Adaptive evolution of plasmid and chromosome contributes to the fitness of a blaNDM-bearing cointegrate plasmid in Escherichia coli

Abstract Large cointegrate plasmids recruit genetic features of their parental plasmids and serve as important vectors in the spread of antibiotic resistance. They are now frequently found in clinical settings, raising the issue of how to limit their further transmission. Here, we conducted evolutionary research of a large blaNDM-positive cointegrate within Escherichia coli C600, and discovered that adaptive evolution of chromosome and plasmid jointly improved bacterial fitness, which was manifested as enhanced survival ability for in vivo and in vitro pairwise competition, biofilm formation, and gut colonization ability. From the plasmid aspect, large-scale DNA fragment loss is observed in an evolved clone. Although the evolved plasmid imposes a negligible fitness cost on host bacteria, its conjugation frequency is greatly reduced, and the deficiency of anti-SOS gene psiB is found responsible for the impaired horizontal transferability rather than the reduced fitness cost. These findings unveil an evolutionary strategy in which the plasmid horizontal transferability and fitness cost are balanced. From the chromosome perspective, all evolved clones exhibit parallel mutations in the transcriptional regulatory stringent starvation Protein A gene sspA. Through a sspA knockout mutant, transcriptome analysis, in vitro transcriptional activity assay, RT-qPCR, motility test, and scanning electron microscopy techniques, we demonstrated that the mutation in sspA reduces its transcriptional inhibitory capacity, thereby improving bacterial fitness, biofilm formation ability, and gut colonization ability by promoting bacterial flagella synthesis. These findings expand our knowledge of how cointegrate plasmids adapt to new bacterial hosts.


Introduction
Bacterial plasmids play a significant role in facilitating horizontal gene transfer and harbor a diverse array of accessory genetic elements that are crucial to the adaptation of bacterial species [1].Cointegrate plasmids, as a representative type of plasmids, contain the genetic information of two or more original plasmids and are characterized by a high abundance of antimicrobial resistance (AMR) genes and mobile elements [2][3][4].Their formation is usually mediated by the role of insertion sequences and recombination between homologous regions in diverse versatile plasmids.Currently, numerous studies have reported the presence of cointegrate plasmids in clinical isolates [5,6] or transconjugants [7,8], thereby posing a potential threat to public health.Despite their increasing abundance, these plasmids exhibit uneven distribution across prokaryotic taxa, implying that they may confer varying levels of potential burden to different species [9].
In addition to some plasmid-chromosome pairs within bacteria that are ecologically compatible [10,11], the acquisition of an external plasmid is generally associated with a fitness cost for the host bacteria [12], which may ultimately lead to the extinction of plasmid.Despite the well-known fitness costs, plasmids are ubiquitous and seem to spread without any obstacles, which is referred to as the "plasmid paradox" [13].The widespread prevalence of plasmids in bacterial genomes poses a challenging conundrum, and solving the plasmid paradox remains an active area of research [14].Recently, multiple solutions to the plasmid paradox have been proposed, including understanding the plasmid fitness costs as well as exploring compensatory evolution that reduces the burden of plasmid carriage [15].An increasing number of studies have focused on the evolutionary dynamics and fates of plasmid-carrying bacteria to elucidate how compensation evolution mitigates the cost of plasmids.For instance, chromosomal gene mutations and specific regions deletions could facilitate the persistence of multidrug-resistant IncHI2 plasmids in Salmonella Typhimurium [16].In an Escherichia coli strain MG1655 carrying the multidrug-resistant plasmid RP4, chromosomal mutations and transcriptional modifications facilitated phenotypic changes, thereby enhancing colonization ability and plasmid transferability [17].Because these identified critical changes are strongly associated with improved bacterial fitness, the investigation of the molecular mechanisms underlying bacterial adaptive evolution is crucial in mitigating the dissemination of multidrug-resistant plasmids.
In a previous study, we identified pL53T as a cointegrate plasmid resulting from the homologous recombination of two IS26 elements on an IncX3 plasmid and a multidrug-resistant IncFII plasmid, and it carried AMR genes, including fosA3, bla NDM-5 , bla CTX-M-55 , and bla TEM-1B , which poses a significant threat to public health [18].Given the widespread prevalence of IncX3 and IncFII plasmids in E. coli [19,20], the convergence of such MDR plasmids pose a significant concern.To learn the evolution and underlying molecular mechanisms of the cointegrate plasmid, we introduced the plasmid into E. coli C600 and conducted a long-term evolution experiment under selection pressure to decipher the molecular mechanisms driving the adaptative evolution.Our findings reveal the role of psiB in facilitating horizontal transferability of the cointegrate plasmid.In addition, the sspA mutation in the chromosome enhances bacterial fitness, biofilm formation, and gut colonization ability.This adaptive evolutionary strategy broadens our understanding of plasmid adaptation to bacterial hosts in different ecological niches and enables us to identify critical targets for curtailing plasmid transmission.

Bacterial strains, plasmids, and culture conditions
The E. coli C600 strain, a prototypical K-12-derived laboratory strain, is widely recognized as an optimal material for conducting molecular microbiology and bacterial physiology research [21].The cointegrate plasmid pL53T was introduced into the C600 via electroporation to construct ancestral strain C600-pL53T.A detailed description of pL53T can be found in the results section.Unless otherwise specified, all bacteria were cultured in LB broth at 37 • C with shaking at 200 rpm by default.

