The study area is located in the Pusang state-owned forestry farm in Shunchang County, Fujian Province (117°46′E, 26°55′N), with a subtropical monsoon climate. The annual mean temperature is 18.5°C, and annual precipitation is 1,880 mm, with 230 frost-free days and relative humidity of 82%. Chinese fir was planted in 2006 on the same slope direction, at a slope of 22–31° and an elevation of 135–228 m. The initial planting density of Chinese fir is 1,500–1,980 stems ha^–1. The Chinese fir stand was thinned in 2017 to reduce the stocking density to 900, 1,200, and 1,875 stems ha^–1, and the 2 years old broad-leaved trees seedlings of S. superba, P. zhennan, M. macclurei, and T. odorum were planted with the density of 225 stems ha^–1 per species. In each of the three thinned stand, three standard plots of 20 m × 20 m were set up (9 per species of broad-leaved trees). The total C, total N and total P in the soil of Chinese fir forest land before thinning and interplanting were 7.45–27.321, 0.595–1.945, and 13.064–34.313 mg/kg, respectively. The growth of Chinese fir and the four broad-leaved species was recorded annually.

Soil profiles were excavated in the subplots in September 2020 according to the diagonal sampling method (On the diagonal of each standard plot, and then select the upper, middle and lower diagonal, with a total of three points), and soil samples were taken from 0 to 10 cm, 10 to 20 cm, and 20 to 40 cm soil layers with a 100 cm³ cutting rings for the determination of soil physical properties of Chinese fir forests with different thinning reserve densities; at the same time, the chemical soil of each standard plot is collected. After mixing the soil samples of the same layer in the standard plot, that is, as a test soil sample. Each standard plot has three soil samples, a total of 27 soil samples, for the determination of soil nutrients of Chinese fir forest under different thinning reserve densities. In each standard plot, each broad-leaved tree species selects three seedlings with the same growth, collects soil samples within about 50 cm of the bottom of the tree trunk, peels the soil attached to the root, and mixes the same broad-leaved tree rhizosphere soil samples as a rhizosphere soil sample. We collected 36 rhizosphere soil samples from stands with three densities with four broad-leaved trees underplanted. The soil samples were divided into three equal parts, one part was air-dried and ground for soil nutrient content determination, while the remaining part was collected and stored at 4°C for soil nutrient determination. The remaining part was collected and stored at −80°C for total soil DNA extraction.

Soil total C and N contents were measured by elemental analyzer (Elementar, Vario MAX CN, Germany). Total P and total K with hydrofluoric acid–perchloric acid digestion and determined by ICP emission spectrometry (Optima, 7,000 DV, Germany), the available P content was measured by HCl-NH₄F leaching method, and the available K content was measured by the method of ammonium acetate leaching. The soil pH value shall be measured by pH acidimeter after being diluted and shaken by soil: water = 1:2.5 and then standing for 30 min.

Soil DNA was extracted using the CTAB method, DNA quality was detected using 1% agarose gel electrophoresis, and DNA concentration was detected using Nanodrop. The primers were designed to amplify the V3 and V4 regions of the bacterial 16S ribosomal coding sequence, and the primer sequences were (ACTCCTACGGGAGGCAGCAG) and (GGACTACHVGGGTWTCTAAT). The amplification conditions were set at 94°C pre-denaturation for 5 min, 94°C denaturation for 30 s, 55°C annealing for 30 s, 72°C extension for 60 s, 28 cycles, and 72°C extension for 7 min. The amplified PCR products were detected by 1% agarose gel electrophoresis and purified by Agencourt AMPure XP Nucleic Acid Purification Kit, and sequenced by Illumina MiSeq PE300 platform (Beijing Ovison Bioinformatics Co., Ltd., Guangzhou, China).

The Pair-End double-ended Fastq format data obtained from the sequencing were processed using Trimmomatic with a sliding window strategy, with a window size set at 50 bp, an average quality value of 20, a minimum retained sequence length of 120, and sequences with N were removed using Pear software. The reads for each sample were stitched using Flash to obtain the original Tags data. The original sequencing data were deposited in the NCBI SRA database under the accession number PRJNA918639.

