Introduction

Exogenous and endogenous stresses cause various types of DNA damage, among which DNA double-strand breaks (DSBs) are the most deleterious1. To eliminate the threat, life has evolved two general pathways to repair DSBs: homologous recombination (HR) and non-homologous end-joining (NHEJ)2,3. HR is active only in the S/G2 phases following DNA replication4, whereas NHEJ functions throughout the cell cycle5,6,7. NHEJ consists of two sub-pathways, canonical NHEJ (c-NHEJ) and alternative NHEJ (alt-NHEJ). C-NHEJ plays a predominant role in the rapid repair of DSBs, whereas alt-NHEJ mainly participates in the slow repair of DSBs8,9,10. Alt-NHEJ, also referred to as microhomology-mediated end-joining (MMEJ)11,12, is a pathway that utilizes microhomology for the rejoining of DSBs13. However, the microhomology-mediated rejoining of DSBs across the genome often leads to chromosomal translocation, fostering genetic instability that is critical for genome plasticity, evolution, and carcinogenesis11,14. A mechanism similar to alt-NHEJ is present in prokaryotes, known as alternative end-joining (A-EJ)15,16. A-EJ operates independently of the key components of c-NHEJ, namely Ku and Ligase 4. In c-NHEJ-deficient Escherichia coli, A-EJ functions autonomously from HR, but hijacks certain components such as the RecBCD complex, and its DNA ligation involves Ligase-A instead of ligase-B17. Despite limited genetic conservation between A-EJ and alt-NHEJ, they share crucial features including the involvement of their components in DNA replication and recombination, independence from Ku and Ligase 4, a strong reliance on microhomology, and utilization of NAD+-dependent Ligase17.

As a backup pathway, alt-NHEJ is typically employed in cases of HR or c-NHEJ deficiencies, while other factors facilitating alt-NHEJ pathway choice remains to be elucidated. Initiation of alt-NHEJ relies on end resection that is correlated with DSB complexity11,18. DSB complexity is defined by the nature and number of damage in close proximity and can be classified into distinct levels18. Restriction endonucleases generate a simple form of DSBs without chemically altering any constituent moieties, whereas physical or chemical agents generate complex DSBs with more damage in their ends and/or close proximity19. Endogenous non-B DNA structures such as cruciform DNA can also lead to the formation of complex DSBs20. Under laboratory conditions, ionizing radiation (IR) is commonly used for generating complex DSBs, with low-linear energy transfer (LET) irradiation like X- and γ-rays for low-complexity DSBs and high-LET heavy-ion irradiation for high-complexity DSBs19,21. However, the role of DSB complexity in initiating alt-NHEJ repair has been rarely characterized.

Mechanisms of DSB repair are primarily investigated by measuring the repair kinetics of DSBs in various genetically deficient backgrounds22,23,24. This is typically carried out using site-specific reporter systems of plasmids and chromosomes, which are suitable for endonuclease-generated DSBs with simple or chemically modified ends22,24. The alt-NHEJ repair of simple DSBs has been extensively characterized using thus reporter systems in prokaryotes and eukaryotes, revealing several molecular features of DSB rejoining, such as overwhelming microhomology usage, longer sequence deletion, and absence of DNA synthesis17,22. However, due to the stochastic distribution of complex DSBs arising from IR or other stresses throughout the genome, there is currently no available approach to molecularly characterize the alt-NHEJ repair of complex DSBs.

Here, we genetically engineered a strain of E. coli with a single copy of lacI::lacI::lacO::amp integrated into its genome, where the expression of the amp gene is activated by the mutation of lacO element, synchronous mutations of the two lacI genes, or translocation-mediated fusion of the amp gene with other coding genes. Owing to the abundance of coding sequences in the E. coli genome, our reporter system is capable of efficiently detecting A-EJ mediated amp translocation. Utilizing this unique reporter system, we characterized A-EJ repair of DSBs generated by different types of IRs (Fig. 1A).

