Short Chain Fatty Acids and Bacterial Taxa Associated with Reduced Salmonella enterica serovar I 4,[5],12:i:- Shedding in Swine Fed a Diet Supplemented with Resistant Potato Starch

Following the 4-week nursery period in which pigs received diets containing either no in-feed additive (CON) or in-feed treatments of RPS, FAM, or RCS, all pigs were transported to an ABSL-2 facility, challenged with Salmonella I 4,[5],12:i:- (SX 240), and monitored for Salmonella fecal shedding and tissue colonization. The RPS group exhibited significant reductions in Salmonella fecal shedding at 2 and 7 dpi compared with the control animals (P = 0.01 and 0.04, respectively, estimated marginal means, mixed models) (Fig. 1A, B). The reductions in early shedding contributed to the significantly lower area under the log curve (AULC) (cumulative shedding) observed in the RPS group compared with the control animals (P = 0.01), though interindividual variation was evident (Fig. 1C, D). Neither of the other treatment groups (FAM, RCS) had significantly different shedding levels when compared with the control. These data suggest that the use of RPS could be beneficial in reducing Salmonella shedding levels in swine.

For tissues, the estimated colonization of the experimental groups tended to be lower than that of the control group (Fig. 2). A mixed model framework and estimated marginal means contrasts suggested the treatment groups tended to have lower levels of Salmonella colonization compared with the control across most tissues. Significant differences in colonization levels were detected between the FAM and control group in the cecal contents (P = 0.002), while all groups had significantly lower colonization levels in the tonsils compared to the control (P < 0.001).

Many differences in the bacterial communities were detected between the treatments and control group, but the RPS-associated communities exhibited the most differences relative to the controls (Fig. 3). We observed a significant reduction in alpha diversity in the RPS group relative to the control at days 0, 2, 14, and 21, while the FAM group exhibited reduced alpha diversity at day 0 only (Fig. 3A). We detected significant differences in the composition of fecal bacterial communities in all experimental groups compared with the control group prior to inoculation (pairwise PERMANOVA FDR corrected P values < 0.05); however, upon Salmonella challenge, the microbial communities of the RCS and FAM dietary treatment groups became indistinguishable from control associated communities, excluding a small difference in the RCS group at 7 dpi (Fig. 2B, Table S2, Fig. S1). In contrast, the RPS group maintained significantly different microbial communities from the control group at all time points pre- and post-Salmonella challenge.

We investigated within group changes over time by comparing the pre-inoculation communities (D0), to each of the time points, within each of the treatments via pairwise PERMANOVA tests. This analysis indicated that, at 7 dpi the bacterial communities of all groups differed from their D0 pre-inoculation status (Fig. S2, Table S3, Fig. S3). Additionally, all groups except the CON group experienced changes in their bacterial communities that persisted to 21 dpi. Although the RPS group maintained consistently different microbial communities compared to the control animals, the Salmonella inoculation appears to have impacted microbial community structure across all groups.

In the RPS group, a substantial variation in Salmonella shedding levels was observed between the pigs. When the RPS-fed animals were divided into high and low shedders (based on AULC), only the pigs that shed low levels of Salmonella had significantly different microbial communities from the control over the course of the study, including before Salmonella inoculation (Fig. S4, Table S4, Fig. S5). In other words, the microbial communities in the RPS-fed animals with the lowest level of Salmonella shedding differed significantly from the microbial communities in the control pigs, while the microbial communities in the highest shedding animals in the RPS group were indistinguishable from those in the control group including prior to Salmonella inoculation (D0). This suggests that the pigs with bacterial communities that were detectably altered by RPS intake were those that shed less Salmonella, while those pigs with minimal change to their gut communities with RPS intake (i.e., more similar to the CON group) shed more Salmonella. The interindividual variation in response to RPS intake is explored further in subsequent sections.

Although significantly differentially abundant OTUs were detected between each experimental group and the control group, a greater number of significantly different OTUs were identified between the RPS and control group (Fig. 3C). Within the RPS group, OTUs from the orders Clostridiales, Bacteroidales, Selenomonadales, Coriobacteriales, Erysipelotrichales, Bifidobacteriales, Aeromonadales, and “unclassified Gamma-Protoebacteria,” were enriched relative to the control group. In the FAM group, OTUs enriched relative to the control were mainly from the orders Clostridiales, Bacteroidales, and Selenomonadales, while OTUs enriched relative to the controls in the RCS group were mostly from the orders Clostridiales, Bacteroidales, and Erysipelotrichales.

