Experimental kiwifruit (cvs. Deliwoong, NHK0013) were designated in the kiwifruit genetic resource center in Namhae, Republic of Korea (34°48′56.8″N 127°55′42.7″E). Samples of rhizosphere soil, sap and pollen were taken from a same tree with three biological replications. Rhizosphere soil samples were collected from surface of rootlet of the tree depths between 25 to 40 cm. Attached rhizosphere soil was brushed lightly into 50-ml tubes. The rhizosphere soil was collected on March 24, 2017, temperature was 9.3 ± 4°C and a relative humidity was 68.8%, respectively. Sap was collected by drilling the second branch and securing the sap into sterilized 5-ml tubes on April 12, temperature was 15.9 ± 6°C and a relative humidity was 28.8%. Pollen was collected on May 15, 2017 (18.5 ± 5°C and a relative humidity of 40.6%). Total 100-150 flower buds of the kiwifruit were picked from each tree and then dried in a dry chamber at 23°C for 4-5 h. Then the pollen was collected in 100 μm of testing sieve (NAVIMRO, Seoul, Korea) by shaking off (Cwynar et al., 1979). All samples were collected with three biological replications.

Pollen samples (0.1 g) were mixed with 9.9 ml of 1× phosphate saline buffered (PBS; 8 g NaCl, 0.2 g KCl, 1.44 g Na₂HPO₄, 0.24 g KH₂PO₄ per liter final pH 7.4) and sonicated with 35 kHz frequencies for 10 min using an ultrasonic bath (Bandelin electronic GmbH & Co., Berlin, Germany) to detach microbial organisms from pollen. Rhizosphere soil (1 g) or 1 ml of sap was mixed with 9 ml of sterilized water. Three replicates of each rhizosphere soil, sap, and pollen samples were diluted with sterile water from 10⁻¹ to 10⁻⁵, spread on 1/5 tryptic soy agar (TSA; 6 g of tryptic soy broth [TSB], 20 g of agar per liter) and then incubated at 27°C for 5 days. Single colony was picked by a sterilized toothpick and transferred into 1/5 TSB (6 g of TSB per liter) in 96-well-plates (SPL Life Science, Seongnam, Korea), incubated at 27°C for 3 days and then stored as glycerol stock (20% final concentration) at −80°C until further use.

Anti-Psa assay was designed using a modified disk diffusion method (Ahn et al., 2009) and the assay conducted three times per a bacterium. As the first screening, Psa was cultured in TSB for 7 days. The bacterial libraries were inoculated on tray plates, which contained 1/5 TSA, using a 96-pin microplate replicator (Boekel Scientific, Feasterville, PA, USA) and incubated at 27°C for 3 days. Psa was mixed with 0.5% sterilized agar solution as final OD₆₀₀ 0.2 and the mixed solution were overlain on the TSA tray plates as the second layer using a 1-ml pipette. The agar was completely hardened at room temperature, incubated at 20°C for 3 days and evaluated inhibition activity to select anti-Psa strains. The second and third screenings were performed with the selected anti-Psa bacteria from 1st screening (cell density was adjusted OD₆₀₀ 0.2) and Psa on 90-mm Petri dishes (SPL Life Science). At the third screening, diameter of Psa inhibition zone was measured and statistical analysis was carried out by Tukey’s HSD test (P < 0.05).

To identify, 16S rRNA and housekeeping genes (atpD and recA) were sequenced with the selected anti-Psa strains. The strains were cultured on MS agar (mannitol soya; 20 g of mannitol, 20 g of soya, 20 g of agar per liter). Genomic DNA was extracted by modified CTAB methods (Wilson, 2001). Polymerase chain reaction (PCR) was performed with amplification primers in Supplementary Table 1. Each 50 μl reaction mixtures contained 20 ng of template DNA, 20 μl of 2× PCR buffer for KOD FX Neo, 0.16 mM dNTPs, 1 μl of forward and reverse primers (10 pmol/μl) and 0.3 μl of KOD FX Neo (1.0 U/μl) with PCR-grade water. PCR reactions were initial denaturation at 95°C for 5 min, followed by 30 cycles of denaturation at 95°C for 30 s, annealing at primer-pair-specific temperature for 30 s, extension at 72°C for 1 min and final extension at 72°C for 10 min. The PCR products were checked by electrophoresis and cloned into a T-vector (Promega, Madison, WI, USA) then sequencing was performed by Macrogen Co. (Seoul, Korea).

