The study area is located at the forest region in Weichang County, Hebei Province, China (from 41°35′ N to 42°40′N, 116°32′E to 117°14′E) ( Figure 1 ). The elevation ranges from 663 to1935 m. This region has a typical cold-temperate continental monsoon climate, with a mean annual temperature ranging from -1.4 to 4.7°C and a mean annual precipitation of 380 to 560 mm. The main tree species include larch, birch, Picea asperata, and Populus davidiana.

Three typical stand types were selected in the region at altitudes of 1300–1600 m, all with similar slopes and aspects: larch forest (LF), birch forest (BF), and mixed larch and birch forest (MF). The birch forests are natural secondary forests, while the mixed forests typically result from the artificial replanting of larch in the understory of BF. The larch forests studied in this study are first-generation plantations established on land that had experienced severe ecological degradation. Prior to the Qing Dynasty, the region was primarily covered by natural forests and grasslands. However, extensive human disturbances—such as overlogging, intensive grazing, and frequent wildfires—led to significant ecosystem deterioration. By the mid-20th century, these activities had nearly eradicated the native forests. In response, systematic restoration efforts were initiated, with large-scale afforestation projects implemented using larch as the dominant species. Tree ring analysis indicated that the trees are approximately 40–50 years old. For each stand type, seven plots (30 × 30 m) were established, yielding a total of 21 plots ( Figure 1 ). All standing trees with a diameter at breast height (DBH) ≥ 5 cm were numbered, and their species name, DBH, height, and other relevant information were measured and recorded. The details of the sample plots are presented in Table 1 .

The main effects of stand type on soil physicochemical properties, SQI, and microbial alpha diversity were examined using one-way analysis of variance (ANOVA), following tests for normality and homogeneity of variance. Multiple comparisons among different stand types were performed using Tukey’s HSD test (P< 0.05). Microbial beta diversity was assessed by Principal Coordinate Analysis (PCoA) based on the Bray-Curtis distance matrix, using the R package “vegan”. The analysis of similarity (ANOSIM) was conducted with the “anosim” function in R to compare the similarity in soil microbial community structure. Potential ecological functions were predicted using FAPROTAX and FUNGuid. Bacterial and fungal diversity were quantified by z-transforming the Chao 1 and Shannon indices, respectively. These diversity indices were then regressed to explore variation in SQI in relation to bacterial and fungal diversity, microbial network complexity, and the abundance of keystone taxa. All data analyses were conducted using R 4.3.2 software (R Core Team, 2023).

The results indicated that all metrics varied significantly among the stand types (P< 0.05) ( Table 2 ). The contents of SOC, TN, AN, AP, and MWHC in MF were similar to that in BF but higher than in LF. No significant difference was observed in the TP content between LF and BF, although it was significantly lower in MF. BD was higher in LF, while pH was higher in BF. Additionally, stand type had a significant effect on SQI (P< 0.05). MF exhibited the highest soil quality, which did not differ significantly from that of BF, whereas LF had the lowest soil quality.

The Chao1 richness index for soil bacteria and fungi showed no significant differences among the stand types (P > 0.05, Figures 2A, C ). However, the Shannon index diversity for bacteria and fungi in MF and BF was similar, with significant differences observed when compared to LF (P< 0.05, Figures 2B, D ). Specifically, the bacterial Shannon index was lower in MF and BF, while the fungal Shannon index was higher. The abundance distribution of bacterial and fungal phyla varied among the stand type. The dominant bacterial phyla were Actinobacteria, Proteobacteria, and Acidobacteria, collectively accounting for approximately 75% of the total bacterial population ( Figure 2E ). The most abundant fungal phyla were Basidiomycota, followed by Ascomycota, which together accounted for over 75% of the total fungal population ( Figure 2F ). The five most abundant bacterial taxa were norank_c_Subgroup_6, Candidatus Udaeobacter, Xanthobacteraceae, norank_c_KD4-96, and Mycobacterium, which collectively comprised 23.74% of the bacterial community. Similarly, the five most abundant fungal genera were Sebacina, Russula, Exophiala, Mortierella, and Inocybe, which comprised 27.54% of the fungal community, on average ( Supplementary Figure S2 ).

