Gut Microbiome Composition Differs Extensively Between Indian Indigenous Chicken Breeds Originated in Different Geographical Locations and a Commercial Broiler Line, but Breed-specific as Well as Across-breed Core Microbiomes Are Found


 Background: With increasing concern on antimicrobial resistance and given the key role of gut microbiome to the host’s nutrition, productivity, development and function of immune system and competitive exclusion of pathogens, characterization of the chicken gut microbiota of different breeds or lines holds high potential for providing clues for developing alternative to antibiotic growth promoters and reducing pathogen load. Methods: In this study, we analyse and compare the gut microbiome of indigenous Indian Nicobari, Ghagus and Aseel chicken breeds originated in Nicobari island, coastal area and Indian mainland, respectively and a global commercial broiler VenCobb400 line reared without antimicrobials using 16S rDNA V3-V4 hypervariable amplicon sequencing.Results: The dominant gut microbial Phyla were Bacteroidetes and Firmicutes in native breeds whereas in VenCobb 400 the phylum Firmicutes was most dominant. Bacteroidetes/Firmicutes ratio in Ghagus and Nicobari were similar but differed from that of Aseel. The Aseel, VenCobb 400, Ghagus and Nicobari breed or line presented 3461, 857, 1911 and 1767 distance (97% cutoff) based operational taxonomic units (OTUs) with > 2 members. Alpha diversity analysis indicated higher diversity as well as richness in indigenous breeds than in the broiler line. Beta diversity analysis indicated significant difference (P<0.001) in community structure between different breeds or line and clear separation of clusters at different taxonomic levels based on breeds or line. Differential abundance analysis using edgeR detected 88 phylotype OTUs with significant (P<0.05) difference in abundance between groups. Linear discriminant analysis effect size (LEfSe) revealed 82 breed or line specific phylotype OTU level biomarkers. Five phylotype OTUs occurred as core microbiome across all breeds or line representing 13% of total prokaryote count. Seven phylotype OTUs were core microbiome across the three native breeds representing 49.3% of total prokaryotic sequences of native breeds. Correlation analysis indicated existence of complex microbial networks in all breeds or line.Conclusion: The study indicated highest overlap in microbial community composition between the coastal breed (Ghagus) and the island breed (Nicobari) among indigenous breeds and very low overlap between the broiler line and all indigenous breeds. Despite extensive variation, breed specific as well as across breed or line core microbiomes were found. This study provides new insight into the comparative community structure of microbiome of chickens of 4 breeds or line of different genetic background or geographical origin and may facilitate development of breed specific feed additives and alternative to antibiotic growth promoters.


Background
The chicken is the cornerstone of animal agriculture worldwide with a ock population exceeding 40 billion birds/year [1]. Poultry represents one of the most e cient form of animal protein with highly e cient feed conversion. Ever increasing human population, urbanization and income levels are Page 3/39 contributing to huge increase in demand for protein and therefore livestock and poultry. Sustainable poultry meat and egg production is important to provide safe and quality protein sources in human nutrition. Feed e ciency and faster growth are the crucial goals in highly competitive poultry production system. For attaining these goals maintaining a healthy gut is an important prerequisite.
The gastrointestinal (GI) tract of chickens is densely populated with diverse and complex microbiota (Bacteria, fungi, Archaea, protozoa, and virus; dominated by Bacteria) that plays a vital role in digestion and absorption of nutrients, host immune system development and pathogen exclusion, endocrine activity, maintaining normal physiological homeostasis and in uencing gut development, nutirent supply and host metabolism and detoxi cation [2]. Understanding diversity and community structure of gut microbiome is important for devising strategy for improving the chicken gut microbiome. Several studies in rats, human, chicken and other species have indicated that gut microbiome composition is strongly in uenced by host. Further, host immune pathways in gut tissues may in uence microbiome structure.
Similarly, some studies have indicated strong correlation between gut microbiome and feed e ciency. In studies with cattle it has been shown that a heritable set of core gut microbiome in uence dairy cow productivity [3]. It has been suggested that high throughput sequencing tools have huge potential to be used for assessment of microbiome structure in gut of chicken in a comprehensive manner which in turn may further development of strategy for improving growth, feed e ciency, survivability or lower pathogen shedding via the development of host speci c probiotics [4].
