Non-Obligate Pairwise Metabolite Cross-Feeding Suggests Ammensalic Interactions Between Bacillus and Aspergillus Species

Metabolite trade-offs at bacterial-fungal interfaces determine their ecological interactions. We designed a non-obligate pairwise metabolite cross-feeding (MCF) between Bacillus and Aspergillus. Cross-feeding Aspergillus metabolites (MCF-1) affected higher growth and biolm formation in Bacillus. LC-MS-based multivariate analyses (MVA) showed marked variations in the endogenous metabolite proles between the cross-fed and control Bacillus. We observed and validated that Aspergillus-derived oxylipins were rapidly depleted in Bacillus cultures concomitant with lowered secretion of cyclic lipopeptides (CLPs). Conversely, Bacillus extracts cross-fed to Aspergillus (MCF-2) diminished its mycelial growth and conidiation. Fungistatic effects of Bacillus-derived cyclic surfactins were temporally reduced following their hydrolytic linearization. MVA highlighted disparity between the cross-fed (MCF-2) and control Aspergillus cultures with marked variations in the oxylipin levels. We conclude that the pairwise MCF selectively benetted Bacillus while suppressing Aspergillus, which suggests their ammensalic interaction. Widening this experimental pipeline across tailored communities may help model and simulate BFIs in more complex microbiomes. ecological interactions between Bacillus and Aspergillus species under the non-obligate pairwise MCF conditions. This study highlights the ecological implications of MCF in BFIs beyond auxotrophies while underpinning the role of SMs in microbial tness. Oxylipins and surfactins belonging to the Aspergillus and Bacillus species, respectively, mediate ammensalic interactions that selectively benet bacteria and inhibit fungi. Using non-targeted metabolomics, we delineated the impact of pairwise MCF on the metabolic plasticity of the receiver species. Correlating the metabolomic and phenotype data, we also explained how the exogenously cross-fed metabolites perturbed both the endogenous metabolites and phenotypes in receiver species. We designed a tractable cultivation medium to study the pairwise MCF between Bacillus and Aspergillus, however it will be important to understand these BFIs in food fermentative niches where substrate cross-feeding and spatial factors also adds to the systems stochasticity. We believe that the reductionist experimental approach used in this study can be leveraged to design, manipulate, and understand BFIs in various microbiomes. freshly prepared inoculum and incubated for 216 h at 30 °C and 200 rpm. The samples were harvested at 0, 120, 168, and 216 to evaluate mycelial growth and conidiation patterns. The mycelial dry weight (mg/mL) was measured after ltering the fungal biomass through a pre-weighed lter paper (110 mm, 30 μm, Whatman Inc., Clifton, NJ, USA) and drying (65 °C) its overnight. Conidial counts were examined under a compound microscope using the Neubauer hemocytometer. All experiments were performed maintaining three independent biological replicates.


Introduction
Bacterial-fungal interactions (BFI) are widespread in nature and display a broad range of inter-kingdom ecological interactions, including both cooperative and competitive types. Besides physical adhesions, BFI are modulated by contactless chemical means through the release of nutritional compounds, secreted peptides, antibiotics, signaling molecules, and redox ions [1,2]. Considering the BFI, a major corollary is that metabolite cross-feeding (MCF) could reconcile the cooperative interactions contingent on compatible secretions under limiting conditions [3]. Although the evolutionary obligate cooperative interactions among microorganisms are theoretically and empirically interpreted, the non-obligate interactions remained largely unpredictable and poorly understood under non-limiting conditions. Factors like low culturability, high diversity, and lack of microbial model systems imitating the natural environment further complicate the laboratory models for MCF in BFIs [4,5].
