Salmonella persistence in soil depends on reciprocal interactions with indigenous microorganisms

Fresh fruits and vegetables have numerous benefits to human health. Unfortunately, their consumption is increasingly associated with food-borne diseases, Salmonella enterica being their most frequent cause in Europe. Agricultural soils were postulated as reservoir of human pathogens, contributing to the contamination of crops during the growing period. Since the competition with the indigenous soil microbiota for colonization sites plays a major role in the success of invading species, we hypothesized that reduced diversity will enhance the chance of Salmonella to successfully establish in agricultural environments. We demonstrated that the abundance of Salmonella drastically decreased in soil with highly diverse indigenous prokaryotic community, while in soil with reduced prokaryotic diversity, Salmonella persisted for a long period. Furthermore, in communities with low diversity, Salmonella had an impact on the abundance of other taxa. The high physiological plasticity allows Salmonella to use agricultural soils as alternative habitat which might provide a route of animal/human infections. In addition, adjusted transcriptional profile with amino acid biosynthesis and the glyoxylate cycle most prominently regulated suggest an adaptation to the soil environment. Our results underline the importance of the maintenance of diverse soil microbiome as a part of strategy aiming at reduced risk of food-borne salmonellosis outbreaks. This article is protected by copyright. All rights reserved.


Introduction
Salmonella was the most frequent cause of food-related disease in Europe in 2016 (EFSA and ECDC, 2017). Contamination of crop plants in the field can take place via aerial and/or below-ground plant parts (Koukkidis and Freestone, 2018). Several reports showed that human pathogenic bacteria persist in different ecological niches, including soil or organic fertilizer, for extended periods (Islam et al., 2004;Barak and Liang, 2008;Brennan et al., 2014;Jechalke et al., 2019). There is reasonable body of evidence suggesting that Salmonella enterica adapts to different environments such as manure or soil (Semenov et al., 2010;Fornefeld et al., 2017). The fate of Escherichia coli and Salmonella in soil and soil-related environments was investigated in several studies (Ongeng et al., 2014;Mallon et al., 2015;Fornefeld et al., 2017;Wang et al., 2018). Furthermore, a number of studies have shown that the survival and success of an invader in a new (soil) environment depends on various biotic and abiotic factors. For example, the survival of Listeria monocytogenes in soil depends on the acidity and the present microbiota (Locatelli et al., 2013). The soil type, root exudates and organic fertilizer might also influence the persistence of pathogens in soil (Berg and Smalla, 2009;Locatelli et al., 2013;Schlaeppi et al., 2014;Jechalke et al., 2019). However, the relationship between the biotic diversity and biotic resistance towards invading organisms, is not fully understood; it seems that environments with high biological diversity have narrower niches and/or might have a higher chance to harbour natural antagonists and competitors (Chapin III et al., 2000;Loreau and Hector, 2001;Mallon et al., 2015;Li et al., 2018). Accordingly, persistence of E. coli O157:H7 in soil was strongly correlated with a low microbial diversity and a high amount of available resources (Westphal et al., 2011;van Elsas et al., 2012). Nonetheless, a possible failure of an invader cannot lead to the conclusion that the resident microbial diversity remains unchanged. Recently, Mallon et al. (2018) observed shifts in microbial communities after unsuccessful invasions of E. coli.
In this study, we investigated the invasion and persistence of S. enterica in agricultural soil. Moreover, we assessed the influences of reduction in microbial diversity as well as inoculation of S. enterica on the structure of soil prokaryotic community. We hypothesized that higher initial diversity of indigenous soil bacteria would have a negative impact on the invasion and persistence of Salmonella in agricultural soil(s). Furthermore, we hypothesized that inoculation of Salmonella into soil with lower diversity would lead to an enhanced persistence. Finally, we wondered if Salmonella inoculation would shift the prokaryotic community composition. To test those hypotheses, we used an agricultural soil, diluvial sand (DS) soil and three S. enterica strains. We used S. Senftenberg as well as the S. Typhimurium strains 14028s and its derivate LT2, which were isolated from basil, poultry and a common lab strain respectively. The survival of the Salmonella strains was followed by cultivation-dependent and cultivationindependent methods. The diversity of prokaryotic communities in DS soil (untreated DS soil) and DS soil with reduced community (autoclaved DS soil) was analysed by sequencing of 16S rRNA gene amplicons from total community DNA. To evaluate a possible response to the soil environment with high and low microbial diversity we assessed the transcriptome of S. enterica Typhimurium 14028s, exposed to mineral medium and resuspended in untreated and autoclaved DS soil using a newly designed experimental approach.