Fig. 1: Particle irradiation initiates translocation of the amp gene in Escherichia coli.
figure 1

A An overview of the workflow, where n is the total number of AmpR clones collected in this study (Created by BioRender.com). B Schematic diagram of the translocation reporter system. Under normal conditions, lacI (green explosive shape) specifically binds to the lac operator (lacO), hindering lac promoter and suppressing amp gene expression. Activation of amp gene expression can occur through three patterns: I, lacO mutation; II, synchronous mutations of the lacI genes; III, translocation of the amp gene. C Track structures of proton, carbon, and argon irradiation in water on X-Z plane. Track structures on X-Y plane are shown in Supplementary Fig. 1. D Mutation rates of Amp resistance (AmpR) induced by the indicated irradiation types (n = 5; ns, P ≥ 0.05; **P < 0.01). E Percentages of the amp translocation in AmpR clones (n = 3; ns, P ≥ 0.05; **P < 0.01). F Proportional distribution of the three types of amp activation patterns in AmpR clones.

Full size image

Results

Induction of amp gene translocation by complex DSBs

In the reporter strain of E. coli K12 MG1655 (MG1655TR), the restoration of Amp resistance (AmpR) can occur through three different types of genetic alterations: synchronous mutations of the two lacI genes, mutation of the lacO, and ectopic expression of the amp gene (i.e., translocation), as schematically shown in Fig. 1B. The MG1655TR strain was subjected to γ, proton, carbon, and argon irradiation with various LETs and track structures (Fig. 1C and Supplementary Fig. 1). To emphasize the radiation quality, a dose of 100 Gy was applied for all types of irradiation. As shown in Fig. 1D, three types of particle irradiation all led to an increased proportion of AmpR clones compared with γ-irradiation (in all cases, P < 0.05); however, no significant difference was observed among the particle irradiation (in all cases, P > 0.05) (Fig. 1E and Supplementary Table 1). Next, we detected 353 AmpR clones generated by γ-irradiation and found that all the amp genes remained at their original loci. However, the amp genes were removed from their original loci in (9.83 ± 1.89) % (28/288), (80.33 ± 8.96) % (380/475), and (91.66 ± 7.51) % (1224/1344) of AmpR clones generated by proton, carbon, and argon irradiation, respectively (Fig. 1E and F). Based on the AmpR of these clones, it was concluded that the amp genes were not lost but were translocated to other genomic loci, correctly fusing into local encoding genes. Therefore, heavy-ion irradiation led to a higher frequency of translocation than proton irradiation (in both cases, P < 0.01), suggesting that high-complexity DSBs might preferentially initiate translocation. Moreover, we examined two other types of genetic alternations in the remaining AmpR clones. As shown in Fig. 1E, γ and proton irradiation caused a higher proportion of point mutations in the lacO, whereas carbon and argon irradiation led to more mutations in the two lacI genes, including synchronous point mutations, recombinant point mutations, and structural deletions (Supplementary Table 2).

Role of A-EJ in translocation formation

Next, we determined the translocation destinations of the amp genes. Using thermal asymmetric interlaced polymerase chain reaction (TAIL-PCR), we located 21, 133, and 134 translocation targets generated by proton, carbon, and argon irradiation, respectively. Molecular analysis showed that the amp genes were translocated across the whole genome (Fig. 2A and Supplementary Table 3), and no obvious translocation hotspot was observed for each type of irradiation, with independent translocation target in each AmpR clone (Supplementary Table 3). The amp genes were fused into the local encoding genes in three patterns: upstream fusion, downstream fusion, and direct fusion with the promoter. In the three types of particle irradiation, the amp genes were largely fused into the upstream of targeted genes (Fig. 2B). According to the integrality of the amp gene in new loci, fusions are classified into two patterns: intact fusion, where no deletion occurs at the 5′ end of the amp gene, and missing fusion, where a small deletion occurs at the 5′ end of the amp gene. In the three types of particle irradiation, most of the fusions with targeted genes showed the second pattern (Fig. 2C). Among all the translocation targets, 96% (277/288) of junctions harbored 2–12 bp microhomologous sequences, 56% (154/277) of which are ≥ 3 bp (Fig. 2D and Supplementary Table 3). Relative to proton irradiation, translocations generated by carbon and argon irradiation tended to utilize shorter microhomology (2 bp). Moreover, the distributions of GC content in microhomologous junctions were not significantly different among the three types of particle irradiation (in all cases, P > 0.05; Fig. 2E). Surprisingly, 2–21 bp of non-templated additions were observed in translocation junctions, with 2/133 for carbon irradiation and 1/134 for argon irradiation (Supplementary Table 3). The RecBCD complex is responsible for the end resection in A-EJ17. We further detected the proportion of amp translocation in the recB- mutant after carbon irradiation, and found a significantly reduced frequency compared with that in wild-type E. coli (P < 0.05) (Fig. 2F and Supplementary Table 1). We also conducted whole-genome sequencing on 3 AmpR strains induced by carbon irradiation and observed a novel occurrence of translocations/deletions compared to control strains, although there was no statistically significance in the proportion of mutation types (P > 0.05; Fig. 2G, H). The DSBs at these loci were rejoined using microhomology (Supplementary Table 3). The extensive usage of junctional microhomology suggests an important role of A-EJ in translocation formation.

