Inducing novel endosymbioses by implanting bacteria in fungi

Strains and standard culturing methods

The strains and chemicals used are listed in Supplementary Tables 1 and 2. E. coli K-12 BW25113 pDGUV-GFP⁵⁰ was cultured in Luria–Bertani broth (Sigma-Aldrich) supplemented with 100 µg ml^−1 carbenicillin (Roth). R. microsporus strain NH is R. microsporus CBS 631.82; R. microsporus strain EH is R. microsporus ATCC62417. M. rhizoxinica HKI-0454 and R. microsporus were cultured as previously reported⁴² at 28 °C. M. rhizoxinica was fluorescently labelled with cytosolic GFP using pBBR-P12-GFP⁵¹. R. microsporus was grown on potato dextrose agar (ThermoFisher) at 28 °C with 20 µg ml^−1 gentamicin. Respective antibiotics for plasmid retention were used as required for the culture of bacteria in all experiments, including the growth of bacteria inside the fungus, in which case, additionally, gentamicin was used to prevent extracellular bacterial growth.

Bacterial injections with FluidFM

The basic setup of the instruments, FluidFM probe processing steps, probe cleaning and probe coating were as previously described³⁶, but the pressure was controlled with a FlowEZ 7000 (Fluigent), in the range of 0–7,000 mbar, using an AT550-9L compressor (WenLing) as pressure source. The probe apex was sharpened with a Helios 5UX DualBeam focused ion beam scanning electron microscope (ThermoFisher) as described previously^36,52, but to a new shape that resulted in a sharp double point, as shown in Fig. 1, by targeting the probe from the front and shaping a centred point with a 60° central angle. The cantilevers of the used probes had a nominal stiffness of 1.6 N m^−1. The fungal sample was prepared by seeding spores in 50-mm WillCo glass-bottom dishes (WillCo Well) and adding 4 ml potato dextrose broth (PDB) medium (ThermoFisher) containing 34 µg ml^−1 chloramphenicol, incubating at room temperature for 14–16 h, washing 1–3 times with 4 ml PDB, depending on germling concentration, exchanging the medium for protoplasting mix (to soften but not completely disintegrate the cell wall; 1.6 g Cellulase Onozuka R10 (Duchefa Biochemie), 40 mg chitinase (Merck), 40 ml iced MMB (0.5 M mannitol, 0.05 M maleate pH 5.5), filtered through a 0.22-µm syringe filter), and incubating for another 3–5 h at room temperature. If germlings were too dense, the sample supernatant was filtered through a 0.22-µm syringe filter. The E. coli sample was grown in a 10-ml culture in a baffled shake flask overnight at 37 °C, washed three times in Hepes2 buffer (10 mM HEPES, 150 mM NaCl, pH 7.4), and adjusted to an optical density of 2 at 600 nm. The M. rhizoxinica sample was grown in a 2-ml culture in a 12-ml culturing tube for 3 to 5 days at 28 °C and prepared like the E. coli culture. Bacterial suspensions (15 µl) were pipetted into the reservoir of the FluidFM probe. The probe was moved towards the glass surface (z = 0 µm) beside the targeted germling and then retracted to a z-value of +8 to +10 µm. The injection site of the germling was chosen on the basis of direct proximity to the glass surface, preferring thick germtube portions. As force-controlled puncturing was not possible owing to the softness of the probe, the germling was punctured by advancing the probe by 10 µm (corresponding to a nominal z-value of +1 µm to −2 µm), holding this position for 5 s and retracting to a z-value of +1 to +3 µm. Successful puncturing of the germling could be monitored with wide-field illumination by observing cytoplasmic flow. Immediately following puncture of the germling, the microfluidic system was pressurized with 3–4 bar, stopping turgor-induced backflow into the probe. Subsequently, the pressure was increased to 6.5 bar until liquid flow into the germling was noticed, and then quickly reduced to 3–4 bar to prevent bursting of the germling. Injection of bacteria was confirmed by switching to the fluorescence channel, and then the pressure was slowly reduced to 0 mbar to allow recovery of the germling. The injected germling was isolated to a fresh dish filled with recovery medium (3.8 ml MMB, 1 ml PDB, 160 µl 4 M sorbitol) after 3–10 min, or after the germling recommenced growth. Isolation was carried out by lifting the FluidFM and exchanging the sample dish for the recovery dish, with the germling sticking to the probe. Recovery of the germling and the ensuing dynamics were visualized using time-lapse and z-stack images in the wide-field and the fluorescence channels. Some of the injections were carried out with the addition of 1:1,000 calcofluor white (Merck) in the protoplasting mix and recovery medium to visualize the cell wall. Germlings were grown in recovery medium overnight and then detached from the probe using overpressure, lifting of the FluidFM and scraping with a plastic pipette tip. Mycelium was then transferred to a potato dextrose agar plate and incubated at 28 °C.

