Intimate relationships among soil bacteria: actinomycetes and mycolic acid-containing bacteria

Co-culture is an ecient strategy for natural product discovery. We have used mycolic acid-containing bacteria (MACB) Tsukamurella pumonis TP-B0596 to induce secondary metabolism by actinomycetes and have found several natural products. We also observed that MACB attached to the mycelium of Streptomyces lividans forming coaggregates during combined-culture. This stimulated interest in the interactions among actinomycetes and MACB, and we found that soil isolated cultures contained a mixture of actinomycetes and MACB. Our previously observed interactions were the result of selective screening and combination of bacteria in the lab, which warranted investigation of the existence of these interactions in the natural soil environment. Therefore, in this paper, we report the interaction between a co-isolated natural pair of actinomycetes and MACB in terms of morphology and metabolic changes. A natural pair of actinomycetes and MACB co-aggregated in liquid culture and showed metabolic changes. Interestingly, co-aggregated actinomycetes and MACB were re-isolated from soil with no obvious morphological colony differences from the colony of a single strain. The results demonstrate that there is a stochastic chance of picking colonies containing co-aggregated actinomycetes and MACB, which suggests that the pair can exist in co-aggregate form in the soil environment and interact with each other. interspecies aggregates pairs cultured in liquid medium, and cultured xed for SEM same incidental nding of mixed culture stocks containing actinomycetes and MACB, which were isolated from colonies be formed by single bacteria. Actinomycetes and MACB are ubiquitous soil bacteria, that this characteristic interaction takes place in the natural soil environment. in we focused on the interaction between isolated pair of actinomycetes and MACB in terms of morphology and metabolic changes. We collected 71 environmental samples from Hegura Island and obtained 1,041 spore/cell bacterial stock cultures. Partial 16S rRNA gene sequencing of 182 strains identied 32 genera among these bacterial isolates, indicating that the environmental samples from Hegura contain a rich diversity of bacteria. Within the 1,041 stocks, we found at least 11 stocks that contained more than two different strains. All co-isolated bacteria from those 11 stocks could grow in monoculture on general agar medium, indicating that there was no obligatory symbiotic relationship between the individual isolates. Taxonomic analysis based on 16S rRNA genes revealed that the co-isolated strains were not bacteria with characteristic properties but were ubiquitous soil bacteria. This suggested that soil bacteria associate with each other in the natural soil habitat and can thus be co-isolated. Interestingly, besides actinomycetes and MACB, other co-isolated bacterial pairs generated co-aggregates in liquid culture. This indicated that the physical association of bacteria does not require special properties but is facilitated by general interaction in the soil microbiome. relatively dominant species in the rhizosphere. Within the 11 stocks, seven mixed stocks were obtained from rhizosphere soil samples from different locations. might suggest the relatively frequent association of bacteria with other in those habitats. Interspecies co-aggregation facilitates bacterial communication physically short distances. bacterial using small signaling membrane vesicles, 36,37 cell–cell for 0.0025% biotin), in distilled water (pH 7.2). To prevent fungal amphotericin (nal conc. 20 µg/ml) and cycloheximide (nal conc. 20 µg/ml) were dissolved in dimethyl sulfoxide (DMSO) and added to the agar medium after autoclaving.


Introduction
Actinomycetes are Gram-positive soil bacteria that produce pharmaceutically important natural products as secondary metabolites. 1 Each strain contains more than 30 putative biosynthetic gene clusters for secondary metabolites. 2,3 Although they are considered to have potential for producing more than 30 secondary metabolites, the number of detectable products in normal lab culture conditions are very limited. 3,4 Previous reports have shown that responses to environmental factors affect secondary metabolism of actinomycetes, 4 therefore, bacterial interaction is a promising strategy for the activation of the cryptic biosynthetic genes.