Experimental evolution and plasmid stability measurement
In the initial stage of the experiment, the ancestral strain C600-pL53T was serially passaged in triplicate under antibiotic-free LB broth every day (twice a day) for 30 days (60 × log 2 1000) to achieve ∼600 generations calculated as previously described [16].The detailed process is as follows: 5 μl of bacterial cultures were inoculated into 5-ml fresh LB broth every 12 h.The bacterial cultures were stored every 5 days.To evaluate the plasmid stability, the stored cultures were streaked onto antibiotic-free LB agar plates and randomly selected 96 single colonies on each plate.PCR targeting at IncX3 replication gene and two fusion sites (Table S1) was applied to confirm the presence of the plasmid and calculate the proportion of plasmid-carrying bacteria in each generation.
The bacterial cultures preserved in this step can be used as a control group for subsequent evolutionary experiments.
Given that plasmid pL53T was not stable in C600 under antibiotic-free conditions, we therefore subjected the ancestral strain C600-pL53T to serial passages in LB broth containing meropenem (2 μg/ml) and fosfomycin (16 μg/ml) for 30 days in triplicate, among which the concentration of antibiotics was the subinhibitory concentration determined based on MIC results.The detailed procedure is the same as above.After the passage, the bacterial cultures were streaked onto the antibiotic-free LB agar plate.One single colony was randomly selected from each of the three parallel groups, and a total of three evolved clones were used for further research.The plasmid stability procedure for the evolved clones remains unchanged, excepted for the reduction in the passage of 15th day.

Competition assays
We selected the ancestral strain, three evolved strains, C600 sspA-pL53T, and C600-pL53T psiB to determine the improvement in fitness during adaptative evolution and the impact of gene deletion on plasmid fitness cost.The following methods were employed for the bacteria competition assay: cultures of the tested strains and the ancestral plasmid-free strain were diluted to a 0.5 McFarland standard and mixed at a ratio of 1:1 in 5-ml antibiotic-free LB broth.The mixtures were then incubated at 37 • C for 24 h, followed by appropriate dilution and plating onto LB agar with or without antibiotics (2 mg/l of meropenem and 16 mg/l of fosfomycin).The number of colonies of the tested strains and the ancestral plasmid-free strain was counted at both 0 and 24 h.The relative fitness (RF) of the tested strains was determined by calculating the ratio of Malthusian parameters [22], as follows: RF = ln(NRt/NRi)/ln(NSt/NSi), where NRt represents the number of resistant clones at 24 h, NRi represents the number of resistant clones at the initial time point, NSi represents the number of susceptible clones at the initial time point, and NSt represents the number of susceptible clones at 24 h.An RF value lower than 1 indicates the presence of fitness cost, or vice versa.

Measurement of biofilm formation
The biofilm formation abilities of the tested strains were evaluated according to previously described methods [2].Brief ly, overnight cultures were diluted to an approximate 0.5 MacFarland standard, and 2 μl of dilutions was pipetted into 96-well plates containing 198-μl LB broth.After incubation at 37 • C for 48 h, the cultures were discarded and the wells were washed twice with PBS to remove the planktonic bacteria.Subsequently, the biofilms were stained with a 0.1% crystal violet solution for 10 min and rinsed with PBS.After air-drying for 30 min, each well was treated with 30% formic acid to dissolve the biofilms.The biofilm yield was quantified by measuring absorbance at 590 nm.

Determination of plasmid conjugation frequency
Strain C600-pL53T, C600e-pL53T-1e, C600e-pL53T-2e, C600-pL53T-3e, and C600-pL53T psiB were utilized as donor strains for plasmid conjugation frequency measurement, with E. coli J53 (resistant to sodium azide) or J53::psiB serving as the recipient strain.The plasmid conjugation frequency was calculated by solid mating [23].After both the donor and recipient cultures reached an optical density of 0.8 at 600 nm, equal volumes (500 μl) of each culture were mixed.The mixed cultures were centrifuged at 3000g for 5 min, and the supernatants were discarded.The cell pellets were then washed with an equal volume of LB medium and subsequently resuspended in 100 μl of LB before being dropped onto LB agar plates.These plates were incubated at a temperature of 37 • C for a duration of 3 h.The number of transconjugants and recipient cells was determined by enumerating the colonies growing on plate containing sodium azide (200 mg/l) and meropenem (2 mg/l), and plate containing only sodium azide (200 mg/l), respectively.Conjugation frequencies were calculated as the ratio of transconjugants to recipient cells.

Motility test
In order to investigate whether differential expression of f lagellarelated genes contributes to changes in bacterial motility, a motility assay was performed following established protocols with biological triplicates [24].Soft MH agar plates (0.3%) were prepared and cultures adjusted to 0.6 McFarland by PBS were inoculated at the center of the semisolid medium.After incubating at 37 • C for 72 h, the diameter of bacterial growth rings was measured.

Quantitative real-time PCR
The messenger RNA expression was evaluated using RT-qPCR.Bacterial total RNA was extracted with the Bacteria RNA Extraction Kit (Vazyme Biotech Co., Ltd, China), followed by reverse transcription into cDNA by HiScriptR R III RT SuperMix for qPCR Kits (Vazyme Biotech Co., Ltd, China).The chromosomal marker 16S rRNA gene was used as the reference gene in qPCR, and the relative expression level was determined by the comparative C T ( C T ) method [25].The primers used in this experiment are listed in Table S1.

