3.1 Soil physicochemical properties
The selected physicochemical characteristics of investigated soil samples are presented in Fig.1. Compared with other indicators, soil pH was generally stable having values between 5.0 and 5.5, with topsoil values being slightly higher topsoil than subsoil values. In both top- and subsoil, SOC, TN, AN, and AK showed a fluctuation pattern of significantly elevated levels for IM6 and IM20. Soil NH4+-N contents tended to decrease with increasing duration of intensive management in the topsoil, a significant difference (P < 0.05) has been only observed between CK and IM6, while it was just opposite in the subsoil. In contrast to NH4+-N, soil NO3--N increased sharply after IM6 (P < 0.05), and then drop to the level as CK. The variation ranges of soil δ 13C in top- and subsoil are -26.875—-25.935‰ and -26.428—-24.645‰, respectively. IM15 had the highest value of δ 13C compared to other treatments, and significantly higher value (P < 0.05) was observed in CK and IM15 compared to the rest treatments. The variation ranges of soil δ 13N in top- and subsoil are 3.45-5.86‰ and 5.40-6.36‰, respectively. Theδ 13N value for CK was significantly lower than those for treatments of IM in the topsoil (P < 0.05), while it was generally stable in subsoil (F = 2.074, P = 0.159).
The cbbL and nifH gene abundance and their relationship with soil properties
The abundance of cbbL gene along the treatments ranged from 1.33×109 to 2.44×109copies g-1 soil and 5.00 ×108 to 6.81×108 copies g-1 soil for topsoil and subsoil respectively (Fig.2a), and the differences observed between layers was significantly different (P < 0.01). The abundance ofcbbL decreased significantly in IM6, IM10 and IM15 compared to CK in topsoil (P < 0.05) and then recovered to the original level in IM20. Two-way ANNOVA analysis of variance showed that the abundance of cbbL gene was affected significantly by both IM (F = 15.147, P = 0.000) and soil depth (F = 384.892, P = 0.000), but the latter had a much greater effect than the former. ThecbbL gene copy numbers were positively correlated with AK (r = 0.632, P < 0.05) in the subsoil, and with AN (r = 0.688, P < 0.05) and AK (r = 0.690, P< 0.05) in the subsoil, respectively.
The abundance of nifH gene ranged from 1.54×106to 2.31×107copies g-1 soil in the topsoil and from 1.43 ×106 to 1.62×107copies g-1 soil in the subsoil (Fig.2b). As for cbbL , the nifH gene copy numbers in both layers decreased significantly (P < 0.05) at IM6. Copy numbers increased gradually from IM10 to IM20 but remained lower than for the CK (P < 0.05). Two-way ANOVA showed that the abundance ofnifH gene was affected significantly by IM (F = 11.872, P= 0.000) rather than soil depth (Table.2). It was found that thenifH gene copy number was positively correlated with NH4+-N (r = 0.655, P< 0.01) in the topsoil and with C:N (r = 0.628, P< 0.05) in subsoil, while negatively with NH4+-N (r = 0.773, P < 0.01) in subsoil.
Community analysis of the cbbL-andnifH-containing bacteria
A total of 17 T-RFs was identified from all samples and used to comparecbbL -containing bacteria numbers among the treatments (Fig.3a). The six predominant T-RFs, having lengths of 40, 44, 168, 175, 177, and 360 bp, varied in the relative abundance among soils. The T-RF lengths of 40 bp (9.7-24.1%) and 177 bp (22.1-54.4%) were among the most dominant for all treatments. In topsoil, the relative abundance of T-RFs having 44bp and 360bp decreased sharply after six years of intensive management, whereas the T-RF of 360bp decreased sharply in subsoils after 10 years of IM. The T-RF having 168bp was observed only after after 15 years of IM in both topsoil and subsoils. Two T-RFs, 439bp and 488bp, were unique to subsoils. The most dominant T-RF 177bpwas closely related to several species of α -Proteobacteria includingBradyrhizobium sp. (CP013949.1), Rhodospirillum centenum(CP000613.2), Mesorhizobium cicero (CP015064.1), andStarkeya novella (CP002026.1), one species ofβ -Proteobacteria named Stappia meyerae (EF101506.1), and one γ-Proteobacteria named Thioflavicoccus mobilis (CP003051.1). The T-RF 44 bp most closely matched Starkeya novella(CP002026.1). The 360bp T-RF was especially abundant in soil from the CK and IM6 treatments and most closely matched the three speciesActinopolymorpha singaporensis (LT629732.1), Mesorhizobium ciceri (CP015064.1), and Rhodospirillum centenum (CP000613.2). The minor speciesStarkeya novella (CP002026.1) and Bradyrhizobium sp.(CP013949.1) were represented by the T-RF 364bp, whereasThermomonospora curvata (CP001738.1) and Stappia meyerae(EF101506.1) were represented by T-RFs 129 and 439bp, respectively.
A total of 17 T-RFs was identified from all samples and used to analyze communities shift of nifH -containing bacteria (Fig.3b). The relative abundance of the five dominant T-RFs of 68, 154, 177, 180, and 332 bp varied between soils, with the 180bp T-RF (22.1-54.4%) being most dominant in all treatments. The relative abundance of T-RFs 154 and 177 bp were higher in IM6, IM10, and IM15 than in CK and IM20 in both layer of soils, while T-RFs 68 and 332bp exhibited opposite trends. The four T-RFs of 75, 81, 147, 160, 180, and 187bp were most closely related to two groups − Rhizobium sp. (M16710.1) and Azorhizobium doebereinerae (FJ223129.1). The T-RF 47 bp closely matchedDesulfovibrio vulgaris (CP002298.1) (Fig.4b).
The relationship between soil properties and cbbL- andnifH-containing bacterial community
The difference in T-RFs profile of CO2- and N2-fixing bacteria between treatments was confirmed by ANOSIM (P < 0.01). For the former, the two-dimensional NMDS plot revealed that the treatments of all IM clustered closely together and separately from CK in the topsoils. (Fig.5a). However, the treatments were divided into three groups in subsoils, with CK and IM6 comprising the first group, IM10 the second (both groups located in the same side of NMSD1), and IM15 and IM20 the third group (located on the opposite side of NMSD1) (Fig.5b). In contrast, diazotrophic bacteria from the CK and IM20 treatments formed tight clusters separated from IM6, IM10 and IM15 (Fig.5c;5d).
Redundancy analyses (RDA) by Monte Carlo permutation test revealed that AP (P = 0.020), δ 13C (P = 0.044) and NH4+-N (P = 0.042) significantly explained the community shift of cbbL -containing bacteria in topsoils in response to IM duration (Fig.6a). In subsoils, only δ 13C (P =0.039) correlated well with community variations of this bacteria (Fig.6b). There was no correlation observed between soil pH and cbbL -containing bacterial community in either topsoils or subsoils. It is interesting thatδ 13C in both layers of soil was positively correlated with the treatment IM15 and IM20, suggesting increasingδ 13C may responsible the special composition ofcbbL -containing bacteria. The samples were divergent along the first two axis, which explains 59.7 % of the variation. ForcbbL -containing bacteria, the first two axes explained 79.5% and 84.0% of the total variation in topsoil and subsoils, respectively. As for nifH -containing bacterial community, they were significantly affected soil AK (P = 0.037), SOC (P = 0.024), C:N (P = 0.033), and AN (P =0.038) content in topsoils (Fig.6c), and with AP (P = 0.009) and C:N (P = 0.001) in the subsoils (Fig.6d). The first two axes of the RDA accounted for 59.7% of the variance of the diazotrophic community composition, with the first axes accounting for 56.4% of the variance.