Unraveling the Coevolutionary Dynamics of Phage and Bacterial Protein Warfare Occurring in the Drains of Beef-Processing Plants

TL;DR

This study explores the ongoing battle between phages and bacteria in beef-processing plant drains, revealing a coevolutionary arms race involving attack and defense mechanisms.

Contribution

The study provides preliminary evidence of coevolutionary dynamics between phages and bacteria in beef-processing plant drains, including phage counterattacks against bacterial defenses.

Findings

Phages targeting Pseudomonas, Klebsiella, and Enterococcus were identified in drain samples.

Phage contigs contained infective and lysis-related genes, while bacterial contigs encoded CRISPR-Cas and other antiphage defense systems.

Anti-CRISPR proteins in phages suggest a counterattack strategy against bacterial defenses.

Abstract

Phages, the most abundant entities on Earth, exhibit a complex interplay with bacteria, especially within environmental biofilms, resulting in an ecological arms race. This study investigates the interaction between phages and bacteria in the drains of beef-processing plants using high-throughput sequencing and metagenomic analysis. Metagenomic data collected from 75 drain samples from beef-processing plants were analyzed to investigate phage–bacterial interactions. First, assembled contigs were screened to identify viral sequences, which were then taxonomically annotated to determine the viral composition, including phages. Functional annotation of these viral sequences provided information about the viral genes and their roles in bacterial interactions specifically associated with attack and counterattack of bacteria. In parallel, bacterial contigs were examined to identify genes associated with antiphage defense systems, providing insights into the strategies adapted by bacteria to resist phage infection. Taxonomic annotation of viral sequences from the bulk metagenomic data revealed the presence of phages targeting Pseudomonas, Klebsiella, and Enterococcus. The higher abundance of Pseudomonas phages aligns with our previous study, where Pseudomonas was identified as the dominant bacterial genus, suggesting potential copersistence of phages and their hosts. Functional annotation of phage contigs revealed infective and lysis-related genes, highlighting their potential role in bacterial attack. Conversely, bacterial contigs encoded antiphage defense systems, including CRISPR-Cas, restriction–modification, and other defense-related genes. The study also uncovered the presence of anti-CRISPR proteins in phages, suggesting a counterattack on the bacterial defense. These findings provide evidence for phage attack, bacterial defense, and phage counterattack and may showcase the ongoing coevolutionary arms race between phages and bacteria. While this evidence looks promising, these results remain preliminary and further studies are needed to validate these findings. Still, this study provides a foundational understanding of bacteria–phage coexistence in beef-processing plant drains and paves the way for further explorations of these intricate interactions and their possible applications in controlling pathogenic microorganisms within biofilms.

Figures

Full text

1. Introduction

Global estimates suggest ~10^31^ virus particles exist on Earth, the majority of which are bacteriophages due to the ubiquity of bacterial hosts [1,2]. The interactions between bacteria and phages are complex, with them locked in ecological warfare called the coevolutionary arms race [3]. Phages can follow a lytic cycle, where they can use host cell machinery to replicate, eventually rupturing the bacterial cell [4]. Alternatively, phages can follow a lysogenic cycle [5] by integrating into the host genome and replicating with the host.

Both phages and bacteria constantly evolve their protein machinery for their respective invasion and defense strategies. Phages recognize bacterial receptors and utilize them to enter bacterial cells to commence the infection process [6]. Some of the ways bacteria defend themselves from phage infection are receptor modification, abortive infection, and the CRISPR-Cas system [3,7,8]. Phages, on the other hand, can overcome bacterial defenses through counterdefense strategies such as anti-CRISPR genes [9,10].

High-throughput sequencing has provided us with the advantages of sequencing with high speed and efficiency, providing a substantial volume of high-quality genomic data [11]. Shotgun metagenomics offers the benefit of accessing the complete genetic material of a sample, providing insights into both desired and undesired targets [12]. However, it is important to note that DNA extraction protocols involving centrifugation or filtration may reduce the recovery of viral particles, potentially underrepresenting the viral diversity, hence low-speed centrifugations (~5000× g) are usually used to recover virus from the supernatant [13].

