The Salmonella ser. Enteritidis strain P125109 genome (NC_011294) was downloaded from the Genbank and visualized using SnapGene software (v6.2.1). Primers and probes were designed using Primer Express™ Software.

The SEN1990 DNA-binding domain was identified using the online tools CDD-NCBI (Lu et al., 2019) by searching the amino acid sequence of SEN1990 (WP_001095779.1). Protein modeling and functional characterization through structure homology were performed using the I-Tasser or D-I-Tasser online server (Yang and Zhang, 2015) and Phyre2 (Kelley et al., 2015) with default settings. The amino acid sequences of the RepL proteins from plasmids pMO1 (NC_002059), pSN2 (NC_005565), pNE131 (NC_001390), pIM13 (NC_001376), and pE5 (M17990.1) were obtained from the original publications for the RepL annotation (Lampson and Parisi, 1986; Hefford et al., 1997). The MarR protein structure was downloaded from the PDB database (Alekshun et al., 2001). Model rendering and structural analysis were performed and processed in Pymol2 (PyMOL v1.2r3pre). Structural alignment was performed using the ColorByRMSD script, and the electrostatic map was calculated through ABPS Electrostatics using the psb2pqr method. The isoelectric point and molecular weight of SEN1990 were calculated in the online server https://web.expasy.org/compute_pi/.

SEN1990 homologs among Enterobacteriaceae-associated ROD21-like (EARL) GIs were identified using tBLASTx through the BLAST web server and Easyfig v2.2.5. The protein alignments were performed through MUSCLE multiple amino acid sequence alignment using MEGA11 software with the standard configuration. The multiple alignment figures were rendered through the ESPript 3.0 web server using the “%MultAlin” sequence similarity depiction. A phylogenic tree was then calculated using the IQTREE webserver (Trifinopoulos et al., 2016) with 1,000 bootstraps. The JTTDCMut+G4 substitution model was chosen according to the Bayesian Information Criterion calculated by ModelFinder (Kalyaanamoorthy et al., 2017). The tree was rendered using the iTOL webserver (Letunic and Bork, 2021).

The expression plasmid pAPI-SEN1990 was constructed using the low copy number plasmid pET-15b as a backbone, which was linearized through PCR using the primers Lineal_pET-Fw and Lineal_pET-Rv (Supplementary Table 1). Then, using the high-fidelity polymerase Platinum SuperFi II (Invitrogen 12,368,010), the IPTG-inducible promoter Trc was amplified from pTrc-HisB, adding overhangs homologous to pET15b upstream to the promoter and to SEN1990 downstream to the promoter. In the same manner, SEN1990 was amplified by adding overhangs homologous to pET15b downstream of the gene and to ptrC upstream of the gene. The fragments were assembled through Gibson Assembly with the NEBuilder HiFi DNA Assembly buffer (NEB E2621). The fragments were mixed in a proportion of 2:3 (fragment: vector), and the reaction was performed at 50°C for 15 min.

Two Salmonella ser. Enteritidis strains were constructed—a deletion mutant lacking SEN1990 (Salmonella ser. Enteritidis ∆SEN1990::frt) and a deletion mutant with an inducible plasmid expressing SEN1990 ectopically (Salmonella ser. Enteritidis ∆SEN1990::frt pAPI-SEN1990).

Salmonella ser. Enteritidis ∆SEN1990::fer. was constructed through the Lambda Red system (Datsenko and Wanner, 2000) using Salmonella ser. Enteritidis PT4 P125109 as the parental strain. Before the elimination of the resistance cassette with plasmid pCP20 (Datsenko and Wanner, 2000), the mutation on the genome of Salmonella ser. Enteritidis ∆SEN1990::cat was transferred to a clean genetic background through general transduction using P22 phage. The absence of lysogens and pseudolysogens was verified by seeding the bacterium in Evans Blue agar and assessing its susceptibility to infection by the P22 H5 lytic phage.