CRISPR system-based gene knockout and psiB gene complementation
A two-plasmid CRISPR/Cas9 system (Fig. S1) was utilized for psiB and sspA knockout, following the previously described protocol [26].Initially, a donor DNA with two homologous arms was integrated into the sgRNA plasmid to generate an intermediate plasmid.Subsequently, a specific spacer (N20) was inserted between the araB promoter and the gRNA scaffold in the intermediate plasmid using single PCR and single Gibson Assembly.The plasmid pHCY-25A was transformed into chemically competent C600-pL53T cells to generate the strain C600-pL53T-pHCY-25A.The resulting strain C600-pL53T-pHCY-25A was then rendered chemically competent and subjected to transformation with 5 μl of sgRNA plasmid.The transformants were recovered from an LB plate supplemented with kanamycin, chloramphenicol, and glucose after incubation at 30 • C for 12 h.The grown strain was induced with IPTG and L-arabinose before being spread onto an LB plate containing chloramphenicol, kanamycin, and L-arabinose.The primers listed in Table S1 were used to verify the knockout of targeted genes.
To complement the psiB gene, we first amplified a complete fragment of the psiB gene with its promoter using zeropsiB-F and zeropsiB-R primers (Table S1).Following purification of the product, a vector carrying this fragment was constructed using pBackZero-T Vector Cloning Kit (TaKaRa Biological Technology Co., LTD, China) and subsequently chemically transformed into E. coli J53 and verified by PCR.

Whole-genome sequencing and bioinformatic analysis
The ancestral strain and three evolved clones underwent wholegenome sequencing (WGS).Genomic DNA was extracted using the FastPure Bacteria DNA Isolation Mini Kit (Vazyme, Biotech Co., Ltd, China) and sequenced via short-read sequencing (2 × 150 bp) with the HiSeq 2500 system (Illumina).The raw sequences were assembled using SPAdes [27], and contigs < 500 bp were excluded.The cointegrate plasmid sequence obtained in previous study [18] was compared with these assembled data to illustrate the alterations of plasmid structure by CLC Genomics Workbench 12. Single-intergenic nucleotide polymorphism (SNP) analysis was performed using Snippy (4.0.2) against the genome of the ancestral strain [28].The effect of single-base mutations on protein function was predicted by Phyre2 [29], and the sequence profiles and mutations were automatically generated by Phyre Investigator tool.

In vitro transcription assays
The impact of sspA gene mutation on RNA polymerase (RNAP) activity was investigated through in vitro transcription assays using the mango method [30].The reaction system, consisting of 50-nM gadA promoter Mango DNA, 200-nM E. coli σ 70 -RNAP holoenzyme in a total of volume of 80 μl, was performed at 37 • C in a reaction buffer containing 40-mM Tris-HCL, PH 8.0, 10-mM MgCl 2 , 0.5% (vol/vol) glycerol, 100-mM KCl, 1-mM DTT, and 0.1% Tween-20.The reaction was preincubated at 37 • C for 5 min, followed by the addition of 10-μl SspA or its mutant protein (final concentration 0.312, 0.625, 2.5, and 5 μM), the NTP mix (0.1 mM; final concentration), and Tol-biotin (1 μM; final concentration) in order.The reaction system was incubated for 30 min, followed by measurement of f luorescence signals using a plate reader (SPARK, TECAN, Inc.) with excitation wavelength and emission wavelength of 510 and 535 nm, respectively.Further details about the preparation of SspA protein, σ 70 , and RNAP core enzyme can be found in the Supplementary Information.

Galleria mellonella larvae infection assay
According to the previous study, Galleria mellonella larvae infection assay was performed to estimate the virulence potential of the tested strains [32].Brief ly, larvae about 300 mg were resuscitated in a 37 • C incubator for 4 h.Overnight cultures were washed and adjusted to 10 6 CFU/ml using PBS.Ten healthy larvae were grouped and challenged with 10 μl of cultures by the Hamilton syringe of 25 μl, and the group injected with PBS was considered as the negative control.The groups were incubated in sterilized Perti dishes at 37 • C for 3 days and the survival rate of each group was recorded daily.

In vivo competition assays
According to the previous study [17], we conducted a comparative analysis of the competitiveness between the ancestral strain and three evolved clones in mouse livers and spleens.The overnight cultures of the tested strains and the plasmid-free ancestral strain were diluted and mixed at a 1:1 ratio to obtain a 250-μl mixture of 10 9 CFU/ml in PBS.Female BALB/c mice aged between 6 and 8 weeks were purchased from the Comparative Medicine Centre of Yangzhou University (Jiangsu, China), with three mice being challenged with bacteria in each group.These mixtures were intraperitoneally injected into mice.After 24 h of competition, the mice were anesthetized with isof lurane and subsequently euthanized.The spleen and liver were then aseptically extracted, weighed, and homogenized.The processed organs were further plated onto LB plates with or without the addition of antibiotic (rifampin: 200 mg/l, meropenem: 2 mg/l, and fosfomycin: 16 mg/l).The following day, the number of clones grown on each type of plate was enumerated and used to calculate an in vivo RF based on collected data.

