Distinct Bacterial Communities Within the Nonrhizosphere, Rhizosphere, and Endosphere of Ammodendron bifolium Under Winter Condition in the Takeermohuer Desert Authors: Zhining Kou, Jiaqin Liu, Gulpiye Tohti, Xiaoying Zhu, Bei Zheng, Yanlei Zhu, Wei Zhang Published: 2024 ABSTRACT Due to human activities and severe climatic conditions, the population of Ammodendron bifolium, an excellent sand-fixing plant, has gradually decreased in the Takeermohuer Desert. The plant-associated bacteria community can enhance its survival in harsh environments. However, the understanding of A. bifolium-associated bacterial community is still unclear during the harsh winter. We investigated the bacterial community structure from the A. bifolium rhizosphere and nonrhizosphere at different depths (i.e., 0–40 cm, 40–80 cm, 80–120 cm) and from endosphere (i.e., root endosphere and stem endosphere) in winter. At the same time, we analyzed the impact of different compartments and soil factors on the bacterial community structure. Studies have shown that the A. bifolium rhizosphere exhibits higher levels of SOM (soil organic matter), SOC (soil organic carbon), SAN (soil alkaline nitrogen), and SAK (soil available potassium) compared with the nonrhizosphere. The dominant bacterial phyla were Proteobacteria (19.6%), Cyanobacteria (15.9%), Actinobacteria (13.6%), Acidobacteria (9.0%), and Planctomycetota (5.7%) in the desert. Proteobacteria (24.0–30.2%) had the highest relative abundance in rhizosphere, Actinobacteria (18.3–22.6%) had the highest relative abundance in nonrhizosphere, and Cyanobacteria had the highest relative abundance in endosphere. At the genus level, the relative abundance of Pseudomonas (1.2%) in the root endosphere was the highest and the other genera were mostly unclassified. The Chao1 and PD_whole_tree indices showed that the diversity of the bacterial communities decreased from nonrhizosphere, rhizosphere, root endosphere to stem endosphere. Co-occurrence network analyses identified Proteobacteria and Actinobacteria as key species across the three compartments. Additionally, unique keystone species like Cyanobacteria, Verrucomicrobiota, and Desulfobacterota were found only in the endosphere. The bacterial community in the rhizosphere was influenced by factors such as EC (electrical conductivity), STC (soil total carbon), SOM, SOC, STN (soil total nitrogen), SAN, STP (soil total phosphorus), and SAK, while that of the nonrhizosphere was mainly influenced by pH, C/N (STC/STN), SAP, and distance. The study highlighted differences in bacterial community composition, diversity, and influencing factors across the three compartments, which can provide a better understanding of the association/interactions between A. bifolium and bacterial communities and lay a foundation for revealing its adaptability in winter.Supplementary InformationThe online version contains supplementary material available at 10.1007/s00248-024-02462-4. Introduction In the recent era, with the rising global temperatures, climate change has enhanced drought and desertification and, in turn, affected biodiversity and terrestrial ecosystem [1, 2]. The plants that lived in desert region have employed several adaptation strategies in order to escape or tolerate extreme environmental conditions. For example, desert plants establish strong symbioses with a large variety of microorganisms [3, 4]. Numerous studies suggested that the plants growing in given conditions try to recruit/select the best, host-specific microbial community, which may play pivotal functional roles in the growth and fitness of natural vegetation and agronomic plants [5–7]. Given the importance of plant-associated microbes, developing and applying them to resist adversity will become a new ecofriendly and sustainable strategy. Plant-associated microbes are reported to exist in various plant compartments (e.g., rhizosphere, endosphere, phyllosphere) [5, 8]. It is well known that plant species is the main factor driving or shaping the composition and diversity of its associated microbial community [9, 10]. However, for a specific plant, soil characteristics, water availability, and temperature are key factors affecting its microbial communities. Moreover, climate factors can directly modulate the effect of soil properties on microbial diversity, thus having a greater influence on microbial community structure than soil properties [11]. With seasonal change, soil temperature, plant photosynthesis, root exudates, and litterfall will affect soil properties at different levels [12]. The various soil factors that affect the microbial community include pH, nutrient availability, and moisture content. Among these soil properties, the soil pH may be the best predictor for the diversity and composition of the soil bacterial community, but not for the fungal community [13, 14]. While organic carbon, nitrogen, and phosphorus were the major factors driving the structure of soil microbial community [15, 16]. Furthermore, seasonal cycles and global warming change temperature, precipitation, soil physicochemical properties, etc., which in turn cause shifts in structure, composition, and function of microbial community [12, 17, 18]. The Takeermohuer Desert located in western China is mainly characterized by sparse precipitation, dramatic seasonal variations, and large fluctuations in soil temperature and humidity [19]. Ammodendron bifolium (Pall.) Yakovl. growing in this region is a dominant desert shrub with windbreak and sand fixation due to its well-developed root system and strong drought tolerance. Currently, we have already explored the diversity and structure of rhizospheric and endophytic bacterial communities in A. bifolium in summer [20, 21]. Numerous studies postulate that plants may harbor a species-specific set of essential microbes that are not influenced by environmental conditions and plant growth [22, 23]. These microbial communities may be crucial for maintaining plant health and fitness. However, it is not clear that there are stable microbial communities in A. bifolium that are not affected by seasonal variation. Thus, the structure and composition of desert A. bifolium-associated microbial communities need to be studied. Here, based on winter climate condition, we focused on (1) analyzing the structure and composition of bacterial communities in the rhizosphere, nonrhizosphere, root, and stem endosphere in A. bifolium, and (2) exploring the key or core bacterial taxa by cooccurrence network analyses. We postulated that seasonal variation may influence the diversity and composition of A. bifolium-associated bacterial communities, but there were stable bacterial communities present in different compartments of desert A. bifolium. Our findings will contribute to revealing the close relationship between bacterial communities and plants and the impact of bacteria on plant growth and stress resistance, which will lay a theoretical foundation for the biodiversity conservation in desert region. Materials and Methods Study Site Description, Sampling, and Processing The samples were collected in the Takermohur Desert in Huocheng County, Xinjiang, northwest China (43°50′–44°09′N and 80°27′–80°51′E) during the winter season in January 2023. The average temperature in the winter months (December, January, and February) was − 9 °C, and the average precipitation was 12.07 mm (Weather Station of Huocheng County in Xinjiang). In the sampling site, three quadrats separated by 100 m from each other were set up, and three adult A. bifolium plants were selected from each quadrat. Then, the rhizosphere samples adjacent to the root surface (within 2 mm) were collected at