Deazaflavin metabolite produced by endosymbiotic bacteria controls fungal host reproduction

Abstract

The endosymbiosis between the pathogenic fungus Rhizopus microsporus and the toxin-producing bacterium Mycetohabitans rhizoxinica represents a unique example of host control by an endosymbiont. Fungal sporulation strictly depends on the presence of endosymbionts as well as bacterially produced secondary metabolites. However, an influence of primary metabolites on host control remained unexplored. Recently, we discovered that M. rhizoxinica produces F_O and 3PG-F₄₂₀, a derivative of the specialized redox cofactor F₄₂₀. Whether F_O/3PG-F₄₂₀ plays a role in the symbiosis has yet to be investigated. Here, we report that F_O, the precursor of 3PG-F₄₂₀, is essential to the establishment of a stable symbiosis. Bioinformatic analysis revealed that the genetic inventory to produce cofactor 3PG-F₄₂₀ is conserved in the genomes of eight endofungal Mycetohabitans strains. By developing a CRISPR/Cas-assisted base editing strategy for M. rhizoxinica, we generated mutant strains deficient in 3PG-F₄₂₀ (M. rhizoxinica ΔcofC) and in both F_O and 3PG-F₄₂₀ (M. rhizoxinica ΔfbiC). Co-culture experiments demonstrated that the sporulating phenotype of apo-symbiotic R. microsporus is maintained upon reinfection with wild-type M. rhizoxinica or M. rhizoxinica ΔcofC. In contrast, R. microsporus is unable to sporulate when co-cultivated with M. rhizoxinica ΔfbiC, even though the fungus was observed by super-resolution fluorescence microscopy to be successfully colonized. Genetic and chemical complementation of the F_O deficiency of M. rhizoxinica ΔfbiC led to restoration of fungal sporulation, signifying that F_O is indispensable for establishing a functional symbiosis. Even though F_O is known for its light-harvesting properties, our data illustrate an important role of F_O in inter-kingdom communication.

Introduction

Bacterial–fungal interactions are critically important in agriculture and human health, and their relevance to ecology is now seen as the norm [1], with numerous studies focusing on identifying the individual microorganisms residing within bacterial–fungal communities (e.g. The Human Microbiome Project) [2]. Among these interactions, fungal–bacterial endosymbiosis, in which bacteria reside within the cytosol of fungal hyphae, represents the most intimate relationship. Contemporary studies have revealed that endosymbiotic bacteria, commonly found in members of the Mucoromycota, are not unassuming background actors, but rather have considerable influence on fungal pathogenicity, physiology, and ecology [3].

The endosymbiosis between the Mucoromycota fungus Rhizopus microsporus and the bacterium Mycetohabitans rhizoxinica is a unique example of fungal pathogenicity being effectuated by an endosymbiont [4]. In housing M. rhizoxinica, R. microsporus gains the ability to kill rice seedlings through the secretion of the bacterial secondary metabolite rhizoxin [5], which also protects the fungus from soil-dwelling micropredators [6]. A fundamental process that is key for the persistence of the Rhizopus–Mycetohabitans symbiosis is that endosymbionts are translocated into the fungal spores during host reproduction [7]. If endosymbionts are eliminated through antibiotic treatment, R. microsporus is unable to reproduce vegetatively (Fig. 1A) [7]. Thus, the ability of R. microsporus to reproduce asexually via sporulation strictly depends on the presence of endobacteria. This endosymbiont-dependent control of fungal sporulation is known to be mediated by a multitude of bacterially produced symbiosis factors such as a specialized lipopeptide O-antigen [8], transcription activator-like effectors [9], and the secondary metabolite habitasporin [10].

Even though M. rhizoxinica is known for its high potential for secondary metabolite biosynthesis [10, 11], it also produces unusual molecules linked to primary metabolism such as F_O (7,8-didemethyl-8-hydroxy-5-deazariboflavin) and 3PG-F₄₂₀, a derivative of the deazaflavin cofactor F₄₂₀ (Fig. 1B) [12]. Although F_O is the precursor of 3PG-F₄₂₀, both metabolites have entirely distinct physiological roles. F_O plays a key role as a light-harvesting chromophore in DNA photolyases across all three domains of life (Bacteria, Archaea, and Eukarya) [13–15]. F₄₂₀ was first discovered as a cofactor of methanogenesis in anaerobic archaea [16]. Beyond archaea, F₄₂₀ metabolism has only been extensively studied in the bacterial phylum Actinobacteria [17], where it mediates diverse catalytic reactions central to processes such as antibiotic biosynthesis in streptomycetes [18] and drug resistance in Mycobacterium tuberculosis [19]. More recently, genomic studies uncovered a broad distribution of genes encoding enzymes responsible for the biosynthesis of F₄₂₀ in Gram-negative bacteria [20, 21]. Of these, three species were confirmed as F₄₂₀ producers [21].

F₄₂₀ production by Gram-negative bacteria that engage in a symbiotic lifestyle has been reported in two cases—namely, the sponge symbiont Candidatus Entotheonella factor [22] and M. rhizoxinica [12]. So far, the physiological role of the deazaflavin cofactor in these microorganisms remains unknown. The R. microsporus–M. rhizoxinica symbioses provide an ideal model system to investigate the function of F₄₂₀ because M. rhizoxinica can be cultured axenically and a readily discernible phenotype exists that is indicative of a stable symbiosis, i.e. sporulation of the fungal host. Other than the discernment of whether metabolites from M. rhizoxinica contribute to the ability of R. microsporus to reproduce, the Rhizopus–Mycetohabitans model allows earlier aspects of symbiosis establishment to be probed such as bacterial colonization of hyphae and intracellular distribution and survival [23–26].

