The extracellular loop of Man-PTS subunit IID is responsible for the sensitivity of Lactococcus garvieae to garvicins A, B and C

Mannose phosphotransferase system (Man-PTS) serves as a receptor for several bacteriocins in sensitive bacterial cells, namely subclass IIa bacteriocins (pediocin-like; pediocins) and subclass IId ones - lactococcin A (LcnA), lactococcin B (LcnB) and garvicin Q (GarQ). Here, to identify the receptor for three other narrow-spectrum subclass IId bacteriocins - garvicins A, B and C (GarA-C) Lactococcus garvieae mutants resistant to bacteriocins were generated and sequenced to look for mutations responsible for resistance. Spontaneous mutants had their whole genome sequenced while in mutants obtained by integration of pGhost9::ISS1 regions flanking the integration site were sequenced. For both types of mutants mutations were found in genes encoding Man-PTS components IIC and IID indicating that Man-PTS likely serves as the receptor for these bacteriocins as well. This was subsequently confirmed by deletion of the man-PTS operon in the bacteriocin-sensitive L. garvieae IBB3403, which resulted in resistant cells, and by heterologous expression of appropriate man-PTS genes in the resistant Lactococcus lactis strains, which resulted in sensitive cells. GarA, GarB, GarC and other Man-PTS-targeting bacteriocins differ in the amino acid sequence and activity spectrum, suggesting that they interact with the receptor through distinct binding patterns. Comparative analyses and genetic studies identified a previously unrecognized extracellular loop of Man-PTS subunit IID (γ+) implicated in the L. garvieae sensitivity to the bacteriocins studied here. Additionally, individual amino acids localized mostly in the sugar channel-forming transmembrane parts of subunit IIC or in the extracellular parts of IID likely involved in the interaction with each bacteriocin were specified. Finally, template-based 3D models of Man-PTS subunits IIC and IID were built to allow a deeper insight into the Man-PTS structure and functioning.

Mannose phosphotransferase system (Man-PTS) is a major phosphoenolpyruvate (PEP)-dependent sugar transporting system for mannose uptake and its concurrent phosphorylation in Firmicutes and Gammaproteobacteria 1 . It has a fairly broad substrate specificity as, besides mannose, it can also transport glucose, fructose, glucosamine, N-acetylglucosamine and galactosamine 2 . It is a multi-component system, composed of general PTS proteins, enzyme I and HPr (phosphoryl group donors for different PTS permeases), and enzyme II, which is a permease specific for the carbohydrates listed above. Enzyme II consists of the cytoplasmatic subunits IIA and IIB and the membrane subunits IIC and IID. The intracellular components transfer the phosphoryl group from PEP to the incoming sugar substrate while the membrane components form a sugar-specific binding site and translocation channel 1 . It has been proposed that in addition to its basic transport-related function, Man-PTS can also regulate a variety of intracellular processes, including metabolism, mutagenesis and gene expression 3 . Moreover, Man-PTS is also a target for several antimicrobial agents, such as bacteriocins 4 , and bacteriophage lambda 5 . Importantly, Man-PTS is considered as convenient drug target as it is absent in eukaryotic cells 6 .
garvieae IBB3403 cells by electroporation and chromosomal DNA was randomly mutagenized by its integration as described by Maguin et al. 32 , with minor modifications including the use of BHI medium, 1000-fold dilution and incubation at 30 °C for 150 min and then 37 °C for 150 min. Aliquots of 1, 2 and 5 ml of the mutagenized L. garvieae IBB3404:pGh9::ISS1 culture were centrifuged, resuspended in 100 μl of BHI medium and plated onto BHI-agar plates with Ery (5 µg/ml) and GarA (7 µg/ml) preheated to 37 °C and grown for 3 days at 37 °C. Mutant colonies were collected and their level of sensitivity to GarA was estimated using microtiter plates with two-fold bacteriocin dilutions. Next, the DNA rescue cloning procedure was applied as described previously 32 . Briefly, total DNA isolated from the erythromycin-and GarA-resistant pGhost9::ISS1 integration mutants was digested with EcoRI or HindIII (Thermo Fisher Scientific, Waltham, MA, USA), self-ligated with T4 DNA ligase (Thermo Fisher Scientific, Waltham, MA, USA) and used to transform an E. coli repA + strain (EC1000) since pGhost::ISS1 due to its temperature-sensitive replicon (repA − ) is unable to replicate at 37 °C. Clones containing pGhost9::ISS1 with flanking chromosomal DNA fragments (left -EcoRI or right -HindIII) were selected on LB plates containing Ery. Rescued fragments were sequenced directly from the pGhost9::ISS1 plasmid with pairs of primers pIS-S1Eco/pGh9 or pISS1Hind/uni (Supplementary Table S1). The sequences obtained were analysed and compared with the L. garvieae genome sequence using BLAST network service (NCBI; https://blast.ncbi.nlm.nih.gov) and standard parameters.
