Introduction

Members of the genus Burkholderia belong to the class β-Proteobacteria and are widely distributed in the environment. Burkholderia are particularly abundant in soil where they can be associated with a wide range of plants (Elliott et al., 2009; Carlier and Eberl, 2012), invertebrates (Kikuchi et al., 2005) and fungi (Warmink et al., 2009; Uroz et al., 2012; Scherlach et al., 2013). Since Burkholderia are mostly found in acidic soils (Stopnisek et al., 2014), interactions with fungi might be of particular relevance as fungi also favour acidic environments. From in vitro studies, we know that Burkholderia can form either antagonistic or mutualistic interactions with fungi. Antagonistic behaviour of Burkholderia species is well described and is largely due to the production of multiple antifungal compounds (Lewenza and Sokol, 2001; Partida-Martinez and Hertweck, 2007; Schmidt et al., 2009) that can be inhibitory to a wide range of phytopathogenic fungi (Quan et al., 2006; Kilani-Feki et al., 2011; Groenhagen et al., 2013). It has been reported that many environmental Burkholderia strains have beneficial effects on fungi, suggesting symbiotic and/or mutualistic interactions. Benefits from such interactions were mainly studied in a model system comprised of B. terrae and Lyophyllum sp., where it was shown that the bacteria (i) colonize the hyphae and use them for transportation and dispersal (Warmink et al., 2011, ii) survive better in acidic soils (Nazir et al., 2010a) and (iii) use several fungal exudates as nutrients (Warmink et al., 2009; Nazir et al., 2010b, 2013). Along these lines, Drigo et al. (2013) reported that Burkholderia strains were among the main consumers of carbon released from arbuscular mycorrhizal fungi in a rhizosphere community. In addition, it has been suggested that the presence of fungi is essential for colonization of sterile soils by Burkholderia (Nazir et al., 2010a; Warmink et al., 2011).

Despite these reports indicating that Burkholderia benefits from partnering with fungi, the following questions remain unanswered: (i) how widespread are such associations in soils, (ii) how specific are they, (iii) what is their molecular basis and (iv) what benefits do the bacteria receive from the interaction?

To answer these questions, we designed a comprehensive study in which co-occurrence network analysis, cultivation-based methods and proteomics were used. Co-occurrence network analysis was carried out on soils from a continental-scale study to determine the extent, the specificity and the distribution of such interactions in the soil. Model Burkholderia strains previously identified as major soil inhabitants (Stopnisek et al., 2014) were tested further for their ability to translocate and disperse with fungi. Finally, the widespread soil inhabitant B. glathei was grown alone or in the presence of either Fusarium solani or Alternaria alternata and the bacterial proteome was analysed under each situation to identify the molecular and physiological basis of the interaction.

Materials and methods

Co-occurrence analysis

We analysed the microbial communities of 266 soil samples from the Nutrient Network (NutNet) globally distributed experiment (Borer et al., 2014). Details of this data set and the methods used to characterize the bacterial and fungal communities are available in Prober et al. (2015). In total, 37 393 ITS fungal operational taxonomic units (OTUs) and 223 693 16S rRNA bacterial OTUs (with OTUs defined at >97% sequence similarity level for ITS1 and 16S rRNA reads for fungi and bacteria, respectively) were included to the analysis. Co-occurrence patterns between Burkholderia and fungal OTUs were tested using Spearman’s rank correlations between OTUs that occurred in at least 5% of the samples and had a rho of >0.3 and P-value<0.01 adjusted using the false discovery rate method (Barberan et al., 2012).

Monitoring the interactions between fungi and Burkholderia sp.