For spore production, the strains were cultured on ISP4 (10 g of starch, 1 g of K₂HPO₄, 1 g of MgSO₄, 1 g of (NH₄)₂SO₄, 1 g of CaCO₃, 1 mg of FeSO₄, 1 mg of MgCl₂, 1 mg of ZnSO₄, 18 g of Bacto agar per liter, final pH 7.2) media at 30°C for 5 days. Mature bacterial colonies on the plates were prepared as glycerol spore stocks. One mL of 20% glycerol was added on the plates and all the spores were rubbed using a sterile cotton stick. The cotton was pushed into a 15-ml syringe and suspension was expelled into the Eppendorf tube by replacing the syringes’ plunger. The glycerol spore stocks were adjusted OD₆₀₀ 0.2 and prepared at −80°C for the further experiments.

Selected anti-Psa strains were assessed three hydrolytic enzyme activities (cellulolytic, chitinolytic, and proteolytic) and six PGP activities (siderophore production, phosphate solubilization, indole-3-acetic acid (IAA) production, nitrogen fixation, attachment, and antibacterial activity) using solid media. Cellulolytic reaction media contained 4 g of KH₂PO₄, 2 g of CaCl₂, 2 g of NH₄Cl, 1 g of yeast extract, 1 g of carboxymethyl cellulose, 1 g of MgSO₄, 0.05 g of MnSO₄, 0.05 g of FeSO₄, 18 g of agar per liter (final pH 7-7.4). The chitinolytic reaction media contained 5 g of chitin, 5 g of yeast extract, 0.7 g of KH₂PO₄, 0.5 g of K₂HPO₄, 0.3 g of MgSO₄, 0.1 g of FeSO₄, 0.1 g of NaCl, 18 g of agar per liter and adjusted final pH 6.5-7.0 (Kim et al., 2017). Proteolytic reaction media contained 10 g of skim milk powder, 18 g of agar per liter (Deepthi et al., 2012). Bacterial spore stocks (20 μl) were inoculated with 8-mm diameter paper disks and incubated at 27°C for 5 days. Cellulolytic reaction plates were stained with 0.1% aqueous congo red solution, chitinolytic reaction plates were stained with 1% aqueous congo red solution for 15 min and both were washed completely by 1 M NaCl (Kim et al., 2017). Diameter of clear yellow (cellulolytic, chitinolytic) and transparent zones (proteolytic) were measured with three replicates. Statistical analysis was carried out (Tukey’s HSD test, P < 0.05).

For siderophore producing assay, chrome azurol S (CAS) agar media were prepared with three steps (Louden et al., 2011). First, Blue dye solution was prepared with three solutions. Solution 1 was dissolved with 0.06 g of CAS in 50 ml of ddH₂O. Solution 2 was dissolved with 2.7 mg of FeCl₃-6H₂O in 10 ml of 10 mM HCl. Solution 3 was dissolved with 73 mg of hexadecyltrimethylammonium bromide in 40 ml of ddH₂O. Solution 2 (9 ml) was mixed with Solution 1 and then Solution 3. Second, respective mixture solutions were prepared. Minimal media 9 (MM9) salt solution was dissolved with 15 g of KH₂PO₄, 25 g of NaCl, 50 g of NH₄Cl with 500 ml of ddH₂O. Glucose stock (20%) was dissolved with 20 g of glucose with 100 ml of ddH₂O. NaOH stock was dissolved with 25 g of NaOH in 150 ml of ddH₂O as final pH 12. Casamino acid solution was prepared overnight at 4°C with 3 g of casamino acid in 27 ml of ddH₂O and 30 ml of 3% 8-hydroxyquinoline in chloroform and filtered sterilize to remove any trace iron. Finally, CAS agar was prepared as follows. One hundred milliliters of MM9 salt solution was added to 750 ml of ddH₂O and dissolved with 32.24 g of PIPES (piperazine-N,N′-bis [2-ethanesulfonic acid]). Total 15 g of agar was added, autoclaved and cooled to 50°C. Thirty milliliters of sterilized casamino acid solution and 10 ml of 20% sterilized glucose stock was added. Blue dye solution (100 ml) was slowly mixed along the glass wall and poured in the plates. Bacterial spore stocks (20 μl) were inoculated on the CAS agar media and incubated at 27°C for 5 days.