BFIs play a vital role in driving major ecosystems that are indispensable to the environment, health, agriculture, and food [6,7]. BFIs contribute signi cantly to biogeochemical cycles through nutrient and enzyme trade-offs that facilitate biomass recycling [8]. Considering BFIs in health, phenazines secreted from Pseudomonas aeruginosa impair bio lm formation and enhance conidiation in Aspergillus avus, promoting its virulence [9,10]. BFI are important regulators of agriculture, where the plant growth promoting rhizobacteria like Streptomyces and Bacillus selectively promote the ectomycorrhizal fungi while inhibiting the pathogen's growth, and hence modulate the tri-trophic interactions [11]. Secreted metabolites from bacteria may modulate the growth and metabolism of interacting fungi, resulting in effects such as reduced ethanol production in yeast and mitigation of antibacterial defense in Aspergillus [1,12]. BFIs are important determinants in various food fermentative bioprocesses employed toward the production of soy foods, dairy products, alcoholic beverages, and processed meats [13]. Importance of recurring Bacillus and Aspergillus interactions cannot be understated in the soy food fermentations. Understanding the complexity of metabolite mediated cross-feeding interactions between Bacillus and Aspergillus species would be critical toward the design and tractability of a scalable bioprocess. Our study aims to explore the ecological implication of secretory secondary metabolites (SMs) in Bacillus and Aspergillus interactions beyond auxotrophies. We performed pairwise cross-feeding of the late-log phase metabolite extracts between Bacillus and Aspergillus species through growth medium conditioning. Herein, we used a non-targeted MS-based metabolomic approach to identify the key metabolite entities which likely determined the nature of Aspergillus -Bacillus interactions.

MCF in uence growth and developmental phenotypes in receiver species
Aspergillus metabolites promote Bacillus growth and bio lm formation. Late-log phase Aspergillus culture extracts signi cantly promoted Bacillus growth phenotypes. Higher growth indices, including cell turbidity (36-48 h), viability (36 h, p < 0.01), and biomass (12 h and 36 h, p < 0.01) were observed in the MCF-1 treated groups as compared to the controls (Fig. 1a-c). Furthermore, early onset of signi cantly higher bio lm formation was recorded for MCF-treated Bacillus cultures (24-36 h, p < 0.01) compared to the controls (Fig. 1d).
Bacillus metabolites suppressed Aspergillus growth and conidiation. Analysis of the MCF-2 (B d → A r ) results revealed that late-log phase metabolite extracts from Bacillus culture displayed fungistatic effects on Aspergillus ( Fig. 1e and 1f). Following the MCF-2, a signi cantly lower mycelial growth was observed for 120 h (p < 0.01) and 168 h incubated Aspergillus cultures as compared to the controls. Further, we recorded a signi cantly lower conidia density in MCF-2 cross-fed Aspergillus cultures at 168 h (p < 0.01) compared to the control. The inhibitory effects of Bacillus metabolites on Aspergillus conidiation were transient the higher conidia density was recorded for 216 h (p < 0.01) incubated Aspergillus cultures following the MCF-2, as compared to the controls.

MCF modulate endogenous metabolites secreted by receiver species
Aspergillus metabolites reduced the secretion of cyclic lipopeptides (CLPs) in Bacillus. We examined the time-correlated exometabolomes of Bacillus subjected to MCF-1 (A d → B r ) treatment with Aspergillus culture extracts. MVA based on LC-MS datasets displayed a clear disparity between the metabolite pro les of the cross-fed Bacillus cultures and the control sets. The unsupervised principal component analysis (PCA) score plot showed an overall variability of 26.14% (PC1 = 16.50%; PC2 = 9.64%) with the datasets segregated temporally between the cross-fed and control groups during the initial growth stages (up to 12 h). However, the datasets representing later stages (24-48 h) of growth were clustered together (Fig. 2a). The supervised partial least squared-discriminant analysis (PLS-DA) score plot also highlighted temporal segregation between the cross-fed treated and control sets across PLS 1 (Fig. 2b). PLS-DA showed an overall variance of 21.25% (PLS1 = 9.55%; PLS2 = 11.90%) among the datasets and indicated 27 signi cantly discriminant metabolites based on their variable importance projection (VIP) at > 0.7 and p < 0.05. From the signi cantly discriminant variables between the MCF-1 treated and control sets, we characterized 17 metabolites of bacterial origin, mostly CLPs, and three metabolites of Aspergillus origin re-extracted from Bacillus cultures, and 8 non-identi ed (N.I) entities (Supplementary Table 1). The PLS-DA model was evaluated with reliable goodness-of-t parameters, including R 2 X (0.302), R 2 Y (0.987), and Q 2 (0.885).
