All bacterial strains used in this study are listed in Table 1. AR phenotypes of these strains and their transconjugants against cephalosporin and ciprofloxacin (Table 1) were determined using the agar dilution method following the guidelines of the Clinical and Laboratory Standards Institute (CLSI, 2019).

Complete genome sequences were annotated using the RAST server (Overbeek et al., 2014). Tools including EasyFig (Sullivan et al., 2011), BRIG (Alikhan et al., 2011), and BWA-MEM (Li et al., 2018) were used for bioinformatic analysis. Primers for PCR are listed in Supplementary Table 1.

The transferability of MRCP and its representative progeny were investigated by filter mating. The donor and recipient strains of the first-round conjugation assay were SHfr_HI2^R strain PJM1 and E. coli J53, respectively. The J53-transconjugants obtained in the first-round conjugation were used as the donor strains of the second-round conjugation and Salmonella strains (SL1344, ATCC 14028, and M39) as recipient strains. Briefly, when grown to log phase in Luria-Bertani (LB) broth, donor and recipient cultures were mixed in a 1:1 ratio; filtered through 0.22-μm pore size membranes (HuanKai Microbial, Guangzhou, China) placed on pre-warmed LB agar plates; and incubated for 16 h at 30, 37, and 42°C, respectively. For E. coli J53, putative J53-transconjugants were selected from MacConkey agar plates supplemented with cefotaxime (4 mg/l) (Sigma-Aldrich, Guangzhou, China) plus sodium azide (100 mg/l) (Sigma-Aldrich, Guangzhou, China), and for Salmonella SL1344, ATCC 14028, and M39, the second-round transconjugants were selected on xylose lysine desoxycholate (XLD) agar (HuanKai Microbial, Guangzhou, China) plates supplemented with cefotaxime (2 μg/ml). Transconjugants were reconfirmed by XbaI-PFGE (XbaI enzyme, Takara Bio Inc., Beijing, China) and PCR screening for bla_CTX–M–55 and repHI2. Transfer frequencies were calculated as the number of transconjugants obtained per recipient. S1-PFGE (S1 enzyme, Takara Bio Inc., Beijing, China) followed by southern blot hybridization with digoxigenin-dUTP-labeled probes of bla_CTX–M–55 and repHI2 genes were conducted to obtain the localizations and/or profiles of the integrated/free plasmids (Piekarowicz et al., 1968; Yang et al., 2017; Zhang et al., 2019).

The stability of MRCP in strain PJM1 and its progeny in E. coli J53 were studied by continuous passage for 30 days. Specifically, 5 μl of bacteria culture per strain were diluted in 5 ml of antibiotic-free fresh LB broth (1:1,000) every 24 h and incubated at 30°C and 180 rpm. Three independent experiments were conducted in triplicate under each condition. The serially diluted cultures of PJM1 and its J53-transconjugants grown to 20th and 30th days were plated on XLD and MacConkey agar plates, respectively. Fifty colonies were selected, and the presence of cephalosporin resistance phenotypes and HI2 replicon was confirmed by PCR. Plasmid stability was calculated as the proportion of cells harboring both bla_CTX–M–55 and repHI2 in the 50 selected colonies. The growth kinetics of PJM1, OYZ4, E. coli J53, and the J53-transconjugants (PJ-T0, PJ-T7, PJ-T13, PJ-T16, PJ-T17, and C-OYZ4) were studied following the method reported in literature (Lv et al., 2020). The maximum growth rate (μ_max), maximum OD₆₀₀ (OD_max), the area under the growth curve (AUC), and the duration of the lag phase (lag) were calculated to evaluate the growth ability of the J53-transconjugants.

The complete sequences of the plasmids and the chromosomes have been deposited in GenBank under the accession numbers pOYZ4 (MN539018), pPJ-T0-236kb (MN539017), pPJ-T7-250kb (OL598600), pPJ-T16-250kb (OL598601), pPJ-T17-213kb (OL598602), PJ-T1-452kb (OM317741), PJ-T15-320kb (OM317740), PJ-T13-chr (CP087110), and PJM1-chr (CP045038).

