Plasmids pick a bacterial partner before committing to conjugation

Abstract Bacterial conjugation was first described by Lederberg and Tatum in the 1940s following the discovery of the F plasmid. During conjugation a plasmid is transferred unidirectionally from one bacterium (the donor) to another (the recipient), in a contact-dependent manner. Conjugation has been regarded as a promiscuous mechanism of DNA transfer, with host range determined by the recipient downstream of plasmid transfer. However, recent data have shown that F-like plasmids, akin to tailed Caudovirales bacteriophages, can pick their host bacteria prior to transfer by expressing one of at least four structurally distinct isoforms of the outer membrane protein TraN, which has evolved to function as a highly sensitive sensor on the donor cell surface. The TraN sensor appears to pick bacterial hosts by binding compatible outer membrane proteins in the recipient. The TraN variants can be divided into specialist and generalist sensors, conferring narrow and broad plasmid host range, respectively. In this review we discuss recent advances in our understanding of the function of the TraN sensor at the donor-recipient interface, used by F-like plasmids to select bacterial hosts within polymicrobial communities prior to DNA transfer.


INTRODUCTION
Conjugation, like transduction and transformation, is central to bacterial evolution, as it facilitates the acquisition and dissemination of virulence and antimicrobial resistance (AMR) genes ( 1 ).Conjugation is known to take place in a broad range of environments, including soil, the surface of plants and medical devices as well as in the lumen of the gut, which is considered a hot spot of gene exchange in bacteria of significant clinical importance ( 2 , 3 ).Since the discovery of conjugation almost 80 years ago ( 4 ), many functional studies have been performed using donor and recipient Esc heric hia coli with F or F-like (IncF) plasmids, which are predominantly isolated from Enterobacteriaceae ( 5 ).pMAR7, encoding the bundle forming pilus (BFP) in typical enteropathogenic E. coli (EPEC) and pSLT, encoding type III secretion system (T3SS) effectors in Salmonella enterica , are key F-like virulence plasmids ( 6 , 7 ).R100, first identified in an isolate of Shigella flexneri in 1956 , encoding multiple resistance genes, and pKpQIL, found in current high risk Klebsiella pneumoniae sequence types (e.g.ST258 / ST512), encoding the KPC carbapenemase, are classical and contemporary resistance F-like plasmids, respecti v ely ( 8 , 9 ).
It has long been known that tailed Caudovirales bacteriopha ges (pha ge) use receptor-binding proteins (RBPs) at the distal end of their tail structures to bind specific bacterial surface pol ysaccharides (e.g.lipopol ysaccharide (LPS) or capsule) and / or proteinacious receptor / s, prior to injection of their DNA.For example, the Siphoviridae phage binds LamB, while the Myoviridae phages T2 and T4 bind OmpA / FadL and OmpC, respecti v ely (re vie wed in ( 10 )).Recent studies have shown that the K. pneumoniae Siphoviridae phages NPat and BMac bind OmpK36 (OmpC homologue) as well as the K2 capsule ( 11 ), while the Serratia sp.ATCC39006 LC53 phage (T4-like) seems to bind OmpW ( 12 ).These interactions determine transduction host specificity and range.In contrast, conjugation of IncF plasmids has been regarded as a promiscuous mechanism of DNA transfer, with host range being determined w holl y by the r ecipient, downstr eam of plasmid transfer, via replica tion associa ted factors (e.g.incompa tibility groups) ( 13 ), and restriction modification and CRISPR Cas systems ( 14 ).
The transfer ( tra ) genes in conjugati v e IncF plasmids comprise a contiguous operon of a pproximatel y 40 kb ( 15 ) (Figure 1 ).The conjugation process itself can be divided into three phases.The first phase occurs e xclusi v ely within the cytosol of the donor.This includes assembly of the conjugati v e transfer machinery, related to type IV secretion systems (T4SS) ( 16 ), which facilitates assembly of the sex pilus ( 17 ).The second phase involves both the donor and recipient, starting with pilus-media ted ma ting pair formation (MPF) followed by TraN-mediated mating pair stabilization (MPS) ( 18 ).Plasmid transfer and the expression of plasmid-encoded genes in the recipient initiate the third phase of conjugation, which renders the r ecipient r efractory to a second wave of conjugation by the same plasmid.TraT (localized to the outer membrane (OM)) and TraS (found in the inner membrane) participate in this process via mechanisms known as surface and entry e xclusion, respecti v ely ( 19 , 20 ).Surface and entry exclusion protect cells from lethal zygosis, wher e r ecipients ar e killed due to membrane damage when they partake in e xcessi v e conjugation acti vity ( 21 ).