In this study, the phages present in a built environment i.e., the drains of beef-processing plants, were investigated using metagenomic sequence data. Floor drains in food-processing facilities harbor a diverse array of microorganisms, including pathogens, and serve as critical niches for contamination through aerosols and spray water, particularly during sanitation practices [14,15]. Rinse water from animal carcasses, equipment and other food-contact surfaces accumulates in these drains, making them representative of the entire processing environment. Bacteria in these drains can potentially form biofilms that may shelter pathogens and spoilage organisms [15,16]. Therefore, attention to these floor drains is essential due to their direct implications for food safety and quality. Effective strategies to control biofilm formation are crucial [17,18]. While antimicrobial agents are most commonly used to control biofilms, utilizing natural mitigation strategies would be safer as this would reduce the risk of antimicrobial resistance [19]. Phages, with their natural ability to target bacteria, offer a promising approach to biofilm control.

Understanding the co-occurrence between phages and bacteria is a crucial first step toward developing intervention strategies. While bacterial–phage interaction dynamics have been extensively studied in other environments, studies focused specifically on drain-associated biofilms in beef-processing facilities remain limited. This study aims to fill that gap by exploring the co-occurrence between phages and bacteria and look for evidence for the ongoing coevolutionary arms race between them. To explore this, we used metagenomic data to identify genes associated with phage attack, bacterial defense and phage counterattack. This study seeks to provide foundational insights into the bacterial–phage co-occurrence in food-processing environments which would inform us to develop control strategies to eliminate pathogens, ensuring food safety and public health.

2. Materials and Methods

2.1. Beef-Processing Plants, Sample Collection, DNA Extraction and Sequencing

Metagenomic data generated in our previous study [16] were used in the current analysis to investigate the presence and diversity of viral sequences in the drains of beef-processing plants. The DNA extraction protocol and sequencing procedures were originally optimized for bacterial community analysis, particularly to study biofilm-forming ability in these environments. As a result, the protocol may have favored bacterial and intracellular DNA (e.g., prophages) and underrepresented free viral particles [12].

In brief, 75 samples were collected from drains at three beef-processing plants: Plant A (n = 23), Plant B (n = 23), and Plant C (n = 29). The samples were taken from five different locations within each plant, including the hot scale (where pre-rigor carcasses are weighed before chilling; n = 4), hotbox (where carcasses are rapidly chilled for about 18 h; n = 17), cooler (where chilled carcasses are graded and sorted; n = 28), processing (where carcasses are fabricated into whole muscle cuts and trimmings; n = 15), and grind room (where ground beef is produced; n = 11). Sample collection was carried out at two distinct time periods: initially during 2017 and 2018, and later in 2021. To ensure balanced representation and minimize sample loss, the 2017 and 2018 collections were grouped together as Time Point 1 (T1), while the 2021 collection was designated as Time Point 2 (T2). Thus, the sampling periods were classified as T1 (2017–2018) and T2 (2021). The beef-processing facilities involved were geographically distant, with Plants A and B located approximately 200 miles apart, Plants B and C separated by about 400 miles, and Plants A and C around 600 miles apart. No permits were required for sample collection because all sampling occurred within commercial, USDA-inspected beef-processing facilities. Access to field sites and collection of drain samples were conducted under a material transfer agreement (MTA) and non-disclosure agreement (NDA) with cooperating facilities. The research protocol was reviewed and approved by the United States Department of Agriculture—Agricultural Research Service (USDA-ARS) and the Institutional Review Board at the U.S. Meat Animal Research Center (USMARC-IRB-12.2).

All beef-processing facilities performed routine nightly sanitization and cleaning procedures. To capture microorganisms present in the processing environment and those actively transported to floor drains during operations, samples were collected approximately 3–4 h after daily processing activities. Sanitization practices varied among plants, including differences in the types of sanitizers and cleaning agents used, wash water temperatures, and application procedures. In addition, sanitizers were rotated periodically at each facility, so samples were collected under differing sanitization conditions and operational pressures.

Floor drain samples were obtained using cellulose sponges moistened with buffered peptone water [20]. Following centrifugation at 10,000× g, the pelleted cells were resuspended in bead suspension buffer and homogenized using a bead beater (Fastprep96—MP Biomedicals, Irvine, CA, USA). The homogenized suspension was then split into two aliquots for DNA extraction using the DNeasy Powerlyzer PowerSoil kit (Qiagen, Germantown, MD, USA). DNA concentrations were measured with a Qubit 4 Fluorometer (Thermo Fisher Scientific, Waltham, MA, USA). For sequencing, 100 ng of DNA per sample was processed with the Illumina DNA Prep Kit and Nextera™ DNA Indexes (Illumina, San Diego, CA, USA). Library fragment sizes were assessed using the Agilent Bioanalyzer 2100 (Agilent Technologies, Santa Clara, CA, USA), and sequencing was performed on the Illumina MiSeq platform. The resulting FASTQ files were subsequently used for downstream analyses.