Native chicken breeds are gaining popularity across India due to their unique characteristics like desirable avour of meat and eggs and ability to thrive in low input system and hot humid climate. Further, there is a growing demand for eggs and meat of native chicken in cities as compared to the exotic ones due to their perceived better avor and taste. Many producers are rearing indigenous chicken breeds under intensive system to meet increasing demand.
Among 17 indigenous chicken breeds of India breeds like Aseel, Nicobari and Ghagus have been shown to have better egg, meat traits, resistance to infectious diseases with better immune parameters but lower feed conversion e ciency as compared to commercial broiler chickens. Analysis to de ne gut micobiome of native Indian breeds like Asamese chicken [5], Aseel and Kadaknath has been reported recently [6]. But the study on Aseel and Kadaknath was carried out when chickens were supplemented with antibiotic growth promoter. Antibiotics are known to modulate gut microbiota signi cantly and hence information on true composition of gut microbiota without in uence of antibiotics of these breeds remains to be studied.
In the present study we use amplicon sequencing targeting hypervariable (V3-V4) region of 16S rRNA genes to compare gut microbiota of three indigenous Indian breeds originated in diverse geographical regions (Aseel, Ghagus and Nicobari) and a commercial broiler (Vencobb 400) reared under commercial set up but without any antibiotic. The Nicobari breed is an indigenous and endemic breed of chicken of Nicobar Islands which are an internationally acknowledged biodiversity hot spot, off the Indian mainland (got detached from Asian main-lands some 100 million years ago) and produces highest number of eggs among all indigenous chicken breeds of India [7] and is believed to have high disease resistance. Aseel is one of the important and most popular indigenous chicken breeds of Indian mainland and is well known for their meat quality with desirable taste and avor besides ability to thrive under adverse climatic and nutritional conditions [8]. Ghagus is another important native chicken breeds of coastal region of Southern India known for their meat quality with desirable taste and avor, disease resistance [9] and ability to perform under hot and humid climate and under low plane of nutrition.
Our hypothesis was that the gut micro ora would vary signi cantly between chicken breeds and lines, offering clues for development of breed/line speci c probiotics or feed additives for improvement in performance or feed e ciency or health or pathogen exclusion or for targeted genetic improvement by selective breeding for desirable type of gut microbiome.

Chicken breeds and experimental design
In the present study four chicken breeds or lines were chosen for comparison which included three indigenous Indian breeds (Nicobari, Aseel and Ghagus) and one global commercial broiler line (Vencobb 400 Care was taken to ensure that all birds received exactly the similar husbandry to minimise non-host variation. All chickens were fed maize and soybean-based balanced diet as per respective feeding standards for intensive production system with no antimicrobials having been provided while in production. Sequences were aligned to SILVA seed alignment (silva.seed_v138.align ; available in the MOTHUR website). Poorly aligned sequences were removed and overhangs at both ends were trimmed so that they overlap the same region. Unique sequences were screened and further de-noised based on pre-clustered command for up to 2 differences between sequences. Chimera sequences were checked and removed using VSEARCH [12] as implemented in MOTHUR. Sequences were then classi ed using naïve Bayesian classi er against RDP 16S rRNA gene training set (version 16) with bootstrap cutoff of 51% [13]. Sequences classi ed to unrelated taxon were removed. Clean sequences were subjected to operational taxonomic unit (OTU) clustering using DMSC software [14] at 97% similarity cutoff. DMSC output was converted to MOTHUR formatted list le and shared le for further analysis. OTUs were taxonomically classi ed using RDP classi er as implemented in mother using GreenGene Taxonomy (gg_13_8_99.gg.tax) database available in the MOTHUR website. MOTHUR formatted shared le and consenus taxonomy les were converted to biom formatted le and singleton and doubleton OTUs were removed from the BIOM le. The BIOM le along with the sample metadata les were uploaded to MetaCoMet web server [15] for plotting Venn diagram of OTUs. The breed speci c OTU Tables along with metadata were uploaded to METAGENassist website [16] for analysis where data were ltered as per default settings and samples were normalized using total sum (sample vs. Sample) and Pareto scaling (taxon vs taxon; mean centred and divided by the square root of