An antiplasmid system drives antibiotic resistance gene integration in carbapenemase-producing Escherichia coli lineages

Abstract

Plasmids carrying antibiotic resistance genes (ARG) are the main mechanism of resistance dissemination in Enterobacterales. However, the fitness-resistance trade-off may result in their elimination. Chromosomal integration of ARGs preserves resistance advantage while relieving the selective pressure for keeping costly plasmids. In some bacterial lineages, such as carbapenemase producing sequence type ST38 Escherichia coli, most ARGs are chromosomally integrated. Here we reproduce by experimental evolution the mobilisation of the carbapenemase bla_OXA-48 gene from the pOXA-48 plasmid into the chromosome. We demonstrate that this integration depends on a plasmid-induced fitness cost, a mobile genetic structure embedding the ARG and a novel antiplasmid system ApsAB actively involved in pOXA-48 destabilization. We show that ApsAB targets high and low-copy number plasmids. ApsAB combines a nuclease/helicase protein and a novel type of Argonaute-like protein. It belongs to a family of defense systems broadly distributed among bacteria, which might have a strong ecological impact on plasmid diffusion.

Fig. 1. pOXA-48 plasmids induce variable fitness costs and are unstable in three ST38 strains.

Three different variants of pOXA-48 were introduced from three K. pneumoniae isolates into three ST38 E. coli strains. a Alignment of the three pOXA-48 plasmids used in the study and displayed with Easyfig. IS1s are indicated in red and the bla_OXA-48 gene is in pink. The two composite transposons are indicated by double headed arrows. Tn6237 carrying bla_OXA-48 is only present in the pOXA-48_1 variant. Percentages of nucleotide identities are indicated by a gray gradient as indicated in the Figure key. b Comparison of the relative doubling time of plasmid-free strains and transconjugants calculated at exponential phase in the absence of meropenem in LB or M9 glucose minimum media. For each strain/plasmid combination, three to five transconjugants were tested. For each transconjugant, three independent experiments were performed. The relative doubling time is the ratio to the doubling time of the plasmid-free strain. c Plasmid loss during serial passages in the absence of meropenem. The ratio of plasmid-carrying colonies over the whole population at days 0, 5 and 10 was quantified by using the number of colonies growing on meropenem containing plates as a proxy for the number of plasmid-carrying colonies. The results are from three biologically independent experiments. For boxplots in (b) and (c), the median is indicated by the line, the box bounds the 25^th and 75^th quartiles and whiskers bound the minimum and maximum values excluding outliers. Outliers are values > 1.5 × interquartile range. For (b), normal distribution of data was assessed with Shapiro-Wilk normality test. Statistical analysis was performed with pairwise two sample two-sided t-test and p-values were FDR-adjusted. Upper panel: Plasmid-free strains: n = 3; ST38_1 transconjugants: for pOXA-48_1: n = 12; for pOXA-48_2 and pOXA-48_3 n = 15; ****p < 0.0001, p = 1.58e-22, 3.71e-39, 2.35e-22 (from left to right); ST38_2 transconjugants n = 15 for each plasmid, ns: no significant difference; ST38_3 transconjugants: n = 9 for each plasmid; *p < 0.05, **p < 0.01, p = 9.97e-3, 2.9e-2 (from left to right). Lower panel: Plasmid-free strains: n = 3; for each strain/plasmid combination: n = 9; *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001, p = 1.02e-5, 1.05e-11, 4.72e-7, 1.85e-12, 2.34e-4, 9.27e-6, 2.9e-2, 8.67e-3, 4.12e-6 (from left to right).Source data are provided as a Source Data File.

Fig. 2. Experimental evolution of pOXA-48_1 transconjugants selects Tn6237 transposition and pOXA-48 loss in ST38_1.

