Innate Immune Homeostasis by the Homeobox Gene Caudal and Commensal-Gut Mutualism in Drosophila

The mucosal epithelia of all metazoans, such as those found in the gastrointestinal tract, are in intimate contact with a large number of commensal microbiota (1). As a consequence, commensal bacteria are known to influence many aspects of the host gut physiology, including innate immunity, development, and homeostasis (2–7). However, the lack of a genetically amenable animal model has limited in-depth analyses of gut-microbe interactions in vivo.

Recent studies have shown that the Drosophila gut activates host antimicrobial defense through the production of microbicidal reactive oxygen species (ROS) and antimicrobial peptides (AMPs) (8–11). During most gut-pathogen interactions, intestinal redox homeostasis, mediated via infection-induced ROS generation by the dual oxidase enzyme and subsequent ROS elimination by immune-regulated catalase, is critical for host survival (8, 9). The direct contact between gut epithelia and ingested pathogens also activates the immune deficiency (IMD) pathway and subsequent nuclear localization of the p105-like NF-κB, Relish, which in turn leads to de novo synthesis of diverse immune effectors: AMPs and immunosuppressive enzymes such as the peptidoglycan recognition proteins PGRP-SC and PGRP-LB (10, 12–16). Although pathogen-initiated gut immunity is fairly well documented in Drosophila (8–10, 14, 16, 17), the molecular interaction between commensals and the Drosophila gut is poorly understood.

Role of gut commensal microbiota in the IMD-Relish pathway. To investigate the molecular mechanism for commensal-gut interactions, we first asked whether the commensal microbiota could elicit host gut immune responses. We produced germ-free wild-type (GF^WT) and conventionally reared wild-type (CR^WT) animals (fig. S1) to enable examination of the gut IMD-Relish pathway potential in the absence or presence of commensals. A large amount of the nuclear-translocated active form of Relish was detected in the intestine cells in the presence of commensals (CR^WT flies) and was even more pronounced after gut infection with the Drosophila pathogen Erwinia carotovora carotovora-15 (Ecc15) (Fig. 1A).

• Download high-res image
• Open in new tab
• Download Powerpoint

The well-characterized constitutive nuclear localization of Relish in the intestines of CR^WT flies was almost completely abolished in the absence of commensals in GF^WT or in antibiotics-treated CR^WT flies. Similarly, it was not seen in flies carrying a mutation in the IMD-Relish pathway (CR^Dredd flies) (Fig. 1A). The expression levels of Relish-dependent immunosuppressive enzymes such as PGRP-SC and PGRP-LB (12, 13) were significantly higher in the guts of CR^WT flies than in the guts of GF^WT or CR^Dredd flies (Fig. 1B), which confirms that the IMD-Relish pathway was activated by commensals under normal gut conditions. However, the Relish-dependent immune effector molecules, such as AMP genes including Cecropin (Cec) and Diptericin (Dpt), were largely silenced in the CR^WT gut despite chronic Relish activation. Similarly, no significant difference in AMP expression levels was observed between the GF^WT and CR^WT midguts (Fig. 1B). Taken together, these results indicate that although commensal organisms of the gut can chronically induce a high level of local IMD–NF-κB pathway activation, only a subset of target genes (notably excluding AMP genes) are activated.

Role of Caudal in gut AMP repression. We next investigated the potential mechanisms of repression of the gut AMP genes to determine how the selective silencing toward commensals might occur. Currently, the cis-regulatory elements controlling epithelial AMP gene expression are poorly understood. It is known, however, that the κB elements in the promoter regions responsible for nuclear factor kappa B (NF-κB)–Relish binding are essential for inducing epithelial AMP expression (14–16), whereas the CDREs, responsive elements for the homeobox transcription factor Caudal (Cad) responsible for Cad binding, are found to be critical for constitutive AMP expression in certain types of epithelia, such as salivary glands and the ejaculatory duct (18). The homeobox transcription factor Cad was originally identified on the basis of its regulatory role in the anteroposterior body axis formation of the Drosophila embryo (19, 20). Because Cad expression in postembryonic life is known to be mostly restricted to the intestine (19) (fig. S2), we analyzed the contribution of CDREs to gut AMP gene expression in vivo. To accomplish this, we used green fluorescent protein (GFP) reporter–expressing transgenic flies carrying a Cec promoter in which the CDREs were mutated (Cec^CDRE-mut-GFP) and compared GFP expression with that of transgenic flies carrying the wild-type promoters fused to GFP (Cec-GFP) (Fig. 1C). We were unable to detect any Cec reporter activity in the midgut of Cec-GFP flies under normal conditions (Fig. 1C). Interestingly, however, Cec^CDRE-mut-GFP flies were found to exhibit high constitutive expression of Cec reporter activity in the posterior midgut and the proventriculus in the absence of oral infection (Fig. 1C).