Construction of manABCD deletion mutants. The mutants were created by a double crossover between pGhost9 plasmids harboring DNA fragments flanking the manABCD operon or its selected genes and the chromosomal region containing these DNA fragments. DNA fragments flanking the entire manABCD operon, manCD, manC or manD genes were created by amplification with primers pairs manABCDUPfor/rev and manABCDDNfor/rev, manCDUPfor/rev and manABCDDNfor/rev, manCDUPfor/rev and manCDNfor/ rev or manDUPfor/rev and manABCDDNfor/rev, respectively (Supplementary Table S1). To each UPrev and DNfor primer the EcoRI restriction site was added. After amplification, PCR products were purified, digested with EcoRI and ligated with T4 DNA ligase, which resulted in DNA fragments containing joined upstream and downstream DNA regions of the manABCD gene(s) to be deleted. An additional PCR was performed using suitable UPfor and DNrev primers and the amplified product was ligated with pGEM-T Easy vector (Promega, Fitchburg, WI, USA) by TA cloning and then cloned into pGhost9 using the ApaI and NotI (Thermo Fisher Scientific, Waltham, MA, USA). Overnight L. garvieae IBB3403 cultures harboring the above pGhost9 derivatives diluted 10 3 -fold in BHI medium with Ery (5 µg/ml). Homologous recombination was enforced by incubation at 30 °C for 1.5 h and at 37 °C for 2.5 h. Integrants were selected at 37 °C on BHI-agar plates containing Ery (5 µg/ ml). Excision from the chromosome and removing of the integration vector from the L. garvieae IBB3403 strains was performed by culturing the integrants in the absence of antibiotic for at least 100 generations at 30 °C. The genetic structure of the resulting deletion strains was confirmed by colony PCR with manABCDfor/rev primers (Supplementary Table S1), sequencing of the DNA region containing the deleted genes and determination of sensitivity to erythromycin.

Construction of Man-PTS complementing plasmids.
In order to complement the deletion of the manABCD operon or its selected genes, two-plasmid nisin-controlled gene expression system (NICE) with pNZ9530 and pNZ8037 plasmids was used 33,34 . Amplification of the entire manABCD operon, manCD, manC or manD genes was performed using primer pairs complmanfor and complmanrev, complCDfor and complmanrev, complCDfor and complCrev or complDfor and complmanrev, respectively (Supplementary constructs were expressed in L. garvieae 548a, L. lactis B464 and/or L. lactis NZ9000. Additional complementing plasmid was created by amplification of the pNZ8037 plasmid harboring manCD genes with γ+ for/rev primers (Supplementary Table S1). Obtained construct was expressed in L. lactis B464.