Fungi used in this experiment belonged to the species Alternaria alternata, Fusarium solani, Rhizoctonia solani and Lyophyllum sp. Karsten, while Burkholderia species applied were B. glathei, B. hospita, B. fungorum and B. caledonica (Table 1). To visualize the interaction with fungi under the fluorescent microscope, all Burkholderia strains were tagged with either the green fluorescent protein (GFP) or the dsRED protein using the electroporation protocol described by Choi et al. (2006). In short, overnight cultures of Burkholderia strains were washed twice in 0.3 M sucrose and subsequently electroporated with the plasmid pBBR1MCS-2-gfp mut3 Kmr or pin62: DsRed Cmr. Electroporated cells were transferred on Pseudomonas isolation agar and Luria-Bertani (LB) plates with the corresponding antibiotics (50 mg l−1). To follow the interactions, fungi were inoculated on water agar plates (15 g l−1) or LB agar at least 3 days before spotting the bacteria on the mycelium. For that overnight liquid Burkholderia cultures were prepared, washed three times in 0.9% NaCl, OD600 was adjusted to 0.1 and 4 drops of 2 μl were spotted on the fungal mycelium pre-grown for 3 days. Plates were incubated for up to 10 days in the dark and the growth of Burkholderia as well as interactions with fungi were followed daily under the fluorescence microscope (Leica M165 FC, Mannheim, Germany) and binocular (Leica DM6000 B, Mannheim, Germany). To monitor dispersal and attachment of Burkholderia strains on hyphae, a sterile iron ring was placed before pouring in the middle of the plate, which served as a physical barrier for bacterial cells. Both sides were filled up to 5 mm below the edge of the iron circle. Selected fungi and Burkholderia strains were inoculated in the inner side of the ring, so that only bacterial cells which could attach and/or disperse would be detected on the outer side of the ring. All experiments were performed in triplicates.

Table 1 Bacterial and fungal strains used in this study

Proteome analysis

For proteomic analyses, B. glathei LMG14190 and two fungi, A. alternata and F. solani, were used. Co-cultures with A. alternata (AB) and F. solani (FB) as well as the B. glathei alone (B) were grown in triplicates as described above with some modifications: as growth medium 1/3 Difco Malt Extract Agar (MEA) (BD, Franklin Lakes, NJ, USA) with 7 g agar l−1 medium was used and plates were supplemented with a cellophane membrane (Bio-Rad, Hercules, CA, USA; 165-0963-MSDS). After 10 days at room temperature, the biomass was collected and samples were treated as described by Carlier and Eberl (2012) with some modifications: to separate bacterial and fungal cells, mixed samples were homogenized in phosphate-buffered saline solution with a pestle, debris were pelleted at 1000 g for 3 min and the top layer was transferred to 50% Percoll (Sigma-Aldrich, St Louis, MO, USA). After ultracentrifugation at 15 000 g for 3 h, fractions were collected and analysed by phase-contrast light microscopy (Leica DM6000 B) and bacterial fractions were pooled and washed twice in phosphate-buffered saline solution. To extract proteins, a bead-beating protocol was used. Fractions were transferred into 2 ml tubes (Sarstedt, Nürnbrecht, Germany) with 0.25 ml of 0.1 mm glass beads (BioSpec Products, Bartlesville, OK, USA) and placed into FastPrep-24 lyser tubes (MP Biomedicals, Santa Ana, CA, USA). Eight rounds of 30 s lysis of 6.5 m s−1 followed by 5 min incubation on ice were used to extract proteins in concentrations from 2 to 8 mg ml−1. In all, 25 μg of cytosolic protein extract (Roti-Nanoquant, Roth, Karlsruhe, Germany) was resolved on 1D SDS-PAGE, lanes were cut in equidistant pieces and in-gel digested with trypsin as recently described (Grube et al., 2014). Peptide mixtures were separated by RP chromatography using an EASYnLC 1000 (Thermo Fisher Scientific, Waltham, MA, USA) with self-packed columns. Peptides were loaded and desalted on the separating column following resolution with a binary non-linear 76 min-gradient from 5% to 75% ACN in 0.1% acetic acid at a constant flow rate of 300 nl min−1. The liquid chromatography system was coupled online to an Orbitrap Elite equipped with a nanoelectrospray ion source (Thermo Fisher Scientific) performing mass spectrometry (MS) and MS/MS experiments of the eluted peptides. Survey scans were recorded in m/z range from 300 to 1700 with a resolution of 60 000 and with lock mass option enabled. MS/MS experiments were performed for the 20 most intensive precursor ions as determined in the survey scan excluding unassigned charge states or singly charged ions from the MS/MS experiments in the Linear ion trap. Proteins were identified by searching all MS/MS spectra against a database containing protein sequences of Burkholderia glathei LMG14190, Fusarium solani and Alternaria alternata (version 13-Dec-17) with added laboratory contaminants using Sorcerer-SEQUEST (Sequest v. 27 rev. 11, Thermo Scientific) including Scaffold_4.0.5 (Proteome Software, Portland, OR, USA). SEQUEST was searched with a parent ion tolerance of 10 ppm and a fragment ion mass tolerance of 1.00 Da. Up to two missed tryptic cleavages were allowed. Methionine oxidation (+15.99492 Da) was set as variable modification. For protein identification, a stringent SEQUEST filter for peptides was used (Xcorr versus charge state: 2,2, 3,3 and 3,8 for doubly, triply and quadruply charged peptides and deltaCn value greater than 0.10) and at least two peptides per protein were required. Proteins that contained similar peptides and could not be differentiated based on MS/MS analysis alone were grouped to satisfy the principles of parsimony. The protein false discovery rate, based on a decoy database, was 0.4%. Relative quantification was based on spectral counting of exclusive spectra using normalized spectral abundance factors (Zhang et al., 2010). Statistical analysis using MeV v4.8.1 (Saeed et al., 2003) was done for proteins that were present in at least two out of three biological replicates. Hierarchical clustering and Student’s t-test of z-transformed normalized exclusive spectra were performed with the following settings: unequal group variances were assumed (Welch approximation), P-values based on permutation (1000) with P=0.05, significance determined by standard bonferroni correction. Only proteins showing at least twofold changes in addition to statistical significance were considered for further analysis. So-called ‘off/on’ proteins needed to be detected in all replicates of one treatment and absent from all replicates of another treatment. Functional prediction and assignment of proteins to cluster of orthologous groups and TIGRFAMs (Haft et al., 2013), respectively, was accomplished by the in-house developed analysis pipeline ‘Prophane 2.0’ (http://www.prophane.de; Schneider et al., 2011). Voronoi treemaps were generated using Paver (Decodon, Greifswald, Germany, http://www.decodon.com/).