three soil layers of 0–40, 40–80, and 80–120 cm along the principal root direction, and the nonrhizosphere was sampled at three soil layers of the same depth with 50 cm away from the sampling plants (sampling site of the rhizosphere was defined as the starting point of sampling, i.e., a distance of 0 cm from plant rhizosphere, and sampling distance of the nonrhizosphere was 50 cm from the plant rhizosphere). According to a previous report, the nonrhizosphere sample at a distance of about 50 cm from the plant rhizosphere was not affected by the root system [24]. Three parallel samples at each soil layer from each quadrat were then mixed to create a composite sample, and a total of 18 soil samples (three soil layers × two soil types × three replications) were obtained. The samples of three soil layers from the rhizosphere were indicated as WRa, WRb, and WRc, and the nonrhizosphere samples were indicated as WBa, WBb, and WBc. In the sample codes, “W” represented winter, “B” and “R” represented different soil types of rhizosphere and nonrhizosphere, and “a,” “b,” and “c” represented different depths of 0–40 cm, 40–80 cm, and 80–120 cm, respectively. In the meanwhile, the roots and stem segments about 15 cm long were cut with sterile scissors from three plants in each quadrat and put into sterile sampling bags, denoted as WRe and WSe, respectively. Then, three plant samples from the same quadrats were mixed to form one composite sample, providing a total of 8 composite samples (two tissues × four replications). In the two sample codes, “R” and “S” represented the root and stem samples, respectively, and “e” represented the endosphere. Altogether, the 26 samples in this study were immediately stored at 4 °C and transported back to the laboratory within 24 h. The soil samples from rhizosphere and nonrhizosphere were divided into two parts after removing litter and other impurities. One part was stored in a refrigerator at − 80 °C for DNA extraction, and the other was used for measuring the water content (WC) and soil physicochemical properties. The plant tissue samples (roots and stems) were surface-sterilized according to the previous method with minor modifications [25, 26]. In brief, the root and stem tissues were cut into small segments of 5–6 cm and soaked in 75% alcohol for 1 min, then soaked in 2.5% sodium hypochlorite solution for 10 min and rinsed with sterile water for 2–3 times; the last rinse solution was spread on R2A medium for evaluation of disinfection efficacy. Subsequently, the surface-sterilized samples were stored at − 80 °C for DNA extraction. Determination of Soil Physicochemical Properties In this study, the physicochemical properties of the rhizosphere and nonrhizosphere samples (18 soil samples) were determined. The soil pH and electrical conductivity (EC) were measured after creating a 1:2.5 (w/v) and 1:5 (w/v) soil-to-distilled water slurry with a pH meter and an EC meter, respectively [27]. The water content of soil (WC) was determined based on the gravimetric method. Soil organic matter (SOM), soil organic carbon (SOC), and soil total carbon (STC) were determined using the dichromate oxidation method [28] and Elemental Analyzer (Vario EL III, Germany), respectively. Soil total nitrogen (STN) and available nitrogen (SAN) were measured by the Kjeldahl method and alkali-diffusion method, respectively. Soil total phosphorus (STP) and available phosphorus (SAP) were determined using sodium hydroxide melting- and sodium bicarbonate extraction-molybdenum-antimony colorimetric methods, respectively [29]. Soil total potassium (STK) and available potassium (SAK) were measured by sodium hydroxide melting and ammonium acetate extraction-flame photometry, respectively [30]. K+ and Na+ in soil were determined by flame photometry, and STC/STN (C/N) was calculated based on dividing the STC by the STN. Isolation of Culturable Bacteria and 16S rDNA Genes Analysis For bacterial cultivation, a 10 g sample of soils, roots, and stems was ground and suspended in 90 ml sterile phosphate-buffered saline solution. The mixtures were shaken at 200 rpm for 30 min at 30 °C, serially diluted and plated on R2A [31], and then incubated in the dark for 5–7 days at 30 °C. The colony-forming units (CFUs) per gram of sample were determined, and colonies with distinct morphology were picked for re-streaking to obtain pure cultures. Genomic DNA of all isolates was extracted according to the CTAB method. PCR amplification of the 16S rDNA genes was performed using the previously described universal primers 27F (5′-AGAGTTTGATCCTGGCTCAG-3′) and 1492R (5′-AAGGAGGTGATCCAGCC-3′) [32]. The sequencing of the 16S rDNA genes was conducted by Sangon Biotech Co. Ltd., Shanghai, China. The obtained 16S rDNA gene sequence was compared with the GenBank database to identify the isolates. The sequence data have been deposited in the GenBank database under accession numbers PP962439–PP962511. Total DNA Extraction and High-Throughput Amplicon Sequencing Total genomic DNA was extracted from 0.5 g of soil sample using the TGuide S96 Magnetic Soil /Stool DNA Kit (Qiagen Inc.) following the manufacturer’s instructions. For root and stem tissues, the genomic DNA was extracted using the CTAB method [33]. In total, 26 samples were extracted and stored at − 20 °C for use. DNA quantification was performed using a microplate reader (synergy HTX, BioTek, USA) and quality assessment by electrophoresis on 0.8% agarose gels. The V4-V5 region can provide the same information as the full-length accuracy of the 16 s rRNA gene at the phylum level, and further, a large number of studies on soil bacteria have selected the V4-V5 region for identifying bacterial species [34]. The bacterial communities were profiled by targeting the V4-V5 region of the 16S rRNA gene, and the target sequences were amplified using the primer pairs 515F (5 GTGYCAGCMGCCGCGGTAA 3) and 907R (5 CCGYCAATTYMTTTRAGTTT 3). Sequencing was on an Illumina NovaSeq 6000 platform (Illumina, San Diego, CA, USA) at the Wuhan Frasergen Bioinformatics Co. Ltd. (Wuhan, China). The raw sequencing data were deposited in the NCBI Sequence Read Archive (SRA) under the accession numbers: SAMN38479733–SAMN38479758. Data Processing and Statistical Analysis Bacterial Diversity and Taxonomic Composition The raw reads were quality filtered and primers removed using Trimmomatic v0.33 [35] and cutadapt v1.9.1 [36]. The pair-end reads were merged using USEARCH v10 [36], and the chimeric sequences were discarded using UCHIME v8.1 [37]. Then, the high-quality sequences (non-chimeric reads) were obtained for subsequent analysis. These sequences were clustered into operational taxonomic units (OTUs) at 97% sequence similarity using USEARCH v10 [38] and denoised and joined to produce amplicon sequence variants (ASVs) using DADA2 [39]. ASVs taxonomy was assigned against the SILVA database (release138) using QIIME2’s classify-consensus-blast and classify-sklearn [40]. ASVs abundance table was generated using USEARCH v10 at different taxonomic levels. Histograms were used to analyze taxonomic composition at the genus and phylum levels, and Venn diagrams were made to display the number of common or unique ASVs in different groups (rhizosphere, nonrhizosphere, root endosphere, and stem endosphere) using the Venn Diagram package in R [41] Bacterial community abundance was compared by ANOVA (Tukey’s HSD test, p value < 0.05). For alpha diversity, the Chao1 and PD_whole_tree indices were selected based on species richness and evenness for evaluating community diversity. The two indices were calculated based on previous methods [21] and plotted using QIIME2 2020.6 (https://qiime2.org). The indices were