The soil environment from where actinomycetes can be isolated contains many different microorganisms. 5 Actinomycetes respond to environmental factors including biogenic factors of other microorganisms. 4 We previously reported that a group of mycolic acidcontaining bacteria (MACB; e.g. Tsukamurella pulmonis TP-B0596) affected the secondary metabolism of actinomycetes in lab culture. 6 MACB are actinomycetes containing speci c long chain fatty acids (mycolic acids) on the cell wall. 7,8 Using MACB (such as T. pulmonis TP-B0596) as a partner, Streptomyces lividans TK23 was activated to produce undecylprodigiosin (RED) and actinorhodin (ACT), which are not produced under general lab culture conditions. 6 Using T. pulmonis TP-B0596 as a combined-culture partner, a number of natural products have been isolated, [9][10][11][12][13][14][15][16][17] and this strategy was also found to be e cient to enhance production during heterologous expression. [18][19][20] This co-culture method using the combination of actinomycetes and MACB is named combined-culture (see Supplemental Table 1 for the whole list).
The activation of RED and ACT did not occur through provision of culture extracts or killed bacteria, 21 therefore, it was suggested that physical cell contact by living T. pulmonis is required for the activation of RED and ACT by S. lividans. Our previous analysis of cell morphology of S. lividans and several MACB in liquid culture under scanning electron microscopy (SEM) imaging showed that they formed co-aggregates. 21 Streptomyces species generally grow by forming a pellet in liquid culture, 22 and signi cant attachment of MACB cells on Streptomyces mycelium was observed. 21 Co-aggregation and the production of RED and ACT was not observed with addition of killed MACB (e.g. by γ-ray irradiation or formaldehyde xation). 21 This suggested that co-aggregates formation through continuous physical contact might be important for the activation of secondary metabolism by actinomycetes.
Bacteria can form interspecies aggregates in the environment, 23 and these aggregates can also occur between bacteria from different environments. 24 Several theories have been reported to explain the interactions in interspecies aggregates for oral or sludge environments. 23 However, there is limited understanding of the mechanism and function of Streptomyces/MACB aggregates in the general soil bacterial community 25 , and it will be important to elucidate the ecological relevance or bene ts of those bacterial interactions in the future.
The previous observation of the interactions between actinomycetes and MACB was the result of deliberate combinations of bacteria selected by screening in the lab for activation of secondary metabolism. 6,21 However, these results suggested that the characteristic interactions such as forming interspecies co-aggregates and affecting each other including secondary metabolism, might be naturally occurring events in the soil environment. In this study, we report the morphology and metabolic changes of a co-isolated pair of actinomycetes and MACB from environmental soil of Hegura Island. The natural co-isolated pair formed interspecies aggregates and their interaction under combined-culture lab conditions affected the production of secondary metabolites. We performed additional experiments to re-isolate the co-aggregated cell pellets and demonstrated that mixed aggregates of actinomycetes and MACB were co-isolated with no obvious morphological change of the growing colony.

Isolation of actinomycetes from Hegura Island
Hegura Island (N37°51'5", E136°55'7") in Ishikawa prefecture, Japan, is known as a stopover point for migratory wild birds in the middle of the Sea of Japan. We expected that those migratory wild birds might bring rich bacterial diversity to this isolated island. We collected environmental samples from 71 points on Hegura Island in June 2008. General methods were applied for the isolation of actinomycetes (see materials and methods for detail and Supplemental Figure 1); growing single colonies that appeared to be actinomycetes were isolated from the agar plates. We obtained a total of 1,041 bacterial isolates from this process. Partial 16S rRNA gene sequences of 182 isolates were determined to estimate the taxonomic diversity of the obtained in-house culture collection. The Identi cation of co-isolated strains After making the spore/cell stocks of the isolated strains, stock solutions were inoculated onto ISP2 agar plates, which was different from the original isolation medium (Supplemental Figure 1). On those agar plates, we observed at least 11 stocks that appeared to contain more than two different bacteria (Table 1). Although the isolates were thoroughly separated in the isolation process including vigorous agitation by vortex mixer and considerable dilution, we obtained mixed colonies of more than two bacteria. Therefore, it was speculated that those bacteria originally formed tight co-aggregates in soil and thus were isolated as a single strain but were subsequently found to be a mixture of several strains. We de ned those bacteria as co-isolated bacteria and performed further analysis. Table 1 Identi cation of strains from co-isolated colonies. Abbreviations, FA: lamentous actinomycetes, MACB: mycolic acid-containing bacteria, CA: cocci-shaped actinomycetes, GP: Gram-positive bacteria, GN: Gram-negative bacteria. Accession numbers are of the most similar type strains. To identify the strains in the mixture, the dilution method was applied to obtain single colonies from the 11 mixed stocks. This procedure successfully isolated two to three different colony types growing on the agar plate with different colony morphology. We then grew the bacteria in liquid culture to isolate their genomic DNA and determined partial 16S rRNA gene sequences of each strain obtained from the 11 co-isolated colonies. Among these pairs, we identi ed four that were a mixture of actinomycetes and MACB, and seven that were a mixture of actinomycetes and other bacteria (Table 1).
