The gastrointestinal tract of an animal is colonized by a complex and abundant ensemble of microbes (Ley et al., 2008). Over millennia, microbes have evolved to coexist with their host and have a symbiotic role for avoiding elimination by the immune system (Garrett et al., 2010). In our previous studies, the heritability of microbiota was proposed to discern genetic from environmental factors that manage the gut microbiota and to potentially distinguish functionally important microbial members (Meng et al., 2014). In addition to heritabilities (Zhao et al., 2013; Meng et al., 2014; van Opstal and Bordenstein, 2015), we reported significant genetic correlations between specific microbes confirming that the genotype of the host was involved in microbe-microbe interactions (Zhao et al., 2013; Meng et al., 2014). The Christensenellaceae has been identified as the most highly heritable taxon and was significantly enriched in lean than obese humans (Goodrich et al., 2014). The heritability and significant genetic correlations of specific microbes suggested co-microevolution relationships for the host-gut microbiota axis. Therefore, increased attention should be focused on investigating co-microevolution mechanisms between gut microbiota and host genetic variation.

The gut microbiota is shaped by both the environment and genetics of the host, with the immune system in particular, considered to be involved in modulating the gut microbiota (Slack et al., 2009). The mutation of a single gene relating to immune response can have a considerable influence on the structure and composition of gut microbiota. For example, nucleotide-binding oligomerization domain protein 2 (NOD2), which recognizes peptidoglycan to enhance systemic innate immunity, can alter the composition of the gut microbiota (Clarke et al., 2010). The mutation of adaptor molecule MyD88 can impact gut microbial composition (Kubinak et al., 2015) and is associated with both a shift in bacterial diversity and a greater proportion of segmented filamentous bacteria in the small intestine (Larsson et al., 2012). Mice, deficient in activation-induced cytidine deaminase or recombination-activating gene 2 (Rag2) lack IgA (immunoglobulin A), exhibited predominant and persistent expansion of segmented filamentous bacteria (Suzuki et al., 2004). Thus, the immune system can have an important influence on the composition of the microbiota. Conversely, certain members of the microbiota can regulate the development of specific subsets of immune cell populations and establish the set-point for a steady-state immune system (Ivanov and Honda, 2012; Honda, 2015). Colonization of mice with segmented filamentous bacteria (SFB) induced an accumulation of T helper 17 cells and an increase in T helper 1 cells (Gaboriau-Routhiau et al., 2009). Clostridium has the ability for expansion of lamina propria and systemic T cells regulation (Atarashi et al., 2011). Polysaccharide A (PSA) of Bacteroides fragilis induces an IL-10 response in intestinal T cells, which prevents the expansion of T helper 17 cells and potential damage to the mucosal barrier (Round et al., 2011). Considering their long history of co-microevolution, while the gut microbiota and the host immune system bidirectional communicate in mutually beneficial ways, co-microevolution mechanisms between gut microbiota and the immune system of the host remains obscure. Especially how the antibody response involved in humoral immunity and gut microbiota modulate each other still remains unknown.

The factors that influence the mechanisms of the interactions between the host and the gut microbiota are likely to be small, and thus detecting them requires well controlled effects other than those of the host genotype. Thus, choosing a model organism that is maintained in a controlled environment should enhance our understanding of the relationships between gut microbiota and host genetic factors. The chicken, which bridges the evolutionary gap between mammals and reptiles, can serve as such an organism due to the characteristics of its less complex gut microbiota and minimal maternal effect.

To explore interactions between the host immune system and gut microbiota, we used next generation sequencing technology to investigate the composition of gut microbiota of White Leghorn chickens that had undergone long-term, bidirectional selection for a single immune trait (Siegel and Gross, 1980). In brief, 40 generation of selection from a common founder population, was performed for high (HAS) or low (LAS) antibody response 5 days post-injection of a non-pathogenic antigen, sheep red blood cells (SRBC) (Siegel et al., 1982; Boa-Amponsem et al., 1997). In addition, two sublines (HAR and LAR), in which selection was relaxed in generation 23 were also used in this study to investigate the response from gut microbiota after selection was relaxed. QTL mapping of host populations revealed 11 genomic regions associated with antibody response traits (Dorshorst et al., 2011). Furthermore, two SNP markers were significantly associated with day 5 antibody titers and overlap with SEMA5A and TGFBR2, respectively (Dorshorst et al., 2011; Lillie et al., 2017). Additionally, there were different haplotypes among lines in the major histocompatibility complex (MHC). The MHC haplotypes were fixed by generation 32 for B²¹ in HAS and B¹³ in LAS. The long-term divergent selection resulted in a 6.5-fold difference in antibody titers between HAS and LAS with modest regressions to the origin in the relaxed lines (Lillie et al., 2017).

