Compared to the huge number of Gram-positive bacteriocins, microcins are distinguished by a high structural heterogeneity inside a restricted number of identified and well-characterized representatives. A widely accepted classification was proposed by Duquesne et al. (2007a) based on both the peptide size and degree of posttranslational modification (PTM). The known microcins are grouped in two classes, class I with molecular masses below 5 kDa and the presence of extensive PTM and class II with molecular masses between 5 and 10 kDa that can be modified or not (Table 1). A brief description of the microcins from the two classes is provided below to help following the next sections. For more detailed overview of the microcins, see two recent reviews (Baquero et al., 2019; Li and Rebuffat, 2020).

Class I assembles three plasmid-encoded microcins that have been well structurally characterized as RiPPs (Supplementary Figure S1A): microcin C (McC) a nucleotide peptide, microcin B17 (MccB17) a linear azol(in)e-containing peptide, and microcin J25 (MccJ25), a lasso peptide. McC is presently the only nucleotide member of the family. However, similar biosynthetic gene clusters are distributed within bacterial genomes (Bantysh et al., 2014), which suggests an unexplored diversity for such peptides. McC is produced by E. coli cells harboring the mccABCDEF gene cluster (Figure 1) under a mccA-encoded formylated heptapeptide precursor, which is further modified (Guijarro et al., 1995; Severinov and Nair, 2012) and processed into a structural mimic of aspartyl adenylate which is the toxic entity (Kazakov et al., 2008) (Supplementary Figure S1A). MccB17 is produced as a 69 amino acid precursor by E. coli strains bearing the mcbABCDEFG gene cluster (Figure 1). Mature MccB17 contains 43 amino acids that are structured into thiazole and oxazole heterocycles (4 thiazoles and 4 oxazoles rings either isolated or fused into oxazole/thiazole- and thiazole/oxazole-bis-heterocycles) by the PTM enzymes (Li et al., 1996; Ghilarov et al., 2019) (Supplementary Figure S1A). Such heterocycles are also found in hybrid non-ribosomal peptide-polyketide natural products such as the anti-tumor drug bleomycin, as well as in RiPPs such as cyanobactin (McIntosh and Schmidt, 2010) or streptolysin (Mitchell et al., 2009), forming the LAP [also termed thiazole/oxazole-modified microcin (TOMM)] peptide family (Melby et al., 2011). Microcin B-like bacteriocins produced by Pseudomonas, Klebsiella and Rhizobium have been reported (Metelev et al., 2013, 2017b; Travin et al., 2019). MccJ25 was isolated first from the E. coli strain AY25 isolated from an infant feces bearing the mcjABCD gene cluster (Salomón and Farias, 1992) (Figure 1). Its maturation from a 58 amino acid precursor into a 21 amino acid lasso peptide is ensured by two enzymes, McjB and McjC, encoded in the microcin gene cluster (Duquesne et al., 2007b; Yan et al., 2012). This unique lasso topology, which is characterized by threading of the C-terminal tail through a seven to nine lactam ring closed by an isopeptide bond, is locked in place with the two bulky side chains of Phe and Tyr aromatic amino acids for MccJ25 (Rosengren et al., 2003) (Supplementary Figure S1A). It is responsible for the sturdiness of MccJ25 and is required for its antibacterial activity (Rebuffat et al., 2004; Wang and Zhang, 2018). Genome mining approaches have revealed a wide distribution of lasso peptides in Gram-positive and Gram-negative bacteria (Maksimov et al., 2012; Hegemann et al., 2013; Tietz et al., 2017; Cheung-Lee and Link, 2019). Many lasso peptides produced by proteobacteria do not show antibacterial activity (Hegemann et al., 2013). This questions their ecological role or can be due to difficulty to decipher the reasons for their narrow activity spectrum.