Co-aggregation of co-isolated actinomycetes and MACB HEK138, a co-isolated pair from bamboo grass growing in rhizosphere soil using H 2 O dilution and ISP4 agar plate (Supplemental  (Table 1). As expected, cells of M. septicum HEK138M (bacilli-shaped cells) and S. variegatus HEK138A ( lamentous hyphae) generated co-aggregates, where the cell clump of M. septicum HEK138M coexisted with the cell pellet of S. variegatus HEK138A mycelium. (Figure 2a) HEK904, a co-isolated pair from soil using phenol treatment and ISP4 agar plate (Supplemental Table 4), contained the cocci-shaped actinomycete Isoptericola nanjingensis HEK904A 28 and MACB Mycolicibacterium neworleansense HEK904M 29 ( Table 1). The coccishaped actinomycete I. nanjingensis HEK904A and MACB M. neworleansense HEK904M also formed mixed aggregates. Although I. nanjingensis did not have the lamentous cell shape, M. neworleansense cells were bacilli-shaped, and the two strains could be differentiated by cell size (Figure 2b). Therefore, SEM imaging con rmed that these two strains formed a mixed aggregate.
HEK128, a co-isolated pair from an unknown succulent plant rhizosphere soil using H 2 O dilution and humic acid agar plate (Supplemental Table 4), contained the lamentous actinomycete Amycolatopsis tolypomycina HEK128A 30 and lamentous MACB Nocardia veterana HEK128M 31 (Table 1). HEK953, a co-isolated pair from seaweed using H 2 O dilution and humic acid agar plate (Supplemental Table 4), contained the lamentous actinomycete Streptomyces canus HEK953A 32 and lamentous MACB Nocardia vinacea HEK953M (Table 1). For these combinations (HEK128 and HEK953), because both strains had a lamentous cell shape, it was not possible to discriminate between them by SEM imaging (Figure 2c and d). Therefore, we could not conclude whether these pairs formed mixed aggregates.
Co-aggregation of co-isolated actinomycetes and other bacteria Seven other co-isolated colonies were identi ed as mixtures of actinomycetes and other bacteria (Table 1). Four combinations were obtained from the remaining seven co-isolated stocks: Streptomyces-other bacteria (HEK314 and HEK830), cocci-shaped actinomycetes-cocci-shaped actinomycetes (HEK654), Streptomyces-cocci-shaped actinomycetes-other bacteria (HEK652 and HEK282), and Streptomyces-Streptomyces (HEK332 and HEK423) ( Table 1). All the co-isolated bacteria could grow in monoculture, indicating that there was no obligatory symbiotic relationship between the co-isolates. As naturally co-isolated actinomycetes and MACB pairs formed coaggregates, we also tested these combinations for their ability to form coaggregates.
Bacilli-shaped cells were found to attach on the mycelium of S. drozdowiczii HEK314A in co-culture, indicating that they formed coaggregates ( Figure 3a).
HEK830 contained Streptomyces tsukubensis HEK830A and Peribacillus simplex HEK830B. P. simplex HEK830B had bacilli-shaped cells with a wrinkled cell surface, and it was found with lamentous actinomycete S. tsukubensis HEK830A mycelium in co-culture, indicating that they formed co-aggregates ( Figure 3b).
HEK654 contained the cocci-shaped actinomycetes Kineococcus mangrovie HEK654A1 and Pseudokineococcus marinus HEK654A2. Both cells were observed with cocci-shaped cells but the cell sizes were different in monocultures. K. mangrovie HEK654A1 cells were relatively large (1.0 µm) whereas P. marinus HEK654A2 cells were smaller (0.65 µm). Cells with both different sizes were observed in the cell pellet of the co-culture, indicating that they formed co-aggregates ( Figure 3c).
