DNA-PK controls Apollo’s access to leading-end telomeres

Abstract The complex formed by Ku70/80 and DNA-PKcs (DNA-PK) promotes the synapsis and the joining of double strand breaks (DSBs) during canonical non-homologous end joining (c-NHEJ). In c-NHEJ during V(D)J recombination, DNA-PK promotes the processing of the ends and the opening of the DNA hairpins by recruiting and/or activating the nuclease Artemis/DCLRE1C/SNM1C. Paradoxically, DNA-PK is also required to prevent the fusions of newly replicated leading-end telomeres. Here, we describe the role for DNA-PK in controlling Apollo/DCLRE1B/SNM1B, the nuclease that resects leading-end telomeres. We show that the telomeric function of Apollo requires DNA-PKcs’s kinase activity and the binding of Apollo to DNA-PK. Furthermore, AlphaFold-Multimer predicts that Apollo’s nuclease domain has extensive additional interactions with DNA-PKcs, and comparison to the cryo-EM structure of Artemis bound to DNA-PK phosphorylated on the ABCDE/Thr2609 cluster suggests that DNA-PK can similarly grant Apollo access to the DNA end. In agreement, the telomeric function of DNA-PK requires the ABCDE/Thr2609 cluster. These data reveal that resection of leading-end telomeres is regulated by DNA-PK through its binding to Apollo and its (auto)phosphorylation-dependent positioning of Apollo at the DNA end, analogous but not identical to DNA-PK dependent regulation of Artemis at hairpins.


DNA-PK P G G G G G G G G G C C C C C C C C C G G G G G G G G G G G G C C C C C C C C C C C C G G G G G G G G G G G G G G G G G G G G G G G G G G G C C C C C C C C C C C C
Leading-end telomeres

Introduction
DNA-PK comprises a 470-kD kinase subunit (DNA-PKcs) and the heterodimeric Ku70 / Ku80 DNA-end binding factor.DNA-PK's primary function is to mediate canonical nonhomologous end joining (c-NHEJ) of double strand breaks (DSBs) by recruiting DNA ligase IV (Lig4), its partner X-ray repair cross complementing 4 (XR CC4), and XR CC4-like factor (XLF) to DNA ends (reviewed in ( 1 ,2 )).Recent biochemical, single-molecule and cryo-EM structural studies have revealed key mechanistic details of c-NHEJ (3)(4)(5)(6).Following loading of Ku70 / 80 onto the two DNA ends, one copy of DNA-PKcs binds to each Ku heterodimer and promotes the synapsis of the ends into what is termed the XLF-mediated long-range (LR) synaptic complex.This LR complex stabilizes the association of Lig4 / XRCC4 and XLF with DNA-PK, but the DNA ends are inaccessible and not aligned, requiring the transition to the short-range (SR) complex for ligation to take place.Although the details still need to be elucidated, the transition from XLF-mediated LR complex to SR complex is initiated by the autophosphorylation in trans of DNA-PKcs on a cluster of S / T-Q sites called the ABCDE / Thr2609 cluster, which induces the conformational changes required for the release of the kinase from the DNA ends and subsequent ligation of the two strands ( 3 , 5 , 7-10 ).The autophosphorylation in trans of a second cluster of S / T, the PQR / Ser2056 cluster, has also been shown to occur in the XLF-mediated LR complex to limit end processing ( 3 , 5 , 7 , 8 ).Another LR synaptic complex, the Ku-mediated domainswap dimer, is instead formed by the interaction in trans between Ku80 and DNA-PKcs binding the two opposite DNA ends ( 4 ).This complex is essential for the processing of DNA ends that are not directly ligatable, such as closed hairpins formed during V(D)J recombination by RAG recombinase.Such hairpins are opened by Artemis, an endonuclease / 5exonuclease regulated by DNA-PK (11)(12)(13)(14)(15). How DNA-PK promotes Artemis's access to the DNA ends and stimulates its nuclease activity is still an object of investigation.However, recent studies suggest that, first, a short (50 residues) region in the unstructured C-terminal tail of Artemis binds the FAT domain of DNA-PK ( 7 , 16 , 17 ).In this conformation, Artemis's catalytic domain may face outwards, with no access to the DNA.After phosphorylation of the ABCDE / Thr2609 cluster either by DNA-PKcs itself (in cis or in trans ) or by the DNA damage kinase ATM, DNA-PK undergoes a conformational change allowing the Artemis nuclease domains to access the DNA ends inside the DNA-binding groove of DNA-PK and to cleave them ( 7 , 15 , 17-20 ).DNA-PK then promotes the ligation of the ends after the transitions to the XLF-mediated LR and, subsequently, to the SR complex.However, since Artemis blocks DNA-PK in a self-inhibited conformation ( 7 ), the molecular mechanisms underlying DNA-PK (re)activation are still not clear.Furthermore, other end-processing factors, such as the Werner syndrome RecQ helicase (WRN) and the Mre11 / Rad50 / Nbs1 (MRN) complex, can also be activated by DNA-PKcs, but the underlying mechanism is less clear ( 21 ,22 ).
These distinct and context-specific roles of DNA-PK at DNA ends depend on multiple layers of regulation: the presence of both Ku70 / 80 and DNA-PKcs are required for end protection; DNA-PKcs autophosphorylation on the PQR / Ser2056 cluster is needed for c-NHEJ; and DNA-PKcs (auto)phosphorylation on ABDCE / Thr2609 cluster is necessary for Artemis recruitment and hairpin-opening as well as for the release of DNA-PK from the ends and ligation.This functional complexity could explain the differences observed in c-NHEJ and V(D)J recombination when DNA-PK is absent or inhibited ( 18 , 20 , 23-26 ).
The joining of telomeres through c-NHEJ is repressed by TRF2 (Telomeric repeat-binding factor 2), one of the subunits of the telomeric shelterin complex ( 27 ).Deletion of TRF2 results in extensive telomere-telomere fusions mediated by Ku70 / 80 and Lig4 (28)(29)(30)(31)(32)(33).TRF2 is thought to block c-NHEJ by promoting the formation of t-loops, which hide the end of the chromosomes (34)(35)(36), as well as through other mechanisms (37)(38)(39).Essential for the protection of telomeres is the single-stranded G-rich 3 -overhang which invades the telomeric duplex to form the t-loop.While the 3 -overhang is thought to be a natural replication product at lagging-end telomeres, the newly synthesized leading-end telomeres should have (near) blunt ends and need to be processed.Formation of a 3 overhang at leading-end telomeres is mediated by the 5 nuclease activity of Apollo, which is brought to telomeres in S phase by TRF2 (40)(41)(42)(43)(44)(45)(46)(47).Defects in Apollo recruitment and / or nuclease activity cause activation of ATM, reduction in the leading-end telomeric overhang and leading-end telomere fusions mediated by alternative EJ (alt-EJ) ( 42 , 44 , 46-50 ).In G1, Nbs1 competes for Apollo binding to TRF2, whereas in G2, hyper-resection by Apollo is counteracted by POT1 once a short overhang has been formed ( 51 ,52 ).However, it is not known if TRF2 recruitment alone is sufficient for Apollo to act right after replication and / or if there are additional layers of regulation.
Paradoxically for a factor involved in the re-joining of broken DNA ends, DNA-PK is present at telomeres, but its telomeric function has remained enigmatic ( 53 ,54 ).On one hand, DNA-PK deletion does not cause overt telomere deprotection except for occasional telomere fusions (55)(56)(57)(58)(59)(60)(61).On the other hand, chemical inhibition of DNA-PKcs causes fusions of newly replicated leading-end telomeres ( 55 , 62 , 63 ), and cells expressing a DNA-PKcs mutant lacking three phosphorylation sites in the ABCDE / Thr2609 cluster showed activation of the DNA damage response at leading-end telomeres in mitosis and a general increase in telomere fusion events ( 64 ).
Here, we demonstrate that kinase activity of DNA-PK and Apollo act in the same pathway to promote 3 -overhang formation and protection of leading-end telomeres from alt-EJmediated fusions.Accumulation of Apollo at telomeres depends solely on TRF2 and does not require DNA-PK or its kinase activity.However, we identify a domain in Apollo (amino acids 344-360) that promotes Apollo's interaction with DNA-PK and is required for Apollo function at telomeres.This DNA-PK binding motif is similar to that of Artemis.Furthermore, DNA-PK activity and phosphorylation of DNA-PK on the ABCDE / Thr2609 cluster is required for Apollo's function at telomeres.Alpha-Fold multimer version AFsample predicts that Apollo has extensive interactions with DNA-PK that would position the nuclease active site at the DNA end in a manner that depends on phosphorylation on the ABCDE / Thr2609 cluster.Therefore, we propose that DNA-PK activation at telomeres is required to grant Apollo access to the newly replicated leading-end telomeres in a timely manner.