Here, we show that homologous genes encoding the pathway to produce 3PG-F₄₂₀ are conserved in the genomes of several endofungal Mycetohabitans symbionts. Furthermore, we demonstrate that F_O, the precursor of 3PG-F₄₂₀, is a symbiosis factor that is instrumental in the maintenance of the phytopathogenic R. microsporus–M. rhizoxinica alliance.

Materials and methods

Strains and culturing conditions

Eight R. microsporus strains harboring Mycetohabitans spp. endobacteria were used in this study (Supplementary Table 1) [27]. Endobacteria from R. microsporus ATCC62417 were eliminated by continuous antibiotic treatment [28] resulting in apo-symbiotic R. microsporus (RMapo). The absence of endobacteria was confirmed by fluorescence microscopy and an absence of rhizoxin in extracts of the fungal mycelium [5]. Both R. microsporus strains (ATCC62417 and RMapo) were cultivated on Potato Dextrose Agar (PDA; Becton, Dickinson & Company, Sparks, MD) at 30°C. Bacterial endosymbionts were isolated from the mycelium of eight fungal strains as previously reported [29]. Pure cultures of M. rhizoxinica were grown at 30°C in MGY M9 minimal medium (10 g/l glycerol, 1.25 g/l yeast extract, M9 salts: 7 g/l K₂HPO₄, 2 g/l KH₂PO₄, 600 mg/l C₆H₇NaO₇, 1 g/l (NH₄)₂SO₄, and 100 mg/l Mg₂SO₄) or Standard I Nutrient Agar (Merck, Darmstadt, Germany) supplemented with 1% glycerol.

Gene expression studies

RNA was isolated from M. rhizoxinica HKI-454 using the Quick-RNA Fungal/Bacterial Miniprep Kit (Zymo Research, Irvine, CA) following the manufacturers’ recommendations. Because there is no standard method for RNA extraction from endofungal bacteria, we developed our own protocol for reliable RNA extraction from symbiotic M. rhizoxinica (see Supplementary File “Materials and Methods” for details).

Quantitative PCR (qPCR) was used to study expression levels of two genes (fbiC and cofC) in axenic M. rhizoxinica HKI-454 as well as in M. rhizoxinica HKI-454 living in symbiosis with R. microsporus (ATCC62417). The gene fbiC encodes F_O synthase, which catalyzes the key step in F₄₂₀ biosynthesis leading to the formation of the metabolically active precursor F_O (Fig. 2A) [30]. A second important step, the side-chain biosynthesis of 3PG-F₄₂₀-0 is catalyzed by CofC. The M. rhizoxinica rpoB gene was used as an internal control for calculation of expression levels and normalization. All qPCR primer pairs used are listed in Supplementary Table 2A. First, primer efficiencies were calculated from standard curves generated with serial dilutions (ranging from 1 to 10⁻³) of axenic M. rhizoxinica cDNA. All primer pairs have an efficiency of 93% to 96% and were used in subsequent gene expression experiments (Supplementary Table 3) using MyTaq HS Mix (Bioline, London, UK) and EvaGreen® Fluorescent DNA Stain (Jena Bioscience, Jena, Germany). Each sample was run in three technical replicates on a QuantStudio 5 Real-Time PCR machine (Applied Biosystems, Waltham, MA). A control reaction, in which sterile water replaced the cDNA, was carried out in parallel. Amplifications of each template were performed in three biological replicates (n = 3). The cycle threshold (C_t) values were calculated using the Design and Analysis Software v1.5.2 (Applied Biosystems). C_t values were used for quantification of expression levels via the 2^−ΔΔCt method [31] in MS Excel. Statistical analysis was performed in GraphPad Prism 9.5.1 (GraphPad Software, La Jolla, CA, www.graphpad.com). An unpaired t-test with Welch’s correction was used to study the gene expression level of fbiC and cofC in M. rhizoxinica wild type grown in axenic culture in relation to M. rhizoxinica living in symbiosis with R. microsporus.

Extraction of F_O/3PG-F₄₂₀-n from fungal mycelium and axenic Mycetohabitans spp.

Small pieces of fungal mycelium were inoculated in 100 ml nutrient broth medium (Merck Millipore, Darmstadt, Germany) in 500 ml baffled Erlenmeyer flasks and incubated at 30°C and 110 rpm. After 7 days of incubation, fungal mycelium was strained through a 40 μm cell strainer (Corning Inc.) and resuspended in 10 ml ice-cold HPLC-grade methanol (VWR Chemicals, Darmstadt, Germany). Axenic bacterial overnight cultures of Mycetohabitans spp. were inoculated in 50 ml MGY M9 minimal medium in 300 ml baffled Erlenmeyer flasks and cultured at 30°C and 110 rpm. After 7 days of incubation, bacterial cultures were snap-frozen in liquid nitrogen in 500 ml round-bottom flasks and freeze-dried overnight (Christ Alpha loc-1 m, Martin Christ Gefriertrocknungsanlagen GmbH, Osterode am Harz, Germany). Dry Mycetohabitans spp. cultures were then resuspended in 10 ml ice-cold HPLC-grade methanol (VWR Chemicals).

