FLASH Irradiation Spares Lung Progenitor Cells and Limits the Incidence of Radio-induced Senescence

For decades, radiation oncologists have been delivering 2 Gy daily fractions at dose rates around 0.01 Gy/second for up to a total dose reaching the limit of tolerance of healthy tissues. We recently developed another methodology named “FLASH” that consists of delivering ≥ 10 Gy in a single microsecond pulse or in a limited number of pulses of 1–2 Gy each given in ≤ 100 ms temporal sequence (1). In mice, FLASH was found to elicit a dramatic decrease of damage to normal tissues, with protection against lung fibrosis as well as memory loss after brain irradiation while keeping the antitumoral effect unchanged (1, 2). Such protection against normal tissue injury has been confirmed in large mammals (3), thus evoking a strong interest for the FLASH methodology in anticancer radiotherapy.

Radiation-induced damage to the lung leads to an acute pneumonitis followed by abortive regenerative processes that, together with chronic inflammation, foster fibrosis development in the months following radiation (4). Pathophysiology of radiation-induced pulmonary fibrosis is complex and lung regeneration after radiation injury is poorly understood. However, authors (5, 6) have reported the presence of persistent DNA damage foci as well as senescent cells in the months following thoracic irradiation, two factors that may impair lung regenerative processes. Using a variety of analytic approaches, we identified some early and late molecular indicators of the differential response of mouse lung to FLASH versus conventional dose-rate irradiation (CONV). Data show that FLASH minimizes the induction of proinflammatory genes, facilitates radiation recovery and reduces the number of persistent DNA damage foci as well as radiation-induced senescence and fibrogenesis.

Studies were performed in accordance with the recommendations of the European Community (2010/63/UE) for the care and use of laboratory animals. Experimental procedures were specifically approved by the ethics committee of the Institut Curie CEEA-IC #118 (Authorization number APAFIS#5479-201605271 0291841 given by National Authority) in compliance with the international guidelines. Females C57BL/6J mice purchased from Charles River Laboratories at the age of 6 weeks were housed in Institut Curie animal facilities. Terc^−/− mice described previously (7, 8) were purchased from Jackson Laboratories and intercrossed until the third generation, at which time mice became sterile or pups died at the next generation.

Mice were exposed to bilateral thorax irradiation at the age of 10–12 weeks using the 4.5-MeV linear electron accelerator facility (Kinetron; ref. 9). Collimation, time-resolved fluence measurement, chemical dosimetry, depth-dose distribution, mouse immobilization, and irradiation of mouse thorax were described previously (1) except for anesthesia that was carried out with a nose cone using 2.5% isoflurane in air, without adjunction of oxygen. Other details are given under Supplementary Data S1.

For the immunofluorescent determination of radio-induced foci of the TRP53BP1 (53BP1) and γH2AX proteins at sites of DNA damage, we used two nontransformed human lung fibroblast cell lines, namely, MRC5 and IMR-90 and the human lung epithelial carcinoma cell line A-549. Details of cell culture conditions and immunofluorescence assays are given in Supplementary Data S2. For studies of differentiation and survival, we used Human Pulmonary Basal Epithelial Cells (PBEC). The preparation, irradiation, and analysis of PBECs are described under Supplementary Data S3.

The methods used for determination of the number of late, persistent foci of 53bp1 in cells isolated from mouse lungs at one week and three months postirradiation, are described under Supplementary Data S4.

The lower lung right lobe was disrupted in cold lysis buffer with electrical homogenizer and total RNA extracted using the RNEasy Plus Mini Kit (Qiagen #74134). Two micrograms of RNA was then reverse transcribed using High Capacity cDNA Reverse Transcription Kit (Applied Biosystems #4368814). Quantitative gene expression was measured by real-time qPCR using Fast Sybr Green Master Mix (Bio-Rad #4385612). Values are presented as the ratio of target mRNA to β-actin mRNA obtained using the 2^–ΔΔC_t method and normalized to nonirradiated control (fold change).

For histologic analysis, lungs were excised, washed with ice-cold PBS, gently inflated in 4% PFA under mild vacuum (25 Torr, 1 hour at room temperature) followed by 3 hours in the cold room, then embedded in paraffin and cut into 7-μm–thick slices. The preparations were stained by hematoxylin–eosin or Masson trichrome (R.A.L Diagnostics, #361350).