Class II microcins form a more homogeneous group than their class I cousins (Table 1 and Supplementary Figures S1B,C), although they are subdivided into class IIa, encompassing MccL (Pons et al., 2004), MccN/24 (Kaur et al., 2016), MccPDI (Eberhart et al., 2012), MccS (Zschüttig et al., 2012) and MccV (Gratia, 1925), and class IIb (MccE492, MccH47, MccM, Vassiliadis et al., 2010). MccN was formely termed Mcc24 (O’Brien and Mahanty, 1994) and is termed MccN/24 in this review. What distinguishes class IIa from class IIb is the presence or not of a siderophore moiety derived from enterobactin anchored at the peptide C-terminal serine carboxylate (Supplementary Figures S1B,C). This catechol-type siderophore PTM sparked coining the name “siderophore microcins” to class IIb microcins (Rebuffat, 2012). Class II microcins result from a proteolytic processing of a precursor with a leader peptide extension, which occurs at a conserved double-glycine (or Gly-Ala) cleavage site, concomitantly with secretion. They have molecular masses between 5 and 10 kDa and exhibit high amino acid sequence similarities, even between class IIa and IIb (Supplementary Figures S1B,C). For examples, the class IIa unmodified MccV and MccN/24 possess high sequence similarities with the class IIb MccH47 and MccE492, respectively, although they do not carry a C-terminal PTM (O’Brien, 1996; Corsini et al., 2010). It was suggested that the conserved C-terminal sequence of these microcins can direct the presence or not of the siderophore PTM and that the C-terminal regions of MccV and MccH47 can be interchanged (Azpiroz and Laviña, 2007). It was further proposed that both class IIa and IIb microcins possess a modular structure (Azpiroz and Laviña, 2007; Morin et al., 2011).

Class IIa microcins have been characterized from E. coli strains from various origins. The MccN/24 producer is an uropathogenic E. coli (Kaur et al., 2016) and the MccL producer comes from poultry intestine (Sablé et al., 2003), while MccS is produced by a probiotic strain, E. coli G3/10 (Symbioflor2^®; DSM17252) (Zschüttig et al., 2012). The producing strains are in some cases multi-microcin producers, such as E. coli LR05 that secretes MccB17, MccJ25 and the uncharacterized MccD93 in addition to MccL (Sablé et al., 2003). Their gene cluster organization includes the four basic genes only, one structural gene encoding the precursor peptide, two export genes and one immunity gene (Zschüttig et al., 2012) (Figure 1). If the five class IIa microcins are all devoid of PTMs, they are also all except MccN/24, stabilized by one (MccV) or two (MccL, MccPDI, MccS) disulfide bonds (Yang and Konisky, 1984; Sablé et al., 2003; Gerard et al., 2005; Morin et al., 2011; Zschüttig et al., 2012) (Table 1 and Supplementary Figure S1B).

Contrasting with class IIa and class I, class IIb microcins (Supplementary Figure S1C) are chromosome-encoded (Poey et al., 2006). MccE492 is secreted by Klebsiella pneumoniae human fecal strain RYC492 (de Lorenzo, 1984) bearing the mceABCDEFGHIJ gene cluster (Destoumieux-Garzón et al., 2006; Vassiliadis et al., 2007; Nolan and Walsh, 2008) (Figure 1). It is the first siderophore microcin to be characterized (Thomas et al., 2004), although it was primarily described as an unmodified peptide (Wilkens et al., 1997). Actually, it was shown further to be secreted under both modified and less active unmodified forms, due to its PTM process (Vassiliadis et al., 2007). The MccE492 PTM was identified as a glucosylated linear trimer of N-(2,3 dihydroxybenzoyl)-L-serine (DHBS) linked to the C-terminal serine carboxylate (Supplementary Figure S1C). The functions of the enzymes involved in establishment of the MccE492 PTM, MceC, MceD, MceI/MceJ, were identified (Vassiliadis et al., 2007; Nolan and Walsh, 2008). MccH47, initially isolated from the human fecal E. coli strain H47 (Laviña et al., 1990) and MccM were both characterized as siderophore microcins produced by several E. coli strains, including the probiotic strain Nissle, 1917 (Mutaflor^®) (Vassiliadis et al., 2010). MccH47 and MccM carry the same PTM as MccE492 (Vassiliadis et al., 2010). Siderophore microcins possess a modular structure, where the N-terminal region is responsible for their cytotoxicity and the C-terminal region, which carries the siderophore moiety, is involved in recognition and uptake. For an overview on siderophore microcins, see Massip and Oswald (2020).