We observed that two of the three strains in this mixture could form coaggregates, but the three bacteria together did not form a single aggregate (Figure 3d).
HEK282 contained Streptomyces albo avus HEK282A1, actinomycetes Cellulosimicrobium funkei HEK282A2, and Paenibacillus chitinolyticus HEK282B. S. albo avus HEK282A1 was observed in mycelium form. C. funkei HEK282A2 was observed as bacillishaped cells but also with signi cantly elongated cell shape. P. chitinolyticus HEK282B was also observed as bacilli-shaped cells with signi cantly elongated cell shape. Bacilli-shaped cells and lamentous cells were observed in co-culture but it was not possible to discriminate the origin from the SEM image ( Figure 3e). Therefore, we could not conclude whether those pairs formed mixed aggregates.
HEK332 and HEK423 each contained two Streptomyces species. Because they both had lamentous cell shape, it was not possible to discriminate between them by SEM (Supplemental Figure 3). Therefore, we could not conclude whether those pairs formed mixed aggregates.
Combined-culture of a co-isolated pair of actinomycete and MACB Because MACB can affect the secondary metabolites pro le of actinomycetes, we compared the metabolites of monoculture and combined-culture using high performance liquid chromatography with diode array detector (PDA-HPLC). We found signi cant differences in metabolites extracted from the culture after combined-culture of the natural pair of actinomycetes and MACB when compared with monoculture ( Figure 4 and Figure 5). Focusing on newly generated peaks, we found 19 new peaks with absorbance at 300 nm generated during combined-culture of HEK138A and HEK138M (Figure 4a). Speci c growth inhibition was observed for combined-culture extracts against tested Gram-positive bacteria B. subtilis, Staphylococcus aureus, and Micrococcus luteus ( Figure 5, Table 2) and weak semi-inhibition against Gram-negative Escherichia coli. The active compound that showed weak semi-inhibition against E. coli was identi ed as desferrioxamine E by activity guided puri cation (Supplemental Figure 9). For HEK128A and HEK128M, one new peak with absorbance at 330 nm was generated during combined-culture (Figure 4b). Antibiotic activity was slightly increased by combined-culture ( Figure 5, Table 2). For HEK953A and HEK953M, ve new peaks with absorbance at 270 nm were generated during combined-culture (Figure 4c). Antibiotic activity was slightly decreased by combined-culture in this case ( Figure  5, Table 2). For HEK904A and HEK904M, two new peaks with absorbance at 340 nm were generated by combined-culture (Figure 4d).
Antibiotic activity was not observed in these strains ( Figure 5, Table 2).   Figure 4). In the negative control, pellets from co-culture of Sl + Bs did not show signi cant co-isolation rate (0/42 colonies from isolated Sv pellets; Figure 6b, Supplemental Figure 5). Because Bs was killed by coculture with Sv (probably by the antibiotic produced by Sv; Table 2), we could not obtain the co-isolation rate for Sv + Bs (data not shown). Taken together with the SEM images, these results con rmed that signi cant cell attachment occurred for Streptomyces species and MACB when compared with B. subtilis.
We further tested whether the co-aggregated pair could be co-isolated using the isolation protocol performed initially for the bacterial isolation from Hegura Island environmental samples (see materials and methods for details). After the combined-culture, the culture was washed twice with phosphate buffered saline (PBS) and mixed with soil before isolation (Supplemental Figure 2b). The results of Sv + Ms combined-culture, which was veri ed by selective colony PCR, showed several co-isolated colonies (4/36 colonies from isolated Sv pellets; Figure 6d, Supplemental Figure 7). These results suggested that a low proportion of co-aggregated cells could be stochastically co-isolated using the isolation protocol.

Discussion
Mixing actinomycetes and MACB in the same laboratory culture resulted in a high chance of alterations in secondary metabolism.
