Identification of a microbial sub-community from the feral chicken gut that reduces Salmonella colonization and improves gut health in a gnotobiotic chicken model

ABSTRACT A complex microbial community in the gut may prevent the colonization of enteric pathogens such as Salmonella. Some individual or a combination of species in the gut may confer colonization resistance against Salmonella. To gain a better understanding of the colonization resistance against Salmonella enterica, we isolated a library of 1,300 bacterial strains from feral chicken gut microbiota which represented a total of 51 species. Using a co-culture assay, we screened the representative species from this library and identified 30 species that inhibited Salmonella enterica subspecies enterica serovar Typhimurium in vitro. To improve the Salmonella inhibition capacity, from a pool of fast-growing species, we formulated 66 bacterial blends, each of which composed of 10 species. Bacterial blends were more efficient in inhibiting Salmonella as compared to individual species. The blend that showed maximum inhibition (Mix10) also inhibited other serotypes of Salmonella frequently found in poultry. The in vivo effect of Mix10 was examined in a gnotobiotic and conventional chicken model. The Mix10 consortium significantly reduced Salmonella load at day 2 post-infection in gnotobiotic chicken model and decreased intestinal tissue damage and inflammation in both models. Cell-free supernatant of Mix10 did not show Salmonella inhibition, indicating that Mix10 inhibits Salmonella through either nutritional competition, competitive exclusion, or through reinforcement of host immunity. Out of 10 species, 3 species in Mix10 did not colonize, while 3 species constituted more than 70% of the community. Two of these species were previously uncultured bacteria. Our approach could be used as a high-throughput screening system to identify additional bacterial sub-communities that confer colonization resistance against enteric pathogens and its effect on the host. IMPORTANCE Salmonella colonization in chicken and human infections originating from Salmonella-contaminated poultry is a significant problem. Poultry has been identified as the most common food linked to enteric pathogen outbreaks in the United States. Since multi-drug-resistant Salmonella often colonize chicken and cause human infections, methods to control Salmonella colonization in poultry are needed. The method we describe here could form the basis of developing gut microbiota-derived bacterial blends as a microbial ecosystem therapeutic against Salmonella.

prevent colonization of enteric pathogens (1)(2)(3).With the advances in the microbiome research and because of the worldwide increase in the bacterial antibiotic resistance (4), there is high interest in using mature gut microbiome as an alternative means of suppressing enteric infections (5)(6)(7).The ability of the healthy gut microbiota to prevent pathogen colonization was first demonstrated by Nurmi and Rantala in a classic experiment in which inoculation of young chicken with adult chicken feces prevented the colonization of Salmonella (8,9).The same concept was used in recent years to treat recurrent Clostridium difficile infection in humans by fecal transplantation from healthy individuals (10,11).Recently, rather than using the whole fecal microbial community, there were efforts to identify individual species in the microbiome that confer coloniza tion resistance (12,13).This approach of using single species or combinations of species conferring colonization resistance to prevent pathogen colonization and infection is termed as precision microbiome reconstitution (14) or microbial ecosystem therapeutics (15).
Although colonization resistance of the gut microbiota was first demonstrated, Salmonella colonization in poultry continues to be a significant problem even today.Poultry has been identified as the most common food in outbreaks with pathogens in the United States (16,17).The poultry industry has responded to this problem by implementing biosecurity measures that are designed to minimize exposure of chicks to the pathogens (18).Conversely, increased biosecurity and clean conditions in the production system would also decrease the exposure to commensal bacteria and would reduce the microbiome diversity in the chicken gut.As proposed by Rolf Freter in 1983 in his nutrient niche hypothesis (19), reduced exposure to commensal gut microbes would open nutrient niches in the gut that can be easily used by pathogens which increases their colonization risk.To reduce this risk, poultry industry has attempted to reduce the pathogen colonization by inoculating chicken with complex commensal bacterial blends such as the lyophilized mixture of anaerobic bacteria from the cecum of adult chicken (20), collection of more than 200 bacteria from pathogen-free birds (21), bacteria from healthy chicken mucosal scrapping (22), and continuous flow culture of cecal chicken bacteria (23).Due to the complexity of these mixtures, it is difficult to understand their mechanism of action and improve their efficacy.
Recently, a modified Koch postulate was proposed for the mechanistic understanding of the gut microbiota colonization resistance (24,25).As per this proposal, a mecha nistic study of the colonization resistance requires the development of commensals as a pure culture library and then mono-or poly-associate them in a germ-free host to demonstrate the amelioration of disease.In this study, we used this framework to better understand the colonization resistance of chicken gut microbiota.To this end, we developed a commensal bacterial culture library from Salmonella-free feral chicken and formulated a defined bacterial sub-community from Salmonella-inhibiting commensal species.This consortium was tested to demonstrate the Salmonella exclusion capacity using a gnotobiotic chicken model.Our results showed that the defined consortium reduced Salmonella colonization, the severity of intestinal tissue damage, and inflamma tion.With further improvements, the current approach and the blend we demonstrate here could offer a means of formulating defined communities of commensal bacteria as microbial ecosystem therapeutic in poultry.