The complemented strain was constructed by electroporating Salmonella ser. Enteritidis ∆SEN1990::frt with 200 ng of the plasmid pAPI-SEN1990. After 3 h of recovery at 37°C in LB, the culture was seeded in LB agar with 100 μg/ml ampicillin and incubated at 37°C for further selection.

GFP-expressing strains were constructed by electroporation of the pKK233.2-gfp plasmid as described above.

All strains constructed were confirmed by PCR (Supplementary Figures 1, 2). Strains were stored in BHI-Glycerol 20% (v/v) at −80°C and cultured in LB broth at 37°C with agitation and with 10 μg/ml chloramphenicol or 100 μg/ml ampicillin when needed.

DNA amplification by PCR was carried out using either the GoTaq G2 Flexi DNA polymerase (Promega M7801) or the Platinum SuperFi II high-fidelity polymerase. The thermocycler protocols were designed according to the polymerase manufacturers using the primers listed in Supplementary Table 1.

Quantitative PCR (qPCR) amplifications were carried out using either the TaqMan™ Fast Advanced Mastermix (Applied Biosystems 4,444,557) or the qPCR SsoAdvanced Universal SYBR green Supermix (Bio-Rad MH5H1EE8Z) in a StepOne Plus real-time thermocycler, using the protocols provided by their respective manufacturers and the primers listed in Supplementary Table 2.

Gene expression, island excision, and ROD21 relative copy number assays were carried out by diluting overnight (O/N) cultures of the Salmonella ser. Enteritidis WT and the the ΔSEN1990::frt mutant strain in 15 ml of LB broth with 0.1 mM L-arabinose to OD₆₀₀ = 0.01 and incubating for 2.5 h at 37°C with shaking, after which samples for DNA and RNA isolation were obtained. RNA was isolated using TRIzol (Invitrogen 15,596,026) according to the manufacturer’s protocol. Residual DNA was removed with the Turbo DNA-Free kit (Invitrogen AM1907), and cDNA was synthesized with the iScript cDNA synthesis kit (Bio-Rad M87EWZESH). The mRNA quantity of oafB was quantified using SsoAdvanced Universal SYBR Green Supermix, and the relative gene expression was calculated with the 2^-(∆∆Ct) equation, using rpoD as the endogenous control and the WT strain as the reference. For the island copy number and excision assays, genomic DNA was extracted using phenol:chloroform:isoamyl alcohol as previously described (Piña-Iturbe et al., 2022). The qPCR amplifications were carried out using TaqMan™ Fast Advanced Master Mix. The copy number of attB and rpoD in each DNA sample was calculated by interpolating the Ct of both amplicons into their corresponding standard curve made of tenth-fold dilutions of Salmonella Typhimurium genomic DNA (10⁸–10² total chromosomes), as it naturally lacks ROD21. The copy number of SEN1998 was calculated using a standard curve of the plasmid pET15b-SEN1998 (10⁸–10² total plasmid copies). Island excision was calculated by the relation: % Excision = 100 × attB/rpoD in each sample. The relative copy number of ROD21 was calculated by the ratio of SEN1998 copies/rpoD copies. The relative number of episomal ROD21 copies was calculated by normalizing the Ct of attP to rpoD within a sample and then to the WT strain (control sample) through the 2^-(∆∆Ct) equation.