Microcin production takes place in the stationary phase (Baquero and Moreno, 1984) of bacterial growth, with the exceptions of MccE492 (de Lorenzo, 1984) and MccPDI (Eberhart et al., 2012). They are encoded by gene clusters, which exhibit a conserved organization, but contain a variable number of genes ranging from four to ten, according to the presence or not of PTMs on the mature microcin (Figure 1). These gene clusters are generally plasmid-borne, except the chromosomally encoded class IIb microcins. The general biosynthetic pathway of microcins (which also applies to other bacteriocins) starts with the ribosomal synthesis of a precursor peptide that is typically composed of two regions, an N-terminal leader part and a core region. The core peptide of modified microcins, which belong to the wide RiPP family, is the region where the PTMs take place (Montalbán-López et al., 2020). In some cases, such as the siderophore microcins, the modifications may result from the non-ribosomal pathway, making these microcins a rare bridge spanning ribosomal and non-ribosomal biosynthesis pathways (McIntosh et al., 2009). The leader is involved in binding to or activation of many of the PTM enzymes, but also maintains the maturing peptide inactive during the process (Arnison et al., 2013), thus contributing to the protection of producing cells as regard their own toxic microcin. For many modified microcins (MccJ25, McC), this binding involves a peptide binding domain (RiPP precursor peptide recognition element, RRE), also present in a wide proportion of RiPP PTM enzymes and similar to a small protein involved in the biosynthesis of the RiPP pyrroloquinoline quinone (PQQ) (Burkhart et al., 2015; Sikandar and Koehnke, 2019). Recently, the crystal structure of the McbBCD synthetase ensuring the extensive modifications in MccB17 was solved, deciphering the organization and functioning of such a multimeric heterocyclase-dehydrogenase catalytic complex at the molecular level and affording the spatial relationships between the two distinct enzymatic activities and the leader peptide binding site (Ghilarov et al., 2019).

In all but a few cases, and irrespective of if the microcin is modified or not, maturation requires removal of the leader region to give the active bacteriocin (Drider and Rebuffat, 2011). This proteolytic cleavage is performed either before and independently of (class I), or concomitantly with (class II) export of the mature microcin (Beis and Rebuffat, 2019). It can be ensured either (i)- concomitantly with the PTM establishment by one of the dedicated enzymes (MccJ25 leader is cleaved off by the McjB leader peptidase encoded in the microcin gene cluster (Yan et al., 2012), or (ii)- by a protease from the producer, which is not encoded in the microcin gene cluster (MccB17 leader is cleaved off before export by the conserved proteins TldD/TldE which assemble as a heterodimeric metalloprotease to ensure this function) (Ghilarov et al., 2017), or (iii)- by a bifunctional ATP binding cassette (ABC) transporter of the peptidase-containing ATP-binding transporters (PCAT) family, which is encoded in the microcin gene cluster (cleavage of the class II microcin leader peptides is performed simultaneously with export of the maturated microcins by an ABC exporter endowed with an N-terminal protease extension) (Håvarstein et al., 1995; Massip and Oswald, 2020).

Microcin gene clusters vary in the number of genes contained and the presence of genes encoding PTM enzymes, and they all carry genes ensuring self-immunity (Figure 1). Each microcinogenic strain is protected against its arsenal of microcins and the self-immunity mechanisms differ from one microcin to another. For instance the self-immunity mechanism to McC is complex and relies on the products of three genes mccC, mccE, and mccF that ensure export of unprocessed microcin outside the cells (MccC pump) and modification of processed McC (MccE and MccF enzymes) (see section mechanisms of resistance) (Novikova et al., 2010; Agarwal et al., 2011, 2012). By contrast, the immunity mechanism to MccL depends on a single gene mclI that encodes an immunity protein (Sablé et al., 2003). Overall, self-immunity of the producers relies either on specific immunity proteins encoded in the gene clusters that bind to the toxic entities making them inefficient, or on efflux systems, mainly ABC transporters, which ensure export of the microcins to the external medium and simultaneously self-immunity of the producing bacteria. As examples, self-immunity to MccJ25 is provided exclusively by McjD, a highly specific ABC exporter which ensures simultaneously export of the microcin (Beis and Rebuffat, 2019), while full self-immunity to MccB17 requires both an immunity protein McbG and an ABC exporter McbEF (Collin and Maxwell, 2019).

The first obstacle to be overcome by an antimicrobial compound to reach its final target is the bacterial cell envelope (Collet et al., 2020). The extent of this barrier varies according to the target to be reached, the chemical structure of the antimicrobial compound and the bacterial species. For Gram-negative bacteria, antimicrobials have to pass first the outer membrane. Then, they can access the cytoplasmic membrane bilayer (inner membrane) and either insert inside or cross it for those antimicrobials having intracellular targets. Many antibiotics are hydrophilic compounds of low molecular mass and uptake across the outer membrane is ensured by passive diffusion using pores formed by specific β-barrel membrane proteins called porins. Porins are the most abundant proteins of the outer membrane in Gram-negative bacteria. They are classified as non-specific (general porins) and specific (selective porins), according to their threshold size and amino acids lining the aqueous channel (Choi and Lee, 2019). The transport varies according to the size, charge and hydrophilicity of the molecule. Recently, the dual function of the porin OmpF both as receptor and translocator for the pore-forming colicin N, has been elegantly demonstrated (Jansen et al., 2020). However, more hydrophobic or higher molecular mass compounds above the porin threshold require other strategies, among which hijacking receptors or transporters required for vital functions is a major one. Indeed, Gram-negative bacteriocins, colicins and microcins, widely parasitize such receptors to enter the periplasmic space, and particularly those involved in iron import. This receptor hijacking qualifies many microcins as “Trojan horse” compounds, as they mimic vital compounds that require being imported in cells, to penetrate sensitive bacteria (Duquesne et al., 2007a; Nolan and Walsh, 2008; Severinov and Nair, 2012).