Development of the feral chicken gut bacterial library
It is known that microbiota from healthy adult chicken could inhibit the growth of S. enterica in the gut (8).Because of the exposure to a broad range of environmental conditions, feral chicken have more diverse functional gut microbiota (26), and hence, a high possibility of the microbiota members that have inhibitory capacity against S. enterica.To ascertain this, we isolated a bacterial library from the pooled intestinal contents of six feral chicken using anaerobic culture conditions.We used a modified brain heart infusion as the base culture medium which is hereafter referred as BHI-M (Table S1).When a non-selective medium is used for cultivation, it is common that fast-growing bacteria use up space and nutrients in the medium.To avoid this prob lem, we used iterative antibiotic supplementation of BHI-M to suppress bacteria that dominated the base medium (Table S1).For example, from the base BHI-M, when 32 bacterial species were isolated, 5 species (Massiliomicrobiota timonensis, Faecalicoccus pleomorphus, Eubacterium cylindroides, Collinsella sp., and Olsenella sp.) accounted for 52.6% of picked colonies.To suppress the growth of these species, we supplemented BHI-M with gentamicin and kanamycin which allowed isolation of several species that were not isolated from the plain medium.Using 12 such selection conditions, 1,300 isolates were selected.Species identity of 1,023 isolates was determined by either matrix-assisted laser desorption/ionization-time of flight (MALDI-TOF) or 16S rRNA gene sequencing (Table S2). Figure 1 shows an overview of the culture conditions, diversity, and frequency of the isolated species.Overall, we identified 51 species using a cutoff of 97.82% at 16S rRNA gene sequence level (Fig. 1).The identified species were mostly dominated by the phylum Firmicutes (36 species), followed by Bacteroides (5 species), Proteobacteria (5 species), Actinobacteria (4 species), and Fusobacterium (1 species).Altogether, we also captured 11 previously uncultured species which can represent the novel-type strains.

Screening and the selection of defined bacterial consortium that inhibits Salmonella
To determine the species that could inhibit Salmonella in our culture library, we tested the inhibition capacity of representative isolates of 51 species against S. enterica Typhimurium (S.Typhimurium) using a co-culture assay.From the total collection, 30 species showed varying degree of inhibition against S. Typhimurium (Fig. 2A).Since the reduction in pH during bacterial growth is inhibitory to S. Typhimurium, we also determined whether pH was reduced at the end of the co-culture assay (Table S3).The pH range varied between 5.5 and 7.0, and in the majority of the cases, pH did not drop below 6.0.This can be explained that the inhibition of S. Typhimurium by these strains may not be primarily mediated by the production of organic acids that would have lowered the pH of the medium.Interestingly, this screen also showed that 11 species in our collection enhanced the growth of S. Typhimurium (Fig. 2A).Furthermore, we tested whether the Salmonella inhibition capacity of these strains has improved when a subset of strains is pooled together.To reduce the complexity of the pool, 12 inhibitory bacterial strains that are fast growing and maintaining a pH above 5.8 were selected to formulate bacterial blends (Table S4).Since there is the chance that species composition of the blend may positively or negatively influence the S. Typhimurium inhibitory ability, we made several subsets using a combinatorial approach, in which two species were randomly removed from the 12 selected species.With this combinatorial approach, a total of 66 different blends, each of which composed of 10 species, could be generated (Table S4).We then tested the S. Typhimurium inhibitory ability of all these blends using co-culture assay.As shown in Fig. 2B, the blend approach improved the S. Typhimurium inhibition.Out of 66 blends, blend 63 showed the highest inhibition with about 250-fold reduction of S. Typhimurium compared to control.Since blend 63 showed the highest inhibition of Salmonella among all tested blends, we further verified properties of this blend in vivo experiments.This blend which hereafter referred to as Mix10 (Table 1) was composed of Faecalicoccus pleomorphus, Lactobacillus agilis, Staphylococcus saprophyti cus, Bacillus paralicheniformis, Enterococcus durans, Olsenella sp., Megasphaera statonii, Pseudoflavonifractor sp., and Massiliomicrobiota timonensis.Based on a 16S rRNA gene similarity search against EzTaxon and NCBI databases (27), two strains (Olsenella sp. and Pseudoflavonifractor sp.) in this blend represented uncultured organisms of their respective genera, of which a novel species Olsenella lakotia SW165 T was characterized previously (28).These results are consistent with our reasoning that feral chicken gut harbor diversity includes new taxa that are inhibitory against Salmonella.To determine whether Mix10 could inhibit other serotypes of Salmonella that commonly colonize  S1.Species identification was performed using MALDI-TOF or 16S rRNA sequencing.A heat map showed diversity and abundance of bacterial species in a culture library were generated using Morpheus, versatile matrix visualization, and analysis software.The numbers in each circle represent the frequency of isolation of that species.Full list of strains is given in Table S2.
chicken, we tested the inhibition of Mix10 against Salmonella Heidelberg, Salmonella Infantis, Salmonella Enteritidis, and Salmonella Typhimurium using the same co-culture assay.These assays revealed that Mix10 inhibits tested Salmonella serotypes at similar levels (Fig. S1).To determine the mechanism of Mix10 inhibition, cell-free supernatants were tested against Salmonella after heat and proteinase K treatment.Nevertheless, our results show that inhibitory activity is lost in the cell-free supernatant indicating that inhibition is mediated through nutrient competition (Fig. S2).

