Exoproteome Analysis of Antagonistic Interactions between the Probiotic Bacteria Limosilactobacillus reuteri LR1 and Lacticaseibacillus rhamnosus F and Multidrug Resistant Strain of Klebsiella pneumonia

The expansion of multiple drug resistant (MDR) strains of Klebsiella pneumoniae presents an immense threat for public health. Annually, this microorganism causes thousands of lethal nosocomial infections worldwide. Currently, it has been shown that certain strains of lactic acid bacteria (LAB) can efficiently inhibit growth of K. pneumoniae and the formation of its biofilms; however, the active principle of such action remains unknown. In the current article, the growth inhibition of MDR K. pneumoniae by two LAB—Limosilactobacillus reuteri LR1 and Lacticaseibacillus rhamnosus F—is demonstrated, and the nature of this inhibition studied at the level of exoproteome. This article shows that the exoproteomes of studied LAB contains both classically and non-classically secreted proteins. While for L. reuteri LR1 the substantial portion of classically secreted proteins was presented by cell-wall-degrading enzymes, for L. rhamnosus F only one out of four classically secreted proteins was presented by cell-wall hydrolase. Non-classically secreted proteins of both LAB were primarily metabolic enzymes, for some of which a possible moonlighting functioning was proposed. These results contribute to knowledge regarding antagonistic interaction between LAB and pathogenic and opportunistic microorganisms and set new perspectives for the use of LAB to control the spread of these microorganisms.


Introduction
Currently, the large amount of antibiotics used in human medicine as well as in animal farming has resulted in the emergence and uncontrollable expansion of many multiple drug resistant (MDR) strains of pathogenic and opportunistic microorganisms [1]. Causing hospital-acquired (i.e., nosocomial) infections, these microorganisms are responsible for thousands of deaths worldwide [2][3][4]. This situation encourages scientists to search for antimicrobial therapeutic solutions as an alternative to antibiotics. In this respect, the use of beneficial microorganisms (i.e., probiotics) and their products (i.e., postbiotics) to eliminate potential pathogens and restore microbial balance is a recently established and highly promising field of research [5][6][7].
For a long time, lactic acid bacteria (LAB) were primarily studied with respect to food fermentation-the biochemical process that extends shelf life and increases the safety of products (e.g., milk, meat, fish and vegetables) while preserving and/or improving their main nutritional value [8][9][10]. It was originally thought that the main reason behind prolonged shelf-life and increased safety of fermented products is the presence of organic Functional annotation of the predicted protein-coding genes was performed using the orthologous groups database eggNOG [23], and the prediction of signal peptides and possible secretion was performed with SignalP [24] and SecretomeP [25], respectively ( Figure 1, Supplementary Tables S1 and S2). As a result of general functional prediction, 1669 genes (86%) of L. reuteri LR1 and 2207 genes (85%) of L. rhamnosus F received clusters of orthologous groups (COG) functional categories-1345 (69%) and 1718 (66%) upon exclusion of nonspecific COG categories (i.e., R, S and X), respectively. For L. reuteri LR1, 570 genes (29%) were assigned to a specific Enzyme Commission number (EC number), and for L. rhamnosus F this number was received by 743 genes (28%). In the L. reuteri LR1 genome, 88 proteins (4%) containing signal peptides were determined, and for 732 proteins (38%) possible non-classical secretion was predicted (these proteins do not contain signal peptides but can be secreted according to SecretomeP). In the L. rhamnosus F genome, 182 proteins (7%) containing signal peptides were determined, and for 919 (35%) proteins possible non-classical secretion was predicted.

Co-Cultivation of L. reuteri LR1 and L. rhamnosus F with Multidrug Resistant K. pneumonia
To examine antagonistic activity of L. reuteri LR1 and L. rhamnosus F against planktonic cells of the MDR clinical isolate of K. pneumoniae, two two-species co-cultivations, L. reuteri LR1 with K. pneumoniae (ReKl), and L. rhamnosus F with K. pneumoniae (RhKl), and one three-species co-cultivation, L. reuteri LR1 and L. rhamnosus F with K. pneumoniae (ReRhKl), were performed. The single-species cultivation of K. pneumoniae was used as a control. The dynamics of changes in the viable cell count are shown in the Figure 2. Functional annotation of the L. reuteri LR1 and L. rhamnosus F genomes. General results of the eggNOG, SignalP and SecretomeP annotations are summarized on the area-proportional Euler diagrams (please, note that some intersections can be excluded to achieve the trade-off between accuracy of the intersection areas and meaningful arrangement of the ellipses). Information about the clusters of orthologous groups (COG) content of the genomes, excluding nonspecific COG categories (i.e., R, S and X), are summarized on the double-layer donut charts.