Iron acquisition is an essential factor for microbial life. However, under aerobic conditions, free iron availability is limited by the very low solubility of ferric iron, and especially within a host, where iron is competed for by both the microbial community and the host (Wilson et al., 2016). To secure iron, bacteria have evolved to develop efficient Fe(III)-chelating agents (K_a ranging from 10²³ to 10⁵²), termed siderophores, to scavenge iron from their surrounding environment and import it. A study by Lewis et al. (2010) showed that siderophores are sufficient for allowing the culture of bacteria previously unculturable in laboratory conditions. Siderophores are non-ribosomally synthesized (Crosa and Walsh, 2002) and are important for enteropathogen survival (Hantke, 2003). Concomitantly, iron availability has been observed to regulate MccE492 gene expression (Marcoleta et al., 2013). The resulting Fe(III)-siderophore complex is then internalized by the producing strains via high affinity siderophore receptors anchored at the outer membrane, which are specifically involved in this function, but also ensure other strategic roles in microbial communities (Kramer et al., 2019). Siderophore receptors consist of a 22-stranded antiparallel β-barrel with external loops serving as ligand binding sites and an N-terminal globular domain forming a plug that occludes the barrel (Krewulak and Vogel, 2008). They are specific to the different siderophore chemical types, such as FhuA for ferrichrome or Cir, Fiu, and FepA for catechol siderophores in enterobacteria. These receptors are coupled to the TonB-ExbB-ExbD three-component machinery anchored at the inner membrane (TonB system), which transfers the energy source from the proton motive force of the cytoplasmic membrane to the outer membrane (Krewulak and Vogel, 2008), thus permitting active transport.

All microcins, whatever they are of class I or II, use either the siderophore receptor or the porin path to reach their final target (Figure 2B). Siderophore microcins uptake requires the FepA-, Cir-, Fiu-TonB systems, with FepA having the most important role (Destoumieux-Garzón et al., 2006; Azpiroz and Laviña, 2007; Vassiliadis et al., 2010). Unmodified microcins use the Cir-TonB system (MccV, MccL) (Chehade and Braun, 1988; Morin et al., 2011), or the porin OmpF which screens incoming products in a non-specific manner (Sato et al., 2000; Kaeriyama et al., 2006) (MccPDI) (Zhao et al., 2015), while class I microcins either use FhuA (MccJ25) (Pugsley et al., 1986; Salomón and Farías, 1993; Mathavan et al., 2014), or OmpF (MccB17, McC) (Laviña et al., 1986; Novikova et al., 2007) to reach the periplasmic space (Figure 2B). In the case of loss of function of the TonB system, MccE492, MccH47, and MccM retain antimicrobial activity, suggesting the involvement of another translocator, such as the TolA-TolQ-TolC system known to mediate the import of certain colicins (Lazdunski et al., 1998). Similar observations were made for MccL and MccV (Gerard et al., 2005; Morin et al., 2011), suggesting that the function of the ExbB protein could be replaced by its homolog TolQ in TonB-dependent microcin activity. However, although the presence of the siderophore PTM enhances its efficiency, the non-modified form of MccE492 (without the C-terminal siderophore) is also able to kill sensitive bacteria, but at a significantly lower level. On their side, antibiotics, which are essentially low molecular mass hydrophobic compounds, are most often transported inside target bacteria via porin or iron siderophore receptor pathways (Table 2).

MccH47 is bactericidal and targets the membrane bound F₀ proton channel subunits of ATP synthase (Trujillo et al., 2001; Rodriguez and Laviña, 2003; Palmer et al., 2020), causing an unregulated influx of protons. It uses FepA-, Cir-, Fiu-TonB dependent receptors to reach its inner membrane target (Patzer et al., 2003). The mechanism of action of the class IIa MccPDI is poorly identified. It was told to require close bacterial proximity to be cytotoxic, hence the name PDI (Proximity Dependent Inhibition) (Eberhart et al., 2012), since co-cultures of producing and sensitive strains separated by a semi-permeable film inhibit its activity. Why proximity is required for activity is unknown, but it could be only a consequence of a concentration-dependence effect (Lu et al., 2019). MccPDI that uses the porin OmpF to cross the outer membrane (Zhao et al., 2015; Lu et al., 2019) was shown (Zhao et al., 2015) to require a functional ATP synthase for exerting its cytotoxic activity, while (Lu et al., 2019) proposed it would induce membrane damage.