Viral gene delivery
For retro-or lentiviral transduction, a total of 20 μg of plasmid DNA was transfected into Phoenix eco or 293T cells, respectively, using CaPO 4 precipitation.The viral supernatant was filtered through a 0.45μm filter, supplemented with 4 μg / ml polybrene, and used for the transduction of target cells.Cre-mediated deletions were achieved with three infections / day (6-12 h intervals) over 2 days with pMMP Hit & Run Cre retrovirus produced in Phoenix eco cells.Timepoint 0 was set 12 h after the first Hit & Run Cre infection.For stable expression of HA-Apollo, six infections at 6-12 h intervals were performed before selection for 2-4 days in 2-4 μM Puromycin before Hit & Run Cre transduction.
HT1080 pictures were acquired on Leica DMi8 with a 63 ×/ 1.2 objective using LAS X software.2D-maximum intensity projection images were obtained using Fiji software ( 67 ).All other pictures were acquired on a DeltaVision RT microscope system (Applied Precision) with a PlanApo 60 × 1.40 NA objective lens (Olympus America, Inc.) at 1 × 1 binning and multiple 0.2 μm Z-stacks using SoftWoRx software.Images were deconvolved, and 2D-maximum intensity projection images were obtained using SoftWoRx software.Chromatid and chromosome-type fusions were analyzed using Fiji software ( 67 ) after arbitrary assignment of red for both (TTAGGG) 3 -probes and green for both (CCCTAA) 3probes.Semi-automated analysis and quantification of colocalization was performed using CellProfiler ( 68 ) with the following pipeline: image cropping to remove edge artifacts due to deconvolution; channel intensity rescaling to cover the full histogram range value; 'speckle features enhancement' with a size cut-off of 5 pixels to increase detection sensitivity and remove background / artifact aggregates; 'channel-wise primary objects identification' to detect individual nuclei (diameter range: 50-1000 pixels, threshold correction: 1.5) and individual foci (diameter range: 2-15 pixels, threshold correction: 1-4); correlation of the foci coordinates in the different channels and with the respective nuclei to define colocalization events.Nuclei with < 10 or 20 detected PNA foci were discarded for HA-Apollo and γ-H2AX analysis, respectively.
In-gel analysis of single-stranded telomeric DNA Telomeric DNA was analyzed on Clamped homogenous electric field (CHEF) gels as previously described ( 29 ).Briefly, cells were harvested by trypsinization, resuspended in PBS, mixed with 2% agarose (1:1 ratio) at 50 • C and cast in a plug mold 0.7-1 × 10 6 cells / plug.Plugs were digested overnight at 50 • C in 1 mg / ml proteinase K (03115879001; Roche) in digest buffer (100 mM EDTA, 0.2% sodium deoxycholate, and 1% sodium lauryl sarcosine) and washed five times in TE.After DNA digestion overnight at 37 • C by 60U MboI (#R0147; New England BioLabs), plugs were then washed in 0.5 × TBE twice before loading on a 1% agarose / 0.5 ×TBE gel.DNA was resolved by a CHEF-DRII PFGE apparatus (Bio-Rad) for 20 h, with the following settings: initial pulse, 5 s; final pulse, 5 s; 6 V / cm at 14 • C. Gel was dried and hybridized overnight at 50 • C with γ-32 P-ATP end-labeled TelC (AACCCT) 4 probe in Church mix (0.5 M sodium phosphate buffer pH 7.2, 1 mM EDTA, 7% SDS, 1% BSA).After three washes in 4 × SSC and one in 4 × SSC / 0.1% SDS at 55 • C, the gel was exposed for one / two days before acquisition of the single-stranded telomere signal by PhosphoImager.The gel was then denatured with 1.5 M NaCl / 0.5 M NaOH for one hour, neutralized with two washes of 1 h each in 0.5 M Tris-HCl pH 7.0 / 3 M NaCl, prehybridized for 30 min at 55 • C in Church mix, and hybridized overnight at 55 • C before being washed and exposed as described before for the acquisition of the total telomere signal.Quantification of the signals in each lane was done using Im-ageQuant software.After subtraction of the background, the single-stranded signal was normalized to the total telomeric DNA signal in the same lane.The indicated control value was set to 1, and all the other values were given as a percentage of this value.

