DNA polymerase Lambda is anchored within the NHEJ synaptic complex via Ku70/80

28 Non-homologous end joining (NHEJ) is the predominant pathway by which double -strand 29 DNA breaks ( DSBs) are repaired in mammals. To enable final break closure, various NHEJ 30 end-processing factors respond to the chemistry of the damaged DNA ends . Amongst these 31 factors is DNA polymerase lambda (Pol ), a member of the Pol X family. How members of 32 the Pol X family engage with the NHEJ complex is unknown. Here, we present cryo -EM 33 structures of Pol λ in complex with the Ku70/80 DSB sensor whilst engaged with the DNA -34 PK holoenzyme in a long-range synaptic complex. These structures reveal a specific interaction 35 site between Ku70/80 and the Pol λ BRCT domain. The functionality of this interaction is 36 assessed by generating point mutations on either side of the Pol λ BRCT :Ku70/80 interface. 37 Using these mutants in t wo orthogonal assays in cells (live protein recruitment at biphoton 38 laser-damaged nuclear sites and transfection with an original gap-filling reporter plasmid) 39 defines the molecular basis and essentiality of the BRCT domain for the recruitment and 40 activity of the Pol λ within the NHEJ complex . Ultimately, these data explain the role of this 41 interaction in cell survival to DSBs. Finally, we propose a unified model for the interaction of 42 the three Pol X family members bearing BRCT domains with the same site of Ku70/80. 43 44

45 Non-homologous end joining (NHEJ) is the predominant pathway in which double strand DNA 46 (dsDNA) breaks are repaired in mammals. Central to the process of NHEJ are large, multi -47 protein complexes formed by the canonical proteins DNA -dependent protein kinase catalytic 48 subunit (DNA-PKcs), the heterodimer of Ku70/80, DNA ligase IV (LigIV), X-ray repair cross-49 complementing protein 4 (XRCC4) and XRCC4 -like factor (XLF) 1. In addition, PAXX 50 (Paralog of XRCC4 and XLF) is a n accessory NHEJ protein identified more recently which 51 has functional redundancy with XLF in the NHEJ mechanism 2,3. Although this repair pathway 52 is a complex process, it is generally considered to proceed via three major steps; DNA end 53 recognition, DNA end processing and finally DNA ligation 4. 54 The initial step of dsDNA break recognition relies predominantly on the Ku70/80 heterodimer, 55 which engages the DNA ends and subsequently recruits the DNA -PKcs kinase, together 56 forming the DNA-PK holoenzyme 5. Recent cryo-electron microscopy (cryo-EM) structures of 57 DNA-PK assemblies revealed how DNA substrates can be positioned for efficient synapsis via 58 dimerization of this large enzyme. Intriguingly, DNA-PK has two distinct mechanisms to form 59 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 13, 2024. ; https://doi.org/10.1101/2024.08.12.607588doi: bioRxiv preprint 3 a pre-synaptic complex, corresponding to alternate structural dimers of DNA -PK. It was first 60 proposed that DNA-PK can synapse the broken DNA ends using a dimer of DNA-PK mediated 61 by the C -terminal region of Ku80 (herein termed Ku80 -mediated dimer ) 6. Soon after, an 62 alternative DNA-PK dimer complex was described, which can form upon addition of XRCC4, 63 LigIV and XLF 2,7,8. In this second complex, the DNA-PK dimer is crucially bridged by XLF 64 (herein termed XLF-mediated dimer). In both assemblies, the distance between the damaged 65 DNA ends is identical at 115 Å. This distance is not close enough for DNA ligation to proceed 66 without further rearrangement of the assembly , therefore both dimeric assemblies have been 67 referred to as long-range synaptic complexes (LRC). We have shown recently that PAXX can 68 replace XLF for bridging the Ku80-mediated DNA-PK dimer through binding to Ku70 but 69 does not replace XLF in the XLF mediated dimer 2. 70 The second step in the NHEJ mechanism is end processing and is crucial if the DNA ends 71 cannot be readily ligated. Different end processing factors can be recruited to either trim or 72 extend the damaged ends via the action of DNA nucleases or polymerases, as detailed below. 73 The final step of the repair mechanism involves ligation of the phosphodiester backbone with 74 5'-phosphate and 3' -hydroxyl moieties by LigIV, facilitated by XLF and XRCC4. To enable 75 ligation, the LRC is thought to transition to a short -range complex (SRC) following 76 autophosphorylation and removal of DNA-PKcs 9. In this SRC, the damaged DNA ends are in 77 much closer proximity to each other and thereby are primed for DNA ligation. A cryo-EM 78 structure of the SRC has been solved which shows the catalytic domain of LigIV interacting 79 directly with the DNA7. While structural studies have significantly improved understanding of 80 the NHEJ mechanism by illuminating the architecture of several complexes central to the 81 process, how the end processing proteins, in particular polymerases, interact with these 82 complexes has remained an enigma. 83 The end processing factors known to be implicated in NHEJ include nucleases such as Artemis 84 which can trim DNA nucleotides, and polymerases which can be used for end filling or 85 extension. Recent structural data has revealed how Artemis can engage with a DNA -PK 86 monomer via its N -terminal nuclease domain and is positioned between the N -HEAT 87 (Huntingtin, Elongation Factor 3, A subunit of protein phosphatase 2A, Target of 88 Rapamycin/TOR) and M -HEAT domain of DNA -PKcs 10. To date however, the structural 89 mechanism of interaction between DNA polymerases and the NHEJ machinery has not been 90 defined. The polymerases involved in NHEJ belong to the DNA Polymerases X family and 91 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 13, 2024. ; https://doi.org/10.1101/2024.08.12.607588doi: bioRxiv preprint 4 catalyse the addition of nucleotides at the 3’-OH end of DNA. In mammals, three polymerases, 92 namely Pol λ, Pol μ and terminal deoxynucleotidyl transferase (TdT), account for the majority 93 of DNA synthesis during NHEJ, with the latter mainly involved in V(D)J recombination 11,12 94 13. These polymerases share a common domain organisation which consists of a breast cancer 95 gene 1 (BRCA1) C -terminal (BRCT) region at the N -terminus followed by a C -terminal 96 catalytic domain (Figure S1A). The catalytic region facilitates the protein :DNA interaction, 97 thus enabling the function of the enzyme. The chemical nature of the DNA ends is a key 98 determinant of the polymerase activity, with each of the three enzymes having varying degrees 99 of template-dependency. Pol λ appears to display strong activity on DNA ends that have a 100 paired primer terminus 14. Furthermore, due to its accuracy, it has been proposed that Pol λ is 101 given priority for gap filling in most cell types 15. The BRCT domains are thought to direct the 102 polymerases to sites of DNA damage via interactions with complexes comprising DNA and 103 Ku70/80, or larger assemblies including the NHEJ proteins DNA-PKcs, LigIV and XRCC4 104 16,17. While the structure of the BRCT and catalytic domain of these polymerases have been 105 solved in isolation 16,18, how these domains engage with each other, and larger NHEJ 106 complexes has remained unclear. 107 In order to elucidate how the polymerases interact with NHEJ machineries, we collected cryo-108 EM data of the Ku80 mediated dimer of DNA-PK in complex with Pol λ. From our structural 109 data we identify the specific binding region of the BRCT domain of Pol λ with the bridge region 110 of Ku70/80 within DNA-PK complex. Using mutagenesis studies in cells on both Pol λ and 111 Ku70/80, we identified key residues that are critical for Pol λ anchoring in NHEJ complexes, 112 efficient gap-filling activity and ultimately cell survival to DSBs. Finally, we propose that all 113 members of the Polymerase X family share a conserved recognition site on the Ku70/80 114 heterodimer. 115 116

117 Cryo-EM structure of Pol λ bound to the Ku80-mediated DNA-PK dimer 118 In order to understand how Pol λ interacts with the NHEJ Long Range Complex (LRC), cryo-119 EM data was collected on a sample containing DNA -PKcs, Ku70/80, DNA, XRCC4, LigIV, 120 PAXX and Pol λ to allow formation of the Ku80 -mediated dimer of DNA -PK. Following 121 extensive particle classification, a consensus map was obtained of the Ku80-mediated dimer of 122 DNA-PK. The particles comprising this initial map were further characterized to generate maps 123 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 13, 2024. ; https://doi.org/10.1101/2024.08.12.607588doi: bioRxiv preprint 5 of the Ku80 -mediated dimer with and without XRCC4 and DNA LigIV (thereafter termed 124 LX4) engaged (Figure 1, S2-4). At this stage, it was apparent that there were no large areas of 125 additional density that could accommodate the full-length Pol λ protein, when compared to 126 previous maps of this DNA -PK dimer. However, a small region of additional density was 127 apparent at the bridge that defines the thinner part of the Ku70/80 ring region (Figure S5). To 128 focus specifically on this region, particles were re-extracted in smaller boxes corresponding to 129 the monomeric size of DNA-PK, and maps were generated of single protomers from within the 130 dimeric assemblies (herein termed half-dimers) with and without LX4 bound. 