Salmonella Derby adaptation to swine and simultaneous attenuation for humans go through decay of Salmonella Pathogenicity Island I

ABSTRACT We report the identification, within the global population of Salmonella Derby, of a lineage highly diffused and expanding in swine while being significantly under-represented in humans which carries stop mutations in the Salmonella Pathogenicity Island 1 (SPI-1) genes sipA and hilC. Single-cell analysis of invasion, vacuolar load, and cytosolic hyper-replication show that these mutations attenuate S. Derby virulence in human intestinal cells. Identification of this lineage follows the previous detection of another lineage over-represented in swine compared with humans which carries a loss-of-function mutation in hilD, the major SPI-1 regulator, responsible for reduced infection of human cells. Notably, this hilD-mutated lineage represents a sub-population of the wider sipA and hilC-mutated lineages identified in this study, indicating that SPI-1 is subjected to genetic decay in this part of the Salmonella Derby population adapted to swine. Virulence assessment on swine intestinal cells showed drastically lower levels for both wild-type and mutant S. Derby compared with human cells. Still, S. Derby with impaired sipA and hilC showed further attenuation also in swine cells compared with the wild type. These data suggest that invasion of swine intestinal epithelium is not needed by S. Derby to circulate in pigs, leading to the decay of SPI-1, which in turn determines attenuation for humans. Accordingly, a further S. Derby lineage under-represented in humans compared with pigs was found carrying truncated or missing SPI-1 genes, indicating that the S. Derby population is undergoing converging evolution in its adaptation to swine while attenuating for humans through loss of SPI-1 function. IMPORTANCE This study integrated population data with in vitro assessment of virulence phenotypes to unveil that a considerable part of the global population of Salmonella Derby is evolving to enhance its host adaptation to the swine host and that this evolution is simultaneously increasing its attenuation for humans. The study shows that the fixation of deleterious mutations in SPI-1 has a role in this process. This evidence indicates that SPI-1 has a key role for S. Derby virulence in humans but not for its circulation in swine. The results show that genes generally considered essential for Salmonella pathogenesis do not play the same key role for all Salmonella serovars or lineages and/or all hosts. The study helps in understanding the molecular mechanisms underlying the ecology and host adaptation of Salmonella showing that the adaptation process can vary for different types of Salmonella and hosts.

animal species as well as humans.In particular, pork is a major vehicle of salmonellosis, because of its considerable consumption and the high prevalence of swine infection (2).The mostly isolated Salmonella enterica serovars from pigs in the EU are S. Typhi murium (15.3%), monophasic variant of S. Typhimurium (28.2%), and Derby (22.3%).Although similarly present in pigs, these serovars differ substantially in their zoonotic potential.S. Typhimurium and monophasic S. Typhimurium cause 11.4% and 8.8% of salmonellosis cases in humans, respectively, whereas S. Derby is responsible for only 0.93% of human cases (1).This is consistent with the established knowledge that S. Typhimurium and its monophasic variant are generalist serovars infecting multiple hosts, while S. Derby is a swine-adapted serovar commonly found in pigs but only rarely detected in humans (3,4).This unbalance in the incidence of infection by S. Derby in swine compared with humans offers the opportunity to investigate the molecular mechanisms of Salmonella adaptation.In a previous study, we identified a specialized case of host adaptation to swine by a specific S. Derby lineage, featuring even further decreased pathogenicity to humans while maintaining high prevalence in pigs (5).We found that the lineage carries a single loss-of-function mutation in hilD, encoding the master activator of Salmonella Pathogenicity Island 1 (SPI-1) (6), and demonstrated that the hilD mutation was responsible for impaired interaction with human cells much more than with swine cells, potentially explaining the negligible presence of this lineage in humans despite its circulation in swine.This finding represents an example of Salmonella evolution towards host adaptation through the generation of allelic variants by point mutation, a reported mechanism of host adaptation along with genome degradation and horizontal gene transfer (7).SPI-1 is a ~40-kb genomic region which includes genes encoding a Type 3 Secretion System (T3SS), effector proteins, and transcriptional regulators of SPI-1 genes (8).The SPI-1 products are considered essential for each step of Salmonella pathogenesis inside intestinal epithelial cells, i.e., invasion, survival in Salmonella-containing vacuole (SCV), replication inside SCV and hyper-replication into the cytosol of cells (9).
The relatively low impact of the observed hilD mutation on the interaction of S. Derby with swine cells compared with the high impact on human cells suggests that the mutation of SPI-1 genes could be a molecular mechanism for S. Derby to further specialize for the swine host while losing pathogenicity for humans.
In line with this hypothesis, we further observed that the specific lineage with the loss-of-function mutation in hilD also carried stop mutations in the coding sequences of both sipA, encoding the SPI-1 virulence effector SipA, and hilC, encoding the SPI-1 minor transcriptional activator HilC.
SipA is a multifunctional protein that takes part in each step of the enterocyte colonization, i.e., enterocyte invasion, vacuolar survival and replication, and cytosolic hyper-replication (10)(11)(12).Once inside epithelial cells, SipA is cleaved by cellular caspase-3 into two functional domains: the actin-binding C-terminal domain (SipA 426-685 ) involved in cell invasion (13) and the N-terminal domain (SipA 1-425 ) involved in Salmonella survival inside the SCV and in gut mucosa inflammation, by stimulating trans-epithelial migration of polymorphonuclear neutrophils (14)(15)(16).The observed stop codon in sipA is located at the amino acid position 350, resulting in a truncated N-terminal domain.
HilC is an AraC/XylS transcriptional regulator that, together with the other AraC/ XylS regulators HilD and RtsA, controls expression of the master SPI-1 activator hilA (6,17).The system is controlled primarily by HilD, whereas HilC and RtsA amplify and accelerate SPI-1 gene expression (18).HilC has a two-domain structure comprising an N-terminal regulatory domain and a C-terminal highly conserved DNA-binding domain with two helix-turn-helix motifs corresponding to amino acid intervals 212-233 and 259-282 (UniProtKB entry E1WAB2).The observed stop codon in hilC is located at amino acid position 209 of 295 in the region encoding the C-terminal DNA-binding domain, upstream of the first helix-turn-helix motif.
To contribute to the understanding of the molecular mechanisms of S. Derby host adaptation, we investigated the diffusion of the observed mutations in SPI-1 genes in the global population of available genomes of S. Derby from swine (ca.1490 genomes) as reported in the Enterobase database.Furthermore, we assessed the impact of such mutations on virulence by quantifying invasion, vacuolar load, and cytosolic hyper-rep lication in both human and swine intestinal epithelial cells, as these in vitro pheno types have been associated with different pathological outcomes in vivo.In particular, Salmonella survival inside the SCV has been associated with systemic and enduring infections (19,20) while the cytosolic hyper-replication triggers Salmonella intestinal expansion and gut inflammation (21,22).
Moreover, we searched for other mutations in the SPI-1 genes of available S. Derby genomes and evaluated if multiple S. Derby lineages underwent converging evolution in their adaptation to swine through SPI-1 loss of function.