The Salmonella ser. Enteritidis WT and ΔSEN1990::frt strains were cultured O/N at 37°C in LB broth with agitation and diluted in LB to OD₆₀₀ = 0.01. The cultures were incubated at 37°C with agitation until OD₆₀₀ = 0.6, after which samples were centrifugated at 8,000 g for 8 min. The supernatant was discarded, and RNA was isolated using TRizol. The samples were further purified from salts using drop dialysis in MF-Millipore™ membranes (0.22 μm pore size; Millipore GSWP02500) against nuclease-free water for 4 h on ice. RNA-seq was performed by Zymo Research as follows: libraries were prepared using the Zymo-Seq RiboFree Total RNA Library Prep Kit and sequenced on an Illumina NovaSeq platform to a sequencing depth of at least 30 million read pairs (150 bp paired-end sequencing) per sample. The bioinformatic analysis was performed as follows: quality control of raw reads was carried out using FastQC v0.11.9. Adapters and low-quality sequences were trimmed from raw reads using Trim Galore! v0.6.6. and the trimmed reads were aligned to the reference genome using STAR v2.6.1d. BAM file filtering, and indexed using SAMtools. RNAseq library quality control was implemented using RSeQC v4.0.0 and QualiMap v2.2.2-dev. Duplicate reads were marked using Picard tools v2.23.9. Library complexity was estimated using Preseq v2.0.3. Duplication rate quality control was performed using dupRadar v1.18.0. Reads overlapping with coding sequences were assigned to genes using featureCounts v2.0.1. Classification of rRNA genes and their reads was based on annotations and RepeatMasker rRNA tracks from the UCSC genome browser when applicable. Differential gene transcription analysis was completed using DESeq2 v1.28.0. Functional enrichment analysis was achieved using g:Profiler python API v1.0.0. Quality control and analysis results plots were visualized using MultiQC v1.9. BAM indexing and read coverage plots were performed with IGV 2.16.0.

The LPS samples were prepared as previously described (Marolda et al., 2006), with some modifications. Briefly, bacteria were grown in LB at 37°C with agitation (160 rpm) until reaching an OD₆₀₀ = 0.6. An aliquot of each culture was adjusted to an OD₆₀₀ = 2.0 with PBS. Next, 1.5 ml of each suspension was centrifuged, and bacterial pellets were suspended in 90 μl of lysis buffer (1 M Tris–HCl pH 6.8, 2% SDS, 4% β-mercaptoethanol, 10% glycerol, and 0.002% bromophenol blue) and incubated for 10 min in a boiling water bath. Finally, each lysate was supplemented with 10 μl of proteinase K (10 mg/ml) and incubated for 1 h at 60°C. The LPS samples (5 μl) were analyzed in 12% (w/v) acrylamide gels using a Tricine-SDS buffer system (Lesse et al., 1990) and visualized by silver staining (Marolda et al., 2006).

For the mouse survival assays, bacterial doses were prepared by diluting O/N cultures to an initial OD₆₀₀ = 0.01 in 3 ml of LB broth and incubated at 37°C and 160 rpm until OD₆₀₀ = 0.6 was reached. Growth was halted by briefly placing the culture tubes in ice. Then, the exact volume of the culture required for a dose of 1 × 10⁶ CFU (colony forming unit) in 50 μl was diluted in sterile PBS. This volume was calculated considering 3 × 10⁸ CFU/μl in OD₆₀₀ = 0.5, as estimated from previous CFU counts. Then, 6–7 weeks old male C57BL/6 mice were administered intragastrically with 50 μl of SEn WT, SEn ΔSEN1990::frt, SEn-aph, SEn ΔSEN1990::cat, or PBS. The mice were monitored throughout the assay while measuring their daily weight and clinical score from 0 to 15, with 0 being a normal state and 15 the state of maximum distress and suffering. A clinical score equal to 8 or a loss of more than 20% of the initial weight was used as a threshold to apply euthanasia (Sebastián et al., 2022). The clinical parameters evaluated are described in Supplementary Table 3.

For bacterial competition assays, bacterial doses were prepared as described above. Mice were infected intragastrically with 50 μl of a mixture of 5 × 10⁵ CFU of SEn WT-aph and 5 × 10⁵ CFU of SEn ΔSEN1990::cat. The mice were monitored throughout the assay while measuring their daily weight and clinical score. Forty-eight hours after the initial inoculum, the mice were euthanized, and complete organs were extracted, namely blood, liver, spleen, mesenteric lymph node, gallbladder, colon, and ileum. The organ samples were homogenized in sterile PBS, and serial dilutions were carried out. The bacterial load was assessed by culturing the serial dilutions on LB and Salmonella-Shigella agar plates with 50 ng/μl kanamycin or 10 ng/μl chloramphenicol. The CFUs were then counted after 24 h of incubation at 37°C. The Salmonella-Shigella agar was used to confirm that the colonies found in the sterile organs were Salmonella. The competition index was calculated as the ratio of the bacterial load recovered from LB-chloramphenicol plates by the bacterial load recovered from the LB-kanamycin plates (i.e., ΔSEN1990::cat/WT-aph) and normalized with the ratio of the input inoculum to account for differences in the initial CFUs of each strain in the mixture.