To test whether Cad was involved in the negative regulation of Cec expression through CDREs, we generated transgenic flies that carried the UAS-Cad-RNAi construct to mimic the loss of function (fig. S3) via RNA interference (RNAi). A spontaneous activation of all tested AMPs was observed in the gut of CR^Cad-RNAi flies under conventional conditions without microbial infection, which was not the case in the control flies (Fig. 1D). Furthermore, similar AMP derepression was also observed in Cad-RNAi flies with different GAL4 drivers (fig. S4). The role of Cad as a repressor was further confirmed by a strong expression of Dpt reporter activity observed in the gut of CR^Cad-RNAi flies carrying the Dpt-LacZ reporter (fig. S5). Cad-dependent AMP repression is highly tissue-specific because no AMP derepression was observed in the fat body of CR^Cad-RNAi flies (Fig. 1D). Introduction of Cad-RNAi had no effect on the expression of PGRP-SC and PGRP-LB (Fig. 1D), which indicates that any repressive role of Cad is restricted to a distinct subset of NF-κB–dependent genes such as AMPs. When we examined Cad-RNAi–induced AMP derepression in the IMD-Relish pathway mutant genetic backgrounds (CR^{Cad-RNAi + Dredd} or CR^{Cad-RNAi + TAK1} flies) or in the absence of commensals (GF^Cad-RNAi flies), the high level of AMP derepression was completely abolished (Fig. 1E).

These data show that Cad acts as a gut-specific transcriptional repressor exerting its antagonistic role in commensal-induced NF-κB–dependent AMP induction. Furthermore, the overexpression of Cad in the gut could abolish the infection-induced AMP expression (Fig. 1F). Thus, it is likely that a dynamic equilibrium between Cad and Relish is one of the major mechanistic aspects determining the selective deployment of gut IMD-AMP pathway potential (21) (Fig. 1 and fig. S6). Because AMPs have a microbicidal effect against a broad spectrum of microorganisms, Cad-mediated AMP repression is likely required for establishing an optimal environment for the commensal microbiota.

Role of Cad in gut homeostasis. Several recent lines of evidence suggest that deregulation of the intestinal NF-κB pathway may be relevant to the etiology and pathology of many important diseases, including inflammatory bowel diseases (IBDs) (22–29). Given that IBDs typically involve apoptosis of intestinal cells, we examined whether cell death might be occurring in the gut of Cad-RNAi flies. Time-course analyses with adult flies showed that gut epithelial cell apoptosis was detected in Cad-RNAi flies at day 18, but not in control flies of the same age (Fig. 2A). Cell death was also observed in the intestines of CR^Cad-RNAi flies with different GAL4 drivers: c729-GAL4, Cad-GAL4, and Da-GAL4 (Fig. 2, A and B). Interestingly, gut epithelial cell apoptosis was abolished in the GF^Cad-RNAi gut (Fig. 2C), which suggests the involvement of commensal organisms in Cad-RNAi–mediated gut pathology. To determine whether the high apoptosis level seen in the Cad-RNAi gut was due to secondary effects of AMP hyperactivation, we overexpressed two AMP genes (Cec and Dpt) to mimic the Cad-RNAi genotype. AMP overexpression in CR flies (CR^Cec+Dpt) was sufficient to induce a high level of gut epithelial cell apoptosis (Fig. 2C). This was not the case in the absence of commensal microbiota (GF^Cec+Dpt) (Fig. 2C). When we overexpressed only a single AMP (Cec or Dpt) in CR flies (CR^Cec or CR^Dpt), similar gut epithelial cell apoptosis could be also observed (Fig. 2D). Overall, these in vivo results show that gut AMP overexpression in the presence but not in the absence of gut commensal microbes is sufficient to cause gut pathology.