Fig. 2: Alternative end-joining (A–EJ)-mediated translocation of the amp gene.
figure 2

A Circos plots of genome-wide translocation landscape of the amp gene in particle irradiation. Circles from the outside to the inside are GC skew (green), GC ratio (blue to red), gene density (blue to red), and chromosome (gray) with the micro-homology of junctions (red barplot). Individual amp translocations are represented as arcs originating from the llla (lacI::lacI::lacO::amp) site and terminating at the partner site. B Proportional distribution of the fusion patterns of the amp gene into the encoding genes in translocation targets, classified as upstream fusion, downstream fusion, and promoter fusion. C Proportional distribution of the intact fusion and missing fusions of the amp gene with encoding genes in translocation targets. D Proportional distribution of the lengths of micro-homologous sequences in translocation junctions (ns, P ≥ 0.05; *P < 0.05; **P < 0.01). E Violin plot of the proportional distribution of GC content in micro-homologous junctions, which includes median (dark gray dotted line) and the lower and upper quantiles (light gray dotted line) of the data (ns, P ≥ 0.05). F Proportion of amp translocation in recB- mutant following carbon irradiation (**P < 0.01). G Circos plots of genome-wide mutation landscape in AmpR strains subjected to carbon irradiation. Translocations/deletions (SV) are indicated by red fonts, while InDel and SNP are represented by black fonts. H Proportional distribution of three types of genome-wide mutations in bacteria irradiated by carbon ions (ns, P ≥ 0.05).

Full size image

Suppressive effect of homology on translocation formation

The role of A-EJ in translocation formation raises a question regarding the function of HR in repairing complex DSBs. This question can be addressed by providing additional homologous substrates for the amp gene and its flanking sequence. For this purpose, the strain MG1655TR was transfected with two different plasmids: psgRNA that harbors a 44 bp sequence of the amp gene, and pl4sgRNA that includes an additional 465 bp sequence of the lacZ gene (Fig. 3A). Under carbon irradiation, the existence of homologous sequences significantly reduced the mutation rate of AmpR (in both cases, P < 0.05) (Fig. 3B and Supplementary Table 1). We molecularly analyzed 91 AmpR clones from the strain MG1655TR (psgRNA) and 68 AmpR clones from the strain MG1655TR (pl4sgRNA) and found that their translocations were completely suppressed, and the activation of AmpR was largely due to mutations in the lacO and lacI genes (Fig. 3C). Similarly, in the proton-irradiated MG1655TR (psgRNA) and argon-irradiated MG1655TR (pl4sgRNA), no translocation was detected in 153 AmpR clones, except for mutations in the lacO and lacI genes (Fig. 3C). These results suggest that when homology exists, HR may play a predominant role in repairing complex DSBs.

Fig. 3: Suppressive effect of homology on amp translocation.
figure 3

A Translocation reporter strains with psgRNA plasmid that harbors a 44 bp homologous sequence of the amp gene (top, 44 bp sequence indicated in orange) and pl4sgRNA that harbors the 44 bp sequence and an additional 465 bp flanking sequence (bottom, 465 bp sequence indicated in blue). B Carbon irradiation-induced mutation rates of AmpR in the strains MG1655TR (psgRNA) and MG1655TR (pl4sgRNA) (n = 5; **P < 0.01). C Proportional distribution of the three patterns of the amp activation in AmpR clones, in which the percentage of the amp translocation is zero.

Full size image

Competition between A-EJ and HR at low-homology DSB loci

The strains MG1655TR (psgRNA) and MG1655TR (plfsgRNA) contained multiple homologous copies of the amp gene and its flanking sequence. We speculated that A-EJ might compete with HR in the repair of complex DSBs at low-homology loci. In the reporter system, the two lacI genes were linked by a 405 bp spacer in the same orientation (Fig. 1A), providing a single homologous substrate for each other. In the above experiments, we found that the mutations of the two lacI genes occurred mainly in two types of patterns: 1) point mutation of one lacI gene after the recombinant deletion of the other; and 2) point mutation of one lacI gene after the translocation/direct deletion of the other (Fig. 4A).