Gut colonization assays
A total of 25 BALB/c female mice, aged 6-8 weeks, were procured from the Comparative Medicine Centre of Yangzhou University.The mice were allocated randomly into four groups (five per group) and subjected to pretreatment with streptomycin [33].Brief ly, the mice were initially fed with sterile food and drinking water supplemented with streptomycin sulfate (5 g/l) for 5 days, followed by a 5-day period of antibiotic-free sterile water consumption before bacterial colonization.At the end of this stage, fecal samples from these mice were collected and milled.The resulting material was then diluted and plated onto media containing antibiotics (rifampin: 200 mg/l, meropenem: 2 mg/l, and fosfomycin: 16 mg/l) to ensure the absence of bacteria with similar resistance profiles in the intestinal tract.After acclimatization, each mouse was orally challenged with 200 μl of bacterial suspension in PBS solution at a concentration of 10 10 CFU/ml.The control group received only PBS solution.Fresh fecal samples were collected and weighted daily starting from the following day.Following homogenization of 0.1 g feces, the resulting homogeneous liquid was inoculated onto agar plates containing antibiotics (rifampin: 200 mg/l, meropenem: 2 mg/l, and fosfomycin: 16 mg/l).A minimum detection limit of 1000-CFU/g feces was set to ensure accurate quantification of bacterial populations.

Statistical analyses
Statistical analyses were performed using Prism 8.0 software (GraphPad, San Diego, CA, USA).The data were expressed as mean ± standard deviation, and the P values between two groups were calculated by unpaired Student's test (nonparametric).The differences in G. mellonella larvae were compared using the log-rank (Mantel-Cox) test.Statistical significance was set at P < .05.

Cointegrate plasmid pL53T reduced the fitness of E. coli C600
To evaluate the fitness cost of pL53T on C600, we initially performed in vivo and in vitro competition assay between C600-pL53T and the plasmid-free ancestral strain C600.We also compared the virulence and biofilm formation ability of C600-pL53T with those of the plasmid-free ancestral strain C600.The plasmid pL53T impede C600 competitiveness both in vivo and in vitro.The 24-h in vitro RF was determined to be 0.68 ± 0.03 (Fig. 1A).Furthermore, the ratio of drug-resistant bacteria to drug-susceptible bacteria decreased following a 24-h competition in vivo mouse model (the ratio of drug-resistant bacteria to drug-susceptible bacteria in the spleen was reduced to 0.572 ± 0.01 times the initial value, whereas the corresponding value for liver was reduced to 0.668 ± 0.05) (Fig. 1B).Moreover, the plasmid did not exhibit any impact on virulence of C600 (log-rank test; P = .3532)(Fig. 1E), whereas it significantly augmented the biofilm formation ability of C600 (P < .05)(Fig. 1C).

Changes of phenotypic parameters of C600-pL53T after serial passages under antibiotics selection
The stability of plasmids facilitates their rapid dissemination and expansion into new ecological niches [34].Therefore, we performed three independent serial passages for about 600 generations in antibiotic-free LB broth at 37 • C to assess the plasmid stability of pL53T in C600.The results indicated that the plasmid cannot be maintained stably in the absence of antibiotics, as the proportion of cells carrying the plasmid decreased from 100% at initiation to 57.6% on the 600th generations (Fig. 1F).It was observed that plasmid-encoded fosA and bla NDM conferred resistance in host bacteria against fosfomycin (MIC = 128 mg/ml) and meropenem (MIC = 32 mg/ml) (Table S2).Therefore, to simulate bacterial evolution in the presence of antibiotic residues, C600-pL53T was subjected to serial passages in triplicate for 600 generations under the subinhibitory concentration of fosfomycin (16 mg/ml) and meropenem (2 mg/ml)-containing broth.
After the passage, we streaked final generation of three parallel groups onto antibiotic-free LB agar plates and randomly selected three endpoint evolved clones, namely, C600e-pL53T-1e, C600e-pL53T-2e, and C600e-pL53T-3e, and conducted a comprehensive examination of various phenotypes both pre-and postevolution.The fitness of three evolved clones was significantly enhanced as demonstrated by an in vitro competition test, with RF of 1.12 ± 0.02, 1.11 ± 0.06, and 1.08 ± 0.05 for C600e-pL53T-1e, C600e-pL53T-2e, and C600e-pL53T-3e, respectively (Fig. 1A).In addition, similar results were observed in an in vivo competition assay (Fig. 1B), suggesting that adaptive evolution may compensate for the fitness cost of plasmids during antibiotic selection.Furthermore, the stability of plasmids was obviously increased in three evolved clones, as evidenced by a constant ratio of plasmidcarrying bacteria in the population remaining at 100% after 300 generations of serial passage (Fig. 1F).
Other notable indicators underwent changes as well.In C600e-pL53T-2e, the conjugation frequency of evolved plasmid appeared to have significantly decreased (P < .05),whereas in the remaining two strains, the plasmid conjugation frequency was comparable to that of the ancestral plasmid (Fig. 1D).Furthermore, biofilmforming capacity was significantly enhanced in two of the evolved clones (C600e-pL53T-1e: P < .05;C600e-pL53T-3e: P < .01),whereas the third clone showed a trend toward improvement (C600e-pL53T-2e: P = .1589)( Fig. 1C).Additionally, all three evolved clones exhibited improved gut colonization ability compared with the ancestral strain.The cell counts of the ancestral strain rapidly decline in the gut colonization assay, and became undetectable after only 2 days in the gut colonization assay, whereas the evolved clones were able to survive for up to 6-8 days, highlighting their superior survivability (Fig. 1G).

