Bile salts activate the expression of T6SS gene clusters of several bacterial pathogens (Crawford and Gunn, 2014; Sistrunk et al., 2016), including those encoded within SPI-6 and SPI-19 in S. Typhimurium and S. Dublin in which McConkey plate are used to induce T6SS-expression during bacterial growth (Brunet et al., 2015a; Sana et al., 2016; Schroll et al., 2019; Amaya et al., 2022) Therefore, we performed competition assays in either LB or McConkey agar using E. coli DH5α as the prey strain. To avoid cross-complementation among components of both T6SS gene clusters, as documented in other systems (Santos-Moreno et al., 2020), we generated mutant strains lacking the complete T6SS gene clusters of SPI-20 and/or SPI-21 in S. arizonae RSK2980. As shown in Figure 1B, the E. coli prey strain was significantly outcompeted by the S. arizonae wild-type strain only when co-incubated in McConkey agar, suggesting that bile salts promote the T6SS-dependent antibacterial activity of S. arizonae. Notably, strains lacking either T6SS_SPI-20 or T6SS_SPI-21 still outcompeted the prey strain at levels similar to those of the wild-type strain. Nonetheless, S. arizonae lacking both T6SSs was unable to outcompete the prey strain. These findings suggest that T6SS_SPI-20 and T6SS_SPI-21 confer a competitive advantage to S. arizonae RSK2980 over E. coli DH5α, and that the presence of bile salts activates both T6SSs. Of note, having both systems do not confer an additive advantage over having only one, suggesting functional redundancy between them.

To identify novel T6SS effector proteins and their cognate immunity proteins, we performed a systematic analysis of each ORF encoded within the SPI-20 T6SS gene cluster of S. arizonae RSK2980. Firstly, we used the Bastion6 prediction pipeline (Wang et al., 2018a), a bioinformatics tool that employs amino acid sequence profiles, evolutionary information, and physicochemical properties to predict T6SS effectors. Secondly, to identify conserved domains and motifs associated with known T6SS effectors, the PROSITE, NCBI-CDD, Motif-Finder, and Pfam databases were used. In addition, functional predictions were made via HMM homology searches using the HHpred HMM-HMM prediction pipeline (Zimmermann et al., 2017) and through structure-based searches using Foldseek based on the predicted structure of each protein obtained with AlphaFold3 (Abramson et al., 2024). To identify if the predicted candidates could be part of an E/I module, we also performed a screen for putative immunity proteins by detecting signal peptides (SignalP), transmembrane domains (TMHMM), and bicistronic operons (Operon-mapper), as outlined by Taboada et al., (Taboada et al., 2018). In addition, we analyzed the SPI-20 T6SS gene cluster to identify potential unannotated ORFs that could encode putative effectors and cognate immunity proteins.

Our analysis confirmed the prediction of a previously identified putative T6SS effector protein and did not identify novel predicted effectors within SPI-20 (Table 1). This putative effector corresponds to SARI_02727 (Figure 2A), a protein previously identified as a putative homolog of the VgrG2b effector protein from P. aeruginosa (Wood et al., 2019). SARI_02727 encodes an 189 amino acid protein with a modest Bastion6 score (0.611) predicted to be part of a bicistronic unit along with SARI_02726. Structure-based analysis of the AlphaFold3-predicted structure for SARI_02727 confirmed the structural similarity between SARI_02727 and the known structure of the VgrG2b effector protein. As shown in Figure 2A, SARI_02727 displays a high degree of similarity with the catalytic domain of VgrG2b, including the conserved HEXXH motif found in the catalytic site of zinc-dependent metallopeptidases (Amaya et al., 2024). The predicted catalytic amino acid side chains (H95, E96 and H136) in SARI_02727 coordinated with a zinc ion at the center (Figure 2A).

As mentioned, SARI_02727 is predicted to be part of a bicistronic unit with SARI_02726, suggesting that SARI_02726 is the immunity protein of the candidate effector protein. SARI_02726 encodes a 154 amino acid protein with a periplasmic lipoprotein-type signal peptide (Sec/SPII). A sequence analysis revealed that SARI_02726 harbors the DUF7480 domain (IPR055903), not previously linked to T6SS function. We then used AlphaFold3 to model and predict if SARI_02727 and SARI_02726 could interact and constitute a T6SS E/I pair. As shown in Figure 2B, AlphaFold3 predicted an interaction between SARI_02727 and SARI_02726 with a high confidence score (ipTM: 0.87), suggesting that the two proteins could constitute a T6SS E/I pair.