All resuspended samples were sonicated for 20 min (Sonorex RK100 Ultrasonic bath, Bandelin, Berlin, Germany) and then shaken (250 rpm) for 1 h followed by centrifugation (10 000 × g) for 15 min (Centrifuge 5810R, Eppendorf, Hamburg, Germany). The supernatant was filtered through a paper filter into a 500 ml round-bottom flask and the solvent was evaporated on a vacuum rotary evaporator (Laborota 4003, Heidolph, Schwabach, Germany). Dry extracts were redissolved in 2 ml LC–MS-grade water (Carl Roth, Karlsruhe, Germany). The aqueous extracts were loaded on to a 100 mg C18 Chromabond cartridge (Macherey-Nagel, Düren, Germany), which was washed with methanol and conditioned with LC–MS-grade water beforehand. The cartridge was then washed and desalted with 1 ml LC–MS-grade water, followed by elution of the sample in three fractions with methanol (0.5 ml, 1.0 ml, and 0.5 ml). F_O was almost always concentrated in the second fraction. The samples were centrifuged at 10000 × g for 20 min and the supernatant was stored at −20°C until further LC–MS analysis (see Supplementary File “Materials and Methods” for details).

Generation of M. rhizoxinica ΔfbiC and M. rhizoxinica ΔcofC using CRISPR/Cas cytidine base editing and subsequent genetic complementation

To investigate a possible role of F_O and 3PG-F₄₂₀-n in the symbiosis, two genes (fbiC and cofC) were deleted using cytidine base editing. To establish a working CRISPR/Cas cytidine base editing method for M. rhizoxinica HKI-454, we used our knowledge gained from the CRISPR/Cas modification of the Betaproteobacterium Burkholderia gladioli [32], where we utilized a temperature-sensitive, low-copy plasmid (pTsK-CasRed-Bt, Supplementary Table 4) containing among others a codon-optimized cas9^* gene. See Supplementary File “Materials and Methods” for details on the generation of the base editing plasmids pTsK-AnCU-fbiC (targeting fbiC) and pTsK-AnCU-cofC (targeting cofC) and subsequent transfer of the plasmids into M. rhizoxinica by electroporation.

Both M. rhizoxinica ΔfbiC and M. rhizoxinica ΔcofC were checked for deleterious mutations using whole genome sequencing (see Supplementary File “Materials and Methods” for details). Both mutants were subsequently genetically complemented. Briefly, M. rhizoxinica ΔfbiC and M. rhizoxinica ΔcofC were transformed with the plasmids pRANGER-fbiC and pRANGER-cofC, respectively. These expression vectors carry the native promoter controlling expression of each gene, yielding the complemented strains M. rhizoxinica ΔfbiC pRANGER-fbiC and M. rhizoxinica ΔcofC pRANGER-cofC (see Supplementary File “Materials and Methods” for details).

Co-culture / sporulation bioassay

Liquid sporulation bioassays, containing apo-symbiotic R. microsporus and 100 μl of overnight cultures of M. rhizoxinica wild type (HKI-454), M. rhizoxinica ΔfbiC, M. rhizoxinica ΔcofC, M. rhizoxinica ΔfbiC pRANGER-fbiC, or M. rhizoxinica ΔcofC pRANGER-cofC were performed as previously described [9]. Experiments were performed at least four times independently (n ≥ 4 biological replicates) with six technical replicates on each plate. GraphPad Prism 9.5.1 (GraphPad Software) was used for statistical analysis and graphing. Data from spore counts were compared between M. rhizoxinica strains using one-way analysis of variance (ANOVA) and Tukey Honestly Significant Difference (HSD) test function in GraphPad. P values (P) <.05 were considered statistically significant. The Brown–Forsythe test was used to test for equal variance and a P < .05 was considered significant.

To test whether F_O alone could trigger fungal sporulation or whether the presence of bacteria is still necessary, we chemically synthesized F_O (see Supplementary File “Materials and Methods” for details) and added varying concentrations of synthetic F_O (312 ng/ml, 625 ng/ml, and 937 ng/ml) to co-cultures of R. microsporus and M. rhizoxinica ΔfbiC. Apo-symbiotic R. microsporus was also grown in the absence of endobacteria in 48-well plates in either liquid or solid potato dextrose medium that was supplemented with a final concentration of 73 μg/ml synthetic F_O. Spores were harvested and quantified as described previously [9], and the formation of sporangia was visualized using a Zeiss Axio Zoom.V16 Stereomicroscope (Carl Zeiss Microscopy, Oberkochen, Germany).

Imaging of endohyphal M. rhizoxinica strains

To visualize the localization of F₄₂₀-producing and F₄₂₀-deficient M. rhizoxinica strains within the fungal hyphae, strains were transformed with a plasmid encoding green fluorescent protein (GFP) (see Supplementary File “Materials and Methods” for details) and then used in the liquid sporulation assays as described previously [9]. After 5 days of incubation, fungal mycelium was transferred to PDA Petri dishes. Sterile, high-performance cover glasses (D = 0.17 mm +/− 0.005 mm, refractive index = 1.5255 +/− 0.0015, Carl Zeiss Microscopy) were placed on the agar plate in close proximity to the mycelium. Plates were incubated for 1–2 days at 30°C to allow fungal mycelium to grow on top of the cover glass. Once fungal mycelium had grown on top of the cover glass, the cover glasses were carefully removed from agar plates, and the mycelium was fixed for 5 min in fixing solution (3.7% formaldehyde, 25 mM KH₂HPO₄). Samples were washed at least three times with 25 mM KH₂HPO₄ before mounting the cover slide in embedding agent (ProLong Gold Antifade Mountant, Invitrogen) on a microscope slide (superfrost, ground 90°, Thermo Scientific). Fluorescence microscopy was carried out using a fluorescence wide-field microscope (LaserWF mode in the ELYRA 7, Carl Zeiss Microscopy), and bacterial cells were visualized using a laser excitation at 488 nm and an emission bandpass from 495 nm to 590 nm. For superresolution images, the structured illumination mode was used with a grating of 27.5 μm period. Fluorescent wide-field images were used to quantify bacterial cells inside of fungal hyphae (see Supplementary File “Materials and Methods” for details).