2.2. Bioinformatic Analysis

This study used metagenomic assemblies generated in our previous study [16] for all downstream analyses. Two assembly strategies were used depending on the analysis: (i) a single pooled assembly (SPA) and (ii) plant/time point assemblies (PTAs). The SPA was generated by concatenating quality-controlled reads from all 75 drain samples. In contrast, PTAs consisted of separate pooled assemblies for each beef-processing plant at each sampling time point (Plant A-T1, Plant A-T2, Plant B-T1, Plant B-T2, Plant C-T1, and Plant C-T2). Details of the assembly process are described in our previous study [16].

Virsorter2 v2.2.3 [21] was applied to the SPA using default parameters to identify potential viral sequences in the assembly. The output FASTA from Virsorter2 was mapped onto the NCBI Viral Database https://ftp.ncbi.nlm.nih.gov/refseq/release/viral/viral.1.1.genomic.fna.gz (accessed on 23 September 2024) using the BLASTN 2.16.0+ program with the following parameters: %identity = 70%; e-value ≤ 0.005. Viral abundance was calculated by the following steps, (a) the output FASTA from Virsorter2 was indexed using the bwa index command; (b) quality-controlled FASTQ files from all samples were concatenated into a single FASTQ file and mapped onto the created index using the BWA-MEM algorithm; (c) the resulting sorted BAM file was used as an input to the samtools idxstats command which provides the number of reads mapped onto each viral contig; (d) percent relative abundance was calculated by dividing the total number of mapped reads to each individual contig read. Shannon diversity index was calculated using the vegan package in R [22].

Gene calling was performed on the viral contigs using Prodigal v2.6.3 [23] to predict protein sequences. Annotation for the viral protein sequences were carried out using eggNOG mapper v2.1.8 [24] and the PHROGs database [25]. Bacterial defense systems were identified using DefenseFinder v1.2.2 [26,27] on the PTA. The SPA was not suitable for efficient processing with DefenseFinder due to its large size. Although analyzed separately, PTAs still offer meaningful insights into the diversity of bacterial defense systems present across sampling conditions. Clustered regularly interspaced short palindromic repeats (CRISPR) present in the bacterial contigs were detected on the SPA using PILER-CR [28].

Anti-CRISPR proteins in the viral contigs were identified by performing a BLASTN search against the Anti-CRISPRdb V2.2 with the following parameters: %identity ≥ 80%; e-value ≤ 0.001 [29]. Polysaccharide depolymerases were identified by creating a custom BLAST database by downloading all the viral ‘depolymerase’ sequences from the NCBI search. Viral proteins were mapped to the custom database using a ‘blastp’ search and the results were filtered using the following parameters: %identity ≥ 40%; e-value < 0.001; query length and bitscore > 0.7 [30]. Viral genome binning was carried out using vRhyme v1.1.0 using the VirSorter2-derived viral contigs [31]. CheckV v1.0.1 [32] was used to check the quality of the generated bins. Phage genome annotation was carried out using Pharokka v1.3.2 [33]. Restriction sites for the selected high-quality viral bin were identified and the circular genome map was constructed using SnapGene software version 8.1 (www.snapgene.com). Selected figures were created and statistical analyses were performed using RStudio (version 2024.12.1, Build 563).

3. Results

3.1. Sample Information, Sequence Data, Assembly Statistics and Gene Prediction

Details regarding the sample information, sequence data, and bacterial assembly statistics were documented in our earlier study [16]. In summary, we collected a total of 75 drain samples, which provided approximately 55.9 million 250 bp reads (~14 Gb), with an average of 746 K reads per sample. Following the elimination of host-related reads (Bos taurus), we retained 19.3 million potential microbial reads (~4.83 Gb) for further downstream analyses.

For the viral assembly, we obtained a total of 12,003 contigs, with an N50 value of 1544 bp. The largest contig was 72,996 bp and the total assembly length was 15,771,757 bp (Table 1). Gene prediction was performed on the viral contigs predicting 21,452 protein-coding genes. Of these, 3.43% of genes were ≤50 amino acids in length. The average gene length was 164 amino acids with the longest predicted gene spanning 3387 amino acids.