The proliferation rate of lung cells during the week postirradiation was assessed by 5-ethenyl-2′-deoxyuridine (EdU) incorporation. Irradiated mice and controls received daily intraperitoneal injection of EdU (4 mg/kg) and were euthanized 7 days postirradiation. The left lung was fixed in 4% PFA, embedded in paraffin, and EdU was revealed using Click-IT chemistry (Baseclick GmBH, BCK488-IV-IM-S) on 7-μm–thick sections. For EdU quantification, lung cells were dissociated from lung lobes, cytospinned, and stained for EdU incorporation as well as CD45 antibody (eBioscience #11-0451, 1/200 dilution). More than 1,000 cells were analyzed per condition.

Lungs were resected and single-cell suspensions were prepared by enzymatic dissociation and sequential filtering (100 and 40 μm) before red blood cell lysis and cell counting. Single-cell 3′-RNAseq samples were prepared using single cell V2 reagent kit and Chromium controller (10× Genomics, PN-120237). cDNA quality control was assessed by capillary electrophoresis (Bioanalyzer, Agilent) before libraries' preparation and sequencing on HiSeq 2500 (Illumina). Raw and processed data were deposited in GEO (GSE133992, for review use token sfaresuwplurtcb). Data processing and analysis was performed as described under Supplementary Data S5.

Senescence-associated β-galactosidase (SA-β-gal) activity was assessed with a commercially available assay (Cell Signaling Technology #9860) as per manufacturer's instructions. Briefly, the left lobes were fixed and incubated overnight with staining solution at 37°C. Before paraffin embedding, macroscopic images representing 80% of the lobe was taken with a stereomicroscope and, once embedded, sections were cut and counterstained with hematoxylin to visualize tissue architecture. β-gal⁺ cell clusters were counted over more than 15 lung sections per mouse.

Three-dimensional X-ray imaging of lung was performed on the cone beam computed tomography (CBCT) module of the Small Animal Radiation Research Platform (SARRP, Xstrahl), operated at 60-kVp X-ray tube voltage, 0.8-mA tube current and 6 frames/second. Animals were anesthetized with 1.5%–2% isoflurane and maintained on a PMMA vertical stand in the upright position. Using the integrated software Murislice (XStrahl), the 3D reconstruction was calculated from 1,440 projections, with isotropic voxels of size 117 μm. The reconstructed slices were analyzed with ImageJ/FIJI (ImageJ, NIH, Bethesda, MD). First, we normalized the voxel intensity of all the acquisitions on a common scale of density (NDU, Normalized Density Units) in applying a linear contrast stretching based on two ROI, in the air and in the PMMA stand, respectively. The lung was then segmented from the rest of the cavity chest with the FIJI k-means Clustering plugin. We calculated the mean NDU of the voxel intensity distribution within the lung, and the 75th percentile. The latter was used to draw the highest 25% of NDU on the whole 3D image. We determined the fibrosis status of the mouse based on the following criteria: either the total mean NDU of the lung was higher than the upper bound of the 95% confidence interval measured on the nonirradiated mice or the highest 25% of NDU were observed notably at the base of the right superior lobe (Supplementary Fig. S6). Previous experiments showed that all the mice presenting a densification of the lung at the base of the right superior lobe will inevitably develop complete lung fibrosis in the following weeks (Supplementary Fig. S6).

Data analysis was performed using StatEL (AD Software) or Prism software (GraphPad software) and, unless otherwise stated, expressed as mean ± SEM. Two-group comparisons were analyzed either by two-tailed t test for independent samples, Mann–Whitney U test or Fisher exact test (*, P < 0.1; **, P < 0.01; ***, P < 0.001).