On one side, outer membrane porins and inner membrane transporters, which are involved in the uptake of antibiotics and microcins into sensitive cells, and on the other side efflux pumps, which pump the toxic compounds out of the bacteria, both constitute a first line resistance strategy (Ghai and Ghai, 2018). Porins, which ensure passive uptake of substrates across the outer membrane (see section mechanisms of action above), serve as the first gate for many antibiotics and several class I and II microcins. Furthermore, efflux pumps can be specific for a single substrate or can confer resistance to multiple antimicrobials by facilitating their extrusion before they can reach their intended targets (Anes et al., 2015). In Gram-negative bacteria, overexpression of efflux pumps is one of the mechanisms of resistance to β-lactams (Amaral et al., 2014) and to quinolones encoded by qepA and oqxAB genes (Fabrega et al., 2009). Likewise, reduced porin levels, which induce decrease of antibiotic concentration inside sensitive cells, is another mechanism of resistance to β-lactams in Gram-negative bacteria (Pfeifer et al., 2010), including K. pneumoniae (Jacoby et al., 2004) and P. aeruginosa (Li et al., 1994). Besides, mutations and deletions of genes encoding porins induce resistance to antibiotics. Indeed, ompF mutant was resistant to several β-lactam antibiotics in some Gram-negative pathogens, including E. coli and the deletion of OmpA resulted in increased susceptibility to several antibiotics including β-lactams in A. baumannii (Smani et al., 2014).

For microcins, the E. coli ABC exporter of unknown function Yojl, mediates resistance to MccJ25 by pumping the microcin out of the cells with the help of TolC, maintaining its concentration below the toxic concentration (Delgado et al., 2005). Yojl is located at the inner membrane and is coupled to the TolC protein at the outer membrane which ensures the last export step, similar to the MccJ25 gene cluster-encoded ABC exporter McjD, which warrants both microcin export and self-immunity for the producing cells (Bountra et al., 2017; Beis and Rebuffat, 2019). Similarly, McC is expelled from producing cells through a major facilitator superfamily (MFS) efflux pump (Severinov and Nair, 2012). Thus, the activation of several efflux pumps simultaneously could induce a co-resistance to antibiotics and microcins.

The iron-siderophore receptor FhuA is not only required for iron import, but it is also a target for bacteriocins (colicin M, MccJ25) and antibiotics (albomycin, rifamycin). Indeed, FhuA external loops L3, L4, L7, L8, and L11 are involved in the sensitivity to colicin M and the antibiotics albomycin and rifamycin. So far, a further mutation, insertion or deletion in the sequence encoding these loops may induce a cross-resistance between colicin M and these two antibiotics (Wang et al., 2018). Concomitantly, MccJ25 was also shown to require a primary interaction with the FhuA external loops L5, L7, L8 and L11 for its recognition and further internalization via this receptor (Destoumieux-Garzón et al., 2005). The level of sensitivity to MccJ25 also varies depending on the acquisition of specific FhuA, with a maximal sensitivity obtained with E. coli FhuA, while several Salmonella serovars are resistant due to a lack of efficiency of their FhuA receptor for MccJ25 uptake (Vincent et al., 2004; Ben Said et al., 2020). Similarly, various mutations in FhuA, especially in the cork domain, were reported to reduce the uptake and consequently the sensitivity to albomycin (Endriss et al., 2003). It could thus be hypothesized too that cross-resistance can occur between MccJ25 and albomycin. Besides, membrane permeabilization induced by a synthetic cationic peptide (KFF)₃K was shown to induce the sensitivity of MccJ25 resistant clinical isolates, thus making the microcin entry independent of FhuA and SbmA proteins (Pomares et al., 2010), and thus confirming that microcin uptake is the first source of resistance to MccJ25. Therefore, both uptake decrease of the toxic entity and pumping it out of the sensitive cells are efficient mechanisms to confer resistance to MccJ25.

