Proteomic Study of the Interactions between Phages and the Bacterial Host Klebsiella pneumoniae

ABSTRACT Phages and bacteria have acquired resistance mechanisms for protection. In this context, the aims of the present study were to analyze the proteins isolated from 21 novel lytic phages of Klebsiella pneumoniae in search of defense mechanisms against bacteria and also to determine the infective capacity of the phages. A proteomic study was also conducted to investigate the defense mechanisms of two clinical isolates of K. pneumoniae infected by phages. For this purpose, the 21 lytic phages were sequenced and de novo assembled. The host range was determined in a collection of 47 clinical isolates of K. pneumoniae, revealing the variable infective capacity of the phages. Genome sequencing showed that all of the phages were lytic phages belonging to the order Caudovirales. Phage sequence analysis revealed that the proteins were organized in functional modules within the genome. Although most of the proteins have unknown functions, multiple proteins were associated with defense mechanisms against bacteria, including the restriction-modification system, the toxin-antitoxin system, evasion of DNA degradation, blocking of host restriction and modification, the orphan CRISPR-Cas system, and the anti-CRISPR system. Proteomic study of the phage-host interactions (i.e., between isolates K3574 and K3320, which have intact CRISPR-Cas systems, and phages vB_KpnS-VAC35 and vB_KpnM-VAC36, respectively) revealed the presence of several defense mechanisms against phage infection (prophage, defense/virulence/resistance, oxidative stress and plasmid proteins) in the bacteria, and of the Acr candidate (anti-CRISPR protein) in the phages. IMPORTANCE Researchers, including microbiologists and infectious disease specialists, require more knowledge about the interactions between phages and their bacterial hosts and about their defense mechanisms. In this study, we analyzed the molecular mechanisms of viral and bacterial defense in phages infecting clinical isolates of K. pneumoniae. Viral defense mechanisms included restriction-modification system evasion, the toxin-antitoxin (TA) system, DNA degradation evasion, blocking of host restriction and modification, and resistance to the abortive infection system, anti-CRISPR and CRISPR-Cas systems. Regarding bacterial defense mechanisms, proteomic analysis revealed expression of proteins involved in the prophage (FtsH protease modulator), plasmid (cupin phosphomannose isomerase protein), defense/virulence/resistance (porins, efflux pumps, lipopolysaccharide, pilus elements, quorum network proteins, TA systems, and methyltransferases), oxidative stress mechanisms, and Acr candidates (anti-CRISPR protein). The findings reveal some important molecular mechanisms involved in the phage-host bacterial interactions; however, further study in this field is required to improve the efficacy of phage therapy.

phages have developed mechanisms to evade the bacterial CRISPR-Cas system. For instance, through a single-nucleotide substitution or a complete deletion in the protospacer region or in the conserved protospacer-adjacent motif (37). Phages have also developed anti-CRISPR systems, which basically consist of Acr proteins (typically small proteins of 80 to 150 amino acids) that inhibit bacterial CRISPR-Cas activity by binding directly to, and thus inactivating, the Cas protein, so that phages can successfully replicate in the bacterial host (38).
A better understanding of phage-host interaction could lead to the development of more successful therapeutic applications for phages. In this context, the aims of the present study were to analyze the proteins isolated from 21 novel lytic phages of K. pneumoniae in search of defense mechanisms against bacteria and also to determine their infective capacity. In addition, the other aim of this work was to investigate the defense mechanisms of bacteria in response to phage infection.
Phage genome annotation. (i) Phage genome analysis. The phage genome sequencing revealed that all phages under study, available from the GenBank BioProject PRJNA739095 (Table 1), were lytic Caudovirales phages, i.e., dsDNA tailed phages, lacking lysogenic genes such as integrase, recombinase, and excisionase. More specifically, (ii) Genetic defense mechanisms of phages. An in-depth study of the phage genomes with different bioinformatic tools revealed the presence of defense mechanisms ( Table 2): (i) 35 RM system evasion mechanisms located in 16 phages, (ii) six TA systems located in two phages (vB_KpnM-VAC13 and vB_KpnM-VAC66), (iii) one DNA degradation evasion located in phage vB_KpnP-VAC1, (iv) four blocking RM of host bacteria located in phage vB_KpnM-VAC36, (v) seven genes that confer resistance to the Abi system of host bacteria located in three phages (vB_KpnM-VAC13, vB_KpnM-VAC36, and vB_KpnM-VAC66), and, finally, (vi) two possible orphan CRISPR-Cas system were located in the phages vB_KpnS-VAC35 and vB_KpnS-VAC51. In addition, almost all phages possessed a possible anti-CRISPR system, composed by Acr and Aca protein, except for phages vB_KpnS-VAC112 and vB_KpnS-VAC113 (see Table S1 in the supplemental material). An inhibitor of the TA system (protein ID QZE51102.1) was also found in the genome of the phage vB_KpnP-VAC1.
