratios of TRF1/2: telomeric DNA (Figs 1 and 5A).
We propos'e a model (Fig. 6), wherein monoubiquitinated FANCD2 promotes the intramolecular resolution of replication forks that have stalled in ALT telomeric DNA, thereby limiting intermolecular exchanges. FANCD2 may also actively inhibit intermolecular reactions by recruiting structure specific nucleases to cleave downstream recombinational products, similar to its role during ICL repair (36–40).We further hypothesize that FANCD2 limits nuclease digestion of nascent telomeric DNA, thereby promoting fork stabilization/reversal and preventing production of pathogenic ssDNA or replication fork collapse. Supporting this hypothesis, FANCD2-depleted ALT cells accumulate high amounts of ssDNA that colocalizes with telomeric DNA (41) and oppose DNA2 helicase/nucleases activity following cisplatin or formaldehyde treatment (35).
Our data also suggest that FANCD2 restricts the role of BLM in ALT, as co-depletion of BLM suppresses the FANCD2 ALT phenotype, while BLM overexpression results in elevated amounts of ALT telomeric DNA (42). Knockdown of BLM results in loss of total telomeric DNA and decreased ECTR DNA production (26). The pro-recombinational roles of BLM in FANCD2-depleted cells may require phosphorylation of BLM by ATR, an event stimulated by replication stress (84). ATR is required for activation and maintenance of the ALT pathway (85,86), and codepletion of ATR also suppresses the increased telomeric DNA synthesis seen in FANCD2-depleted ALT cells (
The repetitive nature of telomeric DNA and the presence of ssDNA gapped regions within ALT telomeric DNA (79) would facilitate homologous pairing via RAD51-independent single strand annealing (SSA), providing a mechanism for the RAD51-independent telomeric exchanges and DNA production we observe in FAND2-depleted cells. This would be conceptually similar to ALT-like type 2 telomere maintenance in yeast, a RAD51-independent process which is thought to rely on break induced replication coupled to single strand annealing and also requires SGS1, a BLM ortholog (81–83). Surprisingly, in the FANCD2-deficient background these RAD51-independent processes are at least as efficient as RAD51 processes, as RAD51 knockdown does not result in even a partial suppression of the FANCD2 phenotype. One possibility is that the lack of FANCD2 creates a permissive environment that increases the efficiency of RAD51-independent SSA, potentially explaining why in wildtype ALT cells RAD51-independent processes can only partially compensate for knockdown of factors required for strand invasion (52).
Features of the ALT telomere phenotype that become exaggerated when FANCD2 is depleted have also been reported in Fanconi anemia hematopietic cells: Elevated exchanges involving telomeric DNAs, increases in ECTR DNA, TIFs, and fragile telomeres (45,46,48,49). Interestingly, while aberrant telomere phenotypes were not reported in FA patient fibroblast lines or the FANCG mouse model (87) a specific telomere length dependent role for a functional FA core complex in limiting recombination between telomeres and preserving telomere stability has been found in telomerase-deficient FANCC mice with short telomeres (88). Taken together with the results of our study of ALT and telomerase positive human cells, these findings suggest a model wherein FANCD2 is excluded from functional, capped telomeres but plays a critical role in promoting telomere stability when telomeres become uncapped through excessive shortening or, in the case of ALT cells, length independent factors. We propose that aberrant telomere maintenance may contribute to FA disease pathology, as elevated turnover of FA progenitor cells due to telomere length independent factors may result in shorter, improperly capped telomeres, which would show increased instability in the FA background, putting further replicative pressure on remaining progenitor cells.
Materials and Methods
Cell culturing and transfections. GM00847 (GM847), Wi38-VA13/2RA (VA13), U2OS, GM00639 (GM639), HT1080, and PD20 cells were grown in DMEM with 10% fetal bovine serum and penicillin-streptomycin, and were mycoplasma free. Fugene 6 (Roche) and Lipofectamine RNAiMax (Invitrogen) were used according to manufacturer’s instructions for plasmid and siRNA transfections, respectively. pMMP-puro-EGFP-FANCD2 was a gift from Dr. Alan D’Andrea. The Quick change II XL site directed mutagenesis kit was used to introduce a K561R mutation in pMMP-puro-EGFP-FANCD2. SiRNA oligonucleotides were synthesized (Dharmacon) to target the following sequences: FANCD2 (A, 5’-GGAGATTGATGGTCTACTA-3’ (89); B, 5’-CCAGGA AG CAACCACTTTC-3’) FANCA (5’-AAGGGTCAA GAGGGA AAA ATA-3’ (90,91)) BLM (5’-GAGCACATCTGTAAATTAA-3’) RAD51 (5’-GAGCTTGACAAACTACTTC-3’ (92)) and control GL2 (5’-AACGTACGCGGAATACTTCGA-3’ (89). For CO-FISH 1x105 cells were transfected with 100 nM siRNA and analyzed 56-60 hrs post transfection, for all other experiments 1 × 105 cells were transfected with 100nM siRNA at 0 and 48 h, and cells were harvested at 120 h.
