Mobile Integrons Encode Phage Defense Systems

ABSTRACT

Integrons are bacterial genetic elements that capture, stockpile and modulate the expression of genes encoded in integron cassettes. Mobile Integrons (MI) are borne on plasmids, acting as a vehicle for hundreds of antimicrobial resistance genes among key pathogens. These elements also carry gene cassettes of unknown function (gcus) whose role and adaptive value remains unexplored. Recent years have witnessed the discovery of a myriad defense systems against bacteriophages, highlighting that viral infection is a major selective pressure for bacteria. We hence sought to explore if gcus could encode phage defense systems. Using the INTEGRALL database, we established a collection of 129 gcus in pMBA, a vector where cassettes are established as part of a class 1 integron. PADLOC and DefenseFinder predicted four phage defense systems in this collection, comprising Lamassu, CBASS and two ABI (abortive infection) systems. We experimentally challenged all cassettes with phages and found eleven additional candidates that were not detected in silico. We have characterized in depth the 15 gcus against a panel of phages in Escherichia coli confirming their role as phage defense integron cassettes (PICs). We used recombination assays to verify that these are bona fide integron cassettes and are therefore mobile. We show that PICs confer resistance in other clinically relevant species, such as Klebsiella pneumoniae and Pseudomonas aeruginosa. Several PICs also limit prophage activation, providing protection at the population-level. Given the stockpiling capacity of integrons, we explored the additivity of phenotypes and found that integrons with two PICs confer multiphage-resistance. Additionally, when combined with antimicrobial resistance genes, integrons confer simultaneously drug and phage resistance. Crucially, we also show that the position of a pic in the array can strongly decrease its cost. Our results prove a role of integrons in phage defense, acting as highly mobile, low-cost defense islands.

INTRODUCTION

Antimicrobial resistance (AMR) is a major public health concern worldwide, and the emergence of multi-drug (MDR) resistant bacteria is making increasingly difficult to treat infections with antibiotics¹. Alternative therapeutic options are being explored as potential solutions to this problem. Phage therapy -the use of viruses that infect and kill bacteria-is an old approach (it predates the use of antibiotics) that is currently being revisited and building momentum. Several recent reports have yielded positive results for the treatment of life-threatening infections caused by MDR bacteria^2–4. However, the efficacy of phage therapy can be limited by the emergence of phage-resistant bacteria^5,6. Resistance can typically emerge through mutations in which the receptors recognized by the phage are lost, or through the acquisition of defense systems. In recent years, a plethora of new defense systems have been discovered and characterized to varying extents, adding to the classic restriction-modification enzymes or the more recently discovered CRISPR-Cas immune systems⁷. These can be classified according to the mechanism used to sense phage infection into systems that detect nucleic acids, phage proteins, the integrity of the host, and others^8,9. These defense mechanisms have evolved in response to phage predation, and they are often encoded in the vicinity of other systems, creating defense islands. In fact, it was the proximity to known defense systems that allowed the initial discovery of systems such as Lammasu, Kiwa, Thoeris, or Zorya, using a guilt-by-association rationale^9–11. Ever since, we have witnessed a sheer increase in the discovery of novel defense systems. Their abundance in genomes illustrates the ongoing evolutionary arms race between bacteria and their phages and highlights the importance of phage predation as a selective pressure for bacteria¹². Defense systems have been shown to be encoded in mobile genetic elements such as integrative conjugative elements, transposons, or prophages¹². However, they are proven to impact the fitness of the bacterial strains carrying them due to the constitutive expression of their genetic content¹³ highlighting the cost of resistance.

Integrons are genetic elements that play an important role in bacterial adaptation to changing environmental conditions^14–16. They capture and accumulate new genes embedded in integron cassettes (ICs), acting as genetic memories of adaptive functions¹⁷. Integrons typically consist of a conserved platform encoding the integrase gene and the recombination site in the integron (attI)¹⁸; and a variable region containing the cassettes. Cassettes are incorporated into the attI site through site-specific recombination reactions mediated by the integrase and are expressed from the dedicated Pc promoter encoded in the platform¹⁹. The integrase can also reorder cassettes in the array to modulate their expression, modifying their distance to the Pc and the polar effects they are subjected to^20–22. The expression of the integrase is controlled by the SOS response, so that integrons provide to their hosts “adaptation on demand”^23,24.

Mobile integrons (MI) are a subset of integrons that are typically associated with mobile genetic elements, such as plasmids and transposons, which can facilitate their horizontal transfer between bacterial cells^25–27. They are a palpable example of the evolvability these platforms confer to their hosts. They were initially discovered in the first MDR isolates of Shigella spp. during the 1950’s, carrying two resistance genes^16,28. Nowadays MIs are commonplace among key Gram-negative pathogens²⁹ carrying almost 200 resistance genes against most antibiotic families^30–32. Consequently, MIs are generally considered to be devoted almost exclusively to AMR. Nevertheless, among the cassettes found within MIs are gene cassettes of unknown function (gcu) whose importance has generally been overlooked. Yet the working model of integrons suggests that cassettes must be adaptive at the time of integration when they are expressed at the maximum strength²³. We hence reasoned that gcus could be playing an unknown adaptive role. Given the importance of phage predation in the lifestyle of bacteria and in their genome’s content, we sought to explore the possibility that gcus encode phage defense systems.

Here we have selected 129 gcus from the INTEGRALL database³³, and cloned them in pMBA, a plasmid mimicking the genetic environment of a mobile integron³². DefenseFinder³⁴ and PADLOC³⁵ predicted 4 gcus in this collection as potential defense systems. We then used in vitro screening test with phages T4 and T7 combined with structural prediction using AlphaFold and discovered 9 novel defense systems that were not predicted. All cassettes confer resistance to at least one of a panel of 8 phages against E. coli. We have confirmed that all systems are encoded in bona fide (functional) integron cassettes, which we have named Phage defense Integron Cassettes (PICs). They are hence mobile and likely to circulate among other species. Indeed, we demonstrate that they also confer resistance in K. pneumoniae and P. aeruginosa, major pathogens of the ESKAPEE group. As part of an integron, cassettes can be stockpiled in an array. Building arrays, we show that the position of a PIC in the array can strongly diminish its cost and its function. We also show that phenotypes encoded in PICs are additive, conferring multi-phage or phage/drug resistance if combined with other PICs or antimicrobial resistance cassettes (ARCs). An update in DefenseFinder allowed to discover a novel defence system in an array with pic24, confirming that multiphage resistance integrons are already present in clinical isolates. Altogether, our work proves the involvement of integrons in phage defense, acting as repositories of a broad variety of defense systems i.e.: as mobile and low-cost defense islands.