The Histone H2AX and the TRP53-binding protein 1 (53BP1) are rapidly phosphorylated in response to DNA double-strand breaks (DSB) and aggregate into discrete foci that can be analyzed by immunofluorescence microscopy. γH2AX recruitment extends over several megabases from both sides of the DSB. 53BP1 is at the crossroad of recombination and endjoining and promotes microhomology end-joining in G₀–G₁ phase cells (10). We thus performed combined analysis of γH2AX and 53BP1 foci by immunofluorescence microscopy for a comprehensive analysis of DSBs in the early times after FLASH versus CONV irradiation in MRC5 and IMR-90 normal human fibroblasts as well as in A549 human lung adenocarcinoma cells grown on coverslips in vitro. Cells at midconfluence were given 5.2 ± 0.2 Gy delivered in the FLASH or CONV mode, fixed and examined at 5, 30, and 180 minutes postirradiation. In the three cell lines, radiation produced a sharp increase in the number of 53BP1 and γH2AX foci per nucleus culminating at 30 minutes followed by a ≥ 60% decrease at 180 minutes postirradiation (Fig. 1A) in close agreement with published data (11, 12). We then chose the 30-minute time point for systematic investigation of foci formation. The results are shown in Fig. 1B. FLASH elicited significantly less 53BP1 foci than CONV in the two fibroblast cell lines. In contrast, the response of A549 tumor cells in terms of 53BP1 foci was indifferent to the modality of radiation. The number of γH2AX foci was slightly reduced in FLASH relative to CONV conditions in the two fibroblast cell lines, but this tiny difference was not statistically significant. These results suggest that FLASH induces less initial, 53BP1-relevant DNA damage than CONV and that this difference is specific to normal cells.

Next, we used primary human pulmonary bronchial epithelial cells (PBEC) expressing TRP63 (P63), a known marker for upper airway basal stem cells (13), to determine whether cells with essential characteristics of stemness elicited a differential response to CONV versus FLASH irradiation. We used an in vitro air–liquid interface (ALI) system to establish a three-dimensional stratified and polarized bronchial epithelium to compare the effects of the two modalities of irradiation at 2 and 4 Gy on the ability of PBECs to proliferate and differentiate. Upon injury, P63⁺ basal stem cells expand and spread over the denuded basal lamina and proliferate and differentiate into ciliated and secretory cells (ref. 14; Supplementary Data and Supplementary Fig. S3 for full details). We observed that the decrease of P63⁺ PBECs relative to the total number of cells at 21 days postirradiation was less pronounced when irradiated with FLASH (Fig. 1C). The difference was significant at the 4 Gy dose point. Therefore, FLASH spares PBECs grown in vitro from radiation-induced differentiation and cell death at moderate doses of radiation.

We reasoned that induction of proliferation required to replace irreversibly damaged cells might start in the lungs in the days following radiation injury. To test this hypothesis, mice received daily injection of EdU all over the first week postirradiation (Fig. 2A). With this technique, the count of EdU⁺ cells includes all the cells that crossed S-phase, at least once during the week postirradiation. Lungs exposed to CONV irradiation showed an increase in the number of proliferating cells highlighted by the presence of clusters of 2–3 EdU⁺ cells in the lung parenchyma (Fig. 2B and C), from 4% in the nonirradiated controls to 8% after CONV irradiation. Remarkably, the number of EdU⁺ cells one week after FLASH irradiation was not significantly higher than in nonirradiated controls.

Because radiation injury elicits an inflammatory response that may cause an infiltration of proliferating immune cells, we quantified the proportion of EdU⁺ cells that were either CD45⁻ (nonimmune cells) or CD45⁺ 1 week postirradiation. While neither CONV nor FLASH irradiation induced any significant increase in the proportion of CD45⁺ EdU⁺ cells, the CD45⁻ population showed a higher proportion of EdU⁺ cells after CONV irradiation only (Fig. 2D).

To assess the molecular changes induced specifically after FLASH irradiation, we next performed single-cell RNA sequencing (sc-RNAseq) analyses 4 days postirradiation (Fig. 3A). This time point was chosen to avoid acute transcriptional changes induced by DNA damage and apoptosis and, second, because it showed the highest proportion of EdU⁺ proliferating cells over a 24-hour pulse period (data not shown). Using a droplet-based single-cell technology (Chromium, 10X Genomics), we analyzed 1,555 cells from nonirradiated mouse lung, then 2,283 and 2,189 cells after CONV and FLASH irradiation, respectively. Integrated sc-RNAseq analysis with Seurat package (15) allowed us to identify 16 distinct lung cell populations present in the three different conditions (Fig. 3B). A major decrease in the percentage of B and T cells (naïve and activated subpopulations) was observed following both CONV and FLASH irradiation (Fig. 3C).

