Machine learning identifies key microbial signatures for Salmonella enteritidis colonization and persistence in broiler ceca Authors: Sabin Poudel, Fanny Abigail Contreras Zelaya, Eva Guadalupe Guzman Guzman, Farazi Abinash Rahman, Muhammad Naeem, Yagya Adhikari, Shijinaraj Manjankattil, Wilmer Pacheco, Ruediger Hauck, Dianna Bourassa, Kenneth Macklin Published: 2026 ABSTRACT Non-typhoidal Salmonella is a leading cause for bacterial foodborne illnesses in humans, and contaminated poultry products are considered a major source of transmission. Therefore, reducing Salmonella contamination in poultry products is needed to reduce human salmonellosis. This study examined the effects of organic acid (OA) supplementation and Salmonella persistence/cleared in cecal microbiome of Ross YP × Ross 708 male broilers. A total of 900 1-day-old male broilers were randomly distributed to three treatments Control (CN), acetic, lactic and propionic acid blend (ALP), and citric acid (CA); 12 replicate pens/treatment and 25 birds/pen. All birds were challenged with 106 Cfu of Salmonella Enteritidis at d 7. Birds other than in control received continuous supplementation of OA via drinking water. Cecal samples (n = 12/treatment) were collected on d 41; genomic DNA was extracted; and full-length 16S rRNA was sequenced using MinIon (Oxford Nanopore). Organic acid supplementation did not significantly affect alpha and beta diversity indexes, indicating no major changes in overall microbial diversity and community structure. Birds were categorized into Salmonella persistence or clearance groups based on detection of genus Salmonella in 16S rRNA sequence results. Salmonella detection status (persistence vs clearance) did not significantly alter diversity metrics. However, shared and unique bacterial analysis revealed that control birds harbored more unique bacterial species than organic-acid supplemented and Salmonella-clearance birds harbored higher unique bacterial species than Salmonella-persistence birds. These findings suggest that, despite stable diversity indices, compositional shifts occurred in response to organic acid supplementation as well as presence of Salmonella. Using random forest-based machine learning approach, Escherichia coli, Shigella dysenteriae, and Shigella flexneri were identified to be positively correlated to Salmonella, while Romboutsia timonensis was negatively correlated with Salmonella-presence. In conclusion, this study identifies bacterial species associated with persistence or clearance of Salmonella and identifies a promising candidate that could potentially competitive exclude Salmonella colonization and proliferation in ceca. Introduction Foodborne illness resulting from the consumption of contaminated poultry products remains major source of non-typhoidal human Salmonellosis, accounting for approximately 20 % of total Salmonella outbreaks in the USA (Scharff, 2020; IFSAC, 2022; NACMCF, 2024). Although several policy-level strategy changes and different frameworks have been proposed by USDA FSIS (USDA FSIS, 2016; IFSAC, 2022; USDA FSIS, 2022), Salmonella remains a significant public health challenge. Despite the implementation of different pre-harvest and post-harvest interventions to reduce the prevalence and contamination levels in poultry products (Singh et al., 2025; Wang et al., 2025), current national targets for reducing Salmonella incidence rate to 11.5 per 100,000 have not been achieved (Collins, 2022; Ray et al., 2021; CDC, 2021; USDA FSIS, 2020; Marshall et al., 2024; US Department of Health and Human Services, 2022). Control of Salmonella in poultry products is complicated due to its persistence within the chicken gastrointestinal tract, which serves as the primary reservoir for this pathogen (Wang et al., 2023; Manjankattil et al., 2025). During processing of Salmonella-colonized birds, contaminated fecal material serves as major source of carcass contamination and sub-sequent cross-contamination occurs during further handling at processing facilitates (Adhikari et al., 2024, 2025). Therefore, reducing Salmonella colonization at pre-harvest stages is important to minimize the contamination and cross-contamination of chicken carcasses and parts during processing (Machado Junior et al., 2020). Among the various pre-harvest strategies for reducing Salmonella in broilers, organic acid supplementation has emerged as a promising alternative due to its broad-spectrum efficacy against a wide range of intestinal pathogens (Ricke, 2003; Huyghebaert et al., 2011; Adil et al., 2010). Although numerous studies have confirmed its effectiveness under in vitro conditions, however their success in vivo conditions has been limited (Fernandez-Rubio et al., 2009; Van Immersee et al., 2005; Abudabos et al., 2017; Adil et al., 2010), with inconsistent results reported across studies (Ferreira et al., 2025; Hernandez-Patlan et al., 2019, a, b; Adhikari et al., 2020; Sobotik et al., 2021). These inconsistencies may be attributed to factors such as differences in Salmonella serovars and virulence, environmental conditions, bird health status, age and genetics, as well as types, concentrations, and delivery route of organic acids. In this study, we evaluated the efficacy of organic acids mixture of acetic, lactic, and propionic acid blend or citric acid administered through drinking water as a pre-harvest intervention to reduce the prevalence of Salmonella Enteritidis in broiler ceca. Consistent with previous studies, this study showed continuous supplementation of organics acids at the evaluated doses did not significantly reduce the prevalence of Salmonella Enteritidis in ceca and carcass of broilers (Guzman et al., 2025). Therefore, the objective of this study was to characterize the cecal microbiome of broilers fed different organic acids and challenged with Salmonella Enteritidis, using full-length 16S rRNA sequencing. Additionally, we aimed to identify bacterial species that were positively or negatively associated with Salmonella colonization/persistence in ceca to facilitate the development of targeted microbial-based interventions. Materials and methods Experimental design, bird management, and sample collection All the experimental procedures and protocols were approved by the Institutional Animal Care and Use Committee of Auburn University (IACUC Approval No. 2024-5378). A total of 900 1-day-old Ross 708 × YPM male chicks were obtained from commercial hatchery and randomly assigned to three treatments: Control (CN), an acetic, lactic, and propionic acid blend (ALP), and citric acid (CA)] each with 12 replicates and 25 birds/pen. In the ALP, acetic acid was included at 86.1-129.5 ppm and lactic acid 8.6-25.8 ppm resulting in a water pH 4.27. In contract, citric acid was supplemented at 399 ppm, yielding water pH 3.24. The pH of control water was 7.29. Each treatment consisted of 12 replicating pens arranged in complete randomized block design and each pen consisted of 25 birds. Experimental pens were bedded with pine shavings at a depth of approximately 5 −10 cm. The chicks were raised following the standard commercial guidelines and birds were provided with ad libitum access to feed and water. Birds were fed standard (starter, grower, and finisher) diet to meet or exceed specifications recommended for YPM x Ross 708 by Aviagen (2022). Starter diets were pelleted and crumbled, and all other feeds were fed as pellets. Birds were continuously supplied with either control water or water supplemented with organic acids from day 0 through the end of the experiment. All birds were orally inoculated with Salmonella Enteritidis at a dose of 5 × 106 Cfu/bird on day 7. On day 41, 12 birds per treatment (one bird per replicate pen) were randomly selected for necropsy, and cecal contents were collected for microbiome analysis. All birds were humanely euthanized via controlled atmosphere stunning (CO2) for sample collection. Genomic DNA extraction, PCR amplification, and library preparation Total genomic DNA was extracted from cecal contents using QIAamp PowerFecal Pro DNA Kit using QIAGEN QIAcube Connect following the manufacturer’s instructions (QIAGEN, Germany). Extracted genomic DNA was amplified using 27F (AGAGTTTGATCCTGGCTCAG) and 1492R (CGGTTACCTTGTTACGACTT) primers. Amplified PCR products were cleaned using AMPure XP beads (Beckman Coulter, Brea, CA), the library was prepared using Native Barcoding Kit 96 V14 (SQK-NBD114.96) and sequenced using MinION flow cells (FLO-MIN114, R10.4, Oxford Nanopore, UK) (Poudel et al., 2025). Data processing and statistical analysis The raw reads with quality Q-score >10 were demultiplexed and base-called using MinKnow (Oxford Nanopore Technologies, Oxford, UK) with higher accuracy model. The generated raw fastq files having base pair ≤800 and ≥2000 and aligning with host Gallus gallus genome (GCF_016699485.2), were filtered out and taxonomic classification at species-level was done using Emu v 3.4.5 (Curry et al., 2022) on remaining reads. To minimize false positives, low-abundance taxa with fewer than 10 estimated counts were filtered out by default Emu threshold prior to downstream analysis. Overview of sequencing results indicated a total of 2,102,292 reads were obtained for cecal samples, with a mean depth of 62,124 for water CN, 56,648 for the ALP, and 61,596 for CA (S Table 1). Alpha metrics (Observed, Shannon, Simpson, Pielou’s Evenness, Chao1) were computed using phyloseq v. 1.48.0 package (McMurdie and Holmes, 2013) and the vegan package v. 2.7-3 (Oksanen et al., 2014) in RStudio 2026.01.2 (Posit, 2026). Normality of data was assessed using the Shapiro-Wilk test and data were analyzed using Kruskal-Wallis rank-sum