Mix10 consortium confers partial protection against S. Typhimurium infection
We further determined the effect of Mix10 colonization on host health and in vivo inhibition capacity.To this end, we used a gnotobiotic chicken (Gallus gallus) model previously developed by our group (29) and a conventional chicken model.Briefly, we pooled an equal number of each species in Mix10 and used 10 7 colony-forming units (CFU)/bird for inoculation, while we used 10 5 CFU/bird of S. Typhimurium for infection.Birds were euthanized at day 2 and day 5 post-infection (Fig. 3A).Salmonella load was determined from the cecum content.Overall, S. Typhimurium loads in each group trended to rise at day 2 and lower at day 5 post-infection, except for Mix10-colonized gnotobiotic group that showed no difference of S. Typhimurium loads between day 2 and day 5 post-infection (Fig. 3B).At day 2 post-infection, S. Typhimurium loads in Mix10-colonized gnotobiotic group with S. Typhimurium infection were significantly lower than in the gnotobiotic group with S. Typhimurium infection (6.3-fold mean reduction) chicks (Fig. 3B).This implicated that Mix10 supports the resident bacteria to inhibit Salmonella colonization in chicken gut.Additionally, Mix10 also significantly lowered S. Typhimurium load in both gnotobiotic and conventional chicks correspond ing to in vitro experiment (Fig. S3).The reduction of Salmonella load in Mix10-colonized groups was in-line with our expectation that this consortium could inhibit Salmonella in vivo.Then, we examined the effect of Mix10 colonization on intestinal physiology via histopathology.Inflammatory symptoms of cecal tissues were evaluated using histological sections (Fig. 3C).Fibrinopurulent exudate was observed in the lumen of gnotobiotic group with S. Typhimurium infection (Fig. 3Ci).Also, the mucosa was swollen due to mixed inflammatory cell infiltrates such as macrophages, lymphocytes, a Whole-genome sequencing of each species was performed using Illumina MiSeq.Genome assembly was performed using Unicycle that builds an initial assembly graph from short reads using the de novo assembler SPAdes 3.11.1.The open reading frames were predicted using Prodigal 2.6 implemented in the Prokka software package.Species identity of the individual species was determined by searching against EzTaxon using full-length 16S rRNA sequence.
and heterophils in lamina propria.Erosion of mucosa was evident with the loss of mucosal folds indicating granulomatous transmural enteritis.This deeper inflammation is typical of salmonellosis.Under higher magnification, early transmural inflammation with minimal peritonitis was observed in the Mix10-colonized gnotobiotic group with S. Typhimurium infection (Fig. 3Cii).Inflammation of the mucosa was still detected which narrowed the luminal space; nonetheless, the load of S. Typhimurium in the gut was reduced in this group.Mucosal folds were noticeable, but mixed inflammatory cells were still spotted.The mucosa was not eroded, and no exudate was found in the lumen.Mix10-colonized gnotobiotic chicks showed a large empty lumen with a small amount of ingesta (Fig. 3Ciii).Thin mucosa with mucosal folds was protruding into the lumen.Mild cellularity of lamina propria with scattered glands was observed.The intestinal epithelium of this group was improved toward that of conventional chicken (Fig. 3Civ).In conventional chicken model, the highest inflammation symptoms were observed in conventional group with S. Typhimurium infection (Fig. 3Cv).Mix10-colon ized conventional chicken with S. Typhimurium infection exhibited only a small amount of exudate in the lumen, and submucosal and transmural edema with macrophages and heterophils (Fig. 3Cv).When these histopathological images were compared, mucosal inflammation was very high in chicken with S. Typhimurium infection but subsided in groups with Mix10 colonization (Fig. 3D and E). S. Typhimurium infection in gnotobiotic and conventional group showed improved histopathology scores at day 5 post-infection, while Mix10 resulted in the least lesions as depicted by the histopathological scores compared to S. Typhimurium infection.The Mix10-colonized gnotobiotic group showed lower histopathological score compared to gnotobiotic group with S. Typhimurium infection that elevated score at day 5 post-infection.When Mix10 was administrated into gnotobiotic group, low score was observed at day 2 post-infection, and this score was reduced at day 5 post-infection.This suggested that Mix10 has capability to maintain gut physiology of gnotobiotic chicks.Mix10-colonized conventional group reduced histopathological scores compared to gnotobiotic group with S. Typhimurium infection.These results suggested that Mix10 normalizes chicken gut by supporting the development of intestinal tissue and reducing inflammatory symptoms and intact mucosa during S. Typhimurium infection.