For phosphate solubilizing assay, Pikovskaya’s medium was prepared (Pikovskaya, 1948). Pikovskaya’s medium contained 10 g of glucose, 0.5 g of (NH₄)₂SO₄, 0.2 g of KCl, 0.1 g of MgSO₄, 0.002 g of MnSO₄, 0.002 g of FeSO₄, 0.5 g of yeast extract, 5 g of Ca₃(PO₄), 20 g of agar per liter, final pH 6.8-7.0. Inoculated methods of the spore stocks and verified method of clear yellow zones were the same as the enzymes assay. For IAA production assay, Salkowski reagent method was employed (Patel and Parmar, 2013). First, the strains were incubated in tryptophan-containing and free 1/5 PDK broth (2 g of potato dextrose broth, 2 g of Bacto peptone per liter) at 30°C for 14 days as three replicates. Cell was centrifuged at 5,000 rpm for 10 min and the 1 ml of aliquots were transferred into 1.5-ml test tubes. The Salkowski reagent was prepared with 35% of HClO₄ and 0.5 M of FeCl₃. The aliquots were mixed with 2 ml of Salkowski reagent as 2:1 ratio. The color change to pink was observed for 25 min and measured by spectroscopy at 530 nm. For nitrogen fixation assay, semi solid NFb medium was produced according to Baldani et al. (2014). The NFb medium contained 5 g of malic acid, 0.6 g of K₂HPO₄, 0.4 g of KH₂PO₄, 0.01 g of MnSO₄, 0.05 g of MgSO₄, 0.02 g of NaCl, 0.002 g of Na₂MoO₄, 2 ml of Bromothymol blue (0.5% in alcohol) and 1.75 g of agar per liter final pH 6.6-7.0. NFb medium (5 ml) in test tubes were autoclaved and cooled. Each of the spore stocks (OD₆₀₀ 0.2) was vertically inoculated in the middle of NFb medium using a loop and incubated at 27°C for 7 days. Attachment abilities were verified using crystal violet dyeing methods with three replicates (De Jong et al., 2009). To evaluate adherent ability, the strains were cultured in 5 ml of minimal NMMP medium (2 g of (NH₄)₂SO₄, 5 g of casamino acids, 0.6 g of MgSO₄, 50 g of polyethylene glycol 6000, 1 ml of minor elements solution, 15 mM NaH₂PO₄, K₂HPO₄ buffer (pH 6.8), 0.5% glucose per liter) at 27°C for 10 days. Non-adherent cells were removed by washing with 5 ml of 1× PBS. Crystal violet solution (1%, 10 ml) was added to each tube and incubated at room temperature for 3-4 h. All solutions were removed using a pipette and each tube was completely dried at room temperature overnight. Crystal violet was solubilized with 2 ml of 95% high performance liquid chromatography grade ethanol. The turbidity was compared with control sample and OD₅₇₀ was measured by a spectroscopy. Statistical analyses were evaluated by Tukey’s HSD (P < 0.05) in Statistix (v. 8.0). Sporulation assays were tested on 1/5 PDK agar. The spore stocks (5 μl; OD₆₀₀ 0.2) were inoculated, grown at 30°C and sporulation abilities were daily examined for 7 days.

For the verification of colonization and Psa inhibition of the selected strains in planta, intergeneric conjugation between Escherichia coli and the selected Streptomyces sp. strains was performed as Paranthaman and Dharmalingam (2003). E. coli ET12567 (pUZ8002/pIJ10257) and E. coli ET12567 (pUZ8002/pSET152) were grown to OD₆₀₀ 0.4 in 10-ml LB broth (25 g of Luria-Bertani per 1 liter). E. coli was obtained by centrifugation, washed twice with fresh LB broth (10 ml) and resuspended in 500 μl of LB broth. Glycerol spore stocks of Streptomyces sp. (10 μl, OD₆₀₀ 1.0) were resuspended in 500 μl of 2× YET medium (yeast extract tryptone; 10 g of yeast extract, 16 g of tryptone and 5 g of NaCl per liter, final pH 6.8) and incubated at 50°C for 10 min to derive germination. Equal volume (500 μl) of the donor cells and the treated spore stocks were mixed, spread on an MS agar plate and incubated at 30°C for 16 h. The plates were overlaid with 1 ml of sterilized water containing 30 μg/ml of nalidixic acid and 80 μg/ml of hygromycin or 50 μg/ml of apramycin. The plates were incubated at 30°C for 3 days. Several colonies were picked and cultured on the MS agar containing 30 μg/ml of nalidixic acid and 80 μg/ml of hygromycin or 50 μg/ml of apramycin. Respective antibiotic resistance colonies were confirmed, and the colony was prepared as spore stocks for further studies.