Spore collection

Spores were collected 6 ± 1 days after injection or plating from spores. Spore solution (8.5% NaCl, 1% Tween 20) was added to plates (16 ml for square plates; 12 ml for round plates), and the spores were thoroughly detached using a spatula. The remaining mycelium was clumped up and gently pressed with the spatula to release the spore solution from the mycelium. The spore solution was filtered through a 10-µm CellTrics filter (Sysmex). Spores were washed three times with 1 ml Hepes2 (relative centrifugal force of 8,000; 2 min) and stored at 4 °C overnight for FACS, or in 50% glycerol at −20 °C (for working stocks) or −80 °C (for long-term storage).

Flow cytometry and cell sorting

Analysis and sorting of spores were conducted at the ETH Flow Cytometry Core Facility on a FACSAria Fusion BSL2 cell sorter (BD). Single spores were selected using SSC-A–FSC-A and FSC-H–FSC-A gates. Colonization by bacteria was checked using an SSC-A–eGFP-A gate. (See Supplementary Fig. 1 for the gating strategy). For samples for which autofluorescence was suspected to complicate the positioning of the positive gate, an mCherry-A versus PerCP–Cy5-5-A channel was used to check for autofluorescence but was not included in the gating. For determination of the fraction of positive spores, 100,000 to 1,000,000 spores were analysed depending on the size of the fraction. For bulk sorting, spores were sorted into 1.5-ml Eppendorf tubes. For determination of germination success, single spores were sorted into 96-well plates containing 125 µl PDB + 34 µg ml^−1 chloramphenicol per well. For verification of positive gates, bulk-sorted spores were intermittently checked under the microscope. Collected data were analysed using FlowJo v10 software (BD). For sorting of high, medium and low fractions of positive spores, gates were set qualitatively in the positive population as illustrated in Fig. 4b.

Determination of germination success

Single-spore-sorted 96-well plates were incubated at 28 °C and checked visually with a Zeiss SteREO Discovery.V8 microscope (Zeiss) for the appearance of a germling. Per sample, three plates of positive spores and one plate of negative spores were sorted. In round 1 of the adaptive laboratory evolution experiment, five positive plates were sorted per sample. Germlings were counted 1 day and 2 days after sorting, after which point there are no new germlings to be discovered. For positive plates, five germlings were checked microscopically on day 1 to confirm the presence of fluorescent bacteria. If a germling could not be confirmed to have endobacteria, five more germlings were checked. No samples failed this control by having more than 20% of germlings without easily detected endobacteria. To calculate the percentage of delayed germinations, the percentage of germlings not detected on day 1 but detected on day 2 was calculated (with 100% corresponding to the number of germlings detected on day 2).

