Sweet potato tubers were collected from three months old plants during the autumn of 2015 representing 12 sweet potato growing sites in Nepal, six from subtropical and six from temperate regions. Sampling sites, climate and soil properties are presented in Table 1.

The sweet potato samples were washed with tap water, shade dried and kept at room temperature until the isolation of the endophytic bacteria. Soil samples collected from the same field during the spring of 2016 were air-dried and crushed to pass through a 2 mm sieve. The pH was measured using the glass electrode method with a soil: water ratio of 1:2.5 [13] . Total carbon (TC) and total nitrogen (TN) were determined by the dry combustion method using an NC analyzer (MT-700, J-Science, Kyoto, Japan). Available phosphorus (P) was

^a5 years average of maximum, minimum annual temperature and annual rainfall (http://www.dhm.gov.np/); ^bBased on USDA classification [26] ; ^cSoil P in mg kg⁻¹, and Soil C and Soil N in g kg⁻¹.

determined by Olsen’s bicarbonate method [14] .

The sweet potato samples were washed again with running tap water for 10 min. Each sample was cut transversely when its diameter was more than 10 mm otherwise cut longitudinally. Then, the cut surface was stamped on the modified MR agar medium [4] , and incubated for 2 days at 26˚C. The appeared colonies were grouped based on their morphologies and the representative colonies reflecting their relative abundance were purified for further analysis as endophytes.

The partial 16S rRNA genes of the isolated endophytic bacteria were amplified using the universal primers (fD1 and rP2) to the domain bacteria [15] . The PCR mixture was prepared by mixing MilliQ water, 10x reaction buffer, 10 mM dNTPs, Taq DNA polymerase (GENETBIO Inc., Daejeon, Korea), fD1 and rP2 primers together with the template. The PCR reaction was carried out with a pre-run at 94˚C for 3 min, 30 s at 94˚C, 30 s at 50˚C, 1 min at 72˚C for 30 cycles and final run at 72˚C for 5 min. The PCR products were sequenced as described by Adhikari et al. [10] . In brief, the respective PCR products were purified by using SOPE^TM resin (Edge Biosystems Inc. USA) and a Performa Dye Terminator Removal (DTR) Gel Filtration Cartridge (Edge Biosystems Inc. USA). Then, their nucleotide sequences were analyzed by an ABI Prism, 3100-Avant-100D2 (3130 xl/Genetic Analyzer, Hitachi, Tokyo, Japan). Close relatives for each isolate was assigned using the data base (https://www.ddbj.nig.ac.jp/index-e.html) by a BLAST search [16] .

Correlation between bacterial class compositions and the environmental parameters (Table 1) of the sampling sites was analyzed by the principal component analysis (PCA). Bacterial class compositions as expressed by relative percentage were used for the calculation.

Sixty representative isolates of 34 genera in 6 classes were selected by their phylogenies (Table S1 & Figure S1) and used for the characterization of their plant growth promoting traits, antagonistic effect and endophytic traits.

For indole-3-acetic acid (IAA) production assay, the isolates were cultivated in the modified MR liquid media supplemented with 200 μg/ml of L-tryptophan at 26˚C for 3 days with shaking (150 rpm). After centrifugation at 8000 g for 15 min, the supernatant was applied for quantification of IAA according to the method described by Gordon and Weber [17] .

For detection of nitrogen fixing gene (nifH), primers PolF and PolR which were designed to match a broad range of bacterial nifH gene [18] were used for PCR. The PCR components and conditions were as described by Adhikari et al. [10] . Nitrogenase activity of the nifH gene containing isolates was evaluated using the acetylene reduction assay (ARA). The isolates were cultivated in N-free modified MR media [19] for 3 days at 26˚C with shaking (150 rpm). Then, the culture was washed and suspended in sterile distilled water at OD_660nm 0.2. Then, 50 μl of the suspension was inoculated on a slant of semi-solid (1.3%) N-free modified MR agar media in 60 ml vials in triplicate. The vials were sealed with a butyl-rubber and an aluminium stopper, and 10% of the headspace volume was replaced with pure acetylene. Vials without acetylene and without inoculants served as controls. Ethylene concentrations in the vials were measured after 3 days of incubation in dark at 28⁰C, using a gas chromatograph (Shimadzu GC-14B; Shimadzu, Kyoto, Japan) equipped with a flame ionization detector and Porapak N (50/80 mesh; GL Sciences, Tokyo, Japan).