Rhizosphere colonization assays were performed using apramycin resistant W1SF4 and hygromycin resistant W3SF9 strains. Bacterial cell densities were adjusted as 10⁶ colony-forming unit (cfu)/ml in 0.1% Tween 20 solution. Roots of kiwifruit seedlings (cv. Deliwoong, 12-week-old) were washed with tap water and the dipped in the bacterial suspension for 15 min. Untreated controls were used as the roots were dipped in sterilized 0.1% Tween 20 solution. Experimental plants were planted in a deepots (Stuewe & Sons, Inc., Tangent, OR, USA) containing garden soil. The plants were maintained in growth chamber at 25°C, 80% humidity, 12 h light photoperiod. Root colonization was measured by diluting the rhizosphere soils from 10⁻¹ to 10⁻⁵, spreading on 1/5 PDK agar with selective antibiotics and verifying the cfu/g at weekly intervals up to 6 weeks.

Leaf colonization assays were performed using apramycin resistant W1SF4, hygromycin resistant W3SF9 and KPB2. Each bacterial cell density was adjusted at 10⁷ cfu/ml with 10 mM MgCl₂ and 0.02% silwet L-77. Kiwifruits were planted with garden soil in the deepots and the bacterial suspensions were sprayed evenly three times with 2-day intervals onto both adaxial and abaxial sides of the leaves. Leaf colonization was measured by diluting leaves from 10⁻¹ to 10⁻⁷, spreading on 1/5 PDK agar with proper antibiotics and verifying the cfu/g. The deepots were maintained for 21 days and evaluated the bacterial strains colonization on leave.

For canker disease suppression assay, Psa (10⁷ cfu/ml) was sprayed on the leaves. The treated plants were maintained for 21 days. The disease index (DI; 0-5) was determined by observing the percentage of diseased leaves; 0 = no symptoms; 1 = trace to 5%; 2 = 5-15%; 3 = 15-30%; 4 = 35-70%; 5 = more than 70%. Statistical analysis was carried out in Statistix (v. 8.0) by Tukey’s HSD test (P < 0.05).

Genome sequencings of Streptomyces racemochromogenes W1SF4 (rhizosphere strain) and Streptomyces parvulus KPB2 (pollen strain) were performed using the PacBio (Pacific Biosciences) RSII platform (English et al., 2012). Single colony of the strains was cultured in PDK broth at 30°C for 5 days. Cells were harvested by centrifugation at 8,000 rpm for 10 min. Genomic DNA was extracted using the MG genomic DNA purification kit (Doctor Protein, Seoul, Korea). Concentration of genomic DNA was measured by NanoDrop spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA) and PicoGreen for screening quality control (QC) criteria. Genomic DNA (8 μg) was used for library preparation of 20-kb SMRTbell templates. The genomic DNA less than 17 kb was determined the actual size distribution by Bioanalyzer 2100 (Agilent, Santa Clara, CA, USA). The proper genomic DNA greater than 40 kb was purified using AMPurePB magnetic beads (Beckman Coulter Inc., Brea, CA, USA). Total 10 μl library was prepared using PacBioDNA template prep kit 1.0 (Thermo Fisher Scientific). SMRTbell templates were annealed using the PacBio DNA/Polymerase binding kit P6 and the 8 SMRT cells were sequenced using the PacBio DNA sequencing kit 4.0. SMRT cells using C4 chemistry and 240-min movies were captured for each SMRT cell using the PacBio RS II platform by Macrogen Co. Gene functions were assigned by Prokka (v 1.12b) after whole genome was assembled. Genome annotation was conducted by RAST version 2.0 (http://rast.nmpdr.org), antiSMASH bacterial version (https://antismash.secondarymetabolites.org/#!/start) for secondary metabolite analysis and NCBI (http://www.ncbi.nlm.nih.gov) for the accurate gene identification (Aziz et al., 2008; Blin et al., 2018).