Single-cell analysis in bulk (i.e., aggregating all cells to simulate bulk RNAseq data) failed to uncover major radiation-induced transcriptional alterations between CONV and FLASH conditions at the organ level (Fig. 3D). However, zooming-in at the cell populations level revealed molecular changes specific of FLASH-irradiated lung. FLASH was less efficient than CONV irradiation in eliciting upregulation of proinflammatory factors like Tgfb1 in the interstitial macrophage population and Cebpb, the Nf-κB canonical transcription factor in monocytes (Fig. 3E). Differential analysis of the alveolar epithelial AT2 cells, which contain stem/progenitor pools in the adult lung (16–18), showed a less pronounced decrease in Egr1 expression after FLASH than after CONV irradiation (Fig. 3F). Consistent with the known role of Egr1 in the maintenance of hematopoietic stem cell quiescence (19), calculation of the cell-cycle score of AT2 cells based on the expression of cell-cycle–specific genes with the CycleMix package, showed that FLASH is less efficient than CONV in pushing AT2 cells to exit quiescence and progress through S- and G₂-phase (Supplementary Fig. S5).

In conclusion, FLASH proved to be less efficient than CONV irradiation in inducing expression of inflammatory genes and exit from quiescence in lung AT2 cells.

Persisting focal accumulation of the adaptor protein 53bp1 at sites of chromosomal damage has been observed in liver, brain, and lung tissues long after exposure to a sublethal dose of radiation (20). These foci have been associated with the DNA segments with chromatin alterations reinforcing senescence (DNA-SCARS; ref. 5).

We sought to determine by immunofluorescence the fate of such persistent 53bp1 foci in cells isolated from irradiated lung at 1 week and 3 months postirradiation versus nonirradiated lungs collected from mice of the same age (Fig. 4A). With regard to the nonirradiated condition, after one week postirradiation, cells exposed to CONV and FLASH showed a higher damage signal in both the number of cells with 53bp1 foci and the number of foci per cell, which were significantly higher after CONV (Fig. 4B). After 3 months, the number of foci per cell decreased in FLASH-irradiated cells while it increased in the CONV condition. Moreover, while the number of cells bearing DNA damage foci remained stable after 3 months following FLASH, this number significantly increased in the CONV condition, suggesting that DNA damage continues to accumulate in the absence of exogenous injury.

With rare exceptions, 53bp1 foci were arranged in multiplets consisting of individual foci paired in close proximity to each other and bound to the interior face of the nuclear membrane. The multiplicity changed with time (Fig. 4C). At 1-week postirradiation, the percentage of nuclei bearing singlets was not different from that seen in nonirradiated controls. In contrast, nearly half of the nuclei presented with foci assembled into doublets, quadruplets, or sextuplets. The ratio of quadruplets to doublets was slightly higher after CONV than FLASH irradiation and, as said above, the difference between the two modalities in terms of the number of foci per nucleus, was significant. The scene was profoundly altered at 3 months postirradiation. In FLASH-irradiated lungs, the count of nuclei presenting 53bp1 foci was not very different from that measured at 1-week postirradiation, but the total number of foci was significantly smaller due to a large decrease of higher-multiplicity foci, namely, sextuplets, and quadruplets. While the proportion of doublets did not change very much with time, the proportion of singlets was 10-fold higher than at 1-week postirradiation. A similar interconversion from high to low multiplicity was observed in nuclei from CONV-irradiated lungs (Fig. 4C). In the CONV group, however, the number of nuclei-bearing foci as well as the total foci count was significantly higher than at 1-week postirradiation, again suggesting that CONV-irradiated cells accumulate spontaneous DNA damage at distance from irradiation.

Replicative (stem) cell senescence is known to play a leading role in the induction of lung fibrosis (6). To test the hypothesis that FLASH may prevent, at least in part, the induction of senescent cells in the lung after irradiation and therefore limit the development of pulmonary fibrosis, we analyzed the presence of senescent clusters by detecting SA-β-gal activity in lung tissue 4 months postirradiation, that is, at the onset of fibrosis. Whole-mount SA-β-gal staining of the superior lobe revealed the presence of multiple senescent subpleural clusters after CONV irradiation, whereas only one or two clusters were visible after FLASH by whole mount as well as on histologic sections (Fig. 5A and B; Supplementary Fig. S7). Quantitative analysis of the number of SA-β-gal clusters per section indicated that there were half as many SA-β-gal⁺ senescent clusters 4 months after FLASH than after CONV irradiation (Fig. 5C).

To confirm this observation, we quantified by qRT-PCR the mRNA expression of Cdkn2a (p16^INK4A), Serpine1 (PAI-1) and Mmp-2, classical markers of senescence, and proinflammatory secreted associated senescence proteins (SASP) at the time of fibrogenesis, that is, 3 to 5 months postirradiation. In agreement with the histologic results, we observed that the expression of these three senescence markers was comparatively lower after FLASH than after CONV irradiation. This difference was particularly significant for p16^INK4A mRNA, which underwent strong upregulation 4 months after CONV radiation exposure (Fig. 5D).

