During this study, 10 broiler houses which housed a total of 22 flocks were selected and sampled from 2020 to 2021. Poultry farms used in this study were in the North and East of Spain. The procedures carried out in this study were approved by the Ministry of Agriculture of each of the respective communities (based on the CSV: 2AD5M8FG-DALESJC1-7DA55S99). All houses included in this study had persistent Salmonella bacteria even though after their conventional C&D procedure. Furthermore, at the end of each trial, phages were inactivated from farm facilities after application of disinfectants.

All samples were analyzed according to the ISO 6579:2017 (Mooijman et al., 2019) and serotyped following the Kauffman-White-Le Minor technique (Grimont and Weill, 2007). Swab cloths and boot swab samples were diluted 1:10 (w/v) in BPW (Buffered Peptone Water ISO, VWR Chemicals, Barcelona, Spain). Samples were then incubated at 37°C for 18 ± 2 h. The pre-enriched samples were transferred into Modified Semi-Solid Rappaport Vassiliadis agar plate (MSRV, Difco, Valencia, Spain), which was incubated at 41.5 ± 1°C for 24–48 h. Suspicious plates obtained in MSRV were transferred into Xylose–Lysine– Deoxycholate (XLD, Liofilchem, Valencia, Spain), and ASAP (bioMerieux, Madrid, Spain), and then incubated at 37 ± 1°C for 24 h. After the incubation, 5 typical colonies of Salmonella were selected and streaked into nutrient agar plates (Scharlab, Barcelona, Spain) at 37 ± 1°C for 24 ± 3 h (Mooijman et al., 2019). Salmonella strains isolated were serotyped according to the Kauffman-White- Le Minor technique (Grimont and Weill, 2007).

Salmonella target phage strains were characterized according to the ISO 20776–1:2006, by a commercially available microtitre plates Sensititre™ EUVSEC (Thermo Scientific, East Grinstead, United Kingdom). To this end, Salmonella Sensititre Plates (Gram Negative MIC Plate) were used to assess antimicrobial susceptibility of isolated strains. The antibiotics selected were those from the antimicrobial panel determined by the Commission Implementing Decision of November 2020 on the monitoring and reporting of antimicrobial resistance in zoonotic and commensal bacteria (Commission Implementing Decision, 2020). This panel, included, Ciprofloxacin (CIP, 0.015–8 μg/mL) and Nalidixic Acid (NAL, 4–128 μg/mL); 2 B-lactams: Meropenem (MERO, 0.03–16 μg/mL) and Ampicillin (AMP, 1–64 μg/mL), one phenicol: Chloramphenicol (C, 8–128 μg/mL); one pyrimidine: Trimethoprim (TM, μg/mL); one tetracycline: Tetracycline (TET, μg/mL); one macrolide: Azithromycin (AZM, 2–64 μg/mL); one glycylcycline: Tigecycline (TGC, 0.25–8 μg/mL); 2 cephalosporin: Ceftazidime (CAZ, 0.5–8 μg/mL) and Cefotaxime (CTX, 0.25–4 μg/mL); one polymyxin: Colistin (COL, 1–16 μg/mL); one potentiated sulfonamide: Sulfamethoxazole (SMX, 8–1,024 μg/mL), and one aminoglycoside: Gentamicin (GN, 0.5–32 μg/mL). Epidemiological cutoff values (ECOFF) were taken to determine resistance against the antibiotics analyzed. These values were established by the European Committee on Antimicrobial Susceptibility Testing (EUCAST) and recommended by UE Commission Decision 179/2020 (Commission Implementing Decision, 2020). The values not included in this legislation (Azithromycin and Sulfamethoxazole) were assessed following National Committee for clinical Laboratory Standards (NCCLS) criteria (Clinical and Laboratory Standards Institute, 2024). MDR was defined as acquired resistance to at least one agent in 3 or more antimicrobial classes (ECDC, 2016).