RESULTS

Integron gcus contain predicted phage defense systems

To obtain a broad set of gcus in MIs we screened the INTEGRALL database, the largest repository for mobile integron-related data (date: November 2021). We retrieved 303 cassettes annotated as gcus (< 95% nucleotide identity) with sizes ranging from 191 bp to 3 kb. We further analysed and curated this list manually using BLAST. We discarded several cassettes misannotated as gcus in which an insertion sequence had interrupted a cassette of known function. Additionally, gene-less cassettes -that failed to produce hits with any known or unknown protein-were also discarded. We therefore selected a panel of 129 elements with sizes ranging from 284 to 2.950 bp (Figure S1). All elements were synthesized in vitro and cloned in pMBA, a vector providing a biologically sound genetic environment, where cassettes are inserted in first position of a class 1 integron and are under the control of a strong Pc promoter (PcS)³². The sequences of our selected collection were submitted in May 2022 to the two major web platforms specialized in phage resistance gene prediction, DefenseFinder and PADLOC^34,35. As a result of this computational analysis, we identified four distinct gene cassettes (gcus 59, 128, 135 and N) as potential defense mechanisms, each exhibiting homology to a known system involved in phage resistance (Figure 1A).

Figure 1. Identification and Characterization of Phage Defense Systems in Integron Cassettes.

(A) Schematic workflow for the identification and experimental validation of phage defense systems in gcus. A collection of 303 gcus was extracted from the INTEGRALL database, curated to 129 gcus, and cloned into the pMBA vector, placing each gcu under the control of a strong integron promoter (PcS). Predicted phage defense systems were identified using PADLOC and DefenseFinder, (May 2022) and experimental screening was performed using a double-spot assay or growth curves to determine susceptibility or resistance to phage infection. (B) Genetic organization and predicted functions of selected gcus. Predicted function is shown for putative defense mechanisms and other candidate gcus. (C) Heatmap showing the defense profiles of identified Phage Integron Cassettes (PICs) against a panel of eight E. coli phages. The degree of protection (resistance) is indicated by the color scale, with dark green representing high protection, light green representing low protection, and white no protection. (D) Examples of growth curves used to build the heatmap in C: pic24 confers strong resistance to phage T7 at both high and low MOI while picN shows resistance to phage T4 only at low MOI. Growth curves are represented as the mean of three independent replicates. The standard error of the mean is represented as a lighter colour shade. (E) Structure of novel phage defense systems, as predicted by AlphaFold2.

Given their newly identified putative function as phage resistance elements, we propose to rename these gcus as pics, while conserving their initial name for simplicity. Pic59 (formerly Gcu59) was identified as a homolog of the AriAB system³⁶ and is suggested to confer phage resistance through abortive infection. pic128 encodes two ORFs (pic128A and pic128B) and shows similarity to a Lamassu Type 1 system. AlphaFold predicts that Pic128A shares motifs with YfjL-like or AbpA proteins, which are involved in phage defense, while Pic128B contains a putative exonuclease domain. Interestingly, the three-dimensional structure of Pic128B protein displays a remarkable similarity to DdmC, a protein from the Lamassu Type 2 plasmid and phage defense system described in Vibrio cholerae³⁷. pic135 also contains two ORFs and share structural similarity with a CBASS Type 1 system³⁸. The proteins encoded by this cassette, Pic135A and Pic135B, are predicted to play roles in nucleotidyl-transfer and membrane translocation, respectively, as suggested by protein structure predictions. Last, PicN is annotated as an AbiV family protein³⁹, implicated in phage resistance through abortive infection mechanisms.

Functional screening and domain prediction reveal additional phage defense systems

Bioinformatic approaches can fail to detect novel systems that are not similar to any of the known ones. We hence turned to in vitro experiments to detect novel resistance mechanisms within our gcu collection. We employed a double spot screening protocol in solid media and the monitoring of growth curves in liquid media, to challenge the whole collection with different concentrations of phages T4 or T7 (Figure 1A and Figure S2). We successfully detected 7 new candidates (pic24, pic142, pic233, pic217, pic113, picWGS21 and pic76) presenting substantial growth in the presence of phage (Figure S2). Using AlphaFold prediction model to find proteins with potential nuclease domains, that are often linked to phage defense, we additionally identified pic167 and its close homolog pic167.2. in silico structural predictions of proteins encoded in these PICs suggest a diversity of functional roles. Both Pic167 and Pic167.2 are likely involved in nucleotide metabolism, with indications of NUDIX hydrolase and nucleotidase activities. Pic142 is a putative restriction endonuclease, while Pic24 possesses a PIN12 domain that potentially acts as a nuclease⁴⁰. Pic233 is possibly a NADP(H) oxidoreductase; Pic113 is a putative GNAT N-acetyltransferase and Pic217 a putative integral membrane protein with a lectin C domain. AlphaFold did yield relevant homologs with a known function for PicWGS21. Last, Pic76 is a small protein of 76 amino acid that contains a LapA domain that has been related to biofilm formation⁴¹ (Figure 1B). To adhere to the custom in the field, these novel systems are also given the name of a deity of Celtiberian mythology (Figure 1B).

PICs confer protection to a panel of phages Escherichia coli

To characterize in depth the role and profile of PICs in phage resistance, we subcloned the whole library in strain E. coli IJ1862 (an F’ strain that can be infected by plasmids targeting the conjugative pilus⁴²) and conducted infection assays with a panel of phages from different families at varying multiplicities of infection (MOIs). We monitored the optical densities (OD₆₀₀) over time of bacterial cultures in presence and absence of phage MS2, F1, G4, T4, T7, P1, Φ80 and HK544. As a control, we included an isogenic strain carrying the empty vector pMBA. We used the area under the (growth) curve to determine the protective effect of the cassette, as in⁴³ (see Materials and Methods). Our data showed that all PICs conferred resistance to at least one condition (one phage at one MOI) (Figure 1C and Figure S3), confirming their role in phage defense. Resistance profiles were diverse, with some PICs showing broad-spectrum resistance, such as pic167, pic167.2, pic128 and pic142. Others, as pic24, showed a narrower spectrum with low level resistance against several phages, but very high resistance against phage T7 at both high and low MOI (Figure 1D). The resistance phenotype displayed by picN, limited to low MOI for phages MS2, T4, F1 and T7, aligns with the characteristic phenotype of an abortive infection, thereby confirming the predictions made by PADLOC and DefenseFinder. Overall, our results confirm that the gcus identified in our screenings are indeed phage resistance genes with known and characterized mechanisms.