Mix10 modulates gut immunity and reduces Salmonella-induced inflamma tion
In chicken, Salmonella infection is known to trigger inflammation of the gut by the production of proinflammatory cytokines such as interleukin (IL)-1 and IL-6 (30)(31)(32)(33), chemokines such as IL-8, and type 1T helper cell cytokines such as IL-2 and interferon-γ, along with a cascade of other cytokines including tumor necrosis factor-α, IL-12, and IL-15 (34,35).As Mix10 was observed to inhibit S. Typhimurium in vitro and in vivo experiment, we investigated whether Mix10 colonization could ameliorate Salmonellainduced inflammation; the expression level of 84 inflammationassociated genes was measured by quantitative reverse-transcriptase polymerase chain reaction (Q-PCR) array in the chicken cecum (Table S5).As expected, gnotobiotic chicks with S. Typhimurium infection showed multiple fold increase in the expression levels of various proinflamma tory cytokines and chemokines; IL-18, IL-1β, IL-6, and IL-8L1 at day 5 post-infection.Massive expression of these proinflammatory cytokines was correlated to upregula ted expression of other genes such as Toll-like receptors, nucleotide-binding oligome rization domain containing 1, and myeloid differentiation primary response gene 88 which are cell surface pattern receptor recognitions and activators of inflammatory pathways suggesting the capability of S. Typhimurium in the induction of inflamma tion in microbiota-free chicks.Conversely, S. Typhimurium infection in Mix10-colonized conventional chicks exhibited downregulated expression of most genes particularly NFKB1, MAPKs, and IRFs, mediators in the expression of inflammatory cytokines.These results demonstrated that microbiota-induced immunity maturation limits Salmonellainduced inflammation.Furthermore, the chicks colonized with Mix10 did not generate a severe inflammatory response as the expression levels of proinflammatory cytokines (IL-6 and IL-18) was comparatively low.Antimicrobial peptides (AMPs) are crucial for eliminating a broad range of pathogens through pathogen-associated molecular pattern receptors.Two AMPs, cathelicidin2 (CATH2) (36) as well as defensin-beta 1 (DEFB1) (37), have been reported.In this study, Mix10-colonized chicks with S. Typhimurium infection showed a higher level of CATH2 as well as DEFB1 when compared to the group infected with S. Typhimurium (Fig. 4).The results suggested that colonization of Mix10 species in the chicken gut can ameliorate S. Typhimurium-induced inflammation by activating AMPs production and antiinflammatory immune response.

Mix10 in vivo community composition and functional genomic analysis
The microbial community profile of Mix10 colonization in chicken gut was determined by using 16S rRNA amplicon from the cecal samples.Although all species of Mix10 were inoculated in equal proportion to gnotobiotic and conventional chicks, some species reached high abundance, while others have low abundance or did not colonize the gut at all (Fig. 5; Table S6).Olsenella sp., Pseudoflavonifractor sp., and Megamonas funifor mis together constituted more than 70% of Mix10 population in the cecum of Mix10colonized chicks and Mix10-colonized chicks with S. Typhimurium infection.Conversely, Staphylococcus saprophyticus, Bacillus paralicheniformis, and Enterococcus durans were not detected in the samples suggesting that they did not successfully colonize chicken gut.The abundance of Salmonella in all Mix10-colonized groups compared to control groups was lower (Fig. 5).This supported the reduction of Salmonella determined by CFU enumeration (Fig. 3B).However, differences in community composition were observed in Mix10-colonized conventional group compared to Mix10-colonized gnotobiotic group.The abundance of Megamonas was increased, while Pseudoflavonifractor was decreased.Furthermore, Olsenella whose abundance was very high in gnotobiotic group was lost in conventional group.
To decipher the overall functional capabilities of the members of Mix10, the genomes were sequenced and analyzed and are presented in Table 1.Since the presence of functional modules computed using Kyoto Encyclopedia of Genes and Genomes (KEGG) has been used to design defined gut bacterial blends that partially inhibited Salmonella (38), we examined whether the presence of KEGG modules correlated with the in vivo colonization of strains in our study.The presence and completeness of KEGG modules in the strains were annotated, and based on that, a total of 293 KEGG modules were present either completely or partially across 10 species of Mix10 (Fig. 6A; Table S7).Based on the results from amplicon sequencing, only seven organisms, Olsenella, Pseudoflavo nifractor, Megamonas, Megasphaera, Massiliomicrobiota, Faecalicoccus, and Lactobacillus, were able to colonize the gnotobiotic chicken gut with diverse abundance, in which they harbored a total of 234 modules including 122 complete modules.However, out of 293 modules detected across all strains, 243 modules with 159 complete modules were contributed by Bacillus, Enterococcus, and Staphylococcus, which did not colonize the gnotobiotic chicken gut.This indicates that the presence of KEGG modules in the genome of Mix10 species may not be the primary determinant of colonization ability in the chicken gut.This was further evident when all functional modules in colonized and non-colonized strains of Mix10 were compared against the predicted complete modules in the feral chicken fecal metagenome (Fig. 6B).Although colonized strains clustered closer to metagenome of the feral chicken microbiome, presence or absence of KEGG module did not reveal any clear partitioning in this comparison.