To evaluate the infection of the strains individually, the doses were prepared as described above, and mice were inoculated intragastrically with 50 μl of 1 × 10⁶ CFUs SEn WT-aph or 1 × 10⁶ CFU of SEn ΔSEN1990::cat. After 48 h post-infection (hpi), the mice were euthanized, and the organs were collected and processed as described above. The bacterial load was reported as CFU per gram of organ or ml of blood. The same assay was performed using the strains without antibiotic resistance genes, SEn WT, or SEn ΔSEN1990::frt, but the samples were seeded on LB and Salmonella-Shigella agar plates without antibiotics.

For all the analyses, we defined a limit of detection (LOD) as counting a minimum of one CFU in the 30 μl LB agar seeded of each organ sample. Thus, when no CFUs were detected, the LOD was applied.

RAW264.7 cells were incubated at 37°C and CO₂ 5% in DMEM medium supplemented with 10% FBS until confluence. Then, cells were scraped in fresh medium and diluted to a final concentration of 5 × 10⁵ cells/ml. Each well in a 24-well plate was filled with 1 ml of the suspension of cells. The bacterial doses were prepared by inoculating 3 ml of LB medium with O/N-grown Salmonella ser. Enteritidis WT pKK233.2-gfp and ∆SEN1990::frt pKK233.2-gfp to an initial OD₆₀₀ = 0.15 and grown at 37°C and shaken until OD₆₀₀ = 0.6. After this, the RAW264.7 cells were infected at an MOI = 10 bacteria/cell. To ensure infection, the plates were centrifuged at 2,000 g for 3 min after bacterial inoculation and before incubation at 37°C with 5% CO₂. After 2 h of infection, 10 μl of gentamicin (10 μg/ml) was added to the cells, which were further incubated for 1 h at 37°C with 5% CO₂, after which they were centrifuged at 2,000 g for 5 min. The media were replaced with 250 μl of PBS, and the cells were scraped and transferred to a 96-well plate where they were centrifuged and washed again with PBS. Then, the cells were stained with 50 μl of the viability probe (Alexa Fluor 700) for 10 min at 4°C, washed in PBS once, and fixed in 4% paraformaldehyde for 15 min at 4°C. The cells were then washed once with PBS and resuspended in 200 μl of PBS. The samples were stored O/N and analyzed the next day using the flow cytometer BD LSRFortessa™ X-20 with the software BD FACSDiva™. Using control samples, the voltage settings and compensation were adjusted before the main data acquisition. The singlets gating was performed by selecting the events that followed a 1:1 relationship between SSC-A and SSC-H. Then, RAW264.7 cells were gated by their size and complexity detected by the FSC-A and SSC-A, respectively. The filters FITC and Alexa-700A were used for GFP and viability probe detection and gating. Data processing and the gating were carried out using FlowJo V10.

Statistical analyses were performed using GraphPad Prism 10 software. All statistical analyses were performed with α = 0.05, and the error bars on the graphs represented the standard deviation.

The gene expression, relative ROD21 copy number, and island excision statistics were analyzed through a one-way ANOVA or t-test for independent samples. RNA-seq differential gene transcription statistics were assessed by calculating the false discovery rate (FDR) of each gene with a foldchange threshold = 0 and setting the significance as FDR < 0.05.

Competitive indexes were log-transformed and analyzed through one sample t-test with a theoretical mean of 0, which implied that when no statistical differences are detected, the bacterial load of both strains is equal. Bacterial loads in independent infections were analyzed with a t-test for independent samples. Mouse weight and clinical score were analyzed through a two-way ANOVA with time post-infection and strain as factors. Mouse survival curves were compared using the log-rank (Mantel-Cox) test.