Large fragment deletions in pL53T-2e were responsible for the decreased conjugation frequency and improved fitness
WGS was employed to gain deeper insights into the phenotypic changes of evolved clones, with a focus on identifying any gene deletions or additions in the three endpoint evolved clones.We initially mapped the sequencing reads from the evolved clones to the ancestral plasmid.Although only an SNP and a nonsynonymous SNP were identified in plasmids pL53T-1e and pL53T-2e, we observed a significant loss of ∼16 kb in pL53T-2e (Fig. 2A).The deleted region, which mainly contained psiA, psiB, parB, parM, and klcA gene (Fig. 2B), may result in a lower conjugation frequency and fitness cost for pL53T-2e.To test the hypothesis, we introduced the evolved plasmid into the ancestral E. coli strain C600 and performed a competition assay with an isogenic plasmid-free strain.The RF of ancestral C600 carrying pL53T-2e was substantially higher than that of C600-pL53T (Fig. 2C), suggesting that the evolved plasmid with a structural deletion indeed reduced its fitness cost.
We delved deeper into the issue of altered conjugation frequency.Generally, plasmid-mediated horizontal transfer can trigger an SOS response in recipient cells, and the psiB (plasmid SOS interference/inhibition) gene can suppress unwarranted SOS induction during conjugation [35].Therefore, we constructed a mutant plasmid pL53T psiB and performed a conjugation assay.The results showed that the conjugation frequency of pL53T psiB decreased to a level comparable to that of pL53T-2e (compared with pL53T: P < .05;compared with pL53T: n.s.).Complementation of the recipient strain with psiB gene restored the conjugation frequency to that of the wild-type plasmid, demonstrating that the absence of the psiB gene in pL53T-2e was a significant contributing factor to impaired plasmid transferability (Fig. 1D).Nonetheless, the deletion of psiB was found to have no discernible impact on the plasmid fitness cost (Fig. 1A and B), suggesting that psiB solely contributes to the alteration of plasmid conjugation transfer phenotype rather than affecting the plasmid fitness cost.To investigate the potential correlation between the improvement of plasmid fitness cost and plasmid copy number, we measured the plasmid copy number and MICs.It was observed that there was no significant change in either the plasmid copy number or MICs before and after evolution (Table S2), indicating that changes in plasmid fitness cost were independent of plasmid copy number.These findings revealed that although sequence deletion events can minimize the fitness cost of plasmids to host bacteria, they also weaken the horizontal transferability.The molecular mechanism underlying the reduced plasmid fitness cost, however, still needs to be elucidated.
As the structural alteration in plasmid pL53T-2e was detected in the last generation of bacterial population, the dynamics of deletion during adaptive evolution remain uncertain.To elucidate these results, we designed primers targeting the partial loss region and analyzed changes in plasmid structure among 96 monoclonal populations sampled every 100 generations from an evolved bacterial population.The frequency of the sequence in the population declined progressively after the 300th generations, with the deletion rate of 3%, 4%, 8%, and 10% for the 300th, 400th, 500th, and 600th, respectively ( Fig. 2D).

Five conserved SNPs were found on the chromosomes of three evolved clones
Chromosomal mutations, particularly in genes encoding transcriptional regulators, provide insight into the molecular mechanisms underlying bacterial fitness evolution [17].Consequently, we identified SNPs present on the chromosomes of all evolved clones.Among the three evolved clones, a total of 40 chromosomal loci with mutations were identified, including 24 nonsynonymous SNPs, six synonymous SNPs, seven intergenic SNPs, and three multiple SNPs within a gene.Furthermore, five shared non-synonymous SNPs were identified on five genes encoding PTS system trehalose-specific EIIBC component (treB), stringent starvation Protein A (sspA), asmA family protein YhjG (yhjG), serine recombinase (pinR), and glycerol-3-phosphate transporter (glpT), respectively (Fig. 3A).The gene treB is involved in trehalose transport at low osmolarity, which acts an essential role in trehalose uptake [36].The pinR gene is responsible for mediating DNA integration and DNA recombination, whereas the function of yhjG gene remains incompletely elucidated with respect to its association with a cellular component.Function prediction through Phyre2 [29] has revealed that SNPs in glpT and sspA genes may have impact on the structures and functions of their corresponding proteins (Fig. 3B).The glpT gene encodes a transport protein that mediates the absorption of sn-glycerol-3-phosphate [37].The sspA gene encodes an RNAP-associated protein that has been demonstrated to modulate bacterial virulence [38] and quorum sensing [39].Furthermore, a recent study has demonstrated that sspA gene plays a positive role in regulating biofilm formation in Pseudoalteromonas sp.R3 [40], suggesting that mutation in this gene during adaptive evolution may be responsible for the observed phenotypic changes.

Mutation of sspA played an essential role in the regulation of fitness, biofilm formation, and gut colonization of E. coli C600
To confirm the involvement of sspA gene in bacterial fitness regulation, we constructed a sspA knockout mutant C600 sspA-pL53 and conducted both in vitro and in vivo competition assays.The results demonstrated that the absence of sspA compromised the competitiveness of C600-pL53T (Fig. 1A and B).In addition, we have observed that the presence of sspA impacts the biofilm formation and gut colonization abilities of C600-pL53T strain.Both these properties were found to be reduced upon loss of sspA gene (Fig. 1C and G), indicating an indispensable role played by sspA in regulating these phenotypes.
As the SspA protein functions as a transcriptional repressor by binding to the E. coli σ 70-RNAP holoenzyme as a homodimer [30], we aimed to investigate the impact of sspA mutation on evolved clones through a f luorescence-based in vitro transcription assay.As SspA activates the gadA promoter, its inhibitory capacity is expected to be positively correlated with the transcriptional effect.The results indicate that the mutant exhibits a weakened transcriptional inhibitory ability compared with the wild type, as evidenced by its lower transcriptional activity at different concentrations (Fig. 3C and D).Consequently, the mutation of sspA diminishes its capacity for transcriptional inhibition, thereby impacting the modification of diverse phenotypes (improved fitness and reinforced biofilm formation and gut colonization ability) in the three evolved clones.