S. Derby-carrying stop mutations in sipA and hilC represent a diffused and expanding lineage in pigs while remaining rare among humans
To evaluate the abundance of the described stop mutations in sipA and hilC in the global population of S. Derby, we analyzed a set of 1,490 S. Derby genomes of swine origin from Europe and North America (2000-2023 time period) belonging to Sequence Type 40 (ST40) available in the Enterobase database.Sequence type ST40 represents the most diffused ST of S. Derby in swine (4).We found that 24.9% of the ST40 genomes from swine carried both the stop mutations in sipA (sipA stop allele) and hilC (hilC stop allele).No isolate was found with only one of the two mutations.Moreover, the neighbor-joining tree of the ST40 genomes from swine, generated under the Enterobase cgMLST v2 HierCC v1 scheme (23), highlighted that all the genomes carrying the sipA and hilC mutations represented different branches of a single large lineage of the population (see red-encircled nodes in Fig. 1A).This hilC stop and sipA stop -carrying lineage included the 15 isolates carrying also the loss-of-function mutation in hilD previously characterized (5).The logistic regression performed on the occurrence of the genomes carrying sipA stop and hilC stop in the ST40 population from swine as a function of time (considering also the continent of isolation) showed a significant increase of the fraction of mutation-carrying genomes over time, both in Europe and North America (Fig. 1B).Moreover, the logistic regression showed that the proportion of mutated genomes was significantly higher in North America (28.4%) compared with Europe (8.4%),P < 0.0001.Furthermore, using the data from Enterobase, we compared the proportion of ST40 genomes carrying the mutations isolated from swine with that observed in the 706 ST40 genomes isolated from humans in Europe and North America in the same period.We found that the proportion in humans was significantly lower than that in pigs (P < 0.0001) both in Europe (4.3% vs 8.4%) and North America (12% vs 28.4%).

The stop mutation in hilC does not alter SPI-1 expression
The stop mutation in hilC generated an activator unable to do its function because of the loss of its DNA-binding domain; therefore, the mutation corresponded to the loss of the entire hilC gene.To assess the effect of the loss of hilC, we analyzed the expression of SPI-1 genes in strain ER1175, naturally carrying the wild-type S. Derby alleles of hilC and sipA, ER1175 deleted for hilC (ER1175ΔhilC) and the strain N11, a representative of the group of isolates naturally carrying hilC stop and sipA stop .For the expression analysis, strains were grown to the early stationary phase as it was previously reported that in this phase, SPI-1 is highly transcribed and significantly downregulated in ER1175 deleted for hilD (encoding the SPI-1 master activator) compared with the wild-type strain (5,24).The expression of SPI-1 genes was not significantly reduced in ER1175ΔhilC and N11 compared with ER1175, indicating that the loss of hilC does not have an appreciable impact on SPI-1 expression in the tested conditions (Table S1).These data are in line with the already reported minor role of HilC in the activation of SPI-1 expression (18).