Bacterial isolation

For isolation of M. rhizoxinica from R. microsporus, 50 ml MGYM9 + 34 µg ml^−1 chloramphenicol was inoculated with spores in a 500-ml baffled shake flask and incubated with shaking at 100 r.p.m. for 5–7 days. Build-up of mycelium on the wall was periodically flushed off by gentle shaking and tilting by hand. Once the medium became turbid, an inoculation loop was inserted in the medium, avoiding mycelial clumps and used for streaking out on agar plates. Plates were checked daily for 3 days and growing fungus was cut out if detected. Plates with bacterial colonies were further cultivated according to standard protocols and were used to make cryogenic stocks. Alternatively, a 2-µm syringe filter (Merck) was used to separate bacteria from mycelium and spores, and the filtrate was used for further cultivation in liquid.

Adaptive laboratory evolution experiment

For the evolution experiment, spores collected in the first round after injection were sorted for determination of germination success, and the remaining spores were bulk sorted for positive spores. The positive spores were split into 10 equal lines, resulting in approximately 300 spores per line. Thereafter, the ten lines were kept separate. The germination success and the fraction of positive spores were determined in every round for every line, and at every round after round 1, some positively and negatively sorted spores were cryopreserved, and bacteria were isolated from 10,000 positively sorted spores (approximately 300 for round 1). The product of the positive fraction and the germination success was calculated to give the fitness index, indicating the percentage of spores that would yield bacteria-populated germlings for the start of a next round without selection. Plating for the evolution experiment was carried out on 120 × 120 mm square Petri dishes (Greiner) to increase the available surface area. Spores to be plated were taken up in 100 µl of buffer, which was pipetted into five parallel lines with equal spacing on the plate. Standard spread plating for high spore numbers had previously proved to lead to inconsistent spore formation. From round 2 on, high densities of spores could be plated for the next round as the fraction of positive spores increased. For making round 3 plates, the highest number of spores was seeded, up to 830,000 spores per plate, whereas for rounds 4 and 5 the numbers varied slightly around 100,000, and from thereon 100,000 were used for all rounds and lines. The number of seeded spores is shown in the Source Data for Fig. 3a. After round 7, only the three best-performing lines according to the fitness index were grown individually (lines 2, 4 and 7), whereas the other seven lines were pooled to line P by mixing equal amounts of positive spores. A scheme indicating the population sizes of spores and the pooling regime can be found in Fig. 3a.

Detection of rhizoxin

A total of 10,000 positively sorted spores per line from round 7 were plated. Strain EH colonized by M. rhizoxinica, the axenic strain NH and a liquid culture of M. rhizoxinica served as controls. Plates were grown for 11 days and extracted with 50 ml ethyl acetate with shaking at 100 r.p.m. overnight at 28 °C. The organic phase was separated, dried with sodium sulfate, filtered through paper filter, and evaporated with a rotary vacuum evaporator. Samples were taken up in 1 ml acetonitrile, centrifuged at 20,000g for 10 min and 800 µl of the supernatant was stored at −80 °C until liquid chromatography with tandem mass spectrometry (LC–MS/MS) analysis. LC separation was carried out with a Thermo Ultimate 3000 UHPLC system (Thermo Scientific) using a C18 reversed-phase column (Kinetex XB-C18 column, particle size 1.7 µm, pore size 100 Å; dimensions 50 mm × 2.1 mm, Phenomenex). Solvent A was 0.1% (v/v) formic acid in water and solvent B was 0.1% formic acid in acetonitrile at a flow rate of 500 µl min^−1. Solvent B was varied as follows: 0 min, 25%; 3 min, 90%; 5 min, 90%; 5.3 min, 25%; subsequently, the column was equilibrated for 2 min at the initial condition. The injection volume was 2 µl.