Transcriptome analysis revealed that the mutated sspA gene facilitated biofilm yield and gut colonization by enhancing flagellar biosynthesis
To get further insight into the global regulation of the sspA gene, we conducted transcriptomic analyses on C600 sspA-pL53T, C600-pL53T, and C600e-pL53T-1e.The comparison between C600 sspA-pL53T and C600-pL53T revealed the upregulation of 383 genes and downregulation of 136 genes ( Fig. 4A).These DEGs were involved in various biological processes, including bacterial metabolism, genetic information processing, environmental information processing, cellular processing, and organismal systems (Fig. 4C).In addition, C600e-pL53T-1e exhibited upregulation of 12 genes and downregulation of 17 genes in comparison to C600-pL53T (Fig. 4B).Considering the diametrically opposite biofilm formation ability, fitness, and gut colonization ability between C600 sspA-pL53T and C600e-pL53T-1e, we especially investigated the pathways that could potentially impact these bacterial properties.As expected, DEGs in both data sets were found to be implicated in bacterial motility (Fig. 4C and D).Furthermore, COG analysis indicated that more genes related to cell motility were downregulated in C600 sspA-pL53T compared with C600e-pL53T-1e (Fig. 4E and F), which may account for the underlying mechanism responsible for the improved fitness, biofilm formation ability, and gut colonization ability of the evolved clones.
To validate the hypothesis, a comparison was made between the expression levels of genes related to f lagellar assembly in C600 sspA-pL53T and C600e-pL53T-1.The expression in all selected genes associated with f lagellar assembly was upregulated in C600e-pL53T-1e, whereas a reverse result was observed in C600 sspA-pL53T (Fig. 5A), which was further confirmed by RT-PCR analysis (Fig. 5B).Consequently, the motility of C600, C600 sspA-pL53T, C600-pL53T, and C600e-pL53T-1e was compared.The results demonstrated that the motility of C600 and C600-pL53T were comparable, whereas the motility of C600e-pL53T-1e was higher than that of C600 sspA-pL53T, as indicated by their respective secretory diameters measuring 1.9 and 1.2 cm, suggesting the mutation in sspA enhanced the bacterial motility during adaptive evolution (Fig. 5C).Moreover, distinct morphological differences were observed between C600e-pL53T-1e and C600 sspA-pL53T through scanning electron microscopy.The surface of C600 sspA-pL53T appeared f lat and smooth without any discernible f lagellar structure, whereas the surface C600e-pL53T-1e was rough and uneven with a conspicuous presence of f lagellar structures (Fig. 5D).These data revealed that the mutant SspA protein facilitated biofilm formation and gut colonization ability by augmenting f lagellar biosynthesis, thereby improving bacterial fitness.