Fig. 4: Competition between homologous recombination (HR) and alternative end-joining (A–EJ) in the repair of double-strand breaks (DSBs) at low-homology loci.
figure 4

A Schematic diagram of the mutation patterns of the lacI genes; the point mutation in one lacI gene is usually accompanied with recombinant deletion (I) and translocation/direct deletion (II) of the other lacI gene. B Proportional distribution of lacO mutation, lacI recombinant deletion, and lacI translocation in AmpR clones of the strain MG1655TR (psgRNA). C Circos plots of the genome-wide translocation landscape of the lacI gene in various types of irradiation. Circles from the outside to the inside are GC skew (green), GC ratio (blue to red), gene density (blue to red), and chromosome (gray) with the micro-homology of junctions (red barplot). Individual lacI translocations are represented as arcs originating from the llla (lacI::lacI::lacO::amp) site and terminating at the partner site. D Proportional distribution of various lengths of micro-homologous sequences in translocation junctions (ns, P ≥ 0.05). E Proportional distribution of lacO mutation, lacI recombinant deletion, and lacI translocation in AmpR clones of the strain MG1655TR (pl4sgRNA), which has the hyper-active λ-Red recombinase gene. F Proportional distribution of translocation, recombination and lacO mutations in recB-deficient bacteria irradiated by carbon ions (**P < 0.01).

Full size image

To evaluate the competition between HR and A-EJ, we examined the ratio of recombination and translocation of the lacI genes in AmpR clones of the strain MG1655TR (psgRNA). Under γ-irradiation, we did not detect translocation of the lacI gene, but observed 35% recombination (125/353). However, we observed 31% recombination (25/81) and 7% translocation (6/81) under proton irradiation, 46% recombination (42/91) and 22% translocation (20/91) under carbon irradiation, and 29% recombination (21/72) and 36% translocation (26/72) under argon irradiation (Fig. 4B and Supplementary Table 1). The distribution of the translocation targets in the genome is shown in Fig. 4C and Supplementary Table 3. Expectedly, 94% (49/52) of translocation junctions harbored microhomology (Fig. 4D and Supplementary Table 3), indicating the involvement of A-EJ in the translocation of the lacI gene. We further examined their ratio in the strain MG1655TR (pRedCas9), in which the pRedCas9 harbored the λ-Red recombinase gene and two homologous sequences of the amp gene (25 bp and 22 bp). The homologous sequences were mainly used to suppress the translocation of the amp gene. The hyperactivity of recombinase enhanced the homologous recombination of the lacI genes, but reduced their translocation, with 87% recombination (85/98) and 2% translocation (2/98) in carbon irradiation and 84% (69/82) recombination and 2% translocation (2/82) in argon irradiation (Fig. 4E and Supplementary Table 1). The RecBCD complex is a key component for both HR and A-EJ. In recB-deficient bacteria, translocation events were completely suppressed, and recombination events also exhibited a 3-fold decrease. Moreover, there was a significant different in the proportion of mutation types between wild-type and recB strains (P < 0.01) (Fig. 4F and Supplementary Table 1). These findings suggest a competition between A-EJ and HR when DSBs occur at low-homology loci.

Discussion

Based on the unique reporter system, we here investigated A-EJ-mediated translocation following various types of IR. The translocation of the amp gene occurs at a lower frequency (10−7) compared with that in site-specific reporter system of E. coli (10−5 for both plasmid and chromosomal DSBs)17. While our system allows translocation targets to be distributed across the whole genome, only the amp gene that is properly translocated to an encoding gene can restore AmpR (Fig. 2A). Thus, the actual level of translocation should be higher than that indicated by the current results. However, this speculation might be somewhat arbitrary because translocation preferentially targets the transcribed genomic loci25. While the IR-generated complex DSBs are more favorable for A-EJ than endonuclease-generated ones11,18, their stochastic distribution also reduces the possibility of DSBs occurring at or near translocation-initiating loci. This might be another reason that our reporter system has a lower translocation frequency than site-specific reporter systems.