Results

Products of conserved F₄₂₀ biosynthetic gene clusters are produced under symbiotic conditions

The discovery that two Mycetohabitans strains encode a complete F₄₂₀ biosynthetic gene cluster [12, 33], despite their highly reduced genomes [10, 34], raised the question as to whether the cofactor might contribute to the integrity of the bacterial–fungal symbiosis. Therefore, we searched the genomes of six additional Mycetohabitans strains that are endosymbionts of globally distributed R. microsporus strains for F₄₂₀ biosynthesis loci (Supplementary Table 1) [27]. We found a complete, highly conserved F₄₂₀ biosynthesis locus, containing the requisite genes for F₄₂₀ biosynthesis, in all Mycetohabitans-symbionts of R. microsporus (Fig. 2A and Supplementary Fig. 1). The gene fbiC encodes F_O synthase, which catalyzes the key step in F₄₂₀ biosynthesis leading to the formation of the metabolically active precursor F_O (Fig. 2A) [30]. A second important step, the side-chain biosynthesis of 3PG-F₄₂₀-0 is catalyzed by CofC and CofD [12, 33]. Finally, a γ-glutamyl ligase (CofE) catalyzes the successive addition of l-glutamate residues (n) to 3PG-F₄₂₀-0 yielding 3PG-F₄₂₀-n [35].

To investigate under which conditions 3PG-F₄₂₀ and its precursor F_O are produced, we extracted and analyzed the following cultures by liquid chromatography coupled with tandem mass spectrometry (LC–MS/MS): (i) eight R. microsporus strains containing their corresponding Mycetohabitans spp. endosymbionts, (ii) one R. microsporus strain (ATCC62417) cured of its endosymbionts (RMapo), (iii) R. microsporus strains that are naturally endosymbiont-free, and (iv) axenic Mycetohabitans spp. strains that were isolated from their corresponding R. microsporus host. The metabolic profiling showed that both F_O and 3PG-F₄₂₀-n species are predominantly produced in Mycetohabitans strains when living as endosymbionts of R. microsporus (Fig. 2B and Supplementary Fig. 2A). In contrast, apo-symbiotic and naturally endosymbiont-free R. microsporus strains do not produce any F_O or 3PG-F₄₂₀-n species (Fig. 2C and Supplementary Fig. 2B). We noted that endosymbiotic bacteria grown outside of their host sporadically produce F_O and 3PG-F₄₂₀-n under the conditions tested (Fig. 2D and Supplementary Fig. 2C). 3PG-F₄₂₀-2 and 3PG-F₄₂₀-3 are the most abundant species in Mycetohabitans strains, although up to four glutamate residues were detected (3PG-F₄₂₀-4) (Supplementary Fig. 2C).

In light of the finding that F_O and 3PG-F₄₂₀-n are predominantly produced under symbiotic conditions, we probed whether F_O and 3PG-F₄₂₀-n production is constitutive or dependent on the presence of the fungal host. Gene expression of fbiC and cofC was monitored in axenic M. rhizoxinica HKI-454 and in M. rhizoxinica HKI-454 living as endosymbiont in R. microsporus ATCC62417. These genes were chosen because they both encode enzymes that catalyze key steps in F₄₂₀ biosynthesis (Fig. 2A). A gene encoding the β-subunit of the bacterial RNA polymerase (rpoB) was used as an internal control for RNA integrity and cDNA synthesis efficiency. Although it is inherently difficult to isolate high-quality RNA from endofungal bacteria, we developed a reliable method for RNA extraction from symbiotic M. rhizoxinica (Supplementary Fig. 3). Using qPCR, we show that fbiC is 1.6-fold upregulated (unpaired t-test: t = 2.87, df = 4.0, P = .0455, Supplementary Table 5) and cofC is 4.6-fold upregulated (unpaired t-test: t = 4.221, df = 4.0, P = .0135, Supplementary Table 6) in symbiotic M. rhizoxinica compared to axenic bacterial cultures (Fig. 2E).

Based on the prevalence of F₄₂₀-biosynthesis genes in the genomes of endofungal Mycetohabitans strains and the increased production of F_O and 3PG-F₄₂₀-n under symbiotic conditions, backed up by elevated transcription levels of F₄₂₀ biosynthetic genes, we deemed it important to further probe the role of these metabolites in the symbiosis.

Generation of F₄₂₀-deficient M. rhizoxinica using CRISPR/Cas-assisted base editing

To facilitate subsequent investigations into the role of F_O/3PG-F₄₂₀ in the Rhizopus–Mycetohabitans symbiosis, we performed targeted gene inactivation of fbiC and cofC. As both genes belong to one operon and are under the control of the same promotor (Fig. 2B), we intended to produce markerless gene inactivations to avoid possible polar effects. Specifically, our goal was to introduce a stop codon into the coding sequences of fbiC and cofC, creating two individual M. rhizoxinica mutant strains. We deemed cytidine base editing a promising method to introduce stop codons [36].