131 The structure without LX4 bound has an overall resolution of 4.53 Å and the map with LX4 132 bound has an overall resolution of 4.26 Å (Figure S2-5). Within both of these half-dimer maps, 133 DNA-PK along with the PAXX Ku-binding motif (P-KBM) could be docked (Figure 1). The 134 PAXX P-KBM can be seen interacting with the von Willebrand-like (vWA) domain of Ku70, 135 as has been previously characterised 2. Additionally, XRCC4 and BRCT tandem repeats of 136 LigIV can be docked into the map with LX4 bound (Figure 1). As seen in our previous 137 structures there is a central helix in DNA-PKcs which blocks the DNA ends, which is only 138 present when LX4 is engaged. Additional density can be observed at the bridge formed by 139 Ku70 and Ku80 in both maps, into which the NMR structure of the Pol λ BRCT domain (PDB: 140 2JW5) could be confidently docked (Figure 1 and S2-5). Although the full-length Pol λ was 141 used, only the BRCT domain could be modelled into the cryo-EM map. This suggests that the 142 catalytic domain has no stable interaction with DNA with the LRC in the cryo-EM assemblies 143 determined. 144 145 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 13, 2024. ; https://doi.org/10.1101/2024.08.12.607588doi: bioRxiv preprint 6 146 Figure 1: Cryo-EM structures of DNA-PK, PAXX and Pol  with and without LX4. A) DNA-PK 147 + PAXX + Pol  dimeric map with extra density for Pol  highlighted in a dashed pink circle, with two 148 orientations of processed half dimer presented below. DNA -PKcs in grey, Ku70 in orange, Ku80 in 149 green, DNA in yellow, PAXX in cyan and Pol  in purple. B) DNA-PK + PAXX + Pol  + LX4 with 150 extra density for Pol  highlighted in a dashed pink circle, with two orientations of processed half dimer 151 presented below. DNA-PKcs in grey, Ku70 in orange, Ku80 in green, DNA in yellow, LigIV in red, 152 XRCC4 in dark blue, PAXX in cyan and Pol  in purple. Inset, zoom in of the Pol  extra density. 153 154 Molecular basis of the Pol λ BRCT interaction with Ku70/80 155 The Pol λ BRCT domain is positioned at the interface formed by Ku70 and Ku80, at the 156 periphery of the DNA binding channel formed by the Ku70/80 heterodimer . An interaction 157 between Pol λ and Ku70/80-LigIV-XRCC4 has been previously suggested, with the residues 158 proposed to be involved in the interaction being situated in α1 helix of the BRCT domain 16. In 159 agreement with this previous study, our structure shows that Pol λ BRCT domain is positioned 160 to allow helix α1 to dock into a groove formed between Ku70 and Ku80 (Figure 2D). 161 Specifically, the amino acid residues Arg57 and Leu60 of Pol  BRCT domain appear to 162 mediate the contact with Ku70/80, in agreement with previous data, which found that mutations 163 of these residues prevented Pol  interaction on DNA with Ku70/80 or Ku70/80 -LigIV-164 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 13, 2024. ; https://doi.org/10.1101/2024.08.12.607588doi: bioRxiv preprint 7 XRCC416,17. With regards to Ku70/80, the specific interaction sites encompass residues 165 between 301 – 310 on Ku80 and residues 301 – 308 on Ku70. 166 167 Mutagenesis of Ku70/80 interaction residues on Pol λ 168 Based on our structural data, we probed how the Pol λ BRCT interaction with Ku70/80 impacts 169 the response to DNA damage in cells. We first analysed the influence on recruitment of Pol λ 170 at laser-induced DNA damage sites. To establish to what extent Pol λ mobilisation to damaged 171 sites was Ku -dependent, we used in -house engineered U2OS cells that allow circumventing 172 the lethality of Ku loss in human cells as previously described 2,19 (Figure 2 and S6A). Briefly, 173 endogenous Ku70 expression was first knocked -down via the constitutive expression of an 174 shRNA associated with cell rescue through expression of Ku70 tagged with a mini -auxin 175 inducible degron (mAID). Within a few hours upon addition of auxin (indole -3-acetic acid, 176 IAA), degradation of endogenous Ku70 occurs with concomitant Ku80 depletion due to the 177 known reciprocal stabilization of both Ku subunits. Full-length Pol λ fused to GFP was rapidly 178 and substantially recruited to DNA damage sites in the presence of Ku. This recruitment of the 179 full-length polymerase was only partly Ku-dependent since it substantially persisted under Ku 180 depletion conditions (Figure 2B, – red line). This may indicate an ability of Pol λ to interact 181 directly with DNA at sites of damage, or may be related to Ku-independent repair functions of 182 Pol λ outside NHEJ 20. 183 In contrast to data obtained with full -length Pol λ, when the fusion was restricted to the N -184 terminal Pol λ BCRT domain ( Figure S1A), its recruitment to micro -irradiated areas was 185 mostly Ku-dependent since it was nearly abolished without Ku (Figure 2C). Guided by our 186 structural data, we next probed the impact of specific Pol λ BRCT mutations on the recruitment 187 to DNA damage sites (Figure 2D). As shown in Figure 2E, all the mutants tested impaired 188 GFP-Pol λ BRCT accrual at laser -induced DNA damage sites to various extents, with 189 mutations at R57 or L60 positions being the most detrimental. 190 Notably, while GFP -Pol λ expression was nuclear with a nucleolar enrichment, the latter 191 disappeared upon Ku depletion (Figure S6B). The Pol λ BRCT nucleolar enrichment was also 192 strongly reduced with all BRCT mutants tested (Figure S6C). Since it is known that nuclear 193 Ku is enriched in the nucleolus in the absence of DNA damage 21, this suggests that GFP-Pol λ 194 localisation to the nucleolus relies on its interaction with Ku , and that this localisation is 195 disrupted by specific mutations within the BRCT domain of Pol λ. 196 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 13, 2024. ; https://doi.org/10.1101/2024.08.12.607588doi: bioRxiv preprint 8 197 Figure 2. Impact of mutations in the BRCT domain of Pol  on its recruitment to DSBs by Ku. A) 198 Principle of the laser micro -irradiation experiment. U2OS cells engineered for auxin (IAA) -induced 199 Ku70 knockdown, rescued or not with wild -type (WT) Ku70 and expressing either GFP -tagged full-200 length Pol  or its BRCT domain, were micro-irradiated with an 800 nm multiphoton laser to generate 201 DSBs in subnuclear areas. 202 B) Left panel: representative images before and 20 s after irradiation (irradiated areas are indicated by 203 white arrows) of nuclei from U2OS expressing GFP -tagged full-length Pol  and depleted or not of 204 Ku70 (±Ku). Right panel: quantification of fluorescence accumulation at laser -induced DNA damage 205 sites. Results are plotted as mean values of at least 20 nuclei ± SEM. 206 C) Same as in B) with U2OS expressing GFP-tagged Pol -BRCT. 207 D) Position of mutated residues in the BRCT domain of Pol  (purple) at the interface with Ku70/Ku80 208 (orange and green, respectively). 209 E) Quantification of fluorescence accumulation at laser -induced DNA damage sites in U2OS cells 210 expressing GFP-tagged WT or mutated BRCT domain of Pol . Results are plotted as mean values of 211 at least 20 nuclei ± SEM. 212 213 214 215 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 13, 2024. ; https://doi.org/10.1101/2024.08.12.607588doi: bioRxiv preprint 9 Mutagenesis of Ku70/80 residues 216 To identify key residues of Ku70/80 residues at the Pol λ BRCT :Ku interface, we replaced 217 endogenous Ku subunits with mutated forms of either Ku70 or Ku80 as guided by our structural 218 data (Figure 3A, S7A-C). We then monitored the impact on Pol λ BRCT recruitment following 219 DNA damage . To ensure that mutants of Ku70/80 did not influence Ku mobilisation at 220 damaged sites per se, we monitored simultaneously the recruitment of mCherry -PAXX co-221 expressed within the same cells, since PAXX recruitment is strictly Ku-dependent 2 (Figure 222 3B-F). Mutations on the positions F303, L310 on Ku70 or E292, E304 on Ku80 nearly 223 abolished Pol λ recruitment (Figure 3D, 3F). Notably, s ome subtitutions on these positions 224 (e.g. Ku70 L310R, Ku80 E304A) also affected PAXX recruitment, although at an intermediate 225 extent compared to the full defect observed for Pol λ BRCT (Figure 3C-E). Considering that 226 PAXX binding to Ku70 is far away from the Ku bridge region, this indicate s that these 227 substitutions likely compromise the stability of the Ku:DNA interaction in addition to directly 228 affecting the Ku :polymerase interaction. Nevertheless, these data establish that Ku70 F303, 229 L310 and Ku80 E292, E304 are key positions for Ku interaction with Pol λ. 230 231 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 13, 2024. ; https://doi.org/10.1101/2024.08.12.607588doi: bioRxiv preprint 10 232 Figure 3. Impact of mutations in the bridge region of Ku70 and Ku80 on the recruitment of the 233 BRCT domain of Pol  at DSBs. A) Position of mutated residues in the bridge domain of Ku70 234 (orange) or Ku80 (green) at the interface with Pol  (purple). B) Principle of the laser micro-irradiation 235 experiment. U2OS cells engineered for auxin (IAA) -induced Ku70 knockdown, rescued with WT or 236 mutated forms of Ku70 and expressing both the GFP-tagged BRCT domain of Pol  and the mCherry-237 tagged PAXX protein, were micro -irradiated and accumulation of each fluorescence was analyzed as 238 in Figure 2. C) Representative images before (upper frames) and 20 s after irradiation (lower frames, a 239 white arrow indicates irradiated area) of nuclei from U2OS expressing either WT Ku70 (left) or the 240 L310R mutant (right) D) Quantification of fluorescence accumulation at laser -induced DNA damage 241 sites of GFP-Pol-BRCT (upper chart) and mCherry-PAXX (lower chart) in U2OS cells expressing the 242 indicated Ku70 constructs. Results are plotted as mean values of at least 20 nuclei ± SEM. E) Principle 243 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 13, 2024. ; https://doi.org/10.1101/2024.08.12.607588doi: bioRxiv preprint 11 of the laser micro-irradiation experiment. U2OS cells engineered for doxycycline (doxy)-induced Ku80 244 knockdown, rescued with WT or mutants of Ku80 and expressing both the GFP-tagged BRCT domain 245 of Pol  and the mCherry -tagged PAXX protein, were micro -irradiated and accumulation of each 246 fluorescence was analyzed as in (D). F) Quantification of fluorescence accumulation at laser -induced 247 DNA damage sites of GFP-Pol-BRCT (upper chart) and mCherry-PAXX (lower chart) in U2OS cells 248 expressing the indicated Ku80 constructs. Results are plotted as mean values of at least 20 nuclei ± 249 SEM. 250 251 Gap-filling assay in cells to assess impact of Ku70/80:Pol λ interface mutations 252 Pol λ has a large spectrum of substrates, accommodating breaks with overhangs (<4 to 12 nt), 253 microhomologies (1-6 nt) and gaps (1-8 nt) 22. Based on these features and to assess the impact 254 of mutations at the Ku:Pol λ interface on Pol λ activity during end-joining, we designed a Cas9-255 targeted plasmid , which once transfected in HEK -293T cells can be used as a gap -filling 256 dependent DNA repair reporter (Figure 4A). Briefly, the reporter is designed such that the Cas 257 family member Cpf1 (Cas12a) induces two staggered DSBs in the mCherry cDNA which have 258 two complementary nucleotides at the very end. Gap -filling at the junction restores mCherry 259 expression, while end-filling enables the expression of a downstream EGFP cDNA which lies 260 in a different reading frame. No expression of mCherry or GFP is expected if ends are trimmed 261 or not ligated . Expression of BFP from a co -transfected plasmid accounts for transfection 262 efficiency. Following Cpf1 -induced DSB, expression of fluorescent proteins is measured by 263 flow cytometry. We first established that gap -filling dependent mCherry expression relied on 264 cells being transfected by the complete Cpf1/gRNA system (Figure 4B), and on NHEJ activity 265 since it was largely inhibited by a DNA-PK inhibitor catalytic activity (NU7441), or upon Ku 266 removal or genetic ablation of the NHEJ factors DNA-PKcs, LigIV, XLF or PAXX involved 267 in break ends tethering and end joining 2 (Figure S7D, E). Moreover, sequencing the junctions 268 from mCherry positive cells revealed that the expected product from gap-filling repair was 269 formed, further validating our reporter system (Figure S7F). Notably, no GFP expression was 270 detected implying that gap -filling is far dominant over end -filling at these staggered DSB 271 (Figure 4B). Then, we evaluated the Pol λ-dependency of the gap-filling activity detected. Pol 272 λ was found dominant over Pol µ in gap -filling at three-nucleotide gaps on 5′ overhang 273 substrates 14,23. Indeed, Pol λ knock -out (KO) led to a ~70% reduction of the activity on our 274 substrate that was fully restored upon re-expression of WT Pol λ, while Pol µ KO alone did not 275 impair gap-filling activity (Figures S7G, H). Notably, Pol µ KO in Pol λ KO cells further 276 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 13, 2024. ; https://doi.org/10.1101/2024.08.12.607588doi: bioRxiv preprint 12 decreased gap-filling activity, suggesting a small compensation by Pol μ in the absence of Pol 277 λ. Finally, expression of a catalytically dead Pol λ (D427A-D429A, 24) (Pol λ dead ) further 278 decreased the remaining gap-filling activity in Pol λ KO cells, supporting a dominant negative 279 effect of the Pol λ dead construct on backup gap-filling enzymes, including Pol µ. 280 281 Characterization of key positions in Pol λ BRCT:Ku interface for Pol λ-dependent gap-filling 282 activity and survival to IR 283 Using the mainly NHEJ - and Pol λ -dependent gap -filling assay described above, we then 284 analysed the impact of mutations in Pol λ BRCT though transfection of Pol λ KO HEK-293T 285 cells complemented with WT or mutant Pol λ. As a control, the same cells were used to evaluate 286 direct end-joining (EJ) activity at blunt-ended breaks using our dedicated Cas9-based reporter 287 plasmid assay described previously 19. As shown in Figure 4C, except for R96A, all other 288 mutations tested in Pol λ BRCT reduced gap-filling efficiency at staggered DSB. The effect of 289 F61G mutation is inconclusive since it lowered the protein expression (Figure S8A). Notably, 290 the extent of defect observed in gap-filling activity of full-length mutant Pol λ correlates with 291 that observed in recruitment of the corresponding mutant GFP -BRCT at laser sites ( Figure 292 2E), again with R57 and L60 positions being the most crucial. Since no detectable repair defect 293 was found at blunt-ended DSB, a readout for direct EJ, this indicates that outside its catalytic 294 function at defined breaks Pol λ is unlikely to fulfill a general function in the overall NHEJ 295 complex assembly and/or stability. Using the same assays, we then evaluated the impact of 296 mutations of Ku70 in the Pol λ:Ku interface by using HEK-293T cells expressing mAID-Ku70 297 fusion and transduced with WT or mutated Ku70 forms (Figure S8B). Upon auxin addition, 298 we analysed in parallel the impact of individual or combined mutations in Ku70 on gap-filling 299 at staggered DSB and on direct EJ at blunt DSB (Figure 4D). Compared to the slight reduction 300 in gap -filling observed with Ku70 mutants at T307 and L310 positions, F303G mutation 301 reduced gap -filling by 60% without impairing repair at blunt -ended breaks, indicating that 302 Ku70 F303 is a crucial position for Ku interaction with Pol λ BRCT but not for Ku interaction 303 with DNA. Then, assessing cell radiosensitivity, we observed that loss of Pol λ marginally 304 reduced cell resistance to ionizing radiation ( IR) that was much more decreased upon 305 expression of the Pol λ dead construct ( Figure 4E), in agreement with gap -filling activity 306 assessed in parallel (Figure S7H). Notably, the R57E mutation that impaired Pol λ recruitment 307 to DSB released the sensitivity associated with expression of the Pol λ dead construct when 308 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 13, 2024. ; https://doi.org/10.1101/2024.08.12.607588doi: bioRxiv preprint 13 both mutations were combined (Figure 4E) and similarly enhanced gap-filling activity up to 309 the level observed in Pol λ KO cells (Figure S8C). Finally, since we observed that the loss of 310 cell viability consecutive to Ku removal is reversible within a 32 h time window, we assessed 311 the consequence of impairing Ku :Pol λ interaction on cell survival to IR (Figure S8D and 312 Figure 4F). We showed that the Ku70 F303G mutant that preserved direct EJ but impaired 313 gap-filling did not fully complement the radiosensitivity observed upon Ku removal. This 314 supports that Ku interaction with gap-filling polymerases promotes cell survival to IR through 315 positioning them at the break ends (Figure 4F). 316 317 318 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 13, 2024. ; https://doi.org/10.1101/2024.08.12.607588doi: bioRxiv preprint 14 Figure 4. Effect of Pol  or Ku mutations on gap-filling activity in cells and survival to IR. A) Gap-319 filling reporter substrate. The gap -filling reporter substrate consists of two consecutive frameshifted 320 (+1) coding sequences for mCherry (mCh) and EGFP, respectively. The mCherry coding sequence is 321 interrupted by a cassette containing the HSV -TK polyadenylation sequence (TK -pA) flanked by two 322 inverted copies of a Cpf1 target sequence (blue characters; the PAM GAAA sequence is shown in 323 orange). Following Cpf1-mediated double cleavage, the HSV-TK polyadenylation sequence is deleted 324 and the resulting 5’ overhang DNA ends can be rejoined in different ways. First, the two overhangs can 325 partially anneal to each other by two G:C pairs, leading to single-stranded gaps than can be filled, thus 326 restoring an intact mCherry coding sequence. Second, the two overhangs can undergo end -filling, 327 leading to blunt ends whose joining disrupts the mCherry reading frame (+2 frameshift) but enables the 328 expression of the downstream EGFP coding sequence. Finally, the two overhangs can be trimmed off 329 before end-joining, leading to a -8 frameshift of the mCherry coding sequence and resulting in neither 330 red nor green fluorescence. B) Gap-filling assay. Top: the gap -filling reporter assay is performed by 331 transfecting HEK-293T cells with the reporter substrate together with a Cpf1/gRNA-expressing vector 332 to cleave the substrate and a BFP -expressing vector to normalize for transfection efficiency. 