Introduction

The co-evolution of the microbiome in animal hosts affects diverse aspects of the host physiology, including immunity, development, metabolism, and behavior1,2,3,4,5. Although many important studies demonstrate the physiological importance of gut microbiome, the lack of a genetic model of “animal-microbiome interaction” has limited the discovery of many important hidden connections between microbiome and host. The Drosophila–microbiome interaction model has proven to be the model of choice due to the advanced genetic toolbox for both major commensal bacteria and host animals6,7,8,9,10. Previously, in a Drosophila model of malnutrition, we and others found that two principal members of the Drosophila gut microbiome, Acetobacter pomorum (AP) and Lactiplantibacillus plantarum (LP), are required for enhanced host growth and development11,12. Importantly, the growth-promoting effect of these commensal bacteria is observed only under conditions of protein malnutrition11,12, indicating the existence of a link among microbiome, nutrition, and host growth. In both commensal bacteria, microbe-induced host growth is mediated by the activation of insulin and insulin-like growth factor (IGF) signaling in Drosophila11,12.

Strikingly, a specific strain of LP originally isolated from Drosophila gut (LPWJL) was found to enhance the juvenile growth not only in a Drosophila model but also in a gnotobiotic mouse model13,14. LPWJL was shown to increase IGF-1 levels, partially via NOD2-dependent bacterial cell wall sensing, indicating evolutionarily conserved interactions between LPWJL and host somatotrophic signaling pathway (e.g., IGF signaling)13,14. However, to date, some long-standing unanswered questions in the field are: “Which bacterial metabolic pathways, more specifically which bacterial metabolites, are capable of inducing host systemic growth?”; “How does host sense (i.e., which organs and which receptors) these microbiome-derived signals?”; and “How does host coordinate microbiome-to-organ-to-organ interactions to achieve systemic growth?”. Research in this direction is of paramount importance in unraveling an unexplained association between the microbiome and the development of a malnourished host.

The present study aimed to identify microbiome-derived growth factors capable of promoting host growth, and understand the molecular mechanism by which these factors impact host physiology. We found that LPWJL emits air-diffusible volatile somatotrophic factors (VSFs) that activate the airway-gut-brain axis in an olfaction-independent manner for host growth under malnutrition. Further, by identifying the chemical nature of VSFs and their cellular receptors, we found that the VSF-activated airway-gut-brain axis involves complex inter-organ signaling interactions including Hippo signaling, FGF-FGFR signaling, peptide hormone signaling, and IGF signaling pathways. Our discovery of airborne interactions involving the microbiome and different organs may provide essential clues to understanding the role of the microbiome in metazoan development during prokaryote-eukaryote co-evolution.

Results

Microbiome-emitted scents promote host growth in an olfaction-independent manner

We initially investigated the microbiome-derived growth-stimulating factors using LPWJL, a well-known Drosophila–microbiome member capable of inducing host growth in a malnutrition-induced stunting model of Drosophila and rodent11,12,13,14. By analyzing the effects of different microbial factors on host growth using a germ-free Drosophila model, we made a serendipitous observation in which LPWJL induced the growth of germ-free animals even without gut colonization (i.e., inhalation of LPWJL volatiles in the absence of physical contact between microbes and host cells), albeit less effective than LPWJL association (Fig. 1a, b). This surprising observation prompted us to hypothesize that LPWJL-emitted scents contain VSFs and that airborne transmission of VSFs promotes host growth.

Fig. 1: Microbiome-emitted scents promote host growth in an olfaction-independent manner.
figure 1

In all cases, germ-free (GF) embryos were treated either without (−) or with (+) inhalation of LPWJL volatiles during entire larval stages. a A schematic diagram of the inhalation experiment involving the treatment with LPWJL volatiles. b Body length of control GF, GF animals with inhalation of LPWJL volatiles (GF + Inhalation), and GF animals monoassociated with LPWJL (GF + monoassociation). Larval sizes were measured at various hours after egg laying (AEL). n = 3 biologically independent experiments. c Relative feeding amounts of control GF (GF) and GF animals with inhalation of LPWJL volatiles (GF + Inhalation). Feeding amounts of control GF animals at 24 h AEL were taken arbitrarily as 1. n = 3 biologically independent experiments. d Body length at 144 h AEL without or with inhalation of LPWJL volatiles. Control, ORCO+/−, and ORCO−/− animals were used. n = 3 biologically independent experiments. e Real-time qPCR analysis of different ORs. Target gene expression in the GF animals without inhalation of LPWJL volatiles was taken arbitrarily as 1. n = 4 biologically independent experiments. f OR42b promoter activity in the anterior midgut (indicated by filled arrowhead) of OR42b-GAL4 > UAS-GFP flies. Representative confocal images of the whole gut (left) and the fluorescence signals in the enterocytes of the anterior midgut (right) are shown. The gut is arranged from anterior to posterior, moving from top to bottom in the photo (right panel). Nuclei were stained with DAPI in blue. Scale bar, 500 μm (left) and 50 μm (right). Data are mean ± s.e.m. P values (*p < 0.05, **p < 0.01, ***p < 0.001, ns not significant) are indicated: one-way ANOVA with Tukey’s post hoc test in (b); two-way ANOVA with Tukey’s multiple comparison test in (d); unpaired two-tailed t-test in (c, e). Genotypes, sample sizes, and statistical analyses are shown in Supplementary Data 1. Source data are provided in the Source Data file.

To ascertain whether the growth gain induced by VSF is not solely attributable to VSF’s stimulation of larval feeding behavior, we compared food intake between VSF-exposed and non-exposed larvae. Given that larvae exposed to VSF for longer than 72 h after egg laying (AEL) are larger in size (e.g., larvae at 96 h AEL) than the non-exposed control larvae, thus exhibiting higher food intake (Fig. 1c), we measured food intake when VSF-exposed and non-exposed larvae were of similar sizes (i.e., at 24, 48, and 72 h AEL). The results indicate that the inhalation of LPWJL volatile does not affect larval feeding behavior, at least during periods when VSF-exposed and non-exposed larvae are of similar size (Fig. 1c).

Recently, it was reported that microbiome-emitted scents influence social behavior, such as mate choice or nestmate recognition15,16,17,18. Because different scent molecules are readily sensed by olfactory receptors (ORs) in OR-neurons (ORNs)19,20,21, we investigated whether ORNs mediated LPWJL-induced host growth. Drosophila larvae have 21 ORNs, each expressing a unique OR complex composed of one to two odor-specific ORs and a single universal OR co-receptor (ORCO)19,22. When well-known smell-blind ORCO−/− animals23,24 were subjected to inhalation of LPWJL volatiles, we surprisingly found that the LPWJL-emitted scents still supported their growth (Fig. 1d). The growth rates of ORCO−/− animals were similar to those observed in wild-type or heterozygous ORCO+/− animals (Fig. 1d). These results demonstrate that host growth induced by LPWJL-emitted scents is independent of olfaction, i.e., independent of scent sensing by ORNs.

OR42b is expressed in the gut and its expression is induced by LPWJL-emitted scents

Although it has been believed that the expression of ORs is restricted to the olfactory organ, recent data have revealed that ORs are found outside the olfactory organ in both Drosophila and mammals25,26. These findings suggest that ectopic ORs have previously unanticipated roles in non-olfactory organs, including the intestine. Based on this, we hypothesized that the microbiome emits VSFs and that enterocytes sense these VSFs via gut-expressing OR. Since stage-specific and tissue-specific expression data suggest that at least twelve ORs are expressed in the larval gut24,27,28,29,30, we assumed that one of these ORs may be the candidate receptor for VSFs in the gut.

To test this hypothesis, we monitored the expression of these ORs in the absence or presence of LPWJL volatiles (Fig. 1e). The result showed that OR42b was greatly upregulated following exposure to LPWJL volatiles (Fig. 1e). Analysis of the intestinal expression of OR42b using reporter flies carrying the OR42b promoter fused to Gal4 (OR42b-Gal4) revealed that OR42b was mainly detected in the enterocytes of the anterior midgut region following inhalation of LPWJL volatiles (Fig. 1f). Additionally, OR42b expression was detected in the boundary region between the proventriculus and midgut, as well as subsets of enteroendocrine cells (EECs) located in the posterior midgut region (Supplementary Fig. 1). Taken together, we conclude that OR42b is expressed in non-olfactory cells, such as enterocytes and EECs, in addition to its known expression in the olfactory organ. Furthermore, intestinal OR42b is induced in response to LPWJL-emitted scents.