standard deviation of each variable). The processed data were used for generation of correlation heatmaps using Spearman rank option. The BIOM (or MOTHUR generated shared and consensus taxonomy les) along with the metadata le and a NJ tree le prepared from the OTU representatives were uploaded to MicrobiomeAnalyst [17] for analysis of alpha diversity, beta diversity, differential abundance, biomarker identi cation and correlation network analysis. For the analysis of alpha diversity and beta diversity (nonmetric multidimensional scaling (NMDS) and principal coordinate analysis (PCoA)) data were normalized by cumulative sum scaling (CSS) method after disabling the default data ltering options for low counts and low variances (however, features appearing only in one sample were removed automatically as there was no option to disable such minimal ltering function). For the analysis of differential abundance (using edgeR), biomarker identi cation (using Linear discriminant analysis effect size or LEfSe) and correlation network data were ltered for low counts and low variances using the default setting to remove less informative features, to focus on important features and to improve downstream statistical analysis (this process removed 127 low count OTUs out of 296 taxonomic OTUs). Beta diversity pro ling and signi cance testing were carried out at different taxonomic levels like OTU, genus, family, class and phylum using PCoA as well as NMDS ordination based on different distance methods like Bray-Curtis dissimilarities, Jensen-Shannon diversion and Weighted Unifrac using statistical methods like permutational multivariate analysis of variance (PERMANOVA) and homogeneity of group dispersion (PERMDISP). Core microbiome analysis were carried out at OTU level where only OTUs with a mean normalized (CSS) relative abundance of at least 0.1% and having within breed prevalence of at least 50% in at least one breed were chosen to focus on the most commonly shared OTUs. On detection of signi cant difference in overall abundance between groups on edgeR analysis followed by Benjamini-Hochberg FDR correction for multiple comparison, groups were compared pairwise using nonparametric Mann Whitney U test (Wilcoxon rank sum test) as implement in SPSS [18]. To identify chicken breed/line-speci c biomarkers at multiple taxonomical levels analysis were performed using Linear discriminant analysis (LDA) effect size (LEfSe) algorithm using Benjamini-Hochberg false discovery rate (FDR) adjusted p value cutoff value of 0.05 and the logarithmic LDA score cutoff of 2 as well as 3.5 [19]. The LEfSe bar plots were created using MicrobiomeAnalyst but the cladogram was created using a standalone version of LEfSe [18]. In all analyses p values were corrected for Benjamini-Hochberg false discovery rate (FDR). BIOM data were rare ed to the minimum library size (at 12151 sequences per sample) and rarefaction analysis were carried out using MOTHUR. Rarefaction curve was visualized by creating plots using R package. Alpha diversity matrices were compared at OTU level using Kruskal-Wallis test followed by Dunn's test as implemented in SPSS [18 ] and Benjamini-Hochberg false discovery rate (FDR) adjustment of p-values. Correlation networks at different taxonomic levels were built based on the pairwise Spearman rank correlation coe cients where each node represents a taxon and two taxa are connected by an edge if the Spearman rank correlation between the two taxa meet the p-value (< 0.05) and correlation (> 0.5) thresholds. Taxonomic assignments were presented as Krona charts from CSS normalized relative abundance data using standalone KronaTools (version 2.7.1) [20].

Microbiome sequencing
High throughput sequencing generated 5.095 million raw reads corresponding to 4.08 Gbp of raw data from gut content of the 32 chickens. After read quality ltering, merging paired end reads, denoising, removing chimeras and ltering low quality sequences, the average number of quality controlled sequences per sample was 113, 028 (range, 50,392 − 295,955) ( Table 1). The 16 s rRNA gene amplicon sequencing results were deposited in the Sequence Read Archive of the NCBI (accession numbers: PRJNA641245 and PRJNA641779). Operational taxonomic unit (OTU) occurrence The indigenous Aseel breed presented the highest number of non-singleton non-doubleton (with > 2 members) OTUs (genetic distance based OTUs at 97% similarity curoff) (3461), followed by Ghagus (1911) and Nicobari (1767) ( Table 1). The commercial broiler line presented the lowest number of OTUs (857). Observed OTU numbers were signi cantly (P < 0.05) higher in the indigenous breeds than that of the commercial broiler line ( Fig. 1 and Additional le 1). Observed OTU numbers in the Aseel breed was higher than those of Ghagus and Nicobari. However, OTU numbers in Ghagus and Nicobari were comparable.