For assay of antagonistic effects, the point inoculation method [20] was used against the fungal pathogen (Pythium ultimum var ultimum Strain OPU744). Briefly, each test strain was streaked on the PDA plate (dextrose: 1 g l⁻¹, peptone: 5 g l⁻¹, yeast extract: 2.5 g l⁻¹ and agar: 15 g l⁻¹). After one day of cultivation at 28˚C, a 5 mm of fungal mycelial disk was placed at approximately 30 mm from the bacterial streaked line and continuously cultivated for 7 days. The corresponding fungal disk without endophyte strain served as control. To test antagonistic effect of endophytes against bacterial pathogens, each test isolates were streaked on a half part of PDA plates and incubated for 2 days at 28˚C. Then each bacterial pathogens (ECa: Erwinia chrysanthemi Strain NARCB200126, AZ9702, causing stem and root rot in sweet potato, ECb: Erwinia chrysanthemi Strain E7725, causing stem rot in potato and ECc: Erwinia chrysanthemi Strain Ech T5-2, causing root rot in Taro); was streaked approximately 5mm to the endophytes and incubated at 28˚C for 9 days. For both assays, antagonistic effects were categorized based on the distance between the test isolates and the pathogen as follows: no (0 mm), weak (1 - 3 mm) and strong (>3 mm) activities.

For the cellulase assay, the isolates were spotted on a carboxymethyl cellulose (CMC) agar medium [21] . Plates were incubated at 28˚C for 6 days. The clear zone around the point of inoculation was examined by staining remaining CMC with Congo red [22] . For the pectinase assay, the test strains were spotted on a nutrient agar (DIFCO laboratories, USA) medium supplemented with 0.5% pectin and incubated at 28˚C for 3 days, then remaining pectin was stained with cetyltrimethylammonium bromide (CTAB) to visualize the clear zone around the bacterial culture [23] . For both assays, the activities were categorized based on the clear zones around the bacterial colony as follows: no (0 mm), weak (1 - 3 mm) and strong (>3 mm) activities.

The endophytes within the same location were selected for the inoculation experiment based on their phylogenies (Table S1). Each strain was cultivated separately in nitrogen containing modified MR liquid medium at 26˚C with shaking for 5 days. The cells were harvested by centrifugation (10,000 rpm for 10 minutes at 15˚C), washed twice and suspended with sterile distilled water, and OD₆₆₀ was adjusted to 0.2. Then, the cell suspensions of the same location were mixed together to make the inoculants in 10 ml final volume. The experiment was conducted using a Leonard jar [24] . The upper pot was filled with water-soaked sterile vermiculite and the lower pot was supplied with 150 ml of sterile 1/5N plant medium [25] . The cotton wick was set to connect the upper pot and the lower reservoir. The whole pot was autoclaved before use.

Micro-propagated sweet potato plantlets cv. Koukei (3 - 4 leaves) was used for the experiments. Two consecutive experiments were conducted in duplicate. After measuring initial fresh weight, vine and root lengths (except for root length in the first experiment due to the absence of roots), the root part was dipped into the inoculants for 3 min and transplanted to the sterile pot, and 5 ml inoculant was poured on the vermiculite around the plant. The inoculated plants were aseptically grown in a plant growth chamber (LH240S, Nippon medical and chemical instruments co., ltd, Japan) with a 14 hour photo period, 28˚C/25˚C (day/night) at 7000 lux, which was provided by white fluorescent tubes. The same plant media without nitrogen was supplied to the bottom pot, as per the requirement. For control, sterile distilled water was inoculated. First experiment was conducted with nine inoculants and the better performing six inoculants were used in the second experiment.

The plants were harvested at 30 days after inoculation. The whole plant was carefully pulled to avoid damage and shaken to release loosely attached vermiculite, then strongly adhered vermiculite was manually removed with tweezers. After blotting excess moisture from the roots with absorbent paper, whole plant fresh weight, vine and root lengths were measured. Then, nitrogenase activity for the fresh roots was assayed by ARA using 100 ml vial. Uninoculated plant roots with/without acetylene served as controls.

All the statistical analyses were carried out using SPSS software package version 16.

All sequences are deposited in the DNA Data Bank of Japan (DDBJ) under accession numbers LC389337 to LC389579 (16S rRNA) and LC389580 to LC389582 (nifH gene).

Different morphologies were observed among the appeared bacterial colonies, ranging from one to nine morphologies in Gulmi and Rupandehi locations, respectively. Two to 43 endophytic bacterial isolates per location, making a total of 243 isolates, were isolated and examined (Table S1).

Based on partial 16S rRNA gene sequencing analysis, 243 endophytic bacterial isolates were assigned to their close relatives, belonging to 34 bacterial genera in 6 classes (Table 2). Among the classes, Bacilli represented the highest relative abundance (28%), and Bacillus sp. was the most dominant genus (25%), followed by Gammaproteobacteria (22%)/Enterobacter sp. (5.3%), Betaproteobacteria (17%)/Burkholderia sp. (8.6%), Actinobacteria (16%)/Microbacterium sp. (6.8%), Alphaproteobacteria (14%)/Rhizobium sp. (6.3%) and Flavobacteriia (4.4%)/Flavobacterium sp. (4.4%).

Endophytic bacterial genera were distributed unevenly among the sweet potato samples (Table 2). Four bacterial genera commonly detected in at least five sampling sites were applied to the distribution analysis (Table 3). Enterobacter sp. and Microbacterium sp. were detected frequently in nutrient rich (copiotrophic) and poor (oligotrophic) soils, respectively. Bacillus sp. showed acidophilic nature while Rhizobium sp. and Microbacterium sp. were alkaliphilic. Similarly, Enterobacter sp. showed neutralophilic property. On the other hand, distribution of these four genera was unaffected by the temperature conditions.