Co-Cultivation of L. reuteri LR1 and L. rhamnosus F with Multidrug Resistant K. pneumonia
To examine antagonistic activity of L. reuteri LR1 and L. rhamnosus F against planktonic cells of the MDR clinical isolate of K. pneumoniae, two two-species co-cultivations, L. reuteri LR1 with K. pneumoniae (ReKl), and L. rhamnosus F with K. pneumoniae (RhKl), and one three-species co-cultivation, L. reuteri LR1 and L. rhamnosus F with K. pneumoniae (ReRhKl), were performed. The single-species cultivation of K. pneumoniae was used as a control. The dynamics of changes in the viable cell count are shown in the Figure 2. Functional annotation of the L. reuteri LR1 and L. rhamnosus F genomes. General results of the eggNOG, SignalP and SecretomeP annotations are summarized on the area-proportional Euler diagrams (please, note that some intersections can be excluded to achieve the trade-off between accuracy of the intersection areas and meaningful arrangement of the ellipses). Information about the clusters of orthologous groups (COG) content of the genomes, excluding nonspecific COG categories (i.e., R, S and X), are summarized on the double-layer donut charts.
While slow growth of K. pneumoniae was observed in its monoculture, in co-cultivations with LAB the viable cell count of K. pneumoniae constantly decreased. The most prominent decrease was observed in the case of ReKl co-cultivation, during which the viable cell count of K. pneumoniae steeply decreased by approximately 4 orders of magnitude after 24 h of co-cultivation and remained at the same level up to 48 h of cultivation. In the RhKl co-cultivation, viable cell count of K. pneumoniae declined less sharply, by 2 and 3 orders of magnitude after 24 and 48 h of co-cultivation, respectively. Surprisingly, during ReRhKl co-cultivation the smallest decrease in viable cell count of K. pneumoniae was observed, as compared with the ReKl and RhKl co-cultivations, by 1 and 2 orders of magnitude after 24 and 48 h of cultivation, respectively.
It may be hypothesized that the less prominent suppression of K. pneumoniae during ReRhKl co-cultivation was related to the antagonistic interactions between LAB. To substantiate this hypothesis, two additional experiments were performed. Firstly, the growth of L. reuteri LR1 and L. rhamnosus F during their liquid-state single-species cultivation was compared with that during the two-species co-cultivation of these LAB (ReRh) ( Figure 3A).
While slow growth of K. pneumoniae was observed in its monoculture, in co-cultivations with LAB the viable cell count of K. pneumoniae constantly decreased. The most prominent decrease was observed in the case of ReKl co-cultivation, during which the viable cell count of K. pneumoniae steeply decreased by approximately 4 orders of magnitude after 24 h of co-cultivation and remained at the same level up to 48 h of cultivation. In the RhKl co-cultivation, viable cell count of K. pneumoniae declined less sharply, by 2 and 3 orders of magnitude after 24 and 48 h of co-cultivation, respectively. Surprisingly, during ReRhKl co-cultivation the smallest decrease in viable cell count of K. pneumoniae was observed, as compared with the ReKl and RhKl co-cultivations, by 1 and 2 orders of magnitude after 24 and 48 h of cultivation, respectively.
It may be hypothesized that the less prominent suppression of K. pneumoniae during ReRhKl co-cultivation was related to the antagonistic interactions between LAB. To substantiate this hypothesis, two additional experiments were performed. Firstly, the growth of L. reuteri LR1 and L. rhamnosus F during their liquid-state single-species cultivation was compared with that during the two-species co-cultivation of these LAB (ReRh) ( Figure  3A). Secondly, the interaction between LAB during their solid-state co-cultivation was tested by the perpendicular streak method ( Figure 3B).
While slow growth of K. pneumoniae was observed in its monoculture, in co-cultivations with LAB the viable cell count of K. pneumoniae constantly decreased. The most prominent decrease was observed in the case of ReKl co-cultivation, during which the viable cell count of K. pneumoniae steeply decreased by approximately 4 orders of magnitude after 24 h of co-cultivation and remained at the same level up to 48 h of cultivation. In the RhKl co-cultivation, viable cell count of K. pneumoniae declined less sharply, by 2 and 3 orders of magnitude after 24 and 48 h of co-cultivation, respectively. Surprisingly, during ReRhKl co-cultivation the smallest decrease in viable cell count of K. pneumoniae was observed, as compared with the ReKl and RhKl co-cultivations, by 1 and 2 orders of magnitude after 24 and 48 h of cultivation, respectively.
It may be hypothesized that the less prominent suppression of K. pneumoniae during ReRhKl co-cultivation was related to the antagonistic interactions between LAB. To substantiate this hypothesis, two additional experiments were performed. Firstly, the growth of L. reuteri LR1 and L. rhamnosus F during their liquid-state single-species cultivation was compared with that during the two-species co-cultivation of these LAB (ReRh) ( Figure  3A). Secondly, the interaction between LAB during their solid-state co-cultivation was tested by the perpendicular streak method ( Figure 3B). Both liquid-and solid-state experiments demonstrated that although L. reuteri LR1 and L. rhamnosus F did not show pronounced antagonism, they inhibited the intensive growth of each other. While during single-species liquid-state cultivation relative changes of viable cell count for L. reuteri LR1 and L. rhamnosus F were 86 and 29 times in 24 h, during ReRh co-cultivation they were 35 and 6 times, respectively ( Figure 3A). When tested on a solid agar medium, both LAB did not show strong zones of growth inhibition. However, near the zone of contact (intersection of two streaks), the least intensive growth was observed for both LAB ( Figure 3B).