Results
GarA, GarB and GarC have a narrow activity spectrum. GarA activity has been examined earlier against strains from the genera Bordetella, Carnobacterium, Enterococcus, Lactobacillus, Lactococcus, Listeria, Pediococcus, Salmonella, Staphylococcus and Streptococcus and found to be limited to L. garvieae 15 . In this study we used an expanded set of strains to determine the activity spectra of GarB and GarC and reexamine it for GarA. In addition to the above, we included several strains from the genera Bacillus, Campylobacter, Candida, Leuconostoc and Pseudomonas (Supplementary Table S2). GarA and GarB showed a narrow spectrum of antimicrobial activity, being highly active only against L. garvieae strains; at a high concentration (1 mg/ml), GarA exhibited minimal activity against L. lactis strains and some strains from the genera Carnobacterium, Enterococcus, Lactobacillus, Leuconostoc and Pediococcus (Supplementary Table S2). GarC exhibited a slightly wider spectrum, being potent against all Lactococcus spp. (L. garvieae, L. lactis and L. raffinolactis) strains and, to a lesser extent, also against some strains from the genera Lactobacillus and Leuconostoc (Supplementary Table S2).  (Table 1). Notably, none of the resistant mutants had mutations in the manAB gene ( Table 1), suggesting that the encoded polypeptide is not directly involved in the sensitivity to the bacteriocins. The second type of mutants were obtained by random integration of the pGhost9::ISS1 32 . To identify the mutated sites in the resulting GarA-resistant mutants, we sequenced regions flanking the integration site of the plasmid. Three integration mutants were obtained (L. garvieae MS1-MS3; Supplementary Table S1) exhibiting MIC GarA values over 1024-fold higher than the wild-type L. garvieae IBB3403 (0.024 µg/ml). The mutants showed full resistance to GarB and GarC, indicating cross-resistance between these bacteriocins. The DNA regions flanking the plasmid integration site in these mutants were successfully cloned, sequenced, and the obtained sequences were compared with the wild-type L. garvieae IBB3403 genome. Also in these strains, the pGhost9:ISS1 insertions occurred at two locations in the operon encoding the Man-PTS system, one between positions 714-715 in the manC gene (L. garvieae MS1) and one between positions 449-450 in the manD gene (L. garvieae MS2 and MS3; Table 1; Supplementary Table S1).

Mutants resistant to
Man-PTS subunits IIC and IID are necessary for GarA-C activity. In order to examine whether the resistance to GarA, GarB and GarC was directly related to the man operon and not to any additional mutations, we deleted the operon in the sensitive L. garvieae IBB3403. Additionally, to evaluate the role of individual Man-PTS membrane subunits in the sensitivity to bacteriocins, we deleted the manC and/or manD genes. All these deletion mutants (L. garvieae B548a, B549a, B550a and B551a; Table S1 in the supplementary file) had MIC GarA values over 1024-fold higher than the wild-type strain, and, all exhibited full resistance to GarB and GarC (Fig. 1). As GarA and GarC but not GarB exhibited also activity against L. lactis strains, we performed additional studies considering this species only with two first bacteriocins. Tested L. lactis B464 strain, which has the mannose-specific PTS operon (ptnABCD) deleted 4 , was insensitive to GarC but not to GarA (Fig. 1), indicating that in this species the Man-PTS system is required for sensitivity to GarC only.
To assess the individual role of Man-PTS subunits in mediating bacteriocin sensitivity, we introduced respective genes in different combinations into the manABCD-deleted L. garvieae B548a strain. Since high expression of genes encoding membrane proteins is often toxic to the host cell, we applied a two-plasmid nisin-controlled gene expression (NICE) system allowing for strict control of protein production 33,34 . We used pNZ8037 plasmid with a nisin-responsive promoter for cloning and expressed it in manABCD-deleted L. garvieae B552a with the pNZ950 plasmid carrying nisK and nisR genes encoding membrane-located histidine kinase NisK and intracellular response regulator NisR, respectively. Unfortunately, even using this system we were unable to express the cloned genes due to plasmid instability. To overcome this problem, we employed the L. lactis B464 strain with ptnABCD deletion, which had been shown to support expression of man-PTS genes from the NICE system 4 . Introduction of manC or manD genes individually into L. lactis B464 (to give respectively L. lactis B559a or B560a; Table S1 in the supplementary file) had no impact on the sensitivity of L. lactis B464 to GarA, GarB and GarC. Only introduction of the entire manABCD operon or the manCD genes together (respectively L. lactis B557a and B558a; Table S1 in the supplementary file) produced clones sensitive to the bacteriocins (Fig. 1). Altogether, the results showed unequivocally that both the IIC and IID subunits of L. garvieae Man-PTS are required and sufficient for sensitivity to GarA-C. To assess the relatedness of GarA, GarB and GarC to those bacteriocins, we compared their amino acid sequences. Neither the leader nor the mature peptides of GarA, GarB and GarC were similar to those of subclass IIa bacteriocins (pediocins) 38 ( Fig. 2A). In contrast, the leader peptides of GarA-C were highly similar to each other and also to the leader peptides of other subclass IId bacteriocins (LcnA, LcnB and GarQ) on their entire length. In the mature peptides similarity was very weak among GarA-C and virtually absent with LcnA, LcnB and GarQ (Fig. 2B), except for 13 N-terminal amino acids 92% identical between GarA and GarB. The predicted secondary structures of GarA-C comprised one long (17 aa; GarA) or two shorter (GarB and GarC) α-helices at the N-terminus and unstructured C-terminal parts (Fig. 3). However, a comparison of their template-based 3D structure models revealed little overall similarity (Fig. 3). Moreover, GarA was predicted to interact with the membrane through its long α-helix parallel to the membrane surface, while the interaction of GarB and GarC was predicted to occur through the unstructured fragments (Fig. 3). These LGA6, LGA13 GTT insertion at 570-571 in manC Val insertion at 190-191 8 32 results suggest that initial electrostatic interaction between cationic bacteriocin and negatively charged bacterial cell membrane may differ between GarA-C. This interaction between cell membrane and N-terminal α-helix of GarA may be also responsible for the observed bacteriocin minimal activity against L. lactis strains and some Carnobacterium, Enterococcus, Lactobacillus, Leuconostoc and Pediococcus strains.