The animal work was reviewed and approved by the Institution Scientific Ethical Committee for Animal and Environmental Care from Pontificia Universidad Católica de Chile under the approval code: 210707009.

Within ROD21, there are 28 ORFs, of which there are only experimental data regarding the function of SEN1970 (Integrase), SEN1993 (H-NS-like), SEN1975 (TcpS), and SEN1998 (RDF). For the remaining 24 ORFs, information regarding their putative function in the RefSeq annotation (NCBI Prokaryotic Genome Annotation Pipeline (PGAP)) is only available for some genes; SEN1976, SEN1977, and SEN1978 encode type IV pili-related proteins (pilus assembly protein PilV-like, phage tail fiber protein, and type IV major pilin), SEN1979 encodes a TradD-like protein, SEN1980 encodes a MobQ family relaxase, SEN1982 and SEN1984 encode lipoproteins, SEN1985 encodes a putative S-adenosyl-L-methionine dependent methyltransferase, SEN1990 encodes a putative plasmid replication/maintenance protein, and SEN1994 and SEN1995 encode a putative conflict system composed of a SLATT domain-containing protein and a nucleotidyltransferase, respectively (Table 1).

To uncover more evidence regarding the function of the ROD21 genes, we used Phyre2 to find structural homologs of each putative encoded protein (Table 1). Although this analysis failed to find any reliable predicted homolog for most ROD21 hypothetical proteins, we found new data regarding the function of the two putative lipoproteins—SEN1982 (WP_000722542.1) and SEN1984 (WP_000734572.1). Additionally, we found data that supports the predicted function of some proteins encoded in ROD21, such as the type IV pili-related proteins SEN1977 and SEN1978, the putative conjugal transfer protein SEN1980, and the putative nucleotidyltransferase SEN1995. The lipoprotein SEN1982 was predicted to contain a motif structurally similar to the wHTH domain of PadR-family transcriptional regulators (Table 1), specifically within the recognition helix and the wing (Ser-58 to Val-99). This suggested that SEN1982 could possibly bind to DNA, whereas the other lipoprotein SEN1984 was predicted to be structurally similar to a β-barrel assembly complex component, BamE (Pro-1 to Arg-109; Table 1). Finally, SEN1990 (WP_004860215.1), which is annotated in the RefSeq Salmonella ser. Enteritidis genome (NC_011294) as a putative replication/maintenance protein from the RepL family, was predicted by Phyre2 to share structural similarity with a LysR-like transcriptional regulator (Table 1). Other hits with a similar confidence score (> 90%) showed structural similarity to other wHTH-containing proteins, such as proteins from the MecR, MarR, and FadR family of transcriptional regulators. This predicted similarity was restricted between Glu-73 to Leu-181 of SEN1990 and the DNA-binding domains of the transcriptional regulators. Further inspection of the predicted domains in SEN1990 through NCBI-CDD showed that the protein contained a putative MarR-like HTH domain between Arg-62 and Leu-121.

