Soil samples were collected from forests of the Qilian Mountains, Gansu province, China (38°25′32″N, 99°55′40″E, 3,150 m). Spruce was the dominant tree species at the altitude of 3,150 m in the Qilian Mountains. The soil sample was serially diluted with sterile 0.9% NaCl (w/v) solution, and dilutions were spread on tryptone soya agar (TSA). All plates were incubated aerobically at 25°C for 7 days. Morphologically different single colonies were randomly picked and further purified. Finally, the purified isolates were preserved as a glycerol suspension (20%, v/v) at −80°C.

The genomic DNA of the isolate was extracted by Bacterial Genomic DNA Extraction kit (TianGen Biotech Co., Ltd., Beijing, China) according to manufacturer’s instructions. The 16S rRNA gene was amplified by PCR using a pair of universal primers, 27F and 1492R, as previous described (Li et al., 2020). The RNA polymerase β-subunit (rpoB) gene, an iconic housekeeping gene of the genus Paenibacillus, was amplified with primers rpoB 1698F (5′-AACATCGGTTTGATCAAC-3′) and rpoB 2041R (5′-CGTTGCATGTTGGTACCCAT-3′) (Yang Y.J. et al., 2018). The nitrogenase reductase (nifH) gene was amplified using the primers POLF (5′-TGCGAYCCSARRGCBGGYATCGG-3′) and POLR (5′-ATSGCCATCATYTCRCCGGA-3′) (Menéndez et al., 2017). The PCR product was purified by PCR purification kit (Sangon Biotech Co., Ltd., Shanghai, China) according to the manufacturer’s instructions. Cloning of the 16S rRNA gene was executed using a pMD 19-T Vector Cloning kit (Takara Bio., Inc., Otsu, Japan). Sequencing was performed by the Sanger method (Beijing AUGCT DNA-SYN Biotechnology Co., Ltd, Beijing, China). Then, the almost-complete 16S rRNA, rpoB, and nifH gene sequences were compiled with the program DNAMAN (version 8.0; Lynnon Biosoft, San Ramon, CA, United States) (Saitou and Nei, 1987). The EzTaxon-e server² (Yoon et al., 2017) was used to calculate the levels of sequence similarity between strain LC-T2^T and related type strains available in GenBank³ (Sayers et al., 2020). The phylogeny of 16S rRNA, rpoB, and nifH sequences was reconstructed by the neighbor-joining (NJ) (Saitou and Nei, 1987), maximum-likelihood (ML) (Felsenstein, 1981), and maximum-parsimony (MP) methods (Fitch, 1971) with MEGA 7.0 program (Kumar et al., 2016). Evolutionary distances were calculated using the Kimura’s two-parameter model, and bootstrap analysis was used to evaluate the tree topology by performing 1,000 replications (Felsenstein, 1985).

The draft genome shotgun project was sequenced using paired-end sequencing technology with the Illumina NovoSeq-PE150 platform (Novogene Biotech Co., Ltd., Tianjin, China). High-quality genomic DNA was carried out using Bacterial Genomic DNA Extraction kit (TianGen Biotech Co., Ltd., Beijing, China) according to standard protocol. The sequencing generated 1-Gb clean data. A de novo assembly of the reads was carried out using SOAPdenovo (version 2.04). The completeness of microbial genomes was assessed using the bioinformatics tool CheckM (Parks et al., 2015). The complete 16S rRNA gene sequence of strain LC-T2^T was annotated via the RNAmmer 1.2 server (Lagesen et al., 2007) from the genome. The draft genome was annotated using the NCBI Prokaryotic Genome Annotation Pipeline (PGAP) (Tatusova et al., 2016; Haft et al., 2018). The predicted coding sequences (CDSs) and functional annotation were generated from the National Center for Biotechnology Information (NCBI) non-redundant database, Kyoto Encyclopedia of Genes and Genomes (KEGG), Cluster of Orthologous Groups of proteins (COG), and Gene Ontology (GO) databases. DNA G + C content was calculated from the draft genome sequence. BLAST algorithm (ANIb) and the MUMmer ultra-rapid aligning tool (ANIm) were used to calculate average nucleotide identity (ANI) by the JSpecies software tool available at the webpage.⁴ The digital DNA–DNA hybridization (dDDH) between strain LC-T2^T and related reference strains was calculated by Genome-to-Genome Distance Calculator 2.1 (GGDC).⁵