333 Fluorescence expression is analyzed 48 h later by flow cytometry. Bottom: representative fluorescence 334 microscopy images of cells non -transfected (NT) or transfected with the full reporter system (full) or 335 omitting the substrate (-substrate) or the gRNA expression vector (-gRNA). PC: phase contrast. C) Gap-336 filling activity (red bars) or direct end-joining activity (orange bars) assessed in parallel in HEK-293T 337 cells knocked-out for Pol  and complemented with either an empty vector (EV) or expression vectors 338 for WT or different mutants of Pol . Results are normalized to the WT condition and plotted as mean 339 values of three to six independent experiments ± SD. P -values from Student’s t-test between the WT 340 condition and the considered mutant, for gap-filling and direct end-joining activities, respectively, are 341 as follows: R96A (0.0731; 0.0583), S116G (0.0007; 0.0176), L60A (0.0156; 0.0165), L60R (0.0089; 342 0.3312), R57A (<0.0001; 0.0025), R57E (0.0021; 0.0252), F61G (0.0005; 0.0528). * P < 0.05, ** P < 343 0.01, *** P < 0.001, **** P < 0.0001, ns: not significant. D) Gap-filling activity (red bars) or direct 344 end-joining activity (orange bars) assessed in parallel in HEK-293T cells knocked-down for Ku70 and 345 rescued with expression vectors for WT or different mutants of Ku70. Results are normalized to the 346 WT condition and plotted as mean values of four experiments ± SD. P -values from Student’s t-test 347 between the WT condition and the considered mutant, for gap-filling and direct end-joining activities, 348 respectively, are as follows: R301A (0.0541, 0.9381), F303G (<0.0001; 0.0841), T307A (0.0006, 349 0.0039), L310G (0.0015; 0.0503), L310R (0.0025; 0.0437). E) Cell survival to X-rays of HEK-293T 350 cells knocked-out for POLL and complemented with an empty vector (EV) or expression vectors for 351 WT or the indicated mutants of Pol . Y axis is log scale. Results are normalized to the untreated 352 condition and plotted as mean values of four to seven experiments ± SD. P-values were calculated at 2 353 Gy using unpaired t -test: Pol -dead versus EV (0.0207); Pol -dead versus WT (0.0092). F) Cell 354 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 13, 2024. ; https://doi.org/10.1101/2024.08.12.607588doi: bioRxiv preprint 15 survival to X -rays of U2OS cells knocked -down for Ku70 (Ku -KD) and complemented or not with 355 expression vectors for WT Ku70 or the F303G mutant. Results are mean values of four experiments 356 and plotted as in (E). P-values at 2 Gy: Ku70-F303G versus Ku-KD (0.0195); Ku70-F303G versus Ku-357 WT (0.0089). 358 359 360

361 We determined the cryo-EM structures of Pol λ in complex with the Ku 70/80 heterodimer 362 bound to DNA whilst engaged with the DNA-PK holoenzyme. These structures show a clearly 363 defined interaction between the BRCT domain of Pol λ and the interface between Ku70/80. 364 We assessed the functionality of this interaction by generating specific site directed mutants on 365 either Pol λ BRCT or Ku70/80. These mutants were used in two orthogonal assays: live protein 366 recruitment at nuclear laser sites and gap -filling activity in cells transfected with an original 367 reporter assay based on Cpf1-generated partially complementary DSB ends. These data define 368 the molecular basis and essentiality of the BRCT domain for recruitment of Pol λ within the 369 NHEJ complex. The data also establishes key positions on Ku 70/80 and Pol λ -BRCT that 370 mediate their interaction , and for the first time position the interaction region at the external 371 face of the Ku70/80 dimer interface. 372 From our cryo-EM structure of the Ku80 mediated DNA -PK dimer bound to LX4, t he LigIV 373 tandem BRCT1 domain can be seen occupying the previously described site on Ku70/80 6,7 374 that is distinct from the Pol λ BRCT interaction site (Figures 1 and 5). The LigIV BRCT1 sits 375 in the groove formed by the Ku70/80 dimer interface, whereas Pol λ BRCT is located at an 376 adjacent site. From our structural data, it is apparent that the two proteins have distinct sites of 377 contact with Ku70/80, and there is no evidence that Pol λ and LigIV form direct interactions 378 with each other. Thus, the previously reported observation that specific Pol λ mutations prevent 379 complex formation with Ku70/80 -LigIV-XRCC4 are likely due to disruption of the Pol 380 λ:Ku70/80 interaction specifically, as has previously been suggested 16. Interestingly, when we 381 collected cryo-EM data of Ku70/80 alone (without DNA-PKcs) with Pol λ, we did not observe 382 any density for the BRCT domain of Pol λ engaged with Ku70/80. Therefore, it is possible that 383 the Pol λ interaction with Ku70/80 is stabilised within the DNA-PK holoenzyme complex. Our 384 cryo-EM structures indicate a possible weak interaction between the BRCT domain of Pol λ 385 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 13, 2024. ; https://doi.org/10.1101/2024.08.12.607588doi: bioRxiv preprint 16 and DNA-PKcs. Although the resolution is low, we would predict that Pol λ (residues 128-386 130) may interact with residues Ser187 and/or Glu188 on DNA-PKcs. 387 Notably, we found that recruitment of the N-terminal Pol λ BRCT domain to micro-irradiated 388 areas was essentially Ku70/80-dependent in contrast with data obtained with the full-length Pol 389 λ. This suggests that Pol λ recruitment to sites of DNA damage may rely on interactions outside 390 of its BRCT domain. This may reflect additional roles of Pol λ outside of classical NHEJ such 391 as in Base Excision Repair where it is proposed to interact with DNA glycosylases involved in 392 the repair of alkylated or oxidized bases 20. As previously reported, we found that Pol λ 393 deficient cell lines showed modest , if any, sensitivity to IR, whereas cells deficient in both λ 394 and µ polymerases were clearly IR sensitive 14,25,26. The expression of a catalytically inactive 395 form of Pol λ concurrently inhibited gap -filling activity in cells (this work) and negatively 396 impacted on cell survival to IR as previously reported 24, to extents higher than the sole deletion 397 of Pol λ. This suggests that Pol λ dead prevents the rescue of the Pol λ defect by Pol µ, likely 398 through occupying and occluding a common interaction site on Ku 70/80, and thereby 399 sustaining a dominant negative effect. Also, we showed that Ku70 F303G mutant cells exhibit 400 sensitivity to IR equivalent to that of a double Pol λ-Pol  KO cells, suggesting again that this 401 mutation impairs Ku70/80 interaction with both Pol X proteins. Indeed, AlphaFold prediction 402 of the structures formed between Ku70/80 with DNA and the BRCT domains of Pol  and TdT 403 indicates the use of similar protein :protein interfaces ( Figure 5B, C). Also, the broad 404 composition and organization of the Pol λ BRCT domain is conserved amongst Pol X members, 405 supporting a conserved mode of interaction with Ku70/80 (Figure 5D). Together with our 406 present cellular data, this allows us to propose a unified model for the interaction of the three 407 BRCT-bearing Pol X with Ku 70/80 on the same site allowing their respective recruitment to 408 the NHEJ complex. 409 The Ku70/80 interaction with Pol X BRCT domain reported here adds to the fascinating list of 410 Ku70/80 sites that contact NHEJ factors. These include Ku80 sites for A-KBM bearing proteins 411 (APLF, WRN, MRI), XLF via its X-KBM 27 and DNA-PKcs 8, Ku70 sites for PAXX via its P-412 KBM 2 and Lig IV via its BRCT1 7,8, and combined sites in Ku80/Ku70 for inositol 413 hexaphosphate 19 and now for Pol X BRCT s (B-KBM). The structures reported here of the 414 Ku70/80:Pol λ interaction represents a paradigm for Pol X mode of binding to Ku70/80 and 415 reinforces the notion that Ku70/80 acts as a central structural hub in the NHEJ mechanism 28. 416 Like other Ku70/80 partners whose KBM motifs are at the N or C-terminus of the protein, Pol 417 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 13, 2024. ; https://doi.org/10.1101/2024.08.12.607588doi: bioRxiv preprint 17 X members are anchored by an N-terminal BRCT domain separated by a flexible region from 418 the catalytic portion of the protein , allowing free movement for the correct positioning at the 419 site of the DSB. 420 The structures of the LRC presented here indicate that the Pol X:Ku70/80 interaction occurs 421 early during the NHEJ process, at a stage where DNA -PKcs is still present and at which the 422 DNA ends are likely not yet accessible to processing enzymes. Moreover, since our cryo -EM 423 structures were obtained using a blunt -ended DNA substrate, this suggests that it is not the 424 nature of the DNA ends that dictates Pol X engagement into the NHEJ process but rather its 425 intrinsic affinity for the initial DNA -PK complex. These structures also illustrate that the 426 evolution of distinct and independent binding sites on the Ku70/80 heterodimer enables the 427 concurrent anchoring of NHEJ proteins (PAXX/XLF, DNA -PKcs, Pol X, XRCC4/LigIV) 428 dedicated to specific activities involved in the repair reaction (synapsis, kinase, polymerase, 429 ligase, respectively ) (Figure 5E). Moreover, the dimeric nature of the LRC allows for a 430 scenario where two different members of Pol X family could be engaged within the same 431 dimeric complex simultaneously. Although our DSB repair assay in cells could theoretically 432 report both gap-filling and end-filling activities, we found that gap-filling is far dominant. This 433 suggests that, when possible, end -annealing on minimal homology is preferred, most likely 434 because it contributes to the stability of the synaptic complex and ultimately to the fidelity of 435 DNA repair . The two -nucleotide overhang substrate used here supports a tight coupling 436 between end annealing and gap filling, as demonstrated elegantly for NHEJ mediated repair in 437 Xenopus egg extracts 23. Within the NHEJ LRC, this coupling is ensured by the concurrent 438 binding of ends , and resynthesis factors by Ku70/80 on both sides of the DSB. Also, the 439 coexistence of LigIV and processing enzymes like Pol X in the synaptic complex at break ends 440 ensures that ligation can proceed as soon a DNA ends are ligatable, limiting unnecessary DNA 441 sequence alteration 23. This illustrates that the unique capability of Ku 70/80 for simultaneous 442 multivalent interaction is crucial for the high adaptability of the NHEJ process 29. 443 444 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 13, 2024. ; https://doi.org/10.1101/2024.08.12.607588doi: bioRxiv preprint 18 445 Figure 5: Comparison of the binding of BRCT domains from NHEJ polymerases X ( ,  and 446 TdT) to Ku70/80 and scheme of Ku as a structural hub. A) Experimentally determined structure 447 from this work showing Pol  binding to Ku70/80, DNA and PAXX. B) AlphaFold 3 prediction of Pol 448  binding to Ku70/80 and DNA. C) AlphaFold 3 prediction of Pol TdT binding to Ku70/80 and DNA. 449 Ku70 is in orange, Ku80 in green, DNA in yellow, PAXX P-KBM in cyan and Polymerase BRCTs in 450 purple. D) Sequence alignment of the BRCT domain of Pol X family members generated by ESPript 30. 451 Structural elements from the Pol  cryo-EM structures are shown above the alignment. Residues 452 mutated in this study are indicated with *. E) Ku70/80 (orange/green) as a structural hub showing the 453 binding of Pol  (purple), Ligase IV (red), PAXX (cyan), XLF (pink) and APLF (brown). 454 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 13, 2024. ; https://doi.org/10.1101/2024.08.12.607588doi: bioRxiv preprint 19