To determine if SARI_02727 contributes to interbacterial competition in S. arizonae, we generated isogenic derivatives lacking either SARI_02727 or the putative E/I module composed by SARI_02727 and SARI_02726, and performed bacterial competition assays on McConkey agar. As shown in Figure 2C, the wild-type strain outcompeted the mutant lacking the putative E/I module, but not the ∆SARI_02727 mutant, in agreement with the prediction that SARI_02726 corresponds to the immunity protein of SARI_02727. In addition, the ∆SARI_02727 mutant was not able to outcompete the mutant lacking the putative E/I module.

Finally, to test the potential antibacterial activity of the predicted metallopeptidase domain of SARI_02727 and the neutralizing effect of the SARI_02726 immunity protein, we performed heterologous expression assays in E. coli. The fundamental premise underlying these assays is that the effector protein should cause a growth impairment when expressed in the appropriate localization in E. coli. This growth impairment should be counteracted, at least partially, by the concomitant production of the cognate immunity protein.

Thus, we transformed E. coli BL21(DE3) cells with derivatives of plasmid pRSFDuet-1 carrying genes encoding either the effector harboring an N-terminal PelB signal for periplasmic localization (pRSFDuet-1_SP_SARI_02727), the immunity protein (pRSFDuet-1_SARI_02726) or both (pRSFDuet-1_SP_SARI_02727/SARI_02726). As shown in Figure 2D, induction of SARI_02727 expression caused a drastic growth impairment in E. coli BL21(DE3). In addition, the co-expression of SARI_02727 and its cognate immunity protein (SARI_02726) partially restored cell growth (Figure 2D). At the same time, there was no apparent growth impairment when cells expressed only the immunity protein or carry the empty vector (Figure 2D). These results indicate that SARI_02727 is toxic to E. coli BL21(DE3) when targeted to the periplasm, and that SARI_02726 could partially neutralize its toxic effect, suggesting that they correspond to an E/I pair.

To date, only one effector protein, VgrG-NucSe1 (SARI_02603), has been predicted to be encoded within the SPI-21 T6SS gene cluster of S. arizonae (Blondel et al., 2009; Ho et al., 2017). VgrG-NucSe1 is an evolved VgrG protein harboring a C-terminal extension with homology to S-type pyocins (Blondel et al., 2009) (Table 1). Although it has been shown that the T6SS of V. cholerae can heterologously deliver the C-terminal region of VgrG-NucSe1 and possesses antibacterial activity (Ho et al., 2017), there is no direct report on the contribution of this effector to interbacterial competition in S. arizonae.

First, we analyzed the predicted HNH nuclease domain within the C-terminal region of VgrG-NucSe1. A Foldseek analysis of the predicted structure generated by AlphaFold3 showed a significant similarity to the DNase domain of the AP41 pyocin from P. aureginosa (Figure 3A), including the three catalytic histidines of VgrG-NucSe1 (H254, H279, and H283 in S. arizonae). AlphaFold3 analysis also predicted with high confidence the interaction between the C-terminal region of VgrG-NucSe1 and its cognate immunity protein, SARI_02602 (Figure 3B).

To test if VgrG-NucSe1 contributes to the antibacterial activity of T6SS_SPI-21 in S. arizonae, we performed interbacterial competition experiments between the wild-type strain and mutant strains lacking either VgrG-NucSe1 or both VgrG-NucSe1 and its cognate immunity protein SARI_02602. As shown in Figure 3C, the wild-type strain successfully outcompeted the mutant strain lacking both VgrG-NucSe1 and SARI_02602, but not the strain retaining the putative cognate immunity protein SARI_02602. In contrast, a mutant strain lacking VgrG-NucSe1 was unable to outcompete a mutant strain lacking the whole E/I module (Figure 3D). Finally, heterologous expression assays in E. coli strain BL21(DE3), showed that expression of the C-terminal region of VgrG-NucSe1 caused significant growth impairment, which was partially recovered by co-expression of the cognate immunity protein SARI_02602 (Figure 3D), in agreement with previous experimental data obtained in V. cholerae (Ho et al., 2017).