compared using T test (between two sample types) and Tukey’s HSD test (between different groups) at p value < 0.05. For beta diversity, Bray–Curtis distances were calculated, and principal coordinates analysis (PCoA) was performed in QIIME software. The PCoA diagrams were visualized using the ggplot2 package in R [42], and non-parametric permutation analysis of variance (PERMANOVA) was performed to determine the statistical significance of dissimilarities in beta diversity using the vegan package in R [43]. Co-occurrence Network Analysis We developed three individual co-occurrence networks to compare their structure in different groups (i.e., rhizosphere, nonrhizosphere, and endosphere). The networks were built using the psych package in R along with the Spearman correlation coefficient (|r|≥ 0.8, p < 0.05), and all network diagrams were visualized in Gephi [44]. We computed the node degree, betweenness centrality, and closeness centrality using the igraph package in R and identified the keystone species by ranking the nodes based on the sum of these parameters [45]. Topology indices of the networks, such as node number, edge number, average path length, and modularity, were also calculated using the same software. Redundancy Analysis The redundancy analysis (RDA) was performed to evaluate the correlation relationships among species and edaphic variables using the R software vegan package [46]. The significance of the relationships between community structure and environmental factors (soil factors and sampling distance) was assessed according to the previous method [21]. Soil properties, the culturable bacterial count, and bacterial community abundance were compared by ANOVA with Tukey’s HSD test using SPSS 20.0 (SPSS Inc., Chicago, USA), and alpha diversity by T-test. P-value less than 0.05 was considered statistically significant in all statistical tests. Results Characterization in Soil Physicochemical Properties The physicochemical properties of rhizosphere and nonrhizosphere soils are profiled in Table 1. From the change of pH value (ranging from 8.59 to 9.45), the soil in the Takeermohuer Desert was slightly alkaline. The pH value in rhizosphere (8.59–9.01) was lower than that in nonrhizosphere (9.40–9.45), and the pH in WRb (8.62) and WRc (8.59) was lower than that of WRa (9.01). In contrast, the EC value of rhizosphere (117.40–262.33 μs cm−1) was significantly higher than that of nonrhizosphere soil (50.93–57.80 μs cm−1), with WRb having the highest EC value. Similarly, the contents of SOM (6.97–9.89 g kg−1), SOC (3.66–5.48 g kg−1), STC (13.50–15.50 g kg−1), STN (0.44–0.51 g kg−1), SAN (33.00–40.72 mg kg−1), and SAK (171.81–230.20 mg kg−1) of the rhizosphere were significantly higher than those of nonrhizosphere, with the contents of 2.09–3.36 g kg−1, 1.12–1.92 g kg−1, 10.46–10.98 g kg−1, 0.23–0.24 g kg−1, 12.18–14.03 mg kg−1, and 112.97–163.33 mg kg−1, respectively. On the contrary, the C/N of rhizosphere was lower than that of nonrhizosphere. Table 1Soil physicochemical properties in the rhizosphere and nonrhizosphere of A. bifolium in winterEdaphic factorsWBaWBbWBcWRaWRbWRcpH9.42 ± 0.04a9.40 ± 0.04a9.45 ± 0.02a9.01 ± 0.02b8.62 ± 0.01c8.59 ± 0.04cEC (us cm−1)57.80 ± 4.90d51.93 ± 0.62d50.93 ± 1.01d117.40 ± 9.30c262.33 ± 4.67a198.67 ± 3.38bSOM (g kg−1)3.36 ± 1.06b2.66 ± 0.93b2.09 ± 0.15b7.99 ± 0.72a6.97 ± 1.07a9.89 ± 1.36aSOC (g kg−1)1.92 ± 0.63c1.12 ± 0.15c1.66 ± 0.47c5.26 ± 0.29a5.48 ± 0.81a3.66 ± 0.43bSTC (g kg−1)10.98 ± 0.60b10.92 ± 0.68b10.46 ± 0.92b14.42 ± 0.21a15.50 ± 0.70a13.50 ± 0.46aSTN (g kg−1)0.23 ± 0.01b0.24 ± 0.01b0.23 ± 0.01b0.51 ± 0.09a0.46 ± 0.03a0.44 ± 0.03aSTP (g kg−1)0.30 ± 0.01a0.29 ± 0.00a0.31 ± 0.01a0.31 ± 0.01a0.31 ± 0.01a0.31 ± 0.02aSTK (g kg−1)20.59 ± 0.54ab20.06 ± 0.48ab19.79 ± 0.32ab20.95 ± 0.38a20.15 ± 0.06ab19.64 ± 0.31bSAN (mg kg−1)14.03 ± 1.96c12.18 ± 1.54c13.91 ± 1.75c39.21 ± 1.24a40.72 ± 2.01a33.00 ± 2.05bSAP (mg kg−1)18.84 ± 0.88a18.63 ± 0.27a18.17 ± 0.17a19.17 ± 2.05a14.32 ± 0.99b15.01 ± 0.35bSAK (mg kg−1)163.33 ± 1.00d127.23 ± 4.30e112.97 ± 1.65f216.41 ± 1.60b230.20 ± 4.19a171.81 ± 1.46cK+ (mg kg−1)8.73 ± 0.41ab8.38 ± 0.60ab8.13 ± 0.51ab7.36 ± 0.45b8.37 ± 0.32ab9.75 ± 0.62aNa+ (mg kg−1)28.93 ± 1.01a26.44 ± 1.39a27.09 ± 1.38a29.21 ± 1.40a26.63 ± 2.10a27.15 ± 0.52aC/N48.03 ± 3.22a45.16 ± 2.31a45.39 ± 3.53a29.60 ± 4.06b33.00 ± 2.99b31.26 ± 2.37bWC (%)3.04 ± 0.29a3.58 ± 0.34a2.02 ± 0.58a2.65 ± 0.30a3.16 ± 0.96a2.35 ± 0.90aThe data is represented as mean ± standard error (n = 3)The same lowercase letters within the same line indicate no significant difference among different soil samples based on Tukey’s HSD test (p < 0.05). WBa, WBb, and WBc represent nonrhizosphere soil in 0–40 cm, 40–80 cm, and 80–120 cm soil layers, respectively; WRa, WRb, and WRc represent rhizosphere soil in 0–40 cm, 40–80 cm, and 80–120 cm soil layers, respectively. EC, electrical conductivity; SOM, soil organic matter; SOC, soil organic carbon; STC, soil total carbon; STN, soil total nitrogen; STP, soil total phosphorus; STK, soil total potassium; SAN, soil alkaline nitrogen; SAP, soil available phosphorus; SAK, soil available potassium; K+, potassium ion; Na+, sodium ion; C/N, the STC divided by the STN; WC, water content Table 1Soil physicochemical properties in the rhizosphere and nonrhizosphere of A. bifolium in winterEdaphic factorsWBaWBbWBcWRaWRbWRcpH9.42 ± 0.04a9.40 ± 0.04a9.45 ± 0.02a9.01 ± 0.02b8.62 ± 0.01c8.59 ± 0.04cEC (us cm−1)57.80 ± 4.90d51.93 ± 0.62d50.93 ± 1.01d117.40 ± 9.30c262.33 ± 4.67a198.67 ± 3.38bSOM (g kg−1)3.36 ± 1.06b2.66 ± 0.93b2.09 ± 0.15b7.99 ± 0.72a6.97 ± 1.07a9.89 ± 1.36aSOC (g kg−1)1.92 ± 0.63c1.12 ± 0.15c1.66 ± 0.47c5.26 ± 0.29a5.48 ± 0.81a3.66 ± 0.43bSTC (g kg−1)10.98 ± 0.60b10.92 ± 0.68b10.46 ± 0.92b14.42 ± 0.21a15.50 ± 0.70a13.50 ± 0.46aSTN (g kg−1)0.23 ± 0.01b0.24 ± 0.01b0.23 ± 0.01b0.51 ± 0.09a0.46 ± 0.03a0.44 ± 0.03aSTP (g kg−1)0.30 ± 0.01a0.29 ± 0.00a0.31 ± 0.01a0.31 ± 0.01a0.31 ± 0.01a0.31 ± 0.02aSTK (g kg−1)20.59 ± 0.54ab20.06 ± 0.48ab19.79 ± 0.32ab20.95 ± 0.38a20.15 ± 0.06ab19.64 ± 0.31bSAN (mg kg−1)14.03 ± 1.96c12.18 ± 1.54c13.91 ± 1.75c39.21 ± 1.24a40.72 ± 2.01a33.00 ± 2.05bSAP (mg kg−1)18.84 ± 0.88a18.63 ± 0.27a18.17 ± 0.17a19.17 ± 2.05a14.32 ± 0.99b15.01 ± 0.35bSAK (mg kg−1)163.33 ± 1.00d127.23 ± 4.30e112.97 ± 1.65f216.41 ± 1.60b230.20 ± 4.19a171.81 ± 1.46cK+ (mg kg−1)8.73 ± 0.41ab8.38 ± 0.60ab8.13 ± 0.51ab7.36 ± 0.45b8.37 ± 0.32ab9.75 ± 0.62aNa+ (mg kg−1)28.93 ± 1.01a26.44 ± 1.39a27.09 ± 1.38a29.21 ± 1.40a26.63 ± 2.10a27.15 ± 0.52aC/N48.03 ± 3.22a45.16 ± 2.31a45.39 ± 3.53a29.60 ± 4.06b33.00 ± 2.99b31.26 ± 2.37bWC (%)3.04 ± 0.29a3.58 ± 0.34a2.02 ± 0.58a2.65 ± 0.30a3.16 ± 0.96a2.35 ± 0.90aThe data is represented as mean ± standard error (n = 3)The same lowercase letters within the same line indicate no significant difference among different soil samples based on Tukey’s HSD test (p < 0.05). WBa, WBb, and WBc represent nonrhizosphere soil in 0–40 cm, 40–80 cm, and 80–120 cm soil layers, respectively; WRa, WRb, and WRc represent rhizosphere soil in 0–40 cm, 40–80 cm, and 80–120 cm soil layers, respectively. EC, electrical conductivity; SOM, soil organic matter; SOC, soil organic carbon; STC, soil total carbon; STN, soil total nitrogen; STP, soil total phosphorus; STK, soil total potassium; SAN, soil alkaline nitrogen; SAP, soil available phosphorus; SAK, soil available potassium; K+, potassium ion; Na+, sodium ion; C/N, the STC divided by the STN; WC, water content Soil physicochemical properties in the rhizosphere and nonrhizosphere of A. bifolium in winter The data is represented as mean ± standard error (n = 3) The same lowercase letters within the same line indicate no significant difference among different soil