Mating pair formation
The sex pilus is an indispensable constituent of F and Flike plasmid conjugation; it is built as a thin fle xib le filament composed of polymerized pilin subunits encoded by traA ( 22 , 23 ).The F plasmid 121 amino acid TraA pro-pilin is cleaved into a 70 amino acid mature pilin subunit by TraQ and TraX ( 24 ).These pilin subunits accumulate in the inner membrane before they are assembled by 11 tra gene products.TraL, E, K, C and G participate in formation of the pilus tip ( 25 ).TraB, V, W, F and H are needed for extension of the pilus, TrbC is necessary for pilus biogenesis, although its precise function is unknown, and TraP stabilizes the extended filament.The TraI relaxase nicks the plasmid at the oriT site leading to formation of a TraI-ssDNA complex, which is recruited to the T4SS by the coupling protein TraD tha t initia tes DNA tr ansfer ( 19 , 20 ); Tr aU also plays a role in DN A transfer, w hile TrbI plays a role in pilus retraction (26)(27)(28)(29)(30)(31).
The structures of the sex pili encoded by the F and Flike plasmids pED208 and pKpQIL have been determined by single-particle cryo-EM ( 32 , 33 ).This re v ealed that the pilin subunits form helical assemblies with phospholipid molecules at a stochiometric ratio of 1:1 ( 32 , 33 ).The lumen of the sex pilus is ∼25 Å in diameter, the external diameter is ∼85 Å and the average length is 20 m ( 32 , 34 ).While the structure of the pilus has been determined, our understanding of its role in conjugation remains incomplete.It is broadly recognised that the pili are important for the initial contacts between the donor and recipient during MPF ( 35 ).Howe v er, the molecular basis of MPF remains unknown as the pilus receptor on the recipient has not yet been identified, partially because the composition of the pilus tip remains undefined.Using li v e-cell fluorescence microscopy, Clarke et al. demonstrated that the pilus is a highly dynamic structure and tha t pilus-media ted interaction between two Hfr bacterial cells triggers its retraction leading to the formation of cell-cell contacts ( 36 ).Babic et al .provided evidence that plasmid DNA can be transferred from a donor to a distant recipient ( 37 ).Recently, Goldlust et al. have shown that distant plasmid transfer occurs through the center of the pilus, answering a longstanding key question in conjugation biology (doi: https: //doi.org/10.1101/2023.06.21.545889 ).

Mating pair stabilization
A key publication by Achtman and colleagues in 1978 had shown that following F pilus retraction, conjugating cells form ma ting aggrega tes tha t ar e r esistant to disruption by shear forces ( 38 ), which bacteria may encounter in different niches, such as peristalsis in the gut ( 39 ).Close inspection of conjugating bacteria re v ealed that they form tight 'mating junctions', characterised by intimate w all-to-w all contact through a process later termed mating pair stabilization (MPS) ( 40 ).Initially, MPS was hypothesized to mediate an interaction between the tip of the pilus in the donor and a recipient receptor.Howe v er, it was later found that mutations in traN and traG affected the formation of mating aggregates without affecting the pilus ( 41 ), suggesting that intimate w all-to-w all contact is distinct from pilus-mediated MPF.
TraG is a multifunctional inner membrane protein ( 42 ).While its N-terminus plays a role in pilus assembly, the Cterminus is r equir ed for MPS; howe v er, the mechanism by which TraG contributes to MPS remains elusi v e. TraN encoded by the F plasmid is a 602 amino acids (aa) outer membrane protein (OMP) that consists of 22 cysteine residues, of which six are important for optimal plasmid transfer ( 43 , 44 ).TraN encoded by the F-like plasmids pED208 and pOX38 was recently reported to contribute quantita- ti v ely to pilus production, conjugation efficiency and pilus extension / retraction dynamics ( 45 ).
While TraN shares little sequence identity with other known OMPs or adhesins, Klimke et al. determined its membrane topology and re v ealed the e xistence of three e xtracellular loops, which wer e pr edicted to be involved in r eceptor r ecognition ( 44 ).These loops corr espond to a r egion spanning around 200 amino acids, which shares low sequence similarity between TraN homolo gues, w hile the N-and C-terminal domains are highly conserved amongst F-like plasmids.(Figure 2 A).Importantly, the TraN homologues of a similar size ( ∼600 aa) are only found in IncF plasmids.In other plasmids MPS is mediated by different mechanisms.
MPS in the Salmonella Typhimurium plasmid R64 (IncI) is mediated by a thin fle xib le type IV pilus (T4P) encoded by the pil locus (located upstream of the tra operon).Donor -recipient interactions are mediated by binding of PilV, loca ted a t the tip of the T4P, to LPS.Inversion within the shufflon can form se v en different PilV adhesins that bind specific LPS moieties on different recipients ( 46 , 47 ).In the Enter ococcus f aecalis pheromone-inducible conjuga tive plasmid pCF10, ma ting aggrega tes are formed by interactions between the plasmid-encoded aggregation substance protein PrgB on the donor and lipoteichoic acid on the receipt ( 48 ).In contrast, no specific recipient factors have been identified for conjugation of the broad host range plasmids R388 (IncW) and RP4 (IncP) ( 49 , 50 ).