DISCUSSION
A mature microbial community in the adult gut is highly diverse and generally prevents the colonization of the pathogens such as Salmonella (39).The pathogen exclusion ability of the gut microbiota has been used to suppress Clostridium difficile infection in humans (7) and to exclude Salmonella in young chicken by fecal transplantation from healthy adults (8).This capacity of the healthy microbiota to exclude or keep the pathogen numbers extremely low in the gut has been termed as competitive exclusion or colonization resistance (40).The incredible complexity of the gut microbiota offers avenues to obtain and identify specific species or defined combination of species that can inhibit pathogens such as Salmonella.Recently, there have been propositions to do so using a modified Koch postulate (24,25,41).As per this proposition, first the commensal organism is isolated as a pure culture, then screened for pathogen exclusion capacity, and in the second stage, the ability of single species or defined communities of the isolated commensals to ameliorate disease is proven by mono-or poly-associating them in a germ-free host (42).
In this study, we used a similar approach to identify Salmonella-inhibiting species from chicken gut microbiota.Since feral chicken has more microbial exposure than commercial chicken, we hypothesized that feral chicken microbiome could contain a high number of species that could provide colonization resistance against Salmonella.Using a culturomics-based approach (43-46), 1,300 strains were isolated, and the species identity of the strains was determined.This collection was composed of 51 species (Fig. 1).From these, a co-culture assay-based screen identified 30 species which inhibited Salmonella in varying degrees (Fig. 2A).A modified Koch postulate to study microbiota function or mechanism proposes the concept of defined bacterial consortium as a unit that could ameliorate disease (12,24,47).Simply put, this paradigm replaces "one pathogen = disease" with "one defined bacterial consortium = no disease." We reasoned that the Salmonella inhibition ability of individual species could be improved if a defined consortium is made by pooling several strains.Consistent with this expectation, a pool of 10 species showed several fold higher Salmonella inhibition capacity (Fig. 2B).What we also observed here is that 2 out of the 66 blends increased the Salmonella growth in vitro and in vivo (Fig. 2B; Fig. S3 and S4).This is a clear demonstration of the fact that bacterial interaction can change individual strain phenotype, for instance, Salmonella inhibition, and is consistent with the concept of "defined bacterial consortium" as a unit to interrogate microbiota function as proposed in the modified Koch postulate for microbiota (24,25).
The blend (Mix10) that showed maximum Salmonella inhibition was then tested in a gnotobiotic and conventional chicken model (29) to determine whether it could exclude Salmonella in the chicken gut and reduce Salmonella-induced disease.Our results clearly showed that Mix10 partially excluded Salmonella (Fig. 3B), reduced tissue damage (Fig. 3C and D), and substantially reduced inflammation (Fig. 4,) in the chicken gut.These results were the validation of the second component in the modified Koch postulate, i.e., a demonstration that defined microbial consortium can ameliorate pathogen-induced disease, such as Salmonella.Our results also showed that this inhibition is not serotype dependent and is applicable to other serotypes of Salmonella that commonly colonize chicken (Fig. S1).Moreover, the cell-free supernatant of the Mix10 failed to inhibit Salmonella (Fig. S2).On the basis of these results, the mechanism of Mix10 against module absent.(B) All complete modules presented in metagenome data were compared to the set of colonized and non-colonized stains of Mix10.The matrix was then used to generate the heat maps using Pearson correlation and average linkage method.The color code indicates the presence and completeness of each KEGG module.A few important KEGG module pathways were indicated in the heatmap.An extended list of KEGG modules and clusters is provided in Table S7.
Salmonella colonization could be nutrient competition of species or improvement of host immune system.When the Mix10 population structure was determined using 16S rRNA amplicon sequencing, we found that out of 10 inoculated species, 3 did not colonize (Staphylococcus saprophyticus, Bacillus paralicheniformis, and Enterococcus durans), while three (Olsenella sp., Pseudoflavonifractor sp., and Megamonas funiformis) dominated the population constituting more than 70% abundance in the cecum.Since these three species dominated all groups, we asserted that they are the key members of Mix10 consortium which produce most of the disease reduction in gnotobiotic chicken.However, loss of colonization may be because of host selection via adhesive molecule such as mucus glycans and immunoglobulin A, and immune system such as antimicro bial compounds, i.e., RegIIIγ and defensins (48)(49)(50)(51) .Overall, our approach of making defined bacterial consortium from a pure culture library and testing them in a native germ-free host offers a means to select a subset of strains for further enhancement by excluding non-colonizers and selecting dominating species in the tested consortium.
Even though Mix10 may not be the best possible consortium of defined bacteria that could exclude Salmonella in chicken, it is possible that many such defined consortia of same or better effectiveness could be formulated from a pure culture library of isolates.As per the redundancy or insurance hypothesis, more than one species is retained in the gut ecosystem to ensure that loss of one species does not result in the loss of function contributed to microbiome by that species (3,52).Therefore, it is more than likely that many species that exist in chicken microbiota contribute to the same function, and their inhibition property can change depending on the composition of the consortium.When the dominating species from several blends such as Mix10 are pooled or stacked, better combinations may emerge.Nevertheless, our results show that choosing a diverse microbiota source is essential to gain Salmonella-inhibiting species because two of three highly dominating species in Mix10 were previously uncultured species, justifying our choice of feral chicken as the input for culture library development.The previous study showed that the use of microbial diversity in the gut as a guide to developing Salmo nella inhibiting defined consortium that was tested in a gnotobiotic mouse model (38).However, we did not find a good correlation between the presence of KEGG modules in the genome and colonization of the strains in vivo.
In the present study, we have tested only representative species in our library.Since other strains in the same species may have the inhibitory capacity, formulating additional blends from such strains in our library, performing combinatorial in vitro inhibition assay followed by testing in the gnotobiotic chicken model, and pooling highly dominating species from multiple mixes may help to design welldefined microbial ecosystem therapeutic against Salmonella.