Experimental evolution was performed by 28-day passages of ST38_1/pOXA-48_1 transconjugants or of the ST38_1 parental strain in LB. For transconjugants, meropenem was added every day (MEM/day) or every three days (MEM/3 days) in cultures at subinhibitory concentration. For each condition five lineages were evolved in parallel. a Evolution of the relative doubling time of ST38_1/pOXA-48_1 transconjugant population and of the parental strain as control. b Quantification of plasmid loss by WGS of whole populations for the evolved lineages in the two experimental conditions. Plasmid loss was estimated by calculating the ratio of the number of reads on three pOXA-48 regions of same size, two outside Tn6237 and one inside (see also Supplementary Fig. 2). c Tn6237-insertion-sites identified in the evolved lineages. d Comparison of the relative doubling time of the original transconjugant and three bla_OXA-48 integrants (Tn6237 integrated in the chromosome at gadW and at two positions in the IncFII endogenous plasmid. The integrant at gadW had an additional mutation in malT frequently present in transconjugant and control evolved colonies. Chr: chromosome). In (a), (b) and (d), boxplots show median, box bounds 25^th and 75^th quartiles, whisker bounds minimum and maximum excluding outliers and outliers are values > 1.5 × interquartile range. For (a) and (b), the results are from five independent evolved lineages, for (d) they are from three biologically independent experiments. Doubling times were determined during exponential growth phase in the absence of meropenem. The relative doubling time was calculated as the ratio of the doubling time to one of the evolved plasmid-free lineages collected on the same day in (a) or to the plasmid-free strain in (d). Normal distribution of data was assessed with Shapiro-Wilk normality test. Statistical analysis was performed with pairwise two sample two-sided t-test and p-values (Source Data) were FDR-adjusted. ns: no significant difference; ****p < 0.0001. a: p = 9.28e-22 and 3.5e-20; d: p = 1.4e-08, 3.5e-06, 6.4e-07 and 2e-07 (from left to right). Source data are provided as a Source Data File.

Fig. 3. Experimental evolution of pOXA-48_transconjugants depends on plasmid structure and plasmid-induced fitness cost.

Experimental evolution was performed by 28-day passages of transconjugants from the three ST38 E. coli strains in LB or in M9 minimal medium. For evolution of the transconjugants, meropenem was added every day (MEM/day) or every three days (MEM/3 days) in cultures at subinhibitory concentration. For each condition five lineages were evolved in parallel. a Evolution of the relative doubling time of ST38_1 pOXA-48_1, pOXA-48_2 or pOXA-48_3 transconjugant populations in the presence of meropenem at subinhibitory concentration and of the parental strain. b Evolution of the relative doubling time from ST38_2 and ST38_3 pOXA-48_1 transconjugants in the presence of meropenem at subinhibitory concentration. c Evolution of the relative doubling time of ST38_2 pOXA-48_1 transconjugants in M9 medium in the presence of meropenem at subinhibitory concentration. For (a), (b) and (c), doubling times were determined during exponential growth phase in the absence of meropenem. The relative doubling time was calculated as the ratio of the doubling time to one of the evolved plasmid-free lineages collected on the same day. Boxplots show median, box bounds 25^th and 75^th quartiles, whisker bounds minimum and maximum excluding outliers and outliers are values > 1.5 × interquartile range. For a, b and c, the results are from five independent evolved lineages. Normal distribution of data was assessed with Shapiro-Wilk normality test. Statistical analysis was performed with pairwise two sample two-sided t-test and p-values (Source Data) were FDR-adjusted. ns: no significant difference; ***p < 0.001, ****p < 0.0001. a: p = 3.51e-19, p = 1.7e-04 (from left to right); c: p = 8.17e-10. Source data are provided as a Source Data File.

Fig. 4. Mutations in the F3141-F3140 operon led to pOXA-48 stabilization.

a Organization of the genomic island containing the F3141-F3140 operon. The organization of the genomic island and its insertion at tRNA-leu is identical for the three ST38 strains. Positions of the mutations of the analyzed mutants named by letters are indicated by vertical dash. In black, IS1 insertion, in green, a deletion of F3141 and in red, SNPs (strain J, L1444* and strain N, V1425W). b Doubling time comparison of ten F3141-F3140 mutant strains as indicated in (a) and of the original transconjugant. Doubling times were determined during exponential growth phase in the absence of meropenem in LB medium for ST38_1 strain and in M9 medium for ST38_2. The relative doubling time is calculated as the ratio to the doubling time of the plasmid-free strain. c, d, e and f. pOXA-48 stability in F3141-F3140 mutants after 10-day passages in LB medium in the absence of meropenem. c pOXA-48_1 stability in three ST38_1 and three ST38_2 mutants. Letters define the mutant strain as in (a) and (b). d Quantification of the carriage of the three pOXA-48 plasmids in ST38_1 and ST38_3 transconjugants deleted for the F3141-F3140 operon. and f. Quantification of pOXA-48 plasmids carriage following complementation of F3141-F3140 operon deletion. A plasmid copy of the complete F3141-F3140 operon (e) or of the operon inactivated for F3141 or F3140 (f) was expressed under the control of the p_BAD inducible promotor in ST38_1∆F3141-F3140 transconjugant. t0 refers to the value obtained following one 24-h passage in LB glucose (0.4%) and t1 to the value determined after two additional passages in LB arabinose (0.02%). pHV7-empty is used as negative control. For a, b, c and d, boxplots show median, box bounds 25^th and 75^th quartiles, whiskers bound minimum and maximum values excluding outliers and outliers are values > 1.5 × interquartile range. For a, b, c and d, the results are from three biologically independent experiments. For b, normal distribution of data was assessed with Shapiro-Wilk normality test. Statistical analysis was performed with pairwise two sample two-sided t-test and p-values (Source Data) were FDR-adjusted. ns no significant difference; *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. b: LB medium: p = 3.12–18, 2.03e-09, 1.5e-6, 3.2e-6, 2.10e-5, 3.73e-5, 1.86e-3, 3.19e-2 (from left to right); M9 medium: p = 2.95e-4, 1.43e-03, 3.24e-4, 1.19e-3,1.58e-3, 1.35e-3, 3.59e-4 (from left to right). Source data are provided as a Source Data File.