Cross-fed Aspergillus metabolites that were re-extracted and characterized from Bacillus cultures include oxylipins 9,12,13-trihydroxyoctadec-10-enoic acid (9,12,13-TriHOME) and 12,13-dihydroxy-9octadecenoic acid (12,. In addition, sphingofungin B and a non-identi ed (N.I. 1) compound of fungal origin were also detected from Bacillus cultures. After 12 h of incubation, we recorded a rapid depletion of the cross-fed Aspergillus metabolites from Bacillus cultures (Fig. 2c). Considering the endogenous metabolites of Bacillus origin, CLPs constituted the largest proportion of the detected compounds, including the iturins, fengycins, and surfactins. Notably, the cross-fed Bacillus cultures displayed a lower relative abundance of most CLPs as compared to the control cultures (Fig. 2c).
However, signi cantly higher levels of linear surfactins, including B-C16 (m/z 1068), A-C15 (m/z 1054), B-C15 (m/z 1040), B-C14 (m/z 1026), and B-C13 (m/z 1012) were recorded for cross-fed Bacillus. Further, a hybrid PK-NRP compound, dihydrobacillaene, was also characterized from Bacillus cultures with its lower relative abundance in cross-fed samples compared to the controls. Non-identi ed metabolites of Bacillus origin including N.I. 2, 3, and 4 were signi cantly lower in cross-fed cultures as compared to the controls. In contrast, N.I. 7 was relatively higher in cross-fed cultures, however the remaining metabolites displayed a similar abundance among both the cross-fed and control Bacillus cultures.
Bacillus metabolites rewired oxylipin production in Aspergillus. The unsupervised PCA score plot indicated a clear segregation between the metabolite pro les of Aspergillus subjected to MCF-2 (B d → A r ) and control groups, with an overall variance of 26.00% (PC1 = 16.00%; PC2 = 10.00%) (Fig. 3a).
Bacillus cross-fed with standard oxylipin (12,13-DiHOME) displayed signi cantly higher growth indices (cell turbidity, viability, and dry weight) than the control sets ( Fig. 4a-c). Unlike MCF-1, we observed a transient increase (≤ 12 h) in bio lm formation for the cross-fed Bacillus. Bio lm formations were signi cantly higher in control Bacillus cultures between 24-48 h (p<0.01) as compared to the oxylipin treated cultures (Fig. 4d). Being a cyclical phenomenon, both the oxylipin treated and control samples displayed lower bio lm formation during the later stages of incubation. Considering the metabolomes, MVA (PCA and PLS-DA) highlighted a signi cant variance between the oxylipin treated and control Bacillus cultures (Supplementary g. 1a and b). Based on the orthogonal projection to latent structuresdiscriminant analysis (OPLS-DA), we selected the endogenous metabolites as the biomarkers signifying high variance and correlations within the datasets representing oxylipin 12,13-DiHOME treated and control Bacillus cultures (Fig. 4e). Following the fast depletion of 12,13-DiHOME during initial growth stages, oxylipin treated Bacillus cultures showed higher relative abundance of cyclic and linear surfactins coupled with lower levels of iturins, dihydrobacillaene, and most fengycins except A-C14 derivative (Supplementary g. 1c).
Cyclic surfactin A-C15 suppressed growth and metabolism in Aspergillus. Based on the metabolite pro ling of the Bacillus extracts cross-fed to Aspergillus cultures in MCF-2, we concluded that surfactins could be the major determinants of Bacillus-Aspergillus interactions. Considering cyclic surfactin A-C15 as the representative of Bacillus CLPs re-extracted primarily from the cross-fed Aspergillus cultures, we tested its effects on the ecological tness of fungal partner.