Coimmunoprecipitation
For Myc-TRF2 pulldown with HA-Apollo, 293T cells were plated in 10-cm dishes 24 h prior to transfection by CaPO 4 precipitation using 10 μg of each indicated plasmid.The medium was changed 12 h after transfection with fresh medium containing DMSO or 12.5 μM NU7441 for 24 h.Cells were harvested using a cell scraper in cold PBS at 4 For DNA-PK pulldown with HA-Apollo, 20 μg of plasmid was used to transfect 293T cells plated in 10 cm dishes 24 h prior to transfection by CaPO 4 precipitation.Cells were harvested 40 h after transfection as previously described.Co-IP was performed as previously described for Artemis with minor modifications ( 14 ).All steps were performed at 4 • C. Briefly, cells were lysed in 0.5 ml lysis buffer (25 mM HEPES pH 7.4, 150 mM KCl, 10 mM MgCl 2 , 10% glycerol, 2 mM DTT, 1 mM EDTA, 8 mM β-mercaptoethanol) supplemented with complete protease inhibitor mix and phosphatase inhibitor and kept 10 min on ice before sonication at 40% amplitude for 30 s; 5 s ON and 5 s OFF.After centrifugation at maximum speed for 30 min, the lysate was pre-cleared with protein-G Sepharose beads in lysis buffer (protein-G Sepharose 4 fast flow; 17061801 Cytiva) for 1 h followed by overnight incubation with anti-HA antibody at 4 • C on an end-to-end nutator.Samples were then incubated with protein-G Sepharose beads pre-blocked in 0.2%BSA / PBS for 4 h at 4 • C on an end-to-end nutator and washed 4 × 5 min with lysis buffer on an endto-end nutator.The immunoprecipitated proteins were then eluted with 40 μl of 2 × Laemmli buffer and boiled for 5 min before separation by SDS-PAGE.

Structural modeling
AFsample ( 69 ), an improved version of AlphaFold-Multimer ( 70 ) deemed the best method in CASP15 ( 71 ) for multimer assembly, was used to predict the interaction of mouse Apollo with mouse DNA-PKcs amino acids 1981-2760.Multiple sequence alignments (MSAs) were constructed using the -reduce_db option, resulting in 5666 and 1070 homologous sequences in the Apollo and DNA-PKcs, MSAs, respectively.Templates to known PDB structures were not used to avoid any bias to different conformational states and experimental conditions.In total, 816 models of Apollo:DNA-PKcs were generated using multimer_v1 neural network weights withdropout enabled at inference to get more diversity in the pool of sampled models.The model with the highest reported ranking confidence was relaxed in the Amber99sb force field ( 72 ) using openMM ( 73 ), and used as a representative model of the interaction.

Sequence alignments
Apollo, Artemis and DNA-PKcs sequences were obtained from Uniprot ( https://www.uniprot.org).Sequences were aligned in SnapGene (GSL Biotech) using Clustal Omega or MUSCLE multiple sequence comparison.Alignments were colored either compared to mouse reference sequence or with Zappo coloring (physico-chemical properties).

Quantification and statistical analysis
Quantification and statistical analysis were performed Graph-Pad on three or more independent experiments, as indicated.For CO-FISH analysis, at least ten metaphases per condition were scored for each experiment.For IF-FISH analysis, > 80 nuclei per condition per experiment were scored.For overhang analysis, the normalized overhang signal was expressed relative to the untreated control and independently for each cell line.Significance was assessed by calculating the p-value using non-parametric Kruskal-Wallis ANOVA for multiple comparisons or Mann-Whitney t -test for single cell analysis and one-or two-way ANOVA for multiple comparisons for population analysis.P -values ≤ 0.05 were considered statistically significant.