Compositional Analysis of Exoproteomes
In order to identify proteins differentially secreted by LABs, cell-free cultural liquid was collected from all mentioned cultivations (i.e., Re, Rh, ReKl, RhKl, ReRh and ReRhKl) at the 24 h time point. The samples were concentrated; proteins were precipitated, separated by 2DE and analyzed by MALDI TOF/TOF MS/MS. The analyzed proteins were matched with those annotated in the genomes of L. reuteri LR1 and L. rhamnosus F.
For all cultivations, analysis of 2D-gels revealed approximately 100 spots that were assigned by MALDI TOF/TOF MS/MS to 28 unique proteins, 16 proteins from the genome of L. reuteri LR1 and 12 proteins from the genome of L. rhamnosus F (Figures 4 and 5). It should be noted that after manual curation two proteins of L. reuteri LR1, MBU5983247.1 and MBU5983476.1 (both are products of incomplete open reading frames), turned out to be parts of a single protein whose gene was misassembled due to the presence of a highly repetitive region. Hence, for further analysis this protein was abbreviated as MBU5983247.1/MBU5983476.1. Importantly, no proteins belonging to K. pneumoniae were detected in any cultivation. By their nature, proteins secreted by L. reuteri LR1 and L. rhamnosus F were extremely different. Only one type of protein, L-lactate dehydrogenase, was secreted by both LAB. For L. reuteri LR1 this protein was secreted during ReKl, ReRh, and ReRhKl co-cultivations, while for L. rhamnosus F it was secreted only during ReRhKl co-cultivation.
For both LAB the greatest number of secreted proteins was detected during ReRhKl co-cultivation, 11 for L. reuteri LR1 and 10 for L. rhamnosus F; while the lowest was detected during their growth in monoculture (i.e., Re and Rh), three for L. reuteri LR1 and two for L. rhamnosus F (Figures 4 and 5). For two-species co-cultivations of L. reuteri LR1, nine proteins were detected during ReRh co-cultivation and eight during ReKl co-cultivation. For both two-species co-cultivations of L. rhamnosus F (i.e., RhKl, ReRh), four proteins were detected. In comparison with L. rhamnosus F, L. reuteri LR1 was characterized by the greater number of proteins present, which was detected in several co-cultivations (eleven vs. five). At the same time, L. rhamnosus F secreted a greater number of co-cultivation specific proteins (seven vs. five for L. reuteri LR1). Importantly, there were no monoculture (i.e., Re and Rh) specific proteins detected for both LAB, and all co-cultivation specific proteins were detected upon the presence of K. pneumoniae (i.e., ReKl, RhKl and ReRhKl).