(iii) Phylogenetic analysis of phages. Phylogenetic analysis was performed by aligning the nucleotide sequence of the large subunit terminase of each phage with MAFF server, followed by the elaboration of the phylogenetic tree with RAxMLHPC-PTHREADS-AVX2 version 8.2.12 (42) under the GTRGAMMA model and 100 bootstrap replicates. This analysis revealed that phages were clustered in the following families:   1A).
Host range assay. The phage infectivity assay was performed in the collection of 47 K. pneumoniae clinical strains (Table 3) by the spot test technique (Fig. 3A). The criteria used to determine the phage infectivity were the presence of clear spots (infection), the presence of turbid spots (low infection or resistance), and the lack of spots (no infection). The results showed a high variability of infectivity between the phages (Fig. 3B). Phage vB_KpnP-VAC1 had the lowest host range, infecting only the strain K2986, while phage vB_KpnM-VAC13 presented the highest range of activity, infecting 27 strains. Phages vB_KpnS-VAC35 and vB_KpnM-VAC36 also exhibited a high host range, infecting both 17 and 18 strains, respectively. Therefore, the following experiments focused on these two phages (vB_KpnS-VAC35 and vB_KpnM-VAC36).
Bacterial genome analysis. The genomes of twenty-seven clinical isolates infected by the vB_KpnS-VAC35 and vB_KpnM-VAC36 phages, according to the host range assay, were analyzed for the presence of CRISPR-Cas systems. The result revealed the presence of CRISPR-Cas systems in 14 strains (Fig. 3B). However, we proceeded to a more detailed study of K. pneumoniae clinical strains K3574 (SAMEA3649560) and K3320 (SAMEA3649520), as they were the ones best infected by phages vB_KpnS-VAC35 and vB_KpnM-VAC36, respectively. The result of this study revealed that both strains had an intact type I-E CRISPR-Cas in their genome, with 35 spacers in the case of K3574 and with 11 spacers in the case of the strain K3320 ( Fig. 4A and B). In turn, five plasmids located in five different contigs were found in the strain K3574, while strain K3320 had four plasmids located in three different contigs (Table 4). On the other hand, only strain K3574 exhibited an RM system, which was type II and functioned as a methyltransferase (Table 4). Finally, five prophages (two intact and three questionable) were detected in strain K3574, and seven prophages (three intact, two incomplete, and two questionable) were detected in strain K3320. However, only the data of the prophages considered intact are shown in Table 5.
Characterization of phages vB_KpnS-VAC35 and vB_KpnM-VAC36. (i) Phage adsorption. Adsorption of phages vB_KpnS-VAC35 and vB_KpnM-VAC36 ( Fig. 5A and B) to the bacterial surface receptor was studied with the previously selected strains K3574 and K3320, at a multiplicity of infection (MOI) of 0.01. Phage vB_KpnS-VAC35 showed a high percentage of adsorption, with 91.28% of phage adsorbed in the strain K3574 after 5 min, while phage vB_KpnM-VAC36 showed slight adsorption in the strain K3320, with 39.02% of phage adsorbed after 2 min ( Fig. 5C and D).
(ii) One-step growth curve assay. The latent period, determined by the one-step growth curve indicating the time taken for a phage particle to reproduce inside an infected host cell, and the burst size, defined as the number of viral particles released in each infection cycle per cell, were 10 min and 45.52 PFU/mL, respectively, for phage vB_KpnS-VAC35 in strain K3574 and 8 min and 2.71 PFU/mL for the phage vB_KpnM-VAC36 in strain K3320 at an MOI of 0.01 ( Fig. 5E and F).