A 236-kb IncHI2 plasmid, which carried chromosomally located bla_CTX–M–55 and qnrS1 in Salmonella Muenster PJM1, was detected in the corresponding J53-transconjugant C-PJM1 (renamed PJ-T0 in this study) as previously described (Zhang et al., 2019). The unique HI2 replicon located in the PJM1 chromosome was subsequently confirmed by southern hybridization with repHI2 probe (Supplementary Figure 1). Interestingly, when we repeated the conjugation experiment, the size of the plasmid transferred to E. coli J53 was variable, but remained unchanged when re-transferred to Salmonella strains (Table 1). In contrast, another circular IncHI2 plasmid in S. Chester OYZ4 (Zhang et al., 2019) remained constant when transferred to E. coli J53 and other receptor strains by conjugation. Considering the facts above, we suspected that there was a multidrug-resistant plasmid (MRCP) integrated in PJM1 chromosome, resulting in the generation of a SHfr_HI2^R strain.

To verify the characteristics of this SHfr_HI2^R strain, 95 and 90 E. coli J53-transconjugants were randomly selected via conjugation at 30 and 37°C, respectively. The cephalosporin and fluoroquinolone co-resistance phenotype, as well as bla_CTX–M–55, qnrS1, and repHI2 genes, was detected in all J53-transconjugants. XbaI-PFGE reconfirmed that the transconjugants were indeed E. coli J53. S1-PFGE profiles of the 185 J53-transconjugants showed that the MRCP in PJM1 was transferred to E. coli J53, yielding a wide range of plasmid sizes (187–459-kb) at 30 and 37°C (Supplementary Figure 2).

Three clusters including 14 subtypes were obtained by clustering analysis on the S1-PFGE profiles (Supplementary Figure 3A). Considering the size error margin of S1-PFGE prediction, plasmids showing a similar size (± 20-kb) were regarded as belonging to the same subtype. Cluster B (circular plasmid, sizes < 260-kb) had the highest yields, while cluster A (circular factors, size > 260-kb) were more frequently detected after conjugation at 30°C. Cluster C (the re-formation of plasmid-chromosomal cointegrates in E. coli J53) only seemed to occur at a very low rate at 37°C (Supplementary Figure 3A). Moreover, siiF was detected on pPJ-T1-452kb and pPJ-T15-320kb of cluster A by southern hybridization (Supplementary Figure 3B). Interestingly, the siiABCDEF gene cluster is present in Salmonella sp. but not in E. coli (Morgan et al., 2004; Barlag and Hensel, 2015). These results prompted us to trace why PJM1 has the capability to yield such a wide range of plasmid sizes and how siiF changed from being exclusive to Salmonella to being shared by E. coli and Salmonella.

To investigate the formation mechanism of the MRCP heterogeneous progeny, the accurate genome structures of parental strain PJM1 and its representative progeny (PJ-T13, plasmid integrated into E. coli J53 chromosome; pPJ-T0-230kb, pPJ-T16-250kb, pPJ-T17-213kb, pPJ-T1-452kb, and pPJ-T15-320kb, element as a 230-, 250-, 213-, 452-, and 320-kb circular molecule in J53, respectively) were obtained using deep DNA sequencing. The PJM1 chromosome was 5,000,356-bp in length and composed of 4,725 predicted coding sequences (CDSs) with 51.9% GC content. BLASTN analysis indicated that this chromosome exhibited high homology (94.4% coverage and 99.9% identity) to S. Muenster str. 0315 (NZ_CP019198) isolated from a fecal sample from a dairy farm in the United States. Comparison with the chromosome of str. 0315 showed that the PJM1 chromosome featured an extra DNA fragment, namely, MRCP (Figure 1). MRCP was 261,307-bp with 46.7% GC and 306 CDSs and displayed 99% identity with 78.5 and 88.2% coverage to the IncHI2 plasmids pDGSE139 (KM198330) and pCFSA1096 (CP033347), respectively (Supplementary Figure 4).

