Succession of the turkey gastrointestinal bacterial microbiome related to weight gain

Introduction

The United States produces more than 250 million turkeys per year (USDA, 2012), resulting in a multi-billion dollar per year industry (USDA, 2007). Growth performance and sustained flock health is of major economic importance to commercial turkey producers. The microbial community of the gastrointestinal tract, or microbiome, is assumed to play a critical role in overall health of turkeys and other poultry, but the composition of bacterial species within the turkey gastrointestinal tract is largely understudied. Furthermore, the role of these microbes in the development of a healthy intestinal tract is not entirely understood.

The onset of next-generation sequencing has resulted in a marked increase in culture-independent studies characterizing the gut microbiome, but much of this work has been focused on humans and other production animals (Danzeisen et al., 2011; Kim et al., 2011; Scalzo et al., 2004). Fewer studies have sought to understand the turkey microbiome. Some work has focused on comparison of the cecal microbiomes of wild and domestic birds (Scupham et al., 2008), or examination of the turkey microbiome in relation to pathogen colonization, such as Campylobacter (Scupham, 2009). These studies identified specific differences in the genera present in the ceca of differing types of turkeys, as well as time-dependent shifts in bacterial populations in the turkey intestinal tract. However, such previous turkey studies relied on culture-based methods or lower output molecular fingerprinting methods, such as T-RFLP (Scupham, 2009; Scupham et al., 2008). These methods present limitations and biases that are not encountered with high-throughput sequencing (Handelsman, 2004), which can be targeted at loci such as the 16S rRNA gene to achieve deep taxonomic coverage and better resolution.

Commercial turkey production in Minnesota is currently facing a widespread challenge in obtaining flock average weights typical of the industry standard, and this condition has been coined “Light Turkey Syndrome” or LTS (personal communication with the Minnesota Turkey Growers Association). Lower than expected market weights associated with LTS seem to be most prevalent in heavy tom flocks, since the problem is likely amplified over the longer growth period of toms versus hens (Calvert, 2012). LTS has been identified in Minnesota turkey flocks for at least five years, and it has been observed that average flock body weights never approach their genetic potential. The gap between observed weights and genetic potential starts immediately at brooding and continues throughout the grow-out phase of the turkey (Calvert, 2012). A number of possible contributing factors have been speculated, including management practices, the presence of known or unknown bacterial or viral pathogens, disruptions of the gastrointestinal microbial communities, problems with nutrient absorption, or dwarfed immune development in poults (Calvert, 2012). This problem has not, to our knowledge, been identified in other states in the USA.

It is unclear if LTS is tied to another condition in turkeys called poult enteritis syndrome, or PES (Jindal et al., 2009). Studies have demonstrated that inoculation of healthy birds with fecal slurries from birds experiencing PES results in significant reductions in body weight compared to controls (Mor et al., 2011). However, a single causative agent responsible for these weight reductions has not been identified. Astroviruses, rotaviruses, and reoviruses have been found in birds experiencing PES, but they are also found in apparently healthy birds (Jindal et al., 2012) or studies examining these viruses lacked negative control groups (Jindal et al., 2010). LTS has been reproduced in controlled animal studies using pooled fecal homogenates, but not individual microorganisms (Mor et al., 2013). Also, no epidemiological associations have been made between the PES and LTS, and LTS often occurs in the absence of PES. Overall, there is no evidence to suggest a single pathogen associated with LTS.

The use of subtherapeutic concentrations of antimicrobial agents in animal agriculture (i.e., growth promoters) is being increasingly scrutinized because of the rise of multidrug resistant pathogens. Therefore, there is an urgent need to identify antibiotic-free alternatives to improving animal health and weight gain. If there is a microbiome association with conditions such as LTS in turkeys, it is therefore necessary to better understand the succession of bacterial communities in the gastrointestinal tract as the bird ages prior to successfully modulating these microbes using antibiotic-free approaches. Thus, the purpose of this study was to examine bacterial community succession in turkeys raised under industry conditions, and to compare the bacterial communities of young turkeys of different average flock weights using 16S rRNA microbiome analysis.