The morphological, physiological, and biochemical characterizations such as growth in different bacteriological media, temperature, pH and NaCl concentrations, the Gram reaction, motility, oxidase, catalase, hydrolysis of Tween 80, DNA, casein, starch, and cellulase were carried out according to Li et al. (2020). Biochemical features were performed using the API 20NE, API ZYM, and API 50CH systems (bioMérieux). GENIII MicroPlates (Biolog) were used to check the utilization of 71 carbon sources as described by the manufacturer’s instructions. Paenibacillus donghaensis KCTC 13049^T, a Xylan-degrading bacterial strain isolated from east sea sediment, and Paenibacillus odorifer JCM 21743^T, a nitrogen-fixing strain isolated from wheat roots, were used as reference strains for comparative taxonomic characteristics (Berge et al., 2002; Choi et al., 2008). The two reference strains were obtained from the Korean Collection for Type Cultures (KCTC) and the Japan Collection of Microorganisms (JCM), respectively. Cells of strain LC-T2^T and the reference strains cultured on Reasoner’s 2A (R2A) agar at 28°C were used for biochemical feature tests. For measurement of nitrogenase activity, strain LC-T2^T and reference strains were grown on nitrogen-free medium (Zhuang et al., 2017). After 48 h at 28°C, strains were incubated in culture bottles with 10% (v/v) acetylene in air for 2 h and then analyzed for ethylene production by 450-GC gas chromatography (Berge et al., 2002).

For chemotaxonomic analysis, cells of strain LC-T2^T and reference strains were routinely cultivated on R2A agar at 28°C and harvested at the mid-exponential growth phase. The fatty acid profiles were analyzed and identified by using the Microbial Identification System (Sherlock version 6.1; midi database, TSBA6) after saponification, methylation, and extraction, according to standard procedures (Sasser, 1990). The polar lipids were extracted and separated by a chloroform/methanol system and one- and two-dimensional thin-layer chromatography (TLC) as described previously (Minnikin et al., 1984; Kates, 1986). Total lipids were detected using molybdatophosphoric acid, aminolipids were detected using ninhydrin reagent, phospholipids were detected using molybdenum blue reagent, and glycolipids were detected using naphthol/sulfuric acid reagent (Minnikin et al., 1984; Kates, 1986). Respiratory quinones were extracted and purified from lyophilized cells, then analyzed by high performance liquid chromatography (HPLC) according to Collins’ method (Collins, 1985).

Plate and pot experiments were used to evaluate plant growth-promoting capabilities of strain LC-T2^T. Double-sterile distilled water (DDW) served as control. Escherichia coli strain DH5α and commercial B. amyloliquefaciens strain GB03 served as positive control. Arabidopsis seeds were surface sterilized with 70% ethanol for 3 min, washed with DDW several times, followed by 1% sodium hypochlorite for 10 min, finally thoroughly washed with DDW for 8–10 times, and then planted on one side of specialized plastic Petri dishes (100 × 15 mm) that contained a center partition; both sides contain half-strength Murashige and Skoog (MS) solid medium with 0.8% (w/v) agar and 1.0% (w/v) sucrose. Seeds were vernalized for 2 days at 4°C in the absence of light. Bacterial suspensions were prepared according to previously described methods (He et al., 2018). Cells were harvested from R2A plates, put into DDW to yield 1.0 × 10⁹ colony forming units (CFU) ml^–1 as determined by optical density, and serially diluted with plate counts. Then, 10 μl of bacterial suspension was spotted at one side of the Petri dish and 2-day-old Arabidopsis seedlings were planted on the other side of the Petri dish. Seedlings were grown under growth chamber (Panasonic, Japan) with a 16/8-h light/dark cycle under 200 μmol m^–2 s^–1 total light intensity, a temperature of 22 ± 2°C, and a relative humidity of 50–55%. The root system was scanned by an EPSON scanner, and morphological parameters were analyzed using the root analysis system WinRHIZO (v5.0, Regent Instruments, Quebec, QC, Canada) after 14 days. In pot experiments, white clover (T. repens L.) seeds (presented by Wanhai Zhou at Gansu Agricultural University, China) were surface sterilized for 1 min in 70% ethanol followed by 10 min in 2% sodium hypochlorite; then, seeds were rinsed with sterile water for 10 times and germinated in filter paper for 3 days. The seedlings with uniform growth were transferred to a plastic pot (diameter 9 cm, depth 10 cm) containing autoclave-sterilized commercial vermiculite–soil mixture and watered with modified half-strength Hoagland’s solution three times per week. White clover seedlings were inoculated with 2 ml of prepared bacterial suspension culture as bacterial treatments or the same volume of DDW as control. Thirty-day-old plants were harvested for plant growth and physiological index measurements. When sampling, the rhizosphere soil samples were collected from the surface of root, and the culturable bacteria in rhizosphere were isolated again by multiple-dilution method to verify the effective inoculation.