Telomere length and proper maintenance are critical to preserve the replication potential of stem/progenitor cells (21). To determine whether the status of telomeres plays a role in normal tissue sparing in the FLASH mode, wild-type and Terc^−/− mice harboring critically short telomeres were exposed to CONV or FLASH irradiation and followed for the development of lung fibrosis by periodical CT scan in the months postirradiation. Mice developed a densification of the lung parenchyma evocative of fibrosis (Fig. 6A and B; Supplementary Fig. S6) and were euthanized at 4 and 5 months postirradiation. Masson trichrome staining of lung sections showed subpleural areas with thickening of alveolar septa and collagen invasion characteristic of lung fibrosis (Fig. 6C). One wild-type FLASH-irradiated mouse out of 8 presented signs of fibrosis while 10 of 11 CONV-irradiated wild-type mice were committed to fibrosis (Fig. 6B), as expected (1). However, 12 of 12 CONV-irradiated and 7 of 8 FLASH-irradiated Terc^−/− mice developed pulmonary fibrosis 4 and 5 months postirradiation. These results indicated that Terc^−/− mice are not spared from fibrosis by FLASH irradiation.

As lack of telomerase induces complex mouse phenotypes (22), we determined whether Terc^−/− mice may be primed toward senescence by evaluating SA-β-gal staining on nonirradiated Terc^−/− lung sections as well as after CONV and FLASH irradiation. In age-matched nonirradiated Terc^−/− lung sections, we did not find any SA-β-gal cluster, whereas, as observed in CONV-irradiated wild-type mice, several SA-β-gal clusters were found in CONV- and FLASH-irradiated Terc^−/− mice (Fig. 6D).

The goal of this study was to pave the way for identification of the mechanistic basis for the protective effect of FLASH to radiation-induced toxicity. Early DNA damage response was investigated in vitro using immunofluorescence characterization of radio-induced γH2AX and 53BP1 foci in normal human lung fibroblasts. The difference between FLASH and CONV with γH2AX as an endpoint was not significant. In contrast, a significant difference was observed at the level of 53BP1 foci. This suggests that the bulk of DNA double-strand breaks does not depend very much on the radiation mode while 53BP1 recognizes specifically a subset of DNA damage that is less abundant in the FLASH mode. FLASH also elicited a higher survival of P63⁺ human PBECs. In mouse lung, we observed a dramatic downregulation of all major indicators of tissue damage including (i) reduced proliferation after radiation injury; (ii) attenuated downregulation of Egr1 in AT2 cells revealed by sc-RNAseq, suggesting a reduced mobilization of the AT2 stem/progenitor pool after radiation injury; (iii) attenuated proinflammatory responses; (iv) a reduction of 53bp1 persistent nuclear foci in lung parenchymal cells; and finally (v) a reduced number of subpleural clusters of senescent cells. Furthermore, the protective effect of FLASH against radiation-induced fibrosis was lost in Terc^−/− mice that present critically short telomeres, thus suggesting that a full potential of replication, most likely of the progenitor cell population, is required for the FLASH effect in vivo. However, even though lungs from nonirradiated Terc^−/− mice did not present any sign of spontaneous senescence, more lung phenotypic analysis is required to confirm this conclusion.

It was recently shown that whole-brain FLASH irradiation preserves neurogenesis in the hippocampus and spares mice from the memory loss subsequent to CONV radiotherapy (2, 23). Actually, radiation injury is known to impact stem cell functions and their regenerative potential in different organs such as the brain and the intestine (24–26). In the lung, recent work from Stripp and colleagues showed that radiation reduces the in vitro colony-forming ability of epithelial progenitor cells from distal airways (27, 28). We recently demonstrated that NOTCH signaling promotes the survival of irradiated P63⁺ basal stem cells through activation of the 53BP1/DNA damage response pathway (29). Whether this also impacts on the protective response of FLASH is not known. Our results indicate that CONV thoracic irradiation leads to an acute proliferation phase in the days following radiation injury and that this pressure to repopulate the irradiated organ is reduced after FLASH irradiation. In line with this result, our sc-RNAseq analysis reveals that after CONV irradiation, the expression of Egr1, a transcription factor whose downregulation is necessary to trigger proliferation of hematopoietic stem cells (19), is reduced specifically in the AT2 compartment which is known to contain a stem/progenitor pool (16, 17, 19). This supports the idea that, owing to an intrinsically lower level of damage, FLASH does not push lung progenitor cells to replicate to the same extent as CONV irradiation, thus preserving the progenitor pool of cells from exhaustion.