The mean observed richness (number of observed OTUs) was highest in the Aseel group followed by Ghagus and Nicobari and lowest in the broiler line. The species richness (or the number of species or OTUs) indices like ACE and Chao1 were higher (P < 0.01) in the indigenous breeds than that of the commercial broiler line. Among indigenous breeds Aseel had signi cantly higher ACE and Chao1 estimates than those of Ghagus or Nicobari.
The diversity (which takes into account both richness and evenness) estimators like Simpson, Shannon and Fisher were also higher (P < 0.01) in the indigenous breeds (Aseel, Ghagus and Nicobari) as compared to the commercial broiler line and these estimators were comparable among the indigenous breeds. Besides these estimators, rarefaction curves based on the Chao1 index were also plotted. The rarefaction curve depicts the correlation between the number of sequences and the number of OTUs and steeper the slope, the higher the diversity [21]. Rarefaction curve also indicated that the broiler line had lower diversity than those of the indigenous breeds (Additional le 3). Rarefaction curve approached asymptotic level for each breed or line, suggesting the availability of su cient reads to rep-resent each microbiome community.

Microbial beta diversity
The beta diversity (the partitioning of biological diversity among breeds or along a gradient, e.g., the number of species shared between two breeds or lines) analysis was undertaken to assess the relationship of microbial communities of different breeds/line using different metrices to calculate the dissimilarity/distance matrix, like Bray-Curtis, Jensen-Shannon, unweighted UniFrac and weighted UniFrac.
The correlation between the distance matrix and metadata categories was tested using PERMANOVA, which reports an R 2 value indicating the proportion of variation explained by this category, and a P value representing the statistical signi cance [22]. Homogeneity of group dispersions were also tested using PERMDISP. Beta diversity was visualized using nMDS as well as PCoA but due to space limitation only plots obtained using nMDS are presented. Results of beta diversity analysis including results of ordination using nMDS or PCOA are presented in Additional le 4.
PERMANOVA tests performed using all beta diversity metrices used in this study showed signi cant (P < 0.001) difference in community structure between different breeds/line both at OTU level and at Phylum level (Additional le 4 Beta diversity plots visualized using ordination methods nMDS at OTU and phylum level using nMDS method of ordination have been presented in Fig. 3 and Additional le 5, respectively. Jaccard index resulted in similar plots in NMDS scaling as that of Bray Curtis distance both at OTU as well as at Phylum Level and hence, plots for Jaccard index have not been presented. The NMDS scaling based on all ve distance metrices showed clear visual separation of breeds/line at OTU level. When Jensen-Shannon or weighted uniFrac distance was used there was high degree of overlap between the indigenous breeds while only minor overlap between broiler and the indigenous breeds was evident at OTU level. When distance metrices like Bray-Curtis or unweighted UniFrac or Jaccard were used for NMDS plotting there was no overlap between the broiler Line and indigenous breeds at OTU level but there were high degree of overlaps among indigenous breeds. At phylum level, there was high degree of overlap between the indigenous breeds but only minor overlap between Broiler and indigenous breeds (Aseel or Nicobari or Ghagus) was observed and extent of overlap between breeds/line varied with the distance metric used. In case of weighted UniFrac distance there was considerable overlap between broiler and Aseel, minor overlap between broiler and Nicobari and no overlap between broiler and Ghagus. While in case of unweighted UniFrac minor overlap between broiler and Nicobari was observed with no overlap between Ghagus and broiler or between Aseel and broiler. In case of Bray -Curtis distance/Jaccard index minor overlap was observed between broiler and Ghagus and no overlap between broiler and Nicobari or between broiler and Aseel. When Jensen -Shannon divergence was used there was minor overlap between broiler and Aseel and no overlap between broiler and Ghagus or broiler and Nicobari. Overall, it is noticeable that even at Phylum level there was very low level of overlap between indigenous breeds and the broiler line indicating almost completely different microbial community compositional distribution pattern between these two categories.