In order to get a deeper insight into the phylogeny of the garvicins and thus their possible mechanism of action, we performed a wide homology search using the NCBI BLAST network service. BLASTp showed that the N-terminal 13 amino acids of GarA and GarB are respectively 85% and 77% identical to a region of TonB-dependent receptor (accession no. WP_091746097.1). The similarity region is a part of the ligand-gated channel and is localized adjacent to the conserved amino acid residues responsible for the ligand binding (Fig. 2C). The same N-terminal 12 amino acids of GarA and GarB are also respectively 83% and 92% identical to a region of the voltage-gated chloride channel protein (accession no. WP_058617448.1). This region is a part of the transmembrane α-helix adjacent to the conserved amino acid residues responsible for chloride ion binding and taking part in the formation of the membrane channel 39,40 (Fig. 2D). Similarity between garvicins and intermembrane fragments of channel forming proteins suggests that studies bacteriocins may act by forming a pore that leads to uncontrolled leakage of intracellular solutes and consequent cell death.
GarA-C target a specific extracellular loop of subunit IID. In order to identify the Man-PTSs features responsible for sensitivity to GarA, GarB and GarC, we compared the amino acid sequences of IIC and IID form sensitive and resistant species. The IIC subunits from L. garvieae strains showed high similarity to those from the other species, especially L. lactis (data not shown). Also the IID subunits showed high conservation, but with notable exception of a stretch of 51 amino acids present only in L. garvieae (Fig. 4). As the additional sequence lies in the extracellular region γ 13 , we propose to call the part of extended region γ, which contains the additional 51 amino acids, as region γ+. A homology search showed that region γ+ is unique to L. garvieae strains. To test whether it may indeed be responsible for the sensitivity of L. garvieae strains to garvicins A, B and C, we deleted in frame an internal part of manD encoding the additional 51 amino acids (γ+) from the garvicin-sensitive L. lactis strain B558a, carrying manCD in trans and with ptnABCD deleted, which gave in L. lactis B561a (supplementary Table S1). Deletion of γ+ conferred partial resistance to GarA and full resistance to GarB and GarC (Fig. 1), without affecting the mannose transport function. This indicates that region γ+ of subunit IID is indispensable for sensitivity to GarB and GarC, and, to a lesser extent, also GarA.
Hybrid Man-PTS system can serve as GarA receptor. As the protein sequences of the Man-PTS IIC subunits of L. garvieae and L. lactis were highly similar, we tested whether they can substitute each other to confer sensitivity to GarA and GarB. Therefore, we expressed L. garvieae genes manC or manD in the GarA-and GarB-resistant L. lactis NZ9000, which harbors the entire ptn operon and nisRK genes on its chromosome. L. lactis carrying manC (B563a) remained resistant to GarA and GarB, while L. lactis carrying manD (B564a) became sensitive to GarA but remained resistant to GarB ( Fig. 1; Table S1 in the supplementary file). This indicates that a hybrid Man-PTS system consisting of L. lactis IIC and L. garvieae IID subunits can serve as a GarA receptor and that the site required for the antimicrobial activity of GarA is localized in the L. garvieae IID subunit. On the other hand, both IIC and IID subunits from L. garvieae are required for GarB activity.