Collectively, these results suggest that the RPS diet had the most striking impacts on gut microbial communities, while the impact of the other in-feed treatments differed only slightly from the control diet. OTUs enriched in the RPS-fed animals were taxonomically diverse, but were from expected gut-associated taxonomic groups. Together, these data provide further evidence that gut microbial community composition and functions may modulate the ability of Salmonella to colonize gut ecosystems and non-antibiotic dietary additives in combination with existing microbial communities can impact pathogen shedding phenotypes.

Many of the OTUs enriched in the RPS group relative to the control group were expected members of the swine gut microbiota and previously associated with gut health and RPS administration (14, 15, 18–24). OTUs belonging to genera Megasphaera, Acidaminococcus, Dialister, Syntrophococcus, Olsenella, Bifidobacteria, and Prevotella_7 were enriched in the RPS group across multiple time points and tissues, with sporadic enrichments of other OTUs (Fig. 4A and B). Some OTUs enriched in the RPS group were relatively lowly abundant, including some in the orders Clostridia, Gammaproteobacteria, Actinobacteria, Spirochaetia, and Methanobacteria. Other OTUs enriched in the RPS group belonged to more highly abundant taxa such as those in the Negativicutes and Bacteroidia orders. In total, the microbes enriched in the RPS group relative to the control group accounted for almost half of the aggregate community (Fig. 4C). Thus, RPS intake impacted a large portion of the total community including both lowly abundant and highly abundant OTUs.

Concomitant to the differences in microbial communities observed between the RPS-fed animals and the control animals, the cecal SCFA profiles of the RPS-fed pigs differed substantially from the control animals. Significantly higher concentrations of butyrate (P = 0.002), valerate (P = 0.008), and caproate (P = 0.042) were measured in the cecal contents of RPS-fed animals relative to the control on dpi 21 (Fig. 5A and B). No significant differences were detected in the other treatment groups compared to controls. When comparing cumulative fecal shedding (AULC) to the cecal concentrations of SCFAs across all treatment groups, pigs that shed the least Salmonella tended to have greater concentrations of butyrate, valerate and caproate in their cecal contents at 21 dpi (Fig. S6). This global correlation structure is at least partially due to the increased concentrations of butyrate, valerate, and caproate and the reduced AULC in the RPS group. When the correlations between these SCFAs and AULC were calculated in the RPS group alone, similar patterns emerged. Considering only the RPS group, pigs that shed the least Salmonella were those that had the highest concentrations of these SCFAs (Fig. 5C). The trend was strongest for valerate (R = −0.70, P = 0.04) and caproate (R = −0.74, P = 0.02), while butyrate (R = −0.45, P = 0.22) was not as strongly associated with reduced cumulative shedding in the RPS group. This recalculation of correlations in the RPS group in isolation also revealed a near significant correlation between increased cecal succinate and reduced AULC (R = −0.65 P = 0.06). These results suggest that increased SCFAs in the RPS group was associated with the reduction in Salmonella fecal shedding, though a direct cause and effect cannot be established.

We sought to determine if specific bacterial OTUs were associated with SCFAs of interest and/or measurements of Salmonella fecal shedding. Using generalized linear models as implemented in DESeq2, we calculated regression coefficients modeling the relationship between OTU abundances and SCFA concentrations in the cecal contents, as well as the associations between OTU abundances and measures of Salmonella shedding (log₁₀ CFU/g feces, and AULC) across all samples. The regression coefficients were calculated taking treatment into account, that is, correcting for differences in SCFAs or Salmonella shedding across treatments. A network displaying significant associations between OTU abundances and covariates of interest was constructed (Fig. S7), with nodes representing features of interest (OTUs, SCFAs, and Salmonella shedding), and edges representing significant regression coefficients. The network was broadly split into two sub-networks: nodes associated with measures of increased Salmonella fecal shedding and decreased SCFA concentrations; and nodes associated with decreased Salmonella shedding and increased SCFA concentrations. Considering the relationships between OTU abundances and fecal shedding or SCFA concentrations across all treatments, many of the OTUs associated with measures of reduced Salmonella shedding were also associated with increased cecal concentrations of butyrate, valerate, and caproate. Additionally, many of the same OTUs were significantly enriched in the RPS-fed animals relative to control animals. A few OTUs held central positions in the network and linked both measures of reduced Salmonella fecal shedding and increased concentrations of SCFAs. These notable OTUs belonged to the genera Megasphaera, Dialister, Prevotella_7, Olsenella, Selenomonas, Pseudoramibacter, and Lachnospiracaea_ge. A Sutterella OTU was significantly associated with both increased Salmonella fecal shedding and reduced butyrate. Overall, the network represents important relationships between OTUs, SCFAs and reduced Salmonella shedding across all groups.