To generate base editing plasmids, a temperature-sensitive, low-copy plasmid containing a rhamnose-inducible and codon-optimized cas9^* gene was used [32]. We added a gene that encodes a fusion protein consisting of APOBEC1 (a cytidine deaminase from rats), nCas9 (Cas9-nickase with a D10A mutation from Streptococcus pyogenes), and a uracil DNA glycosylase inhibitor (UGI) from a Bacillus phage (Fig. 3A).

The base editing plasmids pTsK-AnCU-fbiC and pTsK-AnCU-cofC (Supplementary Table 4) differed only in the synthetic guide RNA (sgRNA) sequences in which the N20 sequences determine the individual target site. Both plasmids were used for precisely targeting fbiC and cofC in the M. rhizoxinica genome to generate the mutants M. rhizoxinica ΔfbiC::stop (M. rhizoxinica ΔfbiC) and M. rhizoxinica ΔcofC::stop (M. rhizoxinica ΔcofC), respectively. Successful base editing in potential mutants was verified via Sanger sequencing (Fig. 3B). Whole genome sequencing of M. rhizoxinica ΔfbiC and M. rhizoxinica ΔcofC revealed no deleterious mutations in either strain, thus confirming the specificity of the chosen sgRNA. CRISPR/Cas-mediated base editing was previously developed for Gammaproteobacteria (Pseudomonas putida) [37] and Alphaproteobacteria (Brucella melitensis) [38]. Here, we established this genome editing strategy for Betaproteobacteria (M. rhizoxinica).

As expected, metabolic profiling of the mutant strains using LC–MS/MS confirmed that F_O and 3PG-F₄₂₀ biosynthesis is abolished in M. rhizoxinica ΔfbiC, whereas M. rhizoxinica ΔcofC is able to produce F_O but no 3PG-F₄₂₀-2 (Fig. 3C). In comparison, M. rhizoxinica wild type produces F_O and 3PG-F₄₂₀-2 in axenic culture under the same conditions. Axenically grown M. rhizoxinica ΔfbiC displays growth behavior similar to the wild type when cultivated on a selection of carbon sources in liquid media (Supplementary Fig. 4). A broader screen for altered utilization of carbon sources using phenotypic microarrays did not reveal any metabolic deficit of F_O-deficient M. rhizoxinica (Supplementary Table 7).

F_O-deficient M. rhizoxinica abolish the sporulation ability of R. microsporus

The ability of R. microsporus to reproduce asexually via sporulation strictly depends on the presence of endobacteria as well as secondary metabolites produced by M. rhizoxinica [7, 10, 26]. Thus, considering that F_O and 3PG-F₄₂₀ are predominantly produced under symbiotic conditions, we aimed to explore the potential involvement of 3PG-F₄₂₀ and/or its precursor F_O in fungal sporulation. The sporulation ability of R. microsporus containing either M. rhizoxinica ΔfbiC or M. rhizoxinica ΔcofC was investigated in a sporulation bioassay [24]. Briefly, apo-symbiotic R. microsporus mycelium is co-cultured with axenic M. rhizoxinica wild type or mutant strains in 48-well plates. After 5–7 days of incubation, the formation of spores can be observed indicating the successful establishment of the symbiosis [7]. Apo-symbiotic R. microsporus does not sporulate (Fig. 4A). Subsequently, the absence/presence of F_O and 3PG-F₄₂₀ in co-cultures was verified via LC–MS/MS analysis.

M. rhizoxinica ΔcofC is able to produce F_O, but not 3PG-F₄₂₀, and readily triggers fungal sporulation to a similar degree as the wild type. (Fig. 4A and B and Supplementary Table 8). In contrast, in the case of co-cultivation with M. rhizoxinica ΔfbiC, spores are absent and the sporulation efficiency is significantly reduced (P < .05) when compared to the M. rhizoxinica wild type (Fig. 4A and Supplementary Table 8). As expected, F_O and 3PG-F₄₂₀ are absent in R. microsporus–M. rhizoxinica ΔfbiC co-cultures (Fig. 4B). Thus, the inability to sporulate appears to be linked to a lack of F_O but not 3PG-F₄₂₀. In addition, growth of the fungal mycelium containing M. rhizoxinica ΔfbiC is slower and fewer aerial hyphae are formed, which is comparable to apo-symbiotic fungal mycelium (Fig. 4A). The observation that F_O-deficiency completely abolishes R. microsporus sporulation is unexpected, as a complete abrogation of fungal sporulation is rarely observed [24].

To confirm that the inability of M. rhizoxinica ΔfbiC to induce fungal sporulation is solely due to disruption of the fbiC gene, we performed an in vivo trans-complementation experiment. To this end, we transformed M. rhizoxinica ΔfbiC and M. rhizoxinica ΔcofC with the plasmids pRANGER-fbiC and pRANGER-cofC, respectively. These expression vectors carry the native promoter controlling expression of each gene, yielding the complemented strains M. rhizoxinica ΔfbiC pRANGER-fbiC and M. rhizoxinica ΔcofC pRANGER-cofC (Supplementary Fig. 5). Recolonization of apo-symbiotic R. microsporus by ΔcofC pRANGER-cofC leads to a level of fungal sporulation similar to that imparted by the wild type (Fig. 4A). In addition, co-incubation of R. microsporus with ΔfbiC pRANGER-fbiC readily triggers sporulation, thus restoring the wild-type phenotype (Fig. 4A). As expected, both complemented strains are able to produce F_O and 3PG-F₄₂₀-2 in co-culture with R. microsporus (Fig. 4B). These results show that a lack of 3PG-F₄₂₀ alone does not affect the ability of R. microsporus to sporulate. However, when 3PG-F₄₂₀-deficiency is coupled with a lack of F_O, fungal sporulation is abolished, signifying that F_O is an essential mediator in the sporulation process of endosymbiont-dependent R. microsporus.