PCoA based on Bray-Curtis distance was performed to assess the isolation of soil bacterial and fungal communities. The results showed no significant difference in bacterial community beta diversity among the three stand types, as confirmed by similarity analysis (ANOSIM, R = 0.036, P = 0.391, Figure 3A ). In contrast, fungal communities from different stand types formed distinct clusters (ANOSIM, R = 0.391, P = 0.002, Figure 3B ). The first two principal components of the PCoA explained 13.95% and 10.12% of the variation in bacterial communities, and 24.72% and 6.66% of the variation in fungal communities, respectively.

To determine the impact of stand types on microbial network complexity, various topological parameters were extracted for networks from different stand types. The LF network consisted of 505 nodes and 2085 edges, the BF network had 851 nodes and 2801 edges, and the MF network comprised 1031 nodes and 12,837 edges. Compared with LF, both MF and BF networks exhibited higher average degrees. In LF co-occurrence networks, microbial interactions were primarily positive, whereas MF and BF networks showed an increased ratio of negative edges compared to LF. Furthermore, the MF network had a lower network diameter and average path length (13 and 4.439, respectively) compared to the other networks ( Figure 4A ; Table 3 ). Overall, network complexity decreased in the order of MF, BF, and LF. Based on random species loss or targeted removal of module hubs, the MF network demonstrated significantly higher robustness (P< 0.05) than LF ( Figures 4B, C ). In line with network complexity, network stability followed the order of MF, BF, and LF.

Bacterial functional prediction was performed using the FAPROTAX database for annotation. As shown in Figure 5 , the proportion of functional taxa involved in N cycling was higher in MF than in the other two stand types, including taxa related to nitrite respiration, nitrification, and denitrification. Furthermore, the overall abundance of chemoheterotrophic taxa was relatively high, especially in MF. The LF stand exhibited a higher abundance of taxa associated with aromatic compound degradation. In contrast, the relative abundance of cellulolysis-related functional taxa involved in carbon fixation was higher in BF. The FUNGuild database was used to predict the ecological functions of fungi, revealing a higher proportion of “saprotrophs” and “pathotrophs” compared to “symbiotrophs”. Furthermore, functional taxa associated with plant pathogens, plant saprotrophs, and animal pathogens were generally more abundance in MF. The functional taxa of ectomycorrhizal and wood saprotroph were more prevalent in LF, while BF exhibited higher abundances of soil saprotrophs, endophytes, and undefined saprotrophs.

Soil bacterial diversity (R ² = 0.30, P< 0.05) was significantly negatively associated with soil quality ( Figure 6A ). However, both fungal diversity (R ² = 0.24, P< 0.05) and network complexity (R ² = 0.68, P< 0.001) were positively associated with soil quality ( Figures 6B, C ). Keystone taxa, which play key roles in shaping the network structure, were identified through the Zi-Pi plot ( Supplementary Figure S1 ). A total of 21 module hubs (nodes with high connectivity to other members in a module) and 24 connectors (nodes linking different modules) were detected as keystone taxa. The number of keystone taxa was higher in MF (23) and BF (17) than in LF (5). At the phylum level, keystone bacterial taxa included Acidobacteria, Proteobacteria, Actinobacteria, Verrucomicrobia, and Chloroflexi, while fungal keystone taxa comprised Ascomycota, Mucoromycota, Basidiomycota, and Mortierellomycota ( Supplementary Table S4 ). Additionally, soil quality was significantly related to the presence of keystone taxa (R ² = 0.30, P< 0.05), with soil quality increasing in proportion to the relative abundance of keystone taxa ( Figure 6D ).