PICs are bona fide integron cassettes

Identification of integrons and their cassettes has not been a straightforward task until the appearance of IntegronFinder⁴⁴. A large part of the content in INTEGRALL predates IntegronFinder, so their classification as integron cassettes needs to be verified. A close analysis of the genetic context, locating integrases in close vicinity and in opposite orientation, can be indirect proof that a gene is indeed a cassette. Most PICs identified here -except for pic135 and pic142- are located within class 1 or class 2 integrons, often in pathogenic strains belonging to the Enterobacteriaceae family (like pic217, pic113, and pic24) or multidrug resistant P. aeruginosa. A compelling example of the latter is the presence of picN next to cassettes bla_IMP, aacA4 and catB6, conferring multidrug resistance to carbapenems, aminoglycosides, and chloramphenicol, respectively (Accession number: AJ223604) (Figure 2A). Yet, given that finding defense systems in integrons has important implications for the mobility of these systems, we sought to verify experimentally that these are bona fide integron cassettes. Cassette recombination is a site-specific reaction that can be evidenced in a suicide conjugation assay that is typical in the field^16,45. Crucially, recombination is semiconservative, involving the bottom strand (but not the top strand) of both the attC sites in cassettes and attI sites of integrons ^46,47. This unique feature can be easily detected because it leads to clear differences in the suicide conjugation assay when the plasmid transfers the bottom or top strand of the cassette⁴⁷. With this in mind, we inserted the putative attC sites of all pics in both orientations in the suicide vector pSW23T and measured the recombination frequency of both strands for each cassette. Recombination events were verified through PCR. As expected, we observed a clear drop (between 100- and 10.000-fold lower) in the recombination frequency of the top strands of all putative attC sites, as would be expected from integron cassettes (Figure 2B). picWGS21 had extremely low recombination frequencies for the bottom strand. Yet recombination of the top strand was only observed in 1 out of 6 replicate experiments, suggesting that it is indeed an integron cassette albeit with a very poor recombination site. Indeed, the folded attC site of picWGS21 has the typical hairpin structure and extra-helical bases of these sites (Figure S4)^47–49. Hence, our data confirms that PICs are bona fide integron cassettes.

Figure 2. Genetic Context, Recombination Frequency, and Phage Protection Profiles of Identified PICs in other species.

(A) Genetic context of PICs as found in databases, highlighting their association with integron integrases (intI1 or intI2, in blue shades). PICs are found in different host species. Each PIC is annotated with its respective identifier (e.g., 59, 128A/B, 135A/B), along with known associated genes such as resistance genes (e.g., blaIMP, aacA4, catB6). (B) Recombination frequency of the attC sites of the identified PICs. The recombination frequency was measured for the bottom (green bars) and top (red bars) strands of the attC sites using a suicide conjugation assay. Bars represent the mean of at least 3 biological replicates. Error bars correspond to the standard error of the mean. (C) Heatmap showing the protection profiles of PICs against phage infection in different host species, including K. pneumoniae (KP5) and P. aeruginosa (PAO1) against a panel of cognate phages (F13, Vs1, Px4, Px5).

Defense cassettes are protective in other species

Given the mobility of integrons and cassettes among important pathogens, we sought to determine if PICs are active in different hosts. To this end, we evaluated the protective effects of PICs in K. pneumoniae and P. aeruginosa, two species of the ESKAPEE group of highly resistant and lethal pathogens where integrons are prevalent. We introduced all PICs into K. pneumoniae KP5, a MDR clinical isolate. We sequenced the genome of this strain and found three plasmids, one of which contains a class 1 integron (the full resistome can be found in Figure S5). The PHASTEST software did not detect prophages in the genome of KP5. We subjected all strains to infection with the Drexlerviridae phage F13 co-isolated with the strain. All PICs, except picN, demonstrated resistance to F13 at low MOI. Most of them also conferred resistance at high MOI (except pic76, pic113, pic217, and picWGS21) (Figure 2C). To introduce PICs in P. aeruginosa PAO1 strain, we changed the p15A origin of replication of pMBA for BBR1, and introduced a tetracycline resistance marker. After several assays, we could only transform picN successfully, while other PICs produced no colonies at all or abnormal (sick) colony morphologies (Figure 2D). This led us to hypothesize that some pics might be toxic or extremely costly in this background. To avoid this, we changed the strong promoter guiding the expression of cassettes in pMBA (PcS) for a weaker version (PcW). This way we could introduce pic24, pic113, pic142 and picWGS21 in PAO1 and were able to measure their activities. We confronted these strains to a collection of 36 phages infecting PAO1 using a double spot screen. We observed growth for picN, pic142 and picWGS21 in presence of phages Vs1, Px4 and Px5. We evaluated the level of resistance by monitoring growth curves for each PIC against the three phages, and confirmed their resistance phenotype (Figure 2C). pic142 - that had a clear phenotype against several phages both in E. coli and K. pneumoniae- confirmed its broad spectrum and host range, conferring resistance to Px5 at both MOI and to Px4 at low MOI; contrarily, picN and picWGS21, that had only showed low resistance against a handful of phages in E. coli, conferred very high resistance against phage Px4 and Vs1 in P. aeruginosa. Altogether, our data proves that PICs can confer resistance to different phages and in different host species. Although some PICs seem to have a defined narrow host range, others, like pic142 were able to confer resistance in the three species against most phages tested, underlying the broad-spectrum activity of this defense mechanism.

PICs protect against prophage activation

An important aspect of the interplay between phages, their hosts, mobile genetic elements and defense systems is the case of prophage activation. Hindering activation of prophages can both protect at the cell and at the population level. Here we asked if a plasmid carrying a mobile integron with a PIC can limit the production of phage particles from a chromosomally encoded prophage once it switches to lytic cycle. To test this, we introduced all PICs in lysogens of E. coli 594 with either HK544 or Φ80 prophages in their genomes. We induced prophage activation with mitomycin and quantitated the phage titer after 6 hours. pic128, pic135 and pic142 showed mild to strong defense against the activation of HK544, with 20 to 1.000 fold decreases in plaque forming units. Pic59 completely abolished the production of HK544 virions, showing >10⁷ fold protection (Figure 3A). Resistance against Φ80 activation was generally low, with only picN, pic167, and pic233 producing statistically significant resistance levels. Interestingly, there is not a clear correlation between defense against the free form of the phage (Figure 1C) and the activation of the prophage. Certain systems, like pic135 and pic142 confer resistance against both forms of a phage (HK544); while others, like pic128, pic167.2 and pic217 only conferred resistance against one. These results show that mobile integrons containing PICs can interfere with prophage activation, highlighting the potential protective effect at the community level, and the complex interplay between mobile genetic elements.

Figure 3. Prophage Induction, Structural Alignment, and Impact of pic128 on Plasmid Stability.