Because only the RPS treatment group exhibited reduced Salmonella fecal shedding, another network was constructed using data only from the RPS-fed pigs to determine if the relationships between OTUs, SCFAs, and Salmonella shedding differed in the context of a RPS diet. To enable easier viewing of important features in the network, only the nodes and edges in the community associated with reduced Salmonella fecal shedding and increased SCFAs are shown (Fig. 6). The full network is available in the supplement (Fig. S8). The RPS-only network revealed many of the same associations evident in the network of all treatment groups; however, some differences were apparent. Several OTUs held central positions in the RPS-only network, meaning they linked different covariates of interest. Nodes with the highest degree (number of edges) are likely to be important members, such as Prevotella_7, Bifidobacterium, Olsenella, and Lachnospiraceae OTUs. Associations between central OTUs and succinate concentrations emerged when considering the RPS-only data. Overall, the RPS-only network analysis suggests the RPS-fed animals with greater abundances of the aforementioned OTUs had lower levels of Salmonella fecal shedding and higher cecal concentrations of butyrate, valerate, and caproate, revealing OTUs that may play an important role in elevating SCFA levels as well as reducing Salmonella shedding in the context of an RPS amended diet.

This study detailed alterations in the swine gut microbiota and SCFA production through the course of a Salmonella challenge in the context of different in-feed additives. Only the RPS-fed group had consistent, meaningful differences when compared with the control group. The RPS-fed animals had a different Salmonella-induced change in their gut microbial communities, an enrichment of RPS- and health-associated microbial taxa, increased concentrations of health-associated SCFAs, and lower measurements of Salmonella fecal shedding relative to the control animals. Within the RPS-fed animals, interindividual variation in response phenotypes was evident. In RPS-fed pigs, those shedding the lowest levels of Salmonella had higher cecal concentrations of butyrate, valerate, caproate, and succinate and higher abundances of many OTUs from genera such as Bifidobacteria, Olsenella, Prevotella_7, and others. Collectively, the data suggest that RPS may be an effective option for limiting Salmonella colonization and shedding (including MDR Salmonella serovar I 4,[5],12:i:-) in swine, provided the appropriate bacterial communities that can utilize RPS as a substrate are present in the gut.

Diets were formulated to industry standards with two diet phases. Phase-1 basal diets were fed from day (d) 1 to d14 and were formulated to contain 3,400 kcal metabolizable energy/kg diet, 1.50% standardized ileal digestible lysine, 0.84% calcium, and 0.45% standardized total tract digestible phosphorus. Phase-2 diets were fed from d15 to d28 and were formulated to contain 3,400 kcal metabolizable energy/kg diet, 1.36% standardized ileal digestible lysine, 0.80% calcium, and 0.40% standardized total tract digestible phosphorus. In each phase, all other amino acids were formulated to meet minimum digestible amino acid:lysine ratios as reported in the NRC (2012). (see Table S1 for complete diet formulations). Weaned pigs, 19 to 21 days of age, obtained from a commercial swine farm were separated into treatment groups within a 96-pen, 480-pig nursery room at the Iowa State University Swine Nutrition Farm, Ames, IA. Treatment groups consisted of 10 pens of five piglets for each dietary treatment, with one pig per pen randomly selected for further inclusion in the inoculation cohort, for a total of 10 piglets in each treatment group (Fig. 7). Treatments consisted of pigs fed a non-amended diet (control, CON), or a similarly formulated diet amended with either: 5% resistant potato starch (RPS; MSP Starch Products Inc., Carberry, MB, Canada), 0.30% fatty acid mix (FAM, SANACORE EN) (salts of sodium butyrate and propionate, palm fat, caprylic, capric, and lauric fatty acids, natural flavoring compounds, and a mixture of steatites and chlorite, Nutriad Inc., Hampshire, IL) or 5% resistant corn starch (RCS, HI-MAIZE 260, Ingredion Inc., Bridgewater, NJ). After pigs received their specific diets for 28 days, the selected inoculation cohort pigs were transferred to isolation rooms at the National Animal Disease Center (NADC, Ames, IA) with one dietary treatment group per isolation room (four rooms total). Pigs were inoculated via the intranasal route (a natural route of Salmonella exposure due to pigs rooting behavior) with 8 × 10⁷ CFU of Salmonella I 4,[5],12:i:- strain SX 240 (swine passaged USDA15WA-1 strain) as previously described (17). Following inoculation, pigs continued to receive their respective phase 2 diets for the remaining 21-days. Fecal sampling intervals for microbiota and Salmonella qualitative and quantitative bacteriology analyses included 0-, 2-, 7-, 14-, and 21-days postinoculation (dpi). At day 0 (prior to inoculation), pigs tested fecal-negative for Salmonella as described previously (45). At 21 dpi, all pigs were euthanized and necropsied to obtain tissue samples (1 g) of the cecum, ileocecal lymph nodes (ICLN), Ileal Peyer’s patch region (IPP), palantine tonsil, and contents of the cecum. All samples were evaluated by quantitative and qualitative bacteriology analyses for Salmonella I 4,[5],12:i:- as previously described (46) using XLT-4 medium (Becton, Dickinson and Company, Sparks, MD) supplemented with 50% tergitol, ampicillin (100 μg/mL), tetracycline (15 μg/mL), novobiocin (50 μg/mL), and streptomycin (50 μg/mL). Salmonella colonies were evaluated on BB CHROMagar Salmonella (Becton, Dickinson and Company) medium for mauve colonies indicative of Salmonella. All experimental procedures involving pigs were in compliance with the recommended principles described in the Guide for the Care and Use of Laboratory Animals by the National Research Council of the National Academies and were approved by the Institutional Animal Care and Use Committee at Iowa State University and the National Animal Disease Center.