PCA of the environmental parameters explained 41.3% and 40.6% of the variation in the first and second principal component factors, respectively, and showed that there are approximately two groupings, first being the high temperature and alkaline soils (Rupandehi, Banke-a and Banke-b) and second with the others (Figure 1(a)). PCA of bacterial class compositions explained 31.0% and

^a: Low P (4.8 - 37 mg kg⁻¹ soil) and high P (87 - 379 mg kg⁻¹ soil), low C (4 - 11 g kg⁻¹ soil) and high C (13 - 20 g kg⁻¹ soil), low N (0.3 - 0.8 g kg⁻¹ soil) and high N (1.0 - 1.7 g kg⁻¹ soil), acidic (pH 5.5 - 5.9), neutral (pH 6.2 - 7.3) and alkaline (pH 8.0 - 8.4), high temperature (34˚C - 39˚C) and low temperature (28˚C - 32˚C).

25.2% of the variation in the first and second principal component factors, respectively, and revealed that the endophytic bacterial composition did not group as the environmental conditions (Figure 1(b)).

Eighty three percent of the bacterial strains presented at least one of the characteristics examined. Within all the strains, 57% produced IAA, 5.0% had nifH gene and showed ARA activity, 37% and 2.0% possessed antagonistic effect against the bacterial and the fungal pathogens, respectively. In addition, 17% and 8.0% showed cellulase and pectinase activities, respectively (Table 4).

Proportions of the bacterial class representing the examined traits were different (Table 4). IAA production was detected in the strains from Bacilli, Actinobacteria, Alpha- and Gamma-proteobacteria classes, and ARA activity was from Beta- and Gamma-proteobacteria. All the classes possessed antagonistic effect against the bacterial pathogens, while only Bacilli class showed the effect against the fungal pathogen. Almost all the bacterial classes showed cellulase activity while Alpa-, Beta- and Gamma-proteobacteria classes had pectinase activity.

Bacterial strains exhibiting at least one of the following plant growth promoting or endophytic traits are presented in Table 5: high IAA production (≥30 μg/ml), ARA activity, strong antagonistic and cellulase/pectinase activities. Among the bacterial strains 8 of them showed higher IAA production potential with Ban-b 4 being the highest (65 μg/ml). Strains belonging to the same genus showed varying levels of IAA producing ability. In addition, Sal 1, Sal 6 and Rol 5 had nifH gene and showed the ARA activity with 54.5 ± 7.3 nmol C₂H₄/h/vial, 39.9 ± 1.9 nmol C₂H₄/h/vial and 8.9 ± 0.8 nmol C₂H₄/h/vial, respectively.

Likewise, 11 bacterial strains showed strong antagonistic effect against at least one pathogen tested. Among them, Chi 2 and Gul 1 possessed strong activity against the tested bacterial pathogens and the latter showed strong activity against the fungal pathogen assayed. However, these two isolates did not show IAA producing ability. On the other hand, 5 bacterial strains showed cellulase activity while Ban-b 6 showed both cellulase and pectinase activity.

Fresh weight (g), vine and root lengths (cm) were considered for the assessment of plant growth promotion. Plantlets used in the experiments were non-uniform in size, and this might affect the parameters. So, times increase as compared to the control were used for the assessment of the plant growth promotion of the inoculants.

In the first experiment, the inoculated sweet potatoes showed higher values than control (Table 6). Mixture of isolates from Salyan, Palpa, Banke gained 3.18 - 3.51 times their initial weight whereas it was 1.6 in control. Likewise, gain in vine length ranged from 1.33 - 1.50 times for Kavre, Chitwan, Banke and Salyan inoculants, while it was 1.17 in control. Finally, the roots were longer in almost all inoculated plants than control.

Further, we selected six inoculants based on the first experiment and again evaluated. In the second experiment, the growth promoting effects were observed in fresh weight and vine length, but root lengths were shorter in the inoculated plants than control (Table 6). Among all the plants assessed, only Salyan

^aIndole-3-acetic acid (μg ml⁻¹); ^bAcetylene reduction activity (nmol/h/vial); ^cErwinia chrysanthemi NARCB200126, AZ9702; ^dErwinia chrysanthemi E7725; ^eErwinia chrysanthemi T5-2; ^fPythium ultimum var ultimum OPU744.

^aIndole-3-acetic acid (μg ml⁻¹) optical density⁻¹; ^bAcetylene reduction activity (nmol/h/vial); ^c-, + and ++ denote no, weak and strong activities, respectively; ^dErwinia chrysanthemi NARCB200126, AZ9702; ^eErwinia chrysanthemi E7725; ^fErwinia chrysanthemi Ech T5-2; ^gPythium ultimum var ultimum OPU744.

^aData from one replication is considered for Sunsari and Chitwan in first and second experiment respectively, as one replication plant was dead; ^bFinal root length (cm).

isolates inoculated plants showed ARA activity (0.09 nmol/h/g).