(A) Prophage induction assays inducing with mitomycin (0,5 µg/ml). The graph shows the concentration of phage (measured as plaque-forming units per millilitre (pfu/ml)) when activating prophages in the presence of pMBA containing PICs over the empty pMBA. ND: not detected (Limits of detection plotted). One sample t and Wilcoxon test; P values are shown when possible; ** < 0.01; *** < 0.001; ****<0.0001. (B) Structural alignment using FoldSeek of Pic128A and Pic128B proteins with known phage and plasmid defense proteins DdmA and DdmC, respectively. Superposition of protein structures is shown, with alignment scores indicating the degree of structural similarity. (C) Plasmid stability assays in E. coli. The percentage of fluorescent cells, representing the stability of pMBA, is shown for strains containing the pBAD plasmid with either the ddmABC or pic128AB systems. While ddmABC significantly reduces plasmid stability, pic128 does not show a significant effect, indicating no anti-plasmid activity. Bars represent the mean of at least 3 biological replicates. Error bars correspond to the standard error of the mean. Statistical significance was assessed through unpaired t-tests and P values are shown.

pic128 does not have anti-plasmid activity

DdmABC is a phage and plasmid defense system. It has been shown to confer phage resistance in E. coli and to cure plasmids in Vibrio cholerae, particularly those with a p15A replication origin³⁷. When modelling protein structures of PICs, we found strong similarities between Pic128B and DdmC (Figure 3B), proteins that belong to (different) Lamassu defense systems. Interested by the interactions between mobile elements, we set out to examine whether the expression of pic128AB could exert a similar influence on plasmid stability. Because pMBA has a p15A replication origin, we decided to investigate the antiplasmid activity of pic128AB from a different vector. We hence cloned pic128AB and ddmABC systems into a pBAD plasmid that possess a pBR322 origin of replication and co-transformed them with an empty pMBA. We induced both systems with arabinose and monitored the loss of pMBA as the presence of non-fluorescent colonies for 3 days (ca. 20 generations). pMBA is stably maintained in the absence of antibiotics, with 99% of the cells in the population being fluorescent after 3 days. As expected, DdmABC strongly decreased the proportion of fluorescent colonies to 19% by day 3 (Figure 3C), while pMBA was stably maintained at >97% in the presence Pic128AB. This suggests that, despite their structural similarities, Pic128AB does not have the anti-plasmid activity detected for DdmABC. Still, both systems could have different plasmid specificities, so we extended our plasmid stability studies to KP5, carrying the three distinct natural plasmids conferring resistance to tetracycline, ertapenem and cefotaxime, respectively. After introducing pMBA-pic128 into this isolate we assessed plasmid stability monitoring the loss of three resistance phenotypes conferred by these replicons over a five-day period (Figure S5). We could not observe a significant plasmid loss in any case, suggesting that pic128 is not an anti-plasmid system.

PICs entail different fitness effects that vary across species

A major question in the field of phage defense is how these systems affect the fitness of the host. This has important implications in the accumulation of such systems in genomes and in their spread between species. Integrons are known to be low-cost platforms that can modulate the expression of the genes in their cargo by changing their position in the array^21,50,51. Yet the cost of each cassette can be different, which is of importance in their dispersal through HGT. To determine the cost of PICs, we performed competition assays in both E. coli and K. pneumoniae (Figure 4). In these experiments two strains that are isogenic -except for the PIC and a fluorescent marker-are mixed and grown together. Using flow cytometry, the ratio of the mixture is measured before and after growth, and changes in this ratio serve to quantitate the cost of cassettes (see Materials and methods). Our results showed a broad distribution of fitness effects among PICs, ranging from large costs (ca. 40%) to virtually costless in both species. In E. coli, pic217, pic128 and picN seemed to have a mildly positive impact in fitness (up to 3%) suggesting that acquisition of certain systems can be costless (Figure 4A). It is of note, that here we have used a strong version of the Pc promoter, which likely maximizes the cost of cassettes. Weaker promoters are also found in MIs, so it is possible that the cost of cassettes is lower in other integron platforms. Interestingly, fitness effects were not conserved between genetic backgrounds. For instance, picWGS21 entailed a small cost (ca. 7%) in E. coli but was very costly (ca. 30%) in K. pneumoniae (Figure 4B). Only a few cassettes, like picN showed very similar cost in both species. The lack of cost of PICs in certain genetic backgrounds, and the differences in fitness effects across backgrounds can be of importance in the accumulation of cassettes and their preferential distribution of PICs among species.

Figure 4. Fitness Effects of PICs.

Fitness effects of PICs in (A) E. coli and (B) K. pneumoniae. Red circles represent fitness values below 1 (cost), while green circles indicate fitness values above 1 (gain). Fitness effects vary in sign and magnitude and are not conserved among genetic backgrounds. (C) Modulation of fitness cost and function by integron position. The graph shows the fitness effect of pic24 in first position (competing pMBA vs pMBA-pic24) and third position (pArrayØ vs. pArray-pic24). The fitness cost imposed by the PIC is reduced in third position behind cassettes with strong polar effects. The cost is related to the function of the PIC since pArray-pic24 does not confer resistance to T7. (D) Polar effects of PICs. All cassettes except pic167 exert polar effects on downstream cassettes, limiting their expression. All assays have been performed at least three times. Error bars represent the standard error of the mean.

In the case of costly defense systems, a key advantage that integrons can provide to the host, compared to encoding them elsewhere, is that they can modulate their expression. Cassette expression dependes on its proximity to the Pc and on the polar effects that cassettes upstream can exert. Because the expression of a gene and its fitness cost generally correlates, we hypothesize that integrons can modulate the fitness of PICs in the array by modifying its position in the array. To test this, we compared the cost entailed by pic24 when located in first position (see above) with the cost it entails when encoded in third position (as it is found in the databases (Figure 2A)). For this, we constructed an array (pArrayØ) containing two antibiotic resistance cassettes, with strong polar effects (aacA54 and aacA8)⁵¹, and introduced pic24 downstream, as a third cassette (pArray-pic24). Competition experiments show that while pic24 conferred a fitness cost of 11% when located in first position, it had no significant cost when in third position (Figure 4C). This was indeed due to the strong repression of its expression, since phage infection experiments showed pArray-pic24 did not protect against T7 infection. Hence, costly PICs can be carried by mobile integrons at no cost to be later reshuffled into first position, providing phage resistance (adaptation-) on demand²⁴ We also show that PICs exert polar effects on downstream cassettes (Figure 4D), participating in the modulation of function and cost that we have previously demonstrated for antimicrobial resistance genes⁵¹.