The actinomycetes and MACB formed co-aggregates when cultured together in the same laboratory culture. This study was initiated by an incidental nding of mixed culture stocks containing actinomycetes and MACB, which were isolated from colonies that appeared to be formed by single bacteria. Actinomycetes and MACB are ubiquitous soil bacteria, therefore it was speculated that this characteristic interaction takes place in the natural soil environment. Therefore, in this paper, we focused on the interaction between the naturally isolated pair of actinomycetes and MACB in terms of morphology and metabolic changes.
We collected 71 environmental samples from Hegura Island and obtained 1,041 spore/cell bacterial stock cultures. Partial 16S rRNA gene sequencing of 182 strains identi ed 32 genera among these bacterial isolates, indicating that the environmental samples from Hegura Island contain a rich diversity of bacteria. Within the 1,041 stocks, we found at least 11 stocks that contained more than two different strains. All co-isolated bacteria from those 11 stocks could grow in monoculture on general agar medium, indicating that there was no obligatory symbiotic relationship between the individual isolates. Taxonomic analysis based on 16S rRNA genes revealed that the co-isolated strains were not bacteria with characteristic properties but were ubiquitous soil bacteria. This suggested that soil bacteria associate with each other in the natural soil habitat and can thus be co-isolated. Interestingly, besides actinomycetes and MACB, other co-isolated bacterial pairs generated co-aggregates in liquid culture. This indicated that the physical association of bacteria does not require special properties but is facilitated by general interaction in the soil microbiome.
Actinomycetes are relatively dominant species in the rhizosphere. 33 Within the 11 stocks, seven mixed stocks were obtained from rhizosphere soil samples from different locations. This might suggest the relatively frequent association of bacteria with other bacteria in those habitats. Interspecies co-aggregation facilitates bacterial communication over physically short distances. 34 Several strategies are known for bacterial communication such as using small signaling molecules, 35 membrane vesicles, 36,37 and cell-cell physically-dependent communication over short distances, 38,39 including interspecies communication. 40 The interactions observed by SEM imaging did not provide evidential proof of known bacterial communication strategies. The ecological bene t of this association requires further study because there could be strong competition for nutrients and other impacts among cells within co-aggregates.
It was suggested that the co-aggregation between actinomycetes and MACB is a naturally occurring association with tight interaction and thus they were isolated together. Thus, we tested whether a co-aggregated bacterial pair could be isolated by forming a single colony. Cell pellets of actinomycetes grown in liquid culture were isolated and incubated on agar plates after washing. The growing colony appeared to be derived from a single strain of actinomycete. However, selective colony PCR indicated that the colony harbored both actinomycetes and MACB. When the cell pellets were isolated by a procedure similar to that of the original isolation from soil, 11.1% (4/36) of colonies were formed by two strains. This demonstrated that co-aggregating strains could be co-isolated under these conditions.
We tested the effect of combined-culture on metabolic changes in the co-isolated pairs. In our previous work using the culture collection obtained from Hegura Island, we isolated natural products from combined-culture with T. pulmonis such as: streptoaminals and 5-alkyl-1,2,3,4-tetrahydroquinolines from Streptomyces nigrescens HEK616, 13,14 dracolactams from Micromonospora wenchangensis HEK797, 12 and streptogramins from Streptomyces albogriseolus HEK740. 41 Therefore we asked whether the original combination of actinomycetes and MACB affects their metabolism. As expected, we observed signi cant differences in the HPLC chromatograms of extracts from combined-culture when compared with monoculture of the naturally isolated pair. The extracted metabolites of HEK138A and HEK138M showed speci c growth inhibition activity indicating potential antibiotic activity. Antibiotic production by the actinomycete (HEK138A) would be bene cial for the MACB (HEK138M), providing protection (by antibiotic) to secure the natural habitat without killing each other. MACB (e.g., Rhodococcus sp.) have a degradative role in the environment, which might provide easily available nutrient sources for the actinomycete (HEK138A). There are no reports to date of the isolation of natural products from Isoptericola species (HEK904A), therefore identi cation of induced metabolites in combined-culture of HEK904A and M would be interesting for future analysis. Taken together, the approach of co-culturing naturally isolated bacterial coaggregates could be an option for screening new bioactive secondary metabolites.