Chemical complementation of F_O-deficient M. rhizoxinica restores sporulation in R. microsporus

The observation that F_O-deficiency completely abolishes fungal sporulation raised the question as to whether F_O alone can trigger R. microsporus sporulation or whether the presence of bacteria is necessary for this process. To answer this question, chemical complementation with synthetic F_O was performed. Apo-symbiotic R. microsporus was co-incubated with M. rhizoxinica ΔfbiC in a sporulation assay as described above. In addition, synthetic F_O was added to individual wells containing apo-symbiotic R. microsporus and M. rhizoxinica ΔfbiC at varying concentrations (312 ng/ml, 625 ng/ml, or 937 ng/ml). When synthetic F_O is added to non-sporulating R. microsporus–M. rhizoxinica ΔfbiC co-cultures, spore formation is restored (Fig. 5A). Although the number of spores is lower than induced by the wild type, the sporulation in chemically complemented co-cultures is significantly increased (P < .002) when compared to M. rhizoxinica ΔfbiC (Fig. 5B and Supplementary Table 9). In fact, we observed a dose-dependent increase in the number of spores with increasing amounts of F_O supplementation (Fig. 5B and Supplementary Table 9).

As these results clearly confirmed the importance of F_O in fungal sporulation, we were curious as to whether F_O itself could induce sporulation in R. microsporus in the absence of endobacteria. Apo-symbiotic R. microsporus was grown in 48-well plates in either liquid or solid potato dextrose medium containing a final concentration of 73 μg/ml synthetic F_O. Spore formation was not detected, even after 2 weeks of incubation (Supplementary Fig. 6). Together with the successful chemical complementation of M. rhizoxinica ΔfbiC, these results indicate that F_O alone is not sufficient to trigger sporulation of the host fungus and that other bacterial traits are involved in this process.

F_O-deficient M. rhizoxinica are able to recolonize R. microsporus hyphae efficiently

A lack of sporulation in F_O-deficient R. microsporus–M. rhizoxinica co-cultures could be caused by the inability of M. rhizoxinica ΔfbiC to recolonize fungal hyphae as has been previously reported for M. rhizoxinica that lack a functional type 2 secretion system [25]. To investigate whether M. rhizoxinica ΔfbiC can enter fungal hyphae, we generated M. rhizoxinica strains that express GFP constitutively. Apo-symbiotic R. microsporus was co-incubated with GFP-expressing M. rhizoxinica wild type, M. rhizoxinica ΔfbiC, or M. rhizoxinica ΔcofC. After 1 week of co-cultivation, superresolution fluorescence microscopy was performed to visualize individual bacterial cells and fungal hyphae. As expected, wild-type M. rhizoxinica is located inside fungal hyphae. M. rhizoxinica ΔfbiC and M. rhizoxinica ΔcofC are also clearly visible inside R. microsporus hyphae confirming successful fungal recolonization by both of these strains (Fig. 6A). Fungal mycelium recolonized by M. rhizoxinica ΔfbiC displays a different morphology, both at the macroscopic and microscopic scale, when compared to that recolonized by M. rhizoxinica wild type and M. rhizoxinica ΔcofC (Figs 4A and 6A). When grown on agar plates, the mycelium containing M. rhizoxinica ΔfbiC grows flat and patchy in contrast to the fluffy, aerial mycelium of the wild type and M. rhizoxinica ΔcofC. A combination of widefield and fluorescence microscopy revealed hyphae with a larger diameter (Fig. 6A) containing a significantly higher number of bacterial cells (P < .0001) than hyphae containing M. rhizoxinica wild type or M. rhizoxinica ΔcofC (Fig. 6B and Supplementary Table 10).

The high bacterial load of M. rhizoxinica ΔfbiC in fungal hyphae is reversible through the addition of synthetic F_O. By adding increasing concentrations of synthetic F_O (312 ng/ml, 625 ng/ml, or 937 ng/ml) to apo-symbiotic R. microsporus–M. rhizoxinica ΔfbiC co-cultures, the bacterial load of M. rhizoxinica ΔfbiC inside the fungal hyphae is comparable to the wild type in all cases (Fig. 6A and Fig. 6B). Quantification of M. rhizoxinica ΔfbiC pRANGER-fbiC (complemented M. rhizoxinica ΔfbiC) reveals significantly fewer countable bacterial cells inside fungal hyphae compared to M. rhizoxinica ΔfbiC (P <.0001, Fig. 6B and Supplementary Table 10). The ability of F_O-deficient M. rhizoxinica to efficiently recolonize their fungal host combined with the abolishment of fungal sporulation shows that F_O is not essential for host colonization but rather supports a process that triggers host sporulation in this bacterial–fungal symbiosis.

