This study was conducted in in the upper reaches of the Minjiang River, western Sichuan Province (31°35′~31°53′ N, 102°2′~102°48′ E), which is located in the outermost part of the fold belt on the eastern Qinghai-Tibetan Plateau (Xu et al., 2021). Its altitude ranges from 2,200 to 5,500 m. The climate with an average annual temperature of 2~4°C, the highest temperature is 23.7°C in summer, and the lowest temperature is -18.1°C in winter. Annual precipitation is 700~1000 mm and concentrated mainly in the growing season (Chen, 2019). The soil in this area is defined as mountain brown soil, mountain brown cinnamon soil and subalpine meadow soil according to the Chinese soil taxonomic classification (Liu, 2010; Feng et al., 2017; Chen, 2019).

Three different primary forests with different mycorrhizal types were selected, including Abies fargesii var. faxoniana primary forest (EcM forest), Cupressus chengiana primary forest (AM forest) and Rhododendron phaeochrysum primary forest (ErM forest) under similar soil and climate conditions. Eight 15 m ×15 m sample plots (≥ 90% the dominant species by basal area in each sample plot) for each forest type were randomly set. In each forest type, the distance between any two sample plots was more than 50 m.

The mineral soil (0-10 cm) samples were collected from the four corners and center of each plot with soil drill from May to November 2019. Five mineral soil samples were mixed into a zipper storage bag and transported to laboratory in an icebox within 3 h. Meanwhile, five polyvinyl chloride collar cores (PVC cores, 15 cm in height and 5 cm in diameter) were buried into depth of 10 cm in the vicinity of each soil sampling location (Idol et al., 2003). In order to separate water and allow gas movement, the top of the PVC core was covered with a permeable plastic film and its bottom covered with gauze (Liu et al., 2021). The difference between soil N H 4 + − N and N O 3 − − N contents each month was used to quantify the rates of soil net ammonification, net nitrification and net N mineralization (Wang et al., 2017), which was the ideal time to estimate changes in soil N (Becker et al., 2015). For the measurement of soil water-filled pore space (WFPS), we followed the methodology of Wang et al. (2010). The temperature of mineral soil was measured with a soil temperature detector in each plot (Liu et al., 2021). Furthermore, the mixed mineral soil was sifted through a 2-mm sieve and gravel and fine roots were removed (Xu et al., 2021). The fresh mineral soil was separated into two parts: one was dried naturally to measure soil pH, soil organic C (SOC) and soil total N (TN); the other was stored at -20°C to analysis enzyme activity and microbial community structure.

Soil N H 4 + − N and N O 3   − − N contents were quantified using an automatic flow injection analyzer (FIAstar 5000 Analyzer, Sweden). Soil pH was determined using the glass electrode meter method, by setting the 1:2.5 (w/v) ratio of soil material to deionized water. The SOC content was measured by applying the wet oxidation method with K₂Cr₂O₇ and H₂SO₄, and FeSO₄ titration. Soil TN content was determined by the Kjeldahl method (Liu et al., 2020). The soil C: N ratio was calculated as the ratio of SOC to soil TN. Finally, the potential activity of peroxidase (PER) and phenol oxidase (POX) which could drive N from soil organic matter (Jian et al., 2016), were determined via microplate fluorescence and photometry (Zheng et al., 2019).