Isolation and purification of phages was done using the bacterial host of each selected farm according to Sevilla-Navarro et al. (2020a,b). Phages were isolated from feces collected from the selected farms by an enrichment procedure, with the aim of applying an AP (Sevilla-Navarro et al., 2018). To do so, 10 g of each feces sample from each farm were diluted in 90 mL of Luria Bertani (LB) (VWR Chemicals, Barcelona, Spain) and incubated along with each specific Salmonella serovar overnight at 37°C. After incubation, 2 mL of this enrichment culture was centrifuged 16,000 × g for 5 min. The supernatant was then filtered through a 0.22 μm membrane. Phages were isolated and purified in a spot test by the double agar method. Briefly, bacterial suspensions of each serovar were adjusted to an optical density (OD) 600 nm of 0.2 nm (~108 CFU/mL) in LB and incubated at 37°C for 4 h. Then, 200 μL of cultures were added to 5 mL of LB agar (LB with 0.6% agar) tempered to 45°C and poured onto previously prepared and dried LB basal agar (with 1.6% agar). Then, 10 μL of each filtrate was spotted onto the surfaces of Salmonella lawns and incubated overnight at 37°C. After the incubation, morphologically different plaques were selected and resuspended in 50 μL of PBS. Ten-fold serial dilutions of the phage suspension were plated by the double agar layer method, and phages that produced clear plaques were selected.

To extract DNA, 10 μL was used of highly concentrated lysates for DNAse I treatment to remove bacterial host. Subsequently, capsids were enzymatically digested using Proteinase K, and the resulting phage DNA was purified using the commercial DNA Clean & Concentrator-5 kit (Zymo Research, USA). Library was prepared with the Nextera XT Library prep kit and used for sequencing using Illumina MiSeq technology (250 bp paired-end reads). Quality control of raw data was done with fastp and resulting fastq files were used to generate a de novo assembly with SPAdes version: 3.13.1 (only-assembler mode) (Bankevich et al., 2012). Contigs were analyzed to discard those shorter than 1 Kbp and coming from bacterial host leading to one contig with high length and k-mer coverage that was taken as the phage genome. A BLAST comparison against the nucleotide database was then performed to determine the closest relative of each genome and assign a preliminary taxonomic classification to each phage (Altschul et al., 1990). The beginning of related representative phages was used to reorder the genomes. Afterwards, the rearranged genomes were corrected with Pilon v1.24 and coverage was evaluated by mapping the trimmed reads with BBMap.sh v38.95 (Bushnell, 2014; Walker et al., 2014). For structural and functional annotation, Pharokka v1.4.1 was used with default parameters and each sequenced coding was associated with a Phrog functional group (Bouras et al., 2023). Phage depolymerases were predicted using the machine learning tool DePP (web version 1.0.0) and protein products with at least a 90% of probability of having depolymerase activity were considered potential depolymerases (Magill and Skvortsov, 2023). Phage safety was assessed using Pharokka in combination with PhageLeads, to determine the presence of temperate markers and antimicrobial resistance and virulence genes (Yukgehnaish et al., 2022). Additionally, phage lifestyle was predicted with PhaTYP (web version from Phage BOX) (Shang et al., 2023). Finally, isolated phage genomes were compared by BLAST and represented with gggenomes v1.0.0 (Hackl et al., 2023). To confirm phage taxonomy, two different approaches were used using Refseq sequences from phages belonging to the preliminary assigned taxonomic rank. Firstly, a single-gene phylogeny using the large subunit of terminase protein was performed using ClustalW v2.1 as the aligner and IQTREE v1.6.12 for phylogenomic inference with 1,000 ultrafast bootstrap replicates (Thompson et al., 1994; Trifinopoulos et al., 2016). The maximum likelihood tree was constructed using ITOL v5 fixing the midpoint as the root and removing branches with a value of bootstrap lower than 90 (Letunic and Bork, 2021). Secondly, VIRIDIC v1.1 was used for calculating viral intergenomic similarity and clustering in species and genera (95 and 70% thresholds respectively) (Moraru et al., 2020). Genus’s clustering was represented with a circle packing using RawGraphs v2.0 (Mauri et al., 2017).

Phages were visualized by transmission electron microscopy (TEM) (FEI Tecnai G2 Spirit Biotwin, 120 KW) in the Centro de Investigación Principe Felipe (CIPF, Valencia, Spain). To this end, 10 μL from the AP with a concentration of 10^8-9PFU/mL was fixed in an aqueous solution of paraformaldehyde (2%). A 7.2 V glow was discharged on samples placed on the MESH Cooper grid and incubated in the grids for 15 min. Then, samples were washed in phosphate buffer 0.1 M for 2 min and fixed with glutaraldehyde (1%). Samples were negatively stained with uracil acetate (1%) and incubated with methyl cellulose (1%) for 30 s. Samples were dried until use (Li et al., 2016; Sevilla-Navarro et al., 2018).