Materials and Methods

Results

A total of 140 samples were analyzed in this study, involving pooled samples for flock-level comparisons (experiment #1) and individual samples for within-flock comparisons (experiment #2). From these samples, a total of 18,117,871 2X250 bp Illumina MiSeq reads were generated, assembled and quality screened using Pandaseq, trimmed of primer/barcode, and rarefied to 2,776,725 sequences. These remaining sequences were analyzed using QIIME (Table 1). After removal of singleton OTUs upon clustering, there were 2,469 OTUs that remained in the dataset representing species-level OTUs across all samples at 97% clustering.

Experiment #1: flock-level comparisons

According to average daily poult weights, flocks were classified into two groups, light (flocks #1–4 and #7) and heavy (flocks #5–6 and #8) which were significantly different from one another based upon weight (P < 0.05; Fig. 1). Subsequent comparisons of sequencing data were based upon these classifications. In total, 396,725 16S rRNA sequences were analyzed for flock-level comparisons (Table 1).

RDP was used to analyze 16S rRNA sequence reads at the class level, using a bootstrap confidence threshold of 50% based upon RDP recommendations for short read lengths (Cole et al., 2009). Using RDP analysis, Firmicutes was the most prevalent phylum in all flocks for the duration of the experiment (Fig. 2), comprised mostly of Bacilli. Gammaproteobacteria made up substantial proportions of the bacterial microbiome at week 1 of age in both the Heavy and Light flocks. Genus-level distribution showed that these populations were largely composed of Pseudomonas species in flocks #1 and #5, but were mostly Escherichia coli in flock #8. Flocks #1 and #2 also showed a higher occurrence of the class Actinobacteria than the other five flocks, which was composed mostly of the genus Bifidobacterium. Overall, the RDP analysis showed on the class level that Bacilli and Actinobacteria were significantly higher (P < 0.05) on average in the Light flocks. The Heavy flocks had higher proportions of an unknown bacteria at weeks 2 and 3 of age, which was later identified as Candidatus division Arthromitus (see below).

Overall, the number of observed OTUs fluctuated from weeks 1 to 3 of age, with no discernible pattern in any of the 7 flocks (Table S1). This is also demonstrated through Chao1, Shannon and Simpson richness and diversity indices, again suggesting an immature and changing gut microbiome during the first three weeks of life. The rarefaction curve (Fig. S1) also suggested that there was no discernable pattern to the curves from weeks 1 to 3 of age in the flocks; if the diversity and richness was increasing over the 21 day time period of this study, one might expect the slope of the curves from weeks 1 to 3 of age to also increase accordingly. Inferred phylogeny of the collective turkey ileum microbiome based upon OTU (Fig. 3) was comparable to the taxonomic classifications performed with RDP, with most of the diversity of the microbiome being attributed to orders Clostridia and Bacilli. Unifrac-based distance matrices demonstrated that there were no significant differences in the overall microbiome compositions of Light versus Heavy flock at any of the three timepoints (Fig. S2). However, OTU-based analysis revealed significant changes in bacterial taxa, as described below.

Among the observed OTUs in the entire dataset, the ten most abundant OTUs represented 78.6% of the total sequences obtained (Table S2). These dominant OTUs were classified using RDP database and BlastN, and included representative sequences with similarity to Lactobacillus salivarius, Lactobacillus acidophilus/crispatus/helveticus (L. delbrueckii group), Lactobacillus aviarius, Lactobacillus johnsonii, Clostridium bartlettii (Clostridium group XI), Lactobacillus reuteri/vaginalis, Candidatus division Arthromitus, Enterococcus species, E. coli, and Roseburia species. In some cases, the V3 region was sufficient for BLAST similarity enabling classification down to the species level at 100% similarity to database sequences, as demonstrated for certain Lactobacillus species in Fig. 4. The dominant OTUs were not evenly distributed among the flocks examined. In fact, a clear trend was observed where the flocks classified as Heavy contained greater proportions of OTUs classified as Candidatus division Arthromitus and C. bartlettii, and lesser proportions of OTUs classified as Lactobacillus species. METASTATS was performed to determine significant enrichments or depletions in OTUs in the flocks classified as Light versus those classified as Heavy (Table S3). We chose METASTATS because it employs a false discovery rate to deal with high-complexity environments and uses Fisher’s exact test to deal with low abundance or sparsely sampled features (White, Nagarajan & Pop, 2009). At week 1 of age, significantly higher proportions (P < 0.05) of OTUs were observed in Heavy flocks with similarity to Anaerobacter species, Clostridium sensu stricto, and Pseudomonas species. In birds two weeks of age, most of the significant changes in Light versus Heavy flocks included higher proportions in OTUs classified as Clostridium bartlettii and Candidatus division Arthromitus and depletions in a number OTUs classified as Lactobacillus spp. in Heavy flocks. At birds three weeks of age, reduced proportions were also observed in OTUs classified as Lactobacillus species along with increases in the proportions of Candidatus division Arthromitus OTU and several OTUs classified as Lactococcus species in Heavy flocks. Light microscopy was used to validate that sequences with similarity to Candidatus division Arthromitus appeared as segmented filamentous bacteria (Fig. 5). Indeed, segmented filamentous organisms were easily identified in samples that contained large proportions of sequences with similarity to Candidatus division Arthromitus, while these organisms were absent in samples lacking sequences with similarity to Candidatus division Arthromitus. The average number of organisms identified in 5 fields per sample correlated with the counts of the OTU similar to Candidatus division Arthromitus (R² = 0.92).