Samples from May, July and November 2019 were selected for soil fungal functional guilds composition analysis. DNA was extracted from 0.25 g of each soil sample by using the PowerSoil^® Kit(100) and the concentration of extracted DNA was then determined by a NanoDrop 1000 spectrophotometer. Soil fungi were amplified fungal ITS2region and selected primers pairs fITS7 (5’- GTGARTCATCGAATCTTTG-3’) and ITS4 (5’-AGCCTCCGCTTATTGATATGCTTAART-3’) (Xiong et al., 2021). The 25 μL PCR reaction mixture contained 8.5 μL of sterile deionized water, 0.75 μL of each primer, 12.5 μL of KAPA Polymerase and 2.5 μL of diluted template DNA. The PCR amplification as followed: begin with 3 min of incubation at 95 °C, then 35 cycles of 98 °C for 30 s, 56 °C for 30 s, 72 °C for 30 s, with a final extension at 72 °C for 10 min. Sterile deionized water was used as a template (negative controls) to determine whether there was contamination in our experiment (Yao et al., 2019). Three PCR reactions were pooled for each sample to reduce random of the PCR reaction (Bödeker et al., 2016). In addition, the PCR products from each soil sample were purified by the Product Gel Purification Kit. An equal quality of purified PCR product from each sample was pooled in and adjusted to 10 ng μL^-1 (Dong et al., 2021). Sequencing was carried out on an Illumina MiSeq PE platform at the Chengdu Institute of Biology, Chinese Academy of Sciences, China.

The chimeras present in the origin sequencing data were removed using the UNITE database, after which the non-chimeric sequence was quality filtered using Usearch. Remaining sequences were clustered into operational taxonomic units (OTUs) at a 97% sequence similarity. Representative fungal sequences were classified using sintax in the UNITE database with a 0.65 confidence threshold (Edgar, 2016). Finally, the number of sequences per sample was normalized to the smallest sample size using the sub.sample function in MOTHUR (Yao et al., 2019) and then we got 10,952 high-quality reads from 72 soil samples. Soil fungal rarefaction curves ( Figures S1 ) were analyzed by R 4.1 software (using the ‘vegan’ package). Soil fungal functional guilds were assigned using the FUNGuild command in Python 3.7 software (Tedersoo et al., 2014; Nguyen et al., 2016). We excluded OTUs that did not belong to a confidence ranking with “probable” or “highly probable” and that were above the genus level (Pfennigwerth et al., 2018).

A two-way ANOVA were used to analyze the effects of forest type, month and their interaction on the changes in soil temperature and soil WFPS. Likewise, we also used this method to analyze differences in the rates of soil net ammonization, soil net nitrification and soil net N mineralization (Liu et al., 2021). For multiple comparisons, Tukey’s HSD method was used. For a given month, differences in soil properties, soil net ammonization rate, soil net nitrification rate, and soil net N mineralization rate and soil enzyme activities among three mycorrhizal forests were performed by a one-way ANOVA.

Pearson correlations were used to analyze the effects of soil enzyme activity, edaphic variables and soil fungal functional guilds classified by FUNGuilds on soil net ammonization rate and soil net nitrification rate among three mycorrhizal forests. Further, the significant influencing factors (P < 0.05) after the correlation analysis, soil net nitrification rate and soil net ammonification rate were selected for a structural equation (SEM) analysis. Model estimation was achieved based on the maximum likelihood method. The adequacy of model fit were determined by non-significant χ2 tests (P > 0.05), comparative fit index (CFI) (values ≥ 0.9) and standardized root mean square residual (SRMR) (values < 0.08) (Ni et al., 2022).

The SEM analysis only selected the data from May, July and November 2019. The ANOVAs and Pearson correlations were carried out in SPSS 26 software. The SEM analysis was performed using R 4.1 software (using the ‘lavaan’ package). Figures were created by Origin 8.0 software.

Seasonal variation of soil temperature and soil WFPS among three mycorrhizal forests showed significant differences during the study period (P < 0.01) ( Figure 1 ). Soil temperature of each mycorrhizal forest followed a single peak curve, being highest in July and lowest in November. Soil temperature was the highest in AM forest (6.7-18.0 °C), followed by EcM forest (2.8-11.6 °C) and ErM forest (1.0-11.3 °C) ( Figure 1A ). Contrary to soil temperature, soil WFPS in AM forest (50.46%-82.56%) was significantly lower than that in EcM forest (62.41%-120.68%) and ErM forest (80.23%-126.30%) ( Figure 1B ).