Burkholderia glathei LMG14190 genome sequencing

High molecular mass DNA was extracted following a standard protocol (Wilson 2001). Paired-end insert libraries were constructed at the FGCZ (Functional Genomics Center Zurich) using the Nextera XT kit (Illumina, San Diego, CA, USA). Sequencing of the 2 × 250 bp inserts was performed on the Illumina MiSeq platform at the FGCZ. The sequencing reads were assembled using the SPADES v2.5 software (Nurk et al., 2013). The resulting 138 contigs were examined and edited for misassemblies in GAP5 (Bonfield and Whitwham, 2010). Ribosomal RNA and tRNA genes were predicted using RNAmmer and tRNAscan (Schattner et al., 2005; Lagesen et al., 2007). The predicted RNA genes were masked using the maskfeat utility of the EMBOSS software package (Rice et al., 2000) before protein-coding gene finding using the Prodigal software (Hyatt et al., 2010). The annotated contigs, containing CDS and RNA gene predictions, were submitted to the RAST online service for functional annotation (Aziz et al., 2008). The annotations were curated using the Artemis software suite (Carver et al., 2008). The annotated sequences were deposited at the European Nucleotide Archive (accession numbers CCNS01000001–CCNS01000139).

Results and Discussion

Burkholderia co-occur with fungi in soils

The frequency of fungi co-occurring with Burkholderia was assessed with co-occurrence analysis from a trans-continental soil sample collection. The analysis supported our hypothesis that members of the genus Burkholderia significantly co-occur with multiple fungal taxa, suggesting broad and frequent associations between fungi and Burkholderia in soils (Figure 1, Supplementary Table S1). Apart from Burkholderia, other bacterial genera also showed high co-occurrence with fungi. Whether these patterns reflect an intimate interaction with the fungi or result from their shared preference for low pH or other environmental parameters remains to be investigated. For Acidobacteria and Verrucomicrobia, two phyla containing OTUs that showed high co-occurrence with fungi, low pH dependency has been demonstrated but to our knowledge there is no evidence for beneficial interactions between members of those phyla and fungi. In contrast, while the genus Burkholderia also shows strong preference for acidic soils (Stopnisek et al., 2014), mounting evidence suggests that many Burkholderia species establish direct associations with fungi and may often live directly on or inside fungal cells (Warmink et al., 2011; Nazir et al., 2012; Uroz et al., 2012). Thus, we speculate that the strong co-occurrence of Burkholderia and fungi indicates that the two organisms interact with each other in the soil, rather than just reflecting shared niche preferences.