3.2. Detection of Phages in Beef-Processing Drains

VirSorter2 identified a total of 12,003 putative viral contigs. Subsequent taxonomic annotation retained 355 high-confidence viral contigs, of which 334 were identified as bacteriophages. Remapping reads to these phage contigs resulted in 122,211 mapped reads (~30.55 Mb), representing ~0.63% of total microbial reads (Figure 1). The calculated Shannon diversity index for this dataset was 0.744.

Taxonomic annotation using BLASTN identified contigs with closest similarity to bacteriophages associated with bacterial genera such as Pseudomonas, Klebsiella, Erwinia, Enterococcus, and Escherichia. All retained BLASTN matches had e-values ≤ 2 × 10^−4^, with the majority of hits showing substantially lower e-values. Reads were subsequently remapped to viral contigs to estimate phage abundance and sequencing depth. Sequencing depth was highly variable across viral contigs, ranging from 1.4× to 1879×, with a mean depth of 17.6× and a median depth of 7.9×. For abundance analysis, phage contigs associated with the same bacterial genus were grouped and their mapped reads were aggregated. Pseudomonas-associated phages demonstrated the highest abundance, with 90,996 reads (74.45% of the total) mapped onto them. Among the Pseudomonas phages, Pseudomonas phage PMBT3 exhibited the most pronounced dominance, accounting for 61,208 reads (50.08% of the total). This finding aligned with our previous study [16] that identified Pseudomonas as the most dominant bacterial genus in the beef-processing drains.

In contrast, the number of reads mapping to the second most dominant phages, Klebsiella-associated phages, was significantly lower, with only 3638 reads, making up merely 2.97% of the total phage profile. Following Klebsiella phages, Erwinia phages were observed as the third most abundant, with 2202 reads (1.8%) mapped onto them. Several other phages exhibited more than 1000 reads mapped onto them, including Pseudoalteromonas (n = 1974), Xanthomonas (n = 1596), Enterococcus (n = 1430), and Ralstonia (n = 1059). Additionally, certain phages were found to be less abundant, with read counts ranging from 100 to 1000, such as Acinetobacter (n = 501), Caulobacter (n = 452), Delftia (n = 346), and Yersinia (n = 138). Lastly, Escherichia phages were the least abundant, with only 68 reads mapped onto them along with other phages with reads even less in number, for example, Lactococcus phages (n = 11) which were collectively represented in the ‘Others’ category in the taxonomy plot (Figure 1) and, due to their limited read counts, their presence could be considered uncertain.

Viral genome binning was performed using vRhyme on the VirSorter2-derived output containing 12,003 contigs, resulting in a total of 258 viral bins (File S6). Quality assessment using CheckV (File S7) identified a high-quality viral bin (Bin 13) corresponding to Pseudomonas phage phCDa. Although Bin 13 comprised three contigs, CheckV analysis indicated that a single contig (72,996 bp) represented a near-complete, high-quality phage genome (Table 2). Read remapping showed that 8592 reads aligned to this contig, corresponding to an estimated sequencing depth of approximately 30x. This contig encoded 104 predicted genes, including 46 viral genes, and no host-associated genes were detected, further supporting its classification as a phage genome. CheckV revealed a completeness of 99.18% and 0% contamination for this contig, meeting the criteria for a high-quality genome under both CheckV and Minimum Information about an Uncultivated Virus Genome (MIUViG) standards [32,34].

3.3. Phage Structure, Interaction with Bacteria and Lysis

To investigate the mechanism by which phages infect bacteria, we annotated the viral contigs to identify potential infective genes. The primary objective was to locate phage structural proteins, which were successfully identified, encompassing proteins associated with head, neck, tail, tail fiber, tail sheath, portal, and capsid components (Figure 2A).

For the phage to enter a bacterial cell, the bacterial cell membrane needs to possess receptors to which the phages can attach using their tail fibers [6]. In our study, we identified a total of 401 genes associated with the outer membrane gene omp, 40 bacterial genes associated with the outer membrane gene ompA, along with 11 genes related to porin ompC and 12 genes associated with porin ompF. Additionally, we detected 55 genes relevant to the lamB receptor protein and 47 fhuA genes (Figure 2B, File S1). Furthermore, we also detected eight phage receptor binding protein (RBP) genes, which play a crucial role in the specific interaction between phages and the bacterial cell surface receptors. Upon entering the bacterial cell, phages have the option to proceed through either a lytic or a lysogenic cycle [35]. Although we identified the presence of DNA integration proteins that could potentially assist in incorporating phage DNA into the bacterial genome, no indications of prophages were determined within the dataset. After phages have undergone replication within the host bacteria, they employ specific proteins to rupture and exit the cell. The search for genes responsible for this lysis unveiled several genes associated with bacterial cell and biofilm breakdown, including eight lysins, five holins, and seven hydrolases (Figure 2A, File S5). The presence of these lysis proteins provides reasonable evidence that the phages might be undergoing lytic cycles. Another group of enzymes identified were polysaccharide depolymerases which degrade the surface polysaccharides of bacteria during phage attacks and may contribute to the inhibition of biofilm formation. Taxonomic annotations of these enzyme-coding sequences pointed to likely phage sources. For instance, polysaccharide depolymerases were predominantly encoded by phages infecting Acinetobacter (4 of 6 genes), while holins, hydrolases and endolysins were mainly associated with Pseudomonas phages (14 of 20 genes), along with contributions from phages targeting Burkholderia, Xanthomonas and Aeromonas.