Results of the growth and physiological parameters were showed as means with standard errors (n = 6). Statistical analysis was assessed by one-way analysis of variance (ANOVA) using SPSS statistical software (Ver. 19.0, SPSS Inc., Chicago, IL, United States). Duncan’s multiple range test was executed to detect a difference between means at a significance level of P < 0.05.

The complete 16S rRNA gene sequence (1,546 bp) was obtained from draft genome (GenBank accession number: OK058271). Comparative analysis built on 16S rRNA gene sequence revealed that strain LC-T2^T was phylogenetically affiliated to the genus Paenibacillus in the family Paenibacillaceae. On the basis of phylogenetic analysis, the highest level of similarity was found between strain LC-T2^T and P. donghaensis KCTC 13049^T (97.4%), followed by P. odorifer JCM 21743^T (96.8%) and other recognized members of the genus Paenibacillus (<96.7%). In the neighbor-joining phylogenetic tree, strain LC-T2^T fell within the cluster comprising the Paenibacillus species and formed a distinct genetic lineage with P. donghaensis KCTC 13049^T (Supplementary Figure 1) and likewise in the tree based on the ML and MP methods (data not shown). The rpoB gene fragment of strain LC-T2^T (GenBank accession number: OK094314) shared 87.2% sequence identity with P. donghaensis KCTC 13049^T and less than 84.8% identity with other members of the genus Paenibacillus (Figure 1). These data further confirmed that target lineage was belonging to the genus Paenibacillus and closely clustered with P. donghaensis KCTC 13049^T. The comparison of the nifH gene sequence of strain LC-T2^T (GenBank accession number: OK094315) with those of the type strains also showed that P. donghaensis KCTC 13049^T was the most closely related living species, with a similarity value of 81.4%, followed by P. odorifer JCM 21743^T (80.1%). The remaining available nifH sequences of the type species of the genus Paenibacillus showed less than 80% similarity to strain LC-T2^T. The phylogenetic analysis of nifH indicated that strain LC-T2^T clustered with P. donghaensis KCTC 13049^T and was phylogenetically divergent from the cluster of any recognized species of the genus Paenibacillus (Supplementary Figure 2).

Draft genome sequencing of strain LC-T2^T (accession number: WJXB00000000) yielded a length of 7,082,651 bp with 35 contigs (total number > 500 bp) and N50 value of 657,675 after assembly. All contigs were larger than 536 bp, and the largest was 1,092,698 bp. The sequencing coverage was about ×200. A total of 6,236 genes were predicted, out of which 5,978 were protein-coding genes, 112 genes for RNA, and 146 pseudo genes (Supplementary Table 1). The DNA G + C content of strain LC-T2^T was 46.0 mol%, which fell within the range given for species of the genus Paenibacillus (Yang D. et al., 2018). The pairwise ANIb, ANIm, and dDDH values between the genome of LC-T2^T and four genomes of related species were 72.6–77.2, 83.3–84.3, and 20.4–23.0%, respectively. These values were obviously lower than the critical value of genomic species identification. The detailed results are displayed in Table 1.