Functional Analysis of Exoproteomes
For the proteins that were identified in all the exoproteomes, the NCBI and eggNOG functional annotations were joined and manually curated ( Table 2). For both LAB approximately 50% of the identified proteins (ten for L. reuteri LR1 and six for L. rhamnosus F) were involved in different metabolic processes. All these proteins were metabolic enzymes, with the only exception being the solute-binding protein (MBU5982880.1) of L. reuteri LR1 that participates in the active transport of amino acids across the cytoplasmic membrane. Most of the determined metabolic enzymes (eight for L. reuteri LR1 and five for L. rhamnosus F) were primarily intracellular. The only extracellular metabolic enzymes were nucleotidase (MBU5982646.1) and β-galactosidase (MBU5977496.1) of L. reuteri LR1 and L. rhamnosus F, respectively. Additionally, possible non-classical secretion was determined for ribonucleoside reductase (MBU5983020.1) of L. reuteri LR1 and glyceraldehyde-3-phosphate dehydrogenase (MBU5978705.1) of L. rhamnosus F.   Figure 5. Cultivations: Re, L. reuteri LR1 monoculture; ReKl, L. reuteri LR1 and K. pneumoniae two-species co-cultivation; Rh, L. rhamnosus F monoculture; RhKl, L. rhamnosus F and K. pneumoniae two-species co-cultivation; ReRh, L. reuteri LR1 and L. rhamnosus F twospecies co-cultivation; ReRhKl, L. reuteri LR1 and L. rhamnosus F and K. pneumoniae three-species co-cultivation. The proteins belonging to L. reuteri LR1 are designated in red and to L. rhamnosus F in green. The result of protein identifications is presented in Figure 5. Cultivations: Re, L. reuteri LR1 monoculture; ReKl, L. reuteri LR1 and K. pneumoniae two-species co-cultivation; Rh, L. rhamnosus F monoculture; RhKl, L. rhamnosus F and K. pneumoniae two-species co-cultivation; ReRh, L. reuteri LR1 and L. rhamnosus F two-species co-cultivation; ReRhKl, L. reuteri LR1 and L. rhamnosus F and K. pneumoniae three-species co-cultivation.  On the UpSet-style plot, the presence of a protein in the corresponding cultivation is marked by a colored circle, if that protein was detected in several cultivations, the corresponding circles are joined by a solid line. Cultivations: Re, L. reuteri LR1 monoculture; ReKl, L. reuteri LR1 and K. pneumoniae two-species co-cultivation; Rh, L. rhamnosus F monoculture; RhKl, L. rhamnosus F and K. pneumoniae two-species co-cultivation; ReRh, L. reuteri LR1 and L. rhamnosus F two-species co-cultivation; ReRhKl, L. reuteri LR1 and L. rhamnosus F and K. pneumoniae three-species co-cultivation. On the UpSet-style plot, the presence of a protein in the corresponding cultivation is marked by a colored circle, if that protein was detected in several cultivations, the corresponding circles are joined by a solid line. Cultivations: Re, L. reuteri LR1 monoculture; ReKl, L. reuteri LR1 and K. pneumoniae two-species co-cultivation; Rh, L. rhamnosus F monoculture; RhKl, L. rhamnosus F and K. pneumoniae two-species co-cultivation; ReRh, L. reuteri LR1 and L. rhamnosus F two-species co-cultivation; ReRhKl, L. reuteri LR1 and L. rhamnosus F and K. pneumoniae three-species co-cultivation.
In the exoproteomes of L. reuteri LR1, the main part of the metabolic enzymes was involved either in the amino acid transport and metabolism (COG E, three proteins) or nucleotide transport and metabolism (COG F, three proteins). In comparison, the main part of the metabolic enzymes secreted by L. rhamnosus F was involved either in energy production and conversion (COG C, three proteins) or carbohydrate transport and metabolism (COG G, three proteins).
In addition to the metabolically active proteins, the exoproteomes of both LAB contained several cell wall-degrading enzymes (three for L. reuteri LR1 and two for L. rhamnosus F). All these enzymes contained signal peptides and, hence, were secreted via classical secretion pathways. The only exception was lytic transglycosylase (MBU5977346.1) of L. rhamnosus F for which possible non-classical secretion was determined.  In the exoproteomes of L. reuteri LR1, two functional groups of proteins absent in the exoproteomes of L. rhamnosus F were identified. The first group contained two proteins participating in cell wall adhesion (i.e., adhesins), and the second group were presented by Chaperonin GroEL that promotes the refolding and proper assembly of unfolded polypeptides. The unique for L. rhamnosus F functional groups of proteins were presented by two proteases and one cold-shock protein that participate in post-translational gene regulation. Importantly, in the exoproteomes of L. rhamnosus F one protein belonging to the viral capsid (MBU5978877.1) was identified.