Notably, the mycelial weights for the surfactin treated Aspergillus cultures remain temporally unvaried and signi cantly lower compared to the control sets between 120-216 h (Fig. 5a). Moreover, the treated sets also showed signi cantly lower conidia density compared to the respective controls, p<0.01 (Fig. 5b). MVA (PCA and PLS-DA) highlighted a marked disparity in the metabolite pro les of the surfactin treated and control Aspergillus cultures (Supplementary g. 2a and b). The OPLS-DA derived S-plot highlighted the metabolite biomarkers which demarcated the observed variance between the surfactin treated and control Aspergillus cultures (Fig. 5c). In corroboration with MCF-2, cyclic surfactin A-C15 (m/z, 1036) concentration was decreased while that of its linear derivative (m/z, 1054) increased temporally in treated Aspergillus cultures. Unlike MCF-2, the endogenous metabolomes for surfactin treated Aspergillus cultures were characterized with lower relative abundance of most oxylipins (5,8-DiHODE, 9,12,13-TriHOME, 12,13-DiHOME, and 13-HODE) except 9-HpODE as compared to the control. Further, we observed lower levels of citreoisocoumarin and asperculin A coupled with signi cantly higher abundance of sphingofungin B for surfactin treated cultures as compared with control cultures (Supplementary g. 2c).
Bivariate correlations recapitulate metabolite mediated interactions. Pearson's correlation networks inferred how the communalities in the metabolomic data, including both the cross-fed metabolites from donor species and the perturbations in endogenous metabolites, in uence phenotypes in receiver species. Cross-fed metabolites are either consumed or transformed, and hence are depleted or enriched, respectively, by receiver species. If the depletion of the cross-fed metabolites was concomitant with higher phenotypes, we assumed their positive effects on the tness of receiver species despite a negative correlation value. However, structural transformation of the cross-fed metabolites resulting in higher abundance of the derivative compounds succeeded by diminished phenotypes would correspond to have negative effect on the receiver's tness despite positive statistical correlations. In contrast, any variation in the endogenous metabolite levels would establish a direct correlation with phenotypes in receiver species. For the endogenous metabolites, positive correlations would be inferred to be phenotype promoting and vice-versa.

Discussion
It is believed that auxotrophies stabilize microbial interactions and turns them obligatory under a nutrient limiting ecosystems [15]. Hence, it can also be hypothesized that the lack of nutritional dependencies, more likely among the prototrophs colonizing a nutrient-rich environment, could result in a non-obligate and transient interaction. To test this hypothesis in the context of BFIs, we excluded the nutritional dependencies between Bacillus and Aspergillus partners by employing a common minimal media (CMM) that could support the growth of each individual species per se but lacks the microbial compounds likely to be exchanged. Both Bacillus and Aspergillus were cultivated stably in CMM, and their growth characteristics were monitored for more than three consecutive generations. To avoid the biases arising from physical interactions between microbes, we opted to cross-feed the metabolite extracts from bacterial and fungal partners through media conditioning in pairwise MCF.
SMs are not necessarily essential for vegetative growth but are important regulators of BFIs owing to their tness functions. SMs are mostly described to function as antibiotics and/or quorum sensing mediators [2]. Herein, we observed that late-log phase Aspergillus extracts promoted Bacillus growth coupled with higher and early onset of bio lm formation. LC-MS analysis of the Aspergillus culture extracts showed a high abundance of oxylipins, among others ( Supplementary Fig. 3a). Pertaining to the considerably high abundance of oxylipins in Aspergillus culture extracts used in cross-feeding (MCF-1), we sought to probe their potential effects on Bacillus tness. Oxylipins constitute an extensive class of oxygenated derivatives of polyunsaturated fatty acids which mediate intra-and inter-species signaling functions across the microbial kingdoms. Oxygenated products of oleic acid (C18:1), linoleic acid (C18:2), and linolenic acid (C18:3) are primarily involved in modulating the growth, development, nitrogen uptake, and species interactions for fungi [16,17]. Oleate oxylipins (10-HOME and 7,10-DiHOME) have also been reported to inhibit agellum-driven motility in Pseudomonas which promotes bio lm formation and virulence [14]. Considering this information, we argue that the oleate oxylipins (9,12,13-TriHOME and 12,13-DiHOME) in Aspergillus extract cross-fed to Bacillus might have affected the higher and early onset of bio lm formation following MCF-1 ( Fig. 1d; Supplementary Fig. 3b). Linoleate-derived oxylipins, including 5,8-DiHODE, 9-HpODE, and 