Experiment #2: within-flock comparisons

In experiment #2, two flocks were followed from ages 1–12 weeks and individual ileum bacterial microbiomes were compared. A total of 119 individual turkey ileum samples were collected from multiple birds at each timepoint examined, resulting in 2,380,000 sequences that were analyzed in QIIME. The dominant OTUs identified in experiment #2 were the same dominant OTUs identified in experiment #1. However, following individual birds over extended periods of time enabled better resolution related to the succession of bacterial populations in the turkey ileum. The flocks followed included a typical commercial turkey flock (designated CF) of 28,000 hens and a research flock (designated RF) replicating commercial farm conditions. Both flocks were sampled at approximately the same timeframe, used the same hatchery as a source of poults, and used similar nutritional plans. The two flocks were strikingly similar in their ileum bacterial microbiome succession, with no significant differences at community-level comparisons based on average Unifrac distances at the same age timepoints (Fig. S3). An average heatmap was constructed to visualize OTU averages for each flock at each timepoint (Fig. 6). In both flocks, birds possessed high proportions of OTUs with similarity to L. salivarius, L. reuteri/vaginalis, and L. acidophilus/crispatus/helveticus (L. delbrueckii group) throughout the 12 weeks of the study. In contrast, OTUs with similarity to L. aviarius and L. johnsonii appeared at 2–4 weeks of age and subsequently increased or maintained their proportions with age in both flocks. An OTU with similarity to C. bartlettii appeared in both flocks at week 5 of age. A number of lower abundance OTUs also appeared over time in a predictable manner in both flocks studied. One key difference between the two flocks studied was the timing of the appearance of the OTU with similarity to Candidatus division Arthromitus, increasing in abundance at 4 weeks of age in the commercial flock and at 2 weeks of age in the research flock. While the proportions of some OTUs varied in individual birds of the same flock and timepoint, overall there were discernable patterns and consistency in the temporal succession of the bacterial microbiome in individual birds (Fig. 7). Sample diversity was assessed temporally using the Shannon index as a measure of community diversity and Chao1 as an estimator of community richness (Fig. S4), and this demonstrated that diversity and richness increased temporally until 8 weeks of age where it plateaued.

Principal coordinate analysis (PCoA) of all samples in the study clearly demonstrated that bird age was more influential than flock or Heavy/Light classification (Fig. 8). Also evident was that as turkey age increased, the individual birds shared greater bacterial community similarity, particularly after the move of birds from brood to grow-out. PCoA coordinate PC2 was plotted against individual bird whole bird weights, and demonstrated that there was a predictable shift in the bacterial community from hatch through approximately 3 kg of weight, after which the ileum microbiome stabilized (Fig. S5). Whole bird weight also was highly correlated with intestinal weight (R² = 0.83; Fig. S6). Based upon two-way hierarchical clustering using the top 50 OTUs in the study, the Neighbor Joining cladogram illustrated that the Heavy flocks in experiment #1 clustered independently from Light flocks (Fig. 9). Overall, the clustering enabled not only clustering based upon bird age, but also clustering based upon flock source.

The core microbiome was also assessed using the data from flocks CF and RF over all timepoints (Table S4). The core microbiome was defined as OTUs present in 100% of all samples at a given timepoint. From these data, a proposed model for the succession of the ileum microbiome was constructed (Fig. 10) involving the dominant bacteria that are considered core at different timepoints as the turkey ages.