Discussion
In summary, the performed two-and three-species co-cultivation of MDR K. pneumoniae, L. reuteri LR1 and L. rhamnosus F demonstrated the antagonistic effects among these bacteria. While L. reuteri LR1 and L. rhamnosus F were able to suppress the growth of K. pneumoniae by 2-4 orders of magnitude during their two-species co-cultivations (i.e., ReKl and RhKl), in the three-species co-cultivation (i.e., ReRhKl) the growth of K. pneumoniae was suppressed only by 1-2 orders of magnitude. The latter was explained by the observed weak antagonistic effect between L. reuteri LR1 and L. rhamnosus F. Hence, the obtained data suggest that, at least in vitro, the mixture of L. reuteri LR1 and L. rhamnosus F is inferior in terms of K. pneumoniae growth suppression to the single-strain probiotic culture; however, whether or not this tendency will be observed during in vivo administration of probiotics must be independently confirmed. Although there is a current tendency to mix as many probiotic strains as possible into a single product, our data strongly support the recent claim that the choice of an appropriate probiotic product should not be based on the number of strains in it but rather on evidence-based trials of the product's efficacy in a given situation [26].
The obtained assembly and annotation of the draft genomes of L. reuteri LR1 and L. rhamnosus F were of comparable quality with those previously published [27]. Although each genome provides an enormous amount of information, in this article the obtained genomes were primarily used as the mean to unambiguously identify proteins present in the studied exoproteomes. Besides the fact that exoproteomes of LAB grown in the presence of K. pneumoniae were never previously reported, the main rationale for its investigation with respect to the antagonistic interactions between studied strains were as follows: (1) since both LAB and K. pneumoniae are acid-tolerant bacteria [28,29], all possible nonspecific acid-related bactericidal effects play a negligibly small role during their interactions; (2) although production of bacteriocins can play a role in antagonistic interactions between studied LAB, the hypothesis of a bacteriocins-related inhibition of K. pneumoniae growth seems unlikely; it is generally accepted that Gram-negative bacteria possess a natural resistance to the bacteriocins synthesized by Gram-positive bacteria [30,31]; (3) there is an ever growing body of evidence that certain secreted proteins of LAB could be responsible, among others, for their pathogen-inhibitory properties [32][33][34].
In general, all proteins detected in the exoproteomes of both LAB can be classified into those that are primarily extracellular (i.e., can undergo classical secretory pathways due to the presence of signal peptides) and those that are primarily intracellular. It should be noted that in the current study cell lysis was not observed; hence, it may be proposed that the detection of primarily intracellular proteins in the exoproteomes is a result of their non-classical secretion, as was previously discussed in [35]. The obtained data show substantial limitations of SecretomeP software [25] for prediction of non-classically secreted proteins in LAB, since it failed to predict this secretion for many apparently non-classically secreted proteins (Table 2).
In the case of L. reuteri LR1, the substantial portion of classically secreted proteins (3 out of 7) was presented by cell wall-degrading enzymes ( Table 2). Although these enzymes play a significant role in the growth and division of L. reuteri LR1, it may be proposed that in co-cultivations they can participate in antagonistic interactions by disrupting the cell walls of neighboring bacteria [36]. Upon the presence of L. rhamnosus F in co-cultivation (ReRh and ReRhKl), L. reuteri LR1 secreted such cell wall-degrading enzymes as Nacetylglucosaminidase (Glycosyl hydrolase family 73, GH73) (MBU5982718.1) and metalloendopeptidase (family M23) (MBU5982939.1) ( Table 2). While N-acetylglucosaminidase (MBU5982718.1) hydrolyses the glycosidic bonds in the peptidoglycan (PG) of both Grampositive and Gram-negative bacteria, peptidase M23 (MBU5982939.1) cuts the peptide bond of the PG cross-link in Gram-positive bacteria ( Figure 6). Although, the involvement of peptidase of family M23 in antagonistic interaction between Lactobacillus salivarius and Staphylococcus aureus has been previously reported [37], this is the first report on the possibility of the involvement of these peptidases in the antagonistic interactions between different LAB species. Upon the presence of K. pneumoniae in co-cultivation (RhKl), L. reuteri LR1 secreted Peptidase P60 (MBU5983126.1) containing NlpC/P60 domain. Previously it has been demonstrated that NlpC/P60 domain-containing peptidases can catalyze the hydrolysis of both Gram-positive and Gram-negative bacterial PG [38] (Figure 6). Among the remaining classically secreted proteins of L. reuteri LR1, two proteins were putative adhesins (MBU5983247.1/MBU5983476.1 and MBU5981906.1), one a solute binding protein and the other a nucleosidase. While the secretion of MBU5983247.1/MBU5983476.1 was linked with the presence of L. rhamnosus F in co-cultivation (ReRh and ReRhKl), the secretion of MBU5981906.1 was observed on every cultivation (Re, ReKl, ReRh and ReRhKl). It may be hypothesized, that the secretion of these adhesins was stimulated by the expected competition over adhesion to the surfaces of possible host cells; hence, in the current experimental setup it can be regarded as an artifact that did not contribute to the observed antagonistic interactions. The presence of solute-binding protein in monoculture (Re) and co-cultivation with L. rhamnosus F (ReRh) can be explained by the active growth of L. reuteri LR1 in the former case and the competition for substrate in the latter. The presence of nucleosidase in three-species co-cultivation (ReRhKl) can also be explained by the competition for the substrate among microorganisms.
vation (ReRh and ReRhKl), the secretion of MBU5981906.1 was observed on every cultivation (Re, ReKl, ReRh and ReRhKl). It may be hypothesized, that the secretion of these adhesins was stimulated by the expected competition over adhesion to the surfaces of possible host cells; hence, in the current experimental setup it can be regarded as an artifact that did not contribute to the observed antagonistic interactions. The presence of solute-binding protein in monoculture (Re) and co-cultivation with L. rhamnosus F (ReRh) can be explained by the active growth of L. reuteri LR1 in the former case and the competition for substrate in the latter. The presence of nucleosidase in three-species co-cultivation (ReRhKl) can also be explained by the competition for the substrate among microorganisms. Figure 6. Cell wall-degrading enzymes of L. reuteri LR1 (red) and L. rhamnosus F (green) determined in this study ( Figure  4, Figure 5 and Table 2).