Chimeric sequences from the UCHIME Algorithm database were compared and removed by the uchime method, short sequences that did not fit the requirements were also removed, and the high-quality Clean Tags sequences were obtained after quality control filtering. Based on the sequence similarity, the Clean Tags sequences were clustered to produce operational taxonomic units (OTUs), and those with more than 97% sequence similarity were classified as an OUT. The UNITE taxonomic database and the RDP Classifier database were used for species annotation of OUT representative sequences and analysis of sample community composition at each taxonomic level to obtain information on bacterial species in different rhizosphere soil samples.

SPSS 22.0 was used to conduct statistical analyses. We used a one-way analysis of variance (ANOVA) to test for differences in soil properties, bacterial community structure, and microbial diversity. The obtained OUT clustering results and species information were analyzed using Mothur software, and the abundance index of Chao 1 and Observed species, diversity index of Simpson and Shannon were calculated, and the statistical results of species composition and relative abundance of the samples were plotted using the Origin 8.5 software package. Pearson correlation analysis was performed using SPSS 22.0 to analyze the correlation between soil nutrient contents and the structure of dominant bacterial community and bacterial diversity in rhizosphere soil. Canoco5 software was used to analyze the relationship between soil nutrients and dominant bacterial community structure and bacterial diversity in rhizosphere soil by RDA constrained sequencing method.

There was no significant difference in the DBH in different stand densities in 2017 (p < 0.05), but the DBH decreased significantly with the increasing of stand densities after thinning (Table 1). The H and GD of S. superba and T. odorum decreased with the increasing of stand density, that is, they grow better under low density, which indicates the biological characteristics of S. superba and T. odorum that like light. In the 900 and 1,200 stems ha^–1 treatments, the H and GD of S. superba and M. macclurei are higher than those of the other two species, indicating they are more suitable for planting under the Chinese fir plantation at the low and middle stocking density (Table 2).

There were significant differences in the contents of total K, available P and available K in the soil of Chinese fir plantation with different thinning reserve densities (Table 3). The contents of total K, available K and available P were higher in the density of 900 stems ha^–1, while the total nitrogen content of 0–10 cm increased with the increasing of stand density. Under the same stocking density, the total K, total N, total C, available K, available P, pH and water content in the soil decreased significantly with increasing of soil layer. Across all stocking density levels, the contents of total K and available P in the 0–10 cm soil layer were significantly higher than those in the 20–40 cm soil layer. In the 900 and 1,200 stems ha^–1 treatments, while the contents of total N, total C and available K in the 0–10 cm soil layer were significantly higher than those in the 20–40 cm soil layer.

There were significant differences in the contents of total C, total N, total K, available K and available P in the rhizosphere soils of the four broad-leaved tree species in the same stocking density, while there were no significant differences in the pH and total P (Table 4). The rhizosphere soil contents of available P, total N, total C and total K increased with the decreasing of stocking density. At low stocking density (900 stems ha^–1), the rhizosphere soil contents of total C, total N, total K and available K increased in the following order: S. superba > M. macclurei > T. odorum > P. zhennan. At medium (1,200 stems ha^–1) and high (1,875 stems ha^–1) stocking density, the rhizosphere soil contents of total C, total N, total K, and available K were significantly higher in M. macclurei than in other species, while the available P content was significantly higher than in T. odorum and P. zhennan. In general, Chinese fir plantations underplanted with S. superba and M. macclurei had higher rhizosphere soil nutrient contents than T. odorum and P. zhennan.

The rhizosphere soil bacterial diversity of S. superba, P. zhennan and T. odorum were significantly different at different densities, but there was no significant difference in M. macclurei (Table 5). The highest rhizosphere soil bacterial diversity of S. superba appeared under the density of 1,200 stems ha^–1. The Chao 1, the number of observed species and the PD whole_tree and Shannon indices of P. zhennan increased with the increasing of stocking density, revealing the rhizosphere bacterial diversity of P. zhennan was richer under high-density Chinese fir plantation. The Chao 1 and the Shannon indices of rhizosphere soils of T. odorum decreased with the increasing of stocking density indicating that T. odorum may be more suitable for high-light environment, and T. odorum planted under the low-density Chinese fir forest has higher rhizosphere bacterial diversity.