MS-product reaction monitoring analysis was carried out with a Thermo QExactive plus instrument (Thermo Fisher Scientific) in the positive Fourier transform mass spectrometry mode. MS level 1 scans were carried out with a mass resolution of 35,000 (m/z = 200) and MS level 2 scans were carried out with a mass resolution of 17,500. Parent ions were isolated at m/z 594.34, 610.337, 612.3531, 626.3323 and 628.348 with unit resolution and fragmented by high-energy C-trap collision dissociation applying a normalized collision energy of 28 eV. A heated electrospray ionization probe was used with the following source parameters: vaporizer temperature, 380 °C; sheath gas, 50; auxiliary gas, 20; sweep gas, 0; RF level, 50.0; capillary temperature, 275 °C. See Supplementary Fig. 2 for relevant spectra.

Genomics

For the generation of sequencing samples of M. rhizoxinica, bacteria were grown according to standard culturing conditions after isolation from the respective time point (Methods, Bacterial isolation), and 4 ml of sample adjusted to an optical density of 1 at 600 nm were pelleted by spinning at a relative centrifugal force of 11,000 for 1 min. Genomic DNA was prepared using the MasterPure DNA Purification Kit (LGC). Genomic DNA was sent on dry ice to BMKGene (Biomarker Technologies) for further processing.

For generation of sequencing samples of R. microsporus, mycelium was grown in 500 ml malt extract broth (Thermo Fisher Scientific) in 2-l shake flasks for Illumina sequencing or in 1.5-l malt extract broth in 5-l shake flasks for PacBio sequencing at 37 °C for 5 days, with the addition of gentamycin and chloramphenicol. The mycelium was then filtered on a 110-mm filter paper and washed thoroughly with double-distilled H₂O and for Illumina samples additionally with 150 ml of 70% ethanol. The mycelium was then removed from the filter paper, packed into a 50- ml screw cap tube and frozen in liquid nitrogen. The samples were then sent to BMKGene (Biomarker Technologies) for further processing.

For genome assembly of R. microsporus CBS 631.82, Pacbio HiFi sequencing with RS II produced a total of 3,867,257,442 base pairs (bp) in 353,300 reads of average length 10.9 kilobases and average quality score 30.3. The reads were assembled with Flye (v2.9.2)⁵³ with the –pacbio-hifi flag, resulting in 118 contigs of total length 55,743,399 bp with an N50 (the shortest contig of the set of the largest contigs making up 50% of the assembly) of 1,370,944 bp. BUSCO (v5.4.7)⁵⁴ was used to check the quality of the assembly, using the lineage dataset mucorales_odb10, giving the following result: C (complete): 97.5%, S (single-copy): 5.1%, D (duplicated): 92.4%, F (fragmented): 1.7%, M (missing): 0.8%, n (number of genes): 2,449. Thus, although nearly complete, the genome also seems to be largely duplicated. The assembly was gene-called with BRAKER (v3.0.6)^{55,56,57,58,59,60,61,62}, using the —fungus flag, and then functionally annotated with eggNOG-mapper (v2.1.12)⁶³ using the option –target_taxa Fungi. BUSCO reported a slightly improved completeness for the called genes: C: 99.8% [S: 1.4%, D: 98.4%], F: 0.1%, M: 0.1%, n: 2,449.

For calling mutations from the adaptive evolution experiment for R. microsporus and M. rhizoxinica, the short-read Illumina sequences provided by BMK were used. The resulting raw reads were cleaned by removing adaptor sequences, low-quality-end trimming and removal of low-quality reads using BBTools v38.18 (https://sourceforge.net/projects/bbmap/). The commands used for quality control can be found on the Methods in Microbiomics web page (https://methods-in-microbiomics.readthedocs.io/en/latest/preprocessing/preprocessing.html). Single nucleotide polymorphisms were called using two different tools—Snippy and bcftools⁶⁴. For variant-calling with bcftools, reads were first aligned to the PacBio assembly of R. microsporus CBS 631.82 or to the M. rhizoxinica reference genome (GCF_000198775.1) using BWA-MEM v0.7 (ref. ⁶⁵). Duplicate reads were marked and removed using GATK4 v4.2 (MarkDuplicates)⁶⁶. Variants were called using the bcftools call command. Single nucleotide polymorphism calls were validated by comparing results obtained by two independent tools. All the variants that were detected in the ancestral sample were filtered out from all of the evolved samples using bcftools-isec to investigate variants that arose during the evolution experiment. The resulting VCF files were filtered using bcftools with the following criteria: -Ov -sLowQual -g5 -G10 -e ‘QUAL < 200 || DP4[2] < 3 || DP4[3] < 3 || (DP4[2] + DP4[3])/sum(DP4) < 0.1 || MQ < 50’. Each variant call produced by each of the tools was manually checked by analysing the read alignments at variant positions using samtools pileup. Mutations were annotated using SnpEff⁶⁷ and InterProScan⁴⁵. STRING was used to identify putative biological processes by searching for interactions on the basis of the closest-related genes for fungi found in the STRING database⁴⁶.