A total of 41 Salmonella isolates were tested, with a focus on S. Infantis strains from different years (2013, n = 9; 2018, n = 10; and 2023, n = 10). Additionally, various Salmonella serovars with likely antigenic formulae isolated in 2023 were selected, including S. Ohio, S. Mikawasima, S. Mbandaka, and S. Newport. The assess the bacterial susceptibility to phages, a spot titration protocol was used that allow us to determine both host range and relative EOP (Gibson et al., 2019). To this end, 10 μL of serial 10-fold dilutions of each phage were spotted on freshly seeded Salmonella spp. lawns and incubated overnight at 37°C. After incubation, the host range and titer were determined by formation of individual plaques within the area of the spot at terminal dilution. EOP was calculated by dividing the titer of the phage at the terminal dilution on the challenge strain by the titer of the same phage on its host strain. Phages were classified as “highly virulent” for ratio > 0.5, “medium virulent” for ratios ≥ 0.1 and ≤ 0.5, “avirulent but active” for ratios >0.001 and < 0.1 and “inefficient” for ratios ≤0.001 or no plaques detected.

A total of 500 mL of each bacteriophage (AP) was multiplied in its specific host to produce high-titer lysates (10^^9–10 PFU/mL). For this, 500 mL of LB medium supplemented with 5 mL of each host culture was grown to mid-log phase (OD600 = 0.2) at 37°C. Subsequently, each phage lysate was added at a multiplicity of infection (MOI) of 0.1. The bacterium-phage mixture was then incubated under agitation (180 rpm) for up to 4 h (vB_Si_CECAV_FGS009, vB_Si_CECAV_FGS017) and up to 6 h (vB_Si_CECAV_FGS029, and vB_Si_CECAV_FGS030), until complete lysis of the bacterial culture was observed. Following incubation, lysates were centrifuged at 10,000 × g for 5 min and filtered using 0.2 μm filters (Nalagene, ThermoFisher) to remove any remaining bacterial cells and debris. Subsequently, 10 μL of each fold dilution were spotted onto double agar layers and incubated overnight at 37°C for quantification of phage concentration via plaque assays and to confirm the absence of viable Salmonella cells. Phages were stored at refrigeration temperatures (4°C) until they were used.

All houses included in this study harbored persistent Salmonella bacteria despite undergoing the standard Cleaning and Disinfection (C&D) procedure. Thus, these farms were subjected to a modified C&D protocol which included an additional step which was the application of the AP twice consecutively at 24 h intervals (C&AP²&D). The standard C&D procedure comprised six steps: (i) initial cleaning to remove bedding, dust, and feces from the utensils within the facility (drinkers, feeders, etc.), followed by (ii) thorough rinsing with water, and (iii) Subsequent cleansing with detergent, rinsing to remove any detergent residue, and allowing the surface to dry before applying the disinfectant. A variety of disinfectants were utilized, with rotation based on the specific requirements of each facility. These included combinations of glutaraldehyde, quaternary ammonium compounds, and peroxides. The AP in each step was applied at a concentration of 10⁸PFU/mL. A total volume of 500 mL of AP diluted in 4.5 L of water was sprayed each time. The steps that were carried out were as follows: (i) Cleaning with detergent (removal of organic matter from the houses wet and dry), (ii) first application of the AP via spray, (iii) Second application of AP (24 h after the first application) and the last step, (iv) application of conventional disinfectants (24 h after the second AP application). To determine whether the houses were positive for S. Infantis, prior to the application of the modified procedure, swab samples from the poultry farms were collected according to the NSCP and analyzed following the ISO 6579:2017 (Mooijman et al., 2019; PNCS, 2020). After the application of each step (i. ii. and iii.) of the procedure C&AP²&D, representative samples from the floor, windows, walls, feeders, drinkers, and fans were taken from the houses. To this end, the farm was divided into 3 identical zones and each of these zones was divided into 3 parts (Figure 1). In this way on each of the sampling days, the surface area floor and walls were taken using boot swabs and swab cloths, respectively. In addition, swabs cloths were used to collect samples from the feeders, drinkers, and fans. To collect these samples, each of the surfaces to be sampled was divided into 3 parts, so that different areas were sampled without swabbing three times the same place. All these samples were analysed following the ISO 6579:2012 (ISO 6579:2012, 2012). Finally, after the last step (iv.) of the procedure C&AP²&D, samples were taken from the houses as laid by the NCSP to verify the absence of Salmonella (PNCS, 2020).