Differential abundances at different taxonomic levels Abundance of few phylotype OTUs such as OTU100296 (family S24-7) and OTU1758401 (family SMB53) were lower in Nicobari than in Aseel or Ghagus. Abundance of OTU4369050 (family Fusobacteriaceae) was higher in Nicobari than in Aseel or Ghagus. Abundance of OTU1066621(genus Prevotella) was signi cantly higher in Ghagus than in Aseel or Nicobari.
Out of 91 genera having mean abundance of ≥ 4 and prevalence of > 20%, twenty four genera were signi cantly different in relative abundance between breeds/line. Genera having signi cant differences in abundance between breeds/line have been presented in Additional le 6. Sequences not assigned to any genera remained major part of total sequences in all groups. The core gut microbiome Phylotype-OTUs with a mean normalized (CSS) relative abundance of at least 0.1% and having within breed prevalence of at least 50% in at least one breed were considered to de ne core microbiome in different breeds/line. Analysis of the prokaryotic community composition at the OTU level indicated that of the 296 phylotype-OTUs detected, only 22 phylotype-OTUs were present in > 0.1% relative abundance in at least 50% of the birds in at least one breed or line (Fig. 6). Twelve, ve and two of the 22 phylotype-OTUs belonged to the phyla Firmicutes, Bacteroidetes and Proteobacteria, respectively with the rest belonging to other diverse phyla like Fusobacteria and Cyanobacteria. Only 5 phylotype-OTUs (OTU1000113 belonging to the order Clostridiales, OTU100567 belonging to the genus Ruminococcus, OTU1010876 belonging to the genus Oscillospira, OTU839684 belonging to the family Lachnospiraceae and OTU98948 belonging to the family Ruminococcaceae) occurred as core microbiome across all breeds or lines and these core OTUs represented 26% of total microbial count (Fig. 6). The OTU839684 belonging to the family Lachnospiraceae alone accounted for 13% of total microbiome (or 50% of core microbiome) count of all chicken breeds/line. Three phylotype-OTUs were unique to the broiler line (OTU181074 belonging to the genus level group cc_115, OTU2229500 belonging to the species Subdoligranulum variable, OTU549991 belonging to the species Lactobacillus helveticus). Two phylotype-OTUs were unique to the Aseel breed (OTU586453 assigned to the family Christensenellaceae and OTU60774 assigned to the species Anaerobiospirillum thomasii). One phylotype-OTU was unique to the Nicobari breed (OTU4369050 belonging to the genus Fusobacterium). Seven phylotype-OTUs (OTU1000062 belonging to the order Bacteroidales, OTU100296 belonging to the family level gut group S24_7, OTU102407 belonging to the genus Bacteroides, OTU1057116 belonging to the phylum Bacteroidetes, OTU1105376 belonging to the genus Suterella, OTU168571 belonging to the species Bacteroidea barnesiae and OTU4324240 belonging to the species Faecalibacterium prausnitzii) were core OTUs across the three indigenous breeds but were not consistently detected in the broiler line. These indigenous chicken speci c core OTUs represented 49.3% of total prokaryote count of indigenous birds.

Correlation analysis
Family level correlations among microbes in Aseel, broiler, Ghagus and Nicobari are shown in Additional

Discussion
Demand for poultry meat and eggs are increasing rapidly in South Asia, where global commercial-type lines contribute most of the meat and egg production but popularity of indigenous chicken grown organically without antibiotic growth promoters are increasing at a relatively faster rate. Gut microbiota plays a key role in the development and functioning of gut including pathogen exclusion and therefore decreasing the expense of energy that birds normally invest in keeping the immune system active against these pathogens. It has been suggested that assessment of microbiome structure in gut in a comprehensive manner may further development of strategy for improving growth, feed e ciency, survivability or lower pathogen shedding via the development of host speci c probiotics [4]. The advent of high throughput sequencing and omics approaches as tools for the study of microbial communities has allowed a detailed characterization of the gut microbiota of chickens in a quick and robust fashion, without the need to culture the microorganisms. This is the rst study exploring gut microbiome of Nicobari and Ghagus breeds of Chicken breeds originated in Nicobari islands (a well known biodiversity hotspot) and coastal India, respectively. Here we have attempted to identify differences in gut microbial community structure of chicken breeds originated in diverse geographical locations and also with a global commercial broiler line.