Salmonella persists better in DS soil with reduced microbial diversity
The persistence of Salmonella was assessed in autoclaved diluvial sand (DS) soil after inoculation with S. enterica serovar Typhimurium strain 14028s (S. Typhimurium 14028s), S. enterica serovar Typhimurium LT2 DSM 18522 (S. Typhimurium LT2) or S. enterica serovar Senftenberg (S. Senftenberg) in three independent experiments ( Fig. 1A-C). In parallel, Salmonella strains were inoculated into untreated DS soil, which served as a control. The initial level of inoculation was about 10 6 , 10 7 and 10 6 colony-forming units/g soil dry weight (CFU (g dw) −1 ) in the first, second and third experimental repetition, respectively. CFU counts revealed that in the untreated DS soil all tested strains declined steadily and were below the quantification limit (10 2 CFU/g soil dry weight (dw)) 63 days post-inoculation (dpi) for S. Typhimurium LT2 as well as S. Senftenberg, and after 77 days for S. Typhimurium 14028s ( Fig. 1; blue lines). In contrast, in the autoclaved DS soil, Salmonella CFU counts increased initially by one to two orders of magnitude to 10 8 CFU (g dw) −1 and even 10 9 CFU (g dw) −1 ( Fig. 1; red lines). After the initial increase, the abundance of Salmonella decreased over the experimental period. However, the Salmonella CFU counts in autoclaved DS soil never dropped below the detection limit (10 1 CFU (g dw) −1 ) and only in the first experiment, S. Senftenberg CFU counts were below the quantification limit at 161 dpi ( Fig. 1; red diamonds). These results, reproducible in all three experiments, suggest that the persistence of Salmonella in autoclaved DS soil was enhanced compared with the untreated soil. Furthermore, we did not observe significant differences between the persistence of different strains (Fig. 1, Supporting Information Tables S1A and S2). The CFU counts of culturable, mostly fast-growing bacteria, were stable throughout the experiments in untreated soil and increased for more than one order of magnitude right after autoclaving of the soil ( Fig. 1; dashed lines, circles). Despite the differences in the number of Salmonella cells inoculated to the soil between the repetitions (~10 6 -10 7 CFU (g dw) −1 ), the influence of autoclaving was significant (Fig. 1).
To further corroborate those observations, we used a cultivation-independent approach, based on quantitative-PCR (qPCR), targeting the Salmonellaspecific invA gene. This approach allows to detect also viable but non-culturable (VBNC) bacteria. The copy number of the invA gene/g dw soil followed the same pattern as observed for the CFU counts (Supporting Information Fig. S1). The amounts of quantified S. Typhimurium 14028s and S. Typhimurium LT2 CFU decreased by factor 1000 in untreated DS soil, and the invA copy number/g dw was between 0.4 and 0.6 log units higher than the CFU (g dw) −1 for all tested Salmonella strains after 7 days (p < 0.05, Student's t-test). Using this culture-independent method, invA was detectable, even in situation when due to low colony numbers on the agar plates, the CFU counts were not quantifiable, for example, in untreated DS soil 91 days after inoculation. In autoclaved DS soil, only 10-fold reduction was observed between day 7 and day 91 for all three strains, and the numbers were not different for S. Typhimurium 14028s and LT2 while the invA copy numbers for S. Senftenberg were 0.7 and 1.3 log units higher than the CFU counts after 7 and 91 days respectively. (Supporting Information Fig. S1).
Taken together, we observed that autoclaving the soil increased the ability of Salmonella to persist in the soil environment. The higher invA copy numbers compared with the CFU data might result from either dead cells or a VBNC state of Salmonella. S. Typhimurium 14028s showed the best persistence in soil in all three experiments and treatments (Fig. 1, Supporting Information Tables S1A and S2). Based on this, S. Typhimurium 14028s was selected for the following experiments.

Diversity influences the ability of Salmonella to persist in soil
In order to further substantiate the results, we asked the question how differences in the diversity level of microbial communities influence the persistence of Salmonella in soil? To this end, we used a modified experimental setup in which S. Typhimurium 14028s was confronted with none, 10% or 100% of the microbial community extracted from untreated DS soil, however, with the same bacterial abundance (assessed on R2A medium and adjusted to 10 6 CFU (g dw) −1 ). Similar to previous observations ( Fig. 1), CFU counts of Salmonella decreased drastically in untreated DS soil to 10 2 CFU (g dw) −1 within 91 days. In autoclaved DS soil, the CFU counts for Salmonella, after an initial increase of 2 orders of magnitudes, stabilized at approximately 10 6 CFU (g dw) −1 (Fig. 2). However, inoculation with 10% of the native DS microbial community had moderate impact on Salmonella persistence ( Fig. 2 and Supporting Information Table S1B), and resulted in a decrease of Salmonella CFU counts from 10 8 CFU (g dw) −1 after the initial increase of 2 orders of magnitude, to 10 5 CFU (g dw) −1 . Whereas after an inoculation with 100% of the native DS soil microbial community, the CFU counts for Salmonella dropped, after the initial increase, to 10 2 CFU (g dw) −1 , very similar to the untreated DS soil (Fig. 2). The CFU counts of culturable, mostly fast-growing bacteria were, after an initial increase, stable throughout the experiments (Supporting Information Fig. S2 and Table S1B). These observations imply that the level of microbial diversity is an influencing factor in the persistence of S. Typhimurium 14028s in the soil environment.

Presence of Salmonella does not influence the prokaryotic diversity in untreated DS soil
The prokaryotic community composition of untreated DS soil inoculated with S. Typhimurium 14028s (10 6 CFU (g dw) −1 ) was analysed and compared with untreated DS soil using 16S rRNA gene amplicon sequencing. We . XLD medium supplemented with rifampicin was used for Salmonella CFU counts (solid lines) and R2A medium for total culturable community CFU counts (dashed lines, circles). The Salmonella enterica serovar Typhimurium strain 14028s (S. Typhiumurium 14028s, triangles), Salmonella enterica serovar Typhimurium strain LT2 DSM 18522 (S. Typhimurium LT2, rectangles) and Salmonella enterica serovar Senftenberg (S. Senftenberg, diamonds) were inoculated on day 0. The Salmonella CFU counts are further plotted for days post-inoculation (dpi). Error bars represent the standard deviation (n = 4). The influencing factors were assessed with ANOVA and the 'multcomb' package, p < 0.05 was considered significant. Slopes of linear regressions were analysed by a generalized linear model using the mixed procedure (Supporting Information Table S1). harvested samples before Salmonella inoculation (day 0), on day 7 (Salmonella successfully persisted in the soil) and on day 91 (Salmonella detected only in autoclaved DS soil). Between 5716 and 114,843 sequence reads were obtained per sample and assigned to 2390 OTUs. The prokaryotic phyla on day 0 represent the initial composition of the prokaryotic community in untreated DS soil ( Fig. 3A and Supporting Information Table S3), this composition changed during the experiment, while the inoculation of Salmonella had no impact on the community composition and diversity (Figs 4, 5 and Supporting Information Table S4). Nonetheless, some taxonomic groups responded to the presence of Salmonella, though without influencing the most abundant groups in these communities (Fig. 3B).