T r aN sensors in the donor cooperate with distinct OMPs in the recipient
Studies intending to discover the sex pilus receptor in the recipient identified mutations in LPS biosynthetic genes, particularly the LPS core, and ompA , encoding the OMP OmpA, as negati v ely af fecting conjuga ti v e uptake of the F plasmid ( 51 ).Three classes of ompA mutations were identified ( 52 ): mutants not expressing OmpA, mutants expressing lower le v els of OmpA and mutants encoding missense mutations including a G154D substitution ( 53 ).Moreover, mutations in ompA were found to specifically affect transfer of the F plasmid but not the related F-like plasmid R100-1 ( 54 ).Building on this specificity, seminal work from the lab of Laura Frost in the 1990s found that OmpA was not the receptor for the sex pilus, as substitution of traA on the F plasmid with traA from the R100-1 plasmid did not abroga te the ef fect of OmpA muta tions ( 55 ).Instead, the ompA muta tions af fected MPS, as dependency was associated with TraN, specifically the highly variable region of the protein ( 55 ).
The cooperation between the F plasmid TraN and OmpA in E. coli conjugation was recently confirmed, howe v er, the F plasmid TraN did not recognise OmpA in a K. pneumoniae recipient ( 56 ).This, together with the finding that substitution of the F plasmid traN with traN of the R100-1 plasmid bypasses the dependency on OmpA ( 55 ), suggested that Tr aN-OMP inter actions mediate conjugation specificity.These results also suggest that the recipient is not merely a bystander but instead participates in MPS.
Analysis of TraN sequences from 824 putati v e conjugati v e IncF-like plasmids in Enterobacteriaceae isolates revealed that 32%, 20% and 22% of plasmids encoded a TraN sharing ≥90% amino acid similarity to TraN of the pKpQIL, R100-1 and F plasmids, respecti v ely ( 56 ).Analysing the remaining 215 plasmids led to the identification of four other TraN variants, one of which was found solely in Salmonella enterica serovars and specifically aligned to TraN from the virulence plasmid pSLT.The other three minor variants, labelled MV1-3, were not associated with well-known plasmids ( 56 ).Of note, while the 22 Cys residues of the F plasmid TraN are conserved across the family, the Cys content of TraN in different F-like plasmids can be > 22 ( 56 ).
In the absence of an experimentally determined structure, the different TraNs were subjected to AlphaFold structural prediction ( 57 ).The TraN structure is composed of two domains: the base and an extended tip, which is composed of the highly variable sequence of the protein (Figure 2 A and  B).The 22 conserved Cys r esidues ar e found within the base domain and predicted to be paired via disulphide bonds, which could stabilise the structure ( 56 ).
The base consists of a conserved amphipathic alphahelix that can potentially anchor the protein to the outer membrane (Figure 2 B).Despite low sequence similarity, the tip folds into a conserved structur e (Figur e 2 A), consisting mostly of ␤-sheets linked to a ␤-sandwich domain (Figure 2 B).Structural differences between the different TraNs are mainly seen within the exposed loops of the tip domain, which functions as the TraN sensor (Figures 2 B and   56), making it the only TraN variant currently known to cooperate with more than one OMP (Figure 3 B).Structural determination of the TraN ␤ -OmpK36 complex and the predicted AlphaFold complex of TraN ␤ -OmpK35 revealed that the unique ␤-hairpin loop of the TraN ␤ sensor is inserted into one of the porin trimer subunits ( 56 , 58 ).This transcellular pr otein-pr otein interaction, which likely represents the molecular basis of MPS and conjugation specificity ( 56 ), parallels the recognition of bacterial hosts by the tail fibre RBPs of Caudovirales phages ( 10 ).
While low membrane abundance of OmpA ( 53) and OmpK35 ( 58 ) affect conjugation efficiency of F and pKpQIL plasmids respecti v ely, lowering the abundance of OmpK36 does not affect conjugation of pKpQIL ( 58 ).
Moreover, a single amino acid difference in loop 3 of OmpW between E. coli (N142) and Citr obacter r odentium (A142) affected their recipient activity.While E. coli OmpW was able to mediate MPS with both TraN ␣1 expressed by R100-1 and TraN ␣2 expressed by pSLT, OmpW of C. rodentium was only able to mediate MPS with TraN ␣2.An N142A substitution in OmpW of C. r odentium was suf ficient to r estor e TraN ␣1-mediated MPS ( 58 ).This is consistent with what has been shown for OmpA, where a single amino acid mutation (G154D) inhibited TraN ␥ -mediated conjugation ( 51 , 52 ).Of note, Ried and Henning showed in 1987 that E. coli expressing the OmpA G154D substitution was also resistant to specific phages ( 53 ), suggesting TraN ␥ and the phage RBPs share the same OmpA binding site.Mechanistically, the AlphaFold models suggest that the OmpW N142A and OmpA G154D substitutions cause steric clashes with TraN ␥ and TraN ␣1, respecti v el y.To gether, this shows that subtle differences in the recipient OMPs affect binding of TraN sensors and phage RBPs, suggesting that potential bacterial hosts can e volv e to resist both phage infections and plasmid conjugation.
It is important to emphasise that while the TraN tip sensors have evolved multiple tertiary structures (Figure 3 A), the structure of the major OMPs is highly conserved between species.Despite the close sequence and structural similarity of K. pneumoniae OmpK36 and E. coli OmpC (Figure 3 C), TraN ␤ mediates MPS specifically with the former.This finding supports the hypothesis that conjugation is not a promiscuous mechanism of DNA transfer, but instead TraN functions as a sensiti v e sensor, enab ling the selection of specific r ecipients.Conversely, OMPs ar e evolving in response to selecti v e pressure, for example due to exposure to antibiotics.In K. pneumoniae , exposure to carbapenems have selected for truncation of OmpK35 and OmpK36 insertion mutants, both of which reduce antibiotic diffusion across the OM ( 59 ).The insertion mutants are characterised by single (D) or double (GD or TD) amino acids insertions into loop 3 of OmpK36, which constrict the porin pore ( 60 ).The OmpK36 loop 3 insertions not only synergise with the pKpQIL-encoded carbapenemase to increase carbapenem resistance, but also, inad vertentl y, reduce conjuga tion ef ficiency due to clashes with the ␤ hairpin of the TraN ␤ tip and the porin ( 56 , 58 ).Interestingly, by and large clinical isolates expressing OmpK36 with loop 3 insertions already contain pKpQIL, suggesting that they may function on behalf of the plasmid as a proxy surface exclusion mechanism.