The microbial communities differ through the chickens gut intestinal tract with particular microbial pro les detected in crop, gizzard, ileum, cecum and colon of broiler chickens [23]. Here we have analysed microbiota from entire hindgut (duodenum to cloaca including caecum) to focus on segments generally considered to be most important for gut health and function.
The initial colonization of the gastrointestinal tract of birds occurs naturally after hatching and can even begin before, by passing of microorganisms through the pores of the eggshell [24]. After the initial colonization of the intestine the species richness and complexity of the population structure of the microbiota increase as the birds grow, until microbiota reaches a state of stabilization. This process normally occurs in commercial broiler chickens around 3 weeks of life [25]. However, development times and succession patterns of intestinal microbiota species can vary depending on the genetic makeup of the birds and management factors. Here, we selected time points late in the production cycle of each breed or line (as each breed matures i.e., at marketing age) to permit assessment of the outcome of colonisation throughout each chickens' life.
In the present study indigenous chicken breed presented higher number of OTUs (1767 to 3461) than that of the commercial broiler line (857). In contrast, in an earlier study the commercial broiler line Cobb 400 was reported to have more OTUs (1273) than Indian indigenous chicken breeds such as Aseel (735 to 1134) or Kadaknath (816 to 833). In general, more number of OTUs were detected in this study as compared to that of Pandit et al. [6]. This may be partly attributed to the fact that in the current study no antibiotic growth promoter (AGP) was used and we have analysed gut content from entire hindgut (duodenum to cloaca including caeca) whereas in the study reported by Pandit et al. [6] only cecal content of birds fed diet containing AGP were analysed.
A total of 135 (out of 296) phylotype-OTUs (with > 2 members) were shared by chickens from all groups.
The gut microbiota of Indian native chicken breeds evaluated here were dominated by sequences belonging to the phyla Bacteroidetes and Firmicutes whereas the gut microbiota of the commercial line was dominated by sequences representative of the phylum Firmicutes. Within Indian indigenous breeds, Firmicutes/Bacteroidetes ratio were more and less similar in Ghagus and Nicobari breed but substantially different from that of Aseel. Our data are in discordance with a previous report by Saxena et al. [5] showing the dominance of Firmicutes in gut microbiome of Asamese breed of Indian Indigenous chicken. Pandit et al. [6] reported that Bacteroidetes was the dominant phyla in most of the gut caecal samples of Indian indigenous breeds like Kadaknath and Aseel and Firmicutes were more common in Cobb 400 samples, which is in concordance with our current study. Generally higher Firmicutes /Bacteroidetes ratios have been shown to be correlated with obesity in human [26]. Both Firmicutes and Bacteroidetes are primarily carbohydrate fermenters. Firmicutes are known to produce both butyrate and propionate, whereas Bacteroidetes primarily produce propionates as fermentation end product [27]. Within Firmicutes, different genera under the order Clostridiales were predominant in different breeds or line. The Bacilli members, like Lactobacillus spp., possessing prebiotic and probiotic activities, were present in very high proportions in the broiler line but were in small proportions in indigenous breeds. However, the genus Faecalibacterium (under order Clostridiales and family Ruminococcaceae), known to produce butyrate and thus having a crucial role in maintaining gut health and host well being [28], were present in signi cant proportion in indigenous breeds but not in the broiler line. The genus Bacteroides and an unknown genus under the order Bacteroidales constituted the top two dominant genera in all the indigenous breeds, whereas Lactobacillus and an unknown genus under the family Lachnospiraceae constituted the top two dominant genera in the boiler line. It has been shown that Faecalibacterium cooccurs with several members of Bacteroidetes in gut [29] and it has been suggested that Faecalibacterium may rely on Bacteroides for cross feeding [28]. Interestingly, as in case of broiler, an unknown genus under the family Lachnospiraceae constituted very high proportion (11%) of microbiota in Aseel but were detected in relatively lower proportion in other indigenous breeds. Gut microbial composition is mediated by many factors such as geographical location, host diet and administration of antibiotics and other medicines. It has been shown that the succession of changes in gut microbiota correlates with changes in the cytokine pro le expressed by host intestinal cells [30].