Discussion

The specialized redox cofactor F₄₂₀ is instrumental for diverse biochemical processes in archaea and Actinobacteria, e.g. biosynthesis of antibiotics [39], methanogenesis [17], or the degradation of chemical pollutants [40]. In recent years, F₄₂₀ has gained considerable interest due to its low redox potential, which allows for F₄₂₀-dependent enzymes to be utilized in biotechnological applications, such as bioremediation and biosynthetic production of antibiotics [41–44]. This has invigorated a search for new F₄₂₀-dependent enzymes leading to the discovery of F₄₂₀ biosynthetic gene clusters in a wide range of Gram-negative bacteria. We have previously shown that the Gram-negative bacterium M. rhizoxinica HKI-454 (Class: Betaproteobacteria) is able to produce an unusual derivative of F₄₂₀ (3PG-F₄₂₀) [12]. However, virtually nothing was known about the role of the products of the F₄₂₀ pathway in M. rhizoxinica and in Gram-negative bacteria in general.

Here, we show that the genetic inventory to produce cofactor 3PG-F₄₂₀, although rather rare in related bacteria, is conserved in the genomes of endofungal Mycetohabitans strains that are known to form a stable symbiosis with their corresponding R. microsporus hosts [27]. Gene expression analysis and metabolic profiling showed that biosynthesis of 3PG-F₄₂₀ and its precursor F_O are particularly pronounced during symbiotic growth. By establishing a CRISPR-assisted cytidine base editing strategy for M. rhizoxinica, we were able to generate M. rhizoxinica strains deficient of F_O and 3PG-F₄₂₀ (M. rhizoxinica ΔfbiC) and deficient of only 3PG-F₄₂₀ (M. rhizoxinica ΔcofC). Our CRISPR-based, markerless gene inactivation method represents an improvement over previous methods based on homologous recombination and counter-selectable markers [24] as it avoids possible polar effects. Co-culture experiments unequivocally demonstrate that sporulation of the host is only triggered by bacteria that produce F_O or when supplemented with F_O. M. rhizoxinica ΔcofC, which produces F_O but not 3PG-F₄₂₀, does not show any obvious effect on sporulation (Fig. 7). These results demonstrate that F_O is a crucial symbiosis factor that controls fungal reproduction.

To probe whether F_O is required by M. rhizoxinica itself, we tested if F_O deficiency negatively impacts bacterial growth in vitro but did not observe any growth defects of M. rhizoxinica ΔfbiC, at least in axenic culture. In vivo, we even observed an increased bacterial cell density of M. rhizoxinica ΔfbiC in R. microsporus hyphae. Thus, it is unlikely that F_O deficiency causes growth impairments or insufficient protective responses under symbiotic conditions, as such traits would result in poor host colonization and reduced intracellular survival of M. rhizoxinica [23]. In fact, an increase in M. rhizoxinica ΔfbiC cell density following fungal recolonization mirrors an unusual phenomenon recently reported in R. microsporus reinfected with M. rhizoxinica lacking Mycetohabitans transcription activator-like effector 1 (MTAL1) [23]. Based on microfluidics and fluorescence microscopy, it was shown that MTAL1-deficient M. rhizoxinica are trapped in hyphae through septa formation leading to high cell numbers. Although we did not observe septa formation and subsequent trapping of M. rhizoxinica in this study, it is of course possible that F_O deficiency causes a similar response by the fungus. Alternatively, F_O may regulate bacterial cell numbers in vivo as a type of quorum sensing molecule. The process of how bacterial numbers are controlled inside hyphae is still unknown as M. rhizoxinica does not produce any of the “classical” quorum sensing molecules [11]. Another possible explanation for the high bacterial load observed in this study is an accumulation of bacterial cells due to the lack of sporulation because wild-type bacteria would normally be packed into spores [7]. In line with this hypothesis, R. microsporus recolonized by M. rhizoxinica ΔcofC shows a normal bacterial load and produces a similar number of spores as the wild type. Because M. rhizoxinica supplies its fungal host with selected amino acids [11], it is conceivable that a bacterial nutrient that is abundant during vegetative growth is consumed by the fungus once the sporulation starts, thus becoming a growth limiting factor. The higher cell density might consequently not be directly related to altered metabolic capacities of the bacteria but rather to an altered metabolism of the fungus.

As symbionts often deliver essential metabolites like vitamins or amino acids to their host [45, 46], F_O might be secreted into the host cytosol and subsequently utilized by the fungus. Indeed, F_O is present in the supernatant of M. rhizoxinica [12, 47] and might, therefore, act as a secreted “vitamin” or “signal” that triggers host sporulation. In support of this hypothesis, M. rhizoxinica is not actively engulfed by R. microsporus during entry into hyphae and is thus lacking a fungal cell membrane around the bacteria [25]. In addition, F_O represents the deazaflavin “head” moiety of 3PG-F₄₂₀, which is redox active [48] and capable of harvesting light [13]. The sporulation-triggering activity of F_O could involve its photochemical properties [13]. The light-harvesting properties of F_O are exploited by some DNA photolyases, which are highly efficient DNA repair enzymes [49]. Because DNA photolyases have a high binding-specificity to DNA, one could imagine a possible activation or repression of fungal genes in a F_O-dependent manner. However, future studies will have to rely on biochemical methods such as chemical proteomics to reveal binding partners of F_O within the endofungal proteome.