Enterocyte OR42b functions as a VSF receptor for growth in an ORCO-independent manner

It is well-known that OR42b is one of the most conserved ORs in many Drosophila species31,32, which suggests that it may play one of the most fundamental functions among ORs. Well-known ligands for OR42b are various chemicals associated with natural Drosophila food sources (i.e., microbe-fermented fruits), such as aliphatic esters (e.g., ethyl acetate, ethyl butyrate, and propyl acetate) and 3-hydroxy-2-butanone-associated metabolites (i.e., 3-hydroxy-2-butanone, diacetyl, and 2,3-butanediol)33,34. Many Drosophila species display attractive behavior toward these chemicals33,34,35, suggesting a conserved role of OR42b in ORN during foraging behavior across the Drosophilidae. Therefore, we initially thought that some of these attractants act as VSFs. We found that inhalation of one of the aliphatic esters or 3-hydroxy-2-butanone-associated metabolites was sufficient to induce larval growth (Fig. 2a), indicating that known OR42b ligands likely act as VSFs.

Fig. 2: OR42b mediates LPWJL volatile-induced host growth in an ORCO-independent manner.
figure 2

a GF embryos were treated either without (+none) or with inhalation of 0.03% (v/v) volatile chemicals (+chemical) during entire larval stages. Larval sizes were measured at 144 h AEL. n = 3 biologically independent experiments. b, c Tetrameric assembly of OR42b. OR42b construct design for expression (b, left top) and SDS-PAGE of purified OR42b eluted from affinity chromatography (b, left bottom). GFP-OR42b fusion protein is indicated with white arrowheads. Fluorescence size-exclusion chromatography screening under LMNG and DDM detergent conditions (b, right). Representative micrograph of OR42b based on negative-stained electron microscopy (c). Notable particles are indicated by white boxes, and enlarged (c, top). 2D class averages are shown at the bottom right (c, bottom). d AlphaFold-predicted model of tetrameric OR42b is presented as side and top views. The putative odorant-binding pocket formed by S2, S3, S4, and S6 helices has a conical shape. Conserved residues in the binding pocket are highlighted as gray spheres and Y156, S158, and S159 are shown in magenta. e, f GF embryos were treated either without (−) or with (+) inhalation of LPWJL volatiles during entire larval stages. Larval sizes were measured at 144 h AEL. GF control animals (Control) and GF animals with gut-specific OR42bDN (Mex1-Gal4 > UAS-OR42bDN) were used in (e); GF Control animals (Control), GF ORCO mutant animals (ORCO−/−) and GF OR42bDN knock-in animals (OR42bDN-KI) were used in (f). n = 3 biologically independent experiments in (e, f). Data are mean ± s.e.m. P values (***p < 0.001, ns not significant) are indicated: one-way ANOVA with Tukey’s post hoc test in (a); unpaired two-tailed t-test in (e, f). Genotypes, sample sizes, and statistical analyses are shown in Supplementary Data 1. Source data are provided in the Source Data file.

Next, we investigated the mechanism by which OR42b acts as a bona fide receptor for VSFs in an ORCO-independent manner. In insects, volatile compounds in scents are normally sensed by a specific OR complexed with a common ORCO, possibly by forming a heterotetramer composed of two OR subunits and two ORCO subunits36. As OR42b seems to act independently of ORCO (Fig. 1d), we first analyzed whether OR42b forms a functional tetramer in the absence of ORCO. Fluorescence size-exclusion chromatography revealed that GFP-tagged OR42b protein forms a complex corresponding to the size of the tetramer (Fig. 2b). Transmission electron microscopy further validated the formation of homo-tetrameric complexes of OR42b (Fig. 2c).

Utilizing the structural insight into ligand recognition in insect ORs37, we could predict the presence of a conical (or funnel-like) cavity for ligand access, created by S2, S3, S4, and S6 helices (Fig. 2d). In this model, we could identify Tyr156, Ser158, and Ser159 of OR42b as potentially critical residues for ligand binding (Fig. 2d). Accordingly, we generated a UAS fly line carrying a ligand-binding mutant allele of OR42b that is deficient in ligand binding (in which three critical amino acids mutated to Leu156, Ala158, and Ala159 residues). Overexpression of this ligand-binding mutant allele by using enterocyte-specific Gal4 driver was sufficient to abolish VSF-induced larval growth, showing that a ligand-binding mutant allele of OR42b acts as a dominant-negative (DN) mutant allele of OR42b (OR42bDN) in the gut (Fig. 2e). These results indicate that critical residues for ligand binding within OR42b are required for VSF-induced growth.

To further validate the role of OR42b in growth and olfaction, we generated an OR42bDN knock-in animal. We observed that VSF-induced larval growth was abolished only in OR42bDN knock-in animals, but not in ORCO−/− animals (Fig. 2f). These results indicate that VSF sensing in the gut for growth is likely OR42b-dependent but ORCO-independent.

LPWJL volatiles induce Btl signaling activation for gut-associated tracheal branching formation

During the dissection of the gut of OR42b-Gal4 > UAS-GFP flies, we routinely observed that OR42b-expressing enterocytes in the anterior midgut are in contact with gut-associated trachea that also express OR42b (Fig. 3a). The tracheal airway, composed of a branched tubular epithelial network throughout the body, acts as a respiratory organ for the passage of gas molecules, such as oxygen, to reach internal tissues38,39. Tracheal epithelial cells express Breathless (Btl, a homolog of the mammalian FGF receptor) and Btl signaling activation induces tracheal branching formation40,41. Btl signaling activation also induces Btl expression, thus forming a positive feedback loop40,42, so we investigated tracheal Btl signaling activation status in the OR42b-expressing gut-associated trachea by examining Btl expression patterns. The results showed that LPWJL volatiles strongly activated tracheal Btl signaling, as evidenced by the upregulation of Btl expression, and subsequent tracheal branching formation (Fig. 3b). In the absence of LPWJL volatiles, we observed only basal Btl expression and reduced branching formation (Fig. 3b).

Fig. 3: LPWJL volatiles Promote tracheal branching formation, gut oxygenation, and gut growth.
figure 3

Germ-free embryos were exposed to LPWJL volatiles throughout larval stages (+) or left untreated (−). Third-instar GF larvae were used. a OR42b expression in the enterocytes (open arrowhead) and gut-associated trachea (filled arrowhead) was observed. b LPWJL volatiles enhanced tracheal Btl signaling activation (left: representative image; right: tracheal coverage normalized to untreated larvae, set to 1). n = 3 biologically independent experiments. c Btl signaling activation is dependent on exposure to LPWJL volatiles. Larvae carrying Btl-Gal4 > UAS-GFP were used. To exclude the effect of larval size on tracheal Btl signaling activation, volatile-exposed larvae and non-exposed larvae of similar size were used (left). Relative tracheal coverage of these animals is shown (right). The tracheal coverage of untreated animals was taken arbitrarily as 1. n = 3 biologically independent experiments. LPWJL volatiles induced enterocyte Bnl expression (d) and tracheal Btl expression (e). f Gut-specific Bnl knockdown reduced tracheal branching formation. Relative tracheal coverage was shown and the tracheal coverage of untreated control animals was taken arbitrarily as 1. n = 4 biologically independent experiments. g LPWJL volatiles increased gut oxygenation, measured using Timer, a fluorescence-based oxygen reporter (left: merged oxidized red/unoxidized green Timer image; right: oxidized red/unoxidized green ratio). n = 4 biologically independent experiments. h The gut lengths, from proventriculus to hindgut, of untreated larvae at 144 h AEL were taken arbitrarily as 1. n = 3 biologically independent experiments. i Gut-specific Bnl knockdown abolishes VSF-induced GF larval growth at 144 h AEL. n = 3 biologically independent experiments. Data are mean ± s.e.m., with P values (*p < 0.05, ***p < 0.001, ns not significant) from unpaired two-tailed t-test (b, c, g, h) or two-way ANOVA with Tukey’s multiple comparison test (f, i). The gut is arranged from anterior to posterior, moving from top to bottom in the photo (a, b, d, e, g). Nuclei were stained with DAPI (blue). Scale bar, 50 μm. Genotypes, sample sizes, and statistical analyses are shown in Supplementary Data 1. Source data are provided in the Source Data file.

As LPWJL volatiles promote larval growth, the VSF-exposed larvae are bigger than the non-exposed larvae at the same age (Fig. 1b). To confirm that enhanced tracheal Btl expression in the VSF-exposed larvae is not solely due to their larger size, we compared VSF-exposed larvae and non-exposed larvae of similar size (i.e., VSF-exposed larvae at 140 h AEL versus non-exposed larvae at 164 h AEL). The results showed that VSF-exposed larvae at 140 h AEL still exhibited higher tracheal branching formation than non-exposed larvae at 164 h AEL (Fig. 3c). These findings suggest that Btl activation for tracheal formation is indeed attributable to the exposure to LPWJL volatiles, rather than differences in larval size or developmental stage.