DNA-PKcs kinase activity and Apollo act in the same pathway to prevent alt-EJ of leading-end telomeres
Previous reports showed that DNA-PK inhibitors IC86621 and NU7026 induce leading-end telomere fusions as does deletion of Apollo, and the severity of the telomere fusion phenotype is similar in these two settings ( 42 , 46 , 48 , 63 , 74 , 75 ).Therefore, we explored the possibility of an epistatic relationship between DNA-PK and Apollo.As expected, we observed leading-end telomere fusions at similar frequencies both in MEFs where Apollo was deleted with Cre, and MEFs treated with NU7441 ( 76 ), one of the most potent and selective DNA-PKcs inhibitor available (hereafter referred to as DNA-PKi) (Figure 1 A,B).DNA-PKi did not exacerbate the leading-end telomere fusion phenotype of Apollo-deficient cells (Figure 1 A,B), suggesting that DNA-PK and Apollo act in the same pathway to protect leading-end telomeres.
Importantly, DNA-PKi reduced the 3 -overhang signals to approximately the same extent as the deletion of Apollo, and the overhang phenotype of Apollo deletion was not exacerbated by DNA-PKi (Figure 1 C,D).Furthermore, DNA-PKi induced a DNA damage response at telomeres similar to Apollo deletion, as evident from the appearance of γ-H2AX-marked telomere dysfunction-induced foci (TIFs), although it did not induce the phosphorylation of Chk2 (P-Chk2) ( Supplementary Figure S1 A-C).Finally, similar to our recent data on Apollo deletion ( 50 ), chromatid fusions induced by DNA-PKi were abolished in the absence of alt-EJ factors Ligase 3 (Lig3) and PolQ ( Supplementary Figure S2 A,  B), but not of c-NHEJ Lig4 ( Supplementary Figure S2 C,D), pointing to alt-EJ as the predominant mechanism promoting telomere fusions in cells with inhibited DNA-PK activity.The reason for the lack of Chk2 phosphorylation upon DNA-PK inhibition is unclear, and we cannot explain the difference with prior reports showing that the other DNA-PKi NU7026, causes telomere fusions mostly dependent on c-NHEJ ( 74 ,75 ).Nevertheless, our data suggest that DNA-PK kinase activity is required for Apollo to form the 3 -overhang and to protect the leading-end telomeres from fusion.
Unexpectedly, we observed that DNA-PKi had an off-target effect on the stability of HA-tagged Apollo expressed through retroviral transduction ( Supplementary Figure S3 ).DNA-PKi strongly reduced the levels of HA-Apollo in a time and dosedependent manner ( Supplementary Figure S3 A,B).Remarkably, the effect of DNA-PKi was independent of the presence of DNA-PK.DNA-PKi-dependent degradation of HA-Apollo was comparable in MEFs expressing or lacking the DNA-PK components DNA-PKcs and / or Ku70, and in MEFs expressing the catalytically inactive DNA-PKcs-D3922A mutant (DNA-PKcs KD / KD ) ( 20), ( Supplementary Figure S3 C,D).This effect could also be observed with other DNA-PK inhibitors (NU7026 and IC86621), even at low concentrations that do not trigger chromatid fusions ( 63 , 74 , 75 ) ( Supplementary Figure S3 E,F).The effect of DNA-PKi was attenuated by Bortezomib ( Supplementary Figure S3 G), indicating that it is, at least partially, due to proteasomal degradation.Since mouse endogenous Apollo could not be detected using commercial antibodies, we could not test whether DNA-PKi also affects the levels of endogenous Apollo.
Given these off-target effects of DNA-PKi in mouse cells, we tested the effect of DNA-PK inhibition in the HT1080 human cancer cell line.In HT1080 cells, DNA-PKi caused only a minimal reduction in endogenous Apollo (Figure 2 A).Nevertheless, DNA-PKi caused similar levels of leading-end telomere fusions as CRISPR / Cas9 deletion of Apollo (Figure 2 A-C).Importantly, the effect of DNA-PKi and Apollo deletion were not additive (Figure 2 C), arguing that the epistatic relationship between Apollo and DNA-PK kinase activity is not due to DNA-PKi inducing the degradation of the endogenous Apollo.Finally, leading-end telomere fusions were also observed in DNA-PKcs KD / KD MEFs, independently of DNA-PKi treatment ( Supplementary Figure 4 A,B) or the presence of an sgRNA targeting mouse Apollo (Figure 2 D,E).These data confirm that DNA-PK kinase activity is required for Apollo to form the 3 -overhang and to protect the leading-end telomeres from fusion.
The target of DNA-PK kinase activity is DNA-PK, not Apollo's S / T-Q sites or recruitment mechanism A possible explanation for the epistatic relationship between Apollo and DNA-PK is that DNA-PK promotes Apollo localization at telomeres together with TRF2.We addressed this question by monitoring the localization of HA-Apollo to telomeres by immunofluorescence (IF).As expected, HA-Apollo was detected at telomeres in cells expressing TRF2 but not in TRF2 F / F Lig4 −/ − MEFs after TRF2 deletion with Cre (Figure 3 A-C).By contrast, the telomeric localization of HA-Apollo appeared unaffected in cells that either lacked DNA-PK or carried the kinase-dead allele of DNA-PKcs (Figure 3 D-F).Consistent with DNA-PK not being required for the TRF2-mediated recruitment of Apollo to telomeres, coimmunoprecipitation (co-IP) experiments showed that the interaction of HA-Apollo and Myc-TRF2 was not affected by DNA-PKi (Figure 3 G).
We also excluded that the effect of DNA-PKi involves the phosphorylation of H2AX or ATM, both known targets of DNA-PKcs kinase activity (77)(78)(79).In fact, DNA-PKi treatment in H2AX or ATM null cells resulted in the reduction of the 3 -overhang signal and the increase in leading-end telomere fusions similar to the control cells ( Supplementary Figure S5 A-D), consistent with previous reports ( 63 ,74 ).
Mouse Apollo contains two S / T-Q sites (T162 and T350) that are considered canonical targets of DNA-PK.Although they have never been shown to be phosphorylated and only one of them is conserved in most mammals ( Supplementary Figure S6 A), we investigated their role in telomere protection.However, an Apollo mutant where both threonine residues are mutated to alanines (Apollo-AA) was indistinguishable from wild-type Apollo (Apollo-WT) in its recruitment at telomeres ( Supplementary Figure S6 B-D), and in the maintenance of the 3 -overhang and the protection of telomeres from fusion (Figure 4 A,B and Supplementary Figure S6 E,F) when expressed in Cre-treated Apollo F / F .Therefore, S / T-Q phosphorylation of Apollo is not required for telomere end-processing.
Finally, we tested whether (auto)phosphorylation of DNA-PKcs is required for the telomeric function of DNA-PK, since MEFs carrying a DNA-PKcs mutant with three threonines out of the five phosphorylation sites in the mouse ABCDE / Thr2609 cluster changed to alanine (DNA-PK 3A / 3A ) have high levels of telomere fusions ( 64 ).Indeed, DNA-PK 3A / 3A MEFs confirmed the fusion phenotype, specifically involving leading-end telomeres (Figure 4 C,D).DNA-PKi treatment increased the leading-end telomere fusion events in DNA-PK 3A / 3A MEFs, but not more than what is observed in the DNA-PK + / + control cells (Figure 4 C,D).Although it is possible that other substrates of DNA-PK and / or other DNA- PKcs phosphorylation sites are involved, this result suggests that, in the context of telomere processing, the relevant target of DNA-PK kinase activity is the ABCDE / Thr2609 cluster of DNA-PKcs.
Phosphorylation on the DNA-PK ABCDE / Thr2609 cluster is predicted to provide Apollo access to DNA ends Autophosphorylation of the ABCDE / Thr2609 opens the DNA-PK groove and exposes the DNA end to Artemis ( 7 ).To investigate the possibility that phosphorylation of DNA-PK could similarly grant Apollo access to the ends, we used AFsample ( 69 ), an improved version of AlphaFold-Multimer ( 70 ) to query potential interactions between full-length Apollo and part of the circular cradle domain (residues 1981-2760) of DNA-PK.This part of DNA-PK contains the PQR / Ser2056 and ABCDE / Thr2609 clusters and interacts with Artemis after autophosphorylation ( 7 ) (Figure 5 A).In total, 816 poten-tial interactions were sampled, and the five highest ranking based on the reported ranking confidence were selected as a representative interaction mode ( Supplementary Figure S7 A).The N-terminal part (residues 1-344) of Apollo and residues 2140-2540 of DNA-PKcs were predicted with high confidence (pLLDT > 0.8) (Figure 5 B-D), while the C-terminal half of Apollo was predicted with low confidence and was removed from further analysis.As expected, the predicted structure of Apollo overlapped substantially (RMSD: 0.692 Å) with the crystal structure of human Apollo (PDB: 5AHO) ( 80 ) with the exception of residues 314-344, which were not resolved in the crystal structure ( Supplementary Figure S7 B).Since these 30 residues are lacking in the crystal structure, it is likely that they are flexible and could potentially be interacting with other regions of DNA-PK not present in the model, as discussed below.
Importantly, the interaction between the catalytic domain of Apollo with DNA-PKcs was predicted with high confidence, indicated by the low predicted alignment error be-   S7 C).Furthermore, in this latter conformation, the inferred interaction between the N-terminal of Apollo and DNA-PK allowed juxtaposition of the catalytic core of Apollo with the DNA end and of Apollo (residues 1-312) with Artemis (residues 1-361; PBD: 7SGL ( 7)), with a calculated RMSD of 2.304 Å and a near identical positioning of the catalytic residues (Figure 5 F,G).These findings support the hypothesis that autophosphorylation of DNA-PK on the ABCDE / Thr2609 cluster promotes the conformational changes required to grant Apollo access to the telomere ends.