GarA-C cause mutations at distinct sites of Man-PTS. The results presented above suggested that
GarA, GarB and GarC may interact with Man-PTS by using different binding patterns. To pinpoint specific amino acids on IICD necessary for the Man-PTS interaction with individual garvicins, we determined the sensitivity to GarA, GarB and GarC of GarQ-resistant and man + (mannose-positive phenotype; Table 1) L. garvieae IBB3403 missense mutants harboring single amino acid substitutions in the transmembrane domains of IIC (PW202 and LGN9) and transmembrane or extracellular domains of IID (PW203, PW204, LGN1, LGN2, LGN4) obtained in a previous study 12 (Supplementary Table S1). In that study we established that spontaneous mutants with nonsense or frameshift mutations in the manABCD operon are unable to ferment mannose (man − phenotype; Table 1) due to premature termination or a change of Man-PTS structure depriving it of its sugar-transporting function. On the other hand, maintained ability to utilize mannose (man + phenotype) by missense mutants indicates that the mutations, although leading to GarQ resistance, had no negative effect on the transmembrane structure of Man-PTS and that the amino acid(s) changed were likely involved in the GarQ -receptor interaction. The mutations turned out to affect also the sensitivity to other garvicins. Crucially for the present study, the effects were different for different garvicins. Thus, the PW202 and LGN1 mutants had, respectively, a 4-fold and 8-fold higher MIC value towards GarA than the wild-type strain, while the other mutants remained fully sensitive to GarA (Table 1). In the case of GarB, the MIC was at least 64-fold higher for PW203, PW204 and LGN1, and only 4-fold higher for LGN2 and LGN4, than that of the wild type (0.78 µg/ml). PW202 remained fully sensitive to GarB (Table 1). For GarC the MIC values were 32-fold for LGN1, 16-fold for LGN2, and 8-fold higher than the wild type (0.39 µg/ml) for PW202, LGN4 and LGN9. The other mutants remained fully sensitive to GarC (Table 1). These results convincingly show that distinct Man-PTS mutations differently affect its interaction with individual garvicins.
To identify other Man-PTS regions or specific amino acids targeted by GarA, GarB or GarC we obtained additional spontaneous bacteriocin-resistant mutants with a preserved Man-PTS functionality (man + phenotype; Table 1), using the described method 12 . Wild-type L. garvieae IBB3403 was cultivated on CDM-agar containing mannose as the sole carbon source supplemented with GarA, GarB or GarC. Eight GarA-resistant man + mutants were obtained (LGA2, LGA3, LGA5; LGA6, LGA8, LGA10, LGA13 and LGA14; Table S1 in (Table 1; Fig. 5). All these GarA-resistant mutants exhibited also between 4-fold and 64-fold higher resistance to GarB and GarC, and between 2-fold and 128-fold higher resistance to GarQ (Table 1; Fig. 5). Ten GarB-resistant mutants were obtained (LGB1-LGB10; Table S1 in the supplementary file) containing only two mutations. Deletion of AsnValValGly from subunit IID, the same as that in the GarA-selected mutant (LGA8) clearly represented a hot spot as it was found in nine GarB-resistant mutants (LGB1-LGB5, LGB7-LGB10). This deletion decreased the sensitivity to GarB 64-fold compared with the wild type, 8-fold to GarC, 4-fold to GarQ,   and 2-fold to GarA (same as for the LGA8 mutant). The remaining mutant (LGB6) also had a mutation in manD resulting in the Gln313 → Lys substitution in the extracellular loop of IID (Fig. 5). It was over 64-fold less sensitive to GarB and 2-fold to GarQ, and was fully sensitive to GarA (Table 1; Fig. 5).
Only two mutants had mutations in the fourth and sixth transmembrane domain of subunit IIC: Ala106 → Val (LGC13) and Gly188 → Val (LGC18), respectively (Table 1; Fig. 5). Most of the GarC-resistant mutants were also resistant, to varying degrees, to the other garvicins tested, especially to GarB, for which thirteen strains had the MIC 64-fold higher than the wild type (Table 1). Altogether, 33 independent resistant mutants were obtained, carrying 13 different mutations (Table 1; Fig. 5). Several features of interest could be observed. The mutants exhibited varying degrees of resistance (between 2-fold and 128-fold higher relative to wild-type) to the selected garvicins and also varying degrees of cross-resistance. Mutations in manD gene were significantly more frequent than in manC gene, extracellular loop of IID being the most common region affected. Deletion of amino acids 262AsnValValGly265 in the γ+ region was particularly frequent, as it occurred independently ten times. These features clearly confirm our earlier conclusion that γ+ region of subunit IID is a major site of interaction with GarA-C and different regions of subunits IIC and IID are critical for the interactions with different garvicins.