In the present study, significant differences in several soil properties were observed among stand types. The contents of SOC, TN, AN, and AP in MF were similar to that in BF, but higher than in LF ( Table 2 ). This result is consistent with the findings of Li et al. (2023), indicating that mixed forests enhance nutrient availability and improve soil nutrient cycle (Liu et al., 2017; Likulunga et al., 2021). The observed effects can be attributed to the quality of plant litter (Thoms et al., 2010). Generally, coniferous forest litter has higher C/N ratios and lignin content. with a higher proportion of recalcitrant compounds compared to broadleaf forest litter (Cornwell et al., 2008). This characteristic leads to slower decomposition rates in larch forests, which may reduce soil fertility (Ge et al., 2023). In contrast, the MF increases litter diversity and accelerates the accumulation of soil humus and other nutrients, thereby promoting soil fertility (Mueller et al., 2015; Wang et al., 2021). Additionally, bulk density (BD) was higher in LF stands and relatively lower in BF and MF stands ( Table 2 ). The lower BD, looser structure, and increased pore space in MF stands contribute to stronger water and nutrient retention capacity (Xiang et al., 2023). In this study, the pH was higher in BF stands ( Table 2 ), likely due to the higher exchangeable cations, such as Mg²⁺ and Ca²⁺, found in broadleaf forest litter, which enhance soil buffering capacity and raise pH (Tóth et al., 2011). In summary, compared to coniferous forests, mixed coniferous and broad forests exhibited higher nutrient availability.

Forest stand types are likely to influence soil microbial community characteristics indirectly by altering soil physicochemical properties. Although the Chao1 index of bacteria and fungi showed no significant differences among stand types, significant differences were found in the Shannon index ( Figures 2A, C ). Specifically, the bacterial Shannon index was lower in MF and BF stands, while the fungal Shannon index was higher in these stands. Soil factors were closely related to the diversity of both soil bacteria and fungi (Yarwood and Hogberg, 2017). Bacterial diversity exhibited a negative correlation with SOC, TN, TP, AN, and AP, while fungal diversity showed a significant positive correlation ( Supplementary Figure S3 ). Soil nutrient content was higher in MF and BF stands compared to LF stands, with elevated levels of SOC, TN, TP, AN, and AP ( Table 2 ). Consequently, MF and BF stands likely provide favorable microhabitats for fungal activity and growth by enhancing soil nutrient availability, which in turn increases fungal diversity. In this study, bacterial community composition did not differ significantly across stand types, but the fungal community was distinctly separated by stand type. Ascomycota contribute to the decomposition of litter and organic matter and are generally considered cellulolytic, with limited lignin-decomposing ability (Osono, 2007). The increase in litter diversity observed in MF stands, which is associated with higher levels of SOC and other nutrients ( Table 2 ), may explain the increased relative abundance of Ascomycota in MF stands. The abundance of Ascomycota in BF stands was similarly high, potentially due to the elevated pH levels ( Figure 2F ; Table 2 ). Previous studies have shown that higher pH levels correlate with greater Ascomycota abundance (Hou et al., 2024). In contrast, Basidiomycota are more strongly correlated with litter biomass and are well-known for their ability to decompose lignin, a component of litter that is otherwise difficult to break down (Lundell et al., 2010). Given that conifer litter is more challenging to decompose than broadleaf litter (Chen et al., 2019a), non-degradable carbon accumulates more readily in LF stands, leading to a higher abundance of Basidiomycota in these stands.

Differences in microbial co-occurrence network structure, including complexity and stability, were observed among the three stand types. The soil microbial network in MF stands was found to be more complex and stable ( Figure 4A ), consistent with the findings of Xu et al. (2023). Additionally, the MF stands had a relatively high ratio of negative edges in bacterial taxa networks, whereas the LF stands displayed a greater ratio of positive edges in fungal taxa networks ( Table 3 ). This pattern suggests that tree species mixing in MF stands increases potential competition among bacterial taxa, while LF stands facilitate greater potential cooperation among fungal taxa. As highlighted in previous studies, the structure of soil microbial communities is susceptible to environmental factors, especially changes in soil physicochemical properties (Nkongolo and Narendrula-Kotha, 2020; Zhang et al., 2023). Specifically, the contents of SOC, TN, TP, BD, and pH mainly affected the bacterial community ( Supplementary Figure S4A ), whereas SOC, TN, AN, MWHC, and pH were the key factors (P< 0.01) influencing the fungal community ( Supplementary Figure S4B ). Consequently, stand types can alter the composition of keystone microbial taxa (Li et al., 2023). More keystone taxa were identified in MF and BF networks compared to the LF network, with most taxa belonging to Proteobacteria, Actinobacteria, Verrucomicrobia, Ascomycota, Basidiomycota, or Mortierellomycota. Furthermore, more keystone fungal taxa than bacterial taxa were observed ( Supplementary Figure S1 ; Supplementary Table S4 ). These keystone taxa, identified based on their topological roles as module hubs and connectors, play central roles in soil biogeochemical processes by mediating inter-module substrate exchange and maintaining network stability (Banerjee et al., 2018; Li et al., 2023). Their removal has been to disrupt both the structure and function of soil microbiomes (Berry and Widder, 2014). Many of these keystone taxa are associated with soil nutrient cycling and transformation. For example, keystone connectors such as the Mesorhizobium genus have been found to enhance soil nitrogen availability by linking nitrogen-fixing modules with root-associated networks (Zhong et al., 2019; Laranjo et al., 2014). Meanwhile, module hubs like the Saitozyma genus contribute to organic matter decomposition by coordinating lignin-degrading modules (Lundell et al., 2010).