Interaction between DNA-PKcs and the C-terminal region of Apollo is required for telomere protection
In addition to the interaction between DNA-PK and Artemis discussed above, Artemis binds to DNA-PK through a short motif just C-terminal of the β-CASP (metalloβ-lactamaseassociated CPSF / Artemis / SNM1 / PSO2) domain in a region of the protein that is largely disordered ( 7 , 16 , 17 ).The predicted structures of full-length Apollo did not show any similarity to Artemis beyond the β-CASP region (Figure 5 F), and this region was not resolved in the X-ray structure of Apollo ( 80 ).However, our sequence alignments identified a short region in Apollo highly conserved in mammals with substantial similarity to the DNA-PK binding site in Artemis (Figure 6 A,B and Supplementary Figure S8 A,B).Indeed, endogenous DNA-PKcs was found to coimmunoprecipitate with HA-Apollo transfected into 293T cells, suggesting a direct interaction (Figure 6 C).Importantly, a version of Apollo with a deletion of amino acids 344-360 (Apollo-PK), removing the  residues most similar to those of the Artemis DNA-PK interaction motif, had a reduced interaction with DNA-PKcs in 293T cells (Figure 6 C).
We next examined the functional consequences of removing these residues from Apollo.Apollo-PK interacted normally with TRF2 and localized to mouse telomeres at the same level as wild-type Apollo (Figure 6 C-F).However, despite its localization to telomeres, Apollo-PK failed to pre-vent telomere fusions when expressed in Apollo-deficient cells (Figure 6 G,H).Although we cannot rule out that the deletion of amino acids 344-360 directly compromises Apollo's catalytic activity, it is important to note that Apollo does not require its C-terminus, including the DNA-PKcs binding region, to process DNA ends in vitro ( 80 ,81 ).Therefore, these data suggest that direct interaction between DNA-PK and Apollo is required for the protection of leading-end telomeres.