Figure 1
figure 1

Co-occurrence of fungal (grey boxes) and Burkholderia (black boxes) OTUs from 266 soil samples. Box size represents the number of significant (rho>0.3, P-value<0.01 after false discovery rate (FDR)) associations for each OTU.

Burkholderia sp. attach, translocate and disperse on fungal hyphae

To test the specificity of the interactions and the ability of Burkholderia to attach, translocate or disperse on hyphae of different fungal species, we designed an in vitro experiment. Four Burkholderia strains that were previously identified as widespread soil inhabitants (Stopnisek et al., 2014), and four fungi that were among those taxa which most frequently co-occurred with Burkholderia were grown in dual cultures in Petri dishes containing an iron ring (Figure 2). The fungus and the bacteria were inoculated together inside the ring and only those bacteria able to attach to the hyphae were able to cross the ring to reach the outer compartment. This experiment was carried out both with water agar, on which bacteria alone could not grow, and with LB as growth media in the outer compartment. Interestingly, even when the rich LB medium was used, Burkholderia strains were mainly found in close vicinity to fungal hyphae, suggesting that the benefits provided by the fungi were not restricted to transportation, but included other goods that caused bacteria to stay on the hyphae rather than spread onto the rich medium. Our experiments showed that (i) for all four tested Burkholderia strains, growth on the hyphae was favoured over growth on the medium (Figure 2) and (ii) all tested Burkholderia strains were able to cross the iron ring and be translocated with the hyphae of the fungi tested, with the exception of Lyophyllum sp. Karsten. Under the conditions used, the mycelium of Lyophyllum showed hydrophobicity, thus inhibiting the formation of a water film around the hyphae, which might be necessary for bacterial attachment and dispersal (Pion et al., 2013). However, Warmink et al. (2011) showed that B. terrae BS001 could migrate together with Lyophyllum sp. Karsten in soil microcosms, suggesting that the physiology of this fungus may change when growing in soil. Our results are consistent with the results reported by Nazir et al. (2012), who found that 19 Burkholderia strains were able to co-migrate with Lyophyllum sp. Karsten, but their abilities differed depending on the soils used. No visible deleterious effects of the bacteria on mycelial growth or morphology were observed.

Figure 2
figure 2

Monitoring Burkholderia–fungi interactions. Plates with sterile iron rings and either water agar, LB or a combination of both were used to monitor the ability to translocate, to disperse and to form biofilms on the hyphae. For microscope observations, Burkholderia cells were tagged with gfp or DsRed (on the pictures B. glathei LMG14190 tagged with DsRed), co-cultivated with fungi (here F. solani) and monitored on a daily basis. Microscopy pictures represent different stages of dispersal on F. solani hyphae over time. (a) Network of hyphae is densely colonised by bacteria (red) in the inner part of the ring. (b) Crossing of the iron ring by fungal hyphae enables Burkholderia cells to cross too when attached to the hyphae. (c) Colonization of the hyphae by transported Burkholderia occurs also on the outer part of the iron ring. The same behaviour was visible when water agar was replaced by LB.

Global analyses to elucidate the molecular basis of Burkholderia–fungi interactions

We used a proteomic approach to characterize the molecular nature of the interaction between Burkholderia strains and fungi. Among the fungal OTUs that co-occurred with Burkholderia, those belonging to the genera Alternaria and Fusarium were among the most frequently detected ones in our analysis (Supplementary Table S1), and thus we selected A. alternata and F. solani as model species. As a representative of the genus Burkholderia, B. glathei LMG14190 (T) (Vandamme et al., 1997) was chosen for these interaction studies, since members of this species were not only the most frequently detected in a previous soil survey (Stopnisek et al., 2014), but have also been demonstrated to interact with fungi (Koele et al., 2009).