test, followed by Dunn’s test for post-hoc pairwise comparison. Beta diversity was accessed using Bray-Curtis, Jaccard and Aitchison distance matrices at both species and genus level. For both Bray-Curtis and Jaccard, cumulative sum scaling (CSS) normalization was done using metagenomeSeq package v 1.46.0 (Paulson et al., 2013), whereas for Aitchison distance, a centered log ratio (CLR) transformation was applied. Compositional differences were visualized using PCoA. PERMANOVA was used to determine the statistical differences between the community clustering using the adonis2 function in vegan package in R. Additionally, homogeneity of multivariate dispersion was tested using PERMDISP test using betadisper function. Pairwise comparisons were performed using the pairwiseAdonis package. Differential abundance analysis of taxa at the species and genus levels was determined using DESeq2 package (v1.44.0) (Love et al., 2014). Global shifts in bacterial composition between treatments were assessed using Likelihood ratio test (LTR) whereas pairwise comparisons between groups were conducted using Wald test. For all tests, P-values were adjusted for multiple comparisons using the Benjamini-Hochberg method, with significance set at P < 0.05. Additionally, Venn diagram was used to determine the shared and unique bacterial taxa among treatment groups. Birds were categorized into Salmonella persistence or clearance groups based on detection of genus Salmonella in 16S rRNA sequence results. To identify the microbial species associated with persistence or clearance of Salmonella in ceca, feature selection was performed using Boruta package (v9.0.0) (Kursa and Rudnicki, 2010) a Random Forest Classifier. Species-level relative abundance was used as the predictor variable, whereas Salmonella persistence or clearance served as the response variable. Spearman’s rank correlation was then used to evaluate the association between the identified important taxa and Salmonella persistence or clearance. The relative abundance of key indicator species between Salmonella- persistence and Salmonella-clearance samples were visualized using box plots. Results Organic acid supplementation did not alter cecal microbial diversity The effects of water treatments on cecal microbiome diversity were evaluated using multiple alpha diversity indices (Observed, Chao1, Shannon, Simpson, and Pielou’s Evenness) at both the genus and species-level (S Figure 1). No significant differences were observed among the treatment groups for majority of the alpha diversity indices (P > 0.05). Although, a global significance was detected for Pielous Evenness at genus level (P = 0.025). However, subsequent pairwise test revealed no significant differences among treatment groups after BH-adjustment (S Table 2). To determine the effect of water treatment on overall microbiome communities, beta diversity was evaluated using three distinct distance matrices at both the genus and species levels (Fig. 1A-C, S Table 3, S Figure 3A). At the species level, no significant differences were detected for any of the metrics (Bray-Curtis P = 0.279; Jaccard P = 0.084; and Aitchison P = 0.214; Fig. 1A-C). At the genus level, while Bray-Curtis (P = 0.223) and Aitchison (P = 0.24) remained non-significant, Jaccard index revealed significant shift in community structure (p = 0.03), as there was a significant difference between control and citric acid supplemented groups (Padjust = 0.044; S Figure 3; S Table 3E). However, these results should be interpreted with caution, as the dispersion test showed significant heterogeneity within-group (P = 0.029). This suggested that the observed difference in Jaccard index may be driven by within-group sample variance rather than a true shift in community composition due to the treatment effects.Fig. 1Beta diversity matrixes at species level were not significantly impacted by supplementation of organic acid via water or presence/absence of Salmonella.(A-F) Principal Coordination Analysis (PCoA) plots illustrate beta diversity patterns of microbial communities from ceca of broiler, based on the Bray-Curtis, Jaccard, and Aitchison distance matrix. The statistical differences in community composition were evaluated using PERMANOVA; P-value is listed at the top of each figure along with multivariate dispersion (PERMDISP) test indicated as Disp P-value. (A) Bray-Curtis Species Organic acid Supplementation (B) Jaccard Species Organic acid Supplementation (C) Aitchison Species Organic acid Supplementation (D) Bray-Curtis Species presence/absence of Salmonella (E) Jaccard Species presence/absence of Salmonella (F) Aitchison Species presence/absence of Salmonella. Beta diversity matrixes at species level were not significantly impacted by supplementation of organic acid via water or presence/absence of Salmonella. (A-F) Principal Coordination