We thank Dr Christos Savva, Dr Emma Hesketh, Dr Claudia Lancey and 455 Dr TJ Ragan from the Midlands Regional Cryo -EM facility for help with grid preparation, 456 screening, data collections and processing support. P.F., N.B., J.C., S.B. V.R, J.B.C and P.C. 457 were supported by the French National Research Agency ( ANR-20-CE11-0026). This work 458 was supported by the Fondation ARC (J.C.). P.C. is a scientist from INSERM. We 459 acknowledge the imaging facility TRI, member of the national infrastructure France -460 BioImaging infrastructure supported by the French National Research Agency (ANR -10-461 INBS-04). JB.C and V.R thank the I2BC Protex platform supported by French Infrastructure 462 for Integrated Structural Biology (FRISBI) ANR-10-INBS-0005. J.C and V.R were supported 463 by ANR-21-CE12-0019-01, ANR-22-CE12-0037 and ANR 23-CE11-0033. 464 Funding: We thank the Lister Institute of Preventative Medicine Prize for support of this 465 research. We would also like to thank the Medical Research Council for the standard research 466 grant (MR/X00029X/1). 467 Author contributions: 468 A.K.C. directed the study. A.K.C. , P.C. and P.F. led the experimental design. H.A., S.Z., 469 A.K.C., S.B., P.F. and P.C. prepared the manuscript. H.A. and S.Z. collected the cryo-EM data 470 and modelled and analysed the structures. C.H . helped with structural analysis and edited 471 manuscript. G.M. Purified proteins and collected initial cryo-EM data. P.F. designed and built 472 the cellular tools. P.F. and N.B. designed, validated and carried out the in vivo repair 473 experiments. P.F. and J.C. performed the multiphoton laser micro -irradiation experiments. 474 A.K.C., S.B. and P.C. acquired funding. S.W.H and D.Y. C advised with cryo -EM data 475 collection set -up and processing . S.W.H helped with structural analysis and manuscript 476 preparation. V.R and J.B.C provided insect cells pellets for human NHEJ factors. 477 Competing interests: 478 The authors declare no competing interests. 479 Data and materials availability: All structural data presented are publicly available. Cryo -480 EM structures and maps are deposited at the PDB and EMDB with accession codes as follows: 481 482 483 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 13, 2024. ; https://doi.org/10.1101/2024.08.12.607588doi: bioRxiv preprint 20

484 Purification of DNA-PKcs and Ku70/80 485 DNA-PKcs and full -length His -tagged Ku70/80 were expressed and purified as previously 486 described 6. 487 Expression and purification of full-length XLF and LX4 488 A construct containing full -length 10xHis-tagged XLF and LX4 (Lig IV and XRCC4) were 489 expressed in insect cells and purified as previously reported 2. 490 DNA annealing. Biotinylated Y-shaped 42-55 bp dsDNA were synthesised and annealed as 491 described previously 6. DNA sequences used for annealing can be found below. 492 Y-shaped DNA Forward 493 Biotin-494 CGCGCCCAGCTTTCCCAGCTAATAAACTAAAAACTATTATTATGGCCGCACGCGT 495 Y-shaped DNA Reverse 496 ACGCGTGCGGCCATAATAATAGTTTTTAGTTTATTGGGCGCG 497 Cryo-EM sample preparation of DNA -PK, LX4, PAXX and Polymerase  structure 498 Proteins were concentrated using a centrifugal filter (Amicon) with a 30 kDa cut-off and buffer 499 exchanged into 20 mM HEPES, pH 7.6, 200 mM NaCl, 0.5 mM EDTA, 2 mM MgCl 2, 5 mM 500 DTT. Purified Ku70/80 full-length was then first mixed with Y-shaped 42-55 bp DNA before 501 being mixed with purified DNA -PKcs, LX4, PAXX and Pol  in respectively a 1:1:2:2:2:6 502 (DNA:DNA-PKcs:Ku:LX4:PAXX:Pol) ratio. 503 Cryo-EM grid preparation 504 Aliquots of 3 μl of ~2.5 mg/ml of the NHEJ complex (based on DNA -PKcs concentration) 505 were mixed with 8 mM CHAPSO to eliminate particle orientation bias (final concentration ; 506 Sigma) before being applied to Holey Carbon grids (Quantifoil Cu R1.2/1.3, 300 mesh), glow 507 discharged for 60 s at current of 25 mA in PELCO Easiglow (Ted Pella, Inc .). The grids were 508 then blotted with filter paper once to remove any excess sample, and plunge -frozen in liquid 509 ethane using a FEI Vitrobot Mark IV (Thermo Fisher Scientific) at 4 oC and 95 % humidity. 510 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 13, 2024. ; https://doi.org/10.1101/2024.08.12.607588doi: bioRxiv preprint 21 Cryo-EM data acquisition 511 The data was collected on a Titan Krios equipped with a Gatan K3 direct electron counting 512 detector at the University of Leicester. All data collection parameters are given in Table S1. 513 Cryo-EM Image processing 514 The classification process for the two datasets is summarised schematically in Figure S2. The 515 final reconstructions obtained had overall resolutions (Table S1), which were calculated by 516 Fourier shell correlation at 0.143 cut-off. 517 Cryo-EM structure refinement and model building 518 The model of the DNA -PK monomer (PDB:7NFE) was used as an initial template and rigid -519 body fitted into the cryo-EM density in UCSF chimera 31 and manually adjusted and rebuilt in 520 Coot 32. LX4 was either removed or manually adjusted depending on whether density was 521 present and PAXX was docked within the central density. The BRCT domain of Polymerase  522 was then manually docked into the density ( PDB: 2JW5 ) in chimera. Models and their 523 corresponding density maps were run through Namdinator 33 before being refined using Phenix 524 real-space refinement 34. 525 Cell lines, cell culture and cell engineering 526 U2OS cells (human osteosarcoma cell line from ECACC, Salisbury, UK) and HEK -293T 527 human embryonic cells, were grown in DMEM (Eurobio, France) supplemented with 10% fetal 528 calf serum (Eurobio, France), 125 U/ml penicillin, and 125 μg/ml streptomycin. Cells were 529 maintained at 37°C in a 5% CO2 humidified incubator. 530 HEK-293T cells knocked-out for POLL (DNA Polymerase ), POLM (DNA Polymerase ), 531 PRKDC (DNA -PKcs, Addgene Plasmid#220493 35, LigIV, XLF and PAXX genes were 532 obtained following cell transfection with the pCAG-eCas9-GFP-U6 vector expressing the 533 corresponding guide RNA (see below) using jetPEI (Polyplus) as a transfection reagent. 534 Following cell sorting, individual clones were isolated and checked by western blot. 535 The generation of U2OS cells expressing an inducible shRNA against Ku80 and the generation 536 of U2OS and HEK-293T cells expressing a mini-auxin-inducible degron-tagged Ku70 protein 537 (mAID-Ku70) in place of the endogenous Ku70 protein have been previously described 19,36. 538 Production of lentiviral particles in HEK-293T cells and transduction of U2OS and HEK-293T 539 cells were performed as previously described 37. 540 541 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 13, 2024. ; https://doi.org/10.1101/2024.08.12.607588doi: bioRxiv preprint 22 Western-blot and antibodies 542 Cell pellets were washed with phosphate-buffered saline (PBS) and resuspended in lysis buffer 543 (50 mM Hepes-KOH, pH 7.5, 450 mM NaCl, 1 mM EDTA, 1% Triton X -100) supplemented 544 with Halt protease inhibitor cocktail (ThermoFisher Scientific). Cells were lysed by four 545 freeze/thaw cycles in liquid nitrogen and 37°C water bath. Lysates were cleared by 546 centrifugation and protein concentrations were determined using the Bradford assay (Bio-Rad, 547 Hercules, CA). Equal amounts of proteins were mixed with concentrated loading sample buffer 548 to 1X final concentration (50 mM Tris.HCl pH 6.8, 10 % glycerol, 1 % SDS, 300 mM 2 -549 mercaptoethanol, 0.01 % bromophenol blue), heat -denatured, separated by SDS -PAGE on 550 Miniprotean TGX stain -free 4-15 % gradient gels (Bio -Rad, Hercules, CA) and blotted onto 551 Protran 0 .45 µm nitrocellulose membranes (GE Healthcare) . Membranes were blocked for 552 60 min with 5% non -fat dry milk in PBS, 0.1% Tween -20 (Sigma-Aldrich) (PBS-T buffer), 553 incubated as necessary with primary antibody diluted in PBS -T containing 1% bovine serum 554 albumin (immunoglobulin- and lipid-free fraction V; Sigma-Aldrich) and washed 3 times with 555 PBS-T. Membranes were incubated for 1 h with HRP -conjugated secondary antibodies 556 (Jackson Immunoresearch Laboratories) in PBS -T and washed five times with PBS -T. 557 Immuno-blots were visualized by enhanced chemiluminescence (Western Lightning Plus -558 ECL; Perkin Elmer) and autoradiography. Primary antibodies used: mouse monoclonal 559 antibodies anti-DNA-PKcs (clone 18.2; Thermo Fisher Scientific), anti-Ku80 (clone 111), anti-560 Ku70 (clone N3H10), anti -Pol lambda (clone E11, Santa Cruz), anti -beta-Actin (clone C4, 561 Santa Cruz); rabbit monoclonal antibodies an ti-LigIV (A11432, Abclonal ), anti -Pol mu 562 (EPR10470(B), Abcam); rabbit polyclonal antibodies anti -XLF (A199957, Abclonal), anti -563 PAXX (NBP1-94172, Novus). 