Fig. 5. F3141-F3140 (ApsAB) acts as an antiplasmid system active against low and high copy number plasmids.

a Activity of F3141-F3140 against non-IncL plasmids with different replicons and copy numbers: pACYt: p15A ori, ca. 10 copies; pBbE8c: ColE1 ori, ca. 20–30 copies; pUC19: pMB1 ori, >100 copies; pBbS8c: pSC101 ori, ca. 5 copies. CNR146C9-pKCP: incFII/IncFIB, ca. 1–2 copies b Comparison of pOXA-48 conjugation frequency between ST38_1 WT (wild type) and ST38_1∆F3141-F3140. c Comparison of transformation efficiency into ST38_1 WT and ST38_1∆F3141-F3140 by using electroporation of pACYt and pBbE8c d Quantification of the ColE1 derivative pBbE8c following induction of the F3141-F3140 operon cloned under the pBAD promoter and inserted into MG1655 E. coli chromosome (TnF3141-F3140). TnzeoR used as control, corresponds to the empty transposon and only carries the zeocin resistance gene. The plasmid copy number was determined on DNA preparations by qPCR by using the ∆Ct method with primers in pBbE8c replication origin and rpsI (chromosomal gene) as reference gene (Supplementary Data 9). For a, b, c and d, boxplots show median, box bounds 25^th and 75^th quartiles, whisker bounds minimum and maximum excluding outliers and outliers are values > 1.5 × interquartile range. For a, b, c, and d, the results are from three biologically independent experiments. For b, c and d, normal distribution of data was assessed with Shapiro-Wilk normality test. Statistical analysis was performed with Log10-transformed data by using a pairwise two-sample two-sided t-test and p-values (Source Data) were FDR-adjusted. ns no significant difference; ***p < 0.001, ****p < 0.0001. b: p = 0.0000198, 0.000648, 0.0000556 (from left to right). d: p = 4.15e- 9, 4.96e-17, 8.47e-15 (from left to right). Source data are provided as a Source Data File.

Fig. 6. bla_OXA-48 integration and pOXA-48_1 loss require the apsAB operon.

a Comparison of the relative doubling time between the original transconjugant and ST38_1∆apsB transconjugant b Quantification of plasmid carrying strains after apsAB deletion in the original transconjugant. c Evolution of the relative doubling time of ST38_1∆apsAB transconjugant whole population and of the parental strain. d Quantification of plasmid loss by whole genome sequencing of whole populations (also see Supplementary Fig. 2). Plasmid loss was estimated by calculating the ratio of the number of reads on two pOXA-48 regions of same size, one outside Tn6237 and one inside. Doubling times in (a) and (c) were determined during the exponential growth phase in the absence of meropenem. The relative doubling time is calculated as the ratio to the doubling time of one of the evolved plasmid-free lineages from the same day for (c) or to the plasmid-free strain in (a). For a, b, c and d, boxplots show median, box bounds 25^th and 75^th quartiles, whisker bounds minimum and maximum excluding outliers and outliers are values > 1.5 × interquartile range. The results are from three biologically independent experiments (a and b) and five independent evolved lineages (c and d). Normal distribution of data was assessed with Shapiro-Wilk normality test. Statistical analysis was performed with Log10-transformed data by using a pairwise two-sample two-sided t-test and p-values (Source data) were FDR-adjusted. ns no significant difference. ****p < 0.0001. a: p = 0.00000584, 0.00000584 (from left to right). Source data are provided as a Source Data File.