In this study, three types of particle irradiation were used to generate complex DSBs, but they made no significant difference in the mutation rates of AmpR (Fig. 1D). AmpR can be activated through three types of mutation patterns: amp translocation, lacO mutation, and synchronous mutations of the lacI genes. Our results show that proton irradiation preferentially mutated the lacO and lacI genes, whereas carbon and argon irradiation mainly generated the translocation of the amp gene (Fig. 1F). The preference of irradiation quality for mutation patterns and their common role to activate AmpR might be a reasonable explanation for the radiation quality-independent activation of AmpR. On the other hand, carbon irradiation has a higher LET (31.46 keV/μm) than argon irradiation (LET, 11.24 keV/μm), but generates a lower proportion of translocations (Fig. 1E). In addition to LET, another major characteristic of heavy-ion irradiation is its radial energy transfer, also known as the track structure. Argon irradiation has a wider radial dose distribution than carbon irradiation (Fig. 1C) and can generate DSBs in a larger volume. According to the ‘contact first’ model for the formation of rejoining-dependent translocations26,27, the spatial distribution of argon irradiation-induced DSBs may be more favorable than that of carbon irradiation-induced DSBs for translocation formation across a wide genome range.

The usage of junctional microhomology is a prominent feature in chromosomal translocation mediated by A-EJ and alt-NHEJ13,17. Single-strand annealing (SSA) has also been reported to utilize microhomology with a length > 20 bp13. Additionally, c-NHEJ occasionally leads to chromosomal translocation by utilizing shorter microhomology13. However, E. coli lacks Ku- and Ligase-D-like proteins and is widely acknowledged as being devoid of c-NHEJ activity, making it an ideal negative control for c-NHEJ experiments in other bacteria17,28,29. Taking into consideration these factors along with the fact that the length of microhomology in our experiments is < 12 bp, it should be acknowledged that in E. coli A-EJ may be responsible for most microhomology-dependent repair events.

The deficiency of c-NHEJ in E. coli also enables us to specifically characterize A-EJ-mediated translocations. Using the E. coli reporter system, we examined the molecular feature of translocation junctions initiated by complex DSBs, such as the overwhelming usage of microhomology (≥2 bp). The distribution of microhomology length is similar between the amp and lacI translocations (Figs. 2D and 4D), indicating that the sequence context at translocation-initiating loci does not significantly affect the usage of microhomology during A-EJ repair. Notably, most of microhomology used in the rejoining of endonuclease-generated DSBs are <3 bp17, whereas approximately half of microhomology used in the rejoining of complex DSBs are ≥ 3 bp, with 56% for the amp translocation and 51% for the lacI translocation (Figs. 2D and 4D). On the other hand, non-templated nucleotide addition is a hallmark of c-NHEJ, and its length is 2–5 bp in prokaryote Mycobacterium15. Interestingly, A-EJ has also been found to utilize some non-templated nucleotide additions (2–21 bp), albeit at a low frequency (~1%) (Supplementary Table 3). These results suggest that A-EJ might employ additional functions to facilitate the repair of high-complexity DSBs. However, it is also likely that alternative repair pathways or their associated components could potentially participate in this process.

While there is limited genetic conservation between A-EJ and alt-NHEJ, they do exhibit similarities17. In the mammalian cells, alt-NHEJ is usually considered to be active only when HR or NHEJ fails, such as in the cancer12,13,14. Thus, alt-NHEJ choice in normal mammalian cells is under debate owing to its tight suppression in wild-type genetic contexts. There has been a general phenomenon that for c-NHEJ-deficient mammalian cells, the relative biological effectiveness (RBE) for inactivation does not increase with the LET of irradiation30,31,32,33,34,35. This indicates that c-NHEJ does not participate in the repair of high-complexity DSBs, at least in no significant way. This is further supported by the fact that resection is essential for the repair of high-complexity DSBs in all phases of the cell cycle in mammalian cells36. Thus, it is likely that mammalian alt-NHEJ might be also highly proficient in repairing high-complexity DSBs, and can bypass the suppression by c-NHEJ and/or HR in wild-type genetic contexts in face with high-complexity DSBs. Some early work supports this speculation. For example, EM9 cells (CHO) deficient in Xrcc1, a component of alt-NHEJ, are more sensitive to carbon irradiation than γ-irradiation35, and G1-type chromosome breaks are efficiently rejoined in normal human fibroblasts (HFL III) during 6–24 h after carbon irradiation34. These findings indicate a necessary role for alt-NHEJ in the repair of high-complexity DSBs in wild-type cells. While end resection-dependent c-NHEJ contributes to chromosomal translocation following DSBs generated by X-irradiation and designer nucleases22,23, carbon irradiation-induced RBE for inactivation is approximately 1 in Ligase 4-deficent 180BR cells (human) and Ku80-deficient xrs6 cells (CHO)32. This indicates a dispensable role of end resection-independent and end resection-dependent c-NHEJ in repairing high-complexity DSBs in normal cells. Therefore, we propose that alt-NHEJ could be activated by high-complexity DSBs in normal mammalian cells. In addition to heavy-ion irradiation, there exists endogenous high-complexity DSBs, such as cruciform structure-derived DSBs19. Alt-NHEJ has been reported to participate in the repair of such DSBs37.