In the case of L. rhamnosus F, only one out of four classically secreted proteins was a cell wall-degrading enzyme, cell wall hydrolase P75 (MBU5979459.1), also known as major secreted protein 1 [39], containing NlpC/P60 domain ( Figure 6). Although this hydrolase was present in monoculture (Rh), in co-cultivations its secretion can be associated with the presence of K. pneumoniae (RhKl and ReRhKl). Also, it is worth mentioning here the secretion, although non-classical, of lytic transglycosylase (MBU5977346.1) in the cocultivation with K. pneumoniae (RhKl). This enzyme can catalyze the hydrolysis of both Gram-positive and Gram-negative bacterial PG ( Figure 6). The presence of a detectable quantity of cell envelope proteinase PrtR only upon co-cultivation with L. reuteri LR1 (ReRh) and β-galactosidase only in the three-species co-cultivation (ReRhKl) may be explained by active competition between microorganisms for the substrates (amino acids and sugars, respectively) during these co-cultivations. As for the adhesins secreted by L. reuteri LR1, the secretion of Zn-dependent protease (peptidase family M10), which is homologous to the human matrix metallopeptidase and is capable of degrading extracellular matrix proteins, in the co-cultivation with K. pneumoniae (RhKl) was most probably stimulated by the expected competition for adhesion to the host cells.
Since the main part of the non-classically secreted proteins found in the exoproteomes of both LAB was primarily metabolic enzymes, it may be proposed that in the observed situation these proteins demonstrated a moonlighting functionality. By definition, moonlighting proteins are proteins that apart from their well-characterized function can  Table 2).
In the case of L. rhamnosus F, only one out of four classically secreted proteins was a cell wall-degrading enzyme, cell wall hydrolase P75 (MBU5979459.1), also known as major secreted protein 1 [39], containing NlpC/P60 domain ( Figure 6). Although this hydrolase was present in monoculture (Rh), in co-cultivations its secretion can be associated with the presence of K. pneumoniae (RhKl and ReRhKl). Also, it is worth mentioning here the secretion, although non-classical, of lytic transglycosylase (MBU5977346.1) in the cocultivation with K. pneumoniae (RhKl). This enzyme can catalyze the hydrolysis of both Gram-positive and Gram-negative bacterial PG ( Figure 6). The presence of a detectable quantity of cell envelope proteinase PrtR only upon co-cultivation with L. reuteri LR1 (ReRh) and β-galactosidase only in the three-species co-cultivation (ReRhKl) may be explained by active competition between microorganisms for the substrates (amino acids and sugars, respectively) during these co-cultivations. As for the adhesins secreted by L. reuteri LR1, the secretion of Zn-dependent protease (peptidase family M10), which is homologous to the human matrix metallopeptidase and is capable of degrading extracellular matrix proteins, in the co-cultivation with K. pneumoniae (RhKl) was most probably stimulated by the expected competition for adhesion to the host cells.
Since the main part of the non-classically secreted proteins found in the exoproteomes of both LAB was primarily metabolic enzymes, it may be proposed that in the observed situation these proteins demonstrated a moonlighting functionality. By definition, moonlighting proteins are proteins that apart from their well-characterized function can perform other physiologically relevant biochemical or biophysical activities, frequently, in different cell compartments and at different times [40][41][42]. Typically, intracellular/extracellular moonlighting proteins are "housekeeping proteins" that normally participate in the central processes of carbohydrate, protein and nucleic acid metabolism [43]. As in the case of many other microorganisms, there is a limited amount of knowledge regarding the moonlighting proteins of LAB. The main part of known LAB moonlighting proteins has been previously associated with the host-probiotic interaction, and the most well-known examples are GAPDH and enolase [44][45][46].
As it was already mentioned with regard to the classically secreted adhesins of L. reuteri LR1 and Zn-dependent protease of L. rhamnosus F, although secretion of moonlighting proteins that promotes host-probiotic interaction was most probably induced by the presence of the neighboring microorganisms, in the current experimental setup this secretion can be regarded as an artifact since no host to interact with was present. Nevertheless, moonlighting activities different from the host-probiotic interaction can be proposed for several proteins based on the previously published data.
For L. reuteri LR1, the possible moonlighting function can be proposed for Cysteine synthase A (CysK) whose secretion can be associated with the presence of K. pneumoniae (ReKl and ReRhKl). CysK typically carries out the second step in the pathway of cysteine biosynthesis, synthesizing cysteine from sulfide (H 2 S) and O-acetyl-L-serine; however, it has been shown that in certain bacteria, including Lactobacillus casei, this enzyme can perform cysteine desulfurization leading to the formation of pyruvate, ammonia (NH 3 ) and H 2 S [47]. Recent investigations demonstrate that the presence of both NH 3 and H 2 S during cultivation significantly influenced bacterial physiology [48,49]. As an example, it has been previously shown that production of NH 3 by Proteus mirabilis gives it a competitive advantage over K. pneumoniae [50].
For L. rhamnosus F, the possible moonlighting function may be proposed for D-lactate dehydrogenase. The D-lactate dehydrogenase (D-LDH/LdhD-1, YP_003169904.1) of L. rhamnosus GG, homologous to that (MBU5977590.1) detected in the exoproteome of L. rhamnosus F during three-species co-cultivation (ReRhKl), was previously recombinantly expressed in Escherichia coli [51]. This protein has a 91.3 % similarity with the D-lactate dehydrogenase (HicDH, AAA25236.1) of Lactobacillus casei, which plays the main role in the formation of D-lactic acid [52]. However, most probably in L. rhamnosus it plays a different role, since its activity by pyruvate is very low [51]. There is a high probability that this enzyme can perform a D-2-hydroxy acid dehydrogenase function producing phenyllactic acid from phenylpyruvate [53,54]. The phenyllactic acid is a natural biological antimicrobial agent that has been previously shown to effectively inactivate Klebsiella oxytoca planktonic and biofilm cells [55].
Among all non-classically secreted proteins of L. rhamnosus F, special attention should be devoted to the viral capsid protein (MBU5978877.1) secretion of which has been linked with the presence of L. reuteri LR1 in co-cultivation (ReRh and ReRhKl). Although in the genome of L. rhamnosus F this protein is part of an integrated prophage, there were no additional phage proteins of this prophage detected in any exoproteome. Hence, it can be concluded that the gene encoding MBU5978877.1 is a "moron locus" [56]. While the majority of phage-encoded genes are repressed during lysogeny, moron loci are often highly expressed and can provide a competitive advantage to the host [57]. Interestingly, the secretion of individual prophage genes was previously reported for L. rhamnosus LRB [58], while L. rhamnosus Pen can spontaneously release entire phage particles that were proposed to increase the survivability of microorganism in their natural ecological niche [59].