Discussion
Our data show that DNA-PK is required for the Apollodependent processing of leading-end telomeres, explaining the telomere fusion phenotype associated with DNA-PKcs inhibition and providing a rationale for the presence of DNA-PK at telomeres.Therefore, we propose a model (Figure 7 ) in which Apollo is recruited at telomeres by binding TRF2, while DNA-PK acts as gatekeeper preventing resection from Apollo or any other nuclease.The processing of the 3 -overhang is then initiated by DNA-PK autophosphorylation on the ABCDE / Thr2609 cluster, which grants Apollo exclusive access to telomeres (Figure 7 A).When Apollo is absent (Figure 7 B) or DNA-PK is inhibited / cannot be autophosphorylated (Figure 7 C), the protective overhang is not formed, and the leading-end telomeres become vulnerable to alt-EJ.
The mechanism by which DNA-PK binds to and facilitates the nucleolytic attack by Apollo has striking similarities to how DNA-PK controls Artemis activity.Apollo seems to bind DNA-PK using a motif positioned just beyond the catalytic domain in the disordered C-terminal half, similar to Artemis.Furthermore, the (auto)phosphorylation of DNA-PKcs on the ABCDE / Thr2609 cluster is essential for DNA processing, while the S / T-Q sites of the nucleases themselves are not direct targets of DNA-PK kinase activity.However, some differences can also be noted.While DNA-PK is required for Artemis recruitment and opening of hairpins ( 20 ), Apollo recruitment to telomeres depends on TRF2 (Figure 3 ) ( 40 , 43 , 44 , 46 , 82 ).Furthermore, in the absence of DNA-PK, telomeres can be resected independently of Apollo ( 50 ), while hairpins can be opened only by Artemis once recruited by DNA-PK.Finally, in V(D)J recombination, DNA-PK kinase activity is not strictly required for hairpin opening by Artemis because it can be compensated for by ATM ( 20 ); in contrast, DNA-PK kinase activity is essential for resection at leadingend telomeres.These differences between Artemis and Apollo recruitment and activation could be explained by the biology of the telomeres themselves.Specifically, at telomeres, TRF2 recruits Apollo independently of DNA-PK while possibly having additional attributes that prevent ATM activation ( 38 ,50 ).
During DSB repair, DNA-PK autophosphorylation on the ABCDE / Thr2609 cluster occurs either in cis , when DNA-PK is monomeric and Artemis is granted access to the DNA ends ( 7 ), or in trans , during the transition from the XLFmediated LR complex to SR ( 10 ).Due to the resemblance between Artemis and Apollo regulation, we suggest that, at newly replicated leading end telomeres, monomeric DNA-PK undergoes autophosphorylation in cis .Such a cis-acting mechanism avoids potentially dangerous synapsis of telomere ends.It is also possible that the autophosphorylation occurs in trans , if the DNA-PK complex is recruited on both the leading-and the lagging-strand telomere ends.After synapsis, the complex on the lagging end could then phosphorylate the one on the leading end and vice versa.In either case, DNA-PK must be activated without promoting c-NHEJ.How c-NHEJ is pre-vented is not known.In addition to DNA-PK acting as a monomer, fusion of telomeres might be prevented in a synaptic setting if the lagging-strand telomere has already been remodeled into the t-loop.It is also possible that the telomeric overhang of the lagging-strand prevents transition to the SR complex.Furthermore, although ATM deficient cells show no clear sign of telomere dysfunction, it is also possible that ATM might directly restrain telomere bound DNA-PK from synapsing and then promoting c-NHEJ, as seen at collapsed replication forks ( 28 , 30 , 83 , 84 ).Finally, TRF2 or one of its binding partners could prevent c-NHEJ independent of t-loop formation as it has been shown in several other settings ( 37 , 39 , 50 ).
Why DNA-PK is at telomeres when it is not strictly required for overhang formation ( 50 ,58-61 ) is not known.It has been suggested that the telomeric role of DNA-PK is the inhibition of alt-EJ ( 85 ), but data shown here and elsewhere ( 50 ,74 ) indicate that alt-EJ can also occur in the presence of DNA-PK.One possibility is that DNA-PK binds newly replicated leading-end telomeres after recognizing them as doublestranded DNA ends.Another possibility is that it is actively recruited to facilitate replication ( 86 ,87 ).In either case, DNA-PK blocks the processing of the blunt-ended telomeres until it autophosphorylates on the ABCDE / Thr2609 cluster and grants Apollo, and no other nucleases, access to the ends.
It will now be of interest to further investigate the interplay of DNA-PK and Apollo in vitro and compare it directly to the interplay between DNA-PK and Artemis.

