Dissecting pOXA-48 fitness effects in clinical Enterobacterales using plasmid-wide CRISPRi screens

Abstract

Conjugative plasmids are the main vehicle for the spread of antimicrobial resistance (AMR) genes in clinical bacteria. AMR plasmids allow bacteria to survive antibiotic treatments, but they also produce physiological alterations in their hosts that commonly translate into fitness costs. Despite the key role of plasmid-associated fitness effects in AMR evolution, their origin and molecular bases remain poorly understood. In this study, we introduce plasmid-wide CRISPR interference (CRISPRi) screens as a tool to dissect plasmid-associated fitness effects. We design and perform CRISPRi screens targeting the globally distributed carbapenem resistance plasmid pOXA-48 in 13 different multidrug resistant clinical Enterobacterales. Our results reveal that pOXA-48 gene-level effects are conserved across clinical strains, and expose the key role of the carbapenemase-encoding gene, bla_OXA-48, as the main culprit for pOXA-48 fitness costs. Moreover, our results highlight the relevance of postsegregational killing systems in pOXA-48 vertical transmission, and uncover new genes implicated in pOXA-48 stability (pri, korC, DNDJGHEP_13 and 14 and H-NS). This study sheds new light on the biology and evolution of carbapenem resistant Enterobacterales and endorses CRISPRi screens as a powerful method for studying plasmid-mediated AMR.

Fig. 1. CRISPRi screens set-up and fitness effects of pOXA-48.

Depiction of the mechanism of CRISPRi screens and fitness costs associated with carrying pOXA-48 in the strains under study. A Schematic representation of CRISPRi gene-silencing and the set-up of CRISPRi screens. Values in pOXA-48 correspond to Kb. Each individual arrow represents an sgRNA targeting the region, and the orientation of the arrow indicates the orientation of dCas9 when binding to the DNA strand (RNAP stands for RNA polymerase). B Schematic representation of CRISPRi screens. Each colour corresponds to bacteria with a different pOXA-48 gene silenced (i.e., carries pFR56apm with an sgRNA programmed against said gene). Bacteria that carry a guide silencing a beneficial gene will decrease in frequency in the population during the screening, lowering the amount of the guide over time (e.g., red cells). On the other hand, bacteria carrying guides that silence costly genes will be enriched during the screening (e.g., light blue cells). C Unrooted phylogenetic tree of the Enterobacterales clinical strains used in this study. Multi Locus Sequence Types (STs) are indicated. D Distribution of pOXA-48 fitness effects (Relative Area Under the Growth Curve (see “Methods”), n = 16, error bars correspond to Standard Deviation) in the clinical strains selected for the CRISPRi screening. Asterisks indicate significant cost of pOXA-48 in each of the strains (two-sided paired t-tests against each relative plasmid-free strain after Bonferroni correction; n = 16; p < 0.05).

Fig. 2. pOXA-48-encoded genes produce similar fitness effects across clinical strains.

Representation of the fitness effects associated with silencing each gene in pOXA-48 individually in the presence and absence of antibiotic pressure, and how conserved those effects are amongst strains. A Heatmap of CRISPRi gene scores at the end of the screening (t = 72 h, ~85 generations). Fitness effects associated with silencing each individual gene / intergenic region in the absence of selection (top panel) and in the presence of ertapenem (bottom panel). The score for each gene / intergenic region corresponds to the median log₂ fold-change of the sgRNAs targeting the element in the population (see “Methods”). Blue shades indicate those genes which silencing is detrimental (i.e., genes that are beneficial in that condition), whereas yellow shades correspond to genes which silencing is beneficial during the experiment (i.e., genes that are costly in that condition). Coding regions of pOXA-48 are shown with different colours indicating common functional clusters as shown on top. Genes and surrounding intergenic regions showing a significant deviation in the gene scores in both conditions are highlighted in the top. A fully annotated version of the figure is available in SFig. 7. B Pearson correlation matrices between strains’ gene scores at the end of the screening without (left) and with ertapenem (right). Blue shades indicate positive Pearson correlation coefficients, meaning similar tendencies in the overall screening results between the strains. Circle sizes indicate correlation absolute values from 0 to 1 (blue) or −1 to 0 (red).

Fig. 3. Role of different plasmid genes in pOXA-48 biology.

Identification of pOXA-48 genes producing significant fitness effects and their effect on plasmid stability. A Volcano plots obtained from two-sided permutation test assays in absence of selection (left) and in the presence of ertapenem (right). The Y axis shows the logarithmic transformation of the adjusted P value (FDR) from the permutation test, while the x axis indicates the median gene score (log₂ fold-change values for sgRNAs targeting each gene from CRISPRi screens). Grey points indicate genes and intergenic regions which silencing was non-significant (i.e., −0.5 < gene score < 0.5; and P > 0.05). Genes and intergenic regions which silencing was detrimental in each condition are indicated in red, whereas genes which silencing was beneficial are coloured in blue. The labels of genes encoding hypothetical proteins are shown in grey. B pOXA-48 stability after 24 h of CRISPRi gene silencing of individual plasmid genes (X axis) in a subset of 5 strains (Y axis; 2 E. coli: EC22 and C288; and 3 K. pneumoniae: KPN18, KPN11 and KPN10) from our screenings (in presence of apramycin to select for pFR56apm). C pOXA-48 Copy Number measured silencing individual plasmid genes relative to control guide values for each strain tested (Log₁₀ scale). Negative values indicate PCN decrease when silencing the indicated gene. Each point shows an individual replicate (n = 6). Bars correspond to median relative PCN values for all the replicates in each of the strains tested. Bars shaded in blue correspond to K. pneumoniae strains, and in pink to E. coli strains.

Fig. 4. Role of PemK in pOXA-48 stability.

The PemI-PemK TA system promotes pOXA-48 vertical transmission. A Stability of pOXA-48 (darker lines) and pOXA-48∆pemK (lighter dashed lines) in the different clinical strains over 5 days in LB culture (with a daily 1:1000 dilution). B pOXA-48 and pOXA-48∆pemK stability after 24 h of CRISPRi gene silencing of individual plasmid genes.

Fig. 5. Role of bla_OXA-48 in pOXA-48 fitness effects.

The bla_OXA-48 gene is responsible for pOXA-48-associated fitness costs. A Two-sided Pearson correlation between the relative Area Under the growth Curve (AUC) of pOXA-48-carrying strains compared to their pOXA-48-free counterparts (shown in Fig. 1D) and the gene score (log₂ fold change) associated with silencing bla_OXA-48 in the CRISPRi screen. Lines indicate the best fit for the correlation of each species. Shadowed areas indicate the 95% CI. Each colour represents a species. B Plasmid cost associated with carrying pOXA-48 and pOXA-48∆bla_OXA-48 in different clinical strains of Enterobacterales measured as relative AUC of pOXA-48-carrying compared to pOXA-48-free clones. Colours of strain names in the x-axis are again indicative of species: in pink E. coli, in blue K. pneumoniae, and in black C. freundii. Asterisks indicate significant cost/benefit of either pOXA-48 or pOXA-48∆bla_OXA-48 in each of the strains (two-sided paired t-tests against each relative plasmid-free strain after Bonferroni correction; n = 8, except for KPN11, KPN10, KPN18, K163, EC22, C288 carrying pOXA-48, in which n = 16; p < 0.05; error bars correspond to Standard Deviation).