Resistance to siderophore microcins which carry a catechol siderophore PTM is also primarily induced by uptake impairment (Thomas et al., 2004; Massip and Oswald, 2020). As seen before, MccE492, MccM and MccH47 are recognized and internalized in sensitive bacteria via the TonB-dependent FepA, Fiu and Cir iron-catecholate receptors. According to Thomas et al. (2004), a fepA, fiu double mutation, the triple cir, fiu, fepA mutation and the tonB mutation induce complete resistance to MccE492, MccM, and MccH47, while deletion of exbB and exbD does not affect the sensitivity to all three siderophore microcins (Vassiliadis et al., 2010). Although it does not carry a siderophore PTM, MccL requires the TonB dependent catecholate receptor Cir for uptake. Mutations/deletions in Cir and TonB, or suppression of the proton motive force, which is required for the TonB function, afford MccL resistance in E. coli and Salmonella, while the proteins involved in serine or sugar transport are not involved (Morin et al., 2011). On the other hand, a mutation in the energy transducer TonB was shown to reduce uptake and confer resistance to ceftazidime. Moreover, ceftazidime-resistant TonB mutants were shown to be cross-resistant to fluoroquinolones and lactivicin, a siderophore-conjugated non-β-lactam antibiotic (Calvopina et al., 2020). Thus, a high probability exists for a possible cross-resistance between these antibiotics and microcins.

Resistance to MccN/24 is afforded by mutations in genes encoding the outer membrane porin OmpF (Jeanteur et al., 1994), or the inner membrane transporter SdaC involved in serine uptake and used for MccV activity (Gerard et al., 2005). Resistance to MccPDI also involves OmpF and more precisely the K⁴⁷G⁴⁸N⁴⁹ amino acid motif found in the predicted outer loop L1 of the porin (Zhao et al., 2015; Lu et al., 2019). In addition, mutations in DsbA and DsbB proteins, presumably involved in the formation of disulfide bonds in OmpF, induce resistance to MccPDI (Zhao et al., 2015). Mutations in ompF and ompR genes encoding OmpF induce a reduced sensitivity to MccB17. Moreover, a mutation in the sbmA gene encoding the inner membrane transporter SbmA, which translocates MccB17 from the inner membrane to the cytoplasm, induces high resistance to MccB17 (Laviña et al., 1986).

As regard the efflux systems involved in resistance to microcins, resistance to MccN/24 is controlled by the multiple antibiotic resistance (mar) operon (Carlson et al., 2001), which modulates efflux pump and porin expression via two encoded transcription factors, MarR and MarA (Sharma et al., 2017). MarA plays an important role in antibiotic resistance by activating the expression of the acrAB-tolC encoded efflux pump (Zhang et al., 2008) and also regulates biofilm formation (Kettles et al., 2019). Resistance to MccN/24 in Salmonella cells appears concomitantly with a multiple antibiotic resistance phenotype to ciprofloxacin, tetracycline, chloramphenicol and rifampicin (Carlson et al., 2001). So far, cross-resistance between MccN/24 and antibiotics raised above is quite possible.

Additional mechanisms involve specific cell wall modifications. Those include surexpression of capsule polysaccharides that can increase resistance to various antimicrobials including both antibiotics, in particular polymixins, and antimicrobial peptides (Campos et al., 2004). Interestingly, capsule polysaccharides are not involved in MccJ25 resistance of the YojI deficient strain (Delgado et al., 2005). Alterations of the LPS resulting in truncated LPS structures promote, among other pleiotropic effects, resistance to antimicrobial peptides and hydrophobic antibiotics (Pagnout et al., 2019).

To allow DNA supercoiling, bacteria use two type II topoisomerases, DNA gyrase and topoisomerase IV, which are both the targets of quinolones. They form a ternary cleavage complex gyrase/DNA/quinolone, thus blocking DNA replication. Mutations in genes encoding DNA gyrase (gyrA, gyrB) and topoisomerase IV (parC, parE) lead to quinolone resistance. Besides, a plasmid-mediated protection of DNA gyrase and topoisomerase IV from the action of quinolones is ensured in a non-specific manner by the gyrase interacting protein Qnr. Qnr is a 218 amino acid pentapeptide repeat protein (PRP) encoded by qnr genes, which blocks the action of quinolones on the DNA gyrase and topoisomerase IV in a lesser extent (Fabrega et al., 2009; Jacoby et al., 2015). Indeed, one of these mutations is the well-known GyrB W⁷⁵¹R mutation which induces resistance to quinolones and is also linked to resistance to MccB17 (Vizán et al., 1991). GyrB Trp⁷⁵¹ is strongly implicated in the interaction of DNA gyrase with MccB17 (Heddle et al., 2001) and gyrB point mutation changing Trp⁷⁵¹ for Arg leads to a protein variant resistant to MccB17 (del Castillo et al., 2001). Additionally, partial resistance to MccB17 is provided by mutations at position 83 in GyrA or 447 in GyrB (Jacoby et al., 2015). Consequently, cross-resistance to MccB17 and quinolones could occur. Otherwise, it is well known that immunity genes are responsible for protecting the producing bacteria from their own bacteriocin. Indeed, three genes mcbE, mcbF, and mcbG are involved in cell protection from endogenous and exogenous MccB17. Interestingly strains harboring these genes are shown to be highly resistant to fluoroquinolones (Tran and Jacoby, 2002). These mechanisms seem to be responsible for co-resistance to MccB17 and quinolones.