The microbiota can induce adaptive changes in host immunity (Farkas et al., 2015), and our results show that the lines selected for differences in immune response contribute to adaptive changes in the microbiota. To further investigate the association of microbiota functions for high vs. low and selected vs. relaxed lines, we used PICRUSt to produce predicted metagenomes from 16S rRNA gene sequence data and then applied STAMP to estimate differences in predicted abundances of KEGG. From this analysis we observed differences between the lines in the endocrine system, excretory system, transport and catabolism, energy metabolism, genetic information processing, and, metabolism of cofactors and vitamins were differed (HA > LA). Pathways allocated for signaling molecules and interactions (LAR), immune system diseases (LAR), glycan biosynthesis and metabolism (LAS), infectious diseases (LAS) and membrane transport (LAS) were different between the lines (HA < LA) (Figure 4C). Furthermore, the three pathways of primary bile acid biosynthesis, secondary bile acid biosynthesis and systemic lupus erythe matosus related to immune system were enriched in LAR (Figure S5b). For selected vs. relaxed lines, microbiota in the former had a greater abundance of functions involved in metabolic pathways involving infectious diseases (LAS line), metabolism of Terpenoid and polyketides, lipid metabolism, transport and catabolism, xenobiotic biodegradation and metabolism, neurodegenerative diseases, metabolism of other amino acids, excretory system, digestive system, cancers, cardiovascular diseases and circulatory system than the later. In contrast, the relaxed lines contained higher energy metabolism, genetic information processing, metabolism of cofactors and vitamins, metabolic diseases, biosynthesis of other second metabolites, nervous system, carbohydrate metabolism, immune system (HAR line) and transcription than their respective selected lines (Figure 7). Taken together, the analysis of microbial function revealed that the pathway of infectious diseases enriched in LAS (Figure 4C). When selection was relaxed, enriched pathway of infectious diseases was replaced by immune system diseases in LAR (Figure 4C).

In this study, the high and low selected and relaxed lines of chickens that originated from a common founder population of chickens (Siegel and Gross, 1980; Siegel et al., 1982; Boa-Amponsem et al., 1997; Dorshorst et al., 2011; Zhao et al., 2012; Farkas et al., 2015; Lillie et al., 2017) provide evidence for humoral immunity-microbe co-microevolution. Dramatic differences in gut microbiota were observed in these lines that had undergone long-term artificial and relaxed selection for high or low antibody titers to SRBC (Figure 1B). Twenty-one genera were identified as significantly different between the high and low selected lines with 8 genera that belonged to Proteobacteria all enriched in the LAS (Table 1). The Proteobacteria is a major phylum of Gram-negative bacteria. They include a wide variety of pathogens and have an outer membrane composed mainly of lipopolysaccharides (Matsuura, 2013) which acts as aprototypical endotoxin caused immunity responses (Conde-Álvarez et al., 2012; O'Neill et al., 2013). The phenomenon is consistent with other studies, such as, human immunodeficiency virus (HIV) infection which is a chronic illness characterized by progressive CD4⁺ T cell loss and associated with increases in Proteobacteria (potentially pathogenic taxa) in the gut microbiota (Dinh et al., 2014; Mutlu et al., 2014; Serrano-Villar et al., 2016). In addition, microbial functions also revealed that pathways of infectious diseases were enriched in LAS, results consistent with those observed at the taxa level. The long-term selection for antibody response to SRBC, resulted in lines that exhibited antibody responses which may inhibit colonization of certain types of bacteria. Then when selection was relaxed, there was considerable shifting of gut microbiota where 35, 39, and 27 genera were significantly different in comparisons of HAR vs. LAR, HAS vs. HAR, and LAS vs. LAR. Concomitantly, variation in antibody titers was also observed, inferring that in the process of co-microevolution, under the artificial selection, variation of antibody titers dominated the microorganisms to colonize in host gut. Conversely, the shift in genera that occurred when selection was relaxed may contribute to reconstructed antibody responses.