As shown in Table 1 , there were significant differences in soil pH and soil TN content, both being highest in AM forest (P < 0.05). SOC content and soil N H 4 + − N content were significantly lower in EcM forest than those in AM forest and ErM forest (P <0.05). Meanwhile, AM forest had the highest soil N O 3 − − N content but there was no significant difference between EcM forest and ErM forest. However, soil C: N was significantly lower in AM forest than the other two forests (P < 0.05).

Forest type, sampling month and their interaction had significant effects on soil net ammonization rate (net R_a), soil net nitrification rate (net R_n) and soil net N mineralization rate (net N_min) (P < 0.01) ( Figure 2 ). Soil net R_a ranged from -0.16 to 0.31 mg kg^-1 d^-1 among three mycorrhizal forests, with the highest rates occurring in July ( Figure 2A ). Soil mean net R_a in ErM forest was negative (-0.09 mg kg^-1 d^-1) being also significantly lower than that in EcM forest (0.10 mg kg^-1 d^-1) and AM forest (0.13 mg kg^-1 d^-1) (P < 0.05) ( Figure 2D ).

The significant effect of mycorrhizal forest types on soil net R_n was detected in May, June and August. Soil net R_n in EcM forest (0.06 mg kg^-1 d^-1) and AM forest (0.24 mg kg^-1 d^-1) peaked in June while that in ErM forest (0.03 mg kg^-1 d^-1) peaked in July ( Figure 2B ). Overall, AM forest had the highest soil mean net R_n (0.06 mg kg^-1 d^-1) whereas it did not differ significantly between EcM and ErM forests ( Figure 2E ).

The trends in monthly variation of soil net N_min were similar to those of soil net R_a among three mycorrhizal forests ( Figure 2C ). There were significant differences in soil mean net N_min among AM forest (0.19 mg kg^-1 d^-1), EcM forest (0.12 mg kg^-1 d^-1) and ErM forest (-0.10 mg kg^-1 d^-1) (P < 0.05) ( Figure 2F ).

In general, 53 fungal classes were identified by excluding the unidentified fungal groups from the 72 soil samples. The top five dominant classes of soil fungi were Agaricomycetes (26.98%-59.96%), Leotiomycetes (9.57%-27.47%), Sordariomycetes (3.55%-15.27%), Eurotiomycetes (3.46%-13.51%) and Archaeorhizomycetes (1.53%-16.13%). By contrast, Dothideomycetes (1.08%-5.06%), Mortierellomycetes (2.36%-8.25%) and Umbelopsidomycetes (0.47%-8.44%) presented low relative abundances in the three forests ( Figure 3 ).

We also obtained three main trophic modes (saprotroph, symbiotroph and pathotroph) and 10 soil fungal functional guilds from the 10952 OTUs. The relative abundance of the unassigned group (40.49%-77.01%) was dominant in our study. Among the assigned OTUs, the relative abundance of EcM fungi (4.14%-52.80%) was the highest, followed by the undefined saprotroph (SAP) (3.80%-15.21%) and ErM fungi (0.48%-5.18%). The relative abundances of the remaining guilds were lower ( Figure 4 ).

Soil PEX and POX activity each differed significantly among three mycorrhizal forests in each month (P < 0.05). Their activity in AM forest and EcM forest peaked in July, while those in ErM forest peaked in November ( Figure 5 ). In general, soil mean PEX activity of EcM forest (41.04 μmol g^-1 h^-1) was significantly higher than the other two forests (P < 0.05) ( Figure 5A ). Soil mean POX activity in ErM forest (18.10 μmol g^-1 h^-1) was significantly lower than in EcM forest (23.13 μmol g^-1 h^-1) and AM forest (22.46 μmolg^-1 h^-1) (P < 0.05) ( Figure 5B ).

The results showed that soil net N_min of the three mycorrhizal forests were affected differently by soil temperature and soil WFPS. Soil temperature and soil WFPS had no significant differences on soil net R_n in EcM forest, but soil net R_a in EcM forest was positively correlated with soil WFPS. Both soil net R_a and net R_n in AM forest were positively correlated with soil temperature but negatively correlated with soil WFPS. In contrast, soil net R_a and soil net R_n in ErM forest were positively correlated with soil WFPS ( Table 2 ).