Figure 1 .
Figure 1.DNA-PK inhibition pre v ents Apollo-dependent leading-end telomere protection.( A ) R epresentativ e chromosome orientation fluorescence in situ hybridization (CO-FISH) of metaphase spreads of Apollo F / F MEFs immortalized with SV40LT 96 h after Hit & Run Cre treatment and / or after 24 h treatment with DNA-PKcs inhibitor NU7441 (DNA-PKi).Leading and lagging-end telomeres were detected with Cy3-(TTAGGG) 3 (red) and Ale xa488-(CCCTAA) 3 (green) probes, respectiv ely.DNA w as stained with DAPI (blue).Arrows indicate leading-end telomere fusions.( B ) Quantification of telomere fusions as shown in (A).Each dot represents the percentage of telomeres fused in one metaphase, aggregated for chromatid-type (in v olving two leading-end telomeres or two lagging-end telomeres) or chromosome-type fusions.Bars represent the median of telomere fusions in 40 metaphases across n = 4 independent experiments (10 metaphases per experiment).( C ) Telomeric overhang assay on Apollo F / F MEFs treated as in (A).Genomic DNA was digested with MboI, and the single-stranded telomeric DNA was detected in-gel with end-labeled 32 P-(AACCCT) 4 (native, left panel).DNA was then denatured, and the gel was rehybridized with the same probe to determine the total telomeric signal (denatured, right panel).The ssTTAGGG signal was normalized to the total telomeric DNA in the same lane.The normalized no Cre value was set to 1 and all the other values were calculated relative to it.( D ) Quantification of the relative overhang signal as detected in (C) for n = 4 independent experiments (indicated by different shades), with means and SDs.Statistical analysis from non-parametric Kruskal-Wallis ANO V A test for multiple comparisons (B) or ordinary one-way ANO V A for multiple comparisons (D), **** P < 0.0 0 01 , ***P < 0.001 , **P < 0.01 , *P < 0.05; n.s., not significant.

Figure 2 .
Figure 2. DNA-PK kinase activity promotes Apollo-dependent leading-end telomere protection.( A ) Immunoblots for Apollo and actin loading control in HT1080 human cancer cells 120 h after lentiviral transduction with sgRNA targeting human Apollo and / or after 24 h treatment with DNA-PKi.Graph represents Apollo le v els normaliz ed o v er actin in n = 4 independent e xperiments, with mean and SD. ( B ) R epresentativ e CO-FISH of metaphase spreads of HT1080 cells treated as described in (A).Leading and lagging-end telomeres were detected with Alexa Fluor 647-OO-(TTAGGG) 3 (red) and Cy3-O O-(CCCTAA) 3 (green), respectiv ely.DNA was stained with DAPI (blue).Arrows indicate leading-end telomere fusions.( C ) Quantification of leading-end telomere fusions as described in (B).Each dot represents the percentage of leading-end telomere fusions in one metaphase.Bars represent the median of telomere fusions in 40 metaphases across n = 4 independent experiments (10 metaphases each).(D and E) CO-FISH metaphase analysis and quantification of leading-end telomere fusions as described in (B) and (C) for DNA-PKcs + / + or DNA-PKcs KD / KD MEFs 108 h after lentiviral transduction with an sgRNA targeting mouse Apollo or the empty vector control.Bars represent the median of telomere fusions in 45 metaphases across n = 3 independent experiments (15 metaphases per experiment).Statistical analysis by ordinary one-way ANO V A for multiple comparisons (A) or by non-parametric Kruskal-Wallis ANO V A test for multiple comparisons (C,E).Statistical significance was indicated by **** P < 0.0 0 01, *** P < 0.001, ** P < 0.01, * P < 0.05; n.s., not significant.

Figure 3 .
Figure 3. DNA-PK kinase activity is not required for Apollo recruitment to telomeres.( A ) Immunoblots for HA-Apollo (anti-HA), TRF2 and phosphorylated Chk2 in SV40LT-immortalized TRF2 F / F Lig4 -/ − MEFs after retroviral transduction transduced with an empty vector (EV) or HA-Apollo, 108 h after Hit & Run Cre-mediated deletion of TRF2 and / or 24 h DNA-PKi treatment.(B and C) Representative immunofluorescence-fluorescence in-situ hybridization (IF-FISH) images and quantification of HA-Apollo localization at telomeres in the same MEFs as described in (A) 120 h after Hit & Run Cre transduction or without any treatment.Apollo and telomeres were detected using anti-HA (red) or Alexa488-OO-(TTAGGG) 3 probe (green), respectively.Approximately 300 nuclei were analyzed for each condition over n = 3 independent experiments (100 nuclei per experiment).Median bar is indicated.( D ) Immunoblots for HA-Apollo (anti-DCLRE1B, Atlas) and actin in SV40-LT-immortalized DNA-PKcs + / + , DNA-PKcs −/ − or DNA-PKcs KD / KD MEFs transduced with an empty vector (EV) or HA-Apollo.* aspecific band.( E and F ) IF-FISH analysis and quantification of HA-Apollo localization at telomeres in the SV40-LT-immortalized DNA-PKcs + / + , DNA-PKcs −/ − or DNA-PKcsK KD / KD MEFs transduced with an empty vector (EV) or HA-Apollo, analyzed as described in (B, C).For HA-Apollo, 300 nuclei o v er n = 3 independent experiments are shown (100 nuclei per experiment), with median bar.As negative control, 296 nuclei (at least 90 nuclei per experiment) were analyzed in parallel for EV. ( G ) Co-immunoprecipitation (Co-IP) of Myc-mTRF2 (WT or F120A) with HA-mApollo in 293T cells without or after 24 h treatment with DNA-PKi.Pull down was performed with anti-HA antibody.HA IP (first and second panels) and Input (third and fourth panels) were analyzed with anti Myc (first and third panels) or HA (second and fourth panels) antibodies.MY C-mTRF2-F1 20A (F1 20A) was used as negative control.Statistical analysis by non-parametric Kruskal-Wallis ANO V A test for multiple comparisons (C, F).Statistical significance was indicated by **** P < 0.0 0 01 , ***P < 0.001 , **P < 0.01 , *P < 0.05 ; n.s., not significant.