Genetic features of B. glathei LGM14190 possibly involved in interactions with fungi

To obtain insights into the reasons for the ubiquitous distribution of B. glathei in soils, as well as for better interpretation of proteomics data, we sequenced the genome of B. glathei LMG14190, which was not available at the time of our analyses. The draft genome is 8'049'485 bp in size, which is consistent with already sequenced strains of the genus Burkholderia (Xu et al., 2013; Carlier et al., 2014, Liu et al., 2014). Annotation of the draft genome predicted 7216 protein coding genes, 50 tRNAs and 3 rRNAs (PRJEB6934).

The genome of B. glathei LMG14190 encodes a large number of genes involved in transport and utilization of various carbon sources. In addition to the utilization of simple sugars, B. glathei LMG14190 has the genetic potential to degrade polymeric carbohydrates such as chitin, cellulose, glycogen, as well as alcohols, including ethylene glycol, ethanol and butanediol. B. glathei LMG14190 contains genes coding for enzymes involved in the degradation of glycerol, which is predicted to be one of the C sources fungi secrete (Nazir et al., 2013). Additionally, fungi produce a number of aromatic compounds that could represent another source of nutrients for Burkholderia (Gutiérrez et al., 1994). Degradation of aromatic compounds is a commonly observed feature of various Burkholderia strains (Pérez-Pantoja et al., 2012; Andreolli et al., 2013; Schamfuss et al., 2013; Xu et al., 2013) and the corresponding genes were also identified in the genome of B. glathei LMG14190. Beside pathways for the utilization of toluene, phenolic, benzoate and catechol compounds, the genome also encodes the complete salicylate degradation pathway. Given the antibacterial and antifungal properties of many of these aromatic compounds, the capacity to degrade such substances could also serve as a protection mechanism for the bacteria and/or their fungal partners.

The type three secretion system has been described as essential to the interaction between B. terrae BS001 and Lyophyllum sp. Karsten (Warmink and van Elsas, 2008). However, homologous genes could not be detected in the genome of B. glathei LMG14190. We also tested for the presence of the hrcR gene, a highly conserved component of the type three secretion system (Warmink and van Elsas, 2008), in the four Burkholderia strains used in our dual culture assays. Surprisingly, the hrcR gene could only be amplified from one of the strains tested, B. hospita, but was not detected in B. glathei, B. fungorum or B. phytofirmans (data not shown). This suggests that type three secretion system is not essential for the interaction with the fungi tested here. Nonetheless, the genome of B. glathei LMG14190 harbours genes that have been described to have a role in biological interactions, such as those involved in the synthesis of the exopolysaccharide cepacian (BGLT00414-BGLT00424), which is produced by most Burkholderia species and is important for interactions with other organisms as well as for resistance to different stresses (Ferreira et al., 2011).

Global changes in the proteome of B. glathei when co-cultivated with fungi

A total of 2406 unique B. glathei LMG14190 proteins and only 3 fungal proteins were detected in the present study, indicating an efficient enrichment of bacterial cells from the co-cultures (Supplementary Table S2). After applying stringent quality filters and considering only proteins identified in at least 2 out of 3 biological replicates, 1583 proteins remained, the expression of which could be semi-quantitatively assessed by spectral counting (Supplementary Table S3). Among those, 923 proteins (almost 60%) were identified in both, single and co-cultures, while 282, 47 and 29 proteins were found to be exclusively expressed during single culturing or co-cultivation with A. alternata and F. solani, respectively (Figure 3). Overall, 742 were differentially expressed (expression change of more than twofold) during co-cultivation compared to the single cultures with the majority of these proteins expressed in lower amounts in the presence of either of the fungi. Most of the proteins identified in this study were assigned to functions associated with the production and conversion of energy, as well as with transport and metabolism of carbohydrates and amino acids (Supplementary Table S2). This corresponds to the high proportion of predicted genes with the same function in the genome of B. glathei LMG14190 (approximately 20% of the whole genome). In contrast, functional classes such as transcription, defence and signalling were under-represented in the B. glathei proteome when considering their contribution to the total number of predicted genes (Supplementary Table S2).