Among the genes with new predicted functions, SEN1990 caught our attention since it was previously reported to have an effect on the persistence of Salmonella ser. Enteritidis in oviduct cells in chickens (Raspoet et al., 2014), and although not statistically significant, in the colonization of internal mouse organs (Silva et al., 2012). Moreover, since this gene is apparently involved in plasmid replication or maintenance, a possible core function of ROD21, its characterization would deepen our knowledge of the biology of the island. Therefore, to initially characterize its function, we gathered structural data for its encoded protein. Since SEN1990 does not have functionally characterized homologs, structural data were scarce. Therefore, we began the in silico structural characterization of SEN1990 by modeling the protein using the deep learning software D-I-TASSER (Zheng et al., 2021), which uses the I-TASSER protein threading pipeline to find structural templates in state-of-the-art protein model databases and generate the model (Yang and Zhang, 2015). The model constructed by the software had a template modeling score (TM-score) of 0.61 on a scale of 0 to 1, with 1 being the highest confidence score in the model (Figure 1A). A TM-score higher than 0.5 indicates a model with correct topology. Additionally, the isoelectric point (IP) calculated for SEN1990 was basic (9.03). Between the residues Lys-71 to Ile-130 of the protein, we identified a domain structurally similar to Helix-Turn-Helix (HTH) domains (Figure 1B). This region of SEN1990 is composed of a bundle of three α-helices (α) arranged triangularly (α5, α6, and α7), common for HTH domains, and three β-Strands (β) with two of them (β4 and β5) having antiparallel orientation that formed a “wing” structure, such as in winged HTH (wHTH) domains (Aravind et al., 2005). Following the predicted secondary structure of the protein, the wHTH was arranged as α5 – β3 – α6 – α7 – β4 – β5. The first β-Sheet (β3) is commonly found on a sub-class of wHTH domains known as three-stranded wHTH (Aravind et al., 2005). Moreover, analysis of the predicted hydrogen bond contacts showed putative interactions between the region connecting the α5 and α6 helices with the positively charged loop in the wing, which is also characteristic of three-stranded wHTH domains (Figure 1C; Aravind et al., 2005). Additionally, in the electrostatic map of SEN1990, there was an accumulation of positive charges in the wHTH domain (Figure 1D). These charges seem to be due to conserved lysine residues in α7, which corresponded topologically with the DNA-recognition helix of wHTH domains, and conserved histidine residues in the wing. The alignment of the putative wHTH region of SEN1990 to MarR (1JGS), a common wHTH domain-containing protein, showed an average distance of 1.74 Å between the aligned C-α atoms (Figure 1B). The alignment distances were as low as 0.33 Å in the region from β3 to β5, whereas α5 was dissimilar to the marR topology. The α5 helix had distances that were as high as 7.6 Å, and even some residues that were not aligned. On the other hand, the rest of SEN1990 (i.e., the N and C terminal regions of the protein) did not share structural similarities with other characterized proteins.

Since ROD21 belongs to the EARL family of EGIs, and no other characterized proteins shared structural homology with SEN1990, we assessed whether genes similar to SEN1990 were carried by other EGIs from this family. Through tBLASTx and loci comparison between ROD21 and EARL EGIs, we identified the presence of possible SEN1990 homologs in 25 EGIs harbored by different bacterial strains. Through multiple sequence alignment and phylogenetic analysis, we identified two groups of SEN1990 homologs (Figure 2A). The first one was composed of homologs closely related to SEN1990 harbored by strains of the animal pathogens E. coli and Salmonella and by the plant pathogen Pectobacterium. Moreover, the predicted proteins from this group were highly conserved, as they were almost identical to SEN1990. On the other hand, the second group of homologs was composed of proteins encoded by EGIs carried by strains of the animal pathogens Klebsiella, Yersinia, Serratia, and Salmonella. Furthermore, the predicted protein homologs from this group had less amino acid similarity with SEN1990, but they still shared some highly conserved residues, especially within the predicted DNA-binding domain, such as the region between the highly conserved motif between Lys-118 and Leu-121 (Figure 2B). Interestingly, the genetic context of all SEN1990 homolog loci was similar, as the genes were located in a homologous position within each EGI. Most of the homologous genes have a relatively long intergenic non-coding region followed by a divergently oriented gene upstream from their initial codon, whereas, in the rest of the loci, there are only small predicted ORFs in the intergenic region.