Conclusion
In this study, we identified a consortium of Mix-10 bacteria which could successfully resist the colonization and protect chicks from Salmonella-induced pathogenesis.We were also successful in formulating a method that could be the basis for developing gut microbiota-derived bacterial blends as a microbial ecosystem therapeutic against Salmonella.

Development of the feral chicken gut microbiota library
Protocols used in this study for sample collection and chicken experiments were reviewed and approved by the Institutional Animal Care and Use Committee (18-058A, 18-063A) at the South Dakota State University, Brookings, South Dakota.For the isolation of bacteria from the feral chicken gut, six intestinal samples were pooled together.The pooled intestinal sample was serially diluted and was plated on modified brain heart infusion agar with 12 different selective conditions (Table S1).All cultures were performed inside an anaerobic chamber (Coy Laboratories) containing 85% CO 2 , 10% H 2 , and 5% N 2 maintained at 37°C.Total of 1,300 colonies was picked from all condi tions and dilutions based on colony morphologies.Selected colonies were streaked on base BHI-M agar, and a single colony was selected for preparing stocks and species identification.Species identity of the isolates was determined using matrix-assisted laser desorption/ionization-time of flight or 16S rRNA gene sequencing.For MALDI-TOF identification, a single colony was smeared on the MALDI-TOF target plate and lysed by 70% formic acid.MALDI-TOF targets were covered with 1 µL of a matrix solution.MALDI-TOF was performed through Microflex LT system (Bruker Daltonics).A MALDI-TOF score >1.9 was considered as positive species identification.Isolates that could not be identified at this cutoff were identified using 16S rRNA gene sequencing.To identify these isolates, genomic DNA of overnight culture from a single colony was extracted using a DNeasy Blood & Tissue kit (Qiagen), according to the manufacturer's instructions.Then 16S rRNA gene sequences were amplified using universal primer set 27F [5′-A GAGTTTGATCMTGGCTCAG-3′ ( 53)]; and 1492R [5′-ACCTTGTTACGACTT-3′ (53,54,54)], and sequenced using a Sanger DNA sequencer (ABI 3730XL; Applied Biosystems) using 27F primer.The 16S rRNA gene sequence was used to verify species using the Gen Bank (www.ncbi.nlm.nih.gov/genbank/) and EZBioCloud (www.ezbiocloud.net/eztaxon)databases (27).All identified isolates were maintained in BHI-M medium with 10% (vol/ vol) dimethyl sulfoxide (DMSO) at −80°C.Aerotolerance of the bacterial species was tested by culturing in aerobic, anaerobic, and microaerophilic conditions.To this end, individual bacteria were first cultured overnight in BHI-M broth at 37°C under anaerobic condition.The optical density at 600 nm (OD 600 ) of the cultures was adjusted to 0.5.Then, 1% of OD 600 adjusted cultures were inoculated in fresh BHI-M media in triplicates.Each replicate of cultures was then incubated under anaerobic, microaerophilic, and aerobic conditions.For microaerophilic condition, a hypoxic box was used to incubate the culture.After 24 hours of incubation, the growth of individual bacteria was deter mined by measuring OD 600.