Further research revealed that the difference in soil net R_a among three mycorrhizal forests was closely related to their soil environmental factors, soil PEX activity and soil fungal functional guilds ( Figures S2 , 6 , 7 ). SEM analysis illustrated that soil temperature, the relative abundances of EcM fungi and ErM fungi had not only direct but also indirect effects on differential soil net R_a among three mycorrhizal forests. Similarly, SAP also indirectly affected the difference of soil net R_a among three mycorrhizal forests by regulating soil PEX activity ( Figures 6A , 7A ).

Different from soil Net R_a, soil environmental factors largely drove the difference found in soil Net R_n among three mycorrhizal forests. Soil temperature had a positive effect on the difference in soil net R_n among three mycorrhizal forests, while soil WFPS and soil N H 4 + − N content had negative effect on it. We also found evidence of an indirect effect of soil net R_a, by changing the soil N H 4 + − N content, on the difference in soil net R_n among three forests ( Figures 6B , 7B ).

There were significant differences in seasonal variations of soil net N_min among three mycorrhizal forests in our research ( Figures 2A-C ). Temporal variation in forest soil net N_min was also suggested in previous studies (Zhao and Li, 2017; Xiao et al., 2022). Overall, our results indicated that soil mean net R_a and soil mean net R_n were positive in EcM forest and AM forest, while there were negative in ErM forest ( Figures 2D-F ). This implied the net production of N H 4 + − N and N O   3 − − N was dominant in EcM forest and AM forest, while the net consumption of N H 4 + − N and N O   3 − − N was dominant in ErM forest. This could be explained by the soil microbial in ErM forest absorbing more N H 4 + − N and N O   3 − − N to maintain the growth and its reproduction of the populations (Miller et al., 2009; Liu et al., 2021).

In our study, soil net R_a and soil net R_n in AM forest decreased gradually after July and June respectively ( Figures 2A, B ). This might be due to the decrease of soil temperature and increase in soil WFPS ( Figure 1 ) which could inhibit soil microbial activity, slowing down soil net R_a and soil net R_n in AM forest (Guntiñas et al., 2012; Hishi et al., 2014). Simultaneously, our findings revealed that the effect of soil temperature on soil net ammonification rate was stronger than that of soil WFPS, while soil WFPS had a stronger effect on soil net nitrification rate than soil temperature in AM forest ( Table 2 ). In contrast to AM forest, soil net R_a and soil net R_n in EcM forest and ErM forest showed trends of increasing in late autumn ( Figures 2A, B ). This was because soil net R_a and soil net R_n in EcM and ErM forests were not determined significantly by soil temperature. Increased soil WFPS promoted both soil net R_a in EcM forest and soil net R_a and soil net R_n in ErM forest ( Table 2 ) which was consistent with previous study (Kou et al., 2018). This might be related to differing ecological tolerance strategies of soil microorganisms to soil temperature and soil WFPS among three mycorrhizal forests (Castro et al., 2010; Placell et al., 2012).

Our results showed that ErM forest had the lowest soil net R_a and there was no difference between EcM forest and AM forest ( Figure 2D ). This might be due to a synergistic effect between soil temperature and different mycorrhizal plants (Tedersoo et al., 2014; Liu et al., 2017; Netherway et al., 2021). In our study, EcM forest had the highest relative abundance of EcM fungi, while the relative abundance of saprophytic fungi was lower ( Figure 4 ). It had been confirmed that the relative abundance of soil EcM fungi increased with the abundance of EcM trees (Cheeke et al., 2017). Previous studies showed that host trees specificity could shape the relative abundances of soil fungi through co-evolution, niche differentiation, and niche conservatism (Tedersoo et al., 2013). EcM fungi have the mutualistic benefits with EcM trees (Velmala et al., 2013), which may be one of the reasons for the relatively high abundance of EcM fungi we found in EcM forest. Moreover, being the main functional groups of soil, EcM and saprophytic fungi would compete for SOC and N, and the strong competitiveness of EcM would limit the relative abundance of saprophytic fungi within the same ecological niche (Chen et al., 2022). Indeed, our results also showed that the relative abundances of soil AM fungi and saprotrophs in AM forests were higher than those in EcM forest and ErM forest ( Figure 4 ), perhaps also due to the fact that the faster nutrient cycling pattern of AM forest (lower C: N, higher inorganic N content) was more suitable for the survival of AM fungi and saprophytic fungi (Bahram et al., 2020).