Bacterial Strains and Pre-Cultivations
Strains of L. reuteri LR1 and L. rhamnosus F were obtained from the Microorganism Collection of the All-Russia Research Institute of the Dairy Industry (VNIMI, Moscow, Russia). The sequences of the 16S ribosomal RNA genes of these strains can be found at the GeneBank accessory numbers MN994628 and MN994629 for L. reuteri LR1 and L. rhamnosus F, respectively.
For inoculum preparation, LAB and hospital strains of K. pneumoniae were cultivated under anaerobic conditions on MRS (De Man, Rogosa and Sharpe) broth at 37 ± 1 • C to achieve the turbidity of 3 McFarland standard and stored at 4 • C until further use; the viable cell counts were measured (as described below in the Sections 4.3 and 4.4) and adjusted to 10 8 CFU·mL −1 by dilution with MRS broth before inoculations.

DNA Isolation, Sequencing and Annotation
For DNA extraction, L. reuteri LR1 and L. rhamnosus F were statically cultivated in 20 mL of MRS broth at 37 ± 1 • C until cloudy in appearance. The approximately 0.1 mL of bacterial cells were separated by centrifugation at 10,000× g for 5 min at 4 • C, pretreated with 20 µL of lysozyme solution (50 mg·mL −1 ) (Serva, Heidelberg, Germany) and 5 µL of RNase A (2 mg·mL −1 ) (Fermentas, St. Leon-Rot, Germany) at 37 • C for 20 min, and total DNA was extracted using DNeasy mericon Food Kit (Qiagen, Valencia, CA, USA), according to the manufacturer's protocol. The quality and quantity of the isolated DNA were checked using an Agilent Bioanalyzer 2100 (Agilent Technologies, Foster City, CA, USA) and Qubit fluorimeter (Thermo Fisher Scientific, Waltham, MA, USA).
The DNA library was prepared using the Ion AmpliSeq library kit 2.0 (Thermo Fisher Scientific, MA, USA) and indexed with an Ion Xpress barcode adapters 1-16 kit (Thermo Fisher Scientific, MA, USA). The quality and quantity of the obtained DNA library were checked using Agilent Bioanalyzer 2100. Whole genome sequencing was carried out using the Ion Torrent Personal Genome Machine (PGM) (Thermo Fisher Scientific, MA, USA). The obtained reads were pre-processed and assembled with CLC Genomics Workbench 11.0 (Qiagen, Valencia, CA, USA). Upon submission, genome annotations were performed using NCBI Prokaryotic Genome Annotation Pipeline (PGAP) [61]. Additionally, annotation with eggNOG [23], SignalP [24] and SecretomeP [25] was performed on the web (Supplementary Tables S1 and S2).