T r aN sensor influences the distribution of IncF plasmids in clinical isolates
A TraN phylogenetic tr ee r e v eals clustering of the different tip variants into distinct clades, which are grouped with the Cys residue content and associated with one or more bacterial genera (Figure 4 ).Analysis of the different tip variants suggests that they could be divided into specialist (TraN ␤ and TraN ␥ ) and generalist (TraN ␣ and TraN ␦) sensors, which exhibit narrow and broad host range, respecti v el y ( 56 ).Importantl y, w hile the specialist TraNs ar e pr edominantly found in a single species (e.g.TraN ␥ is found Figur e 4. A phylo genetic tree of TraN.The tree (made with IQ-Tree) ( 64 ) consists of 639 TraN protein sequences (clustered following sequence alignment with Clustal Omega).These include 632 from Uniprot, filtered with 500-800 amino acids, ≤27 Cys residues and ≥ 30% amino acid similarity to TraN from the F plasmid.An additional se v en TraN sequences from the R100-1 (T raN ␣1), pSLT (T raN ␣2), pKpQIL (T raN ␤1), MV2 (T raN ␤2), F (T raN ␥ ), MV1 (TraN ␦1) and MV3 (TraN ␦2) r efer ence plasmids were included and highlighted.Metadata blocks show the host genus of the plasmid and the number of Cys residues in TraN.The scale bar r epr esents the number of substitutions per site.Each entry is associated with a UniProt accession code, and the structure prediction for each variant is available at: https://alphafold.ebi.ac.uk .An interactive version of the tree is available at: https://micr oreact.org/project/tran .at a frequency of 92% in E. coli ) they are also found at a lower frequency in other species (e.g.TraN ␥ is found at a frequency of 5.6% in S. enterica) ( 56 ).For illustration, the distribution of TraN sensors in a small selection of plasmids found in commensal and pa thogenic Gram-nega ti v e bacteria is shown in T able 1 .T ak en together, these real-w orld distributions suggest that plasmids use TraN sensors to pick partners for dissemination within polymicrobial communities prior to conjugation.