samples based on Tukey’s HSD test (p < 0.05). WBa, WBb, and WBc represent nonrhizosphere soil in 0–40 cm, 40–80 cm, and 80–120 cm soil layers, respectively; WRa, WRb, and WRc represent rhizosphere soil in 0–40 cm, 40–80 cm, and 80–120 cm soil layers, respectively. EC, electrical conductivity; SOM, soil organic matter; SOC, soil organic carbon; STC, soil total carbon; STN, soil total nitrogen; STP, soil total phosphorus; STK, soil total potassium; SAN, soil alkaline nitrogen; SAP, soil available phosphorus; SAK, soil available potassium; K+, potassium ion; Na+, sodium ion; C/N, the STC divided by the STN; WC, water content Furthermore, we found that there was no significant difference in the contents of SOM, SOC, STC, STN, SAN, and C/N among samples of the same soil type. From the change of SAK, the highest SAK content of rhizosphere (230.20 mg kg−1) was observed in WRb, while the SAK contents of nonrhizosphere (112.97–163.33 mg kg−1) tended to decrease with increasing soil depth. The STP content (0.29–0.31 g kg−1) did not exhibit a significant difference among all samples. But we found that the SAP contents in WRb (14.32 mg kg−1) and WRc (15.01 mg kg−1) were the lowest, and the other samples (18.17–19.17 mg kg−1) did not show a difference. The change trend of STK (19.64–20.95 g kg−1), K+ (7.36–9.75 Cmol kg−1), Na+ (26.44–29.21 Cmol kg−1), and WC (2.02–3.58%) contents was not significant. Regardless of the sampling site, the overall levels of SOM, soil nitrogen (STN and SAN), and phosphorus (STP and SAP) were still very low, while the soil potassium (STK and SAK) level was very high. In addition, the soil moisture content indicated that the Takeermohuer Desert exhibited extreme drought. Diversity and Distribution of Bacterial Community by Culture-Dependent Method A total of 73 bacterial strains were isolated from the rhizosphere, nonrhizosphere, and root endosphere of A. bifolium on R2A agar medium (Table S1). These strains belonged to 4 phyla and 17 genera, consisting of 33 rhizosphere strains (4 phyla and 9 genera), 31 nonrhizosphere strains (3 phyla, 12 genera), and 9 root endosphere strains (1 phyla and 4 genera) (Fig. 1a, b). In all samples, the phylum with the highest diversity was Firmicutes, consisting of 6 genera, and the most infrequent phylum was Bacteroidota, with one genus present only in rhizosphere soil (Fig. 1a). When the three soil layers were taken as a whole, the genera with relative proportion greater than 5% were Bacillus, Brevibacillus, and Paenibacillus in both nonrhizosphere and rhizosphere, as well as Bacillus, Brevibacillus, Paenarthrobacter, Paenibacillus, and Peribacillus in the root endosphere (Fig. 1b). Among the three locations, i.e., rhizosphere, nonrhizosphere, and endosphere, the mutual species are higher than the unique species. In addition, the number of bacterial species decreased in nonrhizosphere with increasing soil depth, while the rhizosphere bacterial species in the top two layers of soil were higher than those in the bottom layer (Fig. 1c). Similarly, the plate counting results revealed that the bacterial count varied across different samples (ranging from 10,000 to 550,000 cfu­g−1). The rhizosphere (ranging from 275,000 to 550,000 cfu­g−1) exhibited significantly higher bacterial count compared to nonrhizosphere (ranging from 10,000 to 110,000 cfu­g−1), and root endosphere ranged from 90,000 to 130,000 cfu­g−1 (Fig. 1d). These results indicated that the rhizosphere harbored the highest bacterial population, followed by endosphere, and the nonrhizosphere had the lowest count.Fig. 1Structure of culturable bacterial communities in nonrhizosphere, rhizosphere, and root endosphere bacteria. a, b Composition and distribution of culturable bacterial communities at phylum and genus levels; c the number of culturable bacterial species; d the number of bacteria in different compartments samples. In the sample code, WBa, WBb, and WBc represent the nonrhizosphere soil samples from the three soil layers (a, 0–40 cm; b, 40–80 cm; c, 80–120 cm); WRa, WRb, and WRc represent the rhizosphere samples from the three soil layers (a, 0–40 cm; b, 40–80 cm; c, 80–120 cm); WRe represents the sample from root endosphere; WSe represents the sample from stem endosphere. In d, values are means ± standard errors, and different lowercase letters indicate significant differences (p < 0.05) between different soil samples based on ANOVA and Tukey’s HSD tests Structure of culturable bacterial communities in nonrhizosphere, rhizosphere, and root endosphere bacteria. a, b Composition and distribution of culturable bacterial communities at phylum and genus levels; c the number of culturable bacterial species; d the number of bacteria in different compartments samples. In the sample code, WBa, WBb, and WBc represent the nonrhizosphere soil samples from the three soil layers (a, 0–40 cm; b, 40–80 cm; c, 80–120 cm); WRa, WRb, and WRc represent the rhizosphere samples from the three soil layers (a, 0–40 cm; b, 40–80 cm; c, 80–120 cm); WRe represents the sample from root endosphere; WSe represents the sample from stem endosphere. In d, values are means ± standard errors, and different lowercase letters indicate significant differences (p < 0.05) between different soil samples based on ANOVA and Tukey’s HSD tests Diversity and Distribution of Bacterial Community by Culture-Independent Method The High-Throughput Sequencing Results and Bacterial Community α-Diversity In total, 1,182,071 high-quality sequences were obtained from 26 samples. The number of high-quality sequences in each sample ranged from 31,108 to 58,711. The read lengths ranged from 231 to 419 with an average of 372. A total of 13,938 ASVs were identified in all samples, and the ASVs number ranged from 143 to 1903 in each sample. In the present study, the species accumulation curve (SAC) illustrated that the sequencing results were available and could reflect the bacterial community composition and diversity of the real samples (Fig. S1). As shown in Fig. 2a and b, the Chao1 and PD_whole_tree indices of WBa and WBc in the nonrhizosphere were significantly higher than those of the rhizosphere samples in the same soil layer. Generally speaking, the richness and diversity of the nonrhizosphere (i.e., WBB) were higher than those in the other three compartments (i.e., WRR, WER, and WES), the richness of the rhizosphere (i.e., WRR) was higher than that of the root and stem endosphere, and the diversity of root endosphere (i.e., WRe) was higher than that of stem endosphere (i.e., WSe) (Fig. 2c and d).Fig. 2Diversity of bacterial communities in different groups of samples. The Chao1 index (a, c) and PD_whole_tree index (b, d) were used to estimate the alpha diversity. In a and b: WBa, WBb, and WBc, nonrhizosphere soil samples from the three soil layers (a, 0–40 cm; b, 40–80 cm; c, 80–120 cm); WRa, WRb, and WRc, rhizosphere samples from the three soil layers (a, 0–40 cm; b, 40–80 cm; c, 80–120 cm); WRe, sample from root endosphere; WSe, sample from stem endosphere; the asterisks indicate significant differences between rhizoshpere and nonrhizosphere soil samples at the same depth and between WRe and WSe (T test, p < 0.05). In c and d: WBB, nonrhizosphere sample; WRR, rhizosphere sample; WER, root endosphere sample; WES, stem endosphere sample; the asterisks indicate significant differences between different groups (T test, p < 0.05) Diversity of bacterial communities in different groups of samples. The Chao1 index (a, c) and PD_whole_tree index (b, d) were used to estimate the alpha diversity. In a and b: WBa, WBb, and WBc, nonrhizosphere soil samples from the three soil layers (a, 0–40 cm; b, 40–80 cm; c, 80–120 cm); WRa, WRb, and