A Generalized Linear Model (GLM) was employed due to its flexibility in handling non-normally distributed data and due to the incorporation of multiple variables, providing a robust and flexible approach to assess the efficacy of AP in reducing Salmonella surfaces. To this end, a GLM was performed according to Sevilla-Navarro et al. (2020a). Concentrations (CFU/sample) of Salmonella counts were converted to Log₁₀CFU/sample. Then, to evaluate the impact of the AP² on Salmonella counts following each application, the GLM was fitted with the sampling moment (after cleaning, after 1st AP application, after 2nd AP application) as the response variable and the farm as the fixed effect. Additionally, another GLM was conducted to investigate the effect of each AP application time on Salmonella counts across various sample types, encompassing drinker, feeder, floor, fan, window, and wall. Here, the sampling moment served as the response variable, while the sample type was considered the fixed effect. A significance level of p-value < 0.05 was adopted to denote statistical significance in all analyses. Moreover, antibiotic resistance among Salmonella isolates was compared utilizing a GLM, wherein binomial data were coded as 1 for resistant isolates and 0 for non-resistant ones. The error distribution was specified as binomial, and the probit link function was applied. A significance threshold of p-value < 0.05 was utilized to determine statistical significance. All statistical analyses were performed using SPSS 27.0 software (SPSS Inc., Chicago, IL).

A total of 10 farms were used in this study from which a total of 22 poultry houses were sampled. All Salmonella strains isolated from each farm were tested prior the C&AP²&D procedure. All of them (n = 22) were resistant to at least one of the 14 antimicrobials tested and 72.7 ± 9.7% (16/22) were MDR to three or more of the groups of antimicrobials. The highest percentages of AMR were found to be CIP (95.0 ± 4.4%, 21/22), and NAL (91.0 ± 6.1%, 20/22), followed by TMP and TET (68.0 ± 9.9%, 15/22, each) and SMX (64.0 ± 10.3%, 14/22), and finally AMP (14.0 ± 7.3%, 3/22), TGC (9.0 ± 6.1%, 2/22), and COL, FOT, GEN and TAZ (5.0 ± 4.4%, 1/22, each) (p < 0.05). Resistance to AZI, CHL, and MER was not observed. Overall, 6 different resistance patterns were observed. The combination of QNL-SULF-TE (59.1%, 13/22) was the most frequently observed pattern, followed by QNL alone (13.6%, 3/22), QNL-TE and QNL-B-LAC-SULF-TE (9.1%, 2/22, each), QNL-B-LAC-SULF-POL and B-LAC-AMG (4.5%, 1/22, each).

At each farm, a phage capable of lysing its corresponding propagation strain was isolated. However, after sequencing analysis of the 10 isolated phages, a surprising similarity emerged among them, leading to the characterization of four distinct phages: vB_Si_CECAV_FGS009, vB_Si_CECAV_FGS017, vB_Si_CECAV_FGS029, and vB_Si_CECAV_FGS030.

MiSeq Illumina sequencing run led to 21,250–51,047 cleaned paired end reads that were used for generating the de novo assemblies with SPAdes. After removing bacterial contaminants and short contigs, one long assembly with high coverage was obtained for all the phages. BLAST comparison against the nucleotide database showed vB_Si_CECAV_FGS009, vB_Si_CECAV_FGS017 and vB_Si_CECAV_FGS029 were closed to phages belonging to Tequintavirus genus while vB_Si_CECAV_FGS030 was more related to phages belonging to the genera Felixounavirus. Enterobacteria phage T5 (Accesion Number: NC_005859) and Salmonella phage FelixO1 (Accesion Number: NC_005282.1), type-phages from Tequintavirus and Felixounavirus genus respectively, were used to rearrange the obtained genomes. Reordered assemblies were then corrected with Pilon leading to phage genome sizes bigger than 85 kb, and a GC of content around 39% in all the cases (Table 1).

