TERRA G-quadruplex RNA interaction with TRF2 GAR domain is required for telomere integrity

Telomere dysfunction causes chromosomal instability which is associated with many cancers and age-related diseases. The non-coding telomeric repeat-containing RNA (TERRA) forms a structural and regulatory component of the telomere that is implicated in telomere maintenance and chromosomal end protection. The basic N-terminal Gly/Arg-rich (GAR) domain of telomeric repeat-binding factor 2 (TRF2) can bind TERRA but the structural basis and significance of this interaction remains poorly understood. Here, we show that TRF2 GAR recognizes G-quadruplex features of TERRA. We show that small molecules that disrupt the TERRA-TRF2 GAR complex, such as N-methyl mesoporphyrin IX (NMM) or genetic deletion of TRF2 GAR domain, result in the loss of TERRA, and the induction of γH2AX-associated telomeric DNA damage associated with decreased telomere length, and increased telomere aberrations, including telomere fragility. Taken together, our data indicates that the G-quadruplex structure of TERRA is an important recognition element for TRF2 GAR domain and this interaction between TRF2 GAR and TERRA is essential to maintain telomere stability.

in living cells, we compared the ability of these compounds, along with the PP control, to alter the TRF2-TERRA interaction by RNA-ChIP assay (Fig. 4A,B). For these experiments, we used LOX human melanoma cells that are telomerase positive with long telomeres and high levels of TERRA 74 . For RNA-ChIP, we treated cells with 2 μM of each compound for 24 h. Both NMM and BRACO-19 inhibited TRF2 interaction with TERRA by ~ 50% relative to PP control (Fig. 4B). Neither NMM, PP, nor BRACO-19 had any significant effect on TRF2 binding to telomere DNA by ChIP-assay ( Fig. S6A and B). Remarkably, NMM and to a lesser extent BRACO-19 treatment reduced total TERRA levels relative to PP control as measured by RNA Dot blot (Fig. 4C,D) and Northern blot (Fig. 4E). Furthermore, NMM and BRACO-19 reduced TERRA levels in LOX and ALT positive U2OS cells ( Fig. S6C and D), indicating that the reduction in TERRA levels is not specific to cell type or telomerase expression status. We also found that PDS, which bound TERRA and TeloDNA G4 similar to BRACO-19, also reduced TERRA levels in LOX treated cells ( Fig. S6E and F). Northern blot indicated that TERRA levels were not only reduced by NMM, but accumulated as a shorter form, which was not observed with BRACO-19 or PP treatment (Fig. 4E). To determine if the nascent transcription of TERRA was affected, we treated LOX cells with either PP, NMM, or BRACO-19 for 48 h prior to pulse labeling with ethynyl uridine (EU) for 4 h after prior treatment with either PP, NMM, or BRACO-19 (Fig. 4F). EU labeled RNA was recovered by Click-iT biotin capture and then assayed by RT-qPCR for chromosome specific TERRA RNA. We found that both NMM and BRACO-19, but not PP led to a significant reduction (~ two fold) in nascent TERRA transcripts from chromosomes 10q, XYq and 18q (Fig. 4F). These findings suggest that G4 interacting molecules are inhibiting the nascent transcription of TERRA.