These results revealed that the presence of S. Typhimurium 14028s in untreated DS soil had no impact on the overall prokaryotic community structure.
Salmonella has the ability to influence prokaryotic communities with reduced diversity Furthermore, we tested if findings observed in untreated DS soil hold true for the autoclaved DS soil. Indeed, even if the number of sequence reads obtained directly after autoclaving was not sufficient for further analyses (data not shown), the soil prokaryotic diversity was significantly reduced also 7 and 91 days after autoclaving, if compared with the untreated DS soil (Figs 4, 5 and Supporting Information Table S4). We also observed drastic changes in prokaryotic community composition ( Fig. 3 and Supporting Information Table S3). Most importantly, the presence of Salmonella reduced the prokaryotic diversity in autoclaved soil, as revealed by the alpha-diversities and ANOSIM (Figs 4, 5 and Supporting Information Table S4). This difference was more pronounced at 7 dpi ( Fig. 4 and Supporting Information Table S4). An additional analysis of the diversity using only OTUs not assigned to Salmonella led to the same result, indicating that the influence is independent of the abundance of Salmonella itself (Supporting Information Fig. S3 and Table S4). Taken together, the effect of Salmonella on the diversity of the community was visible only in autoclaved DS soil (Figs 4 and 5).
Because of the differences in the prokaryotic diversity between the treatments, an ANOVA with post-hoc Tukey was performed with the aim to identify the influencing factors. In contrast to the inoculation with Salmonella and the time after inoculation, autoclaving had the strongest impact on the alpha-diversity (Fig. 4). Nevertheless, the time passed after inoculation had an influence on the species richness in both sample types, with and without S. Typhimurium 14028s in autoclaved DS soil, but not in untreated DS soil. The Pielou's evenness was influenced only in samples with S. Typhimurium 14028s (Supporting Information Table S2). Additional diversity analysis indicated higher variations between the replicates of samples with autoclaved DS soil (Fig. 5). While the untreated DS soil samples with and without Salmonella clustered together, a clear distinction between the autoclaved DS soil with Salmonella or without Salmonella was visible. This difference was more pronounced at 7 dpi, suggesting that the influence of Salmonella abundance on the diversity decreased over time (Fig. 5).
When analysing the taxonomic composition of autoclaved DS soil, we observed that Proteobacteria was the major phylum one week after the inoculation with Salmonella ( Fig. 3 and Supporting Information Table S3). The high abundance of Proteobacteria could not be explained by the introduced Salmonella alone (see also below). In contrast, Firmicutes was the most abundant phylum in samples without Salmonella.
Here, most of the OTUs belonged to the genus Bacillus ( Fig. 3 and Supporting Information Table S3). Interestingly, the abundance of the genus Tumebacillus was increased in autoclaved DS soil as well as the genera Arthrobacter and Streptomyces from Actinobacteria. In contrast, the genus Adhaeribacter of the phylum Bacteroidetes was present in very high relative abundance in the autoclaved DS soil samples with Salmonella (Fig. 3B). Taken together, the presence of Salmonella had a significant impact on the Fig. 2. Influence of the diversity on the persistence of Salmonella in soil. Number of colony-forming units (CFU) (g dw) −1 was calculated in untreated DS soil (blue lines, circles) and autoclaved DS soil (red lines), using XLD medium supplemented with rifampicin. The autoclaved soil was supplemented with no (rectangles), 10% (triangles) or 100% (diamonds) of the indigenous soil community diversity at the same bacterial abundance. The Salmonella enterica serovar Typhimurium strain 14028s was inoculated on day 0. The Salmonella CFU counts are plotted for days postinoculation (dpi). Error bars represent the standard deviation (n = 4). Slopes of linear regressions were analysed by a generalized linear model using the mixed procedure (Supporting Information Table S1).
composition of the prokaryotic community in previously autoclaved DS soil.