Fluorescence microscopy. Cells were immunostained as described (42). Primary antibodies against FANCD2 (NB100-182 Novus Biologicals; SC-20022 Santa Cruz), FANCA (gift from Dr. Manual Buchewald), TRF2 (IMG-124 Imgenex), TRF1 (sc-6165 Santa Cruz), PML (AB1370 Chemicon; PML 5E10 gift from Dr. Roel van Driel), 53BP1 (NB 100-304 Novus Biologicals), RAD51 (PC130 EMD Chemicals) and F(ab’)2 affinity purified secondary antibody fragments conjugated to fluorophores (Jackson ImmunoResearch) were used. Images were captured with a Zeiss Axioplan 2 microscope equipped with a Zeiss Plan-Apochromat 40x or 63x/0,95 Korr ∞/0,13-0,21 objective and a Hammamatsu Orca ER camera using Openlab software version 5.5.1 (PerkinElmer). For Q-FISH and Halo-FISH experiments 0.2μm z-stack images were taken and unfocused light was removed by iterative deconvolutions to a 95% confidence level prior to quantitative analysis with Volocity (PerkinElmer). Halo-FISH experiments were analyzed as described (59).
Immunoblotting.Ten micrograms of whole cell lysate (RIPA extraction) were run on NuPAGE 4-12% Bis-Tris gels (Invitrogen) and blotted to PVDF. Primary antibodies to FANCD2 (sc-20022 Santa Cruz), β-tubulin (sc-5274 Santa Cruz), BLM (NB100-214 Novus Biologicals), RAD51 (ab213 Abcam), TRF2 (IMG-124 Imgenex), TRF1 (ab1423 Abcam) were detected with HRP-labelled secondary antibodies (Jackson ImmunoResearch) followed by ECL (GE Healthcare).
FISH. Cells for immunoFISH were immunostained as described above, fixed in 4% PFA for 20 min, then FISH was performed as described (93). Halo-FISH was performed as described (59). Telomere FISH was carried out as previously described (94) with minor modifications. Cells were placed in 75mM KCl (15min at 37 °C), fixed in methanol/acetic acid 3:1, and dropped onto slide following standard protocols. Cells were grown in 0.1ug/ml colcemid for 2–12 h prior to analysis of metaphase chromosomes. Hybridization mixture containing 70% formamide, 0.5ug/ml telomere PNA probe (Rho-(C3TA2)3), 0.5ug/ml FITC-pan-centromeric PNA probe (95), 10mM Tris pH 7.2, 0.1% blocking reagent (Boehringer), MgCl2 buffer (4.1mM Na2HPO4, 0.45mM citric acid, 1mM MgCl2) was preheated for 3 min at 86 °C, added to slides, covered with a coverslip, than slides were heated for 3 min at 81 °C and left for 2 h at room temperature prior to 2 x 15 min washes in 70% formamide, 10mM Tris pH7.2, 0.1% BSA, 3 x 5 min in 100mM Tris pH 7.2, 150mM NaCl, 0.08% Tween 20. For CO-FISH experiments 36 h post siRNA addition (the earliest time point where uniform FANCD2 depletion was observed by immunofluorescence) 7.5uM BrdU, 2.5uM BrdC was added to the media for 24 h (847) and 20h (HT1080) with 0.1ug/ml colcemid added for the last 2 h of growth prior to harvesting. Newly synthesized DNA strands were degraded by treating fixed cells on glass microscope slides with 0.5 μg/ml Hoechst 33258 (Sigma), exposing them to 312 nm light for 45 min while slides were on a 55 °C heating block, and digesting with Exonuclease III (3U/uL NEB) for 12.5 min at room temperature. Slides were dehydrated and processed for FISH as described above.
Statistical Analysis R software was used to determine significance using a general linear model with a Poisson distribution on graphs that display percentages, and a general linear model with a normal error distribution on graphs that display the number of foci.
Electron spectroscopic imaging. Fluorescent microscopy was used on 70 nm sections to identify APBs (PML and TRF2 present) and non-ALT associated PML nuclear bodies (PML NB; PML present, TRF2 absent). Silver enhancement of gold-tagged secondary antibodies was also used to confirm the presence of PML on nitrogen maps. Processing of cells and ESI imaging was carried out as previously described (96,97). For RNase digestion experiments, cells were embedded in hydrophilic LR white resin and sections were picked with non-reactive nickel grids. Nuclei of interest were identified with immunofluorescence imaging on sections as described above, and regions with 3 consecutive sections were chosen. Prior to ESI analysis the middle section digested overnight with RNase A (2mg/mL in 0.1M Tris pH 7.4, Fermentas), and the surrounding sections served as controls.
We wish to thank Dr. Alan D’Andrea for the pMMP-puro-EGFP-FANCD2 construct, Dr. Roel van Driel and Dr. Manual Buch wald for sharing antibodies, and Dr. Paul Bradshaw for constructive comments on the project. This work was supported by grants from the Canadian Cancer Society (#020472); and the Canadian Institute of Health Research.
Conflict of Interest statement. None declared.
This work was made possible, in part, by grants from the Canadian Institutes for Health Research to H.R and M.S.M. as well as grants to M.S.M. from the Canadian Cancer Society Research Institute (Grant #020472) and the National Sciences and Engineering Research Council of Canada (Grant #06222.) Additional resources were provided by a Natural Sciences and Engineering Research Council of Canada grant to D.P.B.-J. (grant number 14511) and the Canada Research Chair in Molecular and Cellular Imaging held by D.P.B.-J.