Figure 3
figure 3

Venn diagram depicting quantified proteins of B. glathei LMG14190 grown in single culture (B) and in co-cultivation with A. alternata (BA) and F. solani (BF), respectively.

To shed light on the molecular mechanisms involved in bacterial–fungal interactions, we focussed mainly on the B. glathei proteins whose abundance changed significantly during fungal co-cultivation. As shown in Figures 4 and 5 co-cultivation affected protein expression across all functional categories. Notably, a particular high percentage of differentially expressed proteins were involved in cell motility and translation (Figures 4 and 5). The putative relevance of the observed protein expression changes during co-cultivation to the biological interaction between B. glathei and fungi is highlighted below.

Figure 4
figure 4

Impact of co-cultivation on proteome composition of B. glathei LMG14190. The percentage of proteins in relation to the theoretical proteome with at least twofold change in co-cultivation with either fungus compared with the control (B. glathei alone) is depicted. BA, co-cultivation of B. glathei with A. alternata; BF, co-cultivation of B. glathei with F. solani. Black bars indicate proteins with changed amount in both co-cultivations, white bars indicate proteins with changed amounts in the BF co-cultivation, grey bars indicate proteins with changed amounts in the BA co-cultivation. Functional annotation is based on cluster of orthologous groups (COG).

Figure 5
figure 5

Voronoi treemaps visualizing changes in the proteome composition of B. glathei LMG14190 during fungal co-cultivation. Expression ratios of B. glathei LMG14190 proteins in single cultures (B) compared to co-cultivation with A. alternata (AB) or F. solani (FB). Functional classification of proteins was carried out by Prophane 2.0 and is based on cluster of orthologous groups (COG). Red colour indicates higher expression of the respective protein in single cultivation and blue colour indicates higher expression in co-cultivations. Each cell represents a single protein, functional classes are separated by thicker black lines.

Attenuation of starvation and stress responses in B. glathei when co-cultivated with fungi

B. glathei grown in the absence of fungi was limited to the nutrients available in the growth medium (1/3 MEA) and this resulted in the expression of many transcriptional regulators related to starvation and stress response. The expression rates of most of these regulators and stress factors were significantly reduced in the co-cultivations, suggesting multiple beneficial effects of fungi on B. glathei (Supplementary Table S3). From a total of 255 predicted proteins in signal transduction mechanisms, 36 and 32 were present in co-cultivations with A. alternata (AB) and F. solani (FB), respectively, with 29 of them shared between the co-cultivations (Supplementary Table S3), suggesting similar metabolic response and regulation of B. glathei when growing with both fungi. Most of these proteins were factors related to stress or starvation linked to important nutrients such as nitrogen, phosphate and carbon, which were all relieved upon co-cultivation. For instance, the attenuation of nitrogen (N) starvation in both co-cultivations was supported by the decrease of the nitrogen regulation protein NRI (BGLT03875) and of the glutamine synthetase (BGLT01582). The nitrogen regulation protein NRI is the major activator of nitrogen-controlled genes such as the glutamine synthetase (Reitzer and Magasanik, 1985; Reitzer et al., 1989). It has been shown that nitrogen starvation in E. coli leads also to an increase of proteins involved in amino acid and polyamine degradation (Zimmer et al., 2000). Decreased expression of various amino-acid transport systems as well as of amino acid and polyamine degradation pathways in B. glathei in co-cultivations additionally supports the idea that the fungi alleviate the N starvation response (Figures 4 and 5). Even though most of the pathways and transport systems showed lower abundances compared with the single culture, subunits of histidine ABC transporters (BGLT00146, BGLT01156) were significantly increased in co-cultivations. Histidine could act as N source provided by fungi to B. glathei, since many enzymes of the histidine degradation pathway, such as histidine ammonia-lyase (BGLT04159), urocanate hydratase (BGLT03664) and N-formylglutamate deformylase (BGLT04163), were detected (Supplementary Table S3).