(iii) Phage kill curve. The infectivity assay in liquid medium to determine the infection curve for phages vB_KpnS-VAC35 and vB_KpnM-VAC36 at an MOI of 1 in the  increase of count after 2 h 30 min of phage infection 1.35 Â 10 5 6 3.54 Â 10 4 (Fig. 6C  and D). Thus, although the appearance of resistant bacteria was not observed in the OD test and the density remained unchanged for both phages, the bacteria did appear in the CFU/mL count test. Finally, we monitored the PFU/mL, and in both cases observed an increase in the number of PFU/mL after 30 min of phage infection, with Consequently, we can conclude that the number of CFU/mL is inversely proportional to the number of PFU/mL. These data confirm that the reduction in CFU is due to multiplication of the phages. NanoUHPLC-Tims-QTOF proteomic analysis: interaction between phages (vB_KpnS-VAC35 and vB_KpnM-VAC36) and clinical strains (K3574 and K3320). The proteomic study conducted by NanoUHPLC-Tims-QTOF (20) analysis revealed a large variety of proteins (listed in Fig. 7A and B, with the respective proportions in each strain): prophage-related proteins, defense, resistance and virulence proteins, oxidative stress proteins, plasmid-related proteins, tRNA, cell wall-related proteins, and membrane proteins, as well as some transport proteins and proteins related to DNA, biosynthesis or degradation of proteins, ribosomes, metabolism, and some of unknown function showing differences in expression compared to the uninfected control ( Fig. 7C and D). In both cases, some prophage-related proteins, as well as a large amount of tRNA, were found. Regarding the defense proteins, we found proteins related to porins, multidrug efflux RND transporter, RM system type I methyltransferase, two-component response regulator system, TA system type II RelE/ParE family, DNA starvation protein, fimbriae, and finally pili. The details of all detected proteins showing differences from the control are summarized in Table S2 in the supplemental material. The presence of some Acr candidates in the phage-infected strains was detected by NanoUHPLC-Tims-QTOF analysis: seven in K3574 infected by vB_KpnS-VAC35 and one in K3320 infected by the vB_KpnM-VAC36 phage ( Table 6).

DISCUSSION
Lytic phage therapy is currently considered one of the best alternatives for treating infections caused by MDR bacterial pathogens (3,4). Phages are known to exhibit some advantages over the use of antibiotics, including the continued warfare between phage and bacteria during the coevolution of both organisms (12). Consequently, phages have developed defense mechanisms to evade the resistance mechanisms of bacteria (26)(27)(28)(29)(30)(31)(32)(33)(34)(35)(36)(37)(38), while at the same time bacteria have developed defense mechanisms to prevent phage infection (43). In this context, the aims of the present study were to analyze 21 new lytic phages in search of defense mechanisms and also to identify the defense mechanisms of two clinical strains K3574 and K3320 when infected by phages, since better knowledge of the latter will lead to improvements in the use of phages to treat infections caused by MDR bacteria.