13-HODE, were not re-extracted from the MCF-1 treated Bacillus cultures, which can be attributed to their lower stability. Heightened growth and bio lm formation in cross-fed Bacillus cultures was concomitant with the rapid depletion of exogenously supplied Aspergillus metabolites in the early log phase. Nevertheless, we continued to observe higher growth and bio lm formation in the cross-fed Bacillus cultures, which can be attributed to the synergistic effects of both the characterized and uncharacterized metabolites of Aspergillus origin. This points to the growth-promoting effects of Aspergillus metabolites, especially the oxylipins. Considering the endogenous metabolites levels perturbed in cross-fed Bacillus cultures (MCF-1), a lower relative abundance of most CLPs was evident except for the hydrolyzed linear derivatives of cyclic surfactins. Both surfactin secretion and bio lm formation are tightly regulated by quorum sensing mechanisms often involving a cascade of secondary metabolites [18]. Lower secretion of surfactin CLPs would result in proportionally higher surface tension between the bacteria and its growth surfaces which restricts swarming motility [19]. We argue that reduced agellar and swarming motility in uenced by cross-fed oxylipins and lower surfactin secretion, respectively, might synergistically affect higher bio lm deposition in Bacillus. Like MCF-1, Bacillus cultures cross-fed with 12,13-DiHOME also displayed enhanced phenotypes except the bio lm formation which suggests the transitory effects of oxylipins on bacterial motility (Fig. 4d). In addition, a signi cantly higher relative abundance of endogenously secreted cyclic and linear surfactins was evident for the oxylipin treated Bacillus cultures. Altogether, this corroborates our previous observation that a condition lacking oleate oxylipins and/or higher abundance of surfactin CLPs was conducive toward a low bio lm deposition. However, further studies are needed to ascertain the role of linear surfactins in Bacillus.
Bacillus CLPs are considered the key regulators of antagonistic interactions and provide the bacterial partner with a competitive edge over fungi. In the present study, Aspergillus cultures cross-fed with Bacillus extracts displayed impaired mycelial growth and conidiation following MCF-2 (B d → B r ). Late-log phase Bacillus culture extracts used in cross-feeding were consisted of three main CLPs classes, including iturins, fengycins, and surfactins (besides some linear derivatives), and an antibiotic compound dihydrobacillaene ( Supplementary Fig. 4a). However, we mainly re-extracted cyclic surfactins and their linear derivatives from Aspergillus cultures following MCF-2 ( Supplementary Fig. 4b). Surfactins are composed of a heptapeptide (L-Glu 1 -L-Leu 2 -D-Leu 3 -L-Val 4 -L-Asp 5 -D-Leu 6 -L-Leu 7 ) connected to the β-OH fatty acid chain (C 12 -C 16 ) through a lactone bond, together forming a CLP [20]. Owing to their amphiphilic structure, surfactins interact readily with lipid bilayers, thereby altering membrane permeability in fungal cells [21]. We observed a temporal decrease in the relative abundance of CLPs and a proportional increase in linear derivatives (only surfactins) in cross-fed Aspergillus cultures. This can be attributed to the growth-linked production of surfactin hydrolases in Aspergillus, as reported previously for bacterial species [22]. Intriguingly, Aspergillus phenotypes displayed strong negative correlations with cyclic surfactins but moderately positive or neutral correlations with linear surfactins. This suggests a loss of antifungal function, more precisely the fungistatic effects of cyclic surfactins following their structural linearization. Among the endogenous metabolites, higher abundance of linoleate oxylipins in cross-fed Aspergillus cultures might be associated to the fungal resilience under the challenged growth conditions. Recently, Niu et al. [23] have suggested the growth modulatory functions of linoleate oxylipin 5,8-DiHODE for regulating the lateral hyphal branching in Aspergillus species. However, we observed a lower abundance of oleate oxylipins in cross-fed Aspergillus cultures suggesting a biosynthetic rewiring of different oxylipins in cross-fed cultures. Together with linoleate (C18:2) and linoleniate (C18:3) derivatives, oleate (C18:1) oxylipins are believed to constitute precocious sexual inducers which determine the asexual/sexual modes of conidiation in Aspergillus [16]. We assume that altered levels of various oxylipins might have signi cantly in uenced the conidia density between the cross-fed and control Aspergillus cultures. As the Bacillus extracts used in MCF-2 had numerous characterized and uncharacterized metabolites, it was di cult to pinpoint how the surfactins alone affected Aspergillus growth and metabolism. Treating Aspergillus with standard cyclic surfactin A-C15 veri ed the fungistatic effects of CLPs as substantiated by the overall down regulation of phenotypes and oxylipin production.