At the phylum level, 11 most abundant bacteria shared by the rhizosphere soil of four tree species (Figure 1). Acidobacterium, Proteobacterium, and Chloroflexi had relative abundances greater than 12% and Actinomycetes and Verrucomicrobia had relative abundances greater than 2%, making them the dominant groups among the rhizosphere soil bacteria. The bacterial community composition structure of different broad-leaved trees under the same stocking density is different, but there is no obvious rule. The bacterial community composition structure of the same broad-leaved tree species under different stocking densities is also different, with the relative abundance of Acidiobacter in the rhizosphere soil of T. odorum decreased with increasing of stocking density, while the opposite trend was observed for the Acidiobacter in P. zhennan. The relative abundance of the phylum Acidiobacter and Chloroflexi in the rhizosphere soil of M. macclurei increased as the stocking density increased. The relative abundance of the phylum Metaphyllobacteria in the rhizosphere soils of M. macclurei decreased with the increasing of stocking density. The genus Acidobacteria had the highest abundance in the four broad-leaved tree species (Figure 2), and Acidiobacter showed an increasing trend in M. macclurei and P. zhennan with the increasing of stocking density, while it decreased in T. odorum and S. superb with increasing of stocking density.

The correlation analysis between the stand soil nutrient contents and rhizosphere bacterial community structure showed that soil effective P, total N and total K contents were negatively correlated with Chloroflexi, the available P content was highly positively correlated with Proteobacteria, the available P content was positively correlated with Gemmatimonadetes, and the soil pH was negatively correlated with Verrucomicrobia (Table 6). Redundancy analysis (RDA) revealed that the structure of rhizosphere bacterial community was related to soil nutrients, and the bacterial community structure is mainly affected by available P, pH and available K (Figure 3). The total interpretation rate of the seven soil physical and chemical indicators reached 74.94%. The interpretation rate and contribution rate of the soil available P content were the highest (37.8 and 47.8%, respectively), indicating that the soil available P content had a great impact on the bacterial community, followed by the soil available K content (17.6 and 22.7%, respectively). There was a significant positive correlation between soil available P and Nitrospirae and Bacteroides, a significant negative correlation between total P and Actinobacia, and a significant negative correlation between soil total N and total K content and Chloroflex. The contents of total N and total K were significantly positively correlated with Gemmatimonadetes.

The correlation analysis between rhizosphere soil nutrient contents and bacterial diversity of the rhizosphere soil showed that the Shannon diversity index was positively correlated with available K, total P and total K content (Table 7), with total N content being highly significant and positively correlated with the Shannon index. Available P content is positively correlated with species number, PD_whole_tree index and Shannon index. Total N and available K content were positively correlated with observed species number. In contrast, total C content and pH were not significantly correlated with bacterial diversity. The bacterial diversity in the rhizosphere soil of broad-leaved trees is mainly affected by the available P, total P, and pH (Figure 4). The total interpretation rate of the seven soil nutrients reached 79.7%, and the interpretation rate and contribution rate of soil available P content were the highest (28.3 and 35.5%, respectively), indicating that the soil available P content had a great impact on bacterial diversity. The bacterial diversity index was positively correlated with soil nutrients, and PD_whole_tree, shannon and observed spries number index were significantly positively correlated with the available P and pH.