To identify novel T6SS effector candidates within SPI-21, we performed a bioinformatic analysis similar to that described above for SPI-20. Our analysis identified one effector protein candidate encoded within SPI-21 (Table 1).

This putative effector corresponds to SARI_02625 (Bastion6 score = 0.933), a 1,032 amino acid protein with three predicted functional domains in its C-terminal region (Figure 4A). These predicted domains correspond to an EF-hand domain (IPR002048), responsible for calcium-binding, a duplicated hybrid motif (IPR002048), frequently found in M23-related peptidases, and an endolysin/autolysin domain (IPR033907) found in peptidoglycan-degrading enzymes (Figures 4A,B). The presence of these three domains suggests that SARI_02625 is an antibacterial effector protein targeted to the bacterial periplasm (Figures 4A,B). In agreement with this, Foldseek analysis of the AlphaFold3-predicted structure of SARI_02625 revealed significant structural similarity of the endolysin/autolysin domain of SARI_02625 to the LysF1 endolysin of the E. coli O157:H7 phage FAHEc1 (Love et al., 2021), including the conserved E[DC]T catalytic triad (E890, D899, and T905 in SARI_02625) (Figure 4B). Our analysis also revealed that SARI_02625 is part of a bicistronic unit along with SARI_02624. This ORF encodes a 276 amino acid protein with a periplasmic-targeting signal peptide (Sec/SPI), suggesting that SARI_02624 is the cognate immunity protein of the putative novel effector SARI_02625. Indeed, AlphaFold3 analysis predicted (although with moderate confidence) that SARI_02625 and SARI_02624 could interact, therefore constituting a T6SS E/I pair (Figure 4D).

To evaluate this, we performed interbacterial competition assays with the wild-type strain and mutant strains lacking either SARI_02625 or both SARI_02625 and SARI_02624. As shown in Figure 4C, the wild-type strain significantly outcompeted the mutant strain lacking the putative E/I pair, but not the strain lacking only the putative SARI_02625 effector protein. Furthermore, this mutant strain lacking the SARI_02625 effector protein could not outcompete the mutant strain lacking the putative E/I pair (Figure 4C). In addition, heterologous expression assays in E. coli showed that expression of the endolysin domain of SARI_02625 targeted to the periplasm, by adding the N-terminal region of PelB, caused significant growth impairment, which was partially recovered when the SARI_RS1220 protein was co-expressed (Figure 4E). Together, our results suggest that SARI_02625 and SARI_02624 conform a T6SS E/I pair encoded within SPI-21 in S. arizonae.

The characterization of new candidate T6SS effectors in S. arizonae, which harbor predicted antibacterial protein domains, prompted us to determine their distribution across Salmonella. To this end, the nucleotide sequence corresponding to the ORF encoding each candidate effector was used in tBLASTx searches in publicly available Salmonella genome sequences deposited in the NCBI database (August, 2025) and the distribution of each effector was determined. Our analysis revealed that SARI_02727, SARI_02603 and SARI_02625 candidate effectors encoded in the SPI-20 and SPI-21 T6SS gene clusters, respectively, are predominantly restricted to isolates of S. enterica subspecies arizonae and diarizonae (Supplementary Table S1).

The bacterial strains used in this study are listed in Table 2. Bacteria were routinely grown in Lysogeny Broth (LB) (10 g/L tryptone, 5 g/L yeast extract, 5 g/L NaCl) at 37°C with aeration. LB medium was supplemented with ampicillin (Amp; 100 μg/mL), kanamycin (Kan; 50 μg/mL), chloramphenicol (Cam; 20 μg/mL), or nalidixic acid (Nal; 15 μg/mL), as needed. LB medium was solidified by the addition of agar (15 g/L). For interbacterial competition assays, bacteria were grown on McConkey agar plates (BD) at 37°C for 24 h. For heterologous expression assays, bacteria were incubated at 37°C for 24 h on LB agar plates supplemented with Kan and/or Cam and the corresponding inducers (i.e., 0.1 mM IPTG and/or 0.4% L-arabinose).