Regarding the results of the whole-genome sequencing (WGS) and annotation, we observed that all phages belonged to the order Caudovirales. Several studies have shown that dsDNA-tailed phages are the most abundant entity on earth (44,45). Most of these phages (61.90%) are members of the Drexerviridae family, 14.29% are members of the Autographiviridae family, 14.29% are members of the Myoviridae family, and 9.52% are members of the Demerecviridae family. Genome annotation has previously shown that all phages are lytic and lacking lysogenic genes such as integrase, recombinase, and excisionase (20). This point is of vital importance for use of these phages in phage therapy (46,47). Most phages were found to have a typical organization of the genome in functional modules, as previously described (2,48,49). In contrast, members of the Myoviridae family, which are included in the "larger phages" (.100 bp), did not present specific lysis blocks, and structural and morphogenesis-related proteins were repeated in several blocks   throughout the genome (41,50). The genomes of all phages had endolysins and holins, proteins that are responsible for degradation of the bacterial cell wall during the infection by the host to facilitate the exit of the phage progeny (51). Genomic annotation revealed the presence of numerous bacterial defense mechanisms: RM system evasion, TA system, DNA degradation evasion, blocking RM of host bacteria, genes that confer resistance to Abi system of host bacteria, a possible orphan CRISPR-Cas system, and almost all the phages possessed a possible anti-CRISPR system. These mechanisms have all already been described (14). The anti-CRISPR, which is composed by operons of Acr and Aca proteins, was first discovered in 2013 in phages and prophages of Pseudomonas aeruginosa (52). Acr-Aca operons are defined as genomic loci fulfilling the following criteria: (i) all genes should be in the same strand, (ii) all intergenic distances should be ,150 bp, (iii) all genes encode proteins shorter than 200 amino acids in length, and, finally, (iv) at least one gene should be homologous to Acr or Aca proteins (53). The main problem of the search of new anti-CRIPSR is that Acr proteins are very poorly conserved, and the best way to discover new anti-CRISPR is therefore to use a "guilt-by-association" approach, which searches for Aca in the genome of phages. Although the function of Acas is not yet understood, these gene often encode a protein containing a helix-turn-helix motif, suggesting that they fulfill a regulatory function (54).
The study of phage infective capacity revealed a large disparity in the infectivity, as previously demonstrated (55,56): phages vB_KpnM-VAC13 and vB_KpnM-VAC66 displayed the highest infectivity capacity (41), whereas phage vB_KpnP-VAC1 displayed the lowest infective capacity (20). The wide host range could be an advantage since it allows infection of a larger number of hosts (57), and this trait could be useful for successful phage therapy. In addition, the "larger phages" vB_KpnS-VAC35 (112.662 bp) and vB_KpnM-VAC36 (169.970 bp) showed a high infectivity for 17 and 18 clinical strains, respectively. Moreover, a possible anti-CRISPR system was detected in their genome. Thus, the presence of the CRISPR-Cas system in the bacterial strains that were successfully infected by these phages was examined to study the possible interaction of both defense mechanisms. The result of this search showed the presence of class I type I-E intact CRISPR-Cas system in the genome of the strain K3574 and K3320. Both phages were examined with their respective host strains. The adsorption curve revealed that phage vB_KpnS-VAC35 displays a higher percentage of adsorption, a higher burst size, and a longer latent period than phage vB_KpnM-VAC36. Moreover, analysis of the infectivity capacity by killing assay measuring the OD 600 , CFU/mL, and PFU/mL revealed that phage vB_KpnS-VAC35 was more effective than phage vB_KpnM-VAC36.
Finally, proteomic studies were conducted with bacterial strains K3574 and K3320 with or without phage infection (vB_KpnS-VAC35 and vB_KpnM-VAC36) to determine any differences at the level of protein expression after phage infection. The pattern of protein expression was found to vary depending on the strain considered. This may be due to the different infection status of the bacterial cell at the time of sample   FIG 7 (A and B) Graphical representation of the proteomics results, showing the abundance of each group of proteins found in the culture with the bacterial strain K3574 infected with phage vB_KpnS-VAC35 and in the culture with the bacterial strain K3320 infected with phage vB_KpnM-VAC36. (C and D) Abundance of proteins with a higher (blue), lower (orange), or undetected (gray) value areas compared to the uninfected control in strains K3574 and K3320 infected with phages vB_KpnS-VAC35 and vB_KpnM-VAC36, respectively.