Overall, the present study has successfully established a reporter system capable of detecting A-EJ repair of complex DSBs. We also demonstrated that the complexity level of DSBs plays a critical role in determining A-EJ pathway choice, as schematically shown in Fig. 5. These findings significantly contribute our understanding of A-EJ/alt-NHEJ pathway choice and provide novel mechanistic insights into the genetic instability initiated by complex DSBs.

Fig. 5: Model for the role of DSB complexity in determining the alternative end-joining (A-EJ) pathway choice.
figure 5

HR homologous recombination.

Full size image

Materials and methods

E. coli strains and growth conditions

The E. coli strains utilized in the present study were all derived from E. coli K-12 MG1655. The strains were cultured at 37 °C or 30 °C in Luria–Bertani (LB) medium (5 g/L yeast extract, 10 g/L tryptone, and 10 g/L NaCl) supplemented with 50 μg/mL apramycin (cat no: A600090, Sangon Biotech, Shanghai, China), 50 μg/mL kanamycin (cat no: A600286, Sangon Biotech), or 50 μg/mL ampicillin (cat no: A610028, Sangon Biotech) when necessary38.

Establishment of translocation reporter strain of E. coli

DNA primers were designed using the web-based software J5 DeviceEditor (https://j5.jbei.org/) (Supplementary Table 4). The DNA fragments were PCR-amplified using KOD-Plus-Neo polymerase (cat no: KOD-401, Takara, Osaka, Japan). The restriction endonuclease BsaI (cat no: ER0291) and T4 ligase (cat no: 15224025) were purchased from Thermo-Fisher Scientific (Waltham, MA, USA). All plasmids used in this study were constructed using the Golden Gate method (Supplementary Table 5)39.

The construction of translocation reporter system was described in our previous study40. In brief, specific fragments of lacZ and mhpR genes were utilized as N20 sequences for insertion into a single guide RNA-encoded plasmid (psgRNA), resulting in the generation of plasgRNA and pmlsgRNA plasmids. The lacI and amp genes were precisely integrated into their respective genome loci using CRISPR-Cas9 editing with these two plasmids. E. coli strain MG1655 was initially transformed with plasmids of pRedCas9 and plasgRNA, followed by induction of Cas9 expression at 30 °C with 2 g/L arabinose. After screening for knock-in strains through PCR amplification and DNA sequencing, the plasmids were eliminated through multiple-generation culture at 37 °C. Subsequently, the resulting strain was transformed with both pRedCas9 and pmlsgRNA. Following similar induction and screening procedures, a transgenic strain carrying the lacI:lacI:lacO::amp cassette, was successfully established and designated as MG1655TR. Similarly, we introduced a recB- mutation into MG1655TR by utilizing a prsgRNA with the recB gene as the N20 sequence. Additionlly, We constructed another psgRNA-derived plasmid pl4sgRNA, which consisted a 44 bp amp gene sequence and an additional 465 bp lacZ gene sequence.

Irradiation of E. coli strains

For irradiation treatments, a single colony of the specified E. coli strains was randomly selected and inoculated into 5 mL LB medium followed by incubatiion at 37 °C until reaching an approximate concentration of 108 colony forming units (CFU)/mL. Subsequently, 1 mL of the dispersed E. coli solution was transferred to a 35 mm petri dish with a depth of approximately 0.3 mm for various types of irradiation treatment. The γ-irradiation was performed using a Biobeam Cs137 irradiator (cat no: GM2000, Gamma-Service Medical, Leipzig, Germany). A 22-MeV proton beam generated at the Beijing Tandem Accelerator Nuclear Physics National Laboratory, China, was utilized for proton irradiation, with an LET of 2.61 keV/μm. Carbon irradiation was conducted at 80 MeV/nucleon with an LET of 31.46 keV/μm while argon irradiation at 310 MeV/nucleon with an LET of 11.24 keV/μm at the Heavy Ion Research Facility in Lanzhou, China.