The long-term consequences of radiation injury have been associated with persistent foci of the DNA damage protein 53bp1 concomitantly with senescence of the stem/progenitor cells (6, 20, 30) including secretion of IL6, an important SASP component (5). FLASH-irradiated cells isolated from whole lung also presented persistent 53bp1 foci and some occasional clusters of senescent cells. However, their number was significantly lower than in CONV-irradiated lungs and declined with time, while additional 53bp1 foci accumulated with aging in CONV-irradiated lungs. In line with our results, a recent study showed that senolytic treatment kills senescent AT2 cells and reverses radiation-induced pulmonary fibrosis, suggesting a causal link between the presence of senescent cells at late stages after irradiation and fibrogenesis (30).

In conclusion, our findings uncover a first mechanistic clue for the beneficial FLASH effect through normal tissue sparing. These results lend support to the design of preclinical studies investigating the therapeutic benefit of FLASH in cancer treatment.

No potential conflicts of interest were disclosed.

Conception and design: C. Fouillade, A. Londoño-Vallejo, P. Verrelle, V. Favaudon

Development of methodology: C. Fouillade, S. Curras-Alonso, L. Giuranno, S. Heinrich, A. Beddok, M. Vooijs, A. Londoño-Vallejo, V. Favaudon

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): C. Fouillade, S. Curras-Alonso, E. Quelennec, S. Heinrich, S. Bonnet-Boissinot, A. Beddok, S. Leboucher, S. Baulande, M. Vooijs, A. Londoño-Vallejo, V. Favaudon

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): C. Fouillade, S. Curras-Alonso, L. Giuranno, E. Quelennec, S. Heinrich, S. Bonnet-Boissinot, A. Beddok, H.U. Karakurt, M. Vooijs, V. Favaudon

Writing, review, and/or revision of the manuscript: C. Fouillade, S. Curras-Alonso, S. Heinrich, M. Vooijs, P. Verrelle, M. Dutreix, A. Londoño-Vallejo, V. Favaudon

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M. Bohec, P. Verrelle, V. Favaudon

Study supervision: C. Fouillade, A. Londoño-Vallejo, V. Favaudon

The authors wish to thank Marie-Noëlle Soler and Claire Lovo from the PICT-IBiSA Orsay Imaging facility of Institut Curie and Christophe Alberti, Frédéric Bertrand, Elodie Belloir, Virginie Dangles-Marie, Casper Caspersen and Isabelle Grandjean from the animal core facility of Institut Curie. The contribution of the bioinformatics core facility at U900 Inserm-Institut Curie is gratefully acknowledged. Thanks are due to Dr. Marie-Catherine Vozenin (Radio-oncology laboratory, Department of radio-oncology, Lausanne University Hospital, Switzerland) and Paul-Henri Roméo (IRCM-SCSR, CEA-DSV, Fontenay-aux-Roses, France) for kind support and helpful discussion. We also thank the dedicated work of Ikbal Rhimi on CT analysis. This work was funded by the Agence Nationale de la Recherche (ANR-14-CE36-0008-02 program), the Comprehensive Cancer Center “SIRIC” program of Institut Curie (Grant INCa-DGOS-4654) and the Nanotherad IDEX (Paris-Saclay University). High-throughput sequencing was performed by the ICGex NGS platform of Institut Curie supported by the grants ANR-10-EQPX-03 (Equipex) and ANR-10-INBS-09-08 (France Génomique Consortium) from the Agence Nationale de la Recherche (Investissements d'Avenir program) and by the Cancéropole Ile-de-France. This project has received funding from the European Union's Horizon 2020 Research and Innovation Program under the Marie Skłodowska-Curie grant agreements No. 666003 (S. Curras-Alonso and H.U. Karakurt PhDs, Institut Curie CO-FUND/IC3i program) and 642623 (L. Giuranno PhD, ITN RADIATE Grant). We are also grateful to the Prolific association for their financial support through a Graine de Chercheur grant to C. Fouillade. Generous financial aid from Mrs. Marie Mahler (Gisikon, Switzerland) is gratefully acknowledged.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.