Molecular Response to Phages in the Host Klebsiella
Microbiology Spectrum processing or due to the inherent proprieties of the bacteria. Therefore, the results revealed the expression of FtsH protease modulator located in prophage of strains K3574 and K3320, which controls the lytic pathway (58), as well as the expression of the cupin protein located in plasmid in the strain K3574, a phosphomannose isomerase involved in lipopolysaccharide (LPS) synthesis, which is an important determinant of pathogenicity and phage susceptibility (59). In addition, proteins related to bacterial defense, resistance, and virulence and also to oxidative stress mechanisms (60) have been observed. The difference in expression compared to the control without phage infection of porins, efflux pumps, LPS, and pilus elements, previously described in the literature as phage receptors (61,62), was also observed. Moreover, proteins involved in the quorum network were observed, e.g., the LuxS that synthesizes AI-2 molecules, or the presence of the CsrA regulator in the strain K3320 infected by phage vB_KpnM-VAC36 (63,64). Indeed, previous studies have associated the quorum network with phage infection (65)(66)(67). In addition, a type II RelE/ParE TA system was expressed in strain K3320. This is a very interesting finding, since phage vB_KpnM-VAC36 did not successfully infect strain K3320. The fundamental role played by TA systems in the inhibition of phage infection has recently been demonstrated (20,(68)(69)(70). Interestingly, an inhibitor of the TA system (protein ID QZE51102.1) was found in the genome of the phage vB_KpnP-VAC1. This type of gene, previously only described in one E. coli phage (35), may play a role in phage defense against bacteria. The methyltransferases, other important proteins that play a key role in phage infection (71), were found in both K3574 and K3320 strains infected by phages. In addition, several Acr candidate proteins were expressed in the infected strains. This is a very interesting finding, because the anti-CRISPR could inhibit the host's CRISPR-Cas system and thus promote infection (54). Therefore, the difference in expression compared to the control of all mechanisms could be due to the phage-host interaction, with the bacteria trying to use all their defense mechanisms in response to the infection.
Conclusion. Phage-host interactions have been examined ever since the discovery of phages a century ago. The present study revealed numerous defense mechanisms both against bacteria by phage (RM system evasion, TA system, DNA degradation evasion, RM block of host, resistance to Abi, anti-CRISPR and CRISPR-Cas system) and against phage infection by bacteria (prophage, plasmid, defense/virulence/resistance, and oxidative stress proteins). However, phage-host bacterial interactions remain poorly understood, and further study is required in order to improve the efficacy of phage therapy.  a The protein header information is presented as listed in the NCBI database. The accession number of the protein as seen in the NCBI database is indicated. "-10LogP" indicates the protein confidence score, and "position(s)" indicates localization in the phage genome.

MATERIALS AND METHODS
was used in this study (Table 3). The sequence type (ST) and the capsular type (K) were determined using the methods available on the Pasteur Institute website (https://bigsdb.pasteur.fr/klebsiella/, accessed between 2018 and the present) and the Kaptive website (https://github.com/kelwyres/Kaptive-Web, accessed in April 2020), respectively. All strains were grown in Luria-Bertani medium (0.5% NaCl, 0.5% yeast extract, 1% tryptone). Isolation, purification, and propagation of lytic phages and TEM. Ten new lytic phages isolated from sewage water samples and twelve lytic phages previously isolated by our research group (20,40,41) were used in this study. Isolation, purification, and propagation of the new phages was performed according to the procedures used in the previously cited articles, using strains of K. pneumoniae as a natural host (Table 1). Next, the 10 new lytic phage solutions were negatively stained with 1% aqueous uranyl acetate before being analyzed by TEM in a JEOL JEM-1011 electron microscope.
Phage DNA extraction and WGS. The phage DNA of the 10 new lytic phages was isolated with the phenol-chloroform method according to a previously published phage-hunting protocol (https://phagesdb .org/media/workflow/protocols/pdfs/PCI_SDS_DNA_Extraction_2.2013.pdf, accessed on 1 February 2021), and WGS was performed as described by Bleriot et al. (20).
Phage genome annotation. (i) Defense mechanisms. All assemblies were initially annotated by sequence homology using Patric 3.6.9 (http://patricbrc.org, accessed on 22 February 2021) and were then manually refined using BLASTX (http://blast.ncbi.nlm.nih.gov, accessed between August and October 2021) and Hhmer (http://hmmer.org, accessed between August and October 2021), as well as the Hhpred tool (https:// toolkit.tuebingen.mpg.de/tools/hhpred, accessed between August and October 2021), which predict functions through protein structure. In addition, to search for phage defense mechanisms against bacteria, the CRISPR Miner 2 (http://www.microbiome-bigdata.com/CRISPRminer/, accessed in March 2022) and PADLOC (https:// padloc.otago.ac.nz/padloc/, accessed in March 2022) tools were used to search for possible CRISPR-Cas systems, as well as the AcrDB tool (https://bcb.unl.edu/AcrFinder/, accessed on October 2021) to search for possible anti-CRIPSR-cas systems with the defect parameter of the website (Aca E value, 0.01; Aca identity, 30%; Aca coverage, 0.8; maximum intergenic distance between genes [bp], 150; operon up/downstream range for MGE-Prophage search [no. of genes], 10). Finally, the family and genus of the different phages were determined by sequence homology with the phage sequences available in the NCBI database. Complete genome sequences were included in the GenBank BioProject PRJNA739095.