Discussion
With the advancement of sequencing technology, the complete structure and formation mechanism of cointegrate plasmids have been well characterized [5,41].Despite extensive exploration into their structural properties, less is known about the factors contributing their long-term persistence and spread within a specific bacterial host in certain ecological niches.In fact, understanding the key factors affecting plasmid dissemination is of greater significance than analyzing the structure and formation mechanisms.Our preliminary findings have confirmed that the initial fitness cost imposed by the cointegrate plasmid could be compensated for through serial passages under selective pressure [2].This phenomenon has also been observed in other types of plasmids, such as the IncHI2 plasmid in Salmonella Typhimurium [16], and IncR or IncC in E. coli [42].Herein, we introduced the cointegrate plasmid pL53T into E. coli C600, followed by long-term evolution to study the mutations or structural alterations.
Our study differs from previous ones in that we have identified compensatory mutations occurring not only in chromosomes but also in plasmids [1,[42][43][44], including deletions of the latter and mutations affecting chromosomal transcriptional regulatory genes, suggesting the diversity of evolutionary routes in diverse host-plasmid combinations.As predicted by an individualbased model from another study [45], the fate of plasmids within bacterial populations can be ultimately summarized as follows: plasmid loss, acquisition of plasmid accessory genes by bacterial chromosomes, and activation of compensatory evolution.The plasmid-borne genes confer benefits to bacteria, but they usually impose an additional burden on the host cells [43].Plasmid will take a number of countermeasures to avoid being abandoned by the host bacteria.The stability of plasmid pL4-3, for example, can be enhanced by acquiring a TA system from an endogenous bacterial plasmid [46], whereas IncHI2 plasmid pJXP9 can offset its cost by eliminating MDR regions and conjugation transfer regions [16].Our study has also identified another strategy through the detection of a large fragment deletion in an evolved clone.This fragment, which has been demonstrated to be widely conserved among conjugative plasmids [ 47,48], was located on the IncFII(pCoo) plasmid, and its absence resulted in decreased bacterial horizontal transferability and fitness costs.Further investigation revealed that the absence of psiB gene was the primary determinant of decreased conjugation frequency.The psiB gene is implicated in the inhibition of SOS response triggered by conjugation [35].Consistent with this finding, a recent study found that plasmid pED208 lacking psiB induced significantly stronger SOS response than its counterpart, which may underline the impediment to horizontal transfer of the plasmid [49].
Although the deletion of the psiB gene results in a significant reduction in conjugation frequency, the decrease in plasmid cost cannot be attributed to it.A possible explanation is that the expression of psiB is inf luenced by the functionality of promoters on a given conjugative plasmid and cellular circumstances.Evidence supports this notion, as activation of the F plasmid psiB promoter appears to occur exclusively during conjugation [50].Because psiB is independent of the plasmid fitness cost reduction, we hypothesized that the loss of genes involved in partitioning may be responsible for the phenomenon.Partitioning proteins have multifunctional roles, and apart from facilitating DNA segregation, the expression of partitioning genes also inf luences plasmid fitness cost [44].Because of the presence of par genes on both daughter plasmids of the cointegrate plasmid and its high cost, the functional redundancy may result in the loss of partitioning genes.The absence of the partitioning system may potentially impact plasmid conjugation as well.In addition to its well-accepted role in plasmid vertical transmission [51], the proteins involving in the partitioning system have been found to interact with proteins from the type IV secretion system, thereby facilitating the assembly of the conjugation machinery [52].
However, we anticipate that this effect might not be as significant as the loss of psiB because of a potential compensatory function provided by partitioning genes on another daughter plasmid.Hence, compared with the dissociated state of two daughter plasmids, the potential selection of fusion plasmids during the evolutionary process may exhibit greater diversity because of the overlap of some gene functions.Furthermore, the plasmid evolutionary pathway observed in this study may be more conductive to vertical rather than horizontal transmission, thereby rendering cross-species transmission improbable in future.This opinion is substantiated by two factors: first, we did not find any relevant reports or wild-type strains with similar fragment deletion on IncFII plasmids, indicating that such deletion event may occur in specific host bacteria or under certain conditions.Moreover, the rate of deletion did not exhibit a significant increase over time.Second, recent research has investigated two plasmids, pKSR100 and pAPR100, which circulated in the same network but exhibited distinct epidemiological characteristics.Plasmid pKSR100 displayed superior conjugation ability compared with the less prevalent plasmid pAPR100; however, bacteria carrying pKSR100 displayed a reduced SOS response in the presence of antimicrobials.Comparative genomic analysis revealed that pKSR100 contained a gene cluster consisting of five genes, including the psiB gene, which could account for the discrepancy in prevalence between the two plasmids [53].
A mutation in the sspA gene located on chromosome of all evolved clones represents an additional mechanism for enhanced fitness.The sspA gene is the first to be implicated in bacterial fitness and functions as a transcriptional repressor for regulating transcription.The SspA protein specifically binds to the E. coli σ 70 -RNAP holoenzyme and interacts with σ 70 Region 4 and zinc binding domain of β subunit, thereby inhibiting transcription initiation by suppressing promoter escape [30].SspA has been demonstrated to modulate bacterial virulence [ 38], the quorum sensing system [39], and acid tolerance [54].To comprehensively address the role of SspA in bacterial fitness regulation, we aim to discuss the potential concerns that may arise among readers.First, because the evolved clone with sspA mutation harbors additional genetic mutations, whether other gene mutations interfere with the improvement of bacterial fitness.According to preliminary findings, compensatory mutations in global regulatory proteins are a mainstream evolutionary trajectory by which bacterial hosts can counteract the cost imposed by horizontally acquired DNA [46,55,56].The outcome led us to shift our attention toward the identification of mutations in transcriptional regulators within the evolved clones.Upon comparison with the ancestral strain, only six genes were found to have missense mutations in C600e-pL53T-1e, which may account for the limited number of DEGs observed in transcriptome analysis (Table S3).Moreover, all the three parallel evolved clones showed improved fitness, and the only shared gene with missense mutations affecting transcriptional regulation was sspA gene.Therefore, it is reasonable to conclude that the sspA gene plays a dominant role in regulating bacterial fitness.Further investigation is needed to determine whether mutations in other genes affect bacterial fitness.Second, What is the underlying mechanism by which the mutated SspA affects bacterial f lagellar synthesis?There are three classes of promoters responsible for the transcription of the f lagellar operons, among which both Class I promoter and Class II promoter are transcribed by σ 70 RNAP [57,58].σ 70 RNAP is the sole target of SspA binding [30], implying sspA mutations may facilitate f lagellar assembly by inf luencing Class I and Class II promoters of the f lagellar operon.Third, as a transcriptional repressor, the reason why sspA knockout mutant exhibits decreased f lagellar synthesis and biofilm formation abilities.This may be attributed to the substation of SspA's function by other transcriptional regulators, as evidenced by significant changes in the transcription levels of certain transcriptional regulators observed in sspA knockout mutant from transcriptomics analyses.In addition, previous research has found that a significant reduction in biofilm formation in the sspA knockout mutant compared with the wildtype strain [40], which is consistent with our findings.Fourth, Whether the f lagellar biosynthesis is related to bacterial fitness.A myriad of studies has demonstrated the indispensable role of f lagella in bacterial biofilm and gut colonization ability.Flagella's primary function in the initial stage of biofilm formation is to facilitate motility, thereby promoting cell-to-surface adhesion by overcoming repulsive forces [59].For gut colonization, f lagellummediated motility assists bacteria in entering the epithelium via the intestinal mucus layer, enabling bacteria to scan epithelial cells for permissive entry sites [60].Furthermore, f lagelladeficient uropathogenic E. coli was outcompeted by wild-type strain in the urinary tract, indicating that f lagella contribute to bacterial fitness [61].Finally, whether the sspA mutation is specifically adapted to the medium conditions or inf luenced by the presence of the plasmid.We think that mutations in sspA may be adapted to both.The reasons are as follows: First, the previous study focused on the adaptation of E. coli to long-term serial passage in LB broth and observed a parallel occurrence of sspA mutation in passaged populations.Furthermore, the sspA mutant exhibited superior competitive ability compared with the wild-type parent strain, which aligns with our findings.However, the underlying mechanisms involved were not addressed in their research [62].This available evidence suggests that mutations in the sspA gene for adapting to medium conditions are common and can substantially improve bacteria fitness.Second, the function of the sspA gene primarily involves regulating the expression level of H-NS [54], a key factor inf luencing the fitness cost of IncX3 plasmid [63].In light of the widespread presence of H-NS on this plasmid, the sspA mutation may potentially adapt to the plasmid by reducing the fitness cost through modulation of H-NS expression.This could be another way of regulating plasmid in addition to the bacterial f lagellar assembly pathway, which warrants further investigation.
Unlike most previous studies that investigated evolutionary dynamics using in vitro evolutionary models [64,65], this study evaluated the competitiveness and gut colonization ability of evolved plasmid-carrying bacteria in vivo, thereby highlighting the likelihood of transmission from environmental sources to mammals.Although the overall fitness effects of plasmids measured in vitro were consistent with those measured in mouse models [66], other experimental models, such as Caenorhabditis elegans, G. mellonella, and mice or pigs, can still be utilized to objectively assess the fitness of plasmid-carrying bacteria within clinically relevant models [12].
This study has revealed that the adaptive evolution to antibiotic exposure can ameliorate the high fitness cost of bla NDMbearing cointegrate plasmid-carrying strains, and this evolution process is jointly driven by both chromosome and plasmid.As a result, the risk of persistence and further transmission of the cointegrate plasmid will inevitably be aggravated.Nevertheless, considering the diversity of cointegrate plasmids and the complexity of evolutionary dynamics under different conditions, further studies are still warranted to explore universal targets that specifically impact the spread of cointegrate plasmids.