CONCLUSIONS AND PERSPECTIVE
Most studies of F-like plasmid conjugation to date have been done using specific donor and recipient pairings (mainly E. coli ) in either solid or liquid rich laboratory me-dia.These investigations have shown that while MPS accelera tes conjuga tion ef ficiency, promiscuous low-frequency transfer can happen e v en in its absence (doi: https://doi.or g/10.1101/2023.06.21.545889 ; 56).Ho we v er, F-like plasmid distribution in the real world suggests that where bacteria are experiencing shear forces, successful conjugation is reliant on MPS ( 56 ).This suggests that under physiological scenarios engagement of the pilus with a recipient is not suf ficient for conjuga tion, but tha t via their TraN sensors plasmids pick their bacterial hosts and guide their own dissemination.Ther efor e, analogous to Caudovirales bacteriophages that use their tailed structures to bind bacterial surface receptors, it seems plasmids have also evolved a mechanism to selecti v ely propagate in specific recipient species within pol ymicrobial comm unities.Interestingl y, tail fibre RBPs and TraN isoform share similar OMP receptors.The questions of why some plasmids seem to avoid potential recipients (considering that that a copy of the plasmid remains in the donor), what are the evolutionary pr essur e that dri v e plasmid specialisation, and the reasons some potential recipients resist plasmid entry, are key questions for future studies.
The realisa tion tha t plasmids hav e e volv ed specific sensors to select their hosts r epr esents a ne w vie wpoint in plasmid biolo gy, w hich could potentiall y be used to predict the spread of emerging resistance and virulence plasmids amongst pathogens.

DA T A A V AILABILITY
The data presented in this manuscript would be made available upon written request to the corresponding author.

Figure 1 .
Figure 1.Genetic arrangement, function and subcellular localization of each gene products encoded by the F plasmid tra operon.

Figur e 2 .
Figur e 2. Conservation anal ysis of TraN.Sequence conservation of TraNs mapped onto the TraN encoded by the pKpQIL, as calculated by Consurf ( 61 , 62 ) ( A ) and the AlphaFold model ( B ).The conservation increases from green to purple.TraN is divided into three functional regions: the base, which anchors the protein to the outer membrane, the scaffold tip, and a distal sensor.The base shows the highest degree of sequence conservation whereas the tip and sensor the least.

(
r epr esented by F), and TraN ␦ (r epr esented by MV1 and MV3) ( 57 ).Subtle differences in host selection have resulted in their classification into subgroups: TraN ␣ of R100-1 and pSLT were classified as TraN ␣1 and TraN ␣2, respecti v el y.Similarl y , T raN ␤ of pKpQIL and MV2 and TraN ␦ of MV1 and MV3 were classified as TraN ␤1 and TraN ␤2 and TraN ␦1 and TraN ␦2, respecti v ely ( 58 ).The e volv ed TraN tip sensors selecti v ely pair with specific OMPs in the recipient: Tr aN ␥ inter acts with OmpA, Tr aN ␣ inter acts with OmpW, Tr aN ␦ inter acts with OmpF and Tr aN ␤ inter acts with both OmpK36 and OmpK35 (the K. pneumoniae OmpC and OmpF homologues respecti v ely) (

Figure 3 .
Figure 3.The TraN sensors.( A ) The predicted structures of the e volv ed F-like plasmid-encoded T raN ␣, T raN ␤, T raN ␥ and T raN ␦ tip sensors.The tip scaffold consists of conserved ␤-sheets, shown by gray ribbons.The colored structural motifs r epr esent the surface exposed TraN sensor, each binding a specific receptor on the surface of the recipient.( B ) The different TraN sensors in the donor (D), w hich reco gnize distinct OMPs in the recipient (R), mediate plasmid spread and conjugation species specificity.( C ) The crystal structures of the K. pneumoniae OmpK36 (PDB ID: 6RD3) ( 59 ) and the E. coli OmpC (PDB ID: 2J1N) ( 63 ) are highly similar, yet the TraN ␤ sensor specifically recognizes recipients expressing OmpK36 but not OmpC.

Table 1 .
TraN in resistance and virulence plasmids a Extended spectrum beta-lactamase