The stop mutations in sipA and hilC impair the ability of S. Derby to infect human cells
Based on the limited diffusion of hilC stop and sipA stop -carrying S. Derby among humans despite its considerable and increasing circulation in swine, we hypothesized that the two stop mutations in sipA and hilC could be associated with possible weakening of the infection mechanisms in the human host and thereby explain the reduced risk for this species observed at population scale.We thus performed cell culture infection assays on human-derived INT-407 intestinal epithelial cells.Automated fluorescence microscopy was used to quantify at the single cell level the extent of invasion, vacuolar load, and cytosolic hyper-replication (25) (Fig. 2A).Strain ER1175, naturally carrying the wild-type S. Derby alleles of hilC and sipA, and strain N11, a representative of the hilC stop and sipA stop -carrying S. Derby, were tested together with ER1175 mutants generated to analyze the impact of each mutation, singularly or in association, on the infection of human cells.The reference strain S. Typhimurium SL1344 was included in the analysis as positive control of infection (Fig S1).
The extent of invasion was calculated as the fraction of infected cells, namely, the proportion of infected cells with ≥0.2% of their area occupied by Salmonella.N11 invaded a significantly lower fraction of cells than ER1175 (0.63 vs 0.90) (Fig. 2B).To determine the specific effect of the stop mutation in sipA on invasion, an ER1175 mutant carrying sipA stop was produced (ER1175::sipA stop ).Also, the deletion mutant ER1175ΔsipA was included in the analysis to evaluate if the sipA stop mutation, causing the loss of expression of the C-terminal domain specifically involved in invasion, conferred the same phenotype as the loss of the entire gene.The results showed that the introduction of sipA stop , as well as the deletion of sipA, did not significantly reduce the fraction of cells infected by ER1175.These results are consistent with the SipA cooperative and redundant role in promoting Salmonella entrance into the host cell (26).The effect of the stop mutation on hilC, corresponding to the loss of the entire gene, was assessed testing the invasion ability of ER1175ΔhilC.The deletion of hilC did not significantly reduce invasion compared with that of wild-type ER1175, in line with the already observed minor effect of HilC on activating SPI-1 expression (Table S1) and with the mild effect of the mutation of hilC on S. Typhimurium invasion (27).A mutant deleted for both sipA and hilC (ER1175ΔsipAhilC) was then tested to evaluate the combined effect of losing both genes, and we observed that ER1175ΔsipAhilC infected a significantly lower fraction of cells compared with ER1175 (0.75 vs 0.90).This result, together with the reduced invasion of N11, indicates that only the double impairment of sipA and hilC can lead to reduced invasion ability, suggesting an additive effect of the two mutations on invasion.
The vacuolar load was quantified as the mean of the percentages of cellular areas occupied by vacuolar Salmonella.Two SPI-1-regulated events contribute to the magnitude of this phenotype, the internalization of multiple Salmonella in a SCV (28), and the survival and replication inside the SCV (29-31) (Fig. 2C).The same pattern observed for invasion was detected for the vacuolar load, with only ER1175ΔsipAhilC and N11 showing a significantly lower mean vacuolar load compared with ER1175 (2.01 and 1.44 vs 2.78).Moreover, strain N11 showed an invasion level and vacuolar load even lower than ER1175ΔsipAhilC, suggesting that also other genetic features of the hilC stop and sipA stop -carrying S. Derby are involved in these phenotypes.
The cytosolic hyper-replication was scored as the proportion of infected cells massively colonized by cytosolic Salmonella, defined as infected cells with ≥20% of their area occupied by cytosolic Salmonella.Not only the double-mutation strains N11 and ER1175ΔsipAhilC but also the sipA-only impaired strains ER1175::sipA stop and ER1175ΔsipA showed reduced ability to hyper-replicate in the cytosol of infected cells compared with ER1175 (Fig. 2D), in accordance with the known involvement of SipA in mediating intra-cytosolic survival and hyper-replication (12,32).The same level of reduced hyper-replication observed for both ER1175::sipA stop and ER1175ΔsipA demonstrates that the stop mutation in sipA generates an effector unable to exert its function.Conversely, ER1175ΔhilC showed no reduction of the hyper-replication rate compared with ER1175, consistently with the unaltered sipA expression observed in hilC-impaired strains ER1175ΔhilC and N11 (Table S1).In this case, N11 showed the same reduction in hyper-replication as sipA-only impaired strains, indicating that the stop mutation in sipA by itself is responsible for the lower ability of N11 to hyper-replicate in the cytosol of host cells.Overall, these results showed that the two stop mutations of sipA and hilC reduce virulence of S. Derby in human cells and likely contribute to explain the reduced zoonotic risk observed in the population for the lineage carrying the two mutations.