Fig. 7. ApsAB associates a protein with helicase and nuclease domains and a novel Argonaute-like protein.

a Predicted structure of ApsA as determined with Alphafold2. The conserved PD-(D/E)XK nuclease domain with active site residues (¹⁴²⁰D…¹⁴³³DV¹⁴³⁵K) in red and the four motifs of the helicase domain in cyan ²¹⁸GFGKS²²², blue ⁵³²DELH⁵³⁵, gray ⁸²³SAT⁸²⁵ and purple ¹¹⁴⁰QAVGRVER¹¹⁴⁷ are indicated. b Effect of mutations replacing key residues of the putative helicase (K221A, E533A) or nuclease (K1435A) domains of ApsA and of the putative 5’ end guide binding motif (K413A) of ApsB on pOXA-48_1 maintenance. Effect on pOXA-48_2 and _3 are shown in Supplementary Fig. 5. apsAB and variants were expressed under the control of the p_BAD inducible promotor in ST38_1∆F3141-F3140 transconjugants. t0 refers to the value obtained following one 24-h passage in LB glucose 0.4%, and t1 to the value determined after two additional passages in LB arabinose 0.02%. pHV7-empty (pHV7) is a negative control. Results are from three biologically independent experiments. Boxplots show median, box bounds 25^th and 75^th quartiles, whisker bounds minimum and maximum excluding outliers, and outliers are values > 1.5 × interquartile range. Source data are provided as a Source Data File. c Predicted structure of ApsB as determined with Alphafold2. The conserved motif acting as 5’-end guide binding motif in the MID domain of pAgos, ⁴⁰⁹Y,⁴¹³K,⁴³¹Q, ⁴⁸⁹K, is indicated as red balls; the RNAse H fold of the PIWI domain and the PAZ domain of pAgos are not conserved in ApsB. d Sequence logo generated by WebLogo from the alignment of ApsB homologs (shown in Supplementary Fig. 6) in the region corresponding to the MID domain, highlighting the conserved residues. A zoom on the ApsB MID domain structure at the ⁴⁰⁹Y,⁴¹³K,⁴³¹Q, ⁴⁸⁹K residues is provided in the lower panel. The C-terminal ⁷⁴³N residue possibly interacting with the putative guide nucleic acid in the binding pocket is also indicated.

Fig. 8. ApsAB belongs to a broad family of putative antiplasmid systems.

a Distance tree of ApsA related sequences retrieved from the NCBI database. Protein sequences were retrieved by using PSI-BLAST and the distance tree was obtained by the Neighbour-joining method. The distance tree separates proteins more related to ApsA (in blue) or to DdmD (in red), or intermediate (in gray). Circles are annotated as the figure key. From inside to outside: the type of associated Argonaute-like protein (ApsB or DdmE), the taxonomic origin of the sequence, and for 19 ApsB-like or DdmE-like proteins with a predicted 3-D structure in Uniprot or NCBI databases, a measure of their structural similarity with ApsB by using the root-mean-square deviation of atomic positions (RMSD)(See also Supplementary Data 8). b Alignment of different families of genomic islands containing a complete or a truncated apsAB-like operon. Other defense systems, such as the SngABC Shango system and the AbiEIAbiEII systems, encoded in some of these genomic islands are indicated. The sequence type (ST) and the strain ID are indicated, and the E. coli phylogroup is given between brackets. When it could be identified, the tRNA in which the genomic island is inserted is also indicated. apsAB variants from CNR36C9 and CNR81D10 tested in (c) have been included. c Quantification of the ColE1 derivative pBbE8c following induction of two homologs of the apsAB operon encoding proteins with 92/92% and 28/25% aa sequence identity with ApsA/B, respectively. Operon was cloned in Tn7 following PCR amplification from the ST219 CNR36C9 and ST10 CNR81D10 isolates under the p_BAD promotor and inserted in MG1655 E. coli chromosome. TnZeoR used as control, corresponds to the empty transposon and only carries the zeocin resistance gene. The plasmid copy number was determined by qPCR by using the ∆Ct method with primers in pBbE8c replication origin and rpsI (chromosomal gene) as the reference gene. The results are from three biologically independent experiments. Boxplots show median, box bounds 25^th and 75^th quartiles, whisker bounds minimum and maximum excluding outliers and outliers are values > 1.5 × interquartile range. The normal distribution of data was assessed with Shapiro-Wilk normality test. Statistical analysis was performed with Log10-transformed data by using a pairwise two-sample two-sided t-test and p-values (Source Data) were FDR-adjusted. ***p < 0.001, ****p < 0.0001.b: p = 4.69e-04, 1.32e-09, 1.77e-11, 8.71e-05 (from left to right). Source data are provided as a Source Data File.