Discussion

The purpose of this study was to identify differences in the bacterial microbiome in commercial turkeys of differing average flock weights, and the succession of the ileum bacterial microbiome in the growing commercial turkey. We hypothesized that differences in bacterial microbiome content could be identified that correlated with Heavy versus Light groups based upon average flock weights. In previous work, we demonstrated that age was the major driving factor in the chicken cecum bacterial microbiome when examining a single research flock, as compared to lesser community effects of in-feed antibiotic treatments (Danzeisen et al., 2011). Age was also a dominating factor in the bacterial microbiomes in this study, but environment also appears to play a key role in the initial stages of turkey bacterial microbiome maturation. Using RDP-based taxonomic classification, discernable patterns between Light and Heavy flocks were identified that included decreases in Bacilli and increases in unclassified bacteria later identified as Candidatus division Arthromitus. OTU-based analysis at 97% similarity enabled much greater resolution of the finer-scale microbiome changes occurring between Light and Heavy flocks, and also microbiome changes as the turkey ages. Our data suggest that certain bacterial taxa can be identified that may be important in the early development of the turkey small intestine.

Some of the predominant OTUs identified in this study (with similarity to Candidatus division Arthromitus, L. aviarius, and L. salivarius) have also been identified as predominant species in the chicken mucosal microbiota (Gong et al., 2007). While Lactobacillus species are commonly used in commercial agricultural probiotics and are considered to be positively correlated with gut health, we observed an inverse correlation between bird performance and Lactobacillus species abundance when comparing Light versus Heavy flocks. There is precedence for this in a previous study involving broiler chickens, where it was found that L. salivarius and L. aviarius were associated with decreased bird performance (Torok et al., 2011). These and other studies therefore suggest that an excess abundance of certain Lactobacillus species in the avian small intestine may actually be indicative of decreased or slower bird development, and this could in fact be due to displacement with other bacteria such as Candidatus division Arthromitus and Clostridium group XI organisms that are critical for gut microbiome development (Danzeisen et al., 2011).

Segmented filamentous bacteria, or SFBs, are known to be indigenous members of developing microbiota in the animal small intestine. These bacteria are visible in the ileum of turkey poults in the early stages of life (Bohorquez, Bohorquez & Ferket, 2011). Previously, SFBs have mostly been associated with disease in poultry (Angel et al., 1990; Goodwin et al., 1991). However, these organisms are quite heterogeneous and belong to multiple bacterial taxa (Thompson, Mikaelyan & Brune, 2013). Therefore, SFBs from different bacterial taxa may exert different effects on the poultry gastrointestinal tract. Some SFBs belong to the candidate taxa known as Candidatus division Arthromitus, more recently referred to as Candidatus Savagella (Snel et al., 1995; Thompson, Mikaelyan & Brune, 2013). Phylogenetically speaking, these organisms form a discrete and distant lineage within the family Clostridiaceae, clustering within Clostridia Cluster I but divergent from other organisms within this cluster (Prakash et al., 2011). Candidatus division Arthromitus organisms are Gram-positive, spore-forming bacteria that are of long filamentous form consisting of segmented structures. Candidatus division Arthromitus organisms are unique from other Clostridia in that they have reduced genomes and are thus highly dependent on their host for many metabolic functions, including synthesis of amino acids and purines and pyrimidines. Because of these dependencies, they have not yet been cultured. Interestingly, Candidatus division Arthromitus organisms have been previously associated with early development of the innate immune system in mice and have been positively associated with gut development in other animal species (Prakash et al., 2011). Therefore, whether directly or indirectly, our data suggest that Candidatus division Arthromitus-like organisms play a positive role in turkey gut development and weight gain.

While we found a clear correlation between the presence of Candidatus division Arthromitus-like organisms and Heavy flock status, it is important to note that the sequence abundance of this OTU was among the most variable identified based upon standard deviations. That is, in individual birds the presence or absence of this organism was highly unpredictable, although the average abundance of this organism using a pooled timepoint approach showed clear trends. This may underscore the reality that substantial inconsistencies exist between poults entering the brood environment that may be contributing to their variable temporal microbiome succession. Future studies are essential that address not only the bacterial microbiome in the turkey gastrointestinal tract, but also the bird immune status that corresponds with these microbiota changes.