Salmonella responds to the soil environment and changes its transcriptional profile
The persistence in soil motivated us to address the question on how Salmonella responds to the soil environment. We performed a transcriptome analysis of S. Typhimurium 14028s exposed for 24 h to the mineral medium (mock control) and untreated DS soil or autoclaved DS soil suspensions. The analysis revealed that the expression of 447 genes was higher in soil suspensions compared with mineral medium of which 84 were upregulated exclusively in untreated DS soil suspension, 80 in autoclaved DS soil suspension and 283 in response to both soil suspensions (Fig. 6). The expression of 136 genes was downregulated in samples exposed to soil suspensions compared with samples exposed to mineral medium, of which 34 were downregulated in samples exposed to DS soil suspension, 8 in samples exposed to autoclaved DS soil suspension and 94 in both soil suspensions (Fig. 6). Among the upregulated genes, in samples exposed to both types of soil suspension compared with the mineral medium, genes assigned to translation and peptide biosynthesis were highly enriched (Supporting Information Table S5). The difference between samples exposed to autoclaved and untreated DS soil was also striking. In particular, the enrichment of genes related to gene ontology (GO) terms associated with the biosynthesis of amino acids was prominent. Genes assigned to GO terms connected to the metabolism of glutamine and alpha-amino acids were enriched in response to Fig. 3. Taxonomic distribution of microbial phyla and treatment responding groups in untreated and autoclaved DS soil. Diluvial sand soil was used as untreated soil (DS) or autoclaved (Autoclaved DS) and inoculated with Salmonella enterica serovar Typhimurium strain 14028s (S. T. 14028s) on day 0. MgCl 2 (10 mM) was added in mock controls (Ctrl). Total DNA was sampled immediately before inoculation or 7 and 91 days post-inoculation (dpi). The relative abundance was calculated as percentage of 16S rRNA gene sequences belonging to a particular phylum in each sample (A). The abundance of significantly responding taxonomic groups (with a p-value of < 0.05) was represented using a colourcode (centre-scaled counts). Negative binomial regression and generalized linear model fit and a likehood ratio test post-hoc false discovery rate multiple correction tests were used. The colour code shows the relative abundance of each taxa (B). Each experimental variant was represented by at least three replicates. untreated DS soil suspension ( Fig. 6 and Supporting Information Table S5). In contrast, Salmonella exposed to suspension of autoclaved DS soil strongly upregulated the expression of genes assigned to GO terms representing synthesis of leucine, valine and isoleucine ( Fig. 6 and Supporting Information Table S5). Very remarkable was the increased expression of genes linked to the glyoxylate cycle in S. Typhimurium 14028s exposed to autoclaved DS soil suspension, a variation of the tricarboxylic acid cycle ( Fig. 6 and Supporting Information Table S5). No enrichment of genes related to specific GO terms was found for downregulated genes (Supporting Information Table S5).
In summary, the transcriptional pattern of S. Typhimurium 14028s indicated a distinct response to different soil environments.
Preadaptation to the soil environment allows a better persistence in soil Changes in the transcriptome observed in response to the soil-related environment further encouraged us to ask if soil pre-adapted bacteria exhibit different physiological properties. To answer this question, Salmonella was preadapted to autoclaved DS soil for 180 days and thereafter soil with the pre-adapted Salmonella was used to inoculate untreated DS soil in different ratios. Salmonella grown in Luria Bertani (LB) medium and inoculated into untreated DS soil was used as control. Similar to previous observations (Fig. 1), the CFU counts of LBoriginated Salmonella declined from initial 10 4 to 10 2 CFU (g dw) −1 within 35 days. If compared with this treatment, bacteria which were previously pre-adapted to soil showed a better persistence in untreated DS soil. Regardless of the used inoculation ratios, Salmonella persisted at 10 3 -10 4 CFU (g dw) −1 over the experimental period of 35 days (Fig. 7). In summary, these results support the notion that Salmonella is able to change its physiology in order to better persist in this environment and to establish in soil.

Discussion
Despite the fact that cases of fresh produce-related salmonellosis are continuously increasing in Europe (EFSA and ECDC, 2017), very little is known about the ecology of S. enterica in agricultural systems. Our study showed that S. Typhimurium 14028s, LT2, as well as S. Senftenberg were able to persist in agricultural soil with reduced diversity for several months. The analysis of factors that influence the survival of Salmonella is crucial to understand its persistence in the soil environment. We designed here a soil inoculation experiment with defined and stable humidity, temperature and pre-growth conditions in order to study an additional factor, namely, the microbial diversity in soil. The diversity of the soil microbiota has been demonstrated to influence the survival of E. coli via its suppressive or supportive abilities (van Elsas et al., 2012;Ma et al., 2013;Mallon et al., 2015). We hypothesized that severe reduction of microbial diversity will improve the invasion and persistence of S. enterica in soil.
This severe reduction of the microbial diversity was achieved by autoclaving, whereby some microorganisms survived and repopulated the soil. The physical and chemical characteristics of soil are altered by autoclaving of the soil (Berns et al., 2008). However, the influence of the bacterial community was clear. We observed a longlasting enhanced persistence of Salmonella in autoclaved DS soil compared with untreated soil in all three experiments. The notion of a reduced prokaryotic community effect was strengthened by the observed enhanced persistence in soil inoculated with only 10% of the microbial community but the same total bacterial abundance. In this case, the observed effect was intermediate between a reduced and natural diversity. Our results indicate that different, possibly low abundant microbial populations, may suppress the persistence of S. enterica in an additive manner. Additional factors influencing the persistence of Salmonella might be the release of nutrients from dead cells and the reduced number of bacteriagrazing Protozoa after autoclaving. In addition, the presence of antibiotic-producing bacteria, such as Arthrobacter or Streptomyces, in autoclaved soil is very intriguing. The produced antibiotics could affect the persistence of other bacteria (and Salmonella) and therefore reshaped the community structure. Different Streptomyces were already reported to be reduced in the presence of Salmonella in soil (Turpin et al., 1992). While Arthrobacter was found to be enriched in all samples, Streptomyces were only enriched in autoclaved DS samples without Salmonella, raising the possibility of a reciprocal effect between Streptomyces and Salmonella.  The Euler diagram shows significantly downregulated genes (teal) and upregulated genes (green) compared with mineral medium (q < 0.05) (A). The bar plot shows GO-terms that were significantly enriched in one treatment compared with the other (false-discovery rate (FDR) < 0.05) (B). Enriched genes connected to specific GO terms were identified using the PANTHER platform using Fisher's exact test with FDR.