LPWJL volatiles induce airway-gut axis activation via Bnl/Btl signaling for oxygen-mediated regulation of growth

We next examined the molecular mechanism by which LPWJL volatiles induce tracheal Btl signaling activation. Branchless (Bnl, a homolog of mammalian FGF) acts as a known ligand for Btl for its downstream signaling activation43. When we examined Bnl expression in the gut in the absence of LPWJL volatiles, we found strong basal expression of Bnl in the enterocytes of the copper region (Bnlcopper) and weak basal expression in the posterior midgut region (Bnlposterior) (Supplementary Fig. 2). Almost no Bnl expression was observed in the anterior midgut region (Bnlanterior) (Supplementary Fig. 2). However, following LPWJL inhalation, we observed a significant induction of Bnlanterior expression, while the intensity of Bnlcopper and Bnlposterior expression remained unchanged (Fig. 3d and Supplementary Fig. 2). The expression of Bnlanterior in this region is not uniform across all cells; weak or no expression is observed in some cell populations (Fig. 3d and Supplementary Fig. 2). These results indicate that Bnlcopper and Bnlposterior expression are constitutive, while Bnlanterior expression is inducible in the same anterior midgut region where OR42b is inducible. Double labeling of Bnl and Btl expression confirmed that both Bnl expression in the enterocytes and Btl expression in the tracheal cells were enhanced following the exposure to LPWJL volatiles (Fig. 3e). Downregulation of Bnl expression in the enterocytes decreased both basal and LPWJL-induced tracheal branching formation (Fig. 3f), indicating that LPWJL-induced Bnl in enterocytes is required for tracheal branching formation. Furthermore, we found that gut oxygen concentration and gut size increased following the exposure to LPWJL volatiles (Fig. 3g, h). Furthermore, we found that downregulation of Bnl expression in the enterocytes abolished the growth-promoting effect of LPWJL volatiles (Fig. 3i). Taken together, these results indicate that LPWJL volatiles induce airway-gut axis activation via Bnl/Btl signaling for oxygen regulation in organs, which facilitates body growth.

LPWJL volatiles promote DILP2 secretion from the brain for systemic growth control

Previous studies have demonstrated a close relationship between tracheal branching and insulin/IGF signaling pathways44,45,46. This insulin/IGF regulates host homeostatic programs controlling developmental rate as well as organ and body size11,12,47,48,49. To determine whether insulin/IGF mediated LPWJL volatile-dependent growth promotion, we first measured the level of circulating insulin/IGF in the insect blood-like hemolymph. Three insulin/IGFs [referred to as Drosophila insulin-like peptides (DILPs) such as DILP2, DILP3, and DILP5] are the major DILPs secreted from insulin/IGF-producing cells (IPCs) of the brain49,50. Although the measurement of the circulating levels of growth hormone is indispensable to understand the physiological context, methods to measure endogenous insulin/IGFs levels in the hemolymph have yet to be established in Drosophila. To measure the circulating levels of each DILP, we generated double epitope-tagged DILP2, DILP3, or DILP5 knock-in animals (Fig. 4a). Enzyme-linked immunosorbent assay showed that DILP2 is the major form of circulating DILP in larvae, representing ~90% of the IPC-secreted DILPs (Fig. 4a). Further, we found that the circulating level of DILP2, but not DILP3 or DILP5, was significantly augmented following the exposure to LPWJL volatiles (Fig. 4a). The IPC-specific DILP2 knockdown was sufficient to abolish LPWJL volatile-induced animal growth (Fig. 4b). Taken together, these results indicate that LPWJL volatiles promote the release of DILP2 from IPC to hemolymph for systemic growth control.

Fig. 4: LPWJL volatiles induce gut-brain somatotrophic axis for systemic growth.
figure 4

Germ-free embryos were treated without (−) or with (+) inhalation of LPWJL volatiles during larval stages. Larvae at 144 h AEL were analyzed. a Schematic of HA/FLAG epitope-tagged DILP2, 3, or 5 knock-in strategy (top) and hemolymph DILP level (bottom) in GF knock-in animals. DILP2 levels of DILP2 knock-in animals without LPWJL volatiles were set as 1. n = 3 biologically independent experiments. b Body length of GF control and GF DILP2-RNAi larvae. n = 3 biologically independent experiments. c Real-time qPCR analysis of gut hormone expression in the GF animals. Control values without LPWJL volatiles were set as 1. n = 3 biologically independent experiments. d AstA+ EEC numbers in GF control and GF CCHA2−/− animals. n = 3 biologically independent experiments. e CCHA2+ EEC numbers in GF control and GF AstA−/− animals. n = 3 biologically independent experiments. f Colocalization of AstA+ EEC with OR42b. Colocalization of AstA receptor 2 (AstAR2) with CCHA2+ EECs (g); and CCHA2+ EEC numbers in GF control and GF AstAR2-RNAi animals (h left, representative image; h right, CCHA2+ EEC number/gut). n = 3 biologically independent experiments. i Relative levels of hemolymph DILP2 in GF control, GF CCHA2-RNAi, and GF AstA-RNAi. DILP2 levels of GF control (DILP2 knock-in) animals without LPWJL volatiles were set as 1. n = 3 biologically independent experiments. j Body length of GF control, GF AstA−/−, and GF CCHA2−/− animals. n = 3 biologically independent experiments. Gut samples were taken from a 0.1 mm2 region to count AstA+ or CCHA2+ cells (d,e, and h). Data are mean ± s.e.m., with P values (*p < 0.05, **p < 0.01, ***p < 0.001, ns not significant); unpaired two-tailed t-test in (c); two-way ANOVA with Tukey’s multiple comparison test in (a, b, d, e, hj). The gut is arranged from anterior to posterior, moving from top to bottom in the photo (dh). Nuclei were stained with DAPI (blue). Scale bar, 50 μm. Genotypes, sample sizes, and statistical analyses are shown in Supplementary Data 1. Source data are provided in the Source Data file.

LPWJL volatiles induce enteroendocrine AstA and CCHA2 peptide hormone expression for brain DILP2 secretion

The molecular mechanism by which LPWJL volatiles can perform long-distance communications with the brain IPC was the focus of our next investigation. Based on our previous observation that gut cells sense the external environment and communicate with the brain via gut hormones5, we hypothesized that enteric hormones that directly respond to LPWJL volatiles may be involved in communication with IPC in the brain. To test this hypothesis, we monitored the expression of gut hormones altered by LPWJL volatiles (Fig. 4c). We found that two gut peptide hormones, allatostatin-A (AstA) and CCHAmide-2 (CCHA2), were significantly upregulated in a subset of EECs by LPWJL volatiles (Fig. 4c–e). We next analyzed the epistatic relationship between AstA and CCHA2 by using AstA−/− and CCHA2−/− animals. We found that LPWJL volatile-induced AstA+ EECs were observed in CCHA2−/− animals (Fig. 4d), whereas LPWJL volatile-induced CCHA2+ EECs were abolished in AstA−/− animals (Fig. 4e and Supplementary Fig. 3). In this condition, we found that the number of EECs does not increase significantly following inhalation of LPWJL (Supplementary Fig. 4). These data indicate that LPWJL volatiles do not affect the total number of EECs, but they do increase the proportion of EECs that secrete AstA or CCHA2. These data also indicate that AstA is required for the inducibility of CCHA2 following the exposure to LPWJL volatiles. Importantly, OR42b is expressed specifically in AstA+ EECs (Fig. 4f). Furthermore, we found that AstA receptor-2 (AstAR2) was colocalized with CCHA2 (Fig. 4g) and that AstAR2 was required for VSF-induced CCHA2 expression (Fig. 4h and Supplementary Fig. 3). Taken together, we concluded that LPWJL volatiles induce AstA expression in OR42b-expressing EECs that in turn induces CCHA2 expression in AstAR2-expressing EECs.

AstA and CCHA2 are found to be expressed in diverse cells, such as fat body cells and EECs, and are known to affect various aspects of host physiology, at least partially by modulating IPCs51,52,53. Therefore, we determine whether LPWJL volatile-induced AstA and CCHA2 mediated DILP2 secretion. For this, we first analyzed the circulating DILP2 levels following EEC-specific knockdown of AstA or CCHA2 expression by using Prospero-Gal4. The results showed that the knockdown of intestinal AstA or CCHA2 abrogated LPWJL volatile-induced DILP2 secretion (Fig. 4i). Consistent with these hormone knockdown data, we found that mutation of one of these hormones was sufficient to abolish LPWJL volatile-induced body growth (Fig. 4j), indicating that both hormones were required for body growth. Taken together, these data indicate that LPWJL volatile-induced intestinal AstA and CCHA2 hormones are required for DILP2 secretion from brain IPC, thereby forming the somatotrophic gut-brain axis.