Analysis (PCoA) plots illustrate beta diversity patterns of microbial communities from ceca of broiler, based on the Bray-Curtis, Jaccard, and Aitchison distance matrix. The statistical differences in community composition were evaluated using PERMANOVA; P-value is listed at the top of each figure along with multivariate dispersion (PERMDISP) test indicated as Disp P-value. (A) Bray-Curtis Species Organic acid Supplementation (B) Jaccard Species Organic acid Supplementation (C) Aitchison Species Organic acid Supplementation (D) Bray-Curtis Species presence/absence of Salmonella (E) Jaccard Species presence/absence of Salmonella (F) Aitchison Species presence/absence of Salmonella. Organic acid supplementation did not alter microbial communities but altered shared bacterial species in ceca The bacterial abundance at the genus and species level were analyzed (Fig. 2 and S Table 4). The cecal microbiome at the species level was dominated by the Faecalibacterium prausnitzii in all three water treatment groups with relative abundance ranging between 9.27 to 13.56 %. (Fig. 2A, S Table 4A). Whereas Lactobacillus reuteri (9.15 %) was the second most dominant bacterial species in control whereas the relative abundance of L. reuteri was much lower in acetic, lactic, and propionic blend (2.79 %) and citric acid (3.59 %).Fig. 2Cecal microbial composition at the species and genus levels.Relative abundance of bacterial species (A, B) and genera (C, D). Panels show composition by Organic acid (OA) supplementation (A, C; Control [CN], n = 11; an acetic, lactic, and propionic acid blend [ALP], n = 12; and citric acid [CA], n = 12). Panels (B, D) shows the presence/absence by Salmonella (Salmonella-negative, n = 32, Salmonella-positive, n = 3). Stacked bars represent individual taxa with a relative abundance ≥ 0.05; those with relative abundance < 0.05 are grouped collectively as "Other.". Cecal microbial composition at the species and genus levels. Relative abundance of bacterial species (A, B) and genera (C, D). Panels show composition by Organic acid (OA) supplementation (A, C; Control [CN], n = 11; an acetic, lactic, and propionic acid blend [ALP], n = 12; and citric acid [CA], n = 12). Panels (B, D) shows the presence/absence by Salmonella (Salmonella-negative, n = 32, Salmonella-positive, n = 3). Stacked bars represent individual taxa with a relative abundance ≥ 0.05; those with relative abundance < 0.05 are grouped collectively as "Other.". Different bacterial genus dominated between treatment groups; Lactobacillus was dominant in control (18.28 %) whereas Blautia was dominant in ALP (14.70 %) and CA (15.14 %). Blautia genus was second most dominant in control (12.77 %), whereas in both ALP (11.23 %) and CA (12.24 %) Faecalibacterium was second most dominant genus (Fig. 2C, S Table 4B). Although there was slight variation in the relative abundance of bacterial genus between the treatments differential abundance analysis showed that there was no significant difference at both species and genus level due to supplementation of the OA via drinking water (S Table 4C and 4D). The majority of bacterial species and genus shared between control, ALP, and CA (Fig. 3A and 3C). Taxonomic overlap analysis revealed that 236 (69 %) of bacterial species and 107 (70.04 %) of bacterial genus were common among the three groups. The water control group exhibited the highest number of unique species, 37 (10.8 %) and genus 15 (9.9 %).Fig. 3Core microbiome composition at species and genus levels.Venn diagrams illustrate shared and unique taxa between groups. Panels represent the core microbiome at the species level based on (A) organic acid (OA) supplementation and (B) Salmonella presence/absence. Panels (C) and (D) show the core microbiome at the genus level for OA supplementation and Salmonella status, respectively. Core taxa were defined as common and unique bacterial species and genus in each group. Core microbiome composition at species and genus levels. Venn diagrams illustrate shared and unique taxa between groups. Panels represent the core microbiome at the species level based on (A) organic acid (OA) supplementation and (B) Salmonella presence/absence. Panels (C) and (D) show the core microbiome at the genus level for OA supplementation and Salmonella status, respectively. Core taxa were defined as common and unique bacterial species and genus in each group. Persistence of Salmonella did not alter cecal microbial diversity To determine the impact/cause of Salmonella persistence in the ceca of certain birds, we compared the microbiome composition between birds that successfully cleared Salmonella from the ceca (Salmonella absence, n = 32) and birds with Salmonella persisted colonization (Salmonella presence, n = 3, originated from ALP = 2, CA = 1). Similar to the supplementation of organic acids, presence