To further support our findings, we identified the arrangement of phage structural proteins, including lysis enzymes like holin and Rz-like spanin, on the assembled Pseudomonas phage phCDa genome, which we obtained through viral binning (Figure 3).

3.4. Bacterial Defense Against Phage Attack

In response to phage attacks, bacteria have evolved various defense mechanisms, including CRISPR defense, restriction modification systems, abortive infection, toxin–antitoxin systems, and several other systems [36]. Using DefenseFinder v1.2.2, we identified bacterial defense systems from assembled microbial (bacterial) contigs across all beef-processing drain assemblies. Analysis of bacterial defense systems across individual assemblies revealed distinct profiles among beef-processing plants and time points. Restriction–modification systems were consistently abundant across all assemblies, encompassing types I, II, III and IV restriction endonucleases and methyltransferases (Figure 4, File S2), with the highest counts observed in Plant A-T2 and Plant C-T2. The ‘Others’ category, which includes less abundant systems, also contributed substantially to total defense gene counts. Notably, CRISPR-Cas systems were most prominent in Plant A-T2, while abortive infection systems were detected across all plants, with higher counts in Plant A-T2 and Plant C-T2. RosmerTA systems were detected at lower frequencies but appeared in multiple assemblies, suggesting their broader distribution. Although these variations appeared substantial, there were no statistically significant differences among the assemblies for any of the defense systems (p = 0.42; Kruskal–Wallis test). In parallel, separate analysis of the Pseudomonas phage phCDa revealed the presence of multiple restriction enzyme recognition sites in the genome, specifically 71 cleavage sites of type II restriction enzymes (Figure 3). To further investigate the presence of CRISPR, we employed the PILER-CR program, which successfully identified 32 putative CRISPR arrays in the pooled assembly (File S3). Furthermore, the presence of curli fibers (n = 39) in the biofilms (File S4) was identified, which act as a first line of defense against phages [37].

3.5. Phages Counterattack

In this study, we also explored the counterattack mechanisms employed by phages to antagonize the CRISPR-Cas immune system, which involves the presence of anti-CRISPR (Acr) genes. Our study revealed the existence of three anti-CRISPR genes present across the viral contigs potentially originating from the bacteria Lactococcus and Acinetobacter (File S5). The collected evidence (Figure 2A) may reflect ongoing interactions between phages and bacteria within the beef-processing plant drain biofilms, suggesting a potential evolutionary arms race between them (Figure 5).

4. Discussion

Bacterial–phage interactions and coexistence play a pivotal role in ecology and evolution [38]. Phages specifically infect bacterial hosts and harness their cellular machinery to either replicate via the lytic cycle or integrate into the host genome through the lysogenic cycle [35]. To combat phage infections, bacteria have evolved various resistance strategies such as surface modification, restriction–modification, abortive infection and CRISPR-Cas [36]. Recent discoveries highlight that phages can also mount a counterattack on bacterial defenses by utilizing anti-CRISPR genes [39], resulting in a molecular arms race between the two. This study aimed to analyze metagenomic data on samples from the drains of three beef-processing plants to identify genomic evidence of the three mechanisms: phage attack, bacterial defense, and phage counterattack. The results provided insights into the complex interplay between phages and bacteria in this specific environment. This data can support the future development of phage therapy to eliminate pathogens and disrupt biofilm formation in critical environments such as food-processing facilities.