Interestingly, S. Chester OYZ4 obtained from poultry in Guangdong Province, China, at the same time as PJM1 in 2017 carried the circular plasmid pOYZ4, which harbored 327 predicted CDSs distributed among 257,945-bp with a 47% GC content. Importantly, although the two strains belonged to different serotypes and XbaI-PFGE profiles (Zhang et al., 2019), pOYZ4 shared a surprisingly high degree of homology (88.46% coverage; 99.71% identity) with MRCP (Supplementary Figure 4 and Figure 2). Similar to pOYZ4, MRCP also possessed an intact IncHI2-ST2 backbone (190-kb), which contained genes for replication (repHIA and repHI2), partition (parAB and parMR), conjugation (tra, trh, and htd), complete or partial toxin–antitoxin systems (hipA/hipB and higB), an SOS mutagenesis correlative operon (umuCD), and two tellurite resistance operons (terY3Y2XY1W and terZABCDEF) (Figure 3 and Supplementary Figure 4). Interestingly, the backbone was flanked by two variable regions (VRs) containing multi-resistance genes and four genes encoding site-specific recombinase (Figure 2).

The left VR (VR-L, 55-kb) clustered numerous mosaic genetic elements. The resistance gene cassettes (cmlA, arr2, and dfrA14) integrated into a class 1 integron were at the end of VR-L adjacent to the plasmid backbone, followed by mph(A), tetR, qnrS1, and a 2,971-bp transposition unit (ISEcp1-bla_CTX–M–55-orf477), which truncated a complete composite transposon Tn2. A partial Tn5393 variant was detected at the other end of VR-L, which consisted of its left terminal inverted repeats (IR_L5393, 81-bp) and carrying the genes strA/B. Importantly, the sequence adjacent to ΔTn5393 seemed to be the residual variant of Tn1721 after losing its basic transposon Tn1722 (ΔTnp1721, 1,789-bp), tetracycline resistance genes, and two 38-bp right terminal repeat variants (IR_RI and IR_RII) at both ends (Figure 2). The right VR (VR-R, 22-kb) contained four IS26 elements, a resolvase gene, and many ARGs including aph (3′)-Ia, floR, and tet(A) (Figure 2). Interestingly, a truncated Tn1721 transposase gene (ΔTnp1721, 848-bp) that seemed to be a part of ΔTn1721 in VR-L was also present in VR-R (Figure 2).

The circular plasmid pPJ-T0-230kb in J53-transconjugant PJ-T0 was 236,068-bp with 46% GC and 311 predicted CDSs (Figures 1, 3 and Supplementary Figure 4). Similarly, the pPJ-T16-250kb in PJ-T16 was 247,473-bp with 47% GC and 298 CDSs, and the pPJ-T17-213kb in PJ-T17 was 213,369-bp with 46% GC and 247 CDSs. All the plasmids showed a high homology with the MRCP (100% identity with 95.17, 97.35, and 90.83% coverage, respectively), and they all shared the same IncHI2-ST2 backbone as expected (Figures 1, 3 and Supplementary Figure 4). Interestingly, the fragment PJ-T13-chr250kb inserted into the J53 chromosomal gene bglH was the same as pPJ-T16-250kb (Figures 1, 3 and Supplementary Figure 4). That is, the SHfr_HI2^R strain could not only form free IncHI2 plasmids but also directly integrate into the chromosome of the recipient, giving place to a new MRCP in E. coli. On the other hand, pPJ-T1-452kb in PJ-T1 was 452,929-bp with 49% GC and 464 CDSs, and pPJ-T15-320kb in PJ-T15 was 331,203-bp with 48% GC and 362 CDSs. Apart from containing the 250-kb MRCP, both pPJ-T1-452kb and pPJ-T15-320kb harbored a segment of PJM1 chromosome (Figure 1), which demonstrated the capability of MRCP to capture and co-transfer adjacent chromosomal segments in PJM1. Therefore, the circular factors (pPJ-T1-452kb and pPJ-T15-320kb) in cluster B corresponded to R_HI2’ variants that consisted of MRCP and a portion of a chromosomal fragment.