Co-culture assays, formulation of the bacterial blends, and determination of inhibitory mechanism
A co-culture assay was used to screen all bacterial species for S. Typhimurium inhibi tion capacity.The growth of each bacterial species was measured following overnight incubation in BHI-M using a spectrophotometer at OD 600 .Subsequently, bacterial cells were maintained by adjusting the OD 600 to 0.5 with 10% (vol/vol) DMSO at −80°C.In this assay, each bacterial stock was anaerobically cultured together with S. Typhimurium in a ratio of 9:1 in 1 mL of BHI-M broth and incubated at 37°C for 24 hours.To quantify the magnitude of S. Typhimurium inhibition by each species, the individual co-cultures were 10-fold serially diluted with 1× anaerobic phosphate buffer saline (PBS) and plated on Xylose Lysine Tergitol 4 (XLT4) agar (BD Difco, Houston, TX).The plates were incubated aerobically at 37°C for 24 hours followed by plating on XLT4 agar, and colony-forming units were enumerated to determine the degree of S. Typhimurium inhibition.
To test whether Mix10 could inhibit multiple serotypes of Salmonella, we determined the inhibitory capacity of Mix10 against four serovars of Salmonella (S.Typhimurium, S. Heidelberg, S. Infantis, and S. Enteritidis) that frequently infect poultry.A co-culture assay was performed as previously described in the methods.
To determine whether Mix10 inhibits Salmonella through nutrient competition or through other mechanisms, cell-free supernatants were tested for Salmonella inhibition.To this end, Mix10 was grown cultured as before for 48 hours.The cell pellets were removed by centrifugation at 3,000 rpm for 1 hours.The supernatant was then filtered through a 0.4-µm filter.The purified supernatant pH was adjusted to 6.5-6.8 using NaOH and HCl.The supernatant was divided into three fractions.The first fraction was untreated, the second fraction was heated at 100°C for 1 hour, while the third fraction was treated with 50 µg/mL of proteinase K for 1 hour at 37°C.Following this, S. Typhi murium was cultured in the three fractions diluted 50% with BHI-M.In parallel, equal dilution of BHI-M with 1× PBS was used to culture Salmonella as a control sample.After 24 hours of incubation, CFU of Salmonella was enumerated as previously described.

Determination of in vivo effect of 10-species consortium using gnotobiotic and conventional chicken model
We used a gnotobiotic chicken model developed by our group to determine the in vivo effect of Mix10 previously.Briefly, the gnotobiotic chicks were hatched and experimented using the previously described protocol (29).Fertile specific pathogen-free eggs were wiped with Sporicidin disinfectant solution (Contec, Inc.), an FDA approved sterilizing solution, followed by washing in sterile water.Furthermore, the eggs were incubated at 37°C and 55% humidity for 19 days.Eggs containing an embryo, confirmed after candling, were dipped in Sporicidin for 15 s and wiped with sterile water before transferring to a biosafety cabinet maintained at 37°C and 65% humidity until hatching.Chicks were orally drenched with 10 7 CFU of Mix10 (best Salmonella-inhibiting mix) (Fig. 2) at days 3, 4, and 5 post-hatching, followed by 10 5 CFU of S. Typhimurium challenge on day 6 post-hatching (Fig. 3A).Similarly, for the conventional group, eggs were not sterilized and were allowed to hatch normally.Considering the fact that conventional chicken microbiome already harbors Salmonella (55), we fed chicks with 10 5 CFU of S. Typhimurium and 10 7 CFU of Mix10 on day 6 post-hatching.Chicks were euthanized by cervical dislocation on day 2 and day 5 post-infection (Fig. 3A).The cecum contents and tissues were aseptically collected for further analysis.S. Typhimurium loads in the cecum contents were determined by plating on Salmonella selective XLT4 agar.

Histopathology
The tissues for histopathology were initially fixed in 10% formalin.The cecum tissues were trimmed and processed into paraffin blocks by routine histopathological methods, i.e., gradual dehydration through a series of ethanol immersion, followed by xylene, and then paraffin wax.They were sectioned at 4 µm and stained with hematoxylin and eosin, followed by scanning of glass slides in a motic scanner.Furthermore, the cecum pathology was evaluated based on scores.

Assessment of immune response using quantitative reverse-transcriptase PCR
To study the chicken immune response after bacterial treatment, gene expression in the cecal tissue was determined.Total RNA from cecal tissue samples was extracted using the TRIzol reagent (Ambion | RNA, Invitrogen) method.Briefly, an average weight of 0.042 g of cecal tissue per sample (n = 7 per group) was used.Tissue samples from each group were pooled and homogenized separately in TRIzol reagent (1 mL per 100 mg of tissue sample).RNA extraction was performed according to manufacturer's protocol.RNA concentration was determined using spectrophotometric optical density measurement (A260/A280) by NanoDrop One (Thermo Fisher Scientific, Wilmington, DE).For Q-PCR, cDNA was synthesized using First-Strand cDNA Synthesis Kit (New England BioLabs, Inc.) according to the manufacturer's protocol.To get enough cDNA for downstream procedures, 4 µg of RNA was used as input in a cDNA synthesis.The dynamics of the chicken immune response was analyzed using RT2 Profiler PCR Array (cat# 8ZA-1214, Qiagen) according to the manufacturer's protocol.Real-time Q-PCR was performed following the manufacturer's protocol using an ABI 7500HT thermal cycler (Applied Biosystems).A cycle threshold cutoff of 0.2 was applied to all gene amplifications and was normalized to ribosomal protein L4 and hydroxymethylbilane synthase as they were stably expressed across all treatment groups from a panel of five housekeeping genes.The fold regulation of genes in four treatment groups was calculated by comparing to control group (gnotobiotic chicks).Data were clustered using Pearson correlation with complete linkage in Morpheus package.