A major limitation of this study was the relatively shallow sequencing depth of the metagenomic data, which constrained high-resolution characterization of phages at the individual sample level. To address this limitation, a single pooled assembly (SPA) was generated by concatenating high-quality reads from all 75 drain samples. While this approach prevents sample-specific inferences, it enables a comprehensive overview of phages present across beef-processing drain environments. Notably, our previous study [16] demonstrated relatively low beta diversity in microbial community composition across drain samples. This indicated that microbial communities were broadly similar across drains and beef-processing plants despite differences in operational procedures, sanitization practices, and sanitizer rotation among the three facilities. This suggested that pooling samples would not substantially bias viral analyses. In addition, the SPA increased sequencing depth, thereby improving taxonomic annotation and mitigating issues of data sparsity and underrepresentation in individual datasets [40].

Phages are ubiquitous in both natural and man-made environments, so their presence in the drains of beef-processing facilities should not come as a surprise. In our previous study [16], we demonstrated the ability of the microbiome in these drains to form biofilms. A biofilm is an aggregation of microbial cells which are attached to any surface enveloped by a self-produced extracellular polymeric substance (EPS) matrix. Additionally, we also discovered phage sequences in the drain biofilms using metagenomic data. This leads to a question about how bacteria and phages coexist within a biofilm environment. While it is common for phages to be present in biofilms, we still lack a clear understanding of the mechanisms governing their interactions.

In the current study, phage diversity was relatively low (Shannon index = 0.74), indicating dominance of a phage targeting a single bacterial genus. This observation was supported by the high relative abundance of Pseudomonas-associated phages detected in our study. Consistent with this pattern, viral genome binning yielded a single high-quality phage genome, Pseudomonas phage phCDa, which represented the most complete and abundant phage genome recovered from the dataset. Combined with the previously reported [16] dominance of the bacterial genus Pseudomonas, this might suggest a phage–host interaction dynamic within this environment. Notably, Pseudomonas-targeting phages, particularly those active against the multi-drug-resistant P. aeruginosa, have been widely studied for their antimicrobial potential [41]. Although our study does not directly explore phage therapy, these findings emphasize the importance of identifying dominant bacterial populations present in polyextremophilic environments like biofilms, which could guide more precise and effective phage-targeting strategies.

Research indicates that bacteria employ a resilient amyloid fiber called curli in the extracellular matrix, which serves to block phages from entering bacterial cells and penetrating the biofilm [37]. Curli proteins are mainly produced in the upper layer of biofilms and can fill the spaces between cells, effectively hindering phage entry into the biofilm [42]. In a study by Vidakovic et al. [43], the importance of curli was highlighted by creating various E. coli mutants, each lacking a specific biofilm matrix component such as flagellar filaments, cellulose, poly-β-1,6-N-acetyl-D-glucosamine (PGA), colonic acid, type I fimbriae, and finally curli fibers. The results demonstrated that mutants lacking curli fibers could not prevent phage entry into the biofilm. Most of the curli proteins identified in our study were possessed by Gammaproteobacteria which encompass the highly abundant genus Pseudomonas, thus suggesting a potential role of Pseudomonas in acting as a first line of defense against phage attacks by blocking their entry into the biofilm. This finding is consistent with previous studies demonstrating the ability of Pseudomonas to produce functional amyloid fibrils [44].

On the contrary, phages have the potential to break down the EPS matrix by utilizing enzymes known as polysaccharide depolymerases, which manifest as tail spike proteins. These enzymes operate by precisely cleaving the polysaccharide repeat units subsequent to attaching to the polysaccharides [45]. A recent study by Wu Y et al. [46] exhibited the capacity of a novel polysaccharide depolymerase (Dep42), encoded by phage SH-KP152226, to cause lysis in Klebsiella pneumoniae. Not only do polysaccharide depolymerases break down biofilms, they also have the capability to degrade the surface polysaccharides of bacteria, including pathogenic Shiga-toxin-producing E. coli [47,48]. Polysaccharide depolymerases identified in this study predominantly belonged to Acinetobacter phages, which are documented to possess polysaccharide depolymerases and are commonly used as a strategy to combat Acinetobacter baumannii [49]. In our previous metagenomic study [16] we also revealed a high abundance of A. baumannii within a cooler of a beef-processing facility. While we do not completely understand this finding given the limited number of examples, this could potentially function as a mechanism by which phages defend against bacterial protection provided by curli fibers.