Mutation of the gene rpoB encoding the β′ subunit of RNAP (see section mechanisms of action above) induces resistance to rifampicin (Campbell et al., 2001; Goldstein, 2014). Likewise, alterations in the 30S or 50S subunit of the ribosome lead to resistance to antibiotics that act on these proteins, mainly tetracycline, chloramphenicol, streptolydigin and aminoglycosides (Kapoor et al., 2017). Similarly, first studies performed to understand the mechanism of action of MccJ25 have shown that a point mutation causing a substitution of Thr⁹³¹ for Ile in the conserved segment of the rpoC gene coding for the largest RNAP subunit β′ conferred resistance to MccJ25, suggesting a mechanism involving occlusion of the RNAP secondary channel (Delgado et al., 2001; Yuzenkova et al., 2002). It was shown further from the crystal structure of the MccJ25-RNAP complex that MccJ25 binds within the RNAP secondary channel and interferes with the traffic of NTPs to the catalytic center (Braffman et al., 2019). Furthermore, additional rpoC mutations affecting amino acids in the conserved segments G, G′ and F and exposed into the RNAP secondary channel, also led to MccJ25 resistance in vivo and in vitro. While MccJ25 acts on the β′ subunit, and rifampicin on the β subunit, streptolydigin acts on both subunits. So far, a cross-resistance between MccJ25 and the above cited antibiotics mainly streptolydigin and rifampicin appears to be highly expected (Yang and Price, 1995; Temiakov et al., 2005).

For other antibiotics and microcins, no specific cross-resistance appears to be predictable. Chromosomally mediated colistin resistance occurs mainly via the addition of cationic moieties onto the negatively charged lipid A, while the plasmid mediated colistin resistance (MCR) is acquired via a plasmid-borne copy of an mcr gene. MCR-1 is the most prevalent MCR enzyme reported for the first time in 2015 followed by nine homologs described to date (Carroll et al., 2019). MCR-1-mediated colistin resistance confers protection against this last resort antibiotic via the presence of modified LPS within the cytoplasmic membrane, rather than the outer membrane (Sabnis et al., 2019). More precisely, the phosphoethanolamine transferase activity of MCR-1 adds a cationic phosphoethanolamine moiety to the anionic lipid domain A of LPS, which results in a net negative charge decrease and thus a lower affinity for the polymyxins.

Fosfomycin inhibits the bacterial cell wall synthesis at the early initiating step of the peptidoglycan synthesis. More specifically, it inhibits UDP-N-acetylglucosamine enolpyruvyl transferase (or MurA), the enzyme involved in transfer of the enolpyruvyl part of phosphoenolpyruvate to the 3′-hydroxyl group of UDP-N-acetylglucosamine, which is the first step in the biosynthesis pathway of peptidoglycan. Mutations in the murA gene confer resistance to fosfomycin due to the replacement of cysteine with aspartate in the active site of MurA, which prevents fosfomycin binding (Falagas et al., 2019). Moreover, resistance to fosfomycin can occur from chromosomal mutations in the structural genes that encode the GlpT and UhpT membrane transporters. GlpT and UhpT transport glycerol-3-phosphate and glycerol-6-phosphate sugars in bacteria, respectively and are used by fosfomycin to facilitate its entry in bacteria. These mutations block fosfomycin cell penetration (Falagas et al., 2019).

On the microcin side, the F₁F₀-ATP synthase has been shown to be the target of MccH47 (Rodriguez and Laviña, 2003) and MccPDI (Zhao et al., 2015). E. coli ATP synthase consists of a membrane-bound F₀ sector, which ensures proton translocation, connected to a cytoplasmic F₁ sector. They form a complex made up of eight different subunits, which are encoded by the atp operon, atpIBEFHAGDC. Three subunits form the F₀ proton channel and five subunits the catalytic F₁ domain. Mutations on genes atpB, atpE, atpF encoding the three subunits F₀a, F₀c, F₀b respectively, which constitute the F₀ proton channel, result in resistance to MccH47 (Rodriguez and Laviña, 2003). Furthermore, deletion of genes encoding subunits in the F₁ and F₀ domains of ATP synthase (atpA and atpF encoding F₁α and F₀b subunits, or atpE and atpH encoding F₀c and F₁δ subunits), result in a loss of susceptibility to MccPDI simultaneously to the loss of ATP synthase function (Zhao et al., 2015). None of these mechanisms appears to be shared between antibiotics and microcins.