In the present study alpha diversity estimators measuring species richness (Chao1and ACE) as well as diversity (Simpson, Shannon and Fisher) and rarefaction curve indicated that the broiler line VenCobb 400 had lower diversity than those of the indigenous breeds, in disagreement with a previous study [6].
Beta diversity analyses were carried out using different types of distance or dissimilarity metrices. The Jaccard distance is based on presence or absence of OTU/species and does not include abundance information. The Bray -Curtis dissimilarity measure is based on abundance of each OTU/species in different communities and it ignore cases in which the species /OTU is absent in both community and is dominated by the abundant species so that rare species add very little to the value of the coe cient [31].
The Jensen-Shannon distance is based on probability distribution of two microbial communities [32] and is a measure of divergence between distributions accounting for both presence and abundances of OTUs.
UniFrac is a distance measure based on phylogenetic information (degree of phylogenetic divergence or distance) of OTUs of microbial communities. Unweighted UniFrac measures the distance between two communities by calculating the fraction of the branch length in a phylogenetic tree that leads to descendants in either, but not both, of the two communities and is considered as a qualitative measure of diversity which takes into account presence/absence of data but does not include abundance data to compare community composition. Weighted UniFrac is a variant of the original unweighted UniFrac measure that weights the branches of a phylogenetic tree based on the abundance of information and is considered as a quantitative measure of diversity, which also take the relative abundance of each type of organism into account [33].
Beta diversity analysis involving different distance metrices has indicated clear separation of microbiome at both OTU as well as phylum levels by chicken breed or line indicating a strong host component in microbiome composition, in agreement with previous studies [6,34,35]. Low levels of overlap between microbiome of the broiler line with those of indigenous breeds also suggest highly different microbial composition across these two categories. Interestingly, NMDS plots also indicates that compositional distribution of gut microbiome of the island breed (Nicobari) have very high degree of similarity with both the coastal breed (Ghagus) and the breed from Indian mainland (Aseel) indicating possibility of strong evolutionary linkage among all these indigenous breeds.
Slightly different results between unweighted unifrac and weighted unifrac beta diversity metrices suggest that the most abundant taxa were more phylogenetically related compared to the low abundant taxa. Weighted unifrac distance considers presence or absence, abundance and phylogenetic relatedness and is considered as a measure of most recent common ancestor and also gives more weightage to deep branches of phylogenetic tree, whereas unweighted unifrac distance only consider presence and absence and no consideration of abundance and gives less importance to deep branches and is in uenced by all of the branches [36].
Differential abundance analysis using edgeR indicated that abundances of several phylotype-OTUs belonging to the order Bacteroidales and an phylotype-OTU under genus Faecalibacterium were signi cantly lower in the broiler line VenCobb 400 but abundances of few phylotype-OTUs belonging to the genera such as Lactobacillus, Bilophila, Clostridium, Weissella, Eggerthella and De uvitalea were higher (P < 0.05) in the broiler line as compared to indigenous breeds.
At order level 18, 6, 12, 14, 17 and 1 orders were differentially abundant between Aseel vs broiler, Aseel vs Ghagus, Aseel vs Nicobari, broiler vs Ghagus, broiler vs Nicobari and Ghagus vs Nicobari, respectively. Similar trend was also observed at other taxonomic levels indicating highest overlap in microbial community composition between the coastal breed (Ghagus) and the island breed (Nicobari) among indigenous breeds and very low overlap between the broiler line and all indigenous breeds. LEfSe biomarkers also indicated a possible strong host genetic in uence on gut microbiome.
On the other hand many breed or line speci c prokaryotes having potential bene cial effects on gut health of hosts (having probiotic or butyric acid producing properties) were also associated as biomarkers such as: Faecalibacterium, Coprococcus, Gemmiger in Aseel; Lactobacillus, Subdoligranulum, Dorea, Blautia and Bi dobacterium in the broiler line; Akkermansia in Ghagus; Barnesiella in Nicobari.