When a phage reaches a bacterial cell, it initiates infection by adhering to the surface receptor and injecting its genetic material. Extensive research spanning several decades has been dedicated to investigating bacterial receptors for phages. The specificity of how phages attach to the bacterial cell membrane depends on the presence of receptors in the bacterial cell itself [50]. In Gram-negative bacteria, outer membrane proteins, especially ompA, play a key role as the preferred receptor for phage entry. Studies have shown that ompA and ompC proteins are crucial for phage penetration [51,52]. Porins, another class of outer membrane proteins, are utilized by phages as entry points into the cell. For instance, phage GH-K3 targets Klebsiella pneumoniae via ompC [53], while ompF is a receptor for T4-like phages [54]. Additionally, the lamB receptor protein facilitates the entry of lambda phages during the initial attachment stage [55]. In our study, the detection of receptor genes such as ompA, ompC, and lamB supports the central role of these receptor proteins in phage–host interactions. The presence of these receptors in these environments may suggest prior or ongoing phage–bacterial interactions, including the potential integration of prophages. However, no intact or high-confidence prophages were detected in our dataset, likely due to the relatively shallow sequencing depth and fragmentation of the metagenomic assemblies. Additionally, it is important to acknowledge that bacteria possess the capability to hinder the attachment of phages to surface receptors. This can occur through mutations in the receptor genes or by employing physical obstacles like capsules or extracellular polymers to mask the receptors [3,7]. However, these factors were not within the focus of this study and therefore remain undetermined.

Once a phage successfully enters the bacterial cell, it follows the lytic cycle to generate multiple copies of itself. The completion of the lytic cycle results in the rupture of the cell and its membrane, allowing the replicated phages to exit the cell. This process is facilitated by two proteins, namely endolysin and holin. Endolysins typically build up in the cytosol while holins simultaneously accumulate in the membrane until the membrane becomes permeable to endolysin, ultimately causing the cell to burst [56,57]. In our study, we identified multiple genes encoding these lytic enzymes, including endolysins and holins, across several phage contigs, indicating the potential for active lytic infections in beef-processing drain biofilms. Notably, the high-quality genome of Pseudomonas phage phCDa contained a lysis module, including an Rz-like spanin, a protein known to be essential for outer membrane disruption during the final stages of phage-mediated lysis in Gram-negative hosts [58]. Therefore, the existence of all the lysis proteins, including the presence in phCDa, offers compelling evidence of the possible occurrence of a phage lytic cycle in the drainage systems of beef-processing plants. Moreover, the effectiveness of phage-derived lysins as interventions to eliminate pathogenic organisms has been recognized. A study by Nelson et al. [59] used a purified streptococcal phage C_1_ lysin in vitro and successfully demonstrated the elimination of pathogenic Streptococci in a murine respiratory model. Similarly, a combination of the Listeria phage endolysin PlyP100 along with the antibacterial peptide nisin effectively reduced Listeria monocytogenes in queso fresco (QF), a fresh cheese product [60]. These findings highlight the promise of using purified lysins as targeted antimicrobial agents. However, the application of such treatments in polyextremophilic environments, such as biofilms, presents additional challenges due to their complex structure and resistance to penetration. Further research is needed to optimize lysin-based interventions for effective use in these settings.

Restriction–modification (RM) is an innate immune strategy that bacteria utilize to shield themselves from the invasion of foreign DNA, including viral infections. This system employs a restriction endonuclease to find and cleave recognition sites within foreign DNA, while a methyltransferase protects bacterial cells by adding methyl groups [61]. The effectiveness of this defense depends on the presence of recognition sites in viral genomes. Our study did confirm the existence of RM systems in high abundance across the SPA, as well as the presence of restriction sites within Pseudomonas phage phCDa, suggesting potential vulnerability to host restriction–modification systems. This higher abundance is not surprising as RM systems are the most fundamental and primitive bacterial immune systems and are highly ubiquitous across bacteria [62]. However, it is important to note that the mere coexistence of these evidential factors does not provide conclusive insights into whether or not the viral infection or the bacterial defense is successful. Besides restriction–modification, evidence for several other bacterial defense systems was identified through our investigation, such as abortive infection (Abi) [8], toxin/antitoxin (TA) system [7] and cyclic oligonucleotide-based antiphage signaling system (CBASS) [63], each following their own mechanism to evade phages. For instance, in the Abi system, the bacterial cell self-destructs to prevent the release of phage virions [8]. Gene hits corresponding to these diverse bacterial defense systems were consistently detected across all PTAs, revealing the multitude of defense capabilities held by the microbial community in the drains of beef-processing facilities. Our observations were based on coassemblies containing contigs from mixed microbial populations, so the presence of diverse defense systems may not be unexpected. However, individual bacterial genomes can also harbor multiple defense mechanisms. For example, Pseudomonas aeruginosa has been reported to encode up to 19 distinct phage defense systems [64].