564 565 Ionizing irradiation and cell survival analysis 566 Two to six thousand U2OS or HEK -393T cells per well were seeded in duplicate in six -well 567 plates. Cells were exposed 24 h later to various doses of X-ray using a Faxitron RX-650 device 568 (130 kV, 5 mA, dose rate 0.3 Gy per min). Six to seven days later, cells were fixed with 7% 569 trichloroacetic acid for 1 h at 4°C. Fixed cells were extensively washed with water and plates 570 were air dried before staining for 15 min with crystal violet (0.1% aqueous solution). Stained 571 cells were further extensively washed with water and plates were air dried. Staining was 572 dissolved with 10% acetic acid solution and absorption was measured at 570 nm (Ultrospec -573 3000 spectrophotometer, Pharmacia Biotech). Results were plotted as mean values of >3 574 independent experiments ± s.d. using Microsoft Excel software. 575 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 13, 2024. ; https://doi.org/10.1101/2024.08.12.607588doi: bioRxiv preprint 23 576 Multiphoton laser micro-irradiation 577 Live cell microscopy and multiphoton laser micro -irradiation were conducted as previously 578 described 27. 579 580 In vivo DNA end-joining assays 581 To assess gap -filling activity, HEK -293T cells were seeded to 20 -40% confluence in 6 -well 582 plates and transfected 24 h later with a mix of Cpf1 -targeted gap-filling reporter substrate, 583 Cpf1/gRNA expressing vector and mTagBFP2 expressing plasmid as an internal control. Cells 584 were trypsinized 2 days post -transfection, washed with PBS and analyzed by flow cytometry 585 on a Fortessa X -20 cell analyzer (BD Biosciences). The integrated red fluorescence signal 586 accounting for gap-filling-mediated repair events (% positive cells x mean fluorescence) was 587 normalized to that of transfection efficiency (BFP). For the repair junction analysis, a variant 588 of the substrate vector was generated in which a sequence encoding HygroR-T2A was inserted 589 in frame, upstream of the disrupted mCherry sequence. This plasmid was stably transfected 590 into HEK -293T cells and, after selection with hygromycin, positive clones were isolated. 591 Following further transient transfection with the Cpf1/gRNA expressing vector, red fluorescent 592 and non-fluorescent clones were isolated. Sequence junctions were analyzed by genomic DNA 593 extraction (SV Genomic DNA Purification System, Promega), PCR amplification with primers 594 mCh-Xba-F and mCh-Xho-R and sequencing (Eurofins Genomics; Ebersberg, Germany). 595 Direct end -joining activity was assessed as described previously with a dedicated reporter 596 substrate in which blunt -ended DSBs were generated with appropriate Cas9/gRNA co -597 expression 19. 598 599 Plasmids and DNA manipulations 600 Cas9 and gRNA expressing vectors for gene knockout were generated by inserting the pre -601 annealed gRNA-F and gRNA-R oligonucleotides of the corresponding targeted gene into the 602 BbsI restriction sites of pCAG-eCas9-GFP-U6-gRNA plasmid (a gift from Jizhong Zou, 603 Addgene plasmid # 79145 ; http://n2t.net/addgene:79145 ; RRID:Addgene_79145). See below 604 the list of oligonucleotides. 605 The previously described pLV3 lentiviral vector 27 was modified as follows to allow insertion 606 of Pol  cDNA downstream a puromycin resistance gene and a sequence encoding the T2A 607 ribosomal skipping peptide: first, a T2A cassette (pre -annealed oligonucleotides kpn2 -T2A-608 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 13, 2024. ; https://doi.org/10.1101/2024.08.12.607588doi: bioRxiv preprint 24 Mlu-F and kpn2 -T2A-Mlu-R) was inserted into the Kpn2I and MluI restriction sites of the 609 pLV3 plasmid; second, a PCR -amplified fragment with a Puro -resistance cDNA sequence 610 (PCR reaction with primers HF -Puro-F and HF -Puro-R on a synthetic DNA molecule as a 611 template) was added by Hot -Fusion 38 at the Kpn2I site, resulting in the pLV3 -Puro-T2A 612 plasmid. A PCR-amplified human Pol  cDNA fragment (primers PolL-Mlu-F and PolL-Bcu-613 R) was then inserted between the MluI and BcuI restriction sites of pLV3-Puro-T2A. 614 Expression vectors for mutant forms of Pol  were obtained in a similar manner following an 615 additional step of overlap extension PCR mutagenesis with the corresponding PolL-mut-F and 616 PolL-mut-R oligonucleotides as mutated inner primers (see below the list of primers). 617 The expression vector for GFP-tagged Pol  constructs (full-length protein or BRCT domain) 618 was obtained by replacing the FLAG-Ku70 cDNA from the previously described pLV3-GFP-619 FLAG-Ku70 vector 2 by a linker cassette (pre-annealed Kpn2-AX-Mlu-F and Kpn2-AX-Mlu-620 R oligonucleotides) between the Kpn2I and MluI sites. The resulting pLV3-GFP plasmid was 621 then used to insert between the MluI and BcuI sites the PCR -amplified human Pol  cDNA 622 fragment described above. Expression vectors for GFP -tagged WT and mutant Pol  BRCT 623 domain (aminoacids 1-136) were obtained in a similar manner following a PCR amplification 624 of the corresponding cDNAs with PolL-Mlu-F and PolL-BRCT-Bcu-R primers and full-length 625 Pol  expression vectors as templates. 626 The lentiviral vector allowing expression of mCherry-tagged human PAXX was obtained, first, 627 by inserting between Kpn2I and MluI restriction sites of pLV3 the coding sequence of 628 mCherry, following PCR amplification from the pmCherry -NLS plasmid (a gift from Martin 629 Offterdinger; Addgene plasmid # 39319 ; http://n2t.net/addgene:39319 ; 630 RRID:Addgene_39319) with primers mCh -Kpn2-F and mCh -Mlu-R, second, by further 631 inserting between MluI and BcuI restriction sites the coding sequence of PAXX, following 632 PCR amplification from the previously described pLV3 -GFP-PAXX plasmid 2 with primers 633 PAXX-Mlu-F and PAXX-Bcu-R. 634 Ku70 lentiviral expression vector was generated by PCR amplification of human Ku70 cDNA 635 with Ku70 -Kpn2-F and Ku70 -Bcu-R primers and subsequent insertion between Kpn2I and 636 BcuI restrictions sites of the pLV3 vector. Expression vectors for mutant forms of Ku70 were 637 obtained in a similar manner following an additional step of overlap extension PCR 638 mutagenesis with the corresponding Ku70-mut-F and Ku70-mut-R oligonucleotides as mutated 639 inner primers (see below the list of primers). 640 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 13, 2024. ; https://doi.org/10.1101/2024.08.12.607588doi: bioRxiv preprint 25 Ku80 lentiviral expression vector was already described 19. Expression vectors for mutant 641 forms of Ku80 were generated following an additional step of overlap extension PCR 642 mutagenesis with the corresponding Ku80-mut-F and Ku80-mut-R oligonucleotides as mutated 643 inner primers (see below the list of primers). 644 The gap -filling reporter substrate was assembled into the pEGFP -N1 vector (Clontech) 645 following sequential insertion of various PCR products, oligonucleotide linkers and synthetic 646 DNA fragments (see Figure 4A for a detailed description). The Cpf1 and guide RNA co -647 expression plasmid used to cleave the reporter substrate was generated by inserting the pre -648 annealed gRNA-GF-F and gRNA -GF-R oligonucleotides into the Esp3I restriction sites of 649 pTE4398 (a gift from Ervin Welker ; Addgene plasmid # 74042 ; http://n2t.net/addgene:74042 650 ; RRID:Addgene_74042). 651 The pNLS -mTagBFP2 plasmid used as an internal control of transfection efficiency was 652 obtained by PCR amplification of the 2xNLS -mTagBFP2 coding sequence with mTagBFP -653 Acc65-F and mTagBFP -Mlu-R primers on the pHAGE -TO-nls-st1dCas9-3nls-3XTagBFP2 654 plasmid template (a gift from Thoru Pederson ; Addgene plasmid # 64512; 655 http://n2t.net/addgene:64512 ; RRID:Addgene_64512). The resulting PCR fragment was 656 inserted into the pEGFP -N1 (Clontech) vector backbone after modification of the multiple 657 cloning site and removal of the GFP coding sequence. 658 All oligonucleotides were purchased from Eurofins Genomics (Ebersberg, Germany). 659 Restriction and modifying enzymes (Phusion and T4 DNA Ligase) were from ThermoFisher 660 Scientific (Illkirch, France). All constructs were checked by sequencing (Eurofins Genomics). 661 662 Oligonucleotides (DNA linkers and PCR primers) 663 gRNA-GF-F agatCAACGTACGGTCTGATATGCAA gRNA-GF-R aaaaTTGCATATCAGACCGTACGTTG gRNA-LigIV-F caccGTTCAGCACTTGAGCAAAAG gRNA-LigIV-R aaacCTTTTGCTCAAGTGCTGAAC gRNA-PAXX-F caccgTGACCGACGCCGCGGAGCTT gRNA-PAXX-R aaacAAGCTCCGCGGCGTCGGTCAc .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 13, 2024. ; https://doi.org/10.1101/2024.08.12.607588doi: bioRxiv preprint 26 gRNA-PRKDC-F caccGGTACCCACCCAGCACCGCG gRNA-PRKDC-R aaacCGCGGTGCTGGGTGGGTACC gRNA-POLL-F caccgTTCACCAGCTGAGCACCCGG gRNA-POLL-R aaacCCGGGTGCTCAGCTGGTGAAc gRNA-POLM-F caccGATGGAAGAGACCTCAGCAG gRNA-POLM-R aaacCTGCTGAGGTCTCTTCCATC gRNA-XLF-F caccGGAGATTATCCAAATGACAG gRNA-XLF-R aaacCTGTCATTTGGATAATCTCC HF-Puro-F GCCTCGAGGTTTAAACTACGGgatcTCCGCcATGACaGAGTACAAGCCaACaGTG HF-Puro-R CCGCATGTTAGCAGACTTCCTCTGCCCTCGGCACCtGGCTTtCtGGTCATGCACC Kpn2-AX-Mlu-F ccggTCAGGATCTGGTAGCGGTTCCGGATCTCCTAGGTCACCCGGGTCTA Kpn2-AX-Mlu-R cgcgTAGACCCGGGTGACCTAGGAGATCCGGAACCGCTACCAGATCCTGA kpn2-T2A-Mlu-F CCGGAGGGCAGAGGAAGTCTGCTAACATGCGGTGACGTCGAGGAGAATCCTG GACCCGGGtcactcA kpn2-T2A-Mlu-R CGCGTgagtgaCCCGGGTCCAGGATTCTCCTCGACGTCACCGCATGTTAGCAGAC TTCCTCTGCCCT Ku70-Kpn2-F ctctcgTCCGGAGCCACCATGTCAGGGTGGGAGTCATATTACAAAACC Ku70-Bcu-R ctcgtcACTAGTTCAGTCCTGGAAGTGCTTGGTGAGGGCTTC Ku70-R301A-F GAAAACCAAGACCgcGACCTTTAATACAAGTACAGGCGGTTTGCT Ku70-R301A-R CTTGTATTAAAGGTCgcGGTCTTGGTTTTCACTGGTTCATTTGT Ku70-F303G-F CCAAGACCCGGACCggTAATACAAGTACAGGCGGTTTGCTTCTGC Ku70-F303G-R GTACTTGTATTAccGGTCCGGGTCTTGGTTTTCACTGGTTCATTT Ku70-T307A-F CGGACCTTTAATACAAGTgCAGGCGGTTTGCTTCTGCCTAGCGATACC .