Culturable bacteria were isolated from kiwifruit ecospheres (rhizosphere, sap, and pollen), total bacteria densities were higher rhizosphere, pollen and sap in order, 10⁹ cfu/g in rhizosphere soil, 10⁷ cfu/g in pollen and 10⁶ cfu/ml in sap, respectively. Of 1,175 bacterial isolates from rhizosphere, 768 isolates from sap and 1,080 isolates from pollen were collected as kiwifruit-associate bacterial library and tested anti-Psa activity (Supplementary Table 2). After the first anti-Psa screening, 56 isolates from rhizosphere, 156 isolates from sap and 106 isolates from pollen were selected. At second screening, 15 isolates from rhizosphere, 35 isolates from sap and two isolates from pollen showed anti-Psa activity. At the final third screening total 11 isolates (9 from rhizosphere, 2 from pollen) were selected as potential anti-Psa microbial agents. The 11 anti-Psa isolates were named accordingly to kiwifruit cultivar, replication number, isolated tissue origin, and location on library; W1SF3, W1SF4, W1SG3, W2SB7, W3SF9, K1SF1, K2SC9, K2SG8, K3SG10, KPB2, and KPP2. First letter W or K indicated kiwifruit cultivar Deliwoong or NHK0013. Number in second place indicated number of replicates. Third letter S or P indicated rhizosphere soil or pollen. Final alphabet and number indicated the storage location in 96-well plates. W1SF4 and W3SF9, both strains were isolated from rhizosphere, showed the largest Psa inhibition zones as 3.53 cm and 3.92 cm. Other strains ranged inhibition zones from 1.34 to 1.96 cm (Fig. 1).

All 11 anti-Psa strains were grown and sporulated on MS media. Phenotypic characteristic of the strains was distinguished as six types according to colony color, shape and color change around the colony (Supplementary Fig. 1). Type A produced powdery and less sporulation on white colony with ocher halo, type B presented powdery and abundant light pink colony, type C showed powdery and rough-surfaced white colony with brown halo and type D produced powdery and flat light pink colony. Type E showed powdery and smooth pink colony and type F produced powdery and smooth dusky pink colony. W1SF4, W3SF9, K2SC9, and K2SG8 were grouped as type A. KPB2 and KPP2 were belonged to type F. For type B (W1SF4), C (W3SF9), D (K2SC9), and E (K2SG8), a single strain for each type was assigned (Supplementary Fig. 1).

Sequence of 16S rRNA of the strains was compared with references strains in GenBank and all the strains showed 100% similarity with various Streptomyces species but fall to identified in species level. However, the strains can be identified in species level with atpD and recA sequences except W3SF9 (Table 1). The strains showed 98% to 100% similarity with reference strains and identified into species level according to their colony phenotypic types except the type C stain. The type A (W1SF4, W3SF9, K2SC9, and K2SG8) strains were identified as Streptomyces virginiae, B type was S. racemochromogenes, C type was Streptomyces sp., D type was Streptomyces nojiriensis, E type was Streptomyces zaomyceticus and F type was S. parvulus (Fig. 2).