LPWJL volatiles inactivate the Hippo pathway leading to Yorkie nuclear localization

We investigated the intracellular signaling pathway by which OR42b mediates VSF-responsive gene expression. Because Yorkie, a transcriptional coactivator of the Hippo pathway, is shown to be required for Bnl expression54, we first analyzed whether the Hippo pathway acted downstream of OR42b for the expression of VSF-responsive genes. In the absence of growth signaling during which the Hippo pathway is constitutively activated, Merlin, in association with Kibra and Expanded, forms a membrane-associated complex for sequential activation of two downstream kinases, Hippo and Warts55,56. Activated Warts inhibit Yorkie nuclear localization by phosphorylating Yorkie57. To assess the subcellular localization of endogenously expressed OR42b, we generated OR42b-FLAG knock-in animals containing C-terminal FLAG epitope-tagged OR42b. By combining OR42b-FLAG knock-in flies with Merlin-YFP flies, we assessed the subcellular localization of Merlin-YFP and OR42b-FLAG before and after the exposure to LPWJL volatiles. The result showed that Merlin and OR42b were colocalized in the membrane under basal conditions (Fig. 5a). Importantly, we found that both proteins rapidly dispersed from the membrane following the exposure to LPWJL volatiles (Fig. 5a). As membrane localization of the Merlin complex is essential for downstream Hippo pathway activation55, we speculated that VSF-induced dispersal of the membrane-localized Merlin complex inactivated the Hippo pathway, leading to nuclear localization of Yorkie. To test this possibility, Yorkie localization was analyzed before and after the exposure to LPWJL volatiles. The results showed that cytoplasmic Yorkie rapidly translocated into the nucleus following the exposure to LPWJL volatiles (Fig. 5b), indicating that VSFs, acting as growth factors, inactivated the Hippo pathway leading to Yorkie nuclear localization.

Fig. 5: LPWJL volatiles modulate the Hippo-Merlin-Yorkie pathway via non-olfactory OR42b.
figure 5

In all cases, germ-free embryos were treated either without (−) or with (+) inhalation of LPWJL volatiles during entire larval stages. Larvae at 144 h AEL were used. a Membrane colocalization of Merlin and OR42b in the absence of LPWJL volatiles, and dispersal of membrane Merlin-OR42b following the exposure to LPWJL volatiles. Enterocytes from OR42b-FLAG knock-in animals carrying Merlin-YFP were stained. Phalloidin staining was performed to visualize the brush border membrane of enterocytes. b Nuclear Yorkie translocation of enterocytes following the exposure to LPWJL volatiles. Staining of Dlg, a septate junction marker, and DAPI were included to facilitate visualization of the cytoplasmic and nuclear regions of individual cells. c LPWJL volatile-independent constitutive membrane localization of OR42bDN. Enterocytes from OR42bDN-FLAG knock-in animals were stained. Membrane localization of Merlin (d) and nuclear localization of Yorkie (e) in enterocytes from GF Da-Gal4 control, GF Da-Gal4 > UAS-OR42bDN, and GF Da-Gal4 > UAS-OR42b-RNAi animals. Nuclear-to-cytoplasmic ratio of Yorkie was shown (e). n = 3 biologically independent experiments. f A model for Hippo/Yorkie pathway in OR42b wildtype, dominant-negative, and RNAi/knockout. Data are mean ± s.e.m. P values (***p < 0.001, ns not significant) are indicated; two-way ANOVA with Tukey’s multiple comparison test in (e). The gut is arranged from anterior to posterior, moving from top to bottom in the photo (ae). Nuclei were stained with DAPI in blue. Scale bar, 50 μm. Genotypes, sample sizes, and statistical analyses are shown in Supplementary Data 1. Source data are provided in the Source Data file.

OR42b is required for the regulation of the Hippo pathway

In contrast to VSF-induced dispersal of membrane-localized OR42b (Fig. 5a), LPWJL volatiles do not affect membrane localization of OR42bDN, showing constitutive membrane localization (Fig. 5c). These data indicate that VSF binding to OR42b likely initiated the dispersal of membrane OR42b. Consistently, OR42bDN overexpression in the gut cells led to a constitutive membrane localization of the Merlin complex (Fig. 5d), which prevented Yorkie nuclear translocation (Fig. 5e). These data suggest that the Merlin complex was tethered to the constitutively membrane-localized OR42bDN to activate the Hippo pathway (see also Fig. 5f). To further investigate whether the physical presence of OR42b on the membrane was indeed required for the maintenance of membrane localization of the Merlin complex, the Merlin localization was analyzed under OR42b-knockdown condition. In these OR42b-knockdown animals, we found that membrane localization of the Merlin complex was abolished even in the absence of LPWJL volatiles (Fig. 5d), showing a constitutive nuclear localization of Yorkie (Fig. 5e). These results indicate that the membrane localization status of OR42b controls Merlin complex localization and corresponding Hippo pathway activity (see also Fig. 5f). OR42b-knockdown animals having dispersed Merlin complex showed constitutive Hippo pathway inactivation (i.e., nuclear localization of Yorkie) whereas OR42bDN-overexpressing animals having constitutive membrane localization of Merlin complex showed constitutive Hippo pathway activation (i.e., cytoplasmic localization of Yorkie) (Fig. 5f).

OR42b-modulated Hippo pathway regulates VSF-responsive gene expression

We next analyzed whether the Hippo pathway-modulated Yorkie controls VSFs-responsive gene expression. We found that LPWJL volatiles failed to induce the expression of all VSF-responsive genes (i.e., Btl, Bnl, and AstA) in Yorkie-knockdown animals (Supplementary Fig. 5). In contrast, all VSF-responsive genes displayed high levels of expression even in the absence of LPWJL volatiles in animals expressing the active form of Yorkie (Supplementary Fig. 5). These results indicate that Hippo pathway activity and subsequent Yorkie activity are crucial for VSF-induced target gene expression following the exposure to LPWJL volatiles.

To further validate the role of OR42b in Hippo/Yorkie pathway regulation, we generated OR42b-knockout (OR42b−/−) animals carrying a null mutant allele of OR42b that lacked all protein-coding sequences. In these OR42b−/− animals, we found LPWJL volatile-independent constitutive activation of Btl and AstA in trachea and EECs, respectively (Fig. 6a, b). Consistent with these data, EEC-specific OR42b-RNAi is sufficient to increase the number of AstA+ EECs (Fig. 6b). The constitutive activation of Btl and AstA observed in OR42b−/− animals indicates constitutive activation of Yorkie. Accordingly, OR42b−/− animals showed enhanced body growth even in the absence of LPWJL volatiles (Fig. 6c), indicating that the absence of OR42b is responsible for Hippo pathway inactivation (i.e., Yorkie activation). The high AstA expression and growth promotion observed in OR42b−/− animals were reduced to the level of control flies when OR42b expression was reintroduced in OR42b-expressing tissues. (i.e., OR42b-Gal4 > UAS-OR42b in OR42b−/− animals) (Fig. 6d–f). Furthermore, the growth promotion observed in OR42b−/− animals was abolished in double mutant animals lacking both OR42b and AstA (Fig. 6g). These results indicate that the constitutive AstA expression and enhanced organ/body growth in the absence of LPWJL volatiles are indeed due to the absence of OR42b, and that AstA is required for the growth promotion observed in OR42b−/− animals. In contrast to OR42b−/− animals, LPWJL volatile-induced Btl, Bnl, and AstA expression (Supplementary Fig. 6a, b), and subsequent body growth were completely abolished in OR42bDN knock-in animals (Fig. 2f), indicating that signal-dependent Yorkie activation was abolished. Importantly, inactivation of OR42b in one of three VSF-responsive organs or cells (by overexpressing OR42bDN in the trachea, enterocytes, or AstA+ EECs) was sufficient to abolish LPWJL volatile-induced host growth (Supplementary Fig. 6c), indicating that OR42b activation in both airway and gut is required for animal growth. Taken together, we could conclude that LPWJL-emitted VSFs induce OR42b dispersal that in turn induces Hippo pathway inactivation leading to Yorkie activation for growth-promoting gene expression.

Fig. 6: OR42b−/− animals showed constitutive Yorkie activation and enhanced body growth even in the absence of LPWJL volatiles.
figure 6

In all cases, germ-free embryos were treated either without (−) or with (+) inhalation of LPWJL volatiles during entire larval stages. Larvae at 144 h AEL were used. a Btl expression (left) and relative tracheal coverage (right) in GF control and GF OR42b−/− animals. The tracheal coverage of untreated GF control animals was taken arbitrarily as 1. The gut is arranged from anterior to posterior, moving from top to bottom in the photo. n = 3 biologically independent experiments. b Quantitative analysis of AstA+ EEC number per gut in GF control and GF OR42b−/− animals (left) or GF OR42b-RNAi animals (right). n = 3 biologically independent experiments. c Body length in GF control and GF OR42b−/− animals. n = 3 biologically independent experiments. Genetic rescue experiment by expressing OR42b in the OR42b knock-out animal. AstA+ EEC number per gut (d), and growth promotion of gut (e) and body (f) were measured. n = 3 biologically independent experiments in (df). Gut sizes of control GF animals were taken arbitrarily as 1 (g). n = 3 biologically independent experiments. Body length in GF control, GF OR42b−/−, GF AstA−/−, and GF OR42b−/−; AstA−/− animals. In each gut sample, an area (0.1 mm2) of a microscopic image from a similar gut region was randomly selected to count the number of AstA+ cells (b, d). Data are mean ± s.e.m. P values (*p < 0.05, **p < 0.01, ***p < 0.001, ns not significant) are indicated; one-way ANOVA with Tukey’s post hoc test in (df); two-way ANOVA with Tukey’s multiple comparison test in (ac, g). Nuclei were stained with DAPI in blue. Scale bar, 50 μm. Genotypes, sample sizes, and statistical analyses are shown in Supplementary Data 1. Source data are provided in the Source Data file.