Since SEN1990 was annotated through the RefSeq pipeline as encoding a protein from a family of replication/maintenance proteins named RepL, and we found evidence that supports its function as a DNA-binding protein, we sought to assess if SEN1990 was involved in the replication/maintenance of ROD21. To do this, we first assessed the protein amino acid sequence similarity to other RepL family proteins by multiple sequence alignment between SEN1990 and the plasmid-encoded proteins from the RepL family (Lampson and Parisi, 1986; Hefford et al., 1997). In these results, SEN1990 seemed to be dissimilar to the RepL proteins since there were only a few conserved residues between all proteins (Figure 3A), and the sequence identity fell below 19% between SEN1990 and RepL proteins. Moreover, residues that were highly conserved in RepL proteins, such as Asn-131 and Pro-132, were not present in SEN1990, nor were they substituted by residues with similar properties. This suggested that SEN1990 might not be part of the RepL family or that it was distantly related to these replication/maintenance proteins. Thus, we experimentally assessed whether SEN1990 was involved in the replication/maintenance of ROD21 despite the protein not sharing strong similarities with RepL proteins. Importantly, the capacity of ROD21 to replicate autonomously and the number of copies in which the excised island can be found within a cell have not been assessed previously. Therefore, we measured the number of copies of ROD21 per chromosome (i.e., total ROD21 copy number) and the relative quantity of excised ROD21. If ROD21 was replicated alongside the chromosome, we could expect an equal amount of ROD21 genes to single copy chromosomal genes, such as rpoD, whereas if ROD21 self-replicated, then there would exist more than one copy of ROD21 genes per rpoD in the bacterial population. The results showed that the copies of a ROD21 gene, SEN1998, were in a ratio of 1 to rpoD in the WT strain, suggesting that ROD21 existed as a single copy of EGI per chromosome. Therefore, it was likely that the island was unable to replicate autonomously, and this process probably relied on chromosomal replication (Figure 3B). Nonetheless, the absence of a protein involved in the EGI extrachromosomal maintenance could lower the amount of excised ROD21 to values under the number of total chromosomes, such as in the case of a population of cells losing ROD21 entirely. Moreover, even when SEN1990 was deleted, we observed that the ratio of ROD21/chromosomes remained 1, indicating that ROD21 was maintained regardless of the presence of SEN1990. Despite this, since the excision of ROD21 in all the strains was only 3% (Figure 3C), the replication of a subset of excised ROD21 might not be identifiable among all the copies of ROD21. Therefore, we measured the relative amount of excised ROD21 by amplifying the attP sequence, which was only present on the circular ROD21. We observed that in the strain lacking SEN1990, the relative quantity of excised ROD21 remained unaltered (Figure 3D), supporting the notion that the protein was not involved in the EGI replication/maintenance.

Since SEN1990 did not seem to play a role in the replication/maintenance of ROD21 and its predicted product shared structural similarities to wHTH-containing transcriptional regulators, we evaluated if the protein could modulate the transcription of genes in Salmonella ser. Enteritidis. Hence, we performed a total RNA-seq of the WT and ΔSEN1990::frt and evaluated the differential gene transcription between them. First, deletion of SEN1990 did not affect the expression of ROD21 genes (Figure 4A), although the gene directly downstream of SEN1990, SEN1991, had an increased expression in the mutant strain (Log₂ Fold change = 1.02) but with an FDR of 0.43. Therefore, we assessed through RT-qPCR the transcription of SEN1991, which, although not statistically significant, showed that the mRNA quantity of ΔSEN1990::frt was higher compared to the WT strain, whereas the complemented strain showed similar expression to the WT strain, hinting at a possible regulation of its own operon (Supplementary Figure 4).

Interestingly, upstream of SEN1990, there was an intergenic non-coding region that also had low transcription compared to the rest of the island, hinting at a possible location of the yet unidentified transfer origin of ROD21. Within the Salmonella ser. Enteritidis genome, the only gene with a statistically significant differential expression besides SEN1990 was oafB (previously known as Family II gtrC), which encodes an LPS modification enzyme located in pathogenicity island SPI-17 (Figure 4A). This gene had a log₂ fold change of −1.18, meaning that it was approximately expressed 0.44 times less in the mutant compared to the WT strain (Figure 4B). Besides oafB, there were several genes downregulated in the mutant strain, although these differences in expression were not statistically significant, and these genes had few reads associated with them. To confirm the RNA-seq result, we performed RT-qPCR of oafB in different samples of Salmonella ser. Enteritidis grown in the same conditions while also assessing the transcription of the gene in the strain complemented ectopically with SEN1990 (Salmonella ser. Enteritidis ΔSEN1990::frt pAPI-SEN1990). The expression of oafB in ΔSEN1990::frt was approximately half of its expression in the WT strain, which supports the results obtained in the RNA-seq analyses. Additionally, the ectopic complementation of SEN1990 recovered the mRNA quantity of oafB near to the levels seen in the WT strain (Figure 4C). Nonetheless, the expression of oafB did not reach statistically significant differences between the deletion and complementation strains. To evaluate if the deletion of SEN1990 could alter the O-antigen of the pathogen, we extracted the LPS from the WT, ΔSEN1990::frt, and complemented strains and compared them in a silver-stained SDS-PAGE, but we could not detect any shifts in the LPS bands between the strains (Supplementary Figure 4B).