Simulation of track structures of particle irradiation

TOPAS-nBio, an extension of TOPAS (version 3.8) was used to simulate the track structures of particle irradiation41. The process was as follows: (1) Monte Carlo geometry: To ensure the accuracy of particle trail structure simulation, the size of the beam incident surface of water phantom was set to 0.9 mm × 0.9 mm and the depth to 1.5 mm. In the simulation, the water phantom was placed in a cube with a side length of 4.0 mm, which was introduced to maintain electron equilibrium conditions. (2) Material settings: the filling material for the above geometries was water. (3) Ion source setting: the beam source used was set as a point source and placed at the isocenter of the water phantom. The number of incident particle was set to one. (4) Physical process settings: the process of track structure simulation of the single-energy ion beam adopted the g4em-livermore physical list, which is based on the common databases of EPDL97, EADL, and EEDL. Electron impact ionization was calculated theoretically according to Seltzer’s method, which combines the modified Weizsacker-Williams formula for soft collisions and the modified Möller formula for hard collisions42.

Detection of ampicillin (Amp)-resistant clones

After irradiation, E. coli were gradiently diluted, and 100 μL of the resulting solutions were respectively plated on control LB agar plates and the selection plates containing 100 μg/mL of ampicillin (Amp). Following overnight incubation at 37 °C, colonies were manually counted using a dissecting microscope (ES-18BZL, Motic, Xiamen, China) to determine the number of CFUs. The AmpR colones collected from ampicillin selection plates were further verified by individual culturing in LB liquid medium supplemented with 100 μg/mL of ampicillin. Mutation frequencies were determined by dividing the number of verified AmpR clones by those on the control LB plates. The presented data represent the average values obtained from five independent experiments.

Location and characterization of amp translocation targets

To determine whether the amp gene remained at the original loci, we designed three pairs of PCR primers for the amp gene and its flanking sequences (Supplementary Table 4). In the AmpR clones, no amplicon for any primer pairs was regarded as amp translocation. The translocation targets in the genome were located using a TAIL-PCR instrument (T100 Thermal Cycler, BIO-RAD, Hercules, USA), in which, a set of three nested primers were specially designed at the 5′ end of the amp gene. The special primers and arbitrary degenerate primers are listed in Supplementary Table 4. The third round amplicons were further sequenced by Sangon Bioteck Company (Shanghai, China), and the translocation targets were located through an alignment between the amplicon and E. coli genome using a National Center for Biotechnology Information (NCBI) alignment tool. The R package “circlize” (v0.4.15) was used to generate the Circos plots of translocation, and the sequence analysis software DNAMAN (https://www.lynnon.com/dnaman.html) was used to determine the sequence characteristics of the translocation junctions.

Detection of recombination and translocation of the lacI genes

PCR was used to determine the recombination of the lacI genes and to preliminarily identify the translocation/direct deletion of the lacI gene. The primers used are listed in Supplementary Table 4. In AmpR clones, the recombination was determined when one lacI gene and the whole spacer sequence were deleted but the genomic sequence flanking the deleted lacI gene remained intact. In the clones containing either the entire spacer or its fragment, the tail-PCR was used to further determine the lacI translocation (Supplementary Table 3).

Identification of mutations across genome of strains

The strains were subjected to whole-genome sequencing using Oxford Nanopore technology conducted by Biomarker Technologies Company (Beijing, China). The filtered reads were assembled using Canu software (1.5 version), followed by cyclizing assembly of genome using Circlator (1.5.5 version). The GenBlastA program (1.0.4 version) was employed to scan the entire genomes for identification of genetic alternations.

Statistical analysis

For the mutation rate of AmpR clones and the proportion of amp translocations, the results are presented as the mean ± standard deviation. Differences between two groups were evaluated by Student’s t-test for normally distributed data or Mann–Whitney u-test for non-normally distributed data (SPSS, 23.0 version, IBM, New York USA). The normality of data was determined using the Kolmogorov–Smirnov test. Differences in length distribution of micro-homology and proportion of mutation types were evaluated by Chi-Square test using the SPSS. Significance was considered at P < 0.05, indicated as *P < 0.05 and **P < 0.01.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.