In addition to Candidatus division Arthromitus, several other OTUs were identified that were associated with Heavy flocks and/or associated with temporal shifts in the turkey ileum microbiome. L. aviarius is a Gram-positive, non-spore-forming strict anaerobe that was first isolated and identified from the chicken (Fujisawa et al., 1984). L. aviarius was found to be suppressed upon Clostridium perfringens challenge in broiler chickens (Feng et al., 2010), and was also suppressed following salinomycin treatment in broilers (Czerwinski et al., 2012). Therefore, the emergence of L. avaiarius in the small intestine appears to be significant in terms of bird gut microbiota development and warrants further study.

Another OTU that was a clear marker of ileum microbiome succession in this study was 100% similar to C. bartlettii. This bacteria belongs to group XI Clostridium (Song et al., 2004), which is a heterogeneous group of Clostridium that also includes Eubacterium and Peptostreptococcus species. C. bartlettii is poorly described in the literature, but it has been associated with a high ability to ferment aromatic amino acids in the gut (Russell et al., 2013). Since L. aviarius and C. bartlettii are strict anaerobes, it is also possible that the gastrointestinal environment dictates the ability of these organisms to colonize and expand in numbers. Whatever the case, these bacteria appear to be good markers of ileum bacterial microbiome development in the turkey.

When examining the succession of the ileum bacterial microbiome in commercial turkeys, there was a clear shift in the microbiome that occurred in all flocks examined. The timing of this shift varied between flocks examined, and in general the shift occurred earlier in research flocks than it did in commercial flocks. Clustering analysis indicated that the largest differences in the timing of this shift occurred in the Heavy research flocks in experiment #1. This is easily explained by the nature of these research flocks, which were pen-based studies, compared to all other flocks that were either actual commercial flocks or research flocks representing commercial conditions. While the timing of bacterial microbiome succession varied depending on flock studied, the succession of microbes was clear (Fig. 10) and a predictable core microbiome was established that was present in 100% of the birds examined at each timepoint (Table S4). Our model for the succession of bacteria in the turkey ileum, based upon this study, indicates that the core predictable bacterial microbiome in the turkey involves early colonization with Gammaproteobacteria at high variability (particularly E. coli). At the same time, consistent and dominant colonization occurs with L. delbrueckii group organisms, L. salivarius, and L. reuteri/vaginalis, along with colonization of lower abundance organisms including Blautia and Roseburia species. A shift occurs at weeks 3–4 of age when some birds are colonized heavily but variably by Candidatus division Arthromitus, which is accompanied by consistent colonization by L. aviarius and L. johnsonii, along with colonization of lower abundance organisms including Bifidobacterium and Corynebacterium species. At weeks 5–7 of age, Clostridium group XI organisms and L. ingluviei begin to consistently colonize the ileum. The overall timing of the key shifts appears to occur earlier in research flocks as compared to commercial flocks. The reasons for this observation are currently unknown. We can at least rule out poult source and diet, since the birds in this study came from the same hatchery and were fed similar diets. Other possible mitigating factors include management approaches and overall bird immune status, and these are variables requiring further examination.

There were some limitations to this study. First, samples were pooled for each flock and timepoint in experiment #1, limiting our ability to assess individual-to-individual variation within a given flock when comparing Light versus Heavy flocks. This was better addressed in experiment #2 addressing individual-to-individual turkey variation. Another limitation is that we performed sequencing on a small portion of the 16S rRNA gene, so the data obtained here are not indicative of functional aspects of the turkey small intestinal microbiota. Furthermore, we have a high level of confidence that the Candidatus Arthromitus-like organisms are SFBs associated with the ileum mucosa through DNA sequencing and light microscopy, but additional tools such as in situ hybridization would be required to confirm these observations. Furthermore, this work did not address viable counts of identified bacteria so it only measured the presence or absence of DNA representing these bacteria.

Conclusions

This study demonstrates that the succession of bacteria in the turkey ileum proceeds in a predictable manner. The timing of bacterial succession appears to be dependent on multiple factors, at least including incoming poult consistency and commercial flock environment. From this work, several candidate markers of turkey bacterial microbiome development have been identified that will aid in future efforts aimed at modulating the microbiome to improve turkey gut health and overall bird growth and development.

Supplemental Information