Firmicutes were present in high abundance 7 days after autoclaving the DS soil. The ability to produce endospores is advantageous, as one of the multiple survival mechanisms upon physical and chemical stresses (Reineke et al., 2013).
In the animal host, Salmonella encounters and adapts to different environmental conditions like temporary acidity, digestive processes of the host and nutritional limitations due to the nutrient uptake by the host (Dandekar et al., 2014). Similar in a soil environment, Salmonella needs to adapt to low nutrient availability. Analysis of short-term transcriptional changes between S. Typhimurium 14028s exposed for 24 h to suspensions of autoclaved or untreated DS soil compared with mineral medium revealed multiple changes in its gene expression pattern. Different water-soluble compounds in the soil suspensions are probably key factors. The expression of genes involved in translation, more precisely protein metabolism and peptide biosynthesis were considerably increased in S. Typhimurium 14028s exposed to either DS soil suspensions, untreated or autoclaved. While more general cellular functions were upregulated in both treatments, the majority of differentially expressed genes between the autoclaved and the untreated soil treatment were mainly assigned to amino acid biosynthesis processes. Genes involved in the biosynthesis of glutamine and alpha-amino acids were strongly upregulated when cells were exposed to untreated DS soil suspensions, genes responsible for leucine valine and isoleucine biosynthesis were strongly regulated in cells exposed to suspension of autoclaved DS soil. Among the alpha-amino acid-related genes, three genes were assigned to the synthesis of tryptophan (trpCBD). The tryptophan metabolism is required for effective biofilm formation on hard surfaces (Hamilton et al., 2009). When exposed to DS soil suspension, the enhanced tryptophan synthesis might provide an advantage for Salmonella to cope with this distinct environment. Besides tryptophan, also expression of genes controlling glutamine biosynthetic processes was upregulated in S. Typhimurium 14028s exposed to untreated soil. Furthermore, genes matching glyoxylate metabolism-related GO terms were enriched after exposure to autoclaved DS soil suspension compared with suspension of untreated DS soil. Wang et al. (2010) demonstrated the importance of amino acid acetylation for the regulation of carbon metabolic processes in Salmonella and particularly the branching between citrate cycle and the glyoxylate bypass. The glyoxylate cycle allows utilizing simple carbon compounds as a carbon source, in a case where more complex sources, such as starch, cellulose, cellobiose, chitin and so on are not available. We assume that the regulations of amino acid biosynthesis and glyoxylate metabolic processes in soil-related environments are interconnected and involved in the adaptation to this ecosystem.
In this study, we show that such an adaptation can take place when untreated DS soil was inoculated with soil containing previously adapted Salmonella. However, such phenomena could be analysed from different perspectives. For example in the case of Pseudomonas fluorescens, the community shifts were stronger when bacteria were soil pre-adapted (Gómez et al., 2016). It appears that also the soil community changes during inoculation with foreign bacteria. Mallon et al. (2018) proposed a legacy effect of the soil community, even after an unsuccessful establishment scenario. Several reports hypothesized that particularly Proteobacteria may occupy niches otherwise taken by Salmonella. Cooley et al. (2003) investigated the influence of soil autoclaving on the colonization of Arabidopsis thaliana with Salmonella Newport and E. coli O157:H7. The authors found that the Gammaproteobacterium Enterobacter asburiae suppressed epiphytic growth of S. enterica under sterile conditions. In contrast, in our study, Alphaproteobacteria were found to be more abundant in all samples with untreated soil compared with samples with autoclaved soil, Rhizobiales being the most abundant order. The results showed, that parts of the community were shifted, which is in line with previously published reports (Horňák and Corno, 2012;Mallon et al., 2018). Taken together, a legacy effect which is in line with the enhanced persistence in soil with low microbial diversity together with the Fig. 7. Persistence of preadapted Salmonella in soil.Number of colony-forming units (CFU) (g dw) −1 was calculated in DS soil, which contained Salmonella for 180 days before the experiment starts. Fresh DS soil was inoculated with either Salmonella obtained from LB plates adjusted to the calculated CFU (blue lines, circles) or with DS soil, which contained Salmonella for 180 days in different ratios. DS soil was incubated for 180 days (red lines) with Salmonella enterica serovar Typhimurium strain 14028s prior the experiment and used to inoculate fresh DS soil (at day 0) with 10% (rectangles), 50% (triangles) or 100% (diamonds) of the Salmonella-containing soil. Salmonella CFU was enumerated using XLD medium supplemented with rifampicin. The Salmonella CFU counts are further plotted for dpi. Error bars represent the standard deviation (n = 4). Slopes of linear regressions were analysed by a generalized linear model using the mixed procedure (Supporting Information Table S1).
pre-adaptation of Salmonella could represent the essential pre-conditions for the successful invasion. However, the used methods have some limitations. Although, we could not amplify sufficient amounts of 16S rRNA gene amplicon sequences directly after autoclaving, suggesting that the 16S rRNA of dead cells is degraded to a high extend, we cannot exclude that 16S rRNA gene amplicon sequences might originate from dead cells. The prevention of strong perturbation seems therefore to be crucial in order to avoid contamination of crops in field.