WRc, rhizosphere samples from the three soil layers (a, 0–40 cm; b, 40–80 cm; c, 80–120 cm); WRe, sample from root endosphere; WSe, sample from stem endosphere; the asterisks indicate significant differences between rhizoshpere and nonrhizosphere soil samples at the same depth and between WRe and WSe (T test, p < 0.05). In c and d: WBB, nonrhizosphere sample; WRR, rhizosphere sample; WER, root endosphere sample; WES, stem endosphere sample; the asterisks indicate significant differences between different groups (T test, p < 0.05) Bacterial Community Structure and Composition The top 10 dominant groups at the phylum and genus are shown in Fig. 3. At the phylum level, the phyla with relative abundance greater than 5% were Proteobacteria (19.6%), Cyanobacteria (15.9%), Actinobacteria (13.6%), Acidobacteria (9.0%), and Planctomycetota (5.7%) based on the average level of 26 samples; these phyla accounted for more than 63.8% of overall bacterial community (Fig. 3a). The phylum Proteobacteria was one of the most abundant phyla across all samples, with a relative abundance ranging from 3.1 to 28.6% (Fig. 3a). The top 3 dominant phyla in the nonrhizosphere were Actinobacteria (18.3–22.6%), Proteobacteria (14.3–19.3%), and Acidobacteria (9.0–11.7%); those of the rhizosphere were Proteobacteria (24.0–30.2%), Acidobacteria (12.5–15.4%), and Planctomycetota (9.5–10.7%); those of the root endosphere were Cyanobacteria (40.7–59.0%), Proteobacteria (18.0–26.1%), and Firmicutes (7.1–16.6%); only one dominant phylum in the stem endosphere was Cyanobacteria, ranging from 73.1 to 75.0% (Fig. 3a). It could be seen that the change trend of the phylum Proteobacteria was as follows: rhizosphere > root endosphere > nonrhizosphere > stem endosphere, and the phylum Cyanobacteria in root and stem endosphere had the largest relative abundance compared with the other samples. In the same type of soil, the relative abundance of the bacterial phyla did not show a clear trend with increasing soil depth.Fig. 3Relative abundance of the dominant groups of bacterial communities at the phylum (a) and genus levels (b) among all samples. In the sample code, WBa, WBb, and WBc represent the nonrhizosphere soil samples from the three soil layers (a, 0–40 cm; b, 40–80 cm; c, 80–120 cm); WRa, WRb, and WRc represent the rhizosphere samples from the three soil layers (a, 0–40 cm; b, 40–80 cm; c, 80–120 cm); WRe represents the sample from root endosphere; WSe represents the sample from stem endosphere Relative abundance of the dominant groups of bacterial communities at the phylum (a) and genus levels (b) among all samples. In the sample code, WBa, WBb, and WBc represent the nonrhizosphere soil samples from the three soil layers (a, 0–40 cm; b, 40–80 cm; c, 80–120 cm); WRa, WRb, and WRc represent the rhizosphere samples from the three soil layers (a, 0–40 cm; b, 40–80 cm; c, 80–120 cm); WRe represents the sample from root endosphere; WSe represents the sample from stem endosphere At the genus level, the genera with relative abundance greater than 1% were Unclassified_Cyanobacteriales, Pseudomonas, Unclassified_Vicinamibacteraceae, Unclassified_Vicinamibacterales, Unclassified_Nitrososphaeraceae, Unclassified_67_14, RB41, and uncultured_compost_archaeon based on the average level of all samples (Fig. 3b). When the three soil layers were taken as a whole, the relative abundance of the genera Unclassified_Gemmatimonadaceae (4.9%), Unclassified_Nitrososphaeraceae (2.8%), Unclassified_67_14 (2.7%), and Pseudomonas (2.5%) in nonrhizosphere was greater than that in the rhizosphere; the relative abundance of Unclassified_Vicinamibacteraceae (3.9%), uncultured_compost_archaeon (2.7%), Unclassified_Vicinamibacterales (2.3%), and RB41 (1.7%) in the rhizosphere was greater than that in the nonrhizosphere; the genus Pseudomonas (1.2%) in the root tissue had greater relative abundance than that in the other samples; the genus Unclassified_Gemmatimonadaceae (7.3%) in the stem tissue had greater relative abundance than that in the other samples. In the same type of soil, the change trend of the relative abundance of bacterial genera was not obvious with increasing soil depth, while the richness of bacterial community in soil samples was significantly greater than that in root and stem tissues. The principal coordinate analysis combined with the PERMANOVA test revealed that community composition was significantly different in eight groups (R2 = 0.8925, p = 0.001) and four locations (R2 = 0.7060, p = 0.001) (Fig. 4). From Fig. 4a, we found that the bacterial community in rhizosphere samples (WRa, WRb, WRc), nonrhizosphere samples (WBa, WBb, WBc), and endosphere samples (WRe, WSe) could be well clustered. It was further confirmed that the differences of the community composition were significant among rhizosphere (WRR), nonrhizosphere (WBB), and endosphere (WER/WES) at p value < 0.01 level, while there was a significant difference between root and stem endosphere at a p value < 0.05 level (Fig. 4b, Table S2). The results indicated that there was a close relationship between the bacterial community composition and the compartments they inhabited.Fig. 4Principal coordinates analyses (PCoA) based on the distance matrix calculated using the Bray–Curtis algorithm for all samples. a The difference of the bacterial community structure among all samples (WBa, WBb, and WBc, nonrhizosphere soil samples from the three soil layers (a, 0–40 cm; b, 40–80 cm; c, 80–120 cm); WRa, WRb, and WRc, rhizosphere samples from the three soil layers (a, 0–40 cm; b, 40–80 cm; c, 80–120 cm); WRe, sample from root endosphere; WSe, sample from stem endosphere). b The difference of the bacterial community structure among different compartments (WBB, nonrhizosphere; WRR, rhizosphere; WER, root endosphere; WES, stem endosphere) Principal coordinates analyses (PCoA) based on the distance matrix calculated using the Bray–Curtis algorithm for all samples. a The difference of the bacterial community structure among all samples (WBa, WBb, and WBc, nonrhizosphere soil samples from the three soil layers (a, 0–40 cm; b, 40–80 cm; c, 80–120 cm); WRa, WRb, and WRc, rhizosphere samples from the three soil layers (a, 0–40 cm; b, 40–80 cm; c, 80–120 cm); WRe, sample from root endosphere; WSe, sample from stem endosphere). b The difference of the bacterial community structure among different compartments (WBB, nonrhizosphere; WRR, rhizosphere; WER, root endosphere; WES, stem endosphere) Comparison of Bacterial Community Distribution and Composition The Venn diagram showed that the ASVs of the bacterial community differed in rhizosphere, nonrhizosphere, root, and stem endosphere. There were 6180 unique ASVs in nonrhizosphere, 4200 unique ASVs in rhizosphere, 1475 unique ASVs in root endosphere, and 363 unique ASVs in stem endosphere, and the shared ASVs were 60 in the four departments (Fig. 5a). The predominant unique phyla in both stem and root endosphere were Proteobacteria, Cyanobacteria, and Firmicutes, while those of both rhizosphere and nonrhizosphere were Proteobacteria, Actinobacteria, Acidobacteria, and Planctomycetota. Besides, the predominant unique phyla in both rhizosphere and root endosphere still included Bacteroidota, and those of nonrhizosphere still included Chloroflexi, Gemmatimonadota, and Thermoplasmatota. Among the bacteria shared by all groups, the phylum Cyanobacteria accounted for the highest proportion, followed by Proteobacteria (Fig. 5b).Fig. 5Comparison of the distribution and composition of the bacterial communities in rhizosphere, nonrhizosphere soil, root, and stem endosphere at the phylum level. a Venn diagram showing the exclusive and shared ASVs of