The truncated N-terminal domain of SipA retains the ability to induce IL-8 expression in human cells
The injection of SPI-1 effectors, and primarily SipA, inside intestinal epithelial cells is required for Salmonella to fully induce the expression of chemokines like IL-8 that stimulates the inflammatory response by inducing polymorphonuclear neutrophilic leukocyte migration, ultimately leading to enteritis and diarrhea (33,34).We measured the level of IL-8 expression induced in human epithelial cells infected by ER1175, ER1175 mutants, and N11 to evaluate the impact of the stop mutations in sipA and hilC on S. Derby ability to induce the inflammatory response (Fig. 2E).The invA-deleted mutant of ER1175 (ER1175ΔinvA) was included in the analysis as negative control of IL-8 induction because invA encodes an essential component of the T3SS-1 (35) and its lack impairs the ability to inject SPI-1 effectors in the host cells and consequently to invade and induce expression of chemokines.
The infection with ER1175 induced the strongest increase in IL-8 expression (6.75-log 2 fold change) in human cells as opposed to ER1175ΔinvA which caused the weakest increase in IL-8 expression (1.28-log 2 fold change), as expected.Compared with ER1175, ER1175::sipA stop showed no significant defect in inducing IL-8 expression.This result indicates that the truncated N-terminal domain of SipA generated by the stop mutation retains the ability, albeit reduced, to induce the inflammatory response.Coherently, the effector encoded by sipA stop still conserves part of the 131-amino acid region of the N-terminal domain (amino acids 294-424) involved in the induction of pro-inflammatory genes (14).The ER1175ΔhilC mutant showed no decrease in IL-8 induction compared with ER1175, which is consistent with the lack of any significant effects of hilC mutations on SPI-1 gene expression (Table S1).Otherwise, the expression of IL-8 in human cells infected by ER1175ΔsipA (4.95-log 2 fold change) as well as ER1175ΔsipAhilC (4.45-log 2 fold change) was significantly lower than that observed for ER1175, in line with the known role of SipA in inducing IL-8 expression.Overall, these results showed that the stop mutations in sipA and hilC individually have no significant effect on reducing the inflammatory response.N11 showed a significant defect in inducing IL-8 expression (4.85-log 2 fold change) compared with ER1175.The reason for this behavior remains not fully explained, but it can be hypothesized that the copresence of the stop mutations in both sipA and hilC affects this phenotype.Notably, N11 was shown to have a lower invasion level (Fig. 2B) compared with ER1175 and a link between invasion defect and the diminished IL-8 induction has been reported (36).
The enhanced adaptation to swine of S. Derby carrying hilC stop and sipA stop goes with the loss of virulence in swine enterocytes S. Derby, a swine-adapted serovar, usually causes asymptomatic infection in swine, as opposed to the generalist serovar S. Typhimurium that can cause enterocolitic forms (37).This difference in pathogenesis in vivo between the two serovars is reflected in a reduced ability of S. Derby to invade and replicate in swine IPEC-J2 intestinal epithelial cells in vitro compared with S. Typhimurium (5).Therefore, we evaluated if S. Derby carrying hilC stop and sipA stop , which appears well adapted to pigs considering its diffusion and expanding trend, was characterized by a further reduction in the ability to colonize swine enterocytes and if the reduction was due to the stop mutations in sipA and hilC.For this, the colonization steps of the swine IPEC-J2 intestinal epithelial cells were analyzed at a single-cell level for S. Typhimurium SL1344, ER1175, ER1175ΔsipAhilC, and N11 (Fig. 3A).
Comparing the different strains for their invasion ability in IPEC-J2 cells (Fig. 3B), a significant reduction was observed for all S. Derby strains compared with S. Typhimurium SL1344, with ER1175, ER1175ΔsipAhilC, and N11 infecting a fraction of swine cells of 0.39, 0.25, and 0.03, respectively, compared with 0.73 of S. Typhimurium.The lower invasion ability of ER1175ΔsipAhilC compared with ER1175 demonstrates that the two genes have a role in the infection of swine cells as well as in human cells.N11 infected swine cells even less than ER1175ΔsipAhilC, indicating that genetic signatures other than the stop mutations in sipA and hilC are involved in its phenotype.
The differences between the strains observed for invasion were reported also for the vacuolar load (Fig. 3C) with S. Typhimurium SL1344 presenting the highest vacuolar load, followed by ER1175, ER1175ΔsipAhilC, and N11 in decreasing order.
A strong reduction in the fraction of infected cells and in the vacuolar load was observed in swine cells compared with human cells for all S. Derby strains (−0.51 for ER1175, −0.50 for ER1175ΔsipAhilC, and −0.60 for N11).This can be partially explained by the fact that the IPEC-J2 cell line is more resistant to Salmonella infection, compared with human cell line (38).However, for S. Typhimurium SL1344, the difference in the fraction of swine vs human cells infected was only −0.24, indicating that the marked reduction observed for S. Derby strains was a serotype-specific feature.The mean vacuolar load, as well as the percentage of infected cells, was generally lower in swine cells compared with human cells.
The hyper-replication rate was not measurable for S. Derby strains in swine cells because of the low frequency of this phenotype along with the low infection rate in swine cells.We were able to quantify hyper-replication only for S. Typhimurium SL1344, which infected a percentage of swine cells much higher than S. Derby strains (Fig S2).

Other lineages of S. Derby carry deleterious mutations in sipA and other SPI-1 genes
Our results showed that the stop mutations in sipA and hilC reduced the virulence of S. Derby both in human and in swine cells.This attenuation is consistent with the limited proportion of human cases associated with this genotype; at the same time, it does not seem to negatively affect its circulation in swine.Under the hypothesis that this mutation process could be a more generalized phenomenon, we looked for other deleterious mutations in sipA and hilC in the framework of adaptation to swine.The same Enterobase data set of genomes used to identify the stop mutations in sipA and hilC was used for this search.No hilC alleles other than that described in this study were found carrying potentially deleterious mutations in the S. Derby ST40 population isolated from European and North American pigs.Conversely, 79 genomes (5.3% of the population) carrying a truncated (68 genomes) or missing (11 genomes) version of the sipA gene and belonging to a unique lineage (see blue encircled nodes in Fig. 4) were detected.Analysis of the SPI-1 sequence of the 68 S. Derby genomes with truncated sipA (Table S2) revealed the presence of an insertion sequence (IS) inside sipA belonging to the IS10 subgroup of the IS4 family.Insertion sequences are known to play an important role in bacterial evolution as transposition inside a gene can potentially inactivate it (39).The IS was located at the sipA nucleotide position 455/2058, corresponding to amino acid residue 152, upstream of the first fragment along the SipA amino acid sequence (defined as amino acids 294-424) known to have a role in Salmonella pathogenesis (14).Consequently, it is very likely that the IS caused the generation of a non-functional protein.In addition, in 49 out of the 68 genomes with truncated sipA, a second IS, belonging to the IS10 subgroup of the IS4 family as the one found in sipA, was detected inside the hilD gene, at nucleotide positions 116/930 and 486/930, corresponding to amino acid positions 39/310 and 162/310, respectively.In both cases, the IS was located upstream of the HilD DNA binding domain, very likely generating a non-functional protein (17).The analysis of SPI-1 in the 11 genomes lacking sipA (Table S2) revealed, in 10 genomes, a deletion between nucleotide position 116/930 of hilD and nucleotide position 455/2058 of sipA.In more detail, the initial part of the same IS previously observed in sipA and hilD was found downstream of hilD nucleotide position 116/930 and its final part was found upstream of sipA nucleotide position 455/2058.Notably, the deletion starts and ends at same positions where we observed IS insertions in sipA and hilD which generated truncated alleles.This observation allows hypothesizing that the deletion was likely the result of the two IS located in sipA and hilD which mobilized the SPI-1 region located between them.Indeed, IS-related genomic deletions were largely documented in Salmonella as well as in many other bacteria (39)(40)(41).Furthermore, in seven of the genomes carrying an IS4 insertion in sipA/hilD genes, a further IS4 insertion was detected in other genes belonging to SPI-1.We next searched for S. Derby genomes carrying the observed IS insertion in sipA among the 706 ST40 genomes from humans already selected from Enterobase.Only one S. Derby genome from humans was found to carry the same insertion (Table S2), strongly indicating that the lineage is rarely represented in S. Derby isolated in humans.
Overall these findings, together with those on the stop mutation in sipA and hilC, demonstrate the existence of S. Derby lineages with diverse deleterious mutations in SPI-1 genes, suggesting that distinct evolutionary events resulting in the loss of function of SPI-1 genes occurred in the population of S. Derby.