Beside N starvation, phosphate (P) starvation seems to have been alleviated in co-cultivation, as indicated by a significant decrease or absence of proteins such as phosphate starvation-inducible protein (BGLT03521) and glycerol-3-phosphate uptake system (BGLT04691, BGLT05290, BGLT00157) in the co-cultivation with fungi, compared with the control. Expression of proteins associated with the UgpBAEC glycerol-3-phosphate uptake system, a member of the ABC superfamily, was among the most strongly decreased in the co-cultivation experiments. In E. coli, glycerol-3-phosphate can be used as C and/or P source and is also an essential intermediate in phospholipid biosynthesis (Schweizer et al., 1982; Boos, 1998). Glycerol-3-phosphate is an insufficient carbon source, but a sufficient P source for E. coli (Boos, 1998). Thus, the Ugp system that is used for scavenging phosphate-containing compounds in E. coli may also serve this purpose in Burkholderia.

The limited amount of carbon contained in the growth medium could explain the detection of proteins indicative of C starvation in the pure culture control. In both co-cultivations, however, most of these proteins were absent or strongly decreased, such as the Stringent starvation protein A, SspA (BGLT04645) for instance. In E. coli, SspA expression is induced by starvation for glucose, nitrogen, phosphate, amino acids as well as by a decreased growth rate (Williams et al., 1994). In addition to SspA, the carbon starvation protein A, CstA (BGLT04027), a protein that is more specifically expressed during C starvation was also detected and significantly decreased in expression when F. solani was present, while no change was detected in co-cultivation with A. alternata. The decreased amount of SspA and other C starvation proteins during co-cultivation suggests a preference of B. glathei for fungal exudates over the C source present in the growth medium (MEA), which is mainly maltose. Indeed, the maltose ABC transporter (BGLT01992), which is responsible for the uptake of maltose, was decreased in co-cultivations. Additionally, a xylose ABC transporter (BGLT06434) was also strongly decreased in co-cultivation. D-xylose is the most abundant sugar in nature after glucose (Ladisch et al., 1983), and it can be utilized by E. coli as a sole carbon source and metabolized through the pentose phosphate pathway (Song and Park, 1998). That xylose was used when B. glathei grew alone on MEA is suggested by the expression of xylose isomerase (BGLT06433), which catalyses the first reaction in the catabolism of D-xylose (Schellenberg et al., 1984) and which was not present in the co-cultivations with either fungus (Supplementary Table S3).

While our results suggest that B. glathei may preferentially utilize fungal exudates in the co-cultivation experiments, we could not unambiguously identify the primary C source of B. glathei in the exudates. A likely candidate appears to be ribose, as the amounts of proteins involved in ribose uptake (BGLT04604, BGLT04606 and BGLT05567) were strongly increased in co-cultivation.

Burkholderia proteins involved in detoxification processes in the presence of both fungi

In contrast to the positive effects of fungi on alleviation of nutrient starvation, proteins indicating stress were also detected when B. glathei was grown with the fungi. An increase in various catalases and peroxidases (BGLT01132, BGLT04737) in co-cultivations suggests that bacteria experienced oxidative stress when grown with fungi (Supplementary Table S3). Increased expression of 5-methyltetrahydropteroyltriglutamate homocysteine methyltransferase, MetE (BGLT01007) in co-cultivation with A. alternaria also suggests oxidative stress, since this protein is very highly expressed in E. coli cells exposed to oxidative stress (Hondorp and Matthews, 2004). Antibacterial compounds released by the fungi could cause oxidative stress (Ammar et al., 1979; Hellwig et al., 2002; Deshmukh et al., 2014; Soltani and Hosseyni Moghaddam, 2014). An increase of the RND (resistance nodulation division) family of efflux pumps (BGLT02584, BGLT02585, BGLT02586, BGLT04622, BGLT05365) in co-cultivations further supports the idea of fungal release of antibacterial molecules. The detected proteins show high homology to members of the NodT family of outer membrane transport proteins from the genus Rhizobium (Rivilla et al., 1995) and the CmeABC multidrug efflux system of Campylobacter jejuni (Lin et al., 2002), which are involved in the export of a wide range of drugs. In addition to drug resistance, RND pumps have an important role in biological interactions such as nodulation, colonization and host persistence (Piddock, 2006). In B. cenocepacia, RND efflux pumps also influence phenotypic traits involved in pathogenesis, such as motility and chemotaxis (Bazzini et al., 2011). In addition to these detoxification systems, other proteins with a putative function in detoxification were found to be increased in co-cultivations: the proteins YagS (BGLT01120), YagT (BGLT01121) and YagR (BGLT00865), which are part of the yagTSRQ operon, were significantly increased in both co-cultivations (Supplementary Table S3). This operon was shown to encode a periplasmatic aldehyde oxidoreductase that oxidizes a broad spectrum of aldehydes to their respective acids, thereby contributing to the detoxification of the cells (Neumann et al., 2009).