Strain LC-T2^T was isolated from high-altitude spruce forests in the Qilian Mountains. As shown in Table 2, the genome of strain LC-T2^T contained a large number of genes that were related to plant growth and habitat adaptation, such as 10 genes coding for phosphate solubilization (pyk, ppc, ackA, citA, citZ, aroK, ldh, phoP, phoR, and nudC), four for auxin biosynthesis (trpA, trpB, trpC, and trpS), one for nitrogen fixation (nifH), and three for other processes of growth promotion (speA, ilvH, and ilvB), suggesting that strain LC-T2^T had the ability to promote plant growth. The genome of strain LC-T2^T contained several genes associated with secretion systems, biofilm formation, or motility. For instance, three genes responsible for flagellar motility (motA, motB, and swrC) and five genes responsible for chemotaxis (cheA, cheY, cheR, cheB, and cheW) indicated that strain LC-T2^T could get attracted to or move toward nutrients and interact with plants. Additionally, extreme conditions, such as low temperature, hypoxia, alpine, strong ultraviolet, erosive forces, and thaw–freezing cycles, prevailed in the Qilian Mountains at high altitudes and shaped abundant extreme microorganisms. The genome of strain LC-T2^T is well equipped with several genes that could alleviate the reactive oxygen species. Some genes responsible for superoxide dismutase [Mn] (sodA), superoxide dismutase [Fe] (sodF), and catalase (katA), demonstrated that strain LC-T2^T could cope with rhizosphere oxidative environments. Notably, several genes of strain LC-T2^T genome responded to extreme temperature at the height of 3,150 m in the Qilian Mountains. Genes coding for cold shock (cspA) and heat shock (Hsp20) showed that strain LC-T2^T was able to adapt the temperature variation. In accordance with the data presented above, the draft genome analysis provided insights into the genetic features to support its plant-associated lifestyle and habitat adaptation.

The cell of strain LC-T2^T was aerobic, Gram-negative (Supplementary Figure 3A), rod-shaped (4.2–4.5 × 0.6–0.7 μm) (Supplementary Figure 3B), and motile via peritrichous flagella (Supplementary Figure 3C). Colonies of strain LC-T2^T on R2A agar were white, round, and smooth with approximately 0.5–1.5 mm in diameter after culture at 28°C for 3 days. It was able to grow aerobically at 4–32°C (optimum at 25–28°C), at pH 6.0–11.5 (optimum at 8.0–8.5), and with 0–1.5% (w/v) NaCl (optimum at 0%). Strain LC-T2^T and the reference strains were positive for catalase and reduction of nitrate to nitrite, but they were negative for oxidase and hydrolysis of DNA, Tween 80, and cellulose. The detailed differential physiological and biochemical characteristics of strain LC-T2^T and its closest type strains of the genus Paenibacillus are given in Table 3 and Figure 2. Strain LC-T2^T was distinguished from the reference strains in API 20NE test strips: assimilation of glucose, mannitol, and N-Acetyl-glucosamine. Strain LC-T2^T also differed from the closely related species in API ZYM test strips: cystine arylamidase, α-chymotrypsin, and acid phosphatase. Meanwhile, strain LC-T2^T also distinguished from the reference-type species in API 50CH test strips: D-ribose and methyl-α-D-glucopyranoside, N-acetyl-glucosamine, and inulin test. In the aspect of nitrogenase activity, the amount of strain LC-T2^T and reference strains, P. donghaensis KCTC 13049^T and P. odorifer JCM 21743^T, that could reduce acetylene to ethylene were 19.7, 15.5, and 25.4 (nmol C₂H₄) (mg protein)^–1 h^–1, respectively (Table 3).

The predominant cellular fatty acids (>10.0% of total fatty acids) of strain LC-T2^T was identified as anteiso-C_15:0 (56.5%) and C_16:0 (12.3%) (Figure 2). The fatty acid profile of strain LC-T2^T was similar to the reference strains, and all three species also contained anteiso-C_15:0 (40.3–56.5%) and C_16:0 (12.3–22.5%) as their major fatty acid. Moreover, the proportion of anteiso-C_15:0 of strain LC-T2^T was 2.4- and 1.2-fold higher than that of P. donghaensis KCTC 13049^T and P. odorifer JCM 21743^T, respectively, whereas the content of C_16:0 of strain LC-T2^T was nearly two-fold lower than that of the reference strains (Figure 2). The polar lipid pattern of strain LC-T2^T was dominated by the presence of large amounts of diphosphatidylglycerol (DPG) and phosphatidylethanolamine (PE) and small amounts of phosphatidylglycerol (PG) and several unidentified ingredients as follows: three unidentified phospholipids (PL1–3), two unidentified aminophospholipids (APL1–2), and one unidentified glycolipid (GL1) (Supplementary Figure 4). MK-7 was detected as the only respiratory quinone in strain LC-T2^T.