Mobile Integrons can accumulate resistance to phages and antibiotics

The fact that integrons carry defense systems is important in the ecology and biology of these systems. Integron cassettes can be stockpiled, and their phenotypes can be additive, as we have observed for multi-drug resistance MIs. It is hence possible that integrons can combine PICs to provide multi-phage resistance; or PICs and antimicrobial resistance cassettes (ARCs) providing phage- and AMR. This is important for phage therapy, since integrons would likely play again -as it occurred with antibiotics-a key role in the evolution of resistance. Hence, we sought to test if phage defense phenotypes can be additive and combined with antimicrobial resistance genes. To do so, we first constructed pArray-1 an integron combining pic24 (T7^R) and bla_OXA-10, a class D β-lactamase gene that confers resistance to carbenicillin. The strain carrying pArray-1 showed high levels of resistance against T7 and carbenicillin both separately and simultaneously (Figure 5A). We then constructed pArray-2 containing pic24 and pic167.2 (Figure 5A) that confer high and intermediate resistance to phage T7 and P1, respectively. Strains carrying pArray-1 showed high resistance against both T7 and P1, highlighting the additivity of phage resistance phenotypes. Our results show that, as predicted by the integron model, integrons can combine phage resistance with AMR. Because MIs are enriched in AMR genes, we can find several natural examples of the former among our PICs, with picN, pic113, pic167, pic217 and picWGS21 located in integrons containing resistance genes against beta-lactams, aminoglycosides, chloramphenicol and trimethoprim (Figure 2A). Because gcus are less abundant, we did not find co-occurrence of PICs within an array. Nevertheless, having described a variety of two-gene PICs, the genetic environment of pic24 caught our attention and was re-evaluated. This PIC is in a class 2 integron, preceded by two cassettes containing two ORFs each. A new search with an updated version of DefenseFinder (June 2024) found homologs of sensor protein ThsB from the Thoeris system in both cassettes (gcu23 and gcu24). It also found a type II restriction modification system downstream the last cassette of the integron (Figure 5C). Both cassettes share 50% bp identity, and the ORFs share 17 and 37% identity in amino acid sequence respectively. Nevertheless, the structure of the proteins in both systems is remarkably conserved. Despite the homology of Pic22A and Pic23A with ThsB from Thoeris, Pic22B and Pic23B are shorter than ThsA (approximately 300 vs. 500 aminoacids of ThsA), show a very low protein identity (ca. 13%) and a very different predicted fold, suggesting that these genes are not homologs. The structure of these proteins is instead similar to that of the Deva system (picWGS21) (Figure 1E).

Figure 5. Additivity of Phage and Antibiotic Resistance Phenotypes in Multi-Cassette Arrays.

Growth curves (OD₆₀₀) of E. coli strains containing (A) pMBA, pic24 and bla_OXA-10 independently, and combined in an array in the absence of phage or antibiotic, and challenged with phage T7, carbenicillin, and a combination of both. The combined array confers resistance to phage and antibiotics simultaneously, and (B) pMBA, pic24 and pic167.2 independently, and combined in an array; either with no phage, or challenged with phage P1, phage T7, and a combination of both. (C) Genetic background of pic24. pic22 and pic23 contain homologs to TshB from Thoeris that confer resistance against phages from E. coli and K. pneumoniae (Kpn). pic23 and pic24 can confer multiphage resistance when in the same array. Growth curves are represented as the mean of three independent replicates. The standard error of the mean is represented as a lighter colour shade.

Both gcu22 and gcu23 were in our initial collection, but were not selected in our screenings. In the case of gcu22 we had not observed resistance in any case, while gcu23 was discarded because it yielded inconsistent results. We investigated this in more depth and sequenced a variety of colonies from our gcu23 stock. This revealed that the frequent interruption of the cassette by the insertion of IS1, which explained the inconsistent results in our screenings. We interpreted this as a sign of high fitness cost, so to test both cassettes we cloned them in a modified pMBA under the control of a weak Pc promoter (PcW) (30-fold less active than the PcS⁵²) and tested them against the panel of phages in E. coli and K. pneumoniae. gcu22 conferred low levels of resistance exclusively against phage F13 in K. pneumoniae, while gcu23 had a broader defense profile, including very high resistance levels against G4, MS2, F1 and T4) (Figure 5D). We hence redefined both cassettes as PICs (pic22 and pic23) and, given the apparent hybrid genetic structure mentioned before, we consider this a novel system that we have called Tragantía (a deity with the upper body of a woman and the lower body of a snake). Finally, we sought to prove experimentally that this natural arrangement of cassettes acts as a defense island. Given that only the phenotypes of pic23 and pic24 are distinguishable, we built pArray-3 with pic23 and pic24 under the control of a PcW and challenged it against T4 and T7. This array conferred multi-phage resistance (Figure 5E), confirming that mobile integrons naturally act as mobile defense islands and that they are already circulating among clinical isolates⁵³.

Defense systems in PICs are found outside integrons

Integron cassettes are exchanged between integrons cohabiting the same cell. Mobile integrons can therefore scan sedentary chromosomal integrons (SCIs) and bring to clinical settings adaptive functions evolved elsewhere in the biosphere⁵⁴. This suggests that PICs are likely found in SCIs. In an article in this issue Darracq et al. provide evidence that the Superintegron in V. cholerae contains indeed a variety of PICs, confirming their role as biobanks of phage defense systems⁵⁵. All the PICs in that work contain systems that are different from the ones described here. We hence sought to investigate the potential origin and distribution of our defense systems. We have searched for homologs in NCBI and examined the genetic context of hits. Certain cassettes show a broad distribution among integrons (Figure S6). Such is the case of picN, found among MIs and SCIs within the family Pseudomonadaceae (Figure 6A), which is consistent with it conferring resistance in P. aeruginosa (Figure 2C) and exemplifies again how integrons connect clinical and environmental genomes. The defense system Cosus was initially identified in pic24 as part of a class 2 integron. Homologs of Cosus are found as PICs in integron arrays in Marinobacter salexigens (Figure 6B). They are also found in the genomes of Vibrio species, outside their SCIs but with a conserved attC site suggesting that this could be a bona fide cassette integrated into an attG site in the chromosome ⁵⁶. Interestingly, homologs of PicN and Pic24 were found isolated in the genomes of Kangiella japonica and Escherichia marmotae without recognisable attC sites despite the sequence conservation along the coding region (Figure 6C and Figure S7). Additionally, a Tragantía system showing 95% aminoacid identity with Pic22A and B, is found in the chromosome of Marinobacter sp. CuT6 (Figure 6D). This isolate has a complete integron and a cassette array lacking an integrase (CALIN), but Tragantía is not in these regions. It is found within a small defense island followed by homologs of Thoeris and Gao⁵⁷. Yet compared to Pic22, this Tragantía system has a pseudo attC. While canonical attC sites have two or three extrahelical bases on one side of the stem -that act as landmarks for the integrase- and a conserved GTT/AAC crossover point in the R box, this site has extrahelical bases at both sides of the stem, and a mismatch in the conserved crossover point, which likely makes it non-recombinogenic (Figure S8). While this does not seem to be a bona fide integron cassette (and it is not recognised by IntegronFinder), it is strikingly close (only a few mutations away) from a canonical one. Altogether, this highlights the high plasticity in the genetic context of the systems encoded in PICs and suggest a recent recruitment of these systems into integrons.

Figure 6. Genetic context of PIC homologs.