• Download high-res image
• Open in new tab
• Download Powerpoint

Role of Cad in the gut commensal community structure. Because constitutive production of AMPs in the guts of Cad-RNAi flies likely affects the ecosystem of normal commensals, we next determined the dominant commensal species in the midgut of control (CR^Cont) and Cad-RNAi (CR^Cad-RNAi) flies. In wild-type flies, five commensal species dominate: Lactobacillus plantarum (LP), Lactobacillus brevis (LB), Acetobacter pomorum (AP), and two novel strains: Gluconobacter sp. strain EW707 (G707), and a bacterium in the family Acetobacteraceae, strain EW911 (A911) (21) (figs. S7 to S9 and table S1). All of these commensal bacteria, but not other bacteria such as Ecc15 and Escherichia coli, could persist in the gut, demonstrating their competences as commensal bacteria (Fig. 3A). Real-time polymerase chain reaction (PCR)–based quantitative analyses of each commensal (21) (figs. S9 to S11) clearly revealed that the commensal community structure of the Cad-RNAi flies differed from that of the control flies (Fig. 3B). Three bacteria—AP, LP, and LB—were commonly dominant in the gut of both control and Cad-RNAi flies (Fig. 3B). However, A911 [a dominant commensal member in controls, ∼1.4 × 10⁵ colony-forming units (CFUs) per gut] was markedly diminished and maintained at a very low level in the Cad-RNAi gut (∼900 CFUs per gut), whereas G707 (a minor commensal member in the control gut, ∼800 CFUs per gut) emerged as a dominant commensal in the Cad-RNAi gut (∼1.7 × 10⁴ CFUs per gut) (Fig. 3B). Time-course analyses with Cad-RNAi flies showed that loss of A911 was visible from as early as day 8, whereas dominance of G707 started at day 13 and reached the maximum level at day 18 (Fig. 3C). Given the high apoptosis levels in the gut epithelial cells of GF^Cont flies fed on the homogenates of the CR^Cad-RNAi gut, but not on homogenates of the CR^Cont gut (fig. S12), we reasoned that the dominant commensals in the Cad-RNAi gut, such as G707, may be involved in the gut pathology.

• Download high-res image
• Open in new tab
• Download Powerpoint

To validate this hypothesis, we introduced each of the isolated commensal bacteria into GF^Cont or GF^Cad-RNAi embryos and maintained these bacteria until the adult stage to generate monoassociated flies. High gut epithelial cell apoptosis was observed in all of the GF flies (GF^Cont or GF^Cad-RNAi) when the single organism was G707 (Fig. 4A), but not when the single organism was one of the other commensal organisms (LP, AP, or A911) (Fig. 4A). Additionally, G707 did not induce the apoptosis in the fat body (Fig. 4A), and G707 was pathogenic to the host when GF flies were subjected to gut infection with G707 (Fig. 4B). These results show that G707 is indeed pathogenic in a gut-specific manner when allowed to become the dominant gut microbe.

• Download high-res image
• Open in new tab
• Download Powerpoint

To further confirm that reorganization of the gut microbiota composition seen in the Cad-RNAi flies was due to the constitutive overexpression of microbicidal AMPs, we tested whether the artificial shift of gut microbiota composition could occur as a result of overexpression of AMPs. We found that overexpression of Cec and Dpt (CR^Cec+Dpt) was sufficient to induce the modification of commensal community (i.e., G707 dominance and a loss of A911), as in the case of the Cad-RNAi guts (Fig. 4C). Similar results could also be observed when we overexpressed only a single AMP (CR^Cec or CR^Dpt) (Fig. 4C). To confirm that the loss of A911 was due to the high sensitivity of this strain to AMP, we performed an in vitro antibacterial test with synthetic Cec A1 and determined the minimum inhibitory concentration (21). The result showed that A911 was highly susceptible to low concentrations of synthetic Cec A1, whereas all other commensals as well as G707 exhibited relatively high resistance to Cec A1 (Table 1). Thus, we conclude that gut AMP overexpression in Cad-RNAi flies acts as a distinct selection pressure on different commensal microbes, resulting in modification of the commensal community structure and pathogenic conditions in the gut.

Role of the normal commensal community in gut homeostasis. Because G707 is normally present at a very low level in the wild-type commensal community, we investigated whether the normal wild-type commensal community structure could antagonize the dominance of G707. To accomplish this, we introduced G707 by oral feeding into the gut of either GF^WT or CR^WT flies and then examined its persistence. The results indicated that G707 bacteria could persist in the absence of other commensal organisms in GF^WT flies but disappeared rapidly and were maintained at a low level in the presence of the normal commensal community in CR^WT flies (Fig. 5A). Consistent with these results, we observed a high-level gut epithelial cell apoptosis in the G707-challenged GF^WT flies, but not in the G707-challenged CR^WT flies (Fig. 5A), which demonstrates the role of normal commensal community structure in gut homeostasis by maintaining G707 at a low level.