To verify the dynamic evolution of MRCP during bacterial division, primers E-F and E-R from the 3’-end and 5’-end of siiE were used to amplify the probable residual regions after the excision of diverse MRCP descendants in the evolutionary PJM1 population (Figure 2A). Three products (1.5, 2.3, and > 5-kb) were obtained in PJM1. DNA sequencing showed that the 1.5-kb product, which was the same as the product in OYZ4, aligned with siiE exactly. The 2.3-kb product consisted of the truncated siiE and a IS26-v1 that interrupted siiE (Figure 2C). Despite many attempts, we still could not obtain the complete sequence of product > 5-kb by Sanger sequencing, which seemed to be a mixture of different sequences with the same adjacent sequence. This phenomenon was consistent with a diversity of J53-transconjugants obtained in the conjugation assays, indicating that the MRCP in the chromosomes of the PJM1 population were in dynamic equilibrium during bacterial division. If MRCP is deleted from the PJM1 chromosome, some sequence traces would be left behind (e.g., if MRCP is deleted in the form of pPJ-T7, it leaves a trace siiEΔ-IS26-v1-ΔsiiE in the PJM1 chromosome; if MRCP is deleted as pPJ-T7, it leaves siiEΔ-IS26-v1-strA/B-tetA/R-ΔTn1721-ISVsa3-floR-ΔsiiE; and in the case of pPJT17, siiEΔ-IS26-v1-strA/B-tetA/R-ΔTn1721-IS1R-sul3-IS26-aac(3)-IId- IS26-v1-Tn2Δ-ΔsiiE would be left, etc.). Due to the heterogeneous forms generated by MRCP, the sequence traces left changed with different deletions. Accordingly, quantitative traces (i.e., decayed MRCP) should theoretically be left in the PJM1 subpopulation. To further verify the dynamic evolution mechanisms of MRCP, the raw data of PJM1 nanopore (ONT) long-reads have also been reanalyzed, but no reads covering the decayed MRCP were found.

In the two VRs of MRCP, 10 copies of IS26 elements (IS26 or IS26 variant, IS26-v1) were identified (Supplementary Table 2), corresponding to the 10 copies in pOYZ4 (Figures 2B, 4). Almost all IS26s copies were intact, except that copy 8 in MRCP and pOYZ4 lacked the right terminal inverted repeat (IR_R), and copy 2 (present in MRCP) featured a single-base deletion (T) at position + 376, which resulted in a frameshift mutation (Figures 2B, 4). None of the IS26s were flanked by target site duplications (TSDs) that are typical for insertion sequences of this family. The 8-bp sequence 5’-CGTCGCGC-3’ (Figure 2B, black lollipop) downstream of IS26 copy 6 and the 7-bp incomplete reverse complement sequence 5’- GC^CGACG-3’ (Figure 2B, white lollipop) upstream of IS26 copy 3/4 of MRCP (“copy 3/4” represents IS26 in MRCP and corresponds with the IS26 copy 3 or IS26 copy 4 in pOYZ4) provided evidence for inverse intramolecular replicative transposition, which reversed the 40-kb segment-designated iU-A + C (Figures 2B, 4A). The IS26-v1 copy 5* (an asterisk was used to distinguish IS26-v1 from IS26) in iU-A + C interacted with the target site (GCATTGGG) in chromosomal siiE, resulting in the integration of the entire plasmid into the PJM1 chromosome to form the MRCP and the splitting of iU-A + C into iU-A and iU-C. Meanwhile, IS26-v1 copies 5*-1 and 5*-2 at each junction between MRCP and the PJM1 chromosome were generated by direct replication of IS26-v1 copy 5* (Figure 4B).