Thinning and interplanting change the species composition of understory vegetation, which then leads to differences in soil physicochemical properties (Pretzsch et al., 2021). Low stocking density decreases the competition between trees, leading to increase in soil nutrients and microbial diversity, which is conducive to the growth of trees and heterophyllous forests (Shu et al., 2021). In this study, the soil physicochemical properties and the growth of Chinese fir were the best under lower thinning reserve density. A short-term of thinning study after 3 years revealed the soil total N increased with the increase of thinning reserve density in Chinese fir plantation (Wang Y. L. et al., 2021). This should be attributed to the effect of thinning on soil nutrition of fast-growing conifer species, which may be more favorable for nutrient accumulation in low-density stands. The soil nutrient content decreased with the increase of forest density, and that the low soil physicochemical properties and medium density of forest were better than those of high density of forest stands (Ding et al., 2021). The pH, total P and physical properties of the surface soil were not significantly affected by thinning intensity after 3 years of thinning on Chinese fir plantation, which may be caused by short-term thinning effect, which is consistent with the results of this study (Xu et al., 2019).

Stand density not only affects the growth of trees, but also significantly affects the growth of underplanted species. With the growth of four broad-leaved trees gradually decreasing as the density of the thinning reserve increased, and the best growth being at 900 stems ha^–1. The growth rates of M. macclureiand and S. superba were significantly higher than those of P. zhennan and T. odorum when the thinning reserve density was 900 and 1,200 stems ha^–1. However, the different underplanted tree species can also affect soil physicochemical properties through their litter input and root exudates (Mayer et al., 2020). Thinning and interplanting not only have an effect on the physicochemical properties of plantations, but also have a significant effect on the rhizosphere soil physicochemical properties of broad-leaved trees underplanted in plantation (Huang et al., 2021). Huang et al. (2021) demonstrated that interplanting of three broad-leaved tree species, Castanopsis hystrix, Mytilaria laosensis, and M. macclureiand under different thinning intensity treatments of eucalyptus plantation improve the nutrient content of eucalyptus continuous cropping forest soil to a certain extent, and the modified forest stands with low stand density have high contents of organic matter, total P, total N, available P, available N and available K. In this study, the rhizosphere soil nutrients of the four broad-leaved trees were the highest under low stand density level (900 stems ha^–1). The available P, total N, total C and total K content of the four broad-leaved trees decreased with the increase of thinning reserve density. The rhizosphere soil nutrients of T. odorum decrease with increasing thinning reserve density, and it grows better at low thinning reserve density, consistent with its biological preference for better light and warmer climates. We believe that the differences in rhizosphere physical and chemical properties among different underplanted broad-leaved tree species could be due to the interspecific differences in root respiration and nutrient uptake among different tree species, resulting in the quantities and quality of litter and root exudates differences. This, in turn, results in changes of bacterial community structure and microbial diversity in rhizosphere soil (Xu M. P. et al., 2021). Thus, thinning and interplanting significantly affected the nutrients of the rhizosphere soil of broad-leaved trees, indicating that these two tending measures had a significant impact on improving the physicochemical properties of rhizosphere soil in Chinese fir plantation.

The amount and nutrient composition of litter and root exudates produced by different broad-leaved tree species are different, so also the litter decomposition and nutrient release rates, which has a selective stimulating effect on soil microbial growth, thereby affecting the microbial community structure and diversity (Zhang et al., 2021). In this study, there were six main types of bacteria in the rhizosphere soil of the four broad-leaved trees, and the abundance of Acidobacteria, Proteobacteria and Chloroflexi were significantly higher across different thinning reserve densities than the other bacterial phyla, which could be related to the larger width of the ecological niche and the low impact on the environment (Xu et al., 2020). Moreover, Acidobacteria has a strong ability to adapt to the environment, and most of them are acidophils (Kalam et al., 2020). As a fast-growing conifer species with continuous planted in the acidic red soil area of south China, the abundance of the Acidobacteria in soil is reasonable. The relative abundance of Acidobacteria in the rhizosphere soil of P. zhennan was significantly higher than that of other broad-leaved trees under different thinning reserve densities, which may be due to the secretion of more acidic substances such as organic acids and protons into the rhizosphere soil. The relative abundance of Acidobacteria decreased with the increase of thinning reserve density in the rhizosphere of T. odorum, while the relative abundance of Acidobacteria showed the opposite trend, indicating that the Acidobacteria in the rhizosphere of T. odorum and M. macclureiand were more sensitive to thinning response.