In addition to the defense mechanisms noted previously, a particularly important one is the discovery of the CRISPR-Cas system, an adaptive immune system that gives bacteria resistance to phages and other invaders [65]. Previous studies estimate that approximately 40% of bacterial genomes carry CRISPR-Cas systems [66], although this percentage can vary across environments depending on phage exposure and selective pressures. For example, a study of groundwater microbial communities found that only 10% of 1724 genomes contained CRISPR-Cas elements [67]. In our study, because the analysis was performed on metagenomic assemblies representing multiple microorganisms rather than individual genomes, it was not possible to directly compare our findings to genome-level statistics. However, this provides us clues about the potential presence of CRISPR-Cas systems in the biofilm communities of beef-processing plant drains.

Interestingly, even with all the defense systems present in the bacteria to resist phage infection, phages can still counterattack to tackle the bacterial defense systems. Anti-CRISPR (Acr) protein is one such counterattack mechanism which phages use to evade the bacterial defense. To date, more than 70 Acr genes have been identified [68], including four Acr genes against the type II CRISPR-Cas system of Listeria monocytogenes, carried by their respective prophages [69]. L. monocytogenes is a major food-borne pathogen found in many foods, including beef, and is one of the leading causes of death from food poisoning in the United States [70]. Although no evidence of Listeria phages was found in our study, the known presence of Acr genes targeting such pathogens is promising for the development of future targeted therapeutic interventions [71]. Furthermore, all Acr genes identified in our study showed sequence homology to bacterial genera such as Lactococcus and Acinetobacter [72]. While this may suggest that these genes may have bacterial origin, further validation would be needed to confirm their source.

The findings from this study provide an initial foundation for understanding the dynamic interplay between phages and bacteria in the drains of beef-processing facilities. Current practices to mitigate biofilm formation in these environments largely rely on the use of chemical sanitizers, particularly the most commonly used food-processing facility sanitizers, quaternary ammonium compounds (QACs). However, prolonged and widespread use of such sanitizers gives rise to the emergence of antimicrobial resistance (AMR) genes [73], including against QACs, as observed in our previous study [16]. This growing concern highlights the need for natural and safer alternatives for sanitation strategies.

Phages present a promising, natural alternative due to their specificity in targeting bacterial pathogens. Prior studies have demonstrated the efficacy of phages in reducing biofilm formation by food-borne pathogens such as Salmonella [74] and Listeria monocytogenes [75], with significant reduction in cells following phage treatment. While these results are promising, most existing studies have focused on single-species biofilms under controlled conditions. In contrast, biofilms within food-processing drains are often composed of diverse microbial communities [16], making them more resilient and challenging to eradicate. Therefore, there is a need for comprehensive studies exploring both bacterial and phage populations, particularly in complex polyextremophilic environments like drain-associated biofilms.

Due to the relatively shallow sequencing depth in this study, a pooled coassembly approach was employed for viral contig identification. Attempting to analyze individual samples separately would have resulted in even lower coverage, potentially compromising the reliability of the results. This approach of pooling samples has been employed in previous metagenomic studies, particularly to address the challenge of low viral titers [76]. Although the proportion of phage reads was relatively lower in our dataset (~0.63%), it remains within the same order of magnitude as reported in previous studies, for instance, a metagenomic study of murine gut microbiomes found phage read proportions ranging from 0.93% to 4.24% [77]. While this strategy limits our ability to study spatial or temporal differences among drain samples, the pooled dataset still offers insights into the comprehensive phage composition in beef-processing drains and the molecular mechanisms used by both bacteria and phages to coevolve.

5. Conclusions

The interplay between phages and bacteria has been a subject of extensive research. Specifically, within confined environments like biofilms, the close proximity can result in even more intricate interactions between phages and bacteria. In this study, we leveraged the metagenomic data collected from the drains of three beef-processing plants in a previous study to reveal the presence of a wide array of protein weaponry possessed by both bacteria and phages, underscoring the ongoing battle between these two entities. This research unveiled compelling evidence for all three key processes: phage attacks, bacterial defenses, and phage counterattacks. Although this study is limited by factors such as low sequencing depth, it offers initial clues about the molecular mechanisms of bacterial defense and phage attack. Continued research in this area will be essential for advancing the development of phage-based interventions to improve food safety and sanitation.