The selected anti-Psa strains’ cellulolytic, chitinolytic, proteolytic, siderophore producing and antibiotic activities were evaluated (Table 2). Most of the strains showed cellulolytic, chitinolytic, proteolytic and siderophore production activities. However, S. virginiae (W1SF3, W1SG3, W2SB7, K1SF1, and K3SG10) and S. nojiriensis K2SC9 were not shown the proteolytic activity. S. zaomyceticus K2SG8 showed the accomplished proteolytic activity but not presented the chitinolytic and siderophore producing activities. S. racemochromogenes W1SF4, Streptomyces sp. W3SF9 and S. parvulus KPB2, KPP2 had outstanding effects for the cellulolytic, chitinolytic, proteolytic and siderophore producing activities. Prominently, Streptomyces sp. W3SF9 had accomplished antibacterial activity against Psa. Sporulation of S. racemochromogenes W1SF4, Streptomyces sp. W3SF9, S. nojiriensis K2SC9, and S. zaomyceticus K2SG8 were 30% in 3 days and 100% in 7 days, respectively. The sporulation of S. parvulus strains KPB2, KPP2 had shown only about 20% and S. virginiae strains W1SF3, W1SG3, W2SB7, K1SF1, K3SG10 were not sporulated at all for 7 days (Table 2). Through enzymatic and PGP abilities, S. racemochromogenes W1SF4, S. virginiae W2SB7, Streptomyces sp. W3SF9 from rhizosphere and S. parvulus KPB2 from pollen were condensed as anti-Psa strains. S. racemochromogenes W1SF4 had not only the anti-Psa activities but also had PGP and sporulation activities. S. virginiae W2SB7 had antibiotic and PGP activities but no sporulation ability. Streptomyces sp. W3SF9 had anti-Psa and PGP activities. S. parvulus KPB2 was the only species selected from pollen and had effects in all the antibiotics, PGP and sporulation (Table 2).

These 4 (W1SF4, W2SB7, W3SF9, and KPB2) strains were evaluated additional PGP assays. The strains showed phosphate solubilization, IAA production and nitrogen-fixing activities (Table 2). The four strains also demonstrated attachment abilities, but W2SB7 produced less spore. S. virginiae W2SB7 had several activities but was excluded due to less proteolytic activity and sporulation activity. Among the four strains, S. parvulus KPB2 was only one species from pollen. The three strains (S. racemochromogenes W1SF4, Streptomyces sp. W3SF9, and S. parvulus KPB2) were selected as probiotics for kiwifruit after evaluated various criteria, which were presented above.

Density of the strains on root and leaf was measured by plate counts with a proper selection antibiotic maker. In kiwifruit rhizosphere, strains W1SF4 and W3SF9 (rhizosphere strains) colonized at a density of 10⁷ cfu/g of soil up to 6 weeks (Fig. 3A). In phyllosphere, S. racemochromogenes W1SF4 colonized at a density of 10⁶ cfu/g, Streptomyces sp. W3SF9 were detected at a density of 10⁵ cfu/g and S. parvulus KPB2 (pollen strain) presented at a density of 10⁸ cfu/g on the kiwifruit leaves. The canker disease suppression abilities were also verified by disease symptoms and DI. Psa infected leaves were formed yellow or brown hallow as the disease symptoms. The symptoms were shown all areas of leaves on Psa treatments but minor or no symptom was observed on the anti-Psa strain treatments leaves (Fig. 4A). The Psa pathogen only treatment showed near DI 5 at 3-weeks after treatment, but all kiwifruit probiotics strains (W1SF4, W3SF9, and KPB2) significantly reduced the disease severity as under DI 1 (Fig. 4B).

Accession numbers of genome analyses for W1SF4 and KPB2 were presented in Supplementary Table 3. The genome of Streptomyces sp. W1SF4 comprised three contigs of 7,272,878 bp, 481,785 bp and 313,480 bp (Table 3, Fig. 5A-C). Contig 1 predicted 6,899 coding sequences (CDS), contig 2 for 505 CDS and contig 3 for 189 CDS. The GC contents of these components were 73.1%, 72.9%, and 74.5% respectively. The genome carried 216 tRNA and 63 rRNA. Total 25 putative antibiotic secondary metabolite gene clusters were predicted by antiSMASH. Similarity was above 60%, two lanthipeptides, one polyketide synthase (PKS), one melanin, two nonribosomal peptide synthetase (NRPS) and one siderophore, two terpenes biosynthesis gene clusters were presented in Table 4. Genome of Streptomyces sp. KPB2 comprised single linear chromosome of 8,082,236 bp and 7,362 predicted CDS (Table 3, Fig. 5D). This strain carried 54 tRNA, 196 rRNA and GC contents was 72.2%. Total 22 antibiotic gene clusters were predicted and major gene clusters were three NRPSs, three terpenes, two PKSs, one siderophore, one melanin and one ectoine (Table 5). Gene ontology information of the two strains was also presented in Supplementary Table 4 and 5.