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 13, 2024. ; https://doi.org/10.1101/2024.08.12.607588doi: bioRxiv preprint 27 Ku70-T307A-R GAAGCAAACCGCCTGcACTTGTATTAAAGGTCCGGGTCTTGGTTTTCAC Ku70-L310G-F CCTTTAATACAAGTACAGGCGGTggGCTTCTGCCTAGCGATACCAAGAGGTC Ku70-L310G-R CTAGGCAGAAGCccACCGCCTGTACTTGTATTAAAGGTCCGGGTCTTGG Ku70-L310R-F CCTTTAATACAAGTACAGGCGGTcgGCTTCTGCCTAGCGATACCAAGAGGTC Ku70-L310R-R CTAGGCAGAAGCcgACCGCCTGTACTTGTATTAAAGGTCCGGGTCTTGG Ku80-R292A-F GAAGATATACAAAAAGcAACAGTTTATTGCTTAAATGATGATGATGAAAC Ku80-R292A-R GCAATAAACTGTTgCTTTTTGTATATCTTCTTTTTTTAGGGTTTTTGC Ku80-D301A-F GCTTAAATGATGATGcTGAAACTGAAGTTTTAAAAGAGGATATTATTCAAG Ku80-D301A-R TAAAACTTCAGTTTCAgCATCATCATTTAAGCAATAAACTGTTTCTTTTTG Ku80-E304A-F GATGATGATGAAACTGcAGTTTTAAAAGAGGATATTATTCAAGGGTTCCGC Ku80-E304A-R CCTCTTTTAAAACTgCAGTTTCATCATCATCATTTAAGCAATAAACTGTTTC Ku80-E304R-F GATGATGATGAAACTcgAGTTTTAAAAGAGGATATTATTCAAGGGTTCCGC Ku80-E304R-R CCTCTTTTAAAACTcgAGTTTCATCATCATCATTTAAGCAATAAACTGTTTC mCh-Kpn2-F ctcgtcTCCGGAGCCACCATGGTGAGCAAGGGCGAGGAGG mCh-Mlu-R ctcctcACGCGTCTTGTACAGCTCGTCCATGCCGCC mCh-Xba-F cctctgTCTAGAgTCCAAGGGCGAAGAAGATAATATGGC mCh-Xho-R cctgtgCTCGAGcaTCACTTGTAAAGTTCGTCCATTCCACC mTagBFP-Acc65-F ctgtctGGTACCTGGAAGTGGAAGCCCAAAGAAAAAGC mTagBFP-Mlu-R ctgtcgACGCGTTTCAATTAAGCTTGTGCCCCAGTTTGCTAGGG PAXX-Mlu-F ctcctgACGCGTTCTGGTAGCGGTTCAGGACTCAGATCCATGGACCCGCTGTC PAXX-Bcu-R cctctcACTAGTTTAGGTCTCATCGAAGTCCACGCCACCAG PolL-Mlu-F ctcctcACGCGTATGGAcCCCAGGGGTATCTTGAAGGCATTTC .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 13, 2024. ; https://doi.org/10.1101/2024.08.12.607588doi: bioRxiv preprint 28 664 665 666 667 668 PolL-Bcu-R cctctcACTAGTTCACCAGTCCCGCTCAGCAGGTTC PolL-BRCT-Bcu-R cctgacACTAGTtcaACTGGGGATGAAGATGCTGAATCCAGCTACATC PolL-dead-F GGCGACCTGTGGTGcTGTCGcCGTGCTCATCACTCACCCAGATGGCC PolL-dead-R GAGTGATGAGCACGgCGACAgCACCACAGGTCGCCTTTCCCCGTC PolL-R57A-F CATTGGACGAGCCgcGGCAGAACTCTTTGAGAAGCAGATTGTTC PolL-R57A-R CAAAGAGTTCTGCCgcGGCTCGTCCAATGCCAGTGCGCACAAC PolL-R57E-F CATTGGACGAGCCgaGGCAGAACTCTTTGAGAAGCAGATTGTTC PolL-R57E-R CAAAGAGTTCTGCCtcGGCTCGTCCAATGCCAGTGCGCACAAC PolL-L60A-F CCGGGCAGAAgcCTTTGAGAAGCAGATTGTTCAGCATGGC PolL-L60A-R CTGCTTCTCAAAGgcTTCTGCCCGGGCTCGTCCAATGCCAGTGC PolL-L60R-F CCCGGGCAGAACgCTTTGAGAAGCAGATTGTTCAGCATGGC PolL-L60R-R CTGCTTCTCAAAGcGTTCTGCCCGGGCTCGTCCAATGCCAGTGC PolL-F61G-F CGGGCAGAACTCggTGAGAAGCAGATTGTTCAGCATGGC PolL-F61G-R CAATCTGCTTCTCAccGAGTTCTGCCCGGGCTCGTCCAATGC PolL-R96A-F GAGCGAGCCCTCgcCCTTCTCAGACTACCCCAGCTGCC PolL-R96A-R GTCTGAGAAGGgcGAGGGCTCGCTCATAGTCCATGCC PolL-S116G-F GAAGTCAGCCTGGCTGgGCTTGTGCCTTCAGGAGAGGAGGCTGG PolL-S116G-R CCTGAAGGCACAAGCcCAGCCAGGCTGACTTCACCAGC .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 13, 2024. ; https://doi.org/10.1101/2024.08.12.607588doi: bioRxiv preprint 29 Supplementary Information 669 Table 1: Cryo-EM data parameters and statistics 670 671 672 673 674 675 676 677 678 679 PK-LigIV-PolL PDB: 9GD7 EMD: 51249 PK-PolL PDB: 9G9L EMD: 51156 Data collection and processing Detector Gatan K3 Gatan K3 Magnification 130k 130k Energy filter slit width (eV) 20 20 Voltage (kV) 300 300 Flux on detector (e/pix/sec) 21.24 21.24 Electron exposure on sample (e–/Å2) 51.96 51.96 Target defocus range (μm) 0.8-2.2 0.8-2.2 Calibrated pixel size (Å) 0.652 (bin 1x to 1.304) 0.652 (bin 1x to 1.304) Symmetry imposed C1 C1 Extraction box size (pixels) 320 320 Initial particle images (no.) 603711 603711 Final particle images (no.) 14964 12656 Refinement Map resolution at FSC=0.143 (Å)* 4.25 4.63 Model composition Non-hydrogen atoms 41717 39800 Protein residues 5158 4944 Nucleotides 50 47 B factor (Å2) Protein 387.65 515.18 DNA 354.09 507.46 R.m.s deviations Bond lengths (Å) 0.004 0.003 Bond angles (°) 0.827 0.803 Validation Molprobity score 2.24 2.34 Clashscore 18.8 23.7 Poor rotamers (%) 0 0.07 Ramachandran plot Favored (%) 92.53 92.44 Allowed (%) 7.25 7.34 Disallowed (%) 0.22 0.22 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 13, 2024. ; https://doi.org/10.1101/2024.08.12.607588doi: bioRxiv preprint 30 680 681 682 683 684 685 686 Figure S1: Domain organization of Pol X family DNA polymerases. The Pol  constructs used in 687 this study correspond to the full -length protein (FL) and the amino -terminal region (residues 1 -136) 688 containing the nuclear localization sequence (NLS) and the BRCT domain (BRCT). 689 690 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 13, 2024. ; https://doi.org/10.1101/2024.08.12.607588doi: bioRxiv preprint 31 691 Figure S2: Single-particle cryo-EM image processing workflow for DNA-PK +PAXX + Pol  with 692 and without LX4. Schematic showing particle picking using WARP and processing including 2D 693 classification and ab initio reconstruction using CryoSPARC. The two main classes generated with the 694 corresponding number of particles is shown and the two maps following non-uniform refinement with 695 resolutions for an FSC of 0.143 are given. Additional focused and composite maps are also shown. 696 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 13, 2024. ; https://doi.org/10.1101/2024.08.12.607588doi: bioRxiv preprint 32 697 Figure S3: Cryo-EM data of DNA-PK with BRCT domain of Pol . a) Example of 2d classes. b) 698 Local resolution map of DNA -PK dimer with BRCT domain of Pol  consensus cryo-EM map. c) 699 Angular distribution calculated in cryoSPARC for particle projections shown as a heat map of the 700 consensus map. d) FSC resolution curves and viewing distribution plot of the consensus map. e) DNA-701 PK dimer with BRCT domain of Pol  consensus cryo-EM map with masking area. f) Local resolution 702 map of DNA -PK dimer with BRCT domain of Pol  locally refined map. g) Angular distribution 703 calculated in cryoSPARC for particle projections shown as a heat map of the locally refined map. h) 704 FSC resolution curves and viewing distribution plot of the locally refined map. The colours 705 corresponding to each resolution are displayed on the specific key chart below the maps. 706 707 708 709 710 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 13, 2024. ; https://doi.org/10.1101/2024.08.12.607588doi: bioRxiv preprint 33 711 Figure S4: Cryo-EM data of DNA -PK with BRCT domain of Pol  and LX4. a) Example of 2d 712 classes. b) Angular distribution calculated in cryoSPARC for particle projections shown as a heat map 713 of the consensus map. c) Local resolution map of DNA -PK dimer with BRCT domain of Pol  and 714 Ligase IV consensus cryo -EM map. d) FSC resolution curves and viewing distribution plot of the 715 consensus map. e) and f) Consensus map with masking area 1 and 2, respectively. g) and h) Local 716 resolution of the two locally refined maps. i) and f) FSC resolution curves and viewing distribution plot 717 of local maps 1 and 2. The colours corresponding to each resolution are displayed on the specific key 718 chart below the maps. 719 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 13, 2024. ; https://doi.org/10.1101/2024.08.12.607588doi: bioRxiv preprint 34 720 Figure S5: Comparison of cryo -EM maps. A) DNA-PK + Pol  (this work) in blue. B) DNA-PK 721 (PDB: 6ZHA), C) Comparison overlay of DNA-PK + Pol  (blue) and DNA-PK (grey). 722 723 724 Figure S6: (A) Western blot on whole cell protein extracts from U2OS cells either unmodified (lane 1) 725 or constitutively expressing an shRNA against endogenous Ku70 and rescued with expression of 726 mAID-tagged Ku70 (lanes 2-4), treated or not with auxin (IAA) for 16 h. Asterisk indicates the position 727 of mAID -Ku70 signal (MW: 77.4 kDa) below that of Ku80. (B) Fluorescence micrographs of 728 U2OS/mAID-Ku70 cells (see (A)) expressing full -length GFP-tagged Pol l, in the presence (+Ku) or 729 the absence (-Ku). (C) Fluorescence micrographs of U2OS cells expressing WT or mutated GFP-tagged 730 Pol l BRCT domain. 731 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 13, 2024. ; https://doi.org/10.1101/2024.08.12.607588doi: bioRxiv preprint 35 732 Figure S7: (A) Western blot on whole cell protein extracts from U2OS/mAID-Ku70 cells (see 733 Suppl. Figure S6A) depleted of endogenous Ku70 and complemented with ectopic expression 734 of either WT or mutated forms of Ku70, as indicated. (B) Western blot on whole cell protein 735 extracts from U2OS cells either unmodified (lane 1) or expressing a doxycycline (doxy) -736 induced shRNA against endogenous Ku80 (lanes 2 -4) and rescued with expression of Ku80 737 (lane4). (C) Western blot on whole cell protein extracts from U2OS/Tet-shKu80 cells (see (B)) 738 depleted of endogenous Ku80 and complemented with ectopic expression of either WT or 739 mutated forms of Ku80, as indicated (D) Gap-filling activity assessed in HEK -293T/mAID-740 Ku70 cells knocked -down for Ku70 when indicated (+IAA), in the presence or not of 3 µM 741 DNA-PK inhibitor (+NU7441 or +NU). Results are normalized to the control condition (full) 742 and plotted as mean values of five to thirteen experiments ± SD. (E) Gap-filling activity 743 assessed in HEK-293T cells knocked-out (KO) for different NHEJ genes, as indicated. Results 744 are normalized to the NHEJ-proficient parental HEK-293T cell line and plotted as mean values 745 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 13, 2024. ; https://doi.org/10.1101/2024.08.12.607588doi: bioRxiv preprint 36 of four to five experiments ± SD. Inset: control western blot on whole cell protein extracts from 746 the different KO cells. (F) DNA junction analysis following gap -filling assay. Left: agarose 747 gel electrophoresis showing PCR products amplified around the junction following stable 748 genome integration of the reporter substrate and gap -filling reaction (see the Materials and 749