Intrigued by the complete abrogation of fungal sporulation in R. microsporus recolonized by M. rhizoxinica ΔfbiC, we investigated whether F_O alone can trigger R. microsporus sporulation and observed that the presence of M. rhizoxinica is a prerequisite for host sporulation. This observation is in line with previous studies suggesting that control of fungal sporulation involves a multitude of bacterially produced symbiosis factors such as transcription activator-like effectors [9], the secondary metabolite habitasporin [10], or a specialized lipopeptide O-antigen [8]. It was shown that a lack of these molecules reduces the sporulation ability of R. microsporus when recolonized by M. rhizoxinica. Here, we report complete abrogation of sporulation after host colonization by M. rhizoxinica ΔfbiC. Such a complete absence of sporulation was previously only reported for M. rhizoxinica strains unable to recolonize the apo-symbiotic host due to the absence of intact type 2 [25] or type 3 secretion systems [24]. Our results demonstrate that F_O, the precursor of 3PG-F₄₂₀, is yet another symbiosis factor that is instrumental for the maintenance of the R. microsporus–M. rhizoxinica alliance.

The purpose of 3PG-F₄₂₀ production in M. rhizoxinica remains mysterious, as M. rhizoxinica ΔcofC, which produces F_O but not 3PG-F₄₂₀, does not show any obvious effect on fungal sporulation or recolonization. It has been shown that 3PG-F₄₂₀ is typically concentrated in bacterial cell pellets [12, 47], indicating that 3PG-F₄₂₀ may still play as-yet-unknown physiological roles in Mycetohabitans species. It is, however, questionable whether 3PG-F₄₂₀ actually acts as a cofactor in Mycetohabitans species, as bioinformatics-based evidence for the presence of genes encoding putative F₄₂₀-dependent oxidoreductases in M. rhizoxinica could not be obtained [12]. This finding speaks against a cofactor role, because F₄₂₀-dependent enzymes typically belong to well-studied enzyme families and at least two such enzymes would be necessary to drive F₄₂₀-dependent redox biochemistry [20, 21]. It is, however, still possible that uncharacterized enzymes catalyze the 3PG-F₄₂₀-dependent reactions. Alternatively, the lack of F₄₂₀-dependent oxidoreductases could potentially be explained by 3PG-F₄₂₀ fulfilling a non-cofactor role in M. rhizoxinica, as has been reported for the cofactor nicotinamide adenine dinucleotide [50, 51] and for the F₄₂₀-derivative factor F₃₉₀ in Methanothermobacter thermautotrophicus [52–54].

In this study, we combined targeted gene inactivation, metabolic profiling, colonization assays, and microscopy to functionally characterize deazaflavins of the endofungal symbiont M. rhizoxinica. We found that F_O, the redox-active moiety of F₄₂₀, is necessary for sporulation of the fungal host, signifying that F_O is an essential symbiosis factor for the Rhizopus–Mycetohabitans symbiosis (Fig. 7). This study represents an unprecedented case of a deazaflavin metabolite controlling the reproduction of a eukaryotic host. It seems plausible that F_O acts as a mediator of fungal sporulation, but the molecular details of how this non-cofactor role is implemented remain to be deciphered. It should be mentioned that addition of the tail moiety to F_O, to form the ultimate product of the biosynthetic pathway, may direct 3PG-F₄₂₀ toward possibly acting as a classical cofactor within M. rhizoxinica. Although the role of atypical metabolites in symbioses and microbial communities in general has been hitherto overlooked, this study showcases the remarkable flexibility of biochemical molecules derived from a relatively simple pathway. Our results should inspire future research avenues into the functional elucidation of F₄₂₀ pathway products involved in cellular communication and symbiosis.

Acknowledgements

We thank the Microverse Imaging Centre (Aurélie Jost and Patrick Then) for providing microscope facility support for data acquisition. We are grateful to Evelyn M. Molloy for insightful discussions and invaluable assistance in proofreading this manuscript.

Author contributions

Ingrid Richter (conceived the idea, developed the study design, preformed experiments, interpreted the data, and wrote the manuscript), Mahmudul Hasan (designed and generated complementation plasmids, performed LC–MS measurements, growth curve analysis, and Biolog assays), Johannes W. Kramer (performed sporulation assays, microscopy and image quantification, and assisted with RNA extractions and qPCR experiments), Philipp Wein (generated CRSIPR/Cas mutant strains), Jana Krabbe (designed and generated GFP expression plasmids and CRSIPR/Cas base editing plasmids), K. Philip Wojtas (generated synthetic F_O), Timothy P. Stinear (assisted in data interpretation and manuscript revision), Sacha J. Pidot (performed whole genome sequencing, variant calling, and revised the manuscript), Florian Kloss (assisted in data interpretation and manuscript revision), Christian Hertweck (assisted in data interpretation and manuscript revision), and Gerald Lackner (conceived the idea, drafted, and revised the manuscript)

Conflict of interest

The authors declare no competing interests.

Funding

This work was funded by the European Union’s Horizon 2020 Research and Innovation Program under the Marie Skłodowska-Curie grant agreement No. 794343 (to I.R.), the Free State of Thuringia with grant number 2019 FGI 0003, and the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) under Germany’s Excellence Strategy—EXC 2051—(Cluster of Excellence “Balance of the Microverse”) Project-ID 390713860 and the SFB 1127 ChemBioSys, Project-ID 239748522, and the Leibniz Award (to C.H.). Part of this work was funded by the DFG under the project number 408113938 (to G.L).

Data availability

All data is included in this manuscript.

References

Author notes