Co-Cultivations of LAB with K. pneumoniae
The MRS broth was inoculated with 1 mL of K. pneumoniae pre-culture and (1) 1 mL of L. reuteri LR1 pre-culture, (2) 1 mL of L. rhamnosus F pre-cultures and (3) 1 mL of L. reuteri LR1 and 1 mL of L. rhamnosus F pre-cultures. The final volume was adjusted with MRS broth to 20 mL, and incubations were performed under anaerobic conditions at 37 ± 1 • C for 24 and 48 h. All cultivations were performed in triplicate.
The viable cell count of K. pneumonia was performed on Nutrient agar (Mikrogen, Moscow, Russia), containing 18 g·L −1 of fish hydrolysate, 8 g·L −1 of NaCl, and 12 g·L −1 of agar, after cultivation under anaerobic conditions at 37 ± 1 • C for 24 h.

Co-Cultivations of LAB with Each Other
The MRS broth was inoculated with (1) 1 mL of L. reuteri LR1 pre-culture, (2) 1 mL of L. rhamnosus F pre-cultures and (3) 1 mL of L. reuteri LR1 and 1 mL of L. rhamnosus F pre-cultures. The final volume was adjusted with MRS broth to 20 mL, and incubations were performed under anaerobic conditions at 37 ± 1 • C for 24. All cultivations were performed in triplicate.

Perpendicular Streak Test
At the first stage, the pre-culture of the first LAB (L. reuteri LR1 or L. rhamnosus F) was streaked on the MRS agar and incubated under anaerobic conditions at 37 • C for 24 h. At the second stage, the pre-culture of the second LAB (L. rhamnosus F or L. reuteri LR1, respectively) was streaked perpendicularly to the first LAB, and the plate was incubated under anaerobic conditions for another 24 h at 37 • C. The antagonistic interactions between LAB were assessed visually by the presence of a growth inhibition zone.

Exoproteomics Study
After 24 h of cultivation, bacterial cells were separated from the culture liquid by centrifugation at 8000× g for 40 min at 4 • C. The cultural liquids from the parallel cultivations were pulled together. All the following procedures were performed as described in [62]. In short, proteins were precipitated and separated by both mass and isoelectric point on two-dimensional (2D) gel. The obtained protein spots were further analyzed using Ultraflex II mass spectrometer (Bruker, Germany) with matrix-assisted laser desorption/ionization ion source and tandem time-of-flight mass analyzer (MALDI TOF/TOF MS/MS). The obtained data (both peptide fingerprints and sequences) were matched against an in-house database of proteins annotated in the genomes of L. reuteri LR1 and L. rhamnosus F (see the Section 4.2).

Conclusions
In conclusion, it was shown that as a part of their complex antagonistic system LAB can secrete proteins with potential antimicrobial activity against K. pneumoniae cells. The performed comparative MS analyses of the exoproteomes from co-cultivations of LAB with K. pneumoniae allowed the identification of several candidate antimicrobial proteins: NlpC/P60 endopeptidases, lytic transglycosylase, Zn-dependent protease (peptidase family M10), nucleoside hydrolase RihC, nucleotidase, ribonucleoside reductase and others. These enzymes are attractive targets for the discovery of novel antimicrobial compounds. Data Availability Statement: The Whole Genome Shotgun projects were deposited at DDBJ/ENA/GenBank under the accessions JAHLXI000000000 and JAHLXH000000000 for Limosilactobacillus reuteri LR1 and Lacticaseibacillus rhamnosus F, respectively. The versions described in this paper are versions JAHLXI000000000.1 and JAHLXH000000000.1 for L. reuteri LR1 and L. rhamnosus F, respectively.

Conflicts of Interest:
The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.