Functional and structural annotation with Pharokka found at least 200 coding sequences in cecav_Si_CECAV_FGS009, cecav_Si_CECAV_FGS017 and cecav_Si_CECAV_FGS029 while 144 coding sequences were identified in phage cecav_Si_CECAV_FGS030 genome. In the four phages more than 200 tRNA and pseudogenes were also detected. Genomic organization of all phages was like other previously described phages belonging to Tequintavirus and Felixounavirus genus (Nicolas et al., 2023) (Figure 2).

Regarding antibiofilm properties of the phages, genetic markers were found. Pharokka annotation found 2 ORFs encoding a lysin in vB_Si_CECAV_FGS009, vB_Si_CECAV_FGS017 and vB_Si_CECAV_FGS029 genomes, and 1 ORF in vB_Si_CECAV_FGS030. Additionally, DePP results showed phage vB_Si_CECAV_FGS030 encoding one potential depolymerase (ORF present in the positive strand), while three potential depolymerases were found in vB_Si_CECAV_FGS017 and four in vB_Si_CECAV_FGS009 and vB_Si_CECAV_FGS029 (Table 2). All these predicted depolymerases were annotated as tail-related proteins by Pharokka. PhageLeads and PhaTYP analysis showed all four phages are strictly lytic (absence of temperate markers and probability of 1.0) and do not contain any sign of bacterial genome integration (absence of virulence and antimicrobial resistance genes).

Regarding taxonomic classification both whole-genome comparison and single-gene phylogeny using large terminase subunit sequence both analyses grouped vB_Si_CECAV_FGS009, vB_Si_CECAV_FGS017 and vB_Si_CECAV_FGS029 together with other phages belonging to the Tequintavirus genus (Figure 3). Specifically, vB_Si_CECAV_FGS009 and vB_Si_CECAV_FGS029 were more similar (88.1% of intergenomic similarity), while vB_Si_CECAV_FGS017 seemed to be slightly different (81.6 and 82.5% of similarity in comparison with vB_Si_CECAV_FGS009 and vB_Si_CECAV_FGS029, respectively). Interestingly, vB_Si_CECAV_FGS030 was more different in comparison with the other three isolated phages, and closer to phages from Felixounavirus genus (Figure 3).

The average head size of Felixounavirus phage was 80.7 nm, whereas for Tequintavirus phages the average head size ranged from 83.3 to 90.1 nm. The tail size of the Felixounavirus phage was 118.4 nm. However, it was 170.6–198.8 nm for Tequintavirus (Figure 4).

None of the phages demonstrated the ability to lyse serovars other than S. Infantis, except in the case of S. Mikawasima, where all phages, excluding vB_Si_CECAV_FGS030, exhibited high virulence against 100% of the isolates. Concerning S. Infantis from 2023 and 2018, phages vB_Si_CECAV _FGS009, vB_Si_CECAV_FGS018, and vB_Si_CECAV_FGS029 highly displayed virulence against 90% (18/20), while phage vB_Si_CECAV_FGS030 demonstrated highly and moderate virulence against 60 and 40% of isolates, respectively. Notably, there was a discernible difference in the infectivity among isolates from 2013, with all phages only able to infect a single strain (Figure 5).

Concerning the assessment of the cleaning and disinfection procedure with AP, statistically significant differences were found between the different steps of the C&AP²&D procedure (p < 0.001). Before the application of the phages, 100% of the houses tested positive for the presence of S. Infantis. Following the initial phage application, a reduction of 27% was observed, resulting in a positivity rate of 73%. After a second application, there was an additional reduction of 64%, bringing the positivity rate down to 36%. Ultimately, after the disinfection process, all houses tested negative for the presence of S. Infantis. Moreover, a noteworthy decline in bacterial presence was evident between procedure steps, leading to the complete elimination of bacteria when the initial count was below 5 logs. Table 3 illustrates the dynamic progression of Salmonella reduction in houses throughout the distinct stages of the C&AP²&D procedure.

In relation to the type of sample tested for Salmonella presence, notable differences were observed post-cleaning step concerning Salmonella prevalence per sample (p = 0.011). The walls exhibited a higher prevalence of the bacterium (5.61 Log₁₀CFU/sample), followed by the floor (5.22 Log₁₀CFU/sample) and drinkers (3.66 Log₁₀CFU/sample). However, after the 1st and 2nd application of the AP², no significant differences were noted between samples. Conversely, significant variations emerged between the reduction of Salmonella in each sample and the sampling moment, with the most substantial reduction occurring after the second application of the AP, as outlined in Table 4.