section). Three positive red -fluorescent clones (R1, R2 and R3) and three non -750 fluorescent clones as a control (C1, C2 and C3) were analyzed. Expected length of the PCR 751 fragments are indicated. Right: the DNA sequences of the PCR fragments are aligned with the 752 reporter substrate sequence using the Snapgene ® software (Dotmatics). (G) Western blot on 753 whole cell protein extracts from HEK-293T cells knocked-out (KO) for POLL, POLM or both 754 genes. When indicated, POLL KO cells were complemented with expression of ectopic WT or 755 catalytic dead Pol . (H) Gap-filling activity assessed in HEK -293T cells knocked -out (KO) 756 for POLL, POLM or both genes. Results are normalized to the parental HEK -293T cell line 757 and plotted as mean values of four to fourteen experiments ± SD. P -values from Student’s t -758 test between the indicated conditions are as follows: 293T versus KO PolL (<0.0001 ****), 759 293T versus KO PolL + Poll -WT (<0.0001 ****), 293T versus KO PolM (0.0002 ***), KO 760 PolL versus KO PolL + Poll -dead (<0.0001 ****), KO PolL+PolM versus KO PolL + Poll -761 dead (0.0001 ***). 762 763 764 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 13, 2024. ; https://doi.org/10.1101/2024.08.12.607588doi: bioRxiv preprint 37 Figure S8: (A) Western blot on whole cell protein extracts from HEK -293T cells knocked-out (KO) 765 for POL and complemented with an empty vector (EV) or expression vectors for wild -type (WT) or 766 the indicated mutants of Pol . (B) Western blot on whole cell protein extracts from HEK-293T/mAID-767 Ku70 cells depleted of endogenous Ku70 in the presence of auxin (+IAA) and rescued with ectopic 768 expression of either WT or mutated forms of Ku70, as indicated. Asterisk indicates the position of 769 mAID-Ku70 signal just below that of Ku80, which persisted after previous hybridization of the 770 membrane with anti -Ku70 antibody. (C) Gap-filling activity (red bars) or direct end -joining activity 771 (orange bars) assessed in parallel in HEK -293T cells knocked-out for POLL and complemented with 772 ectopic expression of wild-type (WT) or different mutants of Pol . Results are normalized to the WT 773 condition and plotted as mean values of four to five experiments ± SD. P -values from Student’s t-test 774 between the considered mutants, for gap -filling and direct end -joining activities, respectively, are as 775 follows: Pol -dead versus Poll -R57E (<0.0001 ****; 0.2862 ns), Poll -dead versus Poll -R57E-dead 776 (<0.0001 ****; 0.0587 ns). (D) HEK-293T/mAID-Ku70 cells were seeded in 12-well plates and treated 777 with auxin (+IAA) for the indicated time. Cell proliferation was then analyzed continuously up to 7 778 days by assessing confluence with an IncuCyte-ZOOM (Essen Bioscience). 779 780 781 782

783 1. Zhao, B., Rothenberg, E., Ramsden, D.A. & Lieber, M.R. The molecular basis and disease 784 relevance of non-homologous DNA end joining. Nat Rev Mol Cell Biol 21, 765-781 (2020). 785 2. Seif-El-Dahan, M. et al. PAXX binding to the NHEJ machinery explains functional redundancy 786 with XLF. Sci Adv 9, eadg2834 (2023). 787 3. Tang, J. et al. Role of Paralogue of XRCC4 and XLF in DNA Damage Repair and Cancer 788 Development. Front Immunol 13, 852453 (2022). 789 4. Stinson, B.M. & Loparo, J.J. Repair of DNA Double -Strand Breaks by the Nonhomologous 790 End Joining Pathway. Annu Rev Biochem 90, 137-164 (2021). 791 5. Jette, N. & Lees -Miller, S.P. The DNA -dependent protein kinase: A multifunctional protein 792 kinase with roles in DNA double strand break repair and mitosis. Prog Biophys Mol Biol 117, 793 194-205 (2015). 794 6. Chaplin, A.K. et al. Dimers of DNA-PK create a stage for DNA double-strand break repair. Nat 795 Struct Mol Biol 28, 13-19 (2020). 796 7. Chen, S. et al. Structural basis of long-range to short-range synaptic transition in NHEJ. Nature 797 593, 294-298 (2021). 798 8. Chaplin, A.K. et al. Cryo-EM of NHEJ supercomplexes provides insights into DNA repair. Mol 799 Cell 81, 3400-3409 e3 (2021). 800 9. Graham, T.G., Walter, J.C. & Loparo, J.J. Two -Stage Synapsis of DNA Ends during Non -801 homologous End Joining. Mol Cell 61, 850-8 (2016). 802 10. Liu, L. et al. Autophosphorylation transforms DNA -PK from protecting to processing DNA 803 ends. Mol Cell 82, 177-189 e4 (2022). 804 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 13, 2024. ; https://doi.org/10.1101/2024.08.12.607588doi: bioRxiv preprint 38 11. Ramsden, D.A. & Asagoshi, K. DNA polymerases in nonhomologous end joining: are there 805 any benefits to standing out from the crowd? Environ Mol Mutagen 53, 741-51 (2012). 806 12. Loc'h, J. & Delarue, M. Terminal deoxynucleotidyltransferase: the story of an untemplated 807 DNA polymerase capable of DNA bridging and templated synthesis across strands. Curr Opin 808 Struct Biol 53, 22-31 (2018). 809 13. Ghosh, D. & Raghavan, S.C. 20 years of DNA Polymerase mu, the polymerase that still 810 surprises. FEBS J 288, 7230-7242 (2021). 811 14. Pryor, J.M. et al. Essential role for polymerase specialization in cellular nonhomologous end 812 joining. Proc Natl Acad Sci U S A 112, E4537-45 (2015). 813 15. Kaminski, A.M. et al. Structural snapshots of human DNA polymerase mu engaged on a DNA 814 double-strand break. Nat Commun 11, 4784 (2020). 815 16. Mueller, G.A. et al. A comparison of BRCT domains involved in nonhomologous end-joining: 816 introducing the solution structure of the BRCT domain of polymerase lambda. DNA Repair 817 (Amst) 7, 1340-51 (2008). 818 17. Craxton, A. et al. PAXX and its paralogs synergistically direct DNA polymerase lambda 819 activity in DNA repair. Nat Commun 9, 3877 (2018). 820 18. DeRose, E.F. et al. Solution structure of polymerase mu's BRCT Domain reveals an element 821 essential for its role in nonhomologous end joining. Biochemistry 46, 12100-10 (2007). 822 19. Kefala Stavridi, A. et al. Structural and functional basis of inositol hexaphosphate stimulation 823 of NHEJ through stabilization of Ku -XLF interaction. Nucleic Acids Res 51, 11732 -11747 824 (2023). 825 20. Braithwaite, E.K. et al. DNA polymerases beta and lambda mediate overlapping and 826 independent roles in base excision repair in mouse embryonic fibroblasts. PLoS One 5, e12229 827 (2010). 828 21. Britton, S., Coates, J. & Jackson, S.P. A new method for high-resolution imaging of Ku foci to 829 decipher mechanisms of DNA double-strand break repair. J Cell Biol 202, 579-95 (2013). 830 22. Chandramouly, G. et al. Pollambda promotes microhomology-mediated end-joining. Nat Struct 831 Mol Biol 30, 107-114 (2023). 832 23. Stinson, B.M., Moreno, A.T., Walter, J.C. & Loparo, J.J. A Mechanism to Minimize Errors 833 during Non-homologous End Joining. Mol Cell 77, 1080-1091 e8 (2020). 834 24. Capp, J.P. et al. The DNA polymerase lambda is required for the repair of non-compatible DNA 835 double strand breaks by NHEJ in mammalian cells. Nucleic Acids Res 34, 2998-3007 (2006). 836 25. Bertocci, B., De Smet, A., Weill, J.C. & Reynaud, C.A. Nonoverlapping functions of DNA 837 polymerases mu, lambda, and terminal deoxynucleotidyltransferase during immunoglobulin 838 V(D)J recombination in vivo. Immunity 25, 31-41 (2006). 839 26. Vermeulen, C., Bertocci, B., Begg, A.C. & Vens, C. Ionizing radiation sensitivity of DNA 840 polymerase lambda-deficient cells. Radiat Res 168, 683-8 (2007). 841 27. Nemoz, C. et al. XLF and APLF bind Ku80 at two remote sites to ensure DNA repair by non-842 homologous end joining. Nat Struct Mol Biol 25, 971-980 (2018). 843 28. Frit, P., Ropars, V., Modesti, M., Charbonnier, J.B. & Calsou, P. Plugged into the Ku -DNA 844 hub: The NHEJ network. Prog Biophys Mol Biol 147, 62-76 (2019). 845 29. Watanabe, G. & Lieber, M.R. The flexible and iterative steps within the NHEJ pathway. Prog 846 Biophys Mol Biol 180-181, 105-119 (2023). 847 30. Robert, X. & Gouet, P. Deciphering key features in protein structures with the new ENDscript 848 server. Nucleic Acids Res 42, W320-4 (2014). 849 31. Pettersen, E.F. et al. UCSF Chimera --a visualization system for exploratory research and 850 analysis. J Comput Chem 25, 1605-12 (2004). 851 32. Emsley, P., Lohkamp, B., Scott, W.G. & Cowtan, K. Features and development of Coot. Acta 852 Crystallogr D Biol Crystallogr 66, 486-501 (2010). 853 33. Kidmose, R.T. et al. Namdinator - automatic molecular dynamics flexible fitting of structural 854 models into cryo-EM and crystallography experimental maps. IUCrJ 6, 526-531 (2019). 855 34. Afonine, P.V. et al. Real-space refinement in PHENIX for cryo-EM and crystallography. Acta 856 Crystallogr D Struct Biol 74, 531-544 (2018). 857 35. Bossaert, M. et al. Identification of the main barriers to Ku accumulation in chromatin. bioRxiv 858 (2024). 859 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 13, 2024. ; https://doi.org/10.1101/2024.08.12.607588doi: bioRxiv preprint 39 36. Seif-El-Dahan, M. et al. PAXX binding to the NHEJ machinery explains functional redundancy 860 with XLF. Sci Adv 9, eadg2834 (2023). 861 37. Cheng, Q. et al. Ku counteracts mobilization of PARP1 and MRN in chromatin damaged with 862 DNA double-strand breaks. Nucleic Acids Res 39, 9605-19 (2011). 863 38. Fu, C., Donovan, W.P., Shikapwashya-Hasser, O., Ye, X. & Cole, R.H. Hot Fusion: an efficient 864

to clone multiple DNA fragments as well as inverted repeats without ligase. PLoS One 865 9, e115318 (2014). 866 867 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 13, 2024. ; https://doi.org/10.1101/2024.08.12.607588doi: bioRxiv preprint