Determination of the population structure of the bacterial consortium in the cecum using 16S rRNA amplicon analysis
We determined the relative abundance of individual species in the Mix10 after coloniz ing the gnotobiotic and conventional chicken using 16S rRNA amplicon sequencing.Genomic DNA from cecal contents was extracted using the PowerSoil DNA Isolation Kit (Mo Bio Laboratories Inc., CA).To ensure even lysis of the microbial community, bead beating was performed on 100 mg of cecal contents for 10 min using a tissue lyser (Qiagen, Germantown, MD).Remaining steps for DNA isolation were performed as per manufacturer's instruction.Final elution of DNA was carried out in 50 µL of nucle ase-free water.The quality of DNA was assessed using a NanoDrop One and quantified using a Qubit Fluorometer 3.0 (Invitrogen, Carlsbad, CA).The samples were stored at −20°C until further use.The samples carrying low DNA yield were removed from the downstream processes.The enrichment of the microbial DNA was performed using the NEBNext Microbiome DNA Enrichment Kit (New England Biolabs Inc., MA) according to the manufacturer's instruction.A total of 31 DNA samples were used for 16S rRNA gene sequencing using the Illumina MiSeq platform with 250 base paired-end V2 chemistry.DNA library preparation was performed using Illumina Nextera XT Library Preparation Kit (Illumina Inc., San Diego, CA) targeting the V3 and V4 region of the 16S rRNA gene sequence .The amplicons were then purified using Agencourt AMPure XP beads (Beckman Coulter).Before loading, libraries were bead normalized and pooled in equal concentration.After sequencing, CLC Genomics Workbench (version 11.0.1)(Qiagen) was used to analyze the 16S rRNA sequence data.An average of 72,749 raw reads per sample (ranging from 34,962 to 100,936) was imported to CLC workbench, and samples with less than 2,000 reads were removed from further analysis.After the initial quality check, reads with low Q30 score were removed by trimming with a quality score limit of 0.01.Paired reads were merged at a minimum alignment score of 40.OTU clustering was performed at the 97% similarity level using a locally downloaded Greengenes database (56) and a custom database of full-length 16S rRNA gene sequence of Mix10 species and Salmonella.Best matches were found at chimera cross over the cost of 3 and kmer size of 6.Finally, on an average 28,759 reads per sample were used to generate OTUs.The abundance table and metadata were then used to create stacked bar plots in Explicet software tool (version 2.10.5)(57).The plot was generated using only those OTUs (genus level) that have more than 0.25% relative abundance across all samples.

Genome analysis of Mix10 species and metagenomic analysis of intestinal content using next-generation sequencing
We used the Bacterial DNA kit (D3350-02, e.Z.N.A, OMEGA bio-tek, USA) to isolate the genomic DNA of individual species.The quality of DNA was assessed using Qubit Fluorometer 3.0.Whole-genome sequencing was performed using Illumina MiSeq platform using MiSeq Reagent Kit V3 chemistry.The reads were assembled using Unicycler that builds an initial assembly graph from short reads using the de novo assembler SPAdes 3.11.1 (58).The quality assessment for the assemblies was performed using QUAST (59).The open reading frames were predicted using Prodigal 2.6 (60) in the Prokka software package (61).To determine the functional modules in the genome, the amino acid sequences were mapped against the Kyoto Encyclopedia of Genes and Genomes database using the BlastKOALA genome annotation tool (62).Each KEGG module was represented on a scale of 0-4 (0 = complete, 1= one block missing, 2 = two blocks missing, and 3 = module absent).The matrix was used for hierarchical clustering using the Morpheus (https://software.broadinstitute.org/morpheus)server for construct ing the heat map using Pearson correlation matrix and average linkage method.As mentioned previously, the strains of culture library were isolated from the pooled intestinal content of six feral chicks.This original sample was used for DNA isolation, sequencing, and analysis of our previous study (29).In this study, the assembled contigs from this inoculum were used to predict the putative protein-coding sequences using FragGeneScan (63).The resulting amino acid sequences were clustered using CD-HIT to reduce the sequence redundancy (64).The clustered proteins were then annotated against the KEGG Orthology database to assign the molecular functions using Ghost KOALA (PMID: 26585406).The complete modules present in the metagenomics sample were compared against the colonized (n = 7) and non-colonized strains (n = 3).

FIG 1
FIG 1Diversity and frequency of bacterial species isolated from the feral chicken gut microbiota in culture library.The abundance and diversity of 51 bacterial species (1,023 isolates) were accounted according to culture conditions.Intestinal content of six feral chicks was pooled, stocked, and cultured using 12 culture combinations given in TableS1.Species identification was performed using MALDI-

FIG 5
FIG 5 Mix10 population structure in chicken cecum determined using 16S rRNA sequencing.The relative abundance of individual species in the Mix10 after colonizing the gnotobiotic and conventional chicken and during S. Typhimurium infection was determined using 16S amplicon sequencing.The OTU clustering was performed at 97% similarity level using CLC Genomics Workbench (version 11.0.1) with the Greengenes database and a custom database of full-length 16S rRNA gene sequences of Mix10 and Salmonella.The stacked bar plots of relative abundance at genus and species level (color code) were generated using Explicet software (version 2.10.5).

TABLE 1
Description of bacterial strains used to formulate Mix10 a