The apparent growth differences of Arabidopsis were observed between LC-T2^T exposure and the other three treatments after 2 weeks of plant growth (Figure 3A). The significant differences were mainly observed on plant roots. Specifically, the total root length was significantly greater for LC-T2^T-exposed roots (P < 0.05) by 61.0, 50.4, and 36.7% compared to control, DH5α, and GB03 exposure, respectively (Figure 3B). The highest root surface area and root projection area were also observed from exposure to LC-T2^T VOCs. The root surface area was increased (P < 0.05) by 61.3, 53.1, and 28.8% (Figure 3C), and the root projection area was enhanced (P < 0.05) by 61.3, 53.1, and 28.7% (Figure 3D) compared to control, DH5α, and GB03 exposure, respectively. The root fork numbers was increased over 1.9-fold (P < 0.05) with LC-T2^T VOCs compared to control and DH5α exposure, respectively. However, the root fork numbers of LC-T2^T were little lower than GB03 exposure (Figure 3E).

The influence of strain LC-T2^T on the growth of white clover was further assessed. Shoot height was increased by 31.5 (P < 0.05), 42.7 (P < 0.05), and 7.3% compared to control, DH5α, and GB03 treatments, respectively (Figure 4C), and root length was increased by 24.4 (P < 0.05) and 10.9% compared to control and DH5α treatments, respectively (Figure 4D).

Plants inoculated with LC-T2^T had a higher biomass than that of control, DH5α, and GB03. Shoot fresh weight was raised with the LC-T2^T group by 147.5 (P < 0.05), 117.6 (P < 0.05), and 2.0% (Figure 4E) and shoot dry weight was increased with the LC-T2^T group by 128.6 (P < 0.05), 127.1 (P < 0.05), and 10.8% compared to control, DH5α, and GB03 treatments, respectively (Figure 4F). Likewise, the root fresh weight of LC-T2^T-inoculated plants was about 216.2 and 89.8% (Figure 4G) and the root dry weight of LC-T2^T-inoculated plants was about 104.7 and 86.5% (Figure 4H) higher than those of the control and DH5α treatments, respectively (P < 0.05). However, the root fresh weight and root dry weight of LC-T2^T were a little lower compared to GB03 exposure (Figures 4G,H).

Strain LC-T2^T can also enhance the accumulation of plant biomass by root activity and chlorophyll content. Root activity was improved with the LC-T2^T group by 57.7 and 60.2% compared to control and DH5α treatments, respectively (P < 0.05; Supplementary Figure 5A). The content of chlorophyll a of LC-T2^T-inoculated plants was increased by 25.9 (P < 0.05), 27.1 (P < 0.05), and 1.8% compared to control, DH5α, and GB03 treatments, respectively (Supplementary Figure 5B), and the content of chlorophyll b was improved by 2.1- (P < 0.05), 1.2- (P < 0.05), and 1.0-fold with LC-T2^T treatment compared to control, DH5α, and GB03 treatments, respectively (Supplementary Figure 5B).

In addition, the photosynthetic rate was enhanced with the LC-T2^T group by 13.7 and 26.5% compared to control and DH5α treatments, respectively (Supplementary Figure 5C). The transpiration rate was raised by 1.2- and 1.5-fold (P < 0.05) with LC-T2^T-inoculated plants (Supplementary Figure 5D), and the stomatal conductance was increased by 17.4 and 39.1% (Supplementary Figure 5E) compared to control and DH5α treatments, respectively. The photosynthetic rate, transpiration rate, and stomatal conductance of LC-T2^T-inoculated plants were slightly lower than those of GB03 treatment, and differences were not statistically significant (Supplementary Figures 5C–E). The water-use efficiency was improved by 1.8- (P < 0.05), 1.2-, and 1.2-fold with LC-T2^T treatment compared to control, DH5α, and GB03 treatments, respectively (Supplementary Figure 5F).