As recently published, when a central fragment is flanked by two identical IS26s, this fragment with one adjacent IS constitutes a translocatable unit (TU), which results in the excision and movement of ARGs between different replicons (from one plasmid to another or to the chromosome, and vice versa) (Harmer et al., 2020). Accordingly, the inversion of iU-A + C also led to the formation of TU-B [IS26-aph (3′)-Ia-IS26-ΔTn1721-IS4321R] in MRCP and in the same orientation of ΔTn1721 within TU-B and iU-C (Figure 4B). A single homologous recombination (HR) event between two ΔTn1721 released pPJ-T0-236kb (Figure 4C). Besides, the HR between two IS26s (copies 1 and 6) mediated the deletion of TU-B also carrying ΔTn1721. Subsequently, two parallel events may have taken place on the chromosome lacking TU-B: (i) The HR between IS26-v1 copies 5*-1 and 5*-2 released pPJ-T7-250kb, resulting in the chromosome modified by the IS26-v1 flanked by the TSD. (ii) IS26-v1 copy 5*-2 downstream of 5′-ΔsiiE interacted with the gene encoding Tn2 resolvase on the same DNA strand giving place to pPJ-T17-213kb (Figure 4C). The IS26-v1 copy 5* on pPJ-T7-250kb, on the other hand, interacted with the bgIH gene on the E. coli J53 chromosome and generated the PJ-T13 cointegrate through intermolecular replicative transposition (Figure 4D). Conversely, the HR between IS26-v1 copies 5*-1 and 5*-2 converted chromosomal fragment PJ-T13-chr250kb into the circular plasmid pPJ-T7-250kb (Figure 4D).

To determine the transferability of the SHfr_HI2^R strain PJM1, conjugation experiments were performed at 30, 37, and/or 42°C. Plasmid pOYZ4 was used as a control and E. coli J53 as recipient. Compared with the other two culture temperatures, both MRCP and pOYZ4 had the highest transfer frequency at 30°C (Figure 5A). However, the transfer frequency of MRCP (10^–4–10^–5 transconjugants per donor) was always significantly lower than that of pOYZ4 (10^–2–10^–3 transconjugants per donor) at selected temperatures (p < 0.0001) (Figure 5A). Interestingly, after 30 days of passage without antibiotics, MRCP and pOYZ4 were stably maintained in parental strain PJM1 and OYZ4, respectively (Figure 5B). Besides, there was no significant difference in the maximum growth rate (μ_max, p > 0.05) and AUC (p > 0.05) between PJM1 and OYZ4, but the duration of the lag phase and maximum OD₆₀₀ of PJM1 was significantly higher than that of OYZ4 (lag, p < 0.01 and OD_max, p < 0.001).

After 30 days of passage, pPJ-T1-452kb, pPJ-T15-320kb, pPJ-T7-250kb, pPJ-T16-250kb, pPJ-T0-236kb, and the chromosomal PJ-T13-chr250kb were stably maintained (plasmid-containing populations fraction > 90%) in their host strains (E. coli J53-transconjugants PJ-T1, pPJ-T15, PJ-T7, PJ-T16, and PJ-T0), whereas the stability of plasmid pPJ-T17-213kb was 66% maintained in its host strain PJ-T17 (Figure 5B). As for the growth ability, although there were no significant differences in the maximum growth rate (μ_max, p > 0.05) between J53-transconjugants (PJ-T0, PJ-T7, PJ-T13, PJ-T16, PJ-T17, and C-OYZ4) and E. coli J53, there were significant differences between PJ-T13 and E. coli J53 in the duration of the lag phase (lag, p < 0.05), AUC (p < 0.05), and maximum OD₆₀₀ (OD_max, p < 0.001), implying no significant differences between the other J53-transconjugants and E. coli J53 during 10-h assessment, except for a significant difference between PJ-T13 and E. coli J53 (Figure 5C). To test the transferability of different progenies of MRCP to other host strains, second-round conjugation assays were performed at 37°C with the representative J53-transconjugants as the donor strains and three Salmonella strains (SL1344, str.14028s, and M39) as the recipient strains (Table 1). Plasmids pPJ-T7-250kb, pPJ-T16-250kb, and pPJ-T0-236kb in J53-transconjugants were transferred to the aforementioned Salmonella strains; pPJ-T1-452kb, pPJ-T15-320kb, and pPJ-T17-213kb were transferred to SL1344 and str.14028s, but not M39, while chromosomal MGE PJ-T13-chr250kb could not be transferred to any of the three.