DISCUSSION
Different serovars of Salmonella and different lineages within the same serovar can differ in their host range.An example is represented by the generalist serovar Typhimurium, which includes pathovariants adapted to different hosts (42).Similarly, in a previous study, we identified a specific lineage of S. Derby widespread in swine but significantly under-represented in humans compared with the other S. Derby lineages, indicating its adaptation to swine associated with reduced virulence to humans.This lineage showed reduced ability to infect enterocytes in vitro compared with other S. Derby.The reduction was much higher in human cells than in swine cells, and this behavior was due to a loss-of-function mutation in HilD which caused loss of SPI-1 expression.These in vitro results, together with the significantly higher presence of that lineage in swine than in humans, suggest that SPI-1 is crucial to cause disease in humans, but not essential for S. Derby to circulate in swine, its main host.Accordingly, the literature reports that evolution to host adaptation leads to the loss of functions that are not essential to colonize a specific host (43).Based on this, we further investigated the HilD-impaired lineage detecting two other mutations in SPI-1 genes, namely, a stop mutation in sipA and one in hilC, indicating that SPI-1 is subjected to genetic decay in this lineage.
The present study shows that the previously characterized lineage, carrying the hilD loss-of-function mutation together with sipA and hilC stop mutations, represents part of a more widespread lineage carrying only the sipA and hilC stop mutations.This large lineage encompasses c.a. 25% of the S. Derby ST40 genomes of swine origin in Enterobase, representing a significant proportion of S. Derby from this host.Moreover, this lineage appears to be growing over time among both European and American pigs, while its presence among human cases of salmonellosis is half of that in pigs in both continents.These epidemiological data show that a lineage of S. Derby accumulating mutations in genes of SPI-1 is expanding in the swine population and is weakening its virulence to humans.The in vitro study of the hilC stop and sipA stop -carrying lineage in human cells shows that introduction of sipA stop into a wild-type S. Derby reduces cytoplasmic hyper-replication at the level observed for the hilC stop and sipA stop -carrying lineage represented by strain N11.The stop mutation in sipA confers the same reduction in cytoplasmic hyper-replication of the deletion of the entire gene, indicating that the missing part of SipA due to the stop mutation is that involved in this phenotype.Cytosolic hyper-replication plays an important role in gut inflammation by inducing extrusion of infected epithelial cells out of the monolayer, inflammatory cell death, and release in the gut lumen of invasion-competent Salmonella which infect secondary cells (21).Therefore, the reduced ability of the hilC stop and sipA stop -carrying lineage to hyper-replicate in human cells due to the stop mutation in sipA could contribute to the reduced virulence for humans of this lineage.
Regarding the invasion and vacuolar load, N11 has significantly reduced values compared with the wild-type S. Derby.N11 showed even lower invasion and vacuolar load than the still-attenuated wild-type strain deleted for both sipA and hilC, indicating that these phenotypes likely entail also additional genetic features than the observed stop mutations in sipA and hilC.In line with this, we found non-synonymous/stop mutations other than those in hilC and sipA in N11 (Table S3).Some mutations were shared by all genomes belonging to the hilC stop and sipA stop -carrying lineage; others were found only in a sub-population of that lineage.Three non-synonymous/stop mutations were located in citB, ssaC, and narZ genes specifically induced in intracellu lar S. Typhimurium (44) and can be involved, likewise hilC and sipA, in the analyzed phenotypes.The induction of expression of the pro-inflammatory cytokine IL-8 by N11 is significantly reduced compared with wild-type S. Derby, and this phenotype could not be fully explained by the tested mutants.Only the sipA-deleted mutant, not the mutant carrying sipA stop , shows reduced IL-8 expression, indicating that the truncated N-terminal domain generated by the stop mutation retains the ability to induce the inflammatory response.The reduced induction of IL-8 expression by N11 requires further investigation.
Overall, these results indicate attenuation for human cells in vitro of the lineage of S. Derby carrying the stop mutations in sipA and hilC, which is expanding in the swine host while showing limited virulence to humans.
With regard to swine cells, we observed a progressive decrease in invasion level and vacuolar load moving from S. Typhimurium to S. Derby without SPI-1 mutations, to S. Derby missing hilC and sipA genes, and to S. Derby belonging to the hilC stop and sipA stop -carrying lineage.The latter lineage thus confirms its attenuation also in swine cells.Notably, a strong reduction in invasion and in the vacuolar load was observed in swine cells compared with human cells for all S. Derby strains compared with S. Typhimurium.In addition, hyper-replication was only measurable for S. Typhimurium and not for S. Derby strains because of its too low frequency.The infection of the gut mucosa and especially cytosolic hyper-replication are crucial steps in inducing inflammation and thus enteritis; therefore, this in vitro evidence is in line with the fact that in vivo S. Derby causes asymptomatic infection in swine whereas S. Typhimurium can cause enterocolitic forms (36).Our findings in swine cells confirm this previous general knowledge on reduced infectivity of S. Derby and add that a widespread lineage belonging to this serovar has even decreased infectivity.This suggests that S. Derby does not need to invade the swine intestinal epithelium with high effectiveness in order to circulate in this host and thus can accumulate loss-of-function mutations in genes involved in invasion such as those in SPI-1.As a confirmation of this finding, a second S. Derby lineage was found with IS insertion causing truncation and/or loss of one or more SPI-1 genes, indicating that the S. Derby population is undergoing converging evolution in the adaptation to swine through loss of function of SPI-1 and the associated attenuation.Salmonella can infect and persist in pigs in the intestine, gut-associated lymphoid tissue, and tonsils (37,45).For example, tonsils are known as an important reservoir of Salmonella as it can persist in that compartment for long periods.Boyen et al. (45) demonstrated that S. Typhimurium SPI-1 genes promote intestinal but not tonsillar colonization in pigs.We therefore hypothesize that the infection of tonsils or other lymphoid districts could be a way for S. Derby to persist and spread in pigs.
Overall, this study shows how a host-adapted Salmonella serovar, S. Derby, is further adapting to its main host, swine, by reducing its virulence in intestinal epithelial cells.The literature reports that Salmonella serovars have adapted to specific hosts by losing the ability to replicate in the intestine of those hosts in favor of disseminating to systemic sites (46).However, serovars usually defined as host adapted, like the swine-adapted S. Choleraesuis and the bovine-adapted S. Dublin, are associated with high virulence, including mortality, in their respective host reservoirs (47) compared with generalist serovars, whereas S. Derby causes asymptomatic infection in swine.Therefore, two different routes to host adaptation seem to exist in Salmonella, one leading to increased virulence for the target host, the other one leading to attenuation.