the bacterial community among different groups (WBB, nonrhizosphere soil; WRR, rhizosphere; WER, root endosphere; WES, stem endosphere). b These ASVs were used to create the abundance bar plots based on the phylum level (all, shared by all groups; WBB_only, only in the nonrhizosphere soil; WRR_only, only in the rhizosphere; WER_only, only in the root endosphere; WES_only, only in the stem endosphere) Comparison of the distribution and composition of the bacterial communities in rhizosphere, nonrhizosphere soil, root, and stem endosphere at the phylum level. a Venn diagram showing the exclusive and shared ASVs of the bacterial community among different groups (WBB, nonrhizosphere soil; WRR, rhizosphere; WER, root endosphere; WES, stem endosphere). b These ASVs were used to create the abundance bar plots based on the phylum level (all, shared by all groups; WBB_only, only in the nonrhizosphere soil; WRR_only, only in the rhizosphere; WER_only, only in the root endosphere; WES_only, only in the stem endosphere) Bacterial Interactions in the Rhizosphere, Nonrhizosphere, and Endosphere To understand potential interactions among bacterial community members associated with three different compartments, we constructed co-occurrence networks for the three locations (Fig. 6a). As shown in Table 2, the nonrhizosphere network contained 627 nodes and 11,656 edges, the rhizosphere network contained 552 nodes and 3832 edges, and the endosphere network contained 137 nodes and 1141 edges. It could be seen that the number of nodes and edges in three networks gradually declined from nonrhizosphere and rhizosphere to endosphere, while the endosphere network had a larger positive edge proportion, degree centralization, betweenness centrality, and density compared with the rhizosphere and nonrhizosphere networks, and the rhizosphere network presented higher average path length and diameter (Table 2 and Fig. 6b, c). In addition, the top three keystone species of the nonrhizosphere network were the phyla Proteobacteria, Acidobacteria, and Actinobacteria; those of the rhizosphere network were Proteobacteria, Acidobacteria, and Gemmatimonadota; those of the endosphere network were Cyanobacteria, Firmicutes, and Actinobacteria (Fig. 6d). We further found that the keystone species in three locations were the phyla Proteobacteria and Actinobacteria, and the unique keystone species that only existed in the endosphere were the phyla Cyanobacteria, Verrucomicrobiota, and Desulfobacterota (Fig. 6d). Therefore, it can be seen from the network diagram that the key species and dominant species in the three compartments are not basically consistent.Fig. 6Co-occurrence network of the bacterial communities associated with nonrhizosphere, rhizosphere, and endosphere of A. bifolium under winter condition. a Visualization of co-occurrence bacterial networks of nonrhizosphere, rhizosphere, and endosphere. Large modules with ≥ 3% of nodes are shown in different colors, and smaller modules are shown in gray. The pink edge between nodes represents a positive association, and the green edge represents a negative association. b Degree of connection of nodes within each network. c Betweenness centrality of nodes calculated for each network. d Visualization of the first 20 keystone species identified based on the sum of degree, closeness centrality, and betweenness centrality for each node in each network. The results from b to d are shown at the phylum levelTable 2The topological characteristics of the co-occurrence network of bacterial communities in nonrhizosphere soil, rhizosphere, and endosphereTopology indicesNonrhizosphere soilRhizosphereEndosphereNodes number627552137Edges number11,65638321141Positive edges7251 (62.2%)2433 (63.5%)837 (73.4%)Negative edges4405 (37.8%)1399 (36.5%)304 (26.6%)Average path length2.934.583.09Diameter10.613.79.6Degree centralization0.1170.1230.230Betweenness centrality0.0210.0540.265Density0.0590.0250.122Modularity0.4700.3780.377 Co-occurrence network of the bacterial communities associated with nonrhizosphere, rhizosphere, and endosphere of A. bifolium under winter condition. a Visualization of co-occurrence bacterial networks of nonrhizosphere, rhizosphere, and endosphere. Large modules with ≥ 3% of nodes are shown in different colors, and smaller modules are shown in gray. The pink edge between nodes represents a positive association, and the green edge represents a negative association. b Degree of connection of nodes within each network. c Betweenness centrality of nodes calculated for each network. d Visualization of the first 20 keystone species identified based on the sum of degree, closeness centrality, and betweenness centrality for each node in each network. The results from b to d are shown at the phylum level Table 2The topological characteristics of the co-occurrence network of bacterial communities in nonrhizosphere soil, rhizosphere, and endosphereTopology indicesNonrhizosphere soilRhizosphereEndosphereNodes number627552137Edges number11,65638321141Positive edges7251 (62.2%)2433 (63.5%)837 (73.4%)Negative edges4405 (37.8%)1399 (36.5%)304 (26.6%)Average path length2.934.583.09Diameter10.613.79.6Degree centralization0.1170.1230.230Betweenness centrality0.0210.0540.265Density0.0590.0250.122Modularity0.4700.3780.377 The topological characteristics of the co-occurrence network of bacterial communities in nonrhizosphere soil, rhizosphere, and endosphere Correlation Analysis Between Soil Bacterial Community and Physicochemical Factors In this study, we used redundancy analysis to understand the influences of environmental factors on sample distribution and bacterial community structure. As shown in Fig. 7a and b, the rhizosphere community was greatly affected by factors such as EC, STC, SOM, SOC, STN, SAN, STP, and SAK, while nonrhizosphere soil is mainly affected by pH, C/N, SAP, and distance. In the three soil layers, the influence of environmental factors on sample distribution had no obvious correlation. From Fig. 7c and Table S3, there was a positive correlation among the dominant phyla in nonrhizosphere (e.g., Actinobacteria) and pH, C/N, and distance a (0.62 < R < 0.86, p < 0.05), and that among the dominant phyla in rhizosphere soil (such as Proteobacteria, Acidobacteria, and Planctomycetota) and EC, STC, SOM, SOC, STN, SAN, and SAK (0.61 < R < 0.91, p < 0.05). The results indicated that EC, STC, SOM, SOC, STN, SAN, STP, and SAK were the major important factors influencing the bacterial community distribution in the rhizosphere, while pH, C/N, and distance were the major influencing factors in nonrhizosphere.Fig. 7Redundancy and correlation analyses between the bacterial communities (at the phylum level) and the soil physicochemical properties. a Redundancy analysis among all soil samples. In the sample code, WBa, WBb, and WBc, nonrhizosphere soil samples from the three soil layers (a, 0–40 cm; b, 40–80 cm; c, 80–120 cm); WRa, WRb, and WRc, rhizosphere samples from the three soil layers (a, 0–40 cm; b, 40–80 cm; c, 80–120 cm). b Redundancy analysis among the rhizosphere and nonrhizosphere soil samples (WBB, nonrhizosphere soil samples; WRR, rhizosphere samples). c Correlations between the relative abundance of bacterial taxa at the phylum level and physicochemical properties. The color gradient on the right represents Spearman’s rank correlation coefficients, with more positive values (pink) indicating stronger positive correlations and more negative values (light blue) indicating stronger negative correlations. The asterisks denote the significance level of Spearman’s rank correlation coefficients using the “corr.test” function. *p < 0.05, **p < 0.01, ***p < 0.001 Redundancy