Our results demonstrated that S. enterica adapts to agricultural soil conditions and may survive for several months, depending on the resident soil microorganisms. In untreated soil, particular microorganisms prevented the establishment of Salmonella competing for nutrients or via negative interactions (e.g. antimicrobials). As a very promising strategy to prevent the colonization of crop plants in the field, we therefore suggest to sustain the diversity and activity of the soil microbiota. Thus, preserving the diversity and activity of the soil microbiota is an important aspect contributing to soil, plant and human health, as highlighted in the one health concept.

Soil inoculation experiments
Soil microcosm experiments were performed with 25 or 50 g of the soil, filled into 50 ml conical tubes or 100 ml flasks, respectively. Sandy soil, an arenic-luvisol (Rühlmann and Ruppel, 2005) with minor amounts of silty sand and 5.5% clay (diluvial sand, DS), sieved through 2-4 mm was used. The soil parameters can be found in Schreiter et al. (2014). To reduce the microbial diversity and abundance of resident microbes, autoclave bags were filled with DS soil and autoclaved (121 C, 20 min) three times with 1 day in between each autoclaving event, resulting in a soil with a severe reduced microbial community. Subsequently, the soil was adjusted to 50% of its maximum water holding capacity (WHC max ), resulting in final water content of 112 ml kg −1 soil (Schreiter et al., 2014). In the first experiment, soil was inoculated with Salmonella cells resuspended in 10 mM MgCl 2 to obtain final cell-counts of 10 6 CFU (g dw) −1 and incubated in closed bottles at 20 C in four different treatments: (i) untreated DS soil without Salmonella, (ii) untreated DS soil with Salmonella, (iii) autoclaved DS soil without Salmonella and (iv) autoclaved DS soil with Salmonella. Each treatment was represented by quadruplicates in three independent repetitions. The first repetition was conducted over a time course of 161 days, the second and third over 147 days.
In the second experiment, 20 g soil was autoclaved and reinoculated with a suspension of microbiota isolated from the same amount of DS soil (20 g) or 10% of the DS soil (see below). For all treatments, DS soil was inoculated with Salmonella cells resuspended in 10 mM MgCl 2 to obtain final bacterial cell-counts of 10 6 CFU (g dw) −1 and incubated in closed bottles at 20 C in eight different treatments: (i) untreated DS soil without Salmonella, (ii) untreated DS soil with Salmonella, (iii) autoclaved DS soil without Salmonella, (iv) autoclaved DS soil with Salmonella, (v) autoclaved DS soil reinoculated with 10% community without Salmonella, (vi) autoclaved DS soil reinoculated with 10% community with Salmonella, (vii) autoclaved DS soil reinoculated with 100% community without Salmonella and (viii) autoclaved DS soil reinoculated with 100% community with Salmonella.
Next, 180 days after the start of the first experiment, the Salmonella containing soil was used to inoculate untreated soil in the third experiment. As a control DS soil was inoculated with Salmonella cells resuspended in 10 mM MgCl 2 to obtain final cell-counts of 10 4 CFU (g dw) −1 adjusted to the calculated CFU in the samples from the first experiment and all samples were incubated in closed bottles at 20 C in four different ratios (Salmonella containing soil/untreated soil): (i) 0/10 (inoculated control), (ii) 1/9, (iii) 5/5 and (iv) 10/0. The bottles were filled to about one-third v/v and opened under sterile conditions every 14 days to ensure aeration.

Extraction of microbiota from soil
For the preparation of the inoculum used in the second soil microcosm experiment, the water-extractable fraction was used. DS soil was mixed with 4 mm glass beads (Karl Hecht, Germany) and sterile water in the ratio 2:1:2 and shaken for 24 h at 200 rpm. The supernatant was transferred to 50 ml conical tubes and centrifuged at 1000g for 15 min twice and subsequently at 3500g for 30 min. The supernatant was then discarded and the precipitate was used as inoculum. For the 100% inoculum, the microbiota isolated from 100 g soil was reinoculated in 100 g autoclaved soil. For the 10% inoculum, the microbiota of 10 g was reinoculated in 100 g soil, incubated for 7 days at 20 C, extracted again and reinoculated in 100 g soil to ensure a similar total abundance. In order to verify the cell density, the CFU number of total culturable bacteria was assessed on R2A plates incubated for 1 day.

Quantification of Salmonella with cultivation-independent and -dependent methods
To enumerate Salmonella, every second week 1 g of soil was transferred to 50 ml conical tubes and suspended in 9 ml of 10 mM MgCl 2 by vortexing. Quadruplicates of serial dilutions were dropped on R2A agar (Merck, Germany), supplemented with cycloheximide (100 mg l −1 ), to estimate the total culturable aerobic CFU counts and XLD (Carl Roth), supplemented with rifampicin (50 mg l −1 ), to determine the CFU counts of Salmonella. The CFU counts were determined after incubation at 20 C (R2A for total culturable bacteria) and 37 C (XLD for Salmonella) for 24 h. For verification of the results and evaluation of a possible VBNC state of Salmonella, a cultivation-independent enumeration was performed. Total DNA was extracted from 0.5 g soil on days 7 and 91 in the second repetition using the FastDNA SPIN Kit for Soil and further purified using the GENECLEAN SPIN Kit (MP Biomedicals, Germany) according to the manufacturer's instructions. The Salmonella-specific gene invA was quantified by qPCR as described by Zimmermann (2014). The invA fragment from the strain LT2, cloned into the pGEM-T vector (Promega, Germany), was used as quantification standard. The amplification efficiencies were between 82.1% and 98.0% and the R 2 values were between 0.991 and 1. Possible inhibitions were not tested. Both quantification methods were adjusted to the dry mass of the samples. We calculated a specific correction factor to estimate the density of Salmonella in our samples. The gene copy number and the CFU counts were compared immediately after soil inoculation with Salmonella. Since the obtained gene copy number was significantly lower than the corresponding CFU counts, we adjusted the qPCR results to the CFU counts for each Salmonella strain and used the resulting factor in the following calculations (Supporting Information Fig. S1).