Analysis of the structure of the global population of S. Derby ST40
The population structure of S. Derby ST40 was analyzed by using the whole-genome sequences and tools available in the Enterobase database (https://enterobase.war wick.ac.uk), accessed on 7 April 2023.Specifically, to analyze the whole-genome sequencing data, the allelic profile obtained through core-genome MLST (cgMLST) analysis using the cgMLST V2 + HierCC V1 scheme was used (23).The cgMLST data were obtained in Enterobase by selecting the sequences in the database with the release date preceding the accessed date through the following query: the genomes (i) were assigned to sequence type ST40 using the "Achtman 7 Gene MLST" scheme for Salmonella, (ii) had a year of isolation between 2000 and 2023 reported in Enterobase's "Collection Year" field, (iii) could be assigned to a known country of isolation within the European or North American continents, per Enterobase's "Country" and "Continent" fields, respectively, and (iv) could be assigned to the source of isolation "Swine" or "Human, " per Enterobase's "Source Type" and "Source Niche" fields, respectively.
The genomes carrying the described stop mutation in the sipA gene were identi fied through the presence of the allele coded "671" in locus "STMMW_28441" of the cgMLST V2 scheme.The genomes carrying the described stop mutation in the hilC gene were identified through the presence of the alleles coded "671" and "835" in locus "STMMW_28291" of the cgMLST V2 scheme.The population tree of S. Derby ST40 in swine was constructed in the Enterobase workspace using the GrapeTree option, where the Ninja neighbor-joining algorithm was employed under the cgMLST V2 + HierCC V1 scheme.
We assessed the temporal trend in the occurrence of the investigated mutations in sipA and hilC genes in the S. Derby ST40 population in swine by fitting a generalized linear model with binomial error distribution (logistic regression) using the "Collection Year" and the "Continent" as explanatory variables.Moreover, we assessed whether the occurrence of the investigated mutations in sipA and hilC genes is different between S. Derby isolated in swine and S. Derby isolated in human by fitting a logistic regression using the source of isolation and the continent as explanatory variables.
The detection of other lineages of S. Derby carrying potentially deleterious mutations in hilC and sipA gene was performed investigating the allelic variants observed in the "STMMW_28291" corresponding to hilC and "STMMW_28441" corresponding to sipA loci of the cgMLST V2 scheme within the population of the ST40 genomes extracted from Enterobase.Since we were interested in detecting potentially deleterious mutations in the sipA gene that were reasonably diffused in the ST40 population, we genetically characterized only the allelic variants in locus "STMMW_28441" observed in more than the 5% of the genomes.We considered as deleterious mutations missense mutations, stop mutations, and mutations generating a truncated or missing allele.

Bacterial strains
Bacterial strains used in this study are listed in Table S4.Bacteria were cultured in Luria Bertani (LB) Miller medium supplemented with appropriate antibiotics (ampicillin 100 µg/mL, kanamycin 50 µg/mL, and chloramphenicol 20 µg/mL) when needed.

Construction of recombinant strains
Plasmids and primers used in this study are listed in Tables S5 and S6, respectively.Gene deletion and allelic exchange were made using the bacteriophage λ red recombinase system (48).For gene deletion, genes were replaced with a kanamycin or chlorampheni col resistance cassette (kan or cat) amplified from pKD4 and pKD3 template plasmids, respectively.For allelic exchange, constructs were generated by overlap PCR to contain sipA stop and the kan resistance cassette from the pKD4 plasmid.Overlap PCR and transformation were made according to Tambassi et al. (5).