Amino acids involved in the sensitivity to GarA-C are in the channel of IIC and in the extracellular parts of IID. To determine the location of the amino acids altered by the resistance-conferring mutations in the folded IIC and IID subunits we built their template-based 3D models and compared them with the 2D topographic predictions. The two structure predictions coincided with each other fairly well (Figs 5 and 6). In IIC differences were minor and concerned the localization of the N-terminal fragment and the presence of an additional C-terminal transmembrane region. Importantly, the central part of the protein where all the mutations were found had a similar location in the two models (Figs 5 and 6A). Both models predict a transmembrane location of IIC, and the 3D top view exposes a channel for the transport of sugars. The channel is formed by six α-helices and most of the amino acid residues substituted in the garvicin-resistant mutants potentially engaged in garvicin binding (Pro100, Ala102, Ala106 and Gly188) are localized within it (Fig. 6A).
In contrast to IIC, the 3D model of subunit IID does not confirm its transmembrane localization predicted by the topological model but instead indicates that is a monotopic membrane protein only anchored in the outside part of lipid bilayer, that suggests that it can be a first target for bacteriocin binding. Nearly all IID is exposed to the milieu except for amino acids 171-178 and 320-365 (C-terminus), which in the 2D model are part of an intracellular domain and form the third and fourth transmembrane helices, respectively (Figs 5 and 6B). All the amino acids altered by mutations in IID are in its extracellular part (Fig. 6B).

Discussion
Some L. garvieae strains are pathogenic and cause lactococcosis in marine or freshwater fish 41 , mastitis in cows 42 , and pneumonia in pigs 43 . Recently, L. garvieae has also been considered to be a human pathogen as it was isolated from patients with diverticulitis, peritonitis 44 , endocarditis 45 , spondylodiscitis 46 , and acute acalculous cholecystitis 47 . In this study we focused on three bacteriocins, GarA, GarB and GarC, encoded by plasmids from L. garvieae 21881 isolated from the blood of a patient suffering from septicemia 48 . We confirmed GarA activity against L. garvieae 15 and revealed for the first time GarB activity exclusively against L. garvieae and GarC -against Lactococcus spp. All three bacteriocins showed no significant activity against any other lactic acid bacteria tested or diverse pathogenic species, which makes them useful for selective treatment of infections caused by pathogenic L. garvieae strains. At present, the most commonly used method for the identification of bacteriocin receptors or other proteins involved in their action, is genome sequencing. It requires generation of spontaneous resistant mutants followed by whole genome sequencing and identification of mutated genes responsible for the reduced sensitivity 12,[35][36][37] . In this study, we also applied an integrative plasmid pGhost9::ISS1 32 to generate bacteriocin-resistant mutants. We have previously used this approach to identify the gene ccpA 49 and several other genes involved in cellobiose and lactose metabolism 50 in L. lactis IL1403. However, this approach has not been applied to date to search for genes involved in the sensitivity to bacteriocins (e.g., potential receptor genes). Sequencing of the DNA regions surrounding the pGhost9::ISS1 integration sites in the genomes of resistant L. garvieae IBB3403 mutants revealed genes encoding Man-PTS, indicating that it likely is the receptor of the bacteriocins. We confirmed the involvement of these genes by using genome sequencing of spontaneous garvicin-resistant mutants (Table 1). Of the three genes encoding Man-PTS subunits, mutations were found exclusively in manC or manD and not in manAB, indicating that the latter is not required for sensitivity to GarA, GarB or GarC. Those results also showed that pGhost9::ISS1 integration can be used as a cheaper alternative to whole genome sequencing in searching for bacteriocin receptors. Deletion of individual man genes in L. garvieae IBB3403 and subsequent complementation studies confirmed an involvement of Man-PTS subunits IIC and IID in the sensitivity to GarA, GarB and GarC. Similarly, deletion of ptnABCD in L. lactis IL1403 confirmed its involvement in the sensitivity to GarC. An attempt to complement the manABCD deletion using nisin-inducible pNZ9530 and pNZ8037 plasmids was unsuccessful probably due to incompatibility between them and the native L. garvieae IBB3403 plasmids. We therefore decided to use the garvicin-resistant, plasmid-free ∆ptnABCD L. lactis B464 4 as the host for complementation studies. Introduction of the entire manABCD operon resulted in a strain sensitivity to the bacteriocins studied. In parallel, introduction of separately cloned manC and manD coding for IIC and IID confirmed that both these Man-PTS transmembrane subunits are indispensable for GarA, GarB and GarC activity (Fig. 1).