Previously, we have reported that oxylipins decrease signi cantly in A. avus subjected to volatile organic compounds mediated growth stress, signifying their intricate signaling functions in Aspergillus development [24]. Thus, we establish that cyclic surfactins which constitute a considerable portion of Bacillus CLPs effectively inhibit Aspergillus growth and endogenous metabolite production, and hence act as key regulators of ammensalic interactions.
In conclusion, we posit the likely nature of metabolite-mediated ecological interactions between Bacillus and Aspergillus species under the non-obligate pairwise MCF conditions. This study highlights the ecological implications of MCF in BFIs beyond auxotrophies while underpinning the role of SMs in microbial tness. Oxylipins and surfactins belonging to the Aspergillus and Bacillus species, respectively, mediate ammensalic interactions that selectively bene t bacteria and inhibit fungi. Using non-targeted metabolomics, we delineated the impact of pairwise MCF on the metabolic plasticity of the receiver species. Correlating the metabolomic and phenotype data, we also explained how the exogenously crossfed metabolites perturbed both the endogenous metabolites and phenotypes in receiver species. We designed a tractable cultivation medium to study the pairwise MCF between Bacillus and Aspergillus, however it will be important to understand these BFIs in food fermentative niches where substrate crossfeeding and spatial factors also adds to the systems stochasticity. We believe that the reductionist experimental approach used in this study can be leveraged to design, manipulate, and understand BFIs in various microbiomes.
Microbial species and culture conditions. Bacillus amyloliquefaciens KCCM 43033 was procured from the 'Korean Culture Center of Microorganisms' (KCCM), Seoul, Republic of Korea. Bacterial culture was maintained in BEP agar (beef extract, 3 g/L; peptone, 5 g/L). Aspergillus oryzae RIB 40 (KACC 44967), was provided by the 'Korean Agricultural Culture Collection' (KACC) and maintained in Wickerhams antibiotic test medium (WATM) agar [25]. We adopted a common minimal medium (CMM) capable of supporting the growth of both the Bacillus and Aspergillus partners. CMM was partially designed based on the components used in Czapek-Dox medium for fungi and the growth medium employed by Zhi et al. [26] for Bacillus cultivation (Supplementary Method 1). We used CMM for microbial growths, culture harvest for metabolite extractions, and pairwise MCFs.
Culture harvest and extractions for MCF. Actively growing Bacillus seed culture was centrifuged, and the resulting pellet was washed twice with 1X PBS. Pellets were reconstituted in CMM to obtain a predetermined culture density (OD 600 nm ~ 0.8; colony forming units, CFU/mL ~ 10 7 /mL). subsequently, 0.1% of the inoculum was transferred into 100 mL of CMM in 500 mL Erlenmeyer asks (ba ed type). The cultures were incubated at 30 °C and 200 rpm for 48 h in a shaking incubator. Bacillus growth parameters, including the cell turbidity (OD 600 nm), viability (CFU/mL), and biomass (CDW, mg/mL) were recorded at every 12 h interval. The OD was measured using a laboratory spectrophotometer at 600 nm and the CFU was estimated by plating the cultures on BEP agar. Microbial cultures were centrifuged (10 776 x g 10 min), pellets were washed twice with 1X PBS, and the CDW was measured after the overnight drying at 65 °C. Late-log phase Bacillus cultures were harvested, centrifuged (10 776 rpm, 10 min) and the supernatants were subjected to liquid-liquid extraction using solvent mixture containing methanol, dichloromethane, ethyl acetate, and hexane at 1:2:3:1. Extraction was carried out in a shaking incubator (25 °C, 200 rpm) for 12 h, and the supernatant layer was decanted for drying in a vacuum concentrator (Hanil Scienti c, Korea). Dried extracts were weighed and re-constituted (10 000 ppm) in 80% methanol for the LC-MS analysis. Metabolites were again vacuum-dried and re-constituted in CMM prior to the MCF.