or absence of Salmonella in the cecal microbiome did not alter alpha diversity indices (Observed, Chao1, Shannon, Simpson, and Pielous Evenness) at both the genus and species-level (S Figure 2). No significant difference was observed between the Salmonella persistence and clearance groups for all alpha diversity indices (P > 0.05). Additionally, beta diversity indices (Bray-Curtis, Jaccard, and Aitchison) did not significantly changed at the species level (Fig. 1D-F). Similarly, at the genus level Bray-Curtis and Jaccard were not significant (S Figure 3). Although, at the genus level Aitchison distance was significant, showing a difference in community structure in ceca based on Salmonella status (Padjust = 0.012); however, a significant dispersion effect (p = 0.039) between samples was observed indicating community instability and higher inter-sample variations. Persistence of Salmonella alters microbial communities and shared bacterial species in ceca The bacterial abundance at genus and species level were analyzed in presence and absence of Salmonella (Fig. 2B and 2D; S Table 5). Faecalibacterium prausnitzii was the most abundant bacterial species in both Salmonella-persistence (22.24 %) and Salmonella-clearance (13.04 %) samples. Although differential abundance analysis showed no significant differences between the groups, bacterial species including Lactobacillus crispatus (8.16 % vs. 2.55 %), L. reuteris (6.18 % vs. 3.13 %), Streptococcus alactolyticus (6.40 % vs. 4.7 %), and Romboutsia timonensis (3.75 % vs. 0.79 %) displayed notable shifts in mean relative abundance with the Salmonella- clearance group showing higher abundance (Fig. 2B and S Table 5A). The persistence or clearance of Salmonella in ceca did not differentially impact bacterial species and genus (S Table 5C and 5D). The shared bacterial communities were conserved between Salmonella-persistence and Salmonella- clearance groups, with 222 (64.9 %) bacterial species and 102 (67.1 %) bacterial genus shared between groups. However, a higher number of unique species 117 (34.2 %) and genus 48 (31.6 %) where observed in the Salmonella clearance groups compared to the persistence group (Fig. 3B and 3D). In-depth analysis of unique and shared bacterial species between the Salmonella-persistence and Salmonella-clearance groups, shows predominantly bacterial species from genus Staphylococcus (19 species) such as S. epidermidis, S. capitis, and S. milleri; Lactobacillus (12 species) such as L. antri, L backii, L curvatus, and L salivarius; and Clostridium (6 species) such as C. acetobutylicum, C. bornimense, and C. subterminale were present in Salmonella-clearance groups of birds (S Table 6A). Whereas only three unique bacterial species were present in Salmonella-persistence birds, including Salmonella enterica itself, Shigella dysenteriae, and Prevotella spp (S Table 6A). Bacterial species associated with Salmonella persistence and clearance in broiler ceca To identify the specific microbial species associated with Salmonella colonization in the broiler ceca, a machine-learning approach was employed. The analysis identified seven bacterial species significantly associated with Salmonella status (Fig. 4). Bacterial species of Enterobacteriaceae family [Escherichia coli (Sperman’s rank correlation coefficient (ρ) = 0.51, p = 0.002), Shigella dysenteriae (ρ = 0.78, p < 0.001), and Shigella flexneri (ρ = 0.68, p < 0.001)] exhibited a strong positive correlation with Salmonella presence. Additionally, Aminiphilus circumscriptus (ρ = 0.28, p = 0.099), Blautia sp SC05B46 (ρ = 36, p = 0.033) and Hungatella xylanolytica (ρ = 0.43, p = 0.010) were identified as key taxa associated with the presence of Salmonella in ceca and were positively correlated. Conversely, Romboutsia timonensis was the only bacterial species that exhibited a negative correlation trend with Salmonella presence (ρ = −0.28, p = 0.097) (Fig. 4).Fig. 4Bacterial species associated with Salmonella presence.Boxplots representing the relative abundance of bacterial species identified via Boruta feature selection. Panels are labeled by species name and include Spearman correlation coefficients (rho) and p-values to indicate the relationship with Salmonella presence/absence. Bacterial species associated with Salmonella presence. Boxplots representing the relative abundance of bacterial species identified via Boruta feature selection. Panels are labeled by species name and include Spearman correlation coefficients (rho) and p-values to indicate the relationship with Salmonella presence/absence. Discussion The gastrointestinal tract, particularly the ceca of broilers, harbors a diverse microbial community capable of influencing gut function and physiology, pathogen colonization, and nutrient utilization (Jozefiak et al., 2004; Forder et al., 2007; Apajalahti and Vienola, 