LPWJL-emitted (2R,3R)-2,3-butanediol is responsible for the activation of airway-gut-brain axis for animal growth

We next analyzed whether different OR42b ligands (i.e., aliphatic esters or 3-hydroxy-2-butanone-associated metabolites) capable of inducing larval growth (Fig. 2a) were indeed released from the LPWJL. Metabolic pathway analysis revealed that pyruvate was converted into 3-hydroxy-2-butanone which was subsequently converted into diacetyl and three isomers of 2,3-butanediol (Fig. 7a). Among three 2,3-butanediol stereoisomers, i.e., the R form (2R,3R-butanediol), S form (2S,3S-butanediol), and meso form (2R,3S-butanediol), only (2R,3R)-stereoisomer of 2,3-butanediol induced tracheal cell activation and subsequent host growth promotion (Supplementary Fig. 7a, b). The treatment of (2R,3R)-2,3-butanediol volatiles induced OR42b dispersal in the enterocyte in both wildtype and ORCO mutant backgrounds (Supplementary Fig. 7c), indicating that OR42b can sense (2R,3R)-2,3-butanediol in an ORCO-independent manner. As aliphatic ester production is largely dependent on the presence of precursors (e.g., both ethanol and aliphatic acid) in culture media58, our precursor-free culture of LPWJL likely favorized the production of 3-hydroxy-2-butanone and its associated metabolites, rather than aliphatic ester production. Gas chromatography–mass spectrometry (GC-MS) analysis of LPWJL-emitted gas revealed that LPWJL indeed emitted 3-hydroxy-2-butanone, diacetyl, and (2R,3R)-2,3-butanediol (Fig. 7b). Quantitative analyses showed that (2R,3R)-2,3-butanediol was the major VSF emitted from LPWJL (Fig. 7b). Consistently, we observed that treatment with (2R,3R)-2,3-butanediol was sufficient to trigger calcium mobilization in the OR42b-expressing cells such as tracheal cells and AstA+ EECs (Supplementary Movies 1 and 2). In contrast, the S form or meso form of 2,3-butanediol failed to induce calcium mobilization in the trachea (Supplementary Movies 3 and 4), indicating that the calcium response is specific to (2R,3R)-2,3-butanediol. Computational analysis of (2R,3R)-2,3-butanediol binding to OR42b further revealed two potential binding pockets (Supplementary Fig. 8).

Fig. 7: LPWJL-emitted (2R,3R)-2,3-butanediol is responsible for the activation of the airway-gut-brain axis during animal growth.
figure 7

a A biosynthetic pathway of 2,3-butanediol production from pyruvate. b GC-MS for quantitative analysis of different metabolites emitted by LPWJL, LPWJLΔALS, and LPWJLΔALS_ALS strains. n = 3 biologically independent experiments. Representative image of Btl and Bnl expression in trachea and enterocytes, respectively (c, left), relative tracheal coverage (c, right), representative merged image of oxidized red and unoxidized green Timer in enterocytes (d, left), ratio of oxidized red to unoxidized green Timer (d, right), representative image of AstA expression in EECs (e, left), quantitative analysis of AstA+ EEC number (e, right), quantitative analysis of circulating DILP2 level (f), and body length (g) were analyzed. n = 3 biologically independent experiments in (cg). Relative values of GF control animals without inhalation of LPWJL volatiles were taken arbitrarily as 1 (c, f). In all cases, GF embryos were treated either without (+none) or with inhalation of volatiles from LPWJL (+LPWJL) bacteria, LPWJL ΔALS (+LPWJL ΔALS) bacteria, or LPWJLΔALS_ALS (+LPWJLΔALS_ALS) bacteria during entire larval stages. In the case of co-stimulation, GF embryos were treated with inhalation of volatiles from both LPWJL ΔALS bacteria and (2R,3R)-2,3-butanediol (0.3% v/v) (+LPWJL ΔALS+2,3-butanediol). Larvae at 144 h AEL were used. In each gut sample, an area (0.1 mm2) of a microscopic image from a similar gut region was randomly selected to count the number of AstA+ cells (e). The gut is arranged from anterior to posterior, moving from top to bottom in the photo (ce). Data are mean ± s.e.m. P values (**p < 0.01, ***p < 0.001, ns not significant) are indicated; one-way ANOVA with Tukey’s post hoc test in (cg). Nuclei were stained with DAPI in blue. Scale bar, 50 μm. Genotypes, sample sizes, and statistical analyses are shown in Supplementary Data 1. Source data are provided in the Source Data file.

To determine whether LPWJL-emitted (2R,3R)-2,3-butanediol was required for host growth, we generated mutant bacteria incapable of producing (2R,3R)-2,3-butanediol. Toward this end, we deleted the acetolactate synthetase (ALS) gene responsible for α-acetolactate (a precursor molecule for 3-hydroxy-2-butanone) production to generate LPWJLΔALS. GC-MS analysis confirmed that the LPWJLΔALS largely lost its ability to produce all three OR42b ligands including (2R,3R)-2,3-butanediol (Fig. 7b). Importantly, in contrast to LPWJL volatiles, LPWJLΔALS volatiles showed impaired activation of airway-gut-brain axis (including tracheal branching formation, organ oxygenation, gut hormone activation, and DILP2 release) without affecting Hippo pathway (as evidenced by unchanged Merlin membrane localization before and after inhalation), resulting in diminished growth-promoting activity (Fig. 7c–g, and Supplementary Fig. 9). All these defects were reversed by volatiles of LPWJLΔALS_ALS strain (a genetically-rescued strain by re-introducing ALS gene into the LPWJLΔALS shown in the Fig. 7b) or volatiles of LPWJLΔALS together with (2R,3R)-2,3-butanediol (Fig. 7c–g, and Supplementary Fig. 9). These findings demonstrate that commensal LPWJL emitted (2R,3R)-2,3-butanediol as a major VSF, which in turn activated OR42b-dependent airway-gut-brain axis for host developmental homeostasis.

Discussion

Revolutionary discoveries of the role of the microbiome highlight novel and essential aspects of animal physiology. Even extremely complex physiological phenomena, such as individual decision-making and behavior, are also influenced by the microbiome58,59. Amazingly, animals, such as honeybees and hyenas, are known to communicate and identify each other via volatile scent molecules emitted from their microbiome, which acts as a sign of hive membership15,16,60,61. In this case, the olfactory organ may allow the transmission of olfactory signals to the brain, which aids animal-to-animal communication. In the present study, we identified a previously unrecognized olfaction-independent role of microbiome-emitted scents in animal growth via airborne interactions between host and microbe.

It is generally accepted that ORCO is indispensable for the function of individual ORs in sensing scent molecules by forming heterotetramers19,22,36. However, our results revealed that OR42b can function in an ORCO-independent manner (Fig. 1d and Supplementary Fig. 7c). Recent genome sequencing data revealed that a single ORCO gene is present in the genome of all insect orders except Archaeognatha, an evolutionarily primitive insect order such as the jumping bristletail (Machilis hrabei)36,62. Interestingly, Machilis hrabei OR5 (MhOR5) formed homotetramers37, suggesting that, prior to the emergence of ORCO, the ancestral OR is able to be functional without the assistance of ORCO. Our study suggests the presence of an OR42b homotetramer (Fig. 2b, c), with structural resemblance to MhOR5 (Supplementary Fig. 8). Notably, both OR42b and MhOR5 share similarities in the bulky hydrophobic residues located at the interfaces between subunits (Supplementary Fig. 8). These findings collectively support the idea that OR42b has the potential to form a homotetramer independent of ORCO. Based on these data, it can be speculated that the olfaction-independent non-neuronal OR function (e.g., ORCO-independent intestinal and airway OR42b function) may be phylogenetically primitive, which predates olfaction-dependent and ORCO-dependent OR function. In this context, it is not surprising that OR expression is not restricted to ORNs, from Drosophila to humans, but occurs in various non-olfactory organs including the gut. However, the in vivo biological functions of ORs expressed in non-olfactory organs remain largely unknown. The present study presents an extraordinary case of non-olfactory OR function transducing the microbiome signal contributing to essential host physiology, such as growth (Fig. 8).