Fitness without selection pressure

Spores for round 0 were first collected from a plate grown according to the standard conditions from the adaptive laboratory evolution experiment. For this, spores from a fresh F_Anc–B_Anc injection plate were bulk sorted and 100,000 positive spores were plated. For F_Evo–B_Anc, frozen positive spores from an injection plate were grown and then 100,000 positive spores were plated; for F_Evo–B_Evo, frozen spores from the evolution experiment from round 10 line 4 were grown and then 100,000 positive spores were plated. From these plates, spores of round 0 were collected. Then 100,000 spores were bulk sorted gating only for single spores but disregarding the GFP signal intensity of the spores and used for subsequent rounds. Every round, 1,000,000 spores per sample were analysed to determine the fraction of positive spores. The endosymbiosis was considered washed out once the positive fraction fell below the threshold of 1/100,000 spores, at which no positive spore is expected to be plated for the next round. The experiment was stopped after round 5. The theoretical trajectory of the positive fraction was calculated using the following formula:

where p is the positive fraction; x is the round for which the positive fraction is being determined; g is the germination success of positive spores as measured in round 0; e is the germination success of negative spores derived from the long-term average from the evolution experiment (69%); and p₀ is the positive fraction measured in round 0 to describe how many spores formed by bacteria-positive germlings are again bacteria positive.

Determination of bacterial load

Spores sorted for high, medium and low intensity were imaged in the wide-field and GFP channel to create z-stacks. Overlays of 10 spores per population per sample are shown in Extended Data Fig. 2a,b and served as visual confirmation for the correlation of GFP intensity as measured by FACS and bacterial population size. Contrast and brightness were kept the same. The volume of voxels considered to be bacteria was calculated using Matlab2018a (Mathworks) using sections of published code³⁶. To gauge the appropriate threshold, the three-dimensional rendering of the resulting voxel cloud was visually inspected, and the diameter of single particles was checked. An example of the resulting three-dimensional rendering of samples is shown in Extended Data Fig. 2c.

Spore longevity

Samples were grown as described in the section entitled Fitness without selection pressure. Single spores were sorted into 96-well plates with 75 µl of Hepes2 + 34 µg ml^−1 chloramphenicol in each well. Per sample, 15 plates with negative spores and 15 plates with positive spores were sorted. Plates were then incubated at 16 °C. One day before the nominal time point, three plates per condition were activated by adding 125 µl of PDB + 34 µg ml^−1 chloramphenicol to each well and incubating at 28 °C.

Statistical analyses

Statistical tests indicated in the figure legends were run with GraphPad Prism v9.0.0.

Images, videos, plots and figures

Images and videos were edited using Fiji⁶⁸ for contrast, z-stacks, time stamps, overlays and scale bars, as indicated in the respective sections of the Methods and figure legends. Videos were cut together and labelled using iMovie v10.4 (Apple). Plots were generated using GraphPad Prism 9. Figures were assembled and edited using Adobe Illustrator 2020. Illustrations for Fig. 1 were created with BioRender.com.

Reporting summary

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