and correlation analyses between the bacterial communities (at the phylum level) and the soil physicochemical properties. a Redundancy analysis among all soil samples. In the sample code, WBa, WBb, and WBc, nonrhizosphere soil samples from the three soil layers (a, 0–40 cm; b, 40–80 cm; c, 80–120 cm); WRa, WRb, and WRc, rhizosphere samples from the three soil layers (a, 0–40 cm; b, 40–80 cm; c, 80–120 cm). b Redundancy analysis among the rhizosphere and nonrhizosphere soil samples (WBB, nonrhizosphere soil samples; WRR, rhizosphere samples). c Correlations between the relative abundance of bacterial taxa at the phylum level and physicochemical properties. The color gradient on the right represents Spearman’s rank correlation coefficients, with more positive values (pink) indicating stronger positive correlations and more negative values (light blue) indicating stronger negative correlations. The asterisks denote the significance level of Spearman’s rank correlation coefficients using the “corr.test” function. *p < 0.05, **p < 0.01, ***p < 0.001 Discussion Soil Physicochemical Properties The soil properties (pH, STN, SOM, STC, water content, etc.) and climatic factors are the main factors affecting bacterial diversity [47, 48] and growth of plants [49, 50] in desert ecosystems. Previous reports have shown that the high contents of STK (> 2 g kg−1) and SAK (> 155 mg kg−1) in soil contribute to plant resistance in abiotic stress, such as low-temperature, low-nutrient, high-salt, and dry environments [51, 52]. Hence, the high content of STK in soil and SAK in rhizosphere may be beneficial to A. bifolium to resist low-temperature stress. The high levels of Na+, K+, and C/N indicated that microbial decomposition and mineralization under winter condition are slow, which inhibits the absorption of N, P, and other elements by plants [53], such that A. bifolium maintains a low level of metabolism in winter conditions, which is conducive to its resistance to abiotic stress. Generally speaking, the soil in the Takeermohuer Desert is alkaline with very low levels of organic matter, organic carbon, soil nitrogen (STN and SAN), and soil phosphorus (STP and SAP), which indicates that the soil nutrient status is relatively poor. However, the soil potassium (STK and SAK) content is extremely high. Previous reports indicated that the fixation of potassium is more likely to occur at higher soil pH value [54]. And many studies have shown that desert soils with alkaline characteristics contain a large amount of potassium ions, such as the Minqin desert [55, 56], which is consistent with the results of this study. Previous researches reported that SOM, SOC, STC, STN, SAN, and SAK of rhizosphere were significantly higher than those of nonrhizosphere, which may be due to the rhizosphere effect [57]. This suggested that A. bifolium may also have a rhizosphere effect under winter conditions. Root exudates could accelerate the decomposition of organic matter, thereby promoting the absorption of some soluble ions by the root system [55], which may be the reason for the high EC value in the A. bifolium rhizosphere. In addition, the root exudates, such as organic acids, can improve the availability of rhizosphere nutrients through acidification, chelation, ion exchange, and reduction of insoluble nutrients in rhizosphere, thus promoting plant growth and development [58]; this may also explain why the rhizosphere showed lower pH and higher SAK in the A. bifolium rhizosphere. In addition, the contents of Na+ and K+ in the Takeermohuer Desert were much lower than those in summer [21], indicating that salinization was lower in winter, which was conducive to the survival of A. bifolium. Characterization of the Bacterial Community in Different Compartments Previous reports have shown that bacterial diversity varies seasonally [59, 60]. In the deserts, plants provide organic matter and physical barriers for bacteria, thus contributing to an increase in bacterial diversity [61]. Under the influence of the winter climate, the growth and metabolism of A. bifolium gradually reduced, which led to the decrease of bacterial diversity. At the same time, soil temperature significantly also has an important influence on bacteria quantity and diversity [56]. Therefore, the soil total OTU in winter decreased (725,475) due to low temperature, compared with summer (1,425,190) [21]. In addition, the diversity of bacterial communities decreases with increasing soil depth [62], and also from nonrhizosphere, rhizosphere to root interior tissue, which was associated with a selective rhizosphere barrier [63, 64]. In this study, the Chao1 index was used to evaluate the number of species (i.e., richness), and the PD_whole_tree index was used to evaluate the community diversity (i.e., richness and evenness). Under the same species richness, the greater the evenness, the greater the diversity of the community. Therefore, the variation of bacterial community diversity in this study is consistent with the above researches. Researches have shown that diverse phyla, such as Actinobacteria, Firmicutes, Cyanobacteria, Acidobacteria, and Proteobacteria, presented in desert ecosystems [65–67]. These predominant phyla are also observed in the Takeermohuer Desert. In the arid soils, nonrhizosphere and rhizosphere are primarily composed of Actinobacteria and Proteobacteria, respectively [68], as was the case in the present study. However, the two phyla are also dominant groups in both the rhizosphere and nonrhizosphere. Actinobacteria are known as abundant primary producers that support the productivity and resilience of desert ecosystems [69], which may explain its advantage and importance in maintaining Takeermohuer Desert ecosystems. Proteobacteria is a microbial community with strong adaptability and can be widely distributed in different soil types. Although desert soil is nutritionally poor compared to other soils, Proteobacteria has the ability to utilize limited water and nutrients, resulting in its relative fast growth [56, 70, 71]. Ultimately, these characteristics make them predominate in desert ecosystems. For Acidobacteria, it is predominant in both nonrhizosphere and rhizosphere in this study. A previous report has shown that Acidobacteria is sensitive to soil moisture conditions, showing a negative correlation with soil humidity [72], indicating its higher abundance in extremely low moisture levels and drought resistance in the Takeermohuer Desert in winter. In contrast, Cyanobacteria showed the highest abundance in the A. bifolium endosphere. Because it can resist adversity and promote plant growth through the secretion of various bioactive substances such as protective pigments, plant hormones, iron carriers, and others [73, 74]. Consequently, we inferred that a higher abundance of Cyanobacteria in roots and stems can assist A. bifolium in better growth and resistance to adversity. The distribution of bacterial genera represents their adaptation to different microhabitats [68]. Research indicates that Bacillus (55.1%) and Paenibacillus (8.9%) are both clearly defined genera associated with plants [75]. Bacillus is commonly found in arid lands and is typically utilized for the production of biofertilizers resulting from its spore-forming ability and high resilience to various environmental stressors [76]. Paenibacillus can directly promote crop growth and provide protection against herbivorous insects and plant pathogens through biological fixation of nitrogen, phosphate-solubilizing, production of plant hormone indole-3-acetic acid (IAA), and the release of iron carriers [77]. This study finds that these two genera are the most