Fig. 8: Model for the olfactory neuron-independent sensing of the microbiome-emitted (2R,3 R)-2,3-butanediol for the activation of somatotrophic airway-gut-brain axis.
figure 8

Traditional mechanism (i.e., olfactory neuron-dependent) of OR42b-mediated scent sensing (shown in the left panel) is based on the previous studies18,31,32,33,34. In this case, OR42b (shown in blue) expressed in olfactory neurons may form a heterotetramer (possibly two OR42b and two ORCO molecules) with co-receptor ORCO (shown in yellow). An olfactory neuron-independent mechanism of OR42b-mediated scent sensing discovered in the present study is shown in the right panel. In this case, OR42b (non-neuronally expressed in subsets of tracheal airway cells, EECs, and enterocytes) likely forms a homotetramer without the assistance of ORCO. In the absence of microbiome-emitted (2R,3R)-2,3-butanediol, Merlin (Mer) complex composed of Mer-kibra-Expanded (Ex) is tethered to the membrane-localized OR42b to sequentially activate Hippo (Hpo) kinase and Warts (Wts) kinase. Activated-Wts kinase phosphorylates Yorkie (Yki) for its cytoplasmic retention. Binding of (2R,3R)−2,3-butanediol to the OR42b initiates the dispersal of membrane-localized OR42b and Mer complex. Dispersed Mer complex inactivates Hpo-Wts kinases, resulting in Yki dephosphorylation for nuclear localization and target gene activation; Branchless (Bnl, a homolog of mammalian FGF) in the enterocytes, Breathless (Btl, a homolog of the mammalian FGF receptor) in the tracheal airway cells, and Allatostatin A (AstA) in the EECs. Bnl in enterocytes guides the migration of Btl+ tracheal airway cells during tracheal branching formation to deliver oxygen to the target organs for organ growth. Yki-dependent AstA expression in EECs induces CCHA2 peptide expression in AstA receptor-2 (AstAR2)-expressing EECs. CCHA2 peptide hormones in turn act on their receptors for DILP2 secretion from brain insulin-producing cells, thereby forming the somatotrophic gut-brain axis for systemic growth. CCHA2 peptide hormones in turn act on their receptors for DILP2 secretion from brain IPCs, thereby forming the somatotrophic gut-brain axis for systemic growth. Although the CCHA2 receptor (CCHA2R) has been shown to be expressed in larval IPCs51, it remains to be elucidated whether gut-derived CCHA2 directly activates CCHA2R-expressing IPCs or indirectly activates IPCs (e.g., via CCHA2R-expressing enteric neurons).

What are the advantages of air-diffusible growth factors? We found that (2R,3R)-2,3-butanediol can serve as a potent attractant for both Drosophila larvae and adults (Supplementary Fig. 10). We also found that the sensing of (2R,3R)-2,3-butanediol by the olfactory organ is dependent on both OR42b and ORCO (Supplementary Fig. 10). Consistent with these data, we found that Drosophila showed a stronger attraction to LPWJL compared to LPWJLΔALS, and this attraction is also dependent on both OR42b and ORCO (Supplementary Fig. 10). Based on these findings, it is tempting to speculate that the airborne nature of bacterial (2R,3R)-2,3-butanediol enables it to function as an attractant, thereby increasing the chance of encountering Drosophila in their natural habitat and consequently enhancing the probability of colonization in the Drosophila gut. Indeed, Lactiplantibacillus is one of the primary members of gut commensal bacteria found in both laboratory-reared and wild-caught Drosophila12. Once introduced into the gut, (2R,3R)-2,3-butanediol acts as a growth-promoting factor in an OR42b-dependent but ORCO-independent manner. These data indicate that bacterial (2R,3R)-2,3-butanediol serves a dual role as both an attractant and a growth-promoting factor for Drosophila, illustrating a striking example of airborne mutualism between host and microbe.

Considering millions of years of co-evolution between microbiome and host, the mutual benefit in the form of growth stimulation likely represents one of the oldest forms of host-microbiome mutualism. Currently, two commensal members of Drosophila, Lactiplantibacillus, and Acetobacter, are the best-known growth-promoting microorganisms11,12,63. It has been shown that D-alanine esterification on teichoic acids contributes to Lactiplantibacillus-mediated Drosophila larval growth promotion under conditions of chronic undernutrition64. In this study, the LP NC8 wild-type strain (NC8WT) was used as a growth-promoting strain. It was found that mutant bacteria lacking the dlt operon (NC8Δdlt), which results in the loss of D-alanine esterification on teichoic acids, showed a significant growth delay when colonizing GF animals. In contrast to this result, our preliminary results showed that both NC8WT and NC8Δdlt exhibited growth-promoting activity under our experimental inhalation conditions (Supplementary Fig. 11), indicating that the dlt operon is not involved in VSF-mediated growth promotion. Consistent with this idea, the NC8ΔALS and NC8Δdlt_ΔALS strains—(2R,3R)-2,3-butanediol-deficient mutant counterparts of NC8WT and NC8Δdlt, respectively—created by deleting the ALS gene, showed impaired inhalation-dependent larval growth (Supplementary Fig. 11). These findings suggest that LP strains promote host growth in Drosophila through at least two independent mechanisms: dlt operon-mediated D-alanine esterification on teichoic acids and ALS-mediated production of (2R,3R)-2,3-butanediol.

In the case of Acetobacter, metabolic pathway analyses reveal that a growth-promoting strain of Acetobacter, previously isolated from the Drosophila gut11, is unable to synthesize (2R,3R)-2,3-butanediol (Supplementary Fig. 12). However, our preliminary experiments indicate that it can support larval growth under our experimental inhalation conditions (Supplementary Fig. 12), suggesting that other VSFs, besides (2R,3R)-2,3-butanediol, are likely involved in host growth promotion. Further in-depth investigation will be needed to fully understand the Acetobacter-derived VSFs and their signaling pathways.

It is striking to note that VSFs including (2R,3R)-2,3-butanediol are also emitted from the plant root microbiome (e.g., rhizobacteria such as Bacillus subtilis and Bacillus amyloliquefaciens)65. Importantly, Bacillus-emitted (2R,3R)-2,3-butanediol promotes plant growth and enhances crop yields65. Interestingly, we found that (2R,3R)-2,3-butanediol emitted from the plant root microbiome can also promote Drosophila larval growth (Supplementary Fig. 13). These results reveal striking similarities between animal and plant growth systems mediated by microbiome-emitted volatiles, although the exact molecular mechanism of (2R,3R)-2,3-butanediol-mediated plant growth (e.g., the receptor and its downstream signaling pathway in a plant system) remains to be elucidated. In addition to host growth promotion, what are the advantages of (2R,3R)-2,3-butanediol in host-microbe interactions? In the plant system, it has been reported that (2R,3R)-2,3-butanediol can trigger systemic resistance, possibly by boosting plant innate immunity66. Furthermore, (2R,3R)-2,3-butanediol is shown to be required for longer root colonization by augmenting bacterial resistance to the acidic environment around the plant root67. In this context, it would be interesting to examine whether (2R,3R)-2,3-butanediol is involved in bacterial ability for gut colonization or modulation of host innate immunity in the Drosophila system.

Another interesting point is that LPWJL-mediated growth of malnourished hosts appears to be conserved from invertebrates to vertebrates. In a mammalian model of malnutrition, the cell wall component of LPWJL acts as a ligand for the NOD2 receptor, which is necessary for crypt cell proliferation, production of IGF-1, and promotion of postnatal growth13. Given that deregulation of the microbiome is closely associated with the progression of malnutrition-associated disease phenotypes (e.g., stunting)68,69, and that probiotics such as LPWJL improves systemic host growth under chronic malnutrition13,14, the discovery of (2R,3R)-2,3-butanediol as a growth-promoting factor may contribute significantly to understanding of the role of the microbiome in alleviating host malnutrition including humans. Interestingly, microbiome-derived metabolites have been shown to activate ORs expressed in mouse enterochromaffin cells, a subtype of EECs in the mammalian gut70. Further investigation into the presence of enterochromaffin cells responding to (2R,3R)-2,3-butanediol in the mammalian gut may provide novel insights into LPWJL-host interactions.