NMM induced telomere DNA shortening and fragility. We next assayed the effect of these G4 interacting molecules on telomere repeat DNA. Telomere repeat length and signal intensity were first measured by telomere restriction length assay (TLA). This revealed that NMM, and to a lesser extent BRACO-19 and PDS,   5A and S6G). NMM also led to a decrease in average telomere length, which was more significant after 3 days of treatment (Fig. 5A). To determine if G4 telomere DNA was affected by incubation with G4 interacting drugs, we performed ChIP assay with BG4 antibody that is known to be highly specific for G4 DNA, but not RNA 75,76 . We found that LOX cells treated with NMM had significant reduction in G4 telomere DNA relative to PP or PDS treated cells (Fig. S6H). These findings are consistent with the loss of telomere signal measured in TLA assay. We next assayed the effects of PP, NMM, and BRACO-19 on DNA damage signaling by γH2AX or 53BP1 immunofluorescence (IF) combined with telomere FISH. We noted that NMM induced a high level of γH2AX or 53BP1 associated telomere dysfunction-induced foci (TIFs), while BRACO-19 produced a moderate level, and PP was equivalent to background for LOX cells (Fig. 5B,C, and Fig. S7A and B). Scoring for γH2AX colocalization with telomere DNA signal revealed that NMM induced a ~ fivefold increase in telomere-associated γH2AX compared to PP treatment, whereas BRACO-19 induced a ~ three fold increase relative to PP treatment (Fig. 5B,C). We also assayed for telomere aberrations by metaphase chromosome FISH (Fig. 5D,E). We observed that both NMM and BRACO-19 treatment increased the frequency of telomere aberrations, including the appearance of fragile telomere doublets (Fig. 5D,E), which are typically associated with defects in telomere DNA replication 61,77,78 . TRF2 GAR domain is required for TERRA expression. To better understand the function of the TRF2 GAR domain and its potential role in mediating the telomeric effects of these G4 interacting drugs, we generated a stable LOX cell line with a doxycycline (Dox)-inducible TRF2ΔB gene. TRF2ΔB expression was readily detectable within 3 h after Dox-induction (Fig. S8C) and continuously expressed in cells treated for more than 2 weeks (Fig. 6A). The effect of TRF2ΔB on TERRA levels was analyzed by Dot blot (Fig. 6B,C), Northern blot (Fig. 6D)  Error bars indicate SD. Two tailed t test, *p value of < 0.05 relative to PP control. (C) LOX cells were treated with 2 μM PP, NMM or BRACO-19 for 3 days. Total RNA was isolated and assayed by dot blot using probes containing the (CCC TAA ) 4 (for TERRA), or 18S sequence. 3 μg of RNA was used for each sample and RNase A treatment (+) was used to assess possible DNA contamination. (D) Quantification of at least three independent RNA dot blot assays as represented in panel (C). Values are the means and SD (error bars). Two tailed t test, **p < 0.005, ***p < 0.001, relative to PP control. (E) Total RNA (6 μg) isolated from treated LOX cells from panel (C) were analyzed by Northern blot. RNA was detected with probe for TERRA or 18S, as indicated. Ethidium Bromide (EtBr) staining of Northern gel was shown in bottom panel to indicate the stability of total RNA. (F) LOX cells were treated with 2 μM PP, NMM or BRACO-19 for 2 days, and then pulse labeled with 0.2 mM EU for 4 h followed by RNA isolation, Click-iT biotinylation and streptavidin purification of nascent RNA. Nascent RNA was then quantified by RT-qPCR with primers specific for chromosome-specific TERRA from 10q, 13q, XYq, and 18q and normalized relative to U1 RNA. Values are the means and SD (error bars). Two tailed t test, **p < 0.05, ***p < 0.01, ns, not significant relative to PP control.  6E). All assays clearly indicated that TRF2ΔB caused an abrupt loss of TERRA expression, similar to that observed with NMM. Dot blot showed ~ tenfold decrease in total TERRA levels by 2 days after Dox-induction of TRF2ΔB and the decreased TERRA levels were not recovered during the time course Telomere length analysis of LOX cells treated with 2 μM PP, NMM or BRACO-19 for 2 and 3 days. Telomere length and relative amount of telomeric DNA was determined by restriction digestion of genomic DNA with AluI/MboI, followed by PFGE and Southern hybridization with a 32 P-labeled (TTA GGG ) 4 probe (middle panel) followed by a 32 P-labeled Alu probe (right panel www.nature.com/scientificreports/ ( Fig. 6B,C). Northern blot revealed a similar decrease in total abundance with the accumulation of smaller forms of TERRA by 10 days after Dox-induction (Fig. 6D). RNA-FISH at 3 days after Dox-induction confirmed the decrease in TERRA in intact cells (Fig. 6E). To rule out the possibility of an effect of doxycycline on TERRA levels, we assayed LOX and LOX TRF2ΔB cells cultured in the absence or presence of Dox for 2 days. RNA dot blot showed that Dox treatment had no effect on TERRA in LOX cells, indicating that the loss of TERRA depends on TRF2ΔB induction (Fig. S8A). In addition, the reduction of TERRA levels was detectable at 3 h after Dox-induction and showed progressively reduced levels across the 24 h time frame (Fig. S8B). In agreement with the previous finding that TRF2ΔB expression induces growth arrest phenotypes 79 , Western blot indicated the gradual increase of p53 and p21 protein along the course of TRF2ΔB induction, while control total histone H3 or H3K9me3 showed no obvious change (Fig. S8C). To determine if the effect of TRF2ΔB on TERRA levels was not cell type-specific, we assayed TERRA levels by Dot blot (Fig. S8D) and Northern blot (Fig. S8F) in U2OS cells containing the Dox-inducible TRF2ΔB gene. Western blot showed clear TRF2ΔB expression after Dox treatment in U2OS TRF2ΔB cells (Fig. S8E). Consistent with our findings in LOX TRF2ΔB cells, both RNA dot blot and Northern blot revealed that TRF2ΔB induction resulted in a time-dependent decrease of TERRA signals in U2OS cells. To determine if this decrease in TERRA was due to changes in nascent transcription, we induced TRF2ΔB prior to pulse labeling with EU followed by Click-iT affinity purification and RT-qPCR for analysis of chromosome specific telomere transcripts (Fig. 6F). Similar to results with NMM treatment, we found that TRF2ΔB expression led to a loss of nascent TERRA RNA from all chromosomes tested (Fig. 6F). Taken together, these data indicate that TRF2ΔB induces a loss of TERRA nascent transcription that is not dependent on cell type or telomere maintenance mechanism.