Automated analysis of Salmonella intracellular phenotypes
The bacterial infection of INT-407 and IPEC-J2 cells for the quantification of Salmonella intracellular phenotypes was performed according to Berni et al. (25).Briefly, INT-407 and IPECJ-2 cells were seeded at a density of 3 × 10 4 and 1 × 10 4 cells/well, respectively, in antibiotic-free media 20-24 hours prior to infection in 96-well imaging plates (CellVis) coated with collagen I from rat tail (Invitrogen).Salmonella strains used for the infec tion were transformed with the pCHAR-Duo reporter plasmid that was kindly provided by Dr. Olivia Steele-Mortimer.The pCHAR-Duo plasmid allows to distinguish vacuolar from cytosolic Salmonella through the differential expression of mCherry and GFP (49).Salmonella strains were cultured statically for 20 hours at 37°C in LB supplemented with ampicillin 100 µg/mL to reach the stationary phase of growth.Then, 100% of confluent monolayers were washed with PBS and infected for 1 hour with ~1.5 × 10 5 bacteria/mm 2 at 37°C in 5% CO 2 with a breathable sealing membrane (Diversified Biotech BEM-1).A multiplicity of infection (MOI) based on the growth surface, rather than on the epithelial cell number, was applied to expose INT-407 and IPEC-J2 confluent monolayers to the same bacterial load, despite the different size of the two cell lines.After 1 hour of infection, monolayers were washed with PBS, treated with gentamicin 100 µg/mL for 1 hour, then washed again with PBS, and treated with gentamicin 10 µg/mL for the remaining time course of the infection.
After 8 hours of infection, monolayers were washed with PBS and fixed with paraformaldehyde 4% (Sigma-Aldrich) for 20 min at room temperature.Finally, epithelial cells were labelled with the HCS CellMask Blue cytoplasmic⁄nuclear stain (Invitrogen) following the manufacturer's instructions.
Samples were imaged with a motorized Axio Observer Inverted Microscope with Colibri 5/7 light source (ZEISS) using 40×/0.75NA objective for INT-407 and 20×/0.8NA objective for IPEC-J2 because the two cell lines have different size.At least 1,000 epithelial cells were pictured for each biological replicate.The automated fluorescence image analysis was performed with FIJI (NIH) (50).The image analysis workflow and scripts for the automation were described in detail by Berni et al. (25).Briefly, epithelial cells were first segmented and defined as a region of interest (ROI), uniquely labelled with their y-x coordinates.Then, the area and the percentage of the ROI area occu pied by Salmonella expressing the mCherry constitutive reporter only or also the GFP cytosol-responsive reporter were measured for each ROI.The percentage of the cell area occupied by mCherry-only expressing Salmonella was used to calculate the percentage of infected cells (considering only ROIs with a percentage of occupied area > 0.2%) and the mean vacuolar load.The percentage of the cell area occupied by GFP-expressing Salmonella was measured to quantify the cytosolic hyper-replication rate scored as the fraction of infected cells massively colonized by cytosolic Salmonella (considering only cells with a percentage of the occupied area ≥20% and ≥15% for human and swine cells, respectively).The statistical analysis on the microscopy data (i.e., fraction of infected cells, vacuolar load, and hyper-replication rate) was performed by using linear mixed models where the samples act as fixed variables and the technical repetitions within the same biological repetition act as random effects (pseudo-replication).The P-values were obtained from the t-statistics using Satterthwaite's method with Bonferroni post-hoc correction.

RNA extraction and cDNA synthesis
RNA was extracted from Salmonella strains grown in in vitro SPI-1-inducing conditions and from epithelial cells infected by Salmonella strains.
To reach the SPI-1-inducing conditions in vitro, Salmonella strains were grown in 2-mL LB broth for 18 hours at 37°C at 220 rpm, then diluted 300 µL in 10 mL of fresh LB, and grown at 37°C at 200 rpm for ~2 hours until the early stationary phase of growth was reached (corresponding to OD 600 = 2.0) (5,24).Then, 500 µL of bacteria culture was treated with RNA protect (Qiagen) and RNA was extracted using the Nucleospin RNA extraction kit (Macherey Nagel) following the manufacturer's instructions.
For RNA extraction from Salmonella-infected cells, INT-407 and IPEC-J2 were cultured in antibiotic-free media 20-24 hours prior to infection in 24-well plates coated with collagen I from rat tail (Invitrogen).Seeding densities of 3 × 10 5 and 1.5 × 10 5 cells/ well were used to reach 100% confluency.Salmonella isolates were grown statically for 20 hours at 37°C in LB broth.Confluent monolayers were washed with PBS and infected with a MOI based on the growth surface as described previously.After 1 hour of infection, the inoculum was removed and fresh medium was added until the end of the infection (33).After 4 hours of infection, medium was removed and RNA was extracted using the Nucleospin RNA extraction kit (Macherey Nagel) by immediately adding 350 μL of RA1, following the manufacturer's protocol.
For both bacterial and eukaryotic-extracted RNA, in-column and in-solution DNase digestion was performed following Nucleospin RNA extraction kit protocol.Then, RNA was precipitated by mixing with sodium acetate (0.1 vol, 3M), ethanol (2.5 vol, 100%), and glycogen (5 µg/sample).After overnight precipitation at −20°C, RNA was washed twice with 750 µL of 70% ice-cold ethanol and resuspended in nuclease-free water.RNA concentration and quality (A 260 /A 280 ) were determined with a Synergy H1 Hybrid spectrophotometer (BioTek) equipped with Gen5 software.The extracted RNA was reverse transcribed using the High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems).