It was observed that stand type influenced SQI, with SQI being generally higher in MF and BF compared to LF ( Table 2 ). The superior soil properties of MF stands were the primary reason for their highest SQI. Consistent with prior findings, soil with relatively high levels of SOC, TN, AN, and lower BD tend to exhibit improved soil quality (Shao et al., 2020). Plants contribute carbon and energy to the soil through root secretions and litter. In stands with mixed birch and larch, increased tree species diversity and litter return enhance soil quality by providing organic matter inputs and improving nutrient availability (Bu et al., 2019). Microorganisms, as key drivers of soil ecological restoration, are influenced by resource allocation strategies and regulated by shifts in community metabolism (Chen et al., 2022). Thus, stand type can indirectly alter soil microbial communities, thereby affecting soil quality. It was demonstrated that the bacterial Shannon index was lower in MF stands, while the fungal Shannon index was higher. Further analyses indicated that the SQI exhibited a strong negative relationship with soil bacterial diversity ( Figure 6A ), whereas a significant positive association was observed with the increase in fungal diversity ( Figure 6B ), consistent with previous findings (Wan et al., 2024). This pattern may be attributed to the fact that higher bacterial diversity can reduce interspecies cooperation and intensify competition for energy and nutrients (Palmer and Foster, 2022), ultimately exerting a negative impact on soil quality. In contrast, the positive association between fungal diversity and soil quality is likely due to the key roles of fungi in decomposing soil organic matter (Zhong et al., 2018; Yan et al., 2020). Specifically, saprophytic fungi are particularly important in breaking down complex organic materials (Wan et al., 2024), facilitating the release of nutrients. Additionally, many fungi form biotrophic relationships with plant roots (Zhong et al., 2018; Liu et al., 2024), enhancing nutrients absorption by plants and improving soil quality through increased plant litter input (Wan et al., 2024). Moreover, soil nutrient availability and substrate quality are known to influence specific microbial taxa (Xu et al., 2021). In this study, a greater number of keystone taxa were identified in the MF microbial network, and their relative abundance showed a positive relationship with SQI ( Figure 6D ). Soil quality was significantly related to these keystone taxa through their role in shaping the structure and function of microbial co-occurrence networks (Berry and Widder, 2014; Banerjee et al., 2018). Additionally, it was observed that soil quality improved as microbial network complexity increased ( Figure 6C ).

We used FAPROTAX and FUNGuild software to further predict the ecological functions of bacterial and fungal taxa in different forest stand types. The relative abundance of bacterial taxa associated with nitrite respiration, nitrification, denitrification, and chemoheterotroph was higher in MF stands ( Figure 5A ). These bacterial taxa play essential roles in the cycling of organic matter and nitrogen in ecosystems (Gu et al., 2019), enhancing nutrient availability in MF stands ( Table 2 ) and alleviating resource limitations for certain microbial taxes. Regarding fungal taxa, the proportions of “saprotroph” and “pathotroph” were higher in MF and BF stands, whereas ectomycorrhizal and wood saprotroph taxa were abundant in LF stands ( Figure 5B ). This distribution may be attributed to differences in soil nutrient availability, litter quality, tree species characteristics, and the colonization strategies of fungal communities (Xu et al., 2021). As an ectomycorrhizal species, larch produces litter with high lignin content (Cornwell et al., 2008), which promotes the colonization of decomposer communities adapted to lower nutrient conditions (Asplund et al., 2019). In conclusion, stand types could influence microbial network complexity and stability by regulating keystone taxa, thereby affecting soil quality and ecosystem functions.