It has been demonstrated earlier that all subclass IIa bacteriocins and subclass IId LcnA, LcnB and GarQ use Man-PTS as the receptor on target cells 4,12 . Although GarA, GarB and GarC target the same receptor as the other Man-PTS-targeting bacteriocins, a comparison of their activity spectra, amino acid sequences ( Fig. 2A,B) and predicted secondary and tertiary structures (Fig. 3) suggested that various bacteriocins may bind differently to their targets. A comparative analysis of the amino acid sequences of subunits IIC and IID from several garvicin-sensitive and garvicin-resistant species revealed high conservation of subunit IIC and the presence of an additional region γ+ unique to IID of L. garvieae (Fig. 4) likely responsible for the GarA and GarB activity, which is also limited to this species. Moreover, prediction of the IID transmembrane structure showed its external localization. Deletion of the γ+ region resulted in full resistance to GarB and partial resistance to GarA suggesting its key role in the interaction with GarB and a lesser significance in GarA binding. Interestingly, the deletion of the γ+ region also resulted in full resistance to GarC, although, besides L. garvieae, it is also active against L. lactis, which lacks region γ+. This suggests that GarC uses different interaction patterns in different species or that the γ+ deletion resulted in a Man-PTS structure change that prevented its binding.
The results discussed above indicated differences in the binding patterns of GarA, GarB and GarC which likely reflected specific interactions between a given bacteriocin and individual amino acids in Man-PTS IICD. Further confirmation of this notion comes from the fact that IIC and IID from L. garvieae only could serve as a receptor for GarB, whereas GarA recognized the IIC subunit from both L. lactis and L. garvieae (Fig. 1). Testing of the GarA, GarB and GarC activity against a panel of L. garvieae IBB3403 IIC or IID mutants selected as partially resistant to one of the three garvicins (Table 1) allowed the identification of 18 amino acids, six in IIC and twelve in IID, likely involved in specific interactions with the bacteriocins (Fig. 5). Most of those in IIC localized to the sugar-transporting channel and all in IID were extracellular (Fig. 6). The distribution of these amino acids in Man-PTS subunits suggests their diverse functions, probably some directly in the bacteriocin binding and others participating indirectly. The surface location of IID predisposes it to the role of a docking receptor, in which the protruding region γ+ and N-terminal part would serve for the initial contact of a bacteriocin with Man-PTS. The bacteriocin binding could induce IID conformational changes forcing its IIC partner to open the channel. Then, the interior of the open channel of IIC would offer a secondary binding site for the bacteriocin, which would cause a lethal fully open channel structure eventually leading to the cell death. This notion is further supported by the finding that the highly conserved N-terminal parts of garvicins A and B are significantly similar to the transmembrane segments of some bacterial transporters such as TonB and chloride channels (Fig. 2C,D). These proteins take part in the uptake of iron and nickel complexes, vitamin B12 and carbohydrates, and serve as colicin, microcin and bacteriophage receptors (TonB-dependent transporter) or allow passive diffusion of Cl − , the Cl − /H + exchange, water and salt transport, pH and cell volume regulation, and stabilization of the membrane potential (voltage-gated chloride channel protein) 39,40 . It is feasible that the N-terminal fragments of GarA and GarB could act in a similar manner by docking into the IIC channel to form a stably open pore increasing the membrane permeability. A similar sequence of bacteriocin binding has already been suggested for GarQ 12 , but here the picture becomes more detailed with the support of the modeled structures of IID as the surface docking site and IIC as the entry channel.