Figure 4 .
Figure 4. DNA-PK, not Apollo S / T-Q sites, is the target of DNA-PK kinase activity.(A and B) CO-FISH metaphase analysis and quantification of leading-end telomere fusions in SV40LT-immortalized Apollo F / F MEFs transduced with an empty vector (EV), HA-Apollo (WT) or HA-Apollo-AA (AA), 96 h after Hit & Run Cre-mediated deletion of Apollo.The graph represents 45 metaphases over n = 3 independent experiments, with median bars.(C and D) CO-FISH and quantification of leading-end telomere fusions in DNA-PKcs + / + or DNA-PKcs 3A / 3A MEFs with or without 24 h treatment with DNA-PKi for 45 metaphases over n = 3 independent experiments, with median bar.Statistical analysis by non-parametric Kruskal-Wallis ANO V A test for multiple comparisons (B, D).Statistical significance was indicated by **** P < 0.0 0 01 , ***P < 0.001 , **P < 0.01 , *P < 0.05; n.s., not significant.

Figure 5 .
Figure 5. Phosphorylation on the DNA-PK ABCDE / Thr2609 cluster is predicted to allow Apollo access to telomere ends.( A ) Schematics of mApollo and mDNA-PKcs.The met allo β-lact amase (MBL) and the β-CASP domains of Apollo and the N-terminal unit, the Circular cradle unit, the FAT and the Kinase domains of DNA-PKcs are indicated.Highlighted, the Apollo region (residues 1-344) predicted with high confidence by AlphaFold-Multimer modeling using AFsample and the DNA-PKcs region (residues 1981 -27 60) used in the modeling, with PQR and the ABCDE clusters.( B ) The predicted Local Distance Difference Test (pLDDT) per residue position.( C ) A model for the interaction of full-length mApollo and mDNA-PKcs (amino acids 1981 -27 60) as predicted by AFsample.Residues 345-541 in Apollo were removed from the model due to low confidence.( D ) Predicted aligned error (PAE), as expected position error ( x,y ) at residue x if the predicted and true structures were aligned on residue y. ( E ) Superposition of (C) to the cryo-EM str uct ure of autophosphorylated DNA-PK in complex with DNA (PBD: 7SGL) ( 7 ).( F,G ) Comparison between the catalytic domains of hArtemis as in (PBD:7SGL) and mApollo model as superposed to PBD:7SGL ( 7 ) in (E), and enlargement on the catalytic residues.The key catalytic residues in Apollo and Artemis are indicated in orange and blue, respectively.

Figure 6 .
Figure 6.Apollo interaction with DNA-PK mediates telomere protection.( A ) Schematics of mArtemis and mApollo.The MBL, the β-CASP, the known Lig4, PTIP (Pax transactivation-domain interacting protein) and TRF2-interacting domains, and the DNA-PKcs interacting regions are indicated.( B ) MUSCLE alignment of human and mouse Artemis and Apollo C-terminal tails.Amino acids are colored according to ph y sico-chemical properties (Zappo).Highlighted, the contact points (CP) of Artemis with DNA-PKcs ( 17 ) and the predicted DNA-PK-interacting region of Apollo, with a patch of positively charged amino acids f ollo w ed b y a patch of negativ ely charged ones.( C ) Co-IP of endogenous DNA-PKcs and TRF2 with HA-mApollo-WT (WT) or HA-mApollo-PK ( PK) in 293T cells.P ull-do wn w as perf ormed with anti-HA antibody.HA IP and input w ere analyz ed b y immunoblotting with anti-DNA-PKcs (top panels), TRF2 (middle panels) or HA (lo w er panel) antibodies.( D ) Immunoblot of SV40LT-immortalized Apollo F / F MEFs transduced with either EV, HA-Apollo-WT or HA-Apollo-PK 108 h after Hit & Run Cre-mediated deletion of endogenous Apollo.(E and F) IF-FISH analysis and quantification of Apollo localization at telomeres in the same MEFs as in ( D ).Graph represents 100 nuclei for each condition from n = 3 independent experiments (300 nuclei in total), with medians.(G and H) CO-FISH metaphase analysis and quantification of leading-end telomere fusions on MEFs treated as in (D).Graph represents 30 metaphases o v er n = 3 independent experiments with medians.Statistic by non-parametric Kruskal-Wallis ANO V A test for multiple comparisons (F, H).Statistical significance was indicated by **** P < 0.0 0 01 , ***P < 0.0 01 , **P < 0.01 , *P < 0.05; n.s., not significant.

Figure 7 .
Figure 7. Proposed model for Apollo recruitment and access to leading-end telomeres after replication.( A ) Apollo is recruited at telomeres by TRF2 binding.The physical presence of inactive DNA-PK at the blunt DNA end prevents any nucleolytic at tac k of the ends, including Apollo's.After autophosphorylation of DNA-PKcs on the ABCDE / Thr2609 cluster and Apollo interaction with DNA-PKcs, Apollo gains access to the telomere ends and perform the initial resection required for the generation of the 3 overhang and for telomere protection.( B ) In the absence of Apollo, when Apollo cannot be recruited by TRF2 or cannot interact with DNA-PKcs, the leading-end telomeres are not resected and are fused via alt-EJ.( C ) In the absence of DNA-PK autophosphorylation, due to inhibition of the kinase activity of DNA-PK and / or to mutations affecting the kinase activity and / or the ABCDE / Thr2609 cluster of DNA-PK, Apollo cannot get access to the telomere ends, the leading-end telomeres are not resected and are fused via the alt-EJ.