Gene expression analysis
Gene expression analyses were performed by quantitative real-time PCR (qRT-PCR) using the kit GoTaq qPCR Master Mix (Promega Corporation) and the Biorad CFX-96 thermocy cler.Each reaction was performed in two technical replicates for each sample.Primer sequences are indicated in Table S5.
The differential expression of eight SPI-1 genes representative for transcriptional regulators, effectors, needle-complex structural component, and chaperone was calculated for ER1175ΔhilC and N11 vs ER1175.Expression of target genes was normal ized to the reference gene gmk.
INT-407 cells infected by Salmonella strains were analyzed for IL-8 expression.Expression of IL-8 was normalized to the reference gene GADPH.Uninfected epithelial cells were used as control condition.The "Common Base Method" described by Ganger et al. (51) was used for the analysis of qRT-PCR data.Briefly, the Ct values were nor malized for the reaction efficiency (E) as follows: log2(E)•Ct.Differences in gene expres sion among samples were analyzed using unpaired t-test on efficiency-weighted ΔCt assuming unequal variances.

Bioinformatic analysis
S. Derby genome assemblies with missing or truncated sipA according to the allele calling of cgMLST were downloaded from Enterobase.Contigs of each S. Derby were aligned to the SPI-1 sequence of S. Typhimurium SL1344 (GenBank FQ312003.1,from sitA CDS to invH CDS) using MAUVE through Geneious 11.1.5(https://www.geneious.com)to visualize the sipA gene in the context of the whole SPI-1 sequence.The nucleotide sequences found inserted in the sipA gene and in other SPI-1 genes were subjected to BLASTN using the ISfinder database (http://www-is.biotoul.fr)(52).

FIG 1
FIG 1 The hilC stop and sipA stop -carrying lineage in the context of the S. Derby population from swine and its occurrence over time in this host.Analysis was performed on 1,490 genomes of S. Derby ST40 isolated in swine in Europe and North America in the 2000-2023 time period available in Enterobase (accessed date 7 April 2023).(A) Neighbor-joining tree based on S. enterica cgMLST allelic profiles.Nodes encircled in red represent the S. Derby genomes carrying the stop mutations in sipA and hilC genes.(B) Occurrence of S. Derby genomes displaying the stop mutations in sipA and hilC genes (open dots) as a function of time, i.e., isolation year.The solid red and blue lines represent the fraction of genomes over time displaying the stop mutations in North America and Europe, respectively, estimated through the logistic regression model.Dashed lines represent the confidence intervals of the estimates.

FIG 2
FIG 2 Derby interaction with human epithelial cells.(A) Representative image of INT-407 cells infected with S. Derby ER1175 carrying the pCHAR-Duo fluorescence reporter plasmid.Epithelial cells are shown in blue (HCS CellMask Blue), intracellular Salmonella in red (mCherry), and cytosolic hyper-replicating (Continued on next page)

FIG 2 (
FIG 2 (Continued) Salmonella in green (GFP).White scale bars are 50 µm.The extent of invasion, vacuolar load and cytosolic hyper-replication for ER1175, its isogenic mutants, and N11 was calculated.(B) The fraction of infected cells was obtained by dividing the number of cells with a percentage of area occupied by mCherry-only expressing Salmonella > 0.2% by the total number of cells.(C) The mean vacuolar load was obtained by calculating the mean percent area occupied by mCherry-only expressing Salmonella in the infected cells.(D) The hyper-replication rate was calculated by dividing the number of cells with a percentage of the area occupied by GFP-expressing Salmonella ≥ 20% by the total number of infected cells.Data shown in B, C, and D are the pooled data of three biological replicates with three technical replicates.Horizontal bars indicate the mean of biological replicates.Vertical bars indicate standard deviation.Asterisks indicate significant difference versus ER1175.(E) IL-8 expression of human cells infected with S. Derby strains analyzed.Each dot represents the log2 fold change of IL-8 expression of infected versus non-infected cells for each biological replicate.Data from three to four biological replicates are reported.Horizontal bars indicate the mean of biological replicates.Vertical bars indicate the standard error of the mean.Tables report P-values from the t-statistics using Satterthwaite's method with Bonferroni post-hoc correction (*P < 0.05, **P < 0.01, and ***P < 0.001; n.s., not significant).

FIG 3
FIG 3 Derby interaction with swine epithelial cells.(A) Representative image of IPEC-J2 cells infected with S. Derby ER1175 carrying the pCHAR-Duo fluorescence reporter plasmid.Epithelial cells are shown in blue (HCS CellMask Blue), intracellular Salmonella in red (mCherry), and cytosolic hyper-replicating Salmonella in green (GFP).White scale bars are 50 µm.The extent of invasion and vacuolar load for S. Typhimurium SL1344 and S. Derby ER1175, ER1175ΔsipAhilC, and N11 was calculated.(B) The fraction of infected cells was obtained by dividing the number of cells with a percentage of the area occupied by mCherry-only expressing Salmonella > 0.2% by the total number of cells.(C) The mean vacuolar load was obtained by calculating the mean percent area occupied by mCherry-only expressing Salmonella in the infected cells.Data shown in B and C are the pooled data of two to three biological replicates with three technical replicates.Horizontal bars indicate the mean of biological replicates.Vertical bars indicate standard deviation.Tables report P-values from the t-statistics using Satterthwaite's method with Bonferroni post-hoc correction (*P < 0.05, **P < 0.01, and ***P < 0.001).

FIG 4
FIG 4 Neighbor-joining tree of the ST40 S. Derby genomes of swine origin employed under the cgMLST v2 HierCC v1 scheme.Circles represent S. Derby genomes analyzed.Genomes carrying the sipA and hilC stop mutations are highlighted in red; genomes with sipA truncated or missing are highlighted in blue.