Using 0.1% relative abundance and 50% prevalence cutoff criteria, 22 phylotype OTUs out of 296 phylotype OTUs representing 26% of total microbial count were found to qualify as core OTUs in at least one breed or line. In contrast, only 5 phylotype-OTUs representing 13% of total microbiome occurred as core microbiome across all breeds or line.
Interestingly, few potentially pathogenic or harmful ( Sutterella in Aseel and Nicobari, Fusobacterium in Nicobari and Anaerobiospirillum in Aseel) prokaryotes were detected above the abundance and prevalence cutoffs thus qualifying to be part of core microbiome in different breeds or line. This is in line with an earlier report where Campylobacter were detected above the 1.0% cut-off in Kadaknath and Aseel, but not in Cobb400 or Ross 308 [6] .
In contrast many prokaryotes having potential bene cial effects on gut health of hosts (having probiotic or butyric acid producing properties) were also detected above the 0.1% relative abundance and 50% prevalence cutoff in different breeds and lines.
Futher work will be required to ascertain exact role of the pathogenic bacteria detected as biomarkers or core microbiome. However, besides being opportunistic or primary pathogens, in ammatory and immune responses induced by these pathogens have been suggested to in uence the intestinal environment, host immunity and its bacterial communities [37,38].
Despite inter-individual differences in community composition, a core set of microbes shared across individuals of a particular breed suggesting presence of breedwise distinct community composition.
Millions of years of co-evolution between the host and microbes have led to a mutualistic symbiosis in which the microbiota contributes to many host physiological processes and the host, in turn, provides a nutritious and hospitable environment to the microbes. Further, the normal gut symbionts forms a stable community that resists the invasion and colonization of non-native bacteria [39].
Potentially opportunistic pathogens such as those within the families Enterobacteriaceae, Clostridiaceae, Campylobacteriaceae, Fusobacteriaceae. etc exhibited a strong correlation with each other and a negative correlation with bene cial bacteria, having probiotic (competitive exclusion) properties or SCFA production ability, belonging to families such as Christensenellaceae, Lactobacillaceae and Bi dobacteriaceae. Several breed or line speci c microbial family level clusters with strong positive correlations with each others and negative correlations with other clusters were detected in all the breeds or lines suggesting existence of strong interactions among different microbial groups. Correlation network analysis data both at genus and order levels further substantiates existence of diverse and complex microbial networks. This is in line with earlier reports [6,40].

Conclusions
In this study, we carried out the rst comprehensive analysis of chicken gut microbiome of Nicobari and Ghagus breed of Indian native chicken originated in biodiversity hotspots like Nicobari island and coastal India, respectively and also carried out comparative analysis of diversity and composition of gut microbiota of three indigenous breeds and one commercial broiler line. The study presented here provides important insights into chicken breed or line speci c variation in enteric bacterial occurrence, diversity and complex microbial networks. Our amplicon sequencing results emphasized the more similarity of the microbiota within the gut lumen of indigenous breeds as compared to the commercial broiler line VenCobb 400, but exhibited distinctive taxonomic differences between them as well. The study also indicated that among native breeds there is more similarity of gut microbiome of the island breed (Nicobari) with that of the coastal breed (ghagus) as compared to the mainland breed (Aseel). The study presented here indicated existence of breed or line speci c core microbiome as well as across -breed or line core microbiome in chicken and occurrence of both bene cial and potential opportunistic pathogenic microbes as part of core microbiome. A deeper understanding of host-microbiome interactions as emanated from the current study may support development of strategies including development of breed speci c feed additives and probiotics for enhanced productivity from unconventional or low value diets, for prevention of colonisation by pathogenic and zoonotic organisms and for deleopment of alternative to antibiotic growth promoters. Figure 5 Breed or line speci c and across-breed or line core phylotype-OTUs. Phylotype-OTUs with a mean normalized (cumulative sum scaling) relative abundance of at least 0.1% and having within breed or prevalence of at least 50% in at least one breed were taken into account The size of bubbles in the bubble plot indicates normalized (cumulative sum scaling) abundance of each OTU.   The genus and order level correlation networks have been presented in Additional le 15 and Figure 7