Several Gram-negative bacteria produce different enzymes that are able to modify antibiotics and thus induce resistance, such as the very well-known β-lactamases, which disrupt the specific structure of β-lactams (Sawa et al., 2020). β-lactamases are classified into four classes including group 1 (class C) cephalosporinases, group 2 (classes A and D) broad-spectrum, inhibitor-resistant, and extended-spectrum β-lactamases as well as serine carbapenemases, and group 3 (class B) metallo-β-lactamases (Bush and Jacoby, 2010). Other enzymes including aminoglycosides modifying enzymes, such as phosphotransferases (APHs), nucleotidyltranferases (ANTs) and acetyltransferases (AACs), which phosphorylate, adenylate and acetylate these compounds, respectively could also be involved in development of resistance (Ramirez and Tolmasky, 2010).

Acetylation is a widespread and efficient mechanism of resistance against different antibiotics. Modification of the piperazine ring of the fluoroquinolones is induced by an acetylase AAC(6′)-Ib-cr, which provides one of the mechanisms of resistance of bacteria to quinolones (Fabrega et al., 2009). Chloramphenicol is also inactivated by acetylation which is performed by chloramphenicol acetyltransferases (CATs) (Smale, 2010). Acetylation is also a major mechanism of resistance to McC, then suggesting a high risk of cross-resistance between chloramphenicol and McC. Before its ultimate processing by non-specific aminopeptidases, which happens in sensitive cells to release the toxic non-hydrolyzable analog of aspartyl-adenylate, McC is exported outside the producer by the MccC pump and uptaken by sensitive cells using the porin OmpF and the inner membrane transporter YejABEF (see section Mechanisms of action). However, although most of produced McC is efficiently exported, intracellular processing also occurs inside the producing cells that ineluctably leads to the accumulation of the toxic entity that cannot be exported by the MccC pump and results in self-poisoning. Therefore, E. coli mcc gene clusters include genes (mccE and mccF) that encode proteins ensuring the self-immunity of the producer. The MccE acetyltransferase acetylates the α-amino group of processed McC, making it unable to bind to AspRS (Agarwal et al., 2011). So far, MccE makes E. coli simultaneously resistant to albomycin and McC (Novikova et al., 2010). MccE belongs to the general control non-repressible 5-related N-acetyltransferases (GNAT) superfamily, and shows high similarity with chromosomally encoded acetyltransferases RimI, RimJ, and RimL, which acetylate the N-termini of ribosomal proteins S18, S5, and L12 (Salah Ud-Din et al., 2016). Indeed, E. coli RimL induces resistance to McC by acetylating the amino group of the processed McC aspartate by the same mechanism as MccE (Kazakov et al., 2014). Similarly, when overproduced, RimL makes cells resistant to albomycin by acetylating processed albomycin, which contains a pyrimidine nucleotide instead of adenosine. Subsequently, a potential cross-resistance between McC and albomycin is quite possible (Kazakov et al., 2014). The MccF serine protease hydrolyses the carboxamide bond between the C-terminal aspartamide and AMP of both intact and processed McC, thus inactivating the aspartyl-adenylate (Agarwal et al., 2012). Moreover, McC inactivation is also ensured by phosphoramidases belonging to the histidine-triad (HIT) superfamily hydrolases that can either be encoded in certain mcc-like biosynthetic clusters or by genes located elsewhere in bacterial genomes (Yagmurov et al., 2020). Resistance to McC-like compounds produced by S. enterica, Nocardiopsis kunsanensis, P. fluorescens or Hyalangium minutum is conferred by hydrolysis of the phosphoramide bond in the toxic aspartamide-adenylate (Yagmurov et al., 2020).Therefore, it appears that resistance to McC and McC-like microcins by toxin inactivation can occur via both enzymes encoded in the microcin biosynthesis clusters and more generalist and non-specific enzymes sharing structural similarities.

Finally, impairment of the final three-dimensional structure of the antibacterial peptide, such as by preventing the formation of disulfide bridges, could be a last mechanism resulting in resistance to microcins. This has been poorly explored until now, but is however illustrated by MccPDI for which mutations in dsbA, dsbB genes induce resistance to MccPDI (Zhao et al., 2015). Genes dsbA, dsbB encode DsbA and DsbB thiol-redox enzymes that usually catalyze disulfide bond formation for proteins that are transported into the periplasm, and which would be possibly involved in formation of the disulfide bond that stabilizes this microcin.