The Man-PTS amino acids altered in the mutants with reduced sensitivity to garvicins are apparently of different levels of importance. Some of them were essential for one or two garvicins and caused high resistance levels when mutated, while others had a lesser effect, i.e., they produced low resistance levels when mutated. This indicates a possibility of diverse binding strength between receptor-bacteriocin particular amino acids: an indispensable role in garvicins binding of amino acids, which substitutions led to high resistance levels (from 32x for GarA to 1024x for GarQ) and supporting role of amino acids that substitutions resulted in low resistance of mutants obtained (e.g., only 2-8-fold higher resistance of mutants in comparison with wild-type strain). These observations provide a further clue indicating that individual garvicins use different binding patterns. Nevertheless, one cannot at present exclude the possibility that in fact the altered amino acids are not directly involved in bacteriocin binding but instead are important for the Man-PTS structure, changes of which prevent bacteriocin binding without compromising the mannose transport function (man + phenotype of the mutants).
So far, no tertiary structure of a Man-PTS subunit has been reported, which made it difficult to interpret the obtained results in light of the bacteriocin -Man-PTS interactions. We therefore resorted to homology modeling to predict the structures of the two subunits relevant to the present study. Importantly, the overall distribution of transmembrane and extracellular regions corresponded well with the predicted membrane topology of these proteins (Figs 5 and 6). The models confirmed the transmembrane localization of IIC and predicted an unexpected cell-surface localization of subunit IID (Fig. 6). This subunit was earlier proposed to be a transmembrane protein forming a heterotetramer or higher multimer 6 . A cell-surface exposed IID subunit, as predicted here, would greatly facilitate bacteriocin binding via some of the following amino acids: Met34, Ala58, Pro111, Thr123, Ala133, Arg203, Tyr226, Trp256, AsnValValGly262-265, Gly301, Gln313 and Val315, all having an extracellular location and found to be important for bacteriocin sensitivity. According to the present study, GarA at high concentrations exhibited low activity against L. lactis and some strains from the genera Carnobacterium, Enterococcus, Lactobacillus, Leuconostoc and Pediococcus. Also GarC was minimally active against some strains from the genera Lactobacillus and Leuconostoc. Importantly, a minimal GarA activity was also observed after mutation or deletion of the Man-PTS-encoding genes in both L. garvieae and L. lactis (Fig. 1) suggesting that it is not dependent on Man-PTS. As such, it was not the subject of this study. It is possible that at high concentration, after initial unspecific electrostatic interaction between GarA N-terminal α-helix and membrane surface, bacteriocin aggregates and its α-helix penetrates the membrane and form pores without specific receptor. Similarly, it has already been shown that the cyclic bacteriocin garvicin ML exhibits non-specific antimicrobial activity at high concentrations, while at lower concentrations it requires a specific membrane receptor for action 51 . Thus, further studies with different GarA and GarC concentrations are required to evaluate their mode of action at high concentrations [52][53][54][55][56][57][58][59][60][61][62][63][64][65][66][67][68][69] .
In summary, this study provides a proof-of-principle for a convenient alternative to whole genome sequencing for the identification of genes involved in bacterial sensitivity to bacteriocins. Results obtained using both these methods indicate that in L. garviaeae Man-PTS subunits IIC and IID act as receptors for GarA, GarB and GarC, three distinct bacteriocins encoded by plasmids of L. garviaeae 21881 isolated from a clinical case of septicemia. We suggest here that their bactericidal effect, directed mainly against L. garvieae, relies on garvicin binding to specific amino acids of IICD result in IIC channel opening and cell death. The amino acid residues critical for the garvicin action were identified by analyzing resistant mutants, and their location in the extracellular regions of the surface protein IID and in the transmembrane channel of the IIC permease was determined with the help of predicted topology and 3D structure of the two Man-PTS subunits. Hundreds of bacteriocins have been identified so far with different antimicrobial spectra, which is commonly believed to indicate that they recognize distinct receptors on target cells. Here, however, we provide evidence that the different structures and inhibition spectra of bacteriocins do not necessarily mean that they recognize different receptors. We show that a single receptor, the mannose-specific PTS, can serve as a target for a number of non-homologous bacteriocins with greatly different activity spectra. As we expect that bacteriocins binding Man-PTS constitute a much larger family than the four investigated here, our future studies will focus on a search for other bacteriocins targeting the membrane subunits of Man-PTS. Their detailed investigation will allow us to build a full picture of the bacteriocin -Man-PTS interactions and could eventually justify proposing a separate group of bacteriocins besides the currently recognized IIa-IId.