When selection was relaxed, there was opportunity for adaptive changes according to the pressure of natural selection (Figure 1A). Thus, the fragile immune homeostasis constructed by artificial selection had costs with natural selection for an intermediate optimum for immune homeostasis. There is evidence that the gut microbiota can induce adaptive change of host immunity. For example, flagellin stimulated intestinal CD103⁺ CD11b⁺ dendritic cells produce Interleukin 23 (IL-23) which enhance mucosal innate immune defenses (Kinnebrew et al., 2012) and regulate TLR4 gene expression by modifications of the level of methylation in 5′ region of the promotor by commensals (Takahashi et al., 2011). In our study, the butyrate-producing bacteria Oscillospira and Ruminococcaceae (Faecalibacterium, Oscillospira, Anaerotruncus, and Butyricicoccus) were greater in HAR than LAR and HAS (Figures 4A,B, Figure S3b and Table S8), respectively. Butyrate, a main type of short-chain fatty acids (SCFA), has been involved in the regulation of macrophage function by inhibiting histone deacetylases (Chang et al., 2014). In particular, butyrate endows dendritic cells with a superior ability to facilitate regulatory T cells differentiation (Wang et al., 2009; Arpaia et al., 2013), or inhibits histone deacetylase influence regulatory T cells by binding to GPCR43 (Tao et al., 2007; Smith et al., 2013). In some autoimmune diseases with aberrant antibody production, there is a decrease in the function or number of regulatory T cells (de Lafaille and Lafaille, 2002; Balandina et al., 2005; Mqadmi et al., 2005).

Regulatory T cells have a critical role in immune homeostasis and in suppressing excessive immune responses by the host (Sakaguchi et al., 2008). For example, regulatory T cells can inhibit IgG production of systemic lupus erythe matosus (SLE), a form of an antibody-driven autoimmune disorder (Facciotti et al., 2016). In addition, regulatory T cells can indirectly suppress Ig production via regulating T helpers cell in stimulating B cells (Lim et al., 2004), or by directly restraining Ig production dependent on B cell responses accompanied by inhibition of Ig class switch recombination (Lim et al., 2005). The cell wall components of Lactobacillus contain microorganism-associated molecular patterns (MAMPs) that are recognized by specific pattern recognition receptors (PRRs) that are expressed in the intestinal mucosa of the host (Wells et al., 2011). They include Toll-like receptors (TLRs), NOD-like receptors (NLRs) and C-type lectin receptors (CLRs). Through these signal pathways Lactobacillus is involved in the regulation of the immune system of host. For example, Lactobacillus species may deliver signals in dendritic cells (DCs) through TLR-2, thereby promoting the activation of CD4⁺ and CD8⁺ T cells to T-helper cells (Cella et al., 1997; Mohamadzadeh et al., 2005). Depending on the pattern, DCs activate and expand T-helper cells, which in turn induce B-cell growth and antibody production (Dubois et al., 1997; Banchereau and Steinman, 1998). In addition, some Lactobacillus species can increase Ig production (Rizzardini et al., 2012; Nishihira et al., 2016).

Microbiota functions associated with primary bile acid biosynthesis, secondary bile acid biosynthesis and systemic lupus erythe matosus were relatively greater in LAR than the other lines (Figure S5b). Bile acids can be sensed by a subset of nuclear receptors including the farnesoid X receptor (FXR) (Vavassori et al., 2009) and G-protein coupled receptor 5 (TGR 5) (Keitel et al., 2007) to modulate the immunity system. Moreover, systemic lupus erythe matosus is associated by an excessively high production of Ig (Facciotti et al., 2016). Implying that the microbes in LAR may contribute to greater production of Ig and thus the gut microbiota, especially Oscillospira and Lactobacillus may play a key role in immune homeostasis. Through bidirectional selection from a common founder population for humoral immunity, changes occurred in the host microbiota axis which in turn became more balanced when selection was relaxed. The bidirectional interplay demonstrated changes in the microbiota as correlated responses. When selection was relaxed the microbiota coevolved reflecting a coevolution of host and gut microbiota. Such a dynamic linkage between humoral immunity and gut microbial ecology, together with our results, indicates that manipulation of gut microbial communities could be another approach in the treatment of autoimmune diseases.

LY, PS, YZ, and HM performed the experiments. SL, JD, RD, CFH, CH, and KX contributed to some of the experiments. LY, PS, YZ, and HM designed experiments, analyzed data, and wrote the paper.