Analysis of the genetic environments of AbiV (picN) (A) and Cosus (pic24) (B) showing their distribution in mobile and sedentary chromosomal integrons of different species. Integron integrases and arrays are marked where recognised. Some homologs to PICs are also found in outside integrons, and, despite a good sequence conservation in the coding region, they lack attC sites (C). A Tragantia system, homolog of pic22, is found within a small defense island in the chromosome of Marinibacter spp. and contains a pseudo attC site. This isolate contains an SCI and a CALIN.

DISCUSSION

Phage therapy is an old approach with renewed interest in the light of the AMR crisis^58,59. It allows for personalized treatments against multidrug resistant bacteria, with encouraging outcomes. Phage resistance is a rapidly evolving field of research. The constant discovery of new resistance genes, often encoded in defense islands and prophages, reveals the intense arms race between bacteria and their phages, and between mobile genetic elements. We lack knowledge on how defense systems are recruited into these islands and on how frequently they can move through HGT¹³. In contrast, mobile integrons are highly adaptable genetic elements that have significantly contributed to the proliferation of multidrug resistance bringing to our hospitals a plethora of resistance cassettes from the genomes of environmental bacteria. Being shed to the environment at extremely high quantities (10²³ per day)⁶⁰, mobile integrons have established a strong communication route between the genomes of pathogenic and environmental bacteria⁶¹. Hence, the rampant movement of mobile integrons among clinically relevant bacteria ensures a high speed of dissemination of novel adaptive functions throughout our hospitals. It is therefore critical for the future of phage therapy to understand the potential role of integrons in phage defense. In this work we confirm that mobile integrons already contain such systems.

We have successfully identified 15 PICs using bioinformatic tools and a phenotypic screening. As is the case of most defense systems, PICs displayed different specificities conferring narrow to broad immunity in our assays. Nevertheless, because integrons can build arrays of cassettes, they can combine phenotypes to confer multi-phage or phage-drug resistance, the same way they currently confer multidrug resistance. In fact, picN is already found in an array with at least three ARCs in a P. aeruginosa clinical isolate (Figure 2A). This, and other examples of integrons containing PICs at different positions within the array, together with our results showing that they can be carried at no cost, suggest that mobile integrons can become low-cost, highly-mobile defense islands. An extreme case of this is shown by Darracq et al. in a back-to-back paper in this issue, where they describe the presence of PICs in the Superintegron of Vibrio cholerae. Indeed, in a recent work we have shown that the Superintegron is carried at no measurable cost ⁶². Additionally, their findings proves that chromosomal integrons are virtually infinite repositories of PICs from where they can be recollected by mobile integrons. Both Darracq et al. and us show that pics are functional in distant genetic backgrounds. In our work, we show that this is the case in 3 of the 5 Gram negative species and strains of the ESKAPEE group of highly resistant and dangerous pathogens: the kind of bacteria aimed by phage therapy assays. This strongly suggests that the spread of PICs among them will be quick if selective pressure with phages becomes commonplace. This has implications in what are today considered as exploitable trade-offs in phage therapy. For instance, it is known that many K. pneumoniae phages bind the bacterial capsule and that resistant clones can easily arise through capsule loss, albeit at the cost of becoming non-virulent⁶³. Also, in some cases becoming resistant to phage infection through mutations comes at the cost of losing antibiotic resistance^64,65. The acquisition of plasmids containing integrons with PICs can abrogate the exploitability of such trade-offs, rendering virulence and antibiotic resistance perfectly compatible with phage resistance.

The role of mobile integrons in phage resistance exemplifies the interplay between mobile genetic elements. Indeed, many of the PICs described here confer resistance against phages that target the conjugative pilus of E. coli IJ1862. Hence, PICs can be beneficial for conjugative plasmids that protect indirectly protecting their host. This way Mobile Integrons would alleviate the trade-off between acquiring adaptive in plasmids and becoming susceptible to a large variety of phages. Another aspect of this complex interplay is the potential link of pic128 with plasmid stability. Contrarily to the ddmABC system, pic128 does not interact with p15A plasmid or with any natural plasmids present in KP5. This result highlights that not all Lamassu systems have the same activity toward plasmids. The lack of anti-plasmid activity in Pic128AB might have a biological meaning. Given that plasmid acquisition is an important source of antibiotic resistance, it is possible that the lack of anti-plasmid activity of pic128 has been selected for in an environment with high antibiotic pressure. Additionally, the fact that certain pics could not be introduced in P. aeruginosa (even under a weak Pc promoter) suggests that PICs can act as barriers to HGT.

Our work is not exempt of limitations. First, our screening -while successful in detecting nine new genes-was based in the use of only T4 and T7. Given that some pics exhibit high specificity against phages, it is likely that the number of pics present in our collection of gcus is underestimated. This could also apply to prophage-activation protection and plasmid stability assays, if additional prophages and plasmids were tested. Additionally, algorithms to detect defense systems are being constantly improved, so that detection of PICs in gcus will likely be more prolific at the time this work is published.

This study shapes our perception of integrons. The presence of phage defense systems and antibiotic resistance genes, proposes that integrons are protection elements against a large breadth of environmental insults. This can guide the search for novel functions in the virtually infinite reservoir of cassettes in nature. Our data also suggest that defense systems have been recruited recently in integrons, providing for the first time, a multitude of examples of closely related homologs of cassettes that are not part of integrons. We have also found an example of a Tragantía system with a pseudo attC sites that is only a few mutations away of becoming a fully recombinogenic attC. This way, the study of PICs can help understand how genes become cassettes, a long-standing question in the field. Our results have also important implications in the field of phage resistance, where mobile integrons provide extreme mobility at a very low and adjustable fitness cost. Finding defense systems in Mobile integrons highlights the risk that phage resistance genes could follow the same dissemination route as ARG in the past century. The field of phage therapy must be vigilant about the adaptive power of integrons: selective pressure will most likely reveal once again a major role of integrons in the spread of resistance mechanisms. If the importance of integrons in the evolution of phage resistance is neglected, it might come a day when phage hunting against clinical isolates becomes extremely difficult.

MATERIAL AND METHODS

Strains and Phage collection

A list of strains can be found in Table S1. E. coli DH5a and P. aeruginosa PAO1 are laboratory strains. K. pneumoniae KP5 is a clinical isolated at Hospital Fundación Jiménez Diaz, Madrid. E. coli 594⁶⁶, was kindly donated by José Penadés together with phages HK544 and Φ80. Strains were grown at 37 °C in lysogeny broth (LB) or LB agar (1.5 %) (BD, France). Zeocin was added at 100 µg/ml to maintain pMBA and pMBA-derived plasmids in E. coli and K. pneumoniae. Tetracycline at 100 µg/ml was used to maintain pMBA plasmids in P. aeruginosa. Liquid cultures were incubated in an Infors Multitron shaker at 200 rpm (Informs HT, Swiss). Antibiotics were purchased from Sigma Aldrich (Merck, USA) except for zeocin (InvivoGen, USA).