Isolation of actinomycetes
We collected 71 environmental samples from Hegura Island in June 2008. Brie y, samples were suspended in sterilized ddH2O and divided in half. Phenol solution was added to the one half of the suspension to kill phenol-sensitive bacteria. Subsequently the phenol-treated suspension was spread with appropriate dilution on ISP4 agar or humic acid agar with antifungal antibiotics. Ten milliliters of sterilized ddH 2 O was added to 1 g of environmental sample and suspended with agitation (on a shaker at 300 rpm) for 10 min. The suspended solution was further diluted by addition of 9 ml of sterilized ddH 2 O. In the next step, 9 ml of sterilized ddH 2 O or phenol suspension (1 g/140 ml) was added to the 1 ml of diluted suspension and mixed for 10 min (on a shaker at 300 rpm). The bacteria were grown on ISP4 medium or humic acid medium at 30°C for 7-10 days. ISP4 agar medium contained soluble starch Under the stereomicroscope, the apparent single colonies forming on ISP4 or humic acid agar plates were picked by toothpick and sub-cultured on new ISP4 or humic acid agar plates. After subculture, we checked for contamination by other bacteria in this process.
The subcultured actinomycetes that appeared as single bacteria were further subcultured on new ISP4 or humic acid agar plates and incubated for 7-10 days. Grown cells or matured spores were recovered from the agar plates to make 20% glycerol spore/cell stocks and stored at −80°C. A total of 1,041 spore/cell stocks were obtained by these isolation processes.

16S rRNA gene sequencing
Genomic DNA of the isolated bacteria was extracted and puri ed using the cetyltrimethylammonium bromide method. 42 16S rRNA genes of the isolated bacteria were ampli ed by PCR using a thermal cycler ( Samples for SEM were prepared as follows. One milliliter of liquid culture was collected into a 1.5 ml tube. After centrifugation (3000 g, 1 min) and removal of the supernatant, 1 ml of 2.5% glutaraldehyde in 0.1 M sodium phosphate buffer (pH 7.4) was added to the cell pellets and incubated for 1 h at 4°C. After discarding the solution, 1 ml of 0.1 M sodium phosphate buffer (pH 7.4) was added to rinse the sample. Then, 1% OsO4 in 100 mM sodium phosphate buffer (pH 7.4) was added and incubated for 1 h on ice. After discarding the solution, 1 ml of 0.1 M sodium phosphate buffer (pH 7.4) was added to rinse the sample. Subsequently, the sample was dehydrated by stepwise addition of 50%, 70%, 90%, and 100% ethanol (1 ml each), and nally substituted by t-butyl alcohol, and stored at 4°C until use. The sample was dried under vacuum before Pt/Pd sputter coating (E-1030, Hitachi, Japan) and observation by SEM (S-4800, Hitachi, Japan). Stereomicroscope images of bacterial colonies formed on agar plates were obtained by Nikon SMZ25 equipped with a SHR Plan Apo 1.6× lens (Nikon) and DS-Ri2 digital camera (Nikon).
Bacterial co-isolation assay S. lividans (Sl), T. pulmonis (Tp), HEK138A, HEK138M, and B. subtilis (Bs) were inoculated into 10 ml of ISP2 medium and precultured for 2 days at 180 rpm and 30°C. The following ratios of aliquot from preculture were used to inoculate 20 ml of YGGS medium in a 100 ml Erlenmeyer ask for co-culture (Sl : Tp = 1 ml : 300 µl, Sl : Bs = 1 ml : 300 µl, 138A : 138M = 1 ml : 300 µl, 138A : Bs = 1 ml : 300 µl). After culture for 5-6 days (180 rpm, 30°C), 1 ml of culture was used for further analysis. First, the whole culture broth was passed through a stainless-steel sieve (φ125 µm, 7.5 × 2.0 cm, SANPO Co.) to trap the cell pellets. Then 500 ml of PBS buffer was used to thoroughly wash the bacterial pellets that were trapped in the sieve. Subsequently, the bacterial pellets were recovered from the sieve and the washing step was repeated. After the washing step, the bacterial pellets were transferred to the center of an ISP4 agar plate and further incubated for 7 days. The growing colony images were obtained by stereomicroscope and the primers (see supplemental data) were used to amplify speci c regions of the 16S rRNA gene by PCR using GoTaq Green master mix (Promega). Agarose gel (1.5% w/v) was used for DNA electrophoresis.