2016; Broom and Kogut, 2018; Rubio, 2019; Meijerink et al., 2020; Zenner et al., 2021; Zaytsoff et al., 2022; Cheng et al., 2025). Previously, studies have characterized the impact of Salmonella colonization on the cecal microbiome; however, most analyses have relied on short-read sequencing platforms targeting certain hypervariable regions, such as V3-V4 (Park et al., 2017a,b; Liu et al., 2018; Khan and Chousalkar, 2020, 2021; Cazals et al., 2022; Videnska et al., 2013;). Although this approach provides much-needed information about the bacterial community at the genus level, typically short reads have limited ability to precisely identify species-level interactions. Therefore, in this study, we utilized full-length 16S rRNA sequencing to precisely identify the species level microbial community, so that microbiome-based targeted interventions could be identified. In this study, all the broilers were inoculated with Salmonella on d 7 of age, and birds were continuously provided with three water treatments [Control (CN), an acetic, lactic, and propionic acid blend (ALP), and Citric acid (CA)]. Our results indicate that the supplementation of organic acid did not fundamentally shift the overall microbiome community as there was no difference in alpha and beta diversity metrics between the control and organic acid-supplemented groups. Similar to our study, irrespective of sequencing techniques utilized, previous studies have also identified that the supplementation of various forms of organic acids was unable to significantly shift the cecal microbial community of the broiler ceca (Aljumaah et al., 2020; Han et al., 2023). However, while analyzing the core and unique bacteria at species and genus levels between the treatments, revealed that supplementation of organic acid had possibly exerted some selection pressure, leading to the pruning of rare or transient taxa as we observed the higher number of the unique bacterial species and genus on control compared to the organic acid-supplemented groups. In addition to the pruning of rare bacterial species, although not-significant, supplementation of organic acid had reduced the relative abundance of bacterial species such as Lactobacillus reuteris, L. crispatus, and L. vaginalis in ceca. The reduction in the abundance of these specific bacterial taxa, along with pruning of some bacterial species may be attributed to downstream indirect systemic impact of supplementation of organic acid on the upper gastrointestinal tract (crop and gizzard), which limits the proliferation of acid-intolerant population in upper tract and limits its downstream passage to ceca (Yuk et al., 2005; Wall et al., 2007; Beier et al., 2018; Dittoe et al., 2018, 2019a; Wei et al., 2021). To evaluate the impact of the Salmonella colonization in the cecal microbiome, samples were categorized based on the persistence or clearance of Salmonella sequences in samples. No significant differences were observed in alpha and beta diversity metrics, indicating that overall microbiome composition was not altered. These results suggest that the bacterial communities remained stable despite the persistence of Salmonella. In addition to the stability observed in diversity metrics, the shared bacterial community remained stable with 222 (64.9 %) bacterial species shared between Salmonella-persistence and Salmonella-clearance groups. Interestingly, the Salmonella clearance group exhibited a significantly higher number of unique bacterial species (117, 34.2 %) compared to the Salmonella-persistence groups. Similar to our results, previous studies have also observed that overall bacterial diversity of Salmonella challenged chicks were lower (Mon et al., 2015). The potential explanations for these results might be that the persistence of Salmonella may competitively exclude rare taxa, leading to reduction in bacterial community diversity and allowing Salmonella to establish a niche via displacement of bacterial communities. Alternatively, birds harboring these unique bacterial species may have resisted Salmonella colonization through competitive exclusion, whereas birds lacking this microbial diversity were more susceptible to colonization. Additionally, mice model experiments have shown that dominance of E. coli in microbiota increases susceptibility to enteric infections (Stecher et al., 2010; Haag et al., 2012), similarly in our experiment colonization of Salmonella was associated with presence of E. coli (ρ = 0.51, P = 0.002) in Salmonella-positive birds. Furthermore, in depth analysis of the unique bacterial species of Salmonella-clearance group, revealed that most of the beneficial bacterial species from Lactobacillus, Limosilactobacillus, Bacillus, and Clostridium genus had been excluded on Salmonella-persistence birds. Whereas Prevotella sp. and Shigella dysenteriae were uniquely present