Our SEM results indicated that soil PEX activity has the strongest direct effect on soil net R_a ( Figures 6A , 7A ). Previous studies showed that soil EcM fungi and SAP could secret PEX to degrade recalcitrant organic matter which could not be absorbed by plants and promote soil net R_a (Bödeker et al., 2014; Corrales et al., 2016; Ward et al., 2021). However, our results showed that ErM fungi were negatively correlated with soil PEX activity ( Figure 6A ) which might be due to microbial C limitation when soil C: N was low (Midgley and Phillips, 2014; Xu et al., 2022). As such, ErM fungi would reduce the energy available for synthase soil PEX (Carrara et al., 2018). At the same time, soil EcM fungi was superior to SAP in degrading soil organic N (Bahram et al., 2020). EcM fungi could compete with SAP to inhibit the relative abundance of SAP and thus affected soil net R_a (Bödeker et al., 2016; Argiroff et al., 2022). In addition, EcM fungi and ErM fungi could directly acquire low molecular weight organic matter (Read and Perez-Moreno, 2003). This should promote soil net R_a by reducing the products of decomposition processes in the first step of ammoniation (Levy-Booth et al., 2014). The SEM also revealed that soil temperature not only directly enhanced soil net R_a, but also had a strong indirect effect ( Figures 6A , 7A ). This principally arose via soil temperature which could affect soil PEX activity and the relative abundance of soil ErM fungi, thus affected soil net R_a (Tedersoo and Bahram, 2019; Fan et al., 2021).

Different from soil net R_a, the SEM suggested that the difference of soil net R_n among three mycorrhizal forests was mainly determined by soil environmental factors ( Figures 6B , 7B ). On the one hand, soil temperature was positively correlated with soil net R_n, but negatively correlated with soil WFPS, a pattern consistent with other research findings (Borken and Matzner, 2009; Liu et al., 2017). On the other hand, soil temperature and WFPS promoted soil net R_a to produce more N H 4 + − N which affected soil net R_n. More recent studies also showed that soil N H 4 + − N as a substrate would accelerate soil net R_n (Elrys et al., 2021). Yet we found that soil N H 4 + − N content was negatively correlated with soil net R_n ( Figure 6B ). This discrepancy may be related to the net dynamics of soil N O   3 − − N being dependent upon substrate content, as well as being affected by microbial absorption, N-fixation, and N-loss (Gao et al., 2015; Xiao et al., 2022).

In this study, SEM indicated that soil environmental factors and soil net R_a explained 53.7% of the difference of soil net R_n among three mycorrhizal forests ( Figure 6B ). However, soil fungal functional guilds had no effect on soil net R_n ( Figure 7B ). This might be due to soil net R_n was driven primarily by soil bacteria (Levy-Booth et al., 2014). Our results showed that AM forest soil also characterized by a high pH and a low C: N ratio ( Table 1 ), which implied AM forest had a fast nutrient cycling (Phillips et al., 2013; Lin et al., 2017). It was reported that AM forest soil with less acidity (high pH) supported higher heterotrophic bacterial activity and greater potential for nitrification (Mushinski et al., 2020). Further, the nitrification pathway of bacteria was superior to that of fungi in soil with a low C: N ratio (Deng et al., 2018). In addition, AM forest had a higher ratio of bacteria to fungi than EcM forest and ErM forest, which also suggested that AM forest was more inclined to have soil net R_n driven by bacteria rather than fungi (Bahram et al., 2020).