(ii) Phage phylogenetic analysis and genome comparison. Phylogenetic analysis of the 21 phages was performed using the nucleotide sequence of the large terminase subunit of each phage. Alignment was first performed with MAFF server (https://mafft.cbrc.jp/alignment/server/index.html, accessed on 3 January 2022), and a phylogenetic tree was then constructed using RAxMLHPC-PTHREADS-AVX2 v8.2.12 (42) under the GTRGAMMA model and 100 bootstrap replicates. A graphical representation of the comparison of all phage genomes was then constructed with the VipTree website (https://www.genome.jp/ viptree/, accessed in June 2022) according to the previously established phylogenetic relationship.
Host range assay. The phage host spectrum was tested by the spot test technique (72), in a collection of 47 clinical strains of K. pneumoniae. A negative control consisting of SM buffer (0.1 M NaCl, 10 mM MgSO 4 , 20 mM Tris-HCl; pH 7.5) was included in each plate. All determinations were made in triplicate. The criteria used to determine the phage infectivity were the presence of clear spots (infection), the presence of turbid spots (low infection or resistance), and the lack of spots (no infection).
Characterization of phages vB_KpnS-VAC35 and vB_KpnM-VAC36. (i) Phage adsorption. Adsorption of phages vB_KpnS-VAC35 and vB_KpnM-VAC36 to the bacterial surface receptors of clinical strains K3574 and K3320, respectively, was determined from the adsorption curve (73) at an MOI of 0.01. The number of phages mixed with bacterial host cells at time zero was considered 100% free of phages. The adsorption curve analysis was performed in triplicate.
(ii) One-step growth curve assay. A one-step growth curve of phages vB_KpnS-VAC35 and vB_KpnM-VAC36 at an MOI of 0.01, was constructed in clinical strains K3574 and K3320, respectively, to determine the latent period (L) and the burst size (B), according to the procedure of Bleriot et al. (20). One-step growth curve analysis was performed in triplicate.
(iii) Phage kill curve assay in liquid medium. Killing curves were constructed from the selected isolates K3574 and K3320, in accordance with the presence of intact CRISPR-Cas system in their genome. Phages vB_KpnS-VAC35 and vB_KpnM-VAC36 were used to monitor the infection of the strain by optical density measured at a wavelength of 600 nm (OD 600 ) and counts of CFU/mL and PFU/mL. For this purpose, when the strains reached an early logarithmic phase (OD 600 = 0.3 to 0.4), cultures were infected with phages at an MOI of 1. The OD 600 , the CFU/mL, and the PFU/mL were determined every 30 min for 3 h. In all cases, the control was the strain without phage infection. All analyses were performed in triplicate.
NanoUHPLC-Tims-QTOF proteomic analysis: interaction between phages (vB_KpnS-VAC35 and vB_KpnM-VAC36) and clinical strains (K3574 and K3320). NanoUHPLC-Tims-QTOF analysis was performed for quantitative study of the protein profile of strain K3574 and K3320 with or without infection with phages vB_KpnS-VAC35 and vB_KpnM-VAC36. When the cultures reached an early logarithmic phase of growth (OD 600 = 0.3 to 0.4), they were infected with phages at an MOI of 1. After 1 h, the cultures of the strains were harvested by centrifugation at 4,302 Â g for 20 min at 4°C. The pellets were then stored at 280°C to facilitate cell disruption. The next day, the pellet was resuspended in phosphate-buffered saline and sonicated. Finally, the sonicated pellets were centrifuged at 4,302 Â g for 20 min at 4°C, and the flowthrough was analyzed using NanoUHPLC-Tims-QTOF. The equipment, as well as the procedure used, are as described by Bleriot et al. (20).
Data availability. Phage genome sequencing revealed all phages under study (available from GenBank BioProject PRJNA739095).