Scientific Reports
TRF2ΔB induces telomere DNA loss and DNA-damage foci. TRF2ΔB expression also led to a substantial loss of telomere repeat DNA signal and decrease in average length, as measured by pulse field gel analysis of Southern blot (Fig. 6G). Telomere signal loss and average length reduction was apparent by day 7 after Dox induction and decreased further through day 27. In addition, LOX cells expressing TRF2ΔB showed increase in γH2AX signals colocalized with telomere foci (Fig. 6H,I). Quantification of these foci indicated ~ 2.5-fold increase in cells with ≥ 5 TIFs by 5 days after Dox-induction. We also assayed for telomere aberrations by metaphase chromosome FISH in LOX TRF2ΔB cells treated with Dox for 20 days (Fig. S8G). We found that TRF2ΔB induction increased the frequency of telomere aberrations, including the appearance of fragile telomeres and loss of telomere signals 80 . Taken together, these findings indicate that TRF2ΔB phenocopies many of the telomere and TERRA defects observed with NMM treatment.

Discussion
G4 DNA and RNA have been implicated in the regulation of diverse cellular processes, as well as potential for pathogenic roles in human disease 27,[81][82][83][84] . TERRA can form G4 structures, but the biological role of these G4 structures are not well-understood. Here, we show that the G4 structure of TERRA is recognized by the TRF2 basic GAR domain, and that this interaction is important for TERRA stability and telomere DNA maintenance 61 . We also found that the G4-interacting compound NMM has selectivity towards TERRA RNA and inhibits the    85 . Our FP assays showed that NMM preferentially bound G4 TERRA relative to G4 Tel-oDNA (Fig. 1). Both NMM and BRACO-19 showed similar affinities for G4 TeloDNA, but only NMM showed preferential binding to TERRA (Fig. 1). NMM is reported to have high selectivity to G4 DNA over other DNA structures and bind selectively to parallel forms of G4 DNA 54,57,86 . TERRA is reported to form a parallel stranded G4 32,87 , and also to have a high affinity for related porphyrins, such as TMPyP4 88 . In contrast, BRACO-19 and PDS bound G4 DNA preferentially over TERRA, indicating that both G4 RNA and DNA have different affinities for these small molecules. Our FP data also demonstrated that TRF2 GAR peptide binds preferentially to G4 TERRA relative to G4 TeloDNA (Fig. 2). It was reported that the GAR motif facilitates the folding of G4 telomeric DNA 89,90 . The G4 structure of TERRA was shown important for its binding to TRF2 64 . Therefore, the recognition of the TERRA G4 structure by TRF2 GAR is likely to be an important functional interaction in telomere organization and regulation.
G4 interacting molecules are known to induce telomere-specific defects in vivo 46,91,92 . Others have reported that BRACO-19, PDS, and NMM can inhibit telomerase activity through stabilization of telomeric G4 DNA 49,50,70,93 . We found NMM, PDS, and BRACO-19 cause the rapid loss of telomere repeat DNA (Fig. 4 and S6) and a corresponding increase in telomere DNA damage signal marked by colocalization with γH2AX and fragile telomere doublets (Fig. 5). While both NMM and BRACO-19 had similar activity, we observed more potent inhibition of TRF2 interaction with TERRA by RNA-ChIP (Fig. 4A,B) and a more robust loss of TERRA (Fig. 4C-E) with NMM than BRACO-19. NMM also showed a more rapid and extensive loss of telomere repeat DNA (Fig. 5A). While inhibition of telomerase may also contribute to loss of telomere signal, the rapid nature of these effects indicate they are most likely due to perturbation of telomere structure and DNA replication defects. We suggest that the specific effect of NMM on TERRA levels is most consistent with NMM's selective binding to TERRA and disruption of its interaction with the TRF2 GAR domain.