Methods

Fly strains

Fly stocks obtained from the Bloomington stock center were: Bnl-Gal4 (stock no. 62607); Bnl-LexA (stock no. 81576); LexAop-rcd2-RFP, UAS-CD8GFP (stock no. 67093); Btl-LexA (stock no. 66620); CCHA2-Gal4 (stock no. 84602); AstAR2-Gal4 (stock no. 84594); OR42b-Gal4 (stock no. 9972); Btl-Gal4 (stock no. 78328); UAS-Yorkie-Active (stock no. 28817); Actin-Gal4 (stock no. 25708); ORCO−/− (stock no. 23130); Mex1-Gal4 (stock no. 91368); UAS-GCaMP7c (stock no. 79030) and UAS-mCD8GFP (stock no. 27399); UAS-nlsTimer-NA (stock no. 78056)71. Fly stocks obtained from the Vienna Drosophila Resource Center were: UAS-Bnl-RNAi (stock no. 5730); UAS-DILP2-RNAi (stock no. 102033); UAS-CCHA2-RNAi (stock no. 102257); UAS-AstA-RNAi (stock no. 103215); UAS-CCHA2R-RNAi (stock no. 100290); UAS-AstAR2-RNAi (stock no. 108648); UAS-Yorkie-RNAi (stock no. 104523); UAS-OR42b-RNAi (stock no. 101143) and UAS-Or83b-RNAi (stock no. 13386). Other fly stocks used were: Btl Gal472; DILP2-Gal450; Yorkie-GFP-KI73; Merlin-YFP74; AstA-Gal453; and CCHA2−/−51.

Generation of transgenic and mutant flies

The gRNA sequences for each gene used in the experiment are listed in Supplementary Table 1. C-terminal epitope (3x FLAG)-tagged OR42b or OR42bDN was generated using CRISPR/Cas9 genome editing. AstA−/− mutants were generated using the method developed by Kondo and Ueda75. OR42b−/− mutants were generated using a conventional CRISPR/Cas9 genome editing method. HA-epitope fused to the carboxy-terminus of DILP2 B-chain and FLAG-epitope fused to the amino terminus of DILP2 A-chain (referred to as DILP2-HF) (Fig. 2a) were generated using CRISPR/Cas9 genome editing. DILP3-HF and DILP5-HF were also generated using the same methods. HA- and FLAG-epitope tagging sites for DILPs were used exactly as described previously76. P-element-mediated germline transformations were carried out to generate UAS-OR42bDN flies.

Fly rearing

The detailed fly genotypes used in this study are listed in the Supplementary Data 1. The flies were maintained at a temperature of 25 °C with a 12-h dark-light cycle on Bloomington Stock Center’s standard cornmeal-agar medium.

Experiments involving the treatment with volatiles

Germ-free embryos were generated exactly as described previously9 and maintained on a protein malnutrition diet [1% yeast (Lesaffre, France), 6.97% cornmeal (Sunglim Co., South Korea), 9.6% sucrose (Samyang Co., South Korea), 1.5% agar (Hansol Tech., South Korea), 0.03% Bokinin, and 0.5% propionic acid]. Germ-free embryos were treated by gases emitted from a bacterial plate (overnight cultured plate previously inoculated with 6 × 108 c.f.u. of bacterial cells) or from synthetic chemical compounds (0.3% or 0.03 %) in a chamber (Fig. 1a) throughout the development. Different bacterial strains including LP WJL strain (wild-type LPWJL strain, LPWJLΔALS, and LPWJLΔALS_ALS), Bacillus subtilis 168 strain77 (wild-type strain, B. subtilis ΔALS_S, B. subtilis ΔALS_D, and B. subtilis ΔPta) and Bacillus amyloliquefaciens FZB42 strain78 (wild-type strain, Bacillus amyloliquefaciens ALS_S, and Bacillus amyloliquefaciens FZB42ALS_D) were used. Body size or organ size was measured at 144 h AEL using a microscope (Zeiss; stemi 305) and was quantified using Zen 3.1 (blue edition) software.

Real-time PCR

Total RNA was extracted from 10 larval guts using the TRIzol reagent (Invitrogen). Real-time PCR was performed to quantify gene expression using the double-stranded DNA dye SYBR Green (Perkin Elmer). SYBR Green analysis was conducted using an ABI PRISM 7700 system (PE Applied Biosystems, Carlsbad, CA, USA) following the manufacturer’s instructions. All samples were analyzed in triplicate, and the levels of detected mRNA were determined by cycling threshold analysis, followed by normalization using RP49 as a control. The relative expression of the target gene was determined. Primer pairs for detecting gene transcripts are listed in Supplementary Table 2.

Immunohistochemistry

Tissues were dissected in 1x PBS and then fixed for 30 min with 4% paraformaldehyde. The gut tissue samples were washed five times with 0.1% Triton X-100 PBS and incubated with 5% bovine serum albumin in 0.1% Triton X-100 PBS for 1 h. In the case of brain tissues, samples were washed five times with 0.3% Triton X-100 PBS and incubated with 5% bovine serum albumin in 0.3% Triton X-100 PBS (for brains) for 3 h. The samples were then washed five times with 1x PBS and incubated with the primary antibody overnight at 4 °C. Subsequently, the samples were washed five times for 5 min with 0.1% Triton X-100 in PBS (for midguts) or 0.3% Triton X-100 in PBS (for brains), followed by incubation with the secondary antibody for 1 h at room temperature. The samples were washed five times for 5 min with 0.1% Triton X-100 (for midguts) or 0.3% Triton X-100 (for brains) in PBS and then washed five times for 5 min with 1x PBS. Finally, the samples were mounted using a mounting buffer (Vectorshield, Vector Laboratories Inc., Burlingame, CA, USA) and analyzed by confocal microscopy (Carl Zeiss). The primary antibodies used in this study were rabbit anti-GFP (1:500; Invitrogen A11122), mouse anti-Prospero (1:100; DSHB AB_528440), mouse anti-DsRed (1:500; Santa Cruz Biology sc-390909), mouse anti-FLAG (1:500; Sigma-Aldrich F1804), rabbit anti-CCHA2 (1:500; newly generated for this study), mouse anti-dlg (1:500; DSHB AB_528203), and rabbit anti-AstA (1:500; Biorbyt). The secondary antibodies used were Fluor 568 goat anti-mouse IgG (1:500; Invitrogen A11004), Fluor 568 goat anti-rabbit IgG (1:500; Invitrogen A11011), Fluor 488 goat anti-mouse IgG (1:500; Invitrogen A11001), and Alexa Fluor 488 goat anti-rabbit IgG (1:500; Invitrogen A11008). Alexa Fluor 633 Phalloidin (1:30; Invitrogen A22284) and DAPI (1:1000; Roche 10236276001) were used for brush border membrane and nuclear staining, respectively.

Quantification of tracheal coverage

To quantify the tracheal coverage of midguts, images acquired by confocal microscopy (LSM700; Carl Zeiss) at 20× magnification were individually processed using ImageJ. The ROI in the gut was cropped to a size of 300 × 300 pixels. The values of tracheal coverage represent the percentage of the area occupied by the trachea per ROI (i.e., the number of pixels of the trachea per that of the ROI of the gut).

Quantification of oxygen concentration

The oxygen concentration of gut tissue was measured using two-color DsRed FT Timer protein, essentially as described previously71. Briefly, the signal intensities of both green and red fluorescence were measured using ZEN image software within an ROI with a size of 150 × 150 pixels. The red/green fluorescence ratio indicates oxygen concentration.

Quantification of Yorkie translocation

To quantify Yorkie translocations in midguts, images acquired using confocal microscopy (LSM700; Carl Zeiss) at 40× magnification were processed individually with Zen image software. To facilitate visualization of the cytoplasmic and nuclear regions of individual cells, Dlg, a septate junction marker, and DAPI were used for staining. The nucleus-to-cytoplasm GFP intensity ratio was then calculated by dividing the GFP intensity in the nucleus by that in the cytoplasm. Three to five cells were selected from each image to measure Yorkie translocations.

Generation of ALS deletion mutant of LPWJL and LPNC8 strain

Homologous sequences of 1000 bp upstream and 1217 bp downstream of the ALS coding sequence were subcloned into the pGID 023 vector79. Subsequently, the constructed plasmid was transformed into LP via electroporation. A homologous recombination strategy was employed to generate the LPWJLΔALS, LPNC8ΔALS, and LPNC8Δdlt_ΔALS mutant strains. The mutant bacterial strains were confirmed via PCR and sequencing analysis. To generate a rescue strain (LPWJLΔAls_Als), the Als gene was cloned in the pBR 256 expression vector and then introduced into the LPWJLΔAls.

Measurement of circulating DILP2, 3 and 5

The HA and FLAG-tagged DILP2-HF, DILP3-HF, and DILP5-HF knock-in larvae were used to measure the circulating levels of DILP2, DILP3, and DILP5. In each sample, animals (a single larva or a pool of 10 larvae) were crushed in a tube containing 70 μL of PBS to obtain hemolymph. Solution (~70 μL) containing hemolymph was briefly centrifuged and an aliquot (50 μL) was used to quantify circulating DILP-HF level using a sandwich ELISA, as described previously76. FLAG-(GS)-HA peptide (D-Y-K-D-D-D-D-K-G-G-G-G-S-Y-P-Y-D-V-P-D-Y-A-amide) were synthesized (Anygen, South Korea) and used for a standard curve, as described previously76. Each value from a single larva was normalized by the volume of the respective larva.