Analysis of the prokaryotic diversity using 16S rRNA gene amplicon sequencing For 16S rRNA gene amplicon sequencing, we used samples from the second soil inoculation experiment, collected on days 0, 7 and 91 and extracted DNA as described above. The 16S rRNA gene sequences were amplified using primers flanking the V3 and V4 region (Sundberg et al., 2013) and spacer and adapter barcode tags were added (Nunes et al., 2016). Products with sizes smaller than 200 bp were removed using Agencourt AMPure XP beads (Beckman Coulter) according to manufacturer's instructions. The samples were pooled and adjusted to equimolar concentrations, using the DNA Clean and Concentrator-5 kit (Zymo Research). High-throughput amplicon sequencing, 2 × 250 bp, paired-end using the MiSeq platform (Illumina) was carried out following manufacturer's instructions. Raw sequence reads were first trimmed of primer sequences using cutadapt (Martin, 2011) and read pairs for which both primers were found are retained for subsequent analysis. The sequences were then merged, clustered in OTU using UPARSE (Edgar, 2013) with 97% pairwise sequence similarity threshold and chimeras were removed using UCHIME (Edgar et al., 2011). The taxonomic annotation was performed using mothur 'classify.seqs' and 'method = wang' with a minium bootstrap probability of 0.8 (Schloss et al., 2009) against RDP database trainset 16 (Cole et al., 2014). In total 2390 OTUs were identified. Between 5716 and 114,843 sequences were retrieved per sample. The alpha-diversity of the samples was calculated using species richness, Pielou's evenness and Shannon index, where the total reads were rarified to 5716, repeated 100 times and averaged. NMDS was conducted using the 'vegan' package in R 3.4.0 (R Foundation, Austria). The raw sequences were deposited in the Sequence Read Archive database (NCBI) with the Bioproject ID PRJNA510723. The raw 16S sequences were deposited in the Sequence Read Archive database (NCBI) with the Bioproject ID PRJNA510723.

Determination of Salmonella transcriptomes in different environments using RNA-Seq
To determine the transcriptional activity of S. Typhimurium 14028s in soil-related environments, we designed a new experimental approach using suspension of untreated or autoclaved DS soils (see below) and compared the transcriptional activity in each soil suspension to the activity in a mineral medium in which Salmonella was able to persist without changes in the CFU counts over 24 h. Salmonella was adjusted to OD 600nm = 1. Two millilitres of this suspension were pipetted into cellulose ester dialysis tubes with a pore size of 100 kDa (Spectrum Europe, The Netherlands). Closed dialysis tubes were placed in 50 ml conical tubes containing: (i) 30 ml of a mineral medium (1× concentrated M9 salts (Sigma-Aldrich, Germany), 2 mM MgSO 4 and 1.23 mM glucose in sterile deionized water); (ii) 10 g untreated DS soil and 20 ml MgCl 2 ; or (iii) 10 g autoclaved DS soil and 20 ml MgCl 2 . All treatments were performed in triplicates. The conical tubes were incubated at 28 C for 24 h while shaking at 180 rpm. Thereafter, the total RNA in 2 × 0.5 ml from each dialysis tube was stabilized using RNAprotect and extracted using the RNeasy Mini Kit (QIAGEN, Germany) according to the manufacturer's instructions. The rRNA was removed using the Ribominus Kit (Thermo Fisher, Germany) with some modifications concerning the specificity of Salmonella 23S rRNA (Mühlig et al., 2014). For the fragmentation of RNA, a Covaris S220 Focused-ultrasonicator (180 s, 175 W peak power, 10% putty, 200 cycles) was used. The fragments were dephosphorylated and phosphorylated using antarctic phosphatase (NEB, Germany), SUPERase In and T4-PNK (Thermo Fisher) to gain a common phosphorylation state for all fragments. The library preparation was conducted with the TruSeq Small RNA Library Prep Kit (Illumina) according to the manufacturer's instructions using the SuperScript IV cDNA synthesis Kit (Thermo Fisher). The RNA sequences were deposited in the Gene Expression Omnibus database (NCBI) with the accession number GSE132633 (GSM 3884781-GSM3884783 for MM, GSM3497443-GSM3497445 for untreated DS soil, GSM3497458-GSM3497460 for autoclaved DS soil).

Statistical analysis
For the statistical analysis of inoculation experiments, we used ANOVA with post-hoc Tukey conducted with R 3.4.0 (R Foundation, Austria) using the 'multcomb' package (Hothorn et al., 2008). Slopes of linear regressions of Salmonella CFU were analysed by a generalized linear model using the mixed procedure (SAS 9.4; SAS institute Inc., Cary, NC). Tukey tests were performed with the procedure glimmix of SAS. The microbial community composition was statistically analysed with generalized linear model fit and a likehood ratio test post-hoc false discovery rate multiple correction test, and the statistical analysis of the NMDS plots was performed with ANOSIM using the 'vegan' package in R 3.4.0 (R Foundation, Austria). The comparison between CFU counts and the qPCR approach was analysed with Student's t-test.