To better understand the effects of NMM on the TRF2 GAR domain interaction with TERRA, we compared NMM effects with those of ectopic expression of TRF2ΔB (Fig. 6), We found that TRF2ΔB expression phenocopied many of the observed effects of NMM, including the rapid loss of TERRA, loss of telomere DNA length and signal, and increase in telomere DNA damage signaling, including fragile telomeres. The rapid loss of telomeric DNA and the appearance of fragile telomeres supports the model that the TRF2 GAR domain is important for the completion of telomere DNA replication. We have previously reported that TRF2 GAR is required for TERRA binding to recruit ORCs to telomeres and facilitate telomere DNA replication and stability 61 and more recently shown a direct role of TRF2 GAR in initiation of DNA replication within telomere repeat DNA 80 . Both TRF2 GAR and TERRA have been implicated in telomere DNA replication, as well as DNA conformation and heterochromatin formation 27,94-96 . Our new findings suggest that disruption of TRF2 GAR interaction with TERRA leads to a rapid loss of TERRA, followed by the subsequent disruption of telomere replication and consequent telomere repeat loss and DNA damage signaling.
TRF2 has multiple domains with distinct activities. Other studies have shown ectopic expression TRF2 dominant negative homodimerization domain, as well as depletion of TRF2, reduced TERRA expression 18 , consistent with our findings that TRF2 is required for TERRA expression. The TRF2 GAR domain has been shown to bind the 4-way junctions formed at T-loops and restrict promiscuous telomere recombination 96 . TRF2 was also shown to inhibit telomerase expression in normal cells via its interaction with G4 structures in the TERT promoter, although it is not clear if the GAR domain was required for this activity 97 . TRF2 may interact with G4 DNA through domains other than GAR, and the GAR domain is known to have functions in addition to TERRA binding. Another study found that another G4 interacting molecule CK1-14 bound TERRA and disrupted TRF2 binding to telomere repeat DNA 65 . In contrast, we found that NMM did not disrupt TRF2 binding to telomeric DNA, but did inhibit TRF2 binding to TERRA. We also found that NMM and BRACO-19 inhibit the nascent transcription of TERRA. How these multiple interactions and functions are coordinated to regulate telomere structure and replication will require further investigation.
In conclusion, we find that G4 interacting molecules can have selectivity for different G4 structures, including selectivity for TERRA relative to telomere G4 DNA. These and newer generation G4 interacting molecules may be useful as probes to better dissect functions of telomere G4 structures in vivo. A recent study with a different class of G4 interacting molecule was found to bind selectively to TERRA and allosterically inhibit TRF2 binding to telomere DNA to promote apoptosis in cancer cells 65 . Furthermore, small molecule inhibition of TERRA is being explored for cancer treatment 21,27,98 . Our findings suggest that the TERRA interaction with TRF2 GAR is responsive to small molecule inhibition, and may have potential utility for telomere-based therapies.

Materials and methods
Oligos and chemicals. Nucleic acids of TERRA (UUA GGG ) 4   Fluorescence polarization (FP) assay. All the nucleic acids, peptides and chemicals were dissolved in buffer containing 20 mM HEPES pH7.5, 150 mM salt (KCl, NaCl, LiCl) at the stock concentration of 1 mM. For peptide and nucleic acid interactions, 25μL of each nucleic acids with a sequential twofold dilutions starting from 50 μM was added into each well of 96-well plate followed by adding 25μL of 10 nM FITC labeled TRF2 peptide. For compounds and nucleic acid interactions, 20μL of each compounds with a sequential twofold dilutions starting from 50 μM were added into each well of 96-well plate followed by adding 5μL of 20 ng/mL tRNA and 25μL of 10 nM FAM labeled nucleic acids. For inhibition assay, 5μL FITC labeled TRF2 GAR peptide (10 nM final concentration) and 10 μL oligos (5 μM final concentration) were mixed, followed by adding 5μL of each compound with a sequential twofold dilutions starting from 50 μM into each well of 384-well plate. All components were mixed thoroughly, and the plate was incubated at room temperature for 1 h before placing in an Envision Plate Reader for reading at 495/520 nm. All data curves were fitted using Prism 8.0.
Telomere and TERRA analysis. TERRA dot blot and Northern blot analyses were described previously 61 .
Telomere length analysis, telomere IF and metaphase FISH were described previously 61,102 . Additional details are provided in Supplemental Information.

Statistics.
Statistical analyses were carried out by paired two-tail Student's t tests. p values and significance levels are annotated in the figures and described in the figure legends.