The coliphages included T4, T7, P1, MS2, G4, F1 which belong to the Myoviridae, Autographiviridae, Myoviridae, Fiersviridae, Microviridae and Inoviridae, respectively. E. coli phage MS2, F1, G4 and their host E. coli IJ1862⁴² were kindly provided by from Prof. James J. Bull (University of Texas). Phages T4, T7, and P1 were purchased from the DMSZ collection (Leibniz, Germany). The K. pneumoniae phage, named F13, (found in effluents from a hospital) was selected for its ability to lyse a clinical isolate called KP5, which produces the class D carbapenemase OXA-48 and used as its host for this study. We computed the genomic data obtained by Illumina technology with PHASTER to determined that F13 belongs to the Drexlerviridae family. The P. aeruginosa PAO1 strain was used as a host for the Px4, Px5 and Vs1 phages used in this study.

Phage production

We prepared phage solutions using previously published protocols⁶⁷. Briefly, we grew host bacteria in LB broth at 37°C until they reached the exponential phase. At that point, we initiated and monitored infection with the respective phage for 4 (for T7, T4, P1 phages) to 6 hours (for MS2, G4, F1, HK544, and Φ80 phages). When we observed a decrease in optical density (OD), we allowed the lysis process to continue for an additional hour. We then transferred the cultures to polystyrene Falcon tubes for centrifugation and subsequent filtration using a 0.2 μm filter (Sartorius, Spain) to collect the lysate containing the phage. Next, we titered the phage solutions to determine the plaque-forming units (PFU). For plating, we mixed bacterial cultures (OD₆₀₀ ∼1) with top agar medium and poured it onto agar dishes. After allowing it to dry for 30 minutes, we spotted serial dilutions of the phages onto the agar plates and incubated them overnight at 37°C⁶⁸.

In silico analysis

We generated the list of gcus using the INTEGRALL database (integrall.bio.ua.pt) and Integron finder^33,44. To determine if the gcus associated with known phage resistance defense systems, we subjected the sequences of all gcus to a BLAST search against the two main defense system databases: DefenseFinder⁶⁹ and PADLOC³⁵. For gcus that showed a resistance phenotype but remained unidentified by these platforms, we used an AlphaFold2 pipeline known as ColabFold for prediction, followed by Foldseek to gain insights into the protein function of each gcu^70,71. Additionally, we employed the Iterative Threading ASSEmbly Refinement (ITASSER) platform to leverage complementary approaches for protein function prediction⁷². Genetic environments were anlayzed with CAGECAT (cagecat.bioinformatics.nl)⁷³.

We carried out the sequencing of the K. pneumoniae isolate KP5 using a combination of Illumina and Nanopore technologies, with SeqCoast (Portsmouth, USA) performing the sequencing. We performed genome assembly using Unicycler software and conducted subsequent analyses using the Center for Genomic Epidemiology platform. We determined the resistome using the ResFinder web platform, identified the sequence type using MLST, and characterized the plasmids using the PlasmidFinder web platform^74–76. We assessed the presence of prophages in the genome using the PHASTER web tool⁷⁷. General DNA sequence analysis was performed with Geneious and figures were created with BioRender.com.

Cloning gcus in pMBA

All gcus were synthesized as double-stranded DNA by IDT ((Newark, USA), with an addition of a 20 bp homology region corresponding to the targeted cloning site of pMBA located between the intI1 and gfp gene³². We performed the cloning process using the Gibson assembly method⁷⁸ and selected clones on an LB agar plate supplemented with zeocin (50μg/mL). We cloned the entire sequences of the gcus, including the open reading frame (ORF) and their attC region, into the pMBA vector. Our in-house designed pMBA plasmid contains a p15A replication origin, a zeocin resistance marker, and a truncated intI1 gene followed by a gfp gene, acting as a second integron cassette. We inserted the cloned gcus as the first cassette in the vector. To ensure a relevant level of expression for each gcu, we selected a strong version of the Pc promoter, PcS, located within the intI1 gene, for our constructions. For the construction of pArray-1 and pArray-2, we cloned at second position of the integron-like structure of pMBA the bla_OXA-10 gene or the pic167.2 instead of having the gfp gene, respectively. For the construction of pArray-3, we cloned pic23 and pic 24 at first and second position, respectively, of the integron-like structure present in pMBA under the control of a PcW promoter.

The construction of pArrayØ was accomplished using the Gibson assembly protocol, wherein the aacA54 and aacA8 resistance gene cassettes were cloned into the pMBA-Δgfp vector. Subsequently, pArray-pic24 was created by cloning the pic24 gene into the third position of the synthetic integron, downstream of the aacA8 gene cassette.

For the cloning of gcus into Pseudomonas aeruginosa PAO1, we used a modified pMBA vector (pBTZ). The p15A replication origin was replaced with the BBR1 origin to ensure replication in both E. coli and P. aeruginosa species. Additionally, the strong PcS promoter was substituted with its weaker version, PcW. Finally, a tetracycline resistance marker was introduced, as P. aeruginosa exhibits low resistance to zeocin. The list of oligonucleotides used can be found in Table S2. The sequences of PICs can be found in Table S3.

Growth curves and phage lysis monitoring

We inoculated a single colony of a clone carrying a potential phage resistance gene into an overnight culture of Luria Bertani (LB) broth at 37°C. The next day, we adjusted the culture to an optical density of 0.1 and transferred it to a 96-well plate containing LB broth supplemented with zeocin (50 μg/mL). We then exposed the plate to different concentrations of phage to test various multiplicities of infection (MOI). For all tested phages, we defined a high MOI condition as an MOI of 1 and a low MOI condition as an MOI of 0,01. However, for phages T4 and T7, which exhibit high lytic activity, we set the high MOI condition to 0,01 and the low MOI condition to 0,00001. We monitored growth curves using a Biotek HTX synergy plate reader (Agilent, USA) at 600 nm over a 16-hour period. To assess the phage lysis level, we calculated the area under the curve (AUC) for each growth condition, both with and without phage as described previously⁴³. The inhibition score for each sample was calculated using the following formula: where ΔAUC is the difference in AUC between the conditions with and without phage. Specifically, ΔAUC = AUC_{no phage} − AUC_phage We calculated the inhibition scores for both high and low multiplicity of infection (MOI) conditions. These scores were then normalized by comparing them to the inhibition scores obtained from the control samples, which used the empty pMBA vector. The normalization was done by computing the ratio of the inhibition score of the experimental sample to the inhibition score of the control: . By using the “inhibition score”, we normalize the reduction of the area under the curve relative to growth without the phage. This allows for comparison of the relative effects regardless of initial growth differences between strains. The protective effect of the PIC is shown as the inverse of the inhibition score.