in Salmonella-persistence birds, indicating a potential dysbiosis in ceca. Although overall community structure (Beta diversity) did not shift significantly, the presence of these specific species suggests that Salmonella colonization and persistence may create a pro-inflammatory niche that favors pathogens over beneficial commensals (Mon et al., 2015). Both Prevotella sp. and Shigella dysenteriae have been previously associated with the increased intestinal inflammation (Larsen, 2017; Iljazovic et al., 2021; Shi et al., 2014) likely increasing specialized metabolites in ceca helping persistence of Salmonella. Although differential abundance analysis (DESeq2) did not identify the significant differences in bacterial species between the Salmonella-persistence and Salmonella-clearance groups, the application of a machine learning algorithm combined with Spearman correlation identified bacterial species associated with Salmonella colonization in broiler ceca. Our findings identified potential pathogen species including E coli, Shigella dysenteriae, and Shigella flexneri as bacterial species positively associated with Salmonella colonization, suggesting co-colonization and possible contribution to the progression of gut dysbiosis. Interestingly, some beneficial bacterial species such as Blautia sp and Hungatella xylanolytica were also positively associated with Salmonella colonization. These bacteria are typically regarded as beneficial due to their contribution to short chain fatty acids (SCFAs) production. Whereas Romboutsia timonensis was the only bacterial species that was observed to be negatively associated with Salmonella colonization. R. timonensis is generally considered a beneficial commensal bacterium, and has been previously associated with increased body weight, enhanced carbohydrate fermentation, maintenance of intestinal homeostasis, and immune modulation (Madigan-Stretton et al., 2021; Popov et al., 2024; Liu et al., 2023; Gautam et al., 2024; Song et al., 2024; He et al., 2025). The negative association of R. timonensis with Salmonella colonization in the ceca suggest its potential as a candidate for competitive exclusion of Salmonella in broilers. Conclusion In conclusion, organic acid supplementation and the presence of Salmonella in the ceca were not associated with changes in overall cecal microbiome community structure under the conditions of this study. However, supplementation of organic acids in drinking water influenced the microbial landscape by reducing the abundance of rare bacterial species in the ceca. Similarly, Salmonella-persistence was associated with a reduction in bacterial diversity in the ceca and the exclusion of beneficial bacterial species from genus Staphylococcus, Lactobacillus, and Clostridium, potentially decreasing the resilience of cecal environment, facilitating Salmonella persistence. Additionally, this study identified bacterial species that were positively and negatively associated with Salmonella colonization, highlighting the importance of advanced sequencing techniques in identifying targeted microbiome-based intervention to enhance food safety and poultry health. CRediT authorship contribution statement Sabin Poudel: conceptualized the study, developed the methodology, administered the project, curated and analyzed the data, generated the visualizations, and wrote the original manuscript draft and conducted reviews and editing. Fanny Abigail Contreras Zelaya: conducted the lab work for data curation. Eva Guzman Guzman: designed and conducted the bird trail. Farazi Abinash Rahman: conducted the lab work for data curation and review and editing of drafts. Muhammad Naeem: conducted the lab work for data curation and review and editing of drafts Yagya Adhikari: contributed to sample collection, data curation, and review and editing of drafts. Shijinaraj Manjankattil: conducted the lab work for data curation and review and editing of drafts. Wilmer Pacheco: conceptualized the study, project administration for conducting the bird trail, acquired the funding, assisted with interpretation of the findings, critically reviewed and revised the manuscript. Ruediger Hauck: conceptualized the study, acquired the funding, critically reviewed and revised the manuscript. Dianna Bourassa: conceptualized the study, acquired the funding, assisted with interpretation of the findings, critically reviewed and revised the manuscript. Kenneth Macklin: conceptualized the study, acquired the funding, critically reviewed and revised the manuscript. All authors have reviewed and approved the final version and agreed for publication. Declaration of AI During the preparation of this work the author used ChatGPT and Google Gemini to optimize codes for data analysis and data visualization. Disclosures The authors declare no conflicts of interest for this article.