Aspergillus was cultured using freshly harvested conidia from WATM agar plates incubated for 216 h.
Subsequently, 0.1% of the conidia suspension (~10 7 conidia/mL) was inoculated into CMM (100 mL) in 500 mL Erlenmeyer asks (ba ed type) and incubated at 30 °C and 200 rpm for 216 h. Cultures were harvested temporally at 0, 120, 168, and 216 h, and analyzed for conidial density (counts/mL), mycelial dry weight (mg/mL), and metabolite extraction. The growth rates of Aspergillus species were estimated based on conidiation and mycelial dry weight data using the method described by Singh and Lee [27].
Culture broth corresponding to the late-log growth phase was ltered using 0.2 μm bottle-top vacuum lter (Corning Inc. NY, USA). Clear broth was subjected to liquid-liquid metabolite extraction using the same procedure as described above. Dried metabolite extracts were analyzed using the LC-MS prior to MCF.
Design of experiment for MCF. MCF between Aspergillus and Bacillus was performed by CMM conditioning prior to the respective inoculations (Fig. 7). Aspergillus. MCF-2 (B d → A r ) cross-fed and control cultures of Aspergillus were inoculated with 0.1% of freshly prepared inoculum and incubated for 216 h at 30 °C and 200 rpm. The samples were harvested at 0, 120, 168, and 216 to evaluate mycelial growth and conidiation patterns. The mycelial dry weight (mg/mL) was measured after ltering the fungal biomass through a pre-weighed lter paper (110 mm, 30 μm, Whatman Inc., Clifton, NJ, USA) and drying (65 °C) its overnight. Conidial counts were examined under a compound microscope using the Neubauer hemocytometer. All experiments were performed maintaining three independent biological replicates.
Non-targeted metabolite pro ling. Microbial cultures representing the cross-fed and control sets were subjected to metabolite extractions using the method described above. Dried samples were weighed and reconstituted in 80% methanol to achieve the required concentration (10 000 ppm) prior to LC-MS. Details of the LC-MS instrumentation and analyses are described in supplementary method 3.
Metabolite selection and functional annotations. Based on the non-targeted metabolite pro ling data for the pairwise MCF between Bacillus and Aspergillus species, signi cantly discriminant metabolites (VIP>0.7, p<0.05) contributing maximum toward the observed variance were selected. Cross-fed metabolites depleted maximum following the MCF were considered as the candidates most likely to have in uenced the phenotypes in the receiver species (Raw data 1 and 2). We estimated the amount of signi cantly discriminant metabolites based on their relative peak intensities in the culture extracts used in cross-feeding. Oxylipin 12,13-DiHOME (2.40 µg/mL), representing the linoleate-derived hydroxy fatty acids of Aspergillus origin was added to the CMM prior to Bacillus inoculation. Similarly, cyclic surfactin A-C15 (5.1 µg/mL) of Bacillus origin was cross-fed to Aspergillus cultures for validations. Three independent biological replicates were maintained for each cross-fed and control sets.

Data processing and statistical analyses
Phenotype data. Statistical signi cance of the phenotype datasets was examined using the unpaired sample t-tests with PASW Statistics 18 software packages (SPSS Inc. Chicago, Illinois, USA).