Plasmid DNA was isolated using the “QIAprep Spin Miniprep Kit” (QIAGEN, MD, USA). PCR products were purified using the “QIAquick PCR Purification Kit” (QIAGEN, MD, USA). Analysis of DNA samples was performed by electrophoresis in 1% agarose gels in 1X TAE buffer and visualized under UV light after RedGel (Biotium, CA, USA) staining. Primers were designed using the SnapGene software (www.snapgene.com) and are listed in Table 3. PCR reaction mixes contained 1X buffer, 2 mM MgCl₂, 100 nM dNTPs, 100 nM of each primer, 100 ng of template DNA and 0.5 to 1 U of Phusion High-Fidelity DNA Polymerase (NEB, USA). Standard conditions for amplification were: 1 min at 98°C, followed by 30–35 cycles of 98°C for 10 s, 55°C to 72°C (according to the appropriate Tm for each primer pair) for 30 s and 72°C for a suitable time (15–30 s/kb), and a final extension step at 72°C for 10 min. All amplifications were conducted in a “MultiGeneTC9600-G” thermal cycler (LabNet, NJ, USA).

Derivatives of S. arizonae RSK2980 with deletions of the SPI-20 and/or SPI-21 T6SS gene clusters were constructed by the one-step inactivation procedure using the Lambda Red recombination system (Datsenko and Wanner, 2000), with modifications (Santiviago et al., 2009). The oligonucleotides used for mutagenesis (Table 3) were made with 40 bases at the 5′ ends identical to the ends of the corresponding deletion, and 20 bases at the 3′ ends that anneal with the 5′ or 3′ end of a Cam or Kan resistance cassette flanked by Flp recombinase target (FRT) sites present in plasmids pCLF2 (GenBank accession number HM047089) and pCLF4 (GenBank accession number EU629214.1), respectively. pCLF2 and pCLF4 were employed as templates for the corresponding amplification of PCR products. S. arizonae RSK2980 was transformed with plasmid pKD46, which allows the inducible expression of the λ Red recombination system in the presence of L-arabinose. Then, bacteria carrying pKD46 were grown to an OD_600nm of 0.6 at 30°C in LB broth supplemented with Amp and L-arabinose (10 mM). In the next step, bacteria were made electrocompetent through serial washes with ice-cold, sterile 15% glycerol, and transformed via electroporation with 500 to 600 ng of each PCR product. Transformants were selected at 37°C on LB agar supplemented with the corresponding antibiotic. The correct insertion of the corresponding antibiotic resistance cassettes in each mutant was confirmed by PCR amplification using suitable primers (Table 3). To avoid any potential off-target effect of the chromosomal insertion of PCR products, we used generalized transduction of the confirmed mutations to a Salmonella wild-type background by means of the transducing phage P22 HT105/1 int-201.

Competition experiments to determine the ability of attacker strains to outcompete prey bacteria were conducted as described previously (Ma et al., 2018), with modifications. In brief, the attacker and prey bacteria were grown overnight in LB broth at 37°C. An aliquot (1 mL) of each culture was collected by centrifugation at 8,000 rpm for 2 min and the supernatant was discarded. Each bacterial pellet was washed three times in sterile PBS, adjusted to an OD_600nm of 0.5, and mixed at a 1:1 attacker-to-prey ratio. Then, aliquots (25 μL) of this mixture were spotted on McConkey agar plates in triplicate and incubated at 37°C for 24 h. This condition has been reported to induce expression of T6SS gene clusters in Salmonella (Schroll et al., 2019). After incubation, the bacterial spots were scraped from the McConkey agar plates, resuspended in 1 mL of sterile PBS, and colony-forming units (CFU) were determined by plating serial dilutions on LB agar supplemented with suitable antibiotics. CFU obtained from interbacterial competition experiments were used for data analysis to calculate a competition index (CI). The CI was calculated as the mean ratio of logarithmically converted CFU of the prey to attacker strains, normalized to the input ratio. Statistical significance was calculated using GraphPad Prism 9.0 software and a one-way ANOVA test followed by a Tukey’s multiple comparisons test.