16S rRNA gene amplicons of the V4 region were generated from feces, cecal contents, and cecal mucosal samples as described in (46) in accordance with the protocol described by Kozich et al. (47). Briefly, samples were collected, immediately placed on ice, and stored at −80°C until DNA extraction with Qiagen Fecal Microbiome kits (Germantown, MD). Amplicons were sequenced on an Illumina Miseq (La Jolla, CA) using V2 reagent kits (2 × 250 bp read lengths). Operational taxonomic units (OTUs) were generated with mothur (48) and classified using the SILVA v132 taxonomy. Global singletons were removed and samples with fewer than 2,000 total reads were omitted.

One gram of cecal contents was thawed, suspended in 2 mL PBS, vortexed for 1 min, and debris was pelleted by centrifugation at 5,000 × g for 10 min. The resulting supernatant (1 mL) was added to heptanoic acid internal standards. Butylated-fatty acid esters were generated as previously described (49), and analyzed using an Agilent 7890 GC (Agilent, Santa Clara, CA). This assay measures the following 12 SCFAs: formate, acetate, propionate, isobutyrate, butyrate, lactate, isovalerate, valerate, caproate, oxalate, phenylacetate, succinate, and fumarate.

Statistical testing and plotting was conducted in R (50), and tidyverse (51) packages were used to assist in plotting and data wrangling. Calculation of AULC for cumulative Salmonella fecal shedding was performed using a trapezoidal integration method implemented in the trapz function of the pracma package (52). Differences in cecal SCFA concentrations, AULC, and tissue colonization were assessed within an ANOVA framework, 95% confidence intervals and post hoc tests were conducted via Tukey’s honest significant differences method. The lme4 (53), lmerTest (54) and emmeans (55) packages were used to construct mixed models to assess and generate 95% confidence intervals for the estimated differences in Salmonella fecal shedding and alpha diversity between treatment groups and controls at each time point. The R package vegan (56) was used to assess the effect of treatment on community structure similarity through PERMANOVA tests on Bray–Curtis dissimilarities as implemented by the adonis function. NMDS ordinations were generated from these same Bray–Curtis dissimilarities. The Shannon alpha-diversity index was used to quantify alpha diversity. Communities were rarefied to 2,462 reads each before dissimilarity and alpha diversity calculations. The overall effect of treatment and time was assessed with a global test followed by individual pairwise post hoc tests for comparisons of interest (each treatment compared to the controls). P-values were corrected for multiple comparisons using the FDR method. Determination of differential abundance of OTUs relative to the control group and associations between OTU abundances and continuous covariates was accomplished using the DESeq2 package (57) using Wald tests with parametric fits and FDR-corrected P-values. Before model construction, OTUs with fewer than 10 counts globally were removed, and the resulting unrarefied counts were used as the input for DESeq2. Estimated coefficients were shrunk using the apeglm package (58). The package phyloseq (59) was used to assist in data wrangling. Network visualization was accomplished with the geomnet (60) package.

All raw sequence data has been deposited in the SRA and is available under bioproject no. PRJNA638426. All scripts used for analysis and figure generation are available at https://github.com/USDA-FSEPRU/FS12b.git.