Prophage activation assay

We chemically activate the lytic cycle of the prophages HK544 and Φ80 in E. coli 594::pMBA-pic using mitomycin C at a concentration of 0,5 µg/mL. Briefly, we adjusted the culture to an optical density of 0.1, and after 1h we induced prophage activation by adding mitomycin C. Then, we quantitated the production of active phages after 6 hours of induction. To do so, we centrifuged the culture we performed serial dilutions of the supernatant and spotted it on top agar plate containing IJ-pMBA to evaluate the phage concentration⁷⁹.

Phage-resistant clone screening

To identify new phage defense integron cassettes, we employed both liquid and solid media screening strategies. Initially, we cultured clones carrying various gcus overnight in LB broth with zeocin (50 μg/mL). The following day, a 1 OD dilution of each culture (10 μL) was spotted onto LB agar plates supplemented with zeocin. After 1 hour of incubation, we applied 7 μL of phage T4 solution at a concentration of 10⁹ PFU/mL to the bacterial spots. After 16-18 hours of incubation, we selected the clones that exhibited growth despite phage presence for further analysis. Additionally, we conducted a high-throughput screening in 96-well plates, challenging the entire gcu collection with phages T7 and T4 under two different multiplicities of infection (MOI) conditions, following the protocol described previously.

Plasmid stability experiment

We co-transformed E. coli IJ862 with both the empty pMBA plasmid carrying the p15A replicon and the L-arabinose inducible pBAD plasmid carrying either the pic128 or ddmABC defense system. For five days, three replicates were put in culture in LB broth supplemented with carbenicillin (100 μg/ml) and L-arabinose (0,1%). Daily, we transferred 50 μl of each culture into fresh 5 ml LB broth and used another 50 μl to count GFP-expressing colonies on LB agar plates supplemented with carbenicillin.

To assess the stability of the endogenous clinical plasmids in KP5 K. pneumoniae isolate, we cultivated clones harbouring pic128 in LB broth with zeocin (50 μg/ml) over five days. On days 1 and 5, we selected >120 colonies from LB zeocin agar plates and subcultured them onto separate LB media containing either tetracycline (15 μg/ml), cefotaxime (50 μg/ml), or ertapenem (30 μg/ml). This step aimed to detect any loss of resistance, which would suggest the corresponding loss of one of the three natural plasmids.

attC recombination experiments

Recombination experiments were conducted as previously described⁸⁰. Briefly, we cloned the 3’ region of each pic using a pSW23T mobilizable plasmid containing a chloramphenicol resistance marker into E. coli β2163 strain (dapA^-, pir⁺) requiring diaminopimelic acid (DAP) to grow. We used E. coli β2163 as donor strain for conjugation experiments using E. coli DH5α strain carrying both a p3938 plasmid (Carb^R) containing an L-arabinose inducible intI1 gene and a p929 plasmid (Kana^R) containing an attI site as recipient. We performed conjugation experiments onto LB agar supplemented with DAP and L-arabinose (0,2%) overnight. The next day, after cell recover, we performed ten-folded dilutions onto LB agar supplemented with carbenicillin (100 µg/ml) and glucose (1%) or with carbenicillin, glucose and chloramphenicol (25 µg/ml) to select for the total population and recombinants, respectively.

Fitness experiment

To assess the fitness cost of each PIC, we performed in vitro competition assays following the protocol described in⁸¹ with some modifications. Briefly, we subcloned the panel of the 13 PICs into a modified version of the pMBA vector, which includes a knock-out gfp gene (with a stop codon instead of the start codon), along with an empty vector control. We then subjected this new set of clones to a fitness competition experiment against the wild type pMBA vector carrying the intact gfp gene. Briefly, we grew six replicates of each clone overnight at 37°C in a 96-well plate filled with LB broth supplemented with zeocin (50 μg/ml). The next day, we mixed cells containing the target gcu in the pMBA-Δgfp in a 1:1 volume ratio with the clone carrying the pMBA-WT vector. We diluted these mixtures 1:20 in 96-well plates containing NaCl 0,9%. This plate was further diluted 1:20 in a plate with NaCl and another with LB broth supplemented with zeocin. The first plate was used to estimate the percentage of pMBA-WT and pMBA-pic-Δgfp cells in the mixture on day 1, using a Cytoflex S cytometer (Beckman Coulter, USA). The second plate (with LB) was incubated overnight. The next day these cultures were diluted 1:400 in NaCl to determine the proportion of strains in the mixture as explained above. Following the formula described in⁸¹, fitness cost was calculated as the difference in ratios of GFP-fluorescence between day 1 and day 2 for each PIC.

Quantitation of polar effects

To determine the polar effects of cassettes, we measured the fluorescence of the Green Fluorescent Protein (GFP) that acts as a second cassette in pMBA. At least three independent colonies of each pMBA derived strain were inoculated in LB zeocin and incubated at 37°C overnight. Cultures were then diluted 1/400 in filtered saline solution to measure fluorescence using a Cytoflex-S flow cytometer (Beckman Coulter). The 488 nm laser was used to detect GFP expression through 525/40 nm (FITC) band pass filter. 20,000 events were recorded per sample. Data were analysed using CytExpert software (v.2.4; Beckman Coulter).

DATA AVAILABILTY

The genome of strain KP5 has been deposited in NCBI GenBank (accession numbers: CP162381-CP162384)

CONFLICT OF INTERESTS

None to declare.

ACKNOWLEDGEMENTS

We would like to thank Dr. José R. Penadés and Dr. James J. Bull for providing phages and strains, and Tamara Barcos for technical assistance. We thank André Carvalho for critical reading of the manuscript. NK is supported by funding from the European Union’s Horizon 2020 research and innovation programme under the Marie Skłodowska-Curie grant agreement No 847635 for the ANARC research project (No 4230415). The work in the MBA laboratory is supported by the European Research Council (ERC) through a Starting Grant [803375]; Ministerio de Ciencia, Innovación y Universidades [BIO2017-85056-P; CNS2022-135857]; Ministerio de Ciencia e Innovación [PID2020-117499RB-100] and the EU HARISSA JPI-AMR program [PCI2021-122024-2A]; A.H. is supported by the PhD program at UCM. PB is supported by the Juan de la Cierva program (FJC 2020-043017-I). P.D-C. was supported by a Ramón y Cajal contract RYC2019-028015-I and the project PID2020-112835RA-I00, funded by MCIN/AEI/10.13039/501100011033, ESF Invest in your future; MG-Q is supported by the Subprograma Miguel Servet from the Ministerio de Ciencia e Innovación of Spain (CP19/00104), Instituto de Salud Carlos III (Plan Estatal de I+D+i 2017–2020), and co-funded by European Social Fund “Investing in your future”. J.A.E. is supported by the Atracción de Talento Program of the Comunidad de Madrid [2016-T1/BIO-1105, 2020-5A/BIO-19726];

Footnotes

• We have added a section on the genetic context of PICs in other genomes.