Ex vivo calcium imaging using GCaMP

Ex vivo calcium imaging using GCaMP7c was performed essentially as described previously80. Larvae were dissected in the adult hemolymph-like (AHL) solution (108 mM NaCl, 8.2 mM MgCl2, 4 mM NaHCO3, 1 mM NaH2PO4, 2 mM CaCl2, 5 mM KCl, 5 mM HEPES, 80 mM sucrose, pH 7.3). Tissues (trachea or gut) were immobilized using a mesh (Electron Microscopy Science; No. 100) in a magnetic chamber (Chamlide; CM-B18-1PA). The tracheal cells or gut cells bathed in the AHL solution were recorded for 5 min to establish a baseline. Subsequently, the solution was changed to AHL solution supplemented with 0.003% of each stereoisomer of 2,3-butanediol volatiles. Images were captured at a speed of 3 s per frame using LSM700 Carl Zeiss confocal microscopy.

Expression and purification of OR42b

The DNA encoding OR42b was PCR-amplified using a Drosophila cDNA library and subsequently subcloned into an EEV vector containing N-terminal Strep-tag II and EGFP. Baculovirus containing the Strep II-EGFP tagged OR42b sequence was created in SF9 cells (ATCC CRL-1711). Cells were grown in suspension at 37 °C in SF-900 II SFM medium (Gibco) supplemented with 10% (v/v) fetal bovine serum with 5% (v/v) carbon dioxide until they reached a density of 0.8 ~ 1.2 × 106 cells/mL and then infected at a multiplicity of infection of 1. After 48 h, 10 mM sodium butyrate (Sigma-Aldrich) was added to the medium and the temperature was reduced to 30 °C for the remainder of the incubation. The cells were harvested ~48 h after initial infection by centrifugation and were flash-frozen in liquid nitrogen. Pellets were stored at −80 °C until thawed for purification.

For the purification step, frozen cell pellets were thawed on ice and resuspended in 20 mL of lysis buffer per gram of cells. Lysis buffer was composed of 50 mM HEPES (pH 7.5), 375 mM NaCl, 1 μg/mL leupeptin, 1 μg/mL aprotinin, 1 μg/mL pepstatin A, 1 mM phenylmethylsulfonyl fluoride (PMSF) and ~0.1 mg/mL DNase I. After resuspending and grinding the pellet using a tissue grinder (Kimble), OR42b was extracted using 2% (w/v) lauryl maltose neopentyl glycol (LMNG; Anatrace) for 2 h at 4 °C under constant stirring with a magnetic bar. After centrifugation at 20,000 × g, the supernatant was added to 0.2 mL StrepTactin Sepharose resin (GE Healthcare) and rolled at 4 °C for 2 h. The resin was collected and washed with 10 column volumes of 20 mM HEPES/ NaOH, and 150 mM NaCl with 0.01% (w/v) LMNG. OR42b was eluted by adding 2.5 mM desthiobiotin (DTB). The eluted sample was injected onto a Superose 6 Increase column (GE Healthcare) equilibrated in 20 mM HEPES/NaOH, 150 mM NaCl with 0.004% (w/v) LMNG.

Negative staining electron microscopy imaging

The eluted sample (5 μL) from StrepTactin Sepharose resin was applied to glow-discharged holey carbon grids with a thin layer of carbon. After adsorption for 1 min, the sample was stained with two droplets of 1% (w/v) uranyl formate solution and water, blotted gently to remove the residual stain, and air-dried. Specimens were examined under the JEOL 2100 Plus electron microscope equipped with RIO9 Camera and operated at 200 kV acceleration voltage, under a nominal magnification of 40,000 and pixel size of 1.3 Å. A total of 230 particles were manually selected from 10 micrographs and averaged to 2D classification using RELION381.

AlphaFold prediction and sequence comparison of OR42b

The monomeric and tetrameric structure of OR42b were predicted based on the protein sequence by AlphaFold and AlphaFold multimer using the ColabFold website82,83,84. Clustal Omega was used for the sequence alignment of OR42b and Machilis hrabei OR585.

Odorant-binding pocket determination using SwissDock

The monomeric OR42b, predicted by AlphaFold82, was used to investigate the binding of the agonist. Mol2 files for (2R,3R)-2,3-butanediol were obtained from the Zinc database86. Both the monomeric OR42b and the (2R,3R)-2,3-butanediol files were used as inputs in the web-based SwissDock program87. Initially, a blind docking process was performed to identify potential binding pockets for the agonist. Subsequently, (2R,3R)-2,3-butanediol was used to identify both the outer and inner binding sites of OR42b. The quality and pose scores of each binding were assessed based on Full Fitness and ∆G (kcal/mol). The docking poses of each molecule were visually analyzed using UCSF Chimera88.

Analysis of volatile organic compounds by HS-GC/MS

The quantitative analysis of three volatile organic compounds [3-hydroxy-2-butanone, diacetyl, and (2R,3R)-2,3-butanediol] was carried out using headspace (HS) sampler, combined with GC-MS (Shimadzu Corporation, Kyoto, Japan), which was equipped at the Ewha Drug Development Research Core Center. A total volume of 1 mL of each sample was added to separate 20 mL glass headspace vials and immediately capped with a septum-attached metal screw cap to minimize any loss of volatile analytes. Each vial was then transferred into the HS instrument from a rotating carousel. The HS unit was independently controlled by GC and MS, and the overall operation involved three fundamental steps: equilibration, pressurization, and sample transfer/loading to the loop. The optimized HS method was based on an oven temperature of 80°C and a sample/transfer line temperature of 180 °C. Under this condition, vials were equilibrated for 15 min, and subsequently pressurized at 50 kPa for 2 min with high-purity helium (He) gas. Pressurized sample gas in the vial was then transferred to the loop for 0.5 min (loading time) and finally fed into the GC for another 1 min (injection time). Subsequent GC separation was conducted using a 30 m × 0.25 mm i.d., 0.25 μm Stabilwax-MS capillary column with high-purity He at a flow rate of 1 mL/min as the carrier gas (mobile phase). GC oven was initially held at 40 °C for 5 min, then increased by 60 °C/min to 160 °C and held for 3 min. The temperature was once again ramped to 200 °C at a rate of 20 °C/min and held for an additional 5 min. The injection port of GC was maintained at 220 °C in split mode, with a ratio of 1:10. Samples were finally detected via MS under an electron ionization mode. The ion source temperature was set to 220 °C, and the detector voltage was adjusted relative to the tuning results. Final data were obtained in selected ion monitoring (SIM) mode. The ions selected for SIM were: m/z 41, 42, 43, 44, 50, and 86 for diacetyl; m/z 27, 29, 45, 43, and 88 for 3-hydroxy-2-butanone; m/z 45, 47, 55, 57, and 90 for (2R,3R)-2,3-butanediol. Chemical identification and quantitation of each separated peak were conducted with LabSolution software provided by Shimadzu Co.

Olfactory preference assay

The olfactory preference assay in Drosophila was conducted essentially as described previously89, with minor modifications. For the larvae, 20 third-instar larvae were positioned at the center of a petri dish with a diameter of 9 cm. Two different attractants were placed at opposite positions on its periphery. The preference assay was conducted for a period of 5 min. The number of larvae near each attractant was counted. For the adults, fifty 5-day-old female flies were introduced into a container containing two trap vials with different attractants for a period of 16 h. The preference index was calculated from a choice between water and 2,3-butanediol (3% v/v), or between parental LPWJL wildtype and its acetolactate synthase mutant lacking 2,3-butanediol production (~1.8 × 109 c.f.u. each).

Quantification of food intake

The food intake assay was conducted essentially as described previously90. Germ-free embryos were placed on a protein malnutrition diet containing 2.5% (w/v) blue food dye (FD&C Blue Dye No.1), and were treated either without or with inhalation of LPWJL volatiles during entire larval stages. At 24, 48, 72, and 96 h AEL, larvae were washed with autoclaved water to eliminate any residual dye present on their outer surface. In each measurement, three larvae were homogenized in 70 µL of autoclaved water and subsequently centrifuged for 60 s at 16,000 × g. The supernatant (50 µL) was used to measure the absorbance at 629 nm using the Flexstation3 Multi-Mode Microplate Reader (Molecular Devices, LLC).

Statistics and reproducibility

Experiment was repeated independently three times (Figs. 3d, 4f, g, 5a–d and 7c), four times (Figs. 1f and 3a, e), or five times (Fig. 2b), and one representative image was shown. For the 2D class images (Fig. 2c), an average was generated using 200 particles from 10 micrographs. Statistical analysis in this study was performed using GraphPad Prism 9.0. Comparisons between the two samples were conducted using an unpaired two-tailed t-test. For comparisons involving multiple samples, one-way ANOVA with Tukey’s post hoc test or two-way ANOVA with Tukey’s multiple comparison was employed.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.