RND efflux pumps were not the only proteins that were increased and that are associated with cell envelope biogenesis and outer membrane structures: many outer membrane porines and exporters (BGLT02029, BGLT00225, BGLT02300, BGLT00513 and BGLT04032) were also highly increased in co-cultivations. Although the functions of most of these proteins remain unknown, a recent transcriptomic study of B. cenocepacia J2315 stress responses revealed that most of the homologous proteins were increased under nutrient starvation and low oxygen conditions, or when exposed to oxidative stress (Sass et al., 2013). Thus, the expression of membrane-associated proteins observed in co-cultivation might be due to stresses induced by the fungi, although changes in membrane composition could also originate from osmotic stress. The primary response to osmotic stress in most bacteria is the uptake of potassium (Paul 2013). Interestingly, the two homologous potassium uptake systems (BGLT02399- BGLT02401 and BGLT06537-BGLT06539) encoded in B. glathei genome were expressed in both co-cultivations (Supplementary Table S3).

B. glathei proteins involved in motility are not expressed in fungal co-culture

Using fungal hyphae for dispersal has been shown in previous studies to be advantageous in the soil, where heterogeneity of soil particles and the lack of water films represent major obstacles to motility. In our in vitro experiments, dispersal of Burkholderia strains on fungal hyphae was observed but many proteins involved in motility, which were expressed in the control, were undetected in the presence of the two fungi (Figures 4 and 5, Supplementary Table S3). This might indicate that bacterial cells encountering a fungal hypha cease to invest resources into their own motility and rather rely on the fungus for transport.

Conclusion

Evidence has emerged over the past years that Burkholderia are often associated with fungi in soils (Warmink et al., 2009; Lepleux et al., 2012; Nazir et al., 2012). The co-occurrence analysis presented here fully supports this notion and extends our understanding of the ability of Burkholderia to form associations with a broad range of fungal taxa. The widespread soil inhabitants Burkholderia glathei, B. terrae, B. fungorum and B. phytofirmans could all interact and disperse with A. alternata, F. solani and R. solani, suggesting that these broadly occurring interactions might also occur in the environment. Previous reports indicated that the interactions with fungi might be beneficial for the bacteria, yet knowledge of the nature of these benefits is still scarce (Nazir et al., 2010a, 2013; Warmink et al., 2011). Our proteome analysis, which gave a snapshot view of a dynamic interaction, revealed that in a nutrient-limited medium, B. glathei was able to use multiple substrates provided by the fungi, which attenuated the starvation response observed when grown alone (Figure 6). However, B. glathei encountered new stresses when co-cultivated with fungi, as revealed by the higher expression of proteins associated with various defence and tolerance mechanisms. These functions seem important for the successful colonization and persistence of Burkholderia sp. on fungal hyphae. The benefits that Burkholderia appear to gain from the interaction with fungi are likely to outweigh the costs involved in the co-existence and thus these associations might represent a successful strategy to permit the survival of Burkholderia in the soil environment.

Figure 6
figure 6

Schematic overview of the most important changes in B. glathei LMG14190 proteome during co-cultivation with A. alternata and F. solani. The overview includes proteins involved in major pathways that have been detected in our analysis and proteins whose amounts were either decreased (red) or increased (green) in both co-cultivations. Dashed lines represent the effects of hypothetical stress factors (lightning symbols) or substrates that could be produced by fungi. All flagellar proteins were absent in the co-cultivation setups, thus the whole flagella apparatus is marked in red.