LC-MS data. Raw data les were converted to NetCDF format using the built-in software (Thermo Xcalibur 2.2, Waltham, MA, USA) and subjected to alignment for signi cant peak-picking, mass artifact ltration, baseline correction, RT shift corrections, and accurate mass calculation using MetAlign (Version 041012, RIKILT-WUR Institute of Food Safety). Peak aligned and noise subtracted data was examined for class-wise variations among the datasets using MVA in the SIMCA-P+ software (v 12.0, Umetrics, Umea, Sweden). The bivariate Pearson's correlation between metabolite abundance and corresponding phenotypes was estimated endogenous metabolomes for receiver species were estimated using PASW statistics. Correlation networks were visualized using Cytoscape software v3.7.2 [29].

Declarations
Data availability. LC-MS/MS data related to this study is available from the corresponding author upon reasonable request. Source data for microbial phenotypes, LC-MS/MS metabolite pro ling, and associated statistical correlations are presented in raw data les 1-4.
Correspondence. Any request for materials and data should be addressed to the corresponding author. Aspergillus culture extracts promoted Bacillus growth and bio lm formation, while the Bacillus extracts suppressed mycelial growth and conidiation in Aspergillus. Following the MCF-1 (Ad → Br), Bacillus phenotypes (a) cells turbidity, (b) cells viability, (c) cells dry weight, and (d) bio lm formation was signi cantly enhanced in cross-fed cultures compared to the controls. In MCF-2 (Bd → Ar), Aspergillus growth phenotypes (a) mycelial dry weight, and (b) conidia density was signi cantly inhibited in cross-fed cultures compared to the controls. The statistical signi cance for the data were evaluated using the unpaired sample t-test with p<0.01** and p<0.05*. ND: not detected.   Oxylipin 12,13-DiHOME promoted higher growth and transitory bio lm formation, and enhanced surfactin secretion in Bacillus. Effects of 12,13-DiHOME treatment on Bacillus phenotypes: (a) cells turbidity, (b) cells viability, (c) cells dry weight, and (d) bio lm formation, compared to the respective controls. The data represent the mean (±SD) of the values corresponding to three independent biological replicates used in the study. Statistical signi cance for the data were evaluated using the unpaired sample t-test with p<0.01** and p<0.05*. (d) S-plot based on the OPLS-DA model indicating the metabolite biomarkers whose levels varied signi cantly between the treated (oxylipin: 12,13-DiHOME) and control groups. Large & green colored circle indicates oxylipin 12,13-DiHOME whereas small & red colored circles represent endogenous metabolites of Bacillus origin.

Figure 5
Cyclic surfactin A-C15 modulated lower mycelial growth, conidiation, and oxylipins production in Aspergillus. Effects of cyclic surfactin A-C15 on Aspergillus (a) mycelial dry weight, and (b) conidia density as compared to the control sets. The data represent the mean (±SD) of the values corresponding to three independent biological replicates used in the study. The statistical signi cance for the data were evaluated using the unpaired sample t-test with p<0.01** and p<0.05*. (d) S-plot based on the OPLS-DA model indicating the metabolite biomarkers whose levels varied signi cantly between the treated (surfactin) and control groups. Large circles (red: cyclic surfactin A-C15; blue: linear surfactin A-C15) indicate surfactins whereas small & green colored circles represent endogenous metabolites of Aspergillus origin. ND: not detected.

Figure 6
Metabolite abundance correlates with phenotypes in receiver species. Pearson's correlation-based networks highlighting the communalities for (a) the cross-fed Aspergillus metabolites, and (b) the standard oxylipin 12,13-DiHOME with Bacillus phenotypes. Conversely, correlations were estimated for (c) the cross-fed Bacillus metabolites, and (d) the standard cyclin surfactin C15 with Aspergillus phenotypes.
In addition, we also highlighted the bivariate correlations between the endogenous metabolite levels and respective phenotypes. Bacillus and Aspergillus phenotypes are indicated with large & colored circular nodes. Metabolites of Bacillus origin are represented by small & white circular nodes, where those of Aspergillus origin are indicated with hexagon shaped sea-green nodes.

Figure 7
Schematics depicting the design of experiment for the non-obligate pairwise metabolite cross-feeding (MCF) between Aspergillus oryzae and Bacillus amyloliquefaciens. Late log phase culture extracts from Aspergillus as donor (Ad) partner were extracted and cross-fed to Bacillus as receiver (Br) partner through