As stated before, previous studies have shown evidence that SEN1990 might have a role in the pathogenesis of Salmonella ser. Enteritidis through microarray-based screenings (Silva et al., 2012; Raspoet et al., 2014). Moreover, we have shown that SEN1990 is involved in the expression of another pathogenicity island gene that encodes a protein that catalyzes the modification of LPS. Thus, we hypothesized that SEN1990 contributes to the pathogenic potential that ROD21 provides to the bacterium. Therefore, to test this hypothesis, we performed an infection competition assay where mice were inoculated intragastrically with a mixture of equal concentrations of Salmonella ser. Enteritidis WT strain marked with a kanamycin resistance gene inserted in ROD21 (WT-aph) and the ΔSEN1990::cat strain. After 48 hpi, there were no differences in the ratio of CFUs recovered of both strains in any of the organs evaluated (Figure 5A). Nonetheless, we also evaluated the pathogenicity of each strain independently since it is possible that a cooperative phenomenon could mask the differences between the strains when they are co-administered. Thus, we inoculated mice intragastrically with either WT-aph or ΔSEN1990::cat. We observed at 48 hpi that the strain lacking SEN1990 had a statistically significant reduced bacterial load in the liver while also a lower bacterial load in the ileum, colon, and spleen (Figure 5B). No CFUs of either strain could be recovered from the blood, mesenteric lymph node (mLN), and gallbladder. To rule out a bias on the bacterial loads caused by the insertion of the antibiotic resistance genes, such as a metabolic burden, we performed the experiment using the strains without the selection markers, Salmonella ser. Enteritidis WT and ΔSEN1990::frt. We observed no statistically significant differences in the bacterial loads between the strains, although there were fewer CFUs of ΔSEN1990::frt recovered from the blood (where we did not detect the development of bacteremia in most mice infected with this strain) and the liver (Figure 5C). In contrast to the previous experiment, the bacterial load in the spleen was almost identical in both strains, and CFUs were detected in only a few mice in mLN and gallbladder, but no tendencies could be seen (Supplementary Figure 5). Additionally, mice infected with the ΔSEN1990::frt strain showed reduced weight loss and development of symptoms related to the disease compared to the WT strain (Figure 5D). Despite this, survival assays showed that the deletion of SEN1990 did not improve the survival rate of the infected mice (Supplementary Figure 5). To better visualize all of the data gathered from the mice infected with the WT and ΔSEN1990::frt strains, we performed a principal component analysis (PCA), including mice weight reduction and clinical score, bacterial loads in every organ, and ileum and colon length. We observed that the mice infected with the mutant and the mice infected with WT formed distinct clusters, indicating that the strains appeared to behave differently in their pathogenicity (Figure 5E). Additionally, to begin the characterization of how the gene could reduce the colonization of sterile organs, we performed bacterial uptake assays by a murine macrophage cell line (RAW264.7). The cells were inoculated for 2 h with either WT or ΔSEN1990::frt expressing GFP ectopically through the pKK233.2-gfp plasmid and treated with gentamycin to kill the extracellular bacteria. In this assay, we did not find any differences in the bacterial internalization by macrophages between the strains (Supplementary Figure 6).