Culturable bacterial strains have been isolated from the rhizosphere after 20 days of inoculation to verify the existence of inoculated strains. The number of genera from LC-T2^T-inoculated rhizosphere soil was 1.7-fold higher than that of control and DH5α, respectively. In the diversity and number of species, LC-T2^T-inoculated soil was also superior to control and DH5α-inoculated soil (Table 4). Some isolates were similar to strain LC-T2^T and GB03, respectively, except for DH5α treatment (Supplementary Tables 2, 3). The species in the genus Bacillus and Pseudomonas were also isolated from all inoculated soils, whereas a higher abundance of Rhodococcus qingshengii was found in control and DH5α-inoculated soils.

Paenibacillus monticola (mon.ti’co.la. L. n. mons, -ntis mountain; L. suff. -cola, inhabitant; N.L. masc. n. monticola, living in the mountains).

Cells are Gram-stain-negative, rod-shaped (4.2–4.5 × 0.6–0.7 μm) and motile by means of peritrichous flagella. Colonies of strain LC-T2^T were white, round, and smooth with approximately 0.5–1.5 mm in diameter after culture at 28°C for 3 days on R2A agar. The isolate grew well on R2A agar, ISP 2 agar, PYG agar, and TY agar and weakly on NA and TSA, but no growth occurs on MA, LB agar, and MacConkey agar. Growth of strain LC-T2^T occurred at 4–32°C (optimum, 25–28°C), at pH 6.0–11.5 (optimum, 8.0–8.5), and with 0–1.5% (w/v) NaCl (optimum, 0%). Strain LC-T2^T was positive for catalase, the reduction of nitrate to nitrite, and the assimilation of mannitol, but it was negative for oxidase and hydrolysis of cellulose, Tween 80, and DNA. In API ZYM test strips, strain LC-T2^T was as follows: positive for alkaline phosphatase, acid phosphatase, esterase (C4), esterase lipase (C8), leucine arylamidase, valine arylamidase, naphthol-AS-BI-phosphohydrolase, α-galactosidase, β-galactosidase, α-glucosidase, and β-glucosidase activities and negative for lipase (C14), trypsin, β-glucuronidase, N-acetyl-β-glucosaminidase, α-mannosidase, cystine arylamidase, α-chymotrypsin, and α-fucosidase activities. Acid is produced from methyl-β-D-xylopyranoside, D-mannose, N-acetyl-glucosamine, mannitol, D-turanose, L-arabinose, D-cellobiose, D-lactose, D-raffinose, D-turanose, D-sucrose, arbutin, and esculin. The following compounds are utilized as sole carbon sources in the GENIII microplates: Dextrin, D-maltose, D-trehalose, D-cellobiose, gentiobiose, sucrose, D-turanose, stachyose, D-raffinose, α-D-lactose, D-melibiose, N-acetyl-D-glucosamine, α-D-glucose, D-mannose, D-fructose, D-galactose, D-sorbitol, and D-mannitol. The predominant cellular fatty acids (>10.0% of total fatty acids) of strain LC-T2^T were anteiso-C_15:0 and C_16:0. The major polar lipids of strain LC-T2^T were established as DPG, PE, PG, and several unidentified ingredient as follows: three unidentified phospholipids (PL1–3), two unidentified aminophospholipids (APL1–2), and one unidentified glycolipid (GL1). Menaquinone-7 (MK-7) was detected as the only respiratory quinone. The DNA G + C content is 46.0 mol%.

The type strain is LC-T2^T (=CCTCC AB 2019254^T = KCTC 43175^T), isolated from spruce forest in the Qilian Mountains, Gansu province, China (38°25′32″N, 99°55′40″E). The GenBank/EMBL/DDBJ accession number for 16S rRNA, rpoB, and nifH gene sequence and the whole-genome sequence of strain LC-T2^T can be found at: https://www.ncbi.nlm.nih.gov/genbank/, OK058271, OK094314, OK094315, and WJXB00000000, respectively.