Re-isolation of co-aggregated cells from soil HEK138A or HEK138M were inoculated in 10 ml ISP2 medium and precultured for 2 days (180 rpm, 30°C). The following ratio of aliquot from preculture was used to inoculate 20 ml of YGGS medium in a 100 ml Erlenmeyer ask for combined-culture (HEK138A : HEK138M = 1 ml : 300 µl). After culture for 5 days (180 rpm, 30°C), the whole culture broth was transferred to a conical tube (50 ml) and left to stand for 2 min to allow pellet sedimentation. The upper layer was discarded and 20 ml of ddH 2 O was added to wash the cell pellets. This washing step was repeated twice. Then 1 g of autoclaved soil (obtained from Yayoi campus) was added to the collected cell pellets. After addition of soil, the process for bacterial isolation was repeated as described above. (see supplemental data) Primer sequences used for speci c detection in the bacterial co-isolation assay For Sl and Tp combined-culture, the following primers were used to speci cally amplify the partial 16S rRNA gene regions. was applied for each data acquisition. The HPLC system was equipped with a Cosmosil 5C18-AR-II column (4.6 i.d. × 250 mm, Nacalai Tesque). Acetonitrile and H 2 O (MilliQ) containing 0.1% formic acid was used as mobile phase. Acetonitrile at 5% was the mobile phase for the rst 5 min, and then increased to 95% by linear gradient from 5 to 25 min, and then maintained at 95% until 30 min. The column temperature were kept at 40°C and the ow was 1.0 ml/min.

Paper-disc growth inhibition assay
Bacillus subtilis ATCC6633, Staphylococcus aureus 209PJC-1, Micrococcus luteus ATCC9341, and Escherichia coli NIHJ-2 were inoculated in 10 ml Mueller Hinton (MH) medium (BD Biosciences) and precultured for 1 day (180 rpm, 30°C). Preculture (0.6 ml) was added to 6 ml of soft agar (1.6% nutrient broth (BD Biosciences), 0.5 % agar). The solution was applied as an overlay to an agar plate (140 mm × 100 mm. Eiken chemical Co., LTD) containing 15 ml of MH agar. Twenty microliters of samples dissolved in DMSO (extracts from 0.2 ml culture) were transferred to paper discs (8 mm, ADVANTEC) and placed on an agar plate. After incubation at 30°C for 1 day, the zones of growth inhibition were measured.
Phylogenetic tree of isolated bacteria.
Parenthesized numbers indicate the number of strains within the clade. Accession numbers are of the most similar type strains.
Letters A-K indicate the co-isolated bacterial pairs.
Identi cation of strains from co-isolated colonies. Abbreviations, A: actinomycetes, MACB: mycolic acid-containing bacteria, GP: Gram-positive bacteria, GN: Gram-negative bacteria. Accession numbers are of the most similar type strains.
Scanning electron microscope image of co-isolated lamentous or cocci-shaped actinomycetes and mycolic acid-containing bacteria in combined-culture. Paper disc assay.
Clearance zones of Gram-positive Bacillus subtilis ATCC6633, Staphylococcus aureus 209PJC-1, Micrococcus luteus ATCC9341, and Gram-negative Escherichia coli NIHJ-2 were determined in response to extracts from culture of the natural pairs in combined-culture.  Metabolites analyses of naturally co-isolated bacterial pair in combined-culture using high performance liquid chromatography (HPLC). Blank arrows indicate the new peaks and gray arrows indicate the lost peaks that were observed consistently in all three replicates of combined-culture. (a) HPLC pro les for HEK128A and HEK128M monitored at 330 nm. (b) HPLC pro les for HEK138A and HEK138M monitored at 300 nm. (c) HPLC pro les for HEK904A and HEK904M monitored at 340 nm. (d) HPLC pro les for HEK953A and HEK953M monitored at 270 nm. The wavelengths monitored were selected to show the clear differences between combined-culture and monoculture.