Figure 1 .
Figure 1.Phenotypic indicators of C600-pL53T during the process of adaptive evolution.Comparison of (A) in vitro competitive ability among the ancestral strain, three evolved clones, the psiB knockout mutant, and the sspA knockout mutant.(B) In vivo competitive ability among the ancestral strain, three evolved clones, and the sspA knockout mutant.(C) Biofilm formation ability among C600, the ancestral strain, three evolved clones, and the sspA knockout mutant.The y-axis represents the OD590 nm value.(D) Conjugation frequencies of plasmids in the ancestral strain, three evolved clones, the psiB knockout mutant, and the psiB complementary strain.The psiB complementation was achieved by introducing the psiB-carrying plasmid to the recipient strain J53, and then the complementary strain was conjugated with the donor strain C600e-pL53T-2e.(E) Survival rate of Gelleria mellonella larvae infected with C600 and C600-pL53T.(F) Plasmid stability among the ancestral strain, and three evolved clones.(G) Murine-gut colonization of the ancestral strain, three evolved clones, and the sspA knockout mutant.The log-rank (mantel-cox) test was employed for statistical comparison of the survival rate in G. mellonella larvae.Student's t-tests were utilized for other data analysis.ns: nonsignificant, * : P < .05,* * : P < .01,* * * : P < .001,* * * * : P < .0001.The tested strains in the figure correspond to a unique color to facilitate comprehension.

Figure 2 .
Figure 2. A strategy for plasmid evolution.(A) Mutations and fragment deletions observed in the plasmids of three evolved clones.(B) the detailed structure of the deleted fragment was displayed, with emphasis on the major functional genes.(C) RF of C600-pL53T and C600-pL53T-2e.(D) Dynamic deletion changes of the 16-kb sequence.Statistical comparison was done by Student's t-tests.* * : P < .01.

Figure 3 .
Figure 3. Chromosomal SNPs analysis of the three evolved clones and functional verification of sspA mutations.(A) Different SNPs between the three evolved clones and the ancestral strain.(B) Prediction of the effect of single-base mutation on protein function.(C) Transcription activities of the wild-type SspA and its mutant at different concentrations.(D) Relative transcription activities of the wild-type SspA and its mutant at different concentrations.Statistical comparison was done by Student's t-tests.* : P < .05,* * : P < .01,* * * : P < .001,* * * * : P < .0001.

Figure 4 .
Figure 4. Transcriptomic analysis of the sspA knockout mutant, the ancestral strain, and one of the evolved clones C600e-pL53T-1e.(A) Volcano plot of the sspA knockout mutant compared with the ancestral strain.(B) Volcano plot of C600e-pL53T-1e compared with the ancestral strain.(C) KEGG enrichment analysis of the sspA knockout mutant compared with the ancestral strain.(D) KEGG enrichment analysis of C600e-pL53T-1e compared with the ancestral strain.(E) COG analysis of the sspA knockout mutant compared with the ancestral strain.(F) COG analysis of C600e-pL53T-1e compared with the ancestral strain.