To identify putative T6SS effectors encoded within SPI-20 and SPI-21 T6SS gene clusters in S. arizonae RSK2980, each ORF of both pathogenicity islands was analyzed with the Bastion6 pipeline (Wang et al., 2018b) excluding those encoding the 13 structural components of the T6SS. ORFs exhibiting a Bastion6 score of at least 0.6 were designated as potential T6SS effectors. Each Bastion6 prediction was subjected to further analysis using tools implemented in the Operon-Mapper web server (Taboada et al., 2018) to determine its potential inclusion in a single transcriptional unit that also encoded a putative immunity protein. This term refers to a small protein with potential signal peptides (SignalP 6.0) and/or transmembrane domains (TMHMM 2.0). The identification of conserved functional domains and motifs in the candidate T6SS effectors was facilitated by the utilization of the PROSITE, NCBI-CDD, Motif-Finder, and Pfam databases (Kanehisa et al., 2002; Sigrist et al., 2013; Finn et al., 2014; Lu et al., 2019; Blum et al., 2020) which were integrated within the GenomeNet search engine. The E-value cutoff score was set at 0.01. Furthermore, for each effector and immunity protein identified, a biochemical functional prediction was conducted through HMM-based searches using the HHpred HMM-HMM comparison tool (Zimmermann et al., 2017). The analysis of predicted effector proteins is summarized in Table 1.

The growth inhibition caused by SARI_02727 effector in E. coli BL21(DE3) was determined by growth on solid media in the presence or absence of the inducer IPTG. Overnight cultures of E. coli BL21(DE3) cells with pRSFDuet-1 plasmid containing SARI_02727, SARI_02726 or both genes were diluted 4-fold and aliquots (5 μL) were spotted onto LB agar plates containing Kan plus 0.1 mM IPTG to induce the synthesis of either the effector, immunity protein or both proteins, respectively. The plates were incubated at 37°C for 24 h and the inhibition of bacterial growth was visually determined. E. coli BL21(DE3) containing empty pRSFDuet-1 was used as control.

To evaluate the ability of SARI_02625-CT to inhibit bacterial growth and the protection conferred by its cognate immunity protein (SARI_02624), overnight cultures of E. coli BL21(DE3) co-expressing the C-terminal domain of effector SARI_02625 from plasmid pRSFDuet-1_SP_CT12255 and its immunity protein SARI_02624 from plasmid pBAD33.1_12250 were serially diluted in LB broth (1:4) and aliquots (5 μL) were spotted onto LB agar plates containing Kan and Cam, plus 0.1 mM IPTG and/or 0.4% L-arabinose to induce the synthesis of the effector and immunity protein, respectively. The plates were incubated at 37°C for 24 h and bacterial growth inhibition was determined. Derivative strains of E. coli BL21(DE3) containing either pRSFDuet-1 and pBAD33.1 empty vectors; pRSFDuet-1_SP_CT12255/pBAD33.1 empty and pRSFDuet-1 empty/ pBAD33.1_12250 were used as controls.

To evaluate the bacterial toxic activity of the S-Type pyocin domain of SAR_02603 and its neutralization by its cognate immunity protein SARI_02602, a similar approach as described above for SARI_02625-CT/SARI_02624 was conducted. In this case, the heterologous expression assay was performed on LB agar plates containing Kan and Cam, plus 0.05% L-arabinose and/or 1 mM IPTG to induce the synthesis of the effector (pBAD33.1_Stype_pyocin_12160) and immunity protein (pRSFDuet-1_Col_Imm_12155), respectively.

Protein structure models of SARI_02727, SARI_02625 and the C-terminal domain of SARI_02603 were obtained using I-TASSER (Yang et al., 2015), a protein structure homology-modeling server, and AlphaFold3 (Abramson et al., 2024) to predict effector-immunity protein interactions. The best model was used in downstream analyses (i.e., the rank_001 model). Protein structure visualization, template alignment, and superposition were performed using Pymol v3.1 and Mathmaker of UCSF ChimeraX (Pettersen et al., 2021). Protein structure searchers were performed with the Foldseek server (van Kempen et al., 2024).

The DNA sequence encoding each T6SS effector identified in this study was subjected to tBLASTx analyses to find orthologs in all Salmonella genome sequences deposited in the NCBI database (August, 2025). The selection of positive matches was based on a 90% identity and 90% sequence coverage threshold. The conservation of sequences was determined by multiple sequence alignments using T-Coffee Expresso (Notredame et al., 2000), MAFFT (Katoh et al., 2017), and ESPript 3 (Robert and Gouet, 2014). Comparative genomic analyses of T6SS gene clusters were performed using Mauve (Darling et al., 2004) and EasyFig v2.2.5 (Sullivan et al., 2011). The analysis of nucleotide sequences was conducted using Artemis version 18 (Rutherford et al., 2000).