Rhizosphere soil bacterial community structure and diversity are influenced by environmental factors, which focus on the rhizosphere nutrition, pH and carbon source (Gao et al., 2019). In this study, there was a clear correlation between the soil bacterial community diversity and structure and the chemical properties of the rhizosphere soils of the four broad-leaved trees. The soil bacterial diversity of the rhizosphere soils was significantly positively correlated with available K and total N content. Studies on the environmental adaptability of bacterial communities in rice rhizosphere soil showed that available K played a decisive role in regulating bacterial community assemblage (Tian et al., 2022). The study on Camellia sinensis rhizosphere soil from Qinling Mountains revealed N content had the greatest effect on the bacterial diversity at genus level, and C, K and pH were positively correlated with Pedobacter and Mucilaginibacter genera (Wang M. J. et al., 2021). However, our results revealed the relative abundance of Acidobacteria and Proteobacteria positively correlated with total N while the relative abundance of Chloroflexi showed a significant negative correlation with total N. Studies on the soil bacterial diversity of four pinus species forests revealed that Acidobacteria was correlated with pH, and the Chloroflexi was correlated with total N (Deng et al., 2019). Fei et al. (2020) also found that the relative abundance of Proteobacteria was positively correlated with the soil total N content. In this study, the total N content was the highest at lower thinning reserve density, and the relative abundance of Proteobacteria was also the highest. The soil in stands with intermediate (1,200 stems ha^–1) and high (1,875 stems ha^–1) density is poorer, and the relative abundance of the Chloroflexi was higher. In general, the soil available K and total N content has an important effect on rhizosphere bacterial community composition.

The available P content of the rhizosphere soil of the four broad-leaved trees significantly positively correlated with bacteria diversity indices, mainly the number of species, the PD_whole_tree index and the Shannon index. The results of redundancy analysis also showed that the rhizosphere bacterial diversity of four broad-leaved trees was mainly affected by soil available P. It is consistent with previous study that demonstrated a significant positive correlation between available P in subtropical Chinese fir soil and various microbial communities (Guo et al., 2022). The Chinese fir is continuous planted in the subtropical red soil area, where the soil is strongly aluminized and strongly acidic with the pH 4.0–5.5. Most of the available P is fixed by Fe, Al, etc., forming Fe-P, Al-P that are difficult to use by the plant, thus it is the main limiting factor for Chinese fir plantation productivity (Wu et al., 2011). In this study, the available P was revealed as the main environmental factor affecting the community structure of four broad-leaved tree rhizosphere bacteria. In forest ecosystems, the availability of P in soil depends on the mineralization, desorption and release, and the content of availability of P in different soil types and vegetation types varies greatly (Wang et al., 2020). The available P content had a significant effect on bacterial diversity in Chinese fir soil, and bacterial diversity in soil was positively correlated with AP content (Fei et al., 2020). Ding et al. (2018) found that the available P content in the rhizosphere soil of four different tree species in the Yellow River Delta was different, and the flavonoids, phenolic compounds citric acid, malic acid, oxalic acid and ferulic acid in the root exudates of different tree species were considerably different, which then led to differences in bacterial diversity and structure of rhizosphere soil. The symbiosis of different broad-leaved tree species with mycorrhiza (Lu et al., 2022), litter decomposition (Zhang H. et al., 2019) and root exudates (Xia et al., 2019) can also affect the availability of soil P in the rhizosphere (Hu et al., 2006). The carbon and nitrogen nutrients secreted by plant rhizosphere and pH of rhizosphere effectively change the composition of rhizosphere bacteria, but the P solution of rhizosphere soil bacteria can also provide an important source of available P for plant rhizosphere, and this interactive symbiotic relationship affects the composition of plant rhizosphere bacterial community (Ding et al., 2017; Bu et al., 2020; Bulgarelli et al., 2022). Therefore, soil available P content could have an important impact on the bacterial diversity in the rhizosphere of broad-leaved trees, and is the main environmental factor affecting the bacterial community structure in the rhizosphere of these four broad-leaved trees planted under different densities of Chinese fir plantation.