• Download high-res image
• Open in new tab
• Download Powerpoint

Given the numerical inferiority of A911 in G707-dominant gut environments, we investigated whether the presence of A911 was sufficient to suppress G707 dominance in the gut. To accomplish this, we generated monoassociated flies by introducing isolated A911 into GF^WT embryos and maintaining this presence until the adult stage. When G707 was introduced into these monoassociated flies through oral feeding, the G707 bacteria rapidly decreased to a low level in the presence of A911, whereas a high initial G707 level was able to persist in the absence of A911 (in the case of GF^WT flies) (Fig. 5A). After G707 introduction, a high apoptosis index was observed in the gut of the GF^WT flies, but not in the gut of A911-monoassociated flies (Fig. 5A). This result shows that the presence of A911 is sufficient to prevent G707 dominance, and that the loss of A911 in the AMP-overexpressing genotype flies (such as Cad-RNAi flies or Cec-or Dpt-overexpressing flies) is likely to be responsible for G707 dominance.

Taken together, these results reveal that Cad acts as a critical host factor that maintains the immune homeostasis responsible for preservation of the normal commensal community structure. Failure of the balanced regulation of AMP, as in the case of Cad-RNAi flies, can act as a novel selection pressure leading to modification of the gut commensal structure. Because G707 is a highly pathogenic organism (Fig. 4, A and B), the G707-dominating host genotype acts as an initial cause of gut apoptosis, whereas the dominance of G707 acts as a direct cause of gut apoptosis.

Role of Cad in host physiology. The apoptosis of gut cells seen in Cad-RNAi flies with different GAL4 drivers was accompanied by elevated mortality (Fig. 5E and figs. S13 and S14). The high mortality of CR^Cad-RNAi flies was significantly ameliorated in the absence of commensal (GF^Cad-RNAi flies) (Fig. 5E) or in the presence of commensals under the IMD pathway mutant genetic backgrounds (CR^{Cad-RNAi + Dredd}) (fig. S14). The normal survival rate could be also observed in GF^Cad-RNAi flies stably associated with major commensal microbiota (AP, LP, LB, and A911) excluding G707, which implies the involvement of G707 in Cad-RNAi–mediated host mortality (Fig. 5E). Furthermore, the c-Jun N-terminal kinase pathway and interleukin-1β converting enzyme were shown to be involved in the microbiota-induced gut apoptosis and mortality process of the Cad-RNAi flies (fig. S15).

To demonstrate that the initial cause of death in the Cad-RNAi flies was the reduced expression of Cad, we attempted an in vivo rescue experiment. Restoration of the basal AMP level (Fig. 5B), healthy microbiota community structure (Fig. 5C), reduced apoptosis (Fig. 5D), and normal host survival levels (Fig. 5E) could be achieved in the Cad-RNAi flies by genetic reintroduction of Cad (CR^{Cad-RNAi + Cad} flies). Taken together, our results show that the intestinal homeobox Cad gene is responsible for the delicate immune homeostasis in the microbe-contacting gut tissue, which is essential for preservation of the normal microbiota community, gut homeostasis, and host survival.

Discussion. The gut epithelia of virtually all organisms have evolved to form a mutually beneficial strategic alliance with microorganisms (6). However, the role of the host factor in gut homeostasis has been largely overlooked, and little information regarding the molecular principles of gut homeostasis established by the interrelationship between the host immunity and commensal microbiota is available. In mammalian gut epithelia, deregulation of the NF-κBand AMP signaling pathways was found to be implicated in the pathogenesis of chronic IBDs (22–29). However, the enormous diversity of the resident microbiota community of the mammalian gut (e.g., 500 to 1000 different species in the human gut) (30) and the genetic complexity of the host immune system make it difficult to clearly establish the molecular links that would clarify the relations among immune genotype, commensal microbiota structure, and disease phenotype at the organism level.

By using a genetically amenable model organism harboring an extremely simple gut commensal structure, we have shown that the commensal microbiota community structure links the defective immune genotype to the gut disease phenotype. Surprisingly, Drosophila gut epithelia have evolved an immune strategy by recruiting a developmental master control gene, Cad, to maintain appropriate AMP levels for preservation of the normal flora community structure. Defective regulation of the AMP level, as seen in the case of the Cad-RNAi genotype, promotes gut pathology by exerting a selection pressure that favors the dominance of a pathogenic commensal, G707, rather than by acting as a direct cause of the disease phenotype. The emergence of a disease-causing commensal organism under an immune-defective genotype indicates the involvement of a microorganism as an origin of chronic inflammation. Further elucidation of the link between the immune genotype–based commensal community structure and host physiology may provide important insights into the causative role of pathogenic commensal microbes in a variety of chronic inflammatory diseases of the commensal-contacting epithelia.