advantageous genera through culture-dependent methods, which are expected to become potential functional strain resources. In addition, RB41 and Pseudomonas are dominant rhizosphere and nonrhizosphere groups, respectively, which were obtained by culture-independent methods. Research indicates that RB41 participates in carbon cycling through reproduction and respiration processes [78]. We found that RB41 is a dominant rhizosphere bacteria from A. bifolium in summer and winter [21], which may suggest it could form stable relationship with A. bifolium through its carbon cycling. For Pseudomonas, it has been isolated from various locations, including the cold deserts [79] and glacier ice in the Indian Himalayas [80], indicating its resistance to low-temperature environments. Moreover, there were a large number of unclassified genera in culture-independent results, which may suggest that A. bifolium contained specific bacterial groups for maintaining its survival in winter of the Takeermohuer Desert. On the other hand, it also shows that many bacterial resources still need to be explored. Species Associations within the Bacterial Communities in Different Compartments of A. bifolium The ecological importance and potential interactions among core microbial communities were revealed by the co-occurrence network analysis [81]. We found that Proteobacteria and Actinobacteria were identified as key and dominant species in all three compartments in this study, indicating a crucial role in maintaining microbial community functionality and structure [82]. Cyanobacteria is the most dominant and key species in the endosphere, which serves as an input of carbon and nitrogen in the ecosystem [83]. In contrast, Gemmatimonadota does not belong to the dominant species, but it is also a keystone species. A similar case also exists in Verrucomicrobiota and Desulfobacterota in the endosphere. Previous studies have shown that low-abundance taxa are also significantly involved in ecosystem functions [84], such as Gemmatimonadetes. It can reduce the potent greenhouse gas N2O through some ecologically important metabolisms [85]. Higher degree centralization, density, and a lower average path length indicate that the network is more connected and complex [81, 86]. And higher betweenness centrality suggests that such species are likely to have a high influence on other interactions [86]. Our study may suggest that microorganisms in the endosphere not only have a stronger relationship but also a greater complexity compared with rhizosphere and nonrhizosphere. A vast amount of studies have suggested that network complexity correspondingly decreases from soil to plant endosphere with a decrease in bacterial diversity [87]. However, some studies have also reported that plant root systems promote the development of habitats dominated by dominant species, which leads to a simultaneous decrease in diversity, an increase in interactions, and ultimately the emergence of more complex coexistence patterns over time [81]. In addition, endophytic bacteria are able to communicate and interact with the plant more efficiently than rhizobacteria, because their living environment is not affected by the vagaries of changing soil condition (such as soil pH and water content) [88, 89]. In this regard, endophytic bacteria may have a more important effect on the growth and development of A. bifolium in natural environments. However, the function and importance of the microbiota derived from co-occurrence network analysis is only a prediction; their actual function remains to be studied in ecosystems. Influence of Edaphic Factors on Soil Bacterial Community in A. bifolium Habitat Soil physicochemical properties, such as soil moisture and nutrient variations, can impact the distribution and composition of bacterial communities in different compartments [83]. Moreover, specific edaphic factors enrich specific bacterial groups [90]. A study has shown that the bacterial diversity and composition in the rhizosphere are positively correlated with total organic carbon (TOC) and total organic nitrogen (TON) [91]. The phylum Proteobacteria plays key roles in the cycling of C, N, and S and increases with higher content of soil organic compounds [92]. This explains why Proteobacteria in the rhizosphere are positively correlated with STC, SOM, SOC, STN, and SAN. The phylum Acidobacteria is more competitive in nutrient-poor environments and grows well in lower pH [90]. Therefore, Proteobacteria and Acidobacteria much tended to inhabit rhizosphere environments compared with nonrhizosphere in this study. Researchers have proved that the phylum Actinobacteria can tolerate higher pH, which contributes to the decomposition of refractory carbon in soil, resulting in an increase in C/N [12, 93]. Maybe, these results can explain the positive correlation between the dominant phylum Actinobacteria and pH and C/N in the nonrhizosphere. Additionally, we found that this sampling distance of 50 cm from the rhizosphere can distinguish the composition of rhizosphere and nonrhizosphere bacterial communities. The composition of rhizosphere bacteria may be less influenced by dispersal [94], which may be due to an increase in distance from the roots and a decrease in root secretions, resulting in an increase in the recruitment of the microbial communities. However, we cannot completely rule out the effect of dispersal on rhizosphere bacteria in this study. In natural ecosystems, microbial communities show potential nutrient preferences and adaptability to drought and low-nutrient stress, which is further confirmed through the perspective of gene-level transfer and ecological niche differentiation [95]. Previous studies have shown that some bacterial groups (such as Proteobacteria, Acidobacteria, and Actinobacteria) could adapt to cold climates and low temperature [96, 97]. This result indicates that the bacterial groups with higher abundance (e.g., Proteobacteria, Cyanobacteria, Actinobacteria, Acidobacteria, and Planctomycetota) in this study could also adapt to extremely low temperature in Takeermohuer Desert. Therefore, understanding the environmental preferences of bacterial communities contributes to predicting their distribution and further exploring their growth promotion and stress resistance effects on plants, which provide an important theoretical basis for species conservation in desert region. Conclusions In this study, we found that the diversity of A. bifolium bacterial community showed a decline trend from nonrhizosphere, rhizosphere, root endosphere to stem endosphere. And the structures and compositions of the bacterial communities varied in different compartments. The rhizosphere bacterial community was mainly influenced by EC, STC, SOM, SOC, STN, SAN, STP, and SAK, while that of the nonrhizosphere was primarily affected by pH, C/N, distance, and SAP. Furthermore, we also found some valuable bacterial resources, such as Bacillus, which plays a crucial role in soil remediation. Co-occurrence network analyses revealed that Proteobacteria and Actinobacteria are key species in the Takeermohuer Desert. The findings of this study will enhance our understanding of the adaptive mechanisms of A. bifolium in desert winter and promote the implementation of effective species restoration strategies for desert A. bifolium. Supplementary Information Below is the link to the electronic supplementary material.Supplementary file1 (JPG 330 KB)Supplementary file2 (DOCX 24 KB)Supplementary file3 (DOCX 18 KB)Supplementary file4 (DOCX 28 KB) Supplementary file1 (JPG 330 KB) Supplementary file2 (DOCX 24 KB) Supplementary file3 (DOCX 18 KB) Supplementary file4 (DOCX 28 KB)