Supporting Information
Additional Supporting Information may be found in the online version of this article at the publisher's web-site: Figure S1: Comparison between cultivation dependent and independent methods. Quadruplicates of serial dilutions were dropped on XLD agar supplemented with rifampicin (50 mg l −1 ) to determine the CFU counts of Salmonella. The colonies were counted after 24 h incubation at 37 C. For the cultivation-independent enumeration of Salmonella, total DNA was extracted and purified. The Salmonella specific gene invA was quantified by real-time PCR. The red dotted line symbolizes the limit of quantification in the cultivation dependent method. qPCR values were adjusted to the CFU counts measured before and the resulting correction factor was used for further calculations. Figure S2: Enumeration of total culturable bacteria The number of colony-forming units (CFU) per g dw (log CFU per g soil) was calculated for untreated DS soil (blue) and autoclaved DS soil (red). R2A medium with cycloheximide was used. Plates were incubated for 24 h at 28 C. The CFU counts are plotted against time days post-inoculation (dpi). Error bars represent the standard deviation (n = 4). Slopes of linear regressions were analysed by a generalized linear model using the mixed procedure (Supporting Information Table S1). Figure S3: Non-metric multidimensional scaling (NMDS) of the microbial community composition excluding OTUs assigned to Salmonella. Community composition, based on the 16S rRNA gene amplicon sequencing, was assessed in autoclaved DS soil (red symbols), untreated DS soil (blue symbols), 7 days (closed symbols) and 91 days (open symbols) post-inoculation with S. Typhimurium 14028s (triangles). MgCl 2 (10 mM) was used as mock control (circles). The OTUs assigned to Salmonella were manually removed before the analysis. The ellipses show 95% confidence intervals. The Bray-Curtis community dissimilarities are based on OTUs from 16S rRNA gene amplicons. ANOSIM was performed using the 'vegan' package in R 3.4.0 (R Foundation, Austria). Table S1: Statistical analysis of the CFU number in soils during the experiments performed using Tukey test. Number of colony-forming units g −1 dry soil (log CFU g −1 dw −1 ) was calculated in different experiments and plotted against days post-inoculation (dpi). The resulting slopes were compared with each other using Tukey test. Different small letter, indicated in bold, represents differences in estimate values at p < 0.05. The analysis of Salmonella persistence in untreated DS soil and autoclaved DS soil (Fig. 1) is presented in A, the slopes of all repetitions were used in this analysis; Impact of microbial diversity (Fig. 2 and Supporting Information Fig. S2) was analysed in B; Effect of adaptation to soil environment ( Fig. 7) was analysed in C. Table S2: Analysis of Variance (ANOVA) with post-hoc Tukey HSD on the persistence of Salmonella in soil under different conditions. Logarithm of colony-forming units/g dry soil (log CFU (g dw) −1 ) was calculated in untreated DS soil and autoclaved DS soil in three independent experiments. To compare the persistence of Salmonella in the different experiments, the slope was calculated for every sample. The relations between autoclaved and untreated soil, the three Salmonella strains Senftenberg, Typhimurium 14028s and LT2 for each of the three independent experiments were calculated using an ANOVA with post-hoc Tukey HSD. Table S3: Relative abundance of prokaryotic phyla Diluvial sand (DS) soil was used as untreated or autoclaved soil and inoculated with S. Typhimurium 14028s on day 0. MgCl 2 (10 mM) was added to mock control samples without Salmonella. Total DNA was sampled immediately before inoculation or 7 and 91 dpi. The relative abundance of phyla was calculated as percentage of 16S rRNA gene sequences belonging to a particular phylum in each sample. Each experimental variant was conducted in at least three replicas. Table S4: Alpha-diversity in all samples and influence of the different effects. Community diversity, based on the 16S rRNA amplicon sequencing, was assessed in autoclaved DS soil, untreated DS soil, 0, 7 and 91 dpi with S. Typhimurium 14028s. MgCl 2 (10 mM) was used as control. The alpha-diversity was calculated using the OTUs and species richness (SR), Pielou's evenness (Pielou) and Shannon index (Shannon). To show, that the change of diversity was independent of Salmonella's abundance, the Shannon index was calculated without Salmonellaassigned OTUs and compared using a t-test. The relations between the diversity indices in autoclaved and untreated soil, 7 and 91 dpi, and the inoculation of Salmonella or the mock control were calculated using an ANOVA with posthoc Tukey HSD. Table S5: Significantly enriched gene ontology (GO) terms in the different conditions. Transcriptional analysis (RNA-Seq) of Salmonella genes differentially expressed during incubation (24 h) in different conditions: untreated DS soil suspension; autoclaved DS soil suspension; or in mineral medium (MM). Enriched genes related to specific GO terms were identified using the PANTHER Overrepresentation Test using Fisher's exact test with false discovery rate (FDR) correction. The table shows all GO-terms that were significantly enriched in one treatment compared with another (FDR < 0.05). Table S6: Differentially expressed genes. Transcriptional analysis (RNA-Seq) of Salmonella genes differentially expressed during incubation (24 h) in different conditions: untreated DS soil suspension; autoclaved DS soil suspension; or in mineral medium (MM).