3.1 Maize growth and distribution of 15N-labeled fertilizer in plant–soil system
Maize shoot dry weight was higher in the NH4+ and NO3− treatments than in CK, but did not differ significantly between the NH4+ and NO3− treatments (Fig. 1b). The root dry weight did not differ significantly among treatments. The N-recovery efficiency did not differ significantly between the NH4+ and NO3− treatments (Fig. 1c). The amount of soil residual fertilizer N showed a U-shaped pattern with increasing distance from roots in both the NH4+ and NO3− treatments, and was higher in the NH4+ treatment than in the NO3− treatments across the soil compartments from 0.5-cm to 4-cm from the root zone (Fig. 1d).
3.2 Soil chemical properties
Soil residual fertilizer N, soil pH, total N, NH4+-N, and NO3−-N contents were significantly affected by both the N form and the distance from the roots. Soil available K content was significantly affected by only distance from the roots and available P was significantly affected by only the N form (Table S1). The interaction between N form and distance from the roots (D × N) significantly affected soil residual fertilizer N, soil pH, NH4+-N, and NO3−-N, but did not affect total N, available K, and available P.
In the NO3− treatment, soil pH was the highest in the root zone and the 0–0.5 cm soil compartment, and then decreased to similar levels to that in CK with increasing distance from the roots (Fig. 2a). Soil pH was the lowest in the NH4+ treatment in all compartments of the rhizobox. Under both N treatments, the variations in soil total N were similar to those in soil residual fertilizer N, and there was no significant difference in soil total N between the NH4+ and NO3− treatments (Fig. 2b). The soil NH4+-N contents were very low in all compartments of the rhizobox in CK and the NO3− treatments, but increased dramatically with increasing distance from the roots in the NH4+ treatment (Fig. 2c). Soil NO3−-N contents were similar between the root zone and 0–4 cm soil compartments in the NH4+ and NO3− treatments, but were dramatically increased in the 4–9 cm compartment in the NO3− treatment (Fig. 2d). The soil NO3−-N contents were low in all compartments of the rhizobox in CK. In all treatments, the soil available K contents slightly decreased from the root zone to the 0–0.5 cm compartment and then increased with increasing distance from the roots (Fig. 2e). The soil available P contents were similar in all compartments in all treatments, and were generally higher in the NH4+ treatment than in the NO3− treatment (Fig. 2f).
3.3 Bacterial alpha and beta diversity
Compared with CK, both the NH4+ and NO3− treatments reduced the bacterial alpha diversity (Shannon’s index) in root zone soil, but not in other soil compartments (Fig. 3a). With increasing distance from the roots, the bacterial alpha diversity tended to increase in the NH4+ and NO3− treatments but not in CK. These results suggested that bacterial alpha diversity was inhibited by the combination of N fertilization and a rhizosphere effect.
The Bray–Curtis distance is used to indicate differences in microbial community structure between two treatments. The Bray–Curtis distances between CK and NH4+, CK and NO3−, and NH4+ and NO3− were significantly higher in the root zone and 0–0.5 cm compartment than in the other soil compartments (Fig. 3b). This indicated a more pronounced effect of the N form on the formation of bacterial community structures within 0.5 cm of the root zone. The Bray–Curtis distance between the root zone and other soil compartments increased with increasing distance from the root zone, and it was higher in the NH4+ and NO3− treatments than in CK (Fig. 3c). The rhizosphere effect was detected within 1 cm and 0.5 cm of the root zone under NH4+ and NO3− treatment, respectively (Table S2). There were no significant variations in bacterial communities among different soil compartments in CK, indicating that the rhizosphere effect on the soil bacterial community was stimulated by N fertilizer. Therefore, in the NMDS analysis, two distinct groups formed as a result of the N form and rhizosphere effect (Fig. 3d).
The bacterial community composition was also influenced by the N form and rhizosphere effect (Fig. 3e). Gemmatimonadales was the dominant bacterial order in all soil compartments in CK. Under NH4+ treatment, Burkholderiales became the dominant bacterial order in the root zone, and its abundance decreased with increasing distance from the roots. Under NO3− treatment, Sphingomonadales became the dominant bacterial order in the root zone, and its abundance also decreased with increasing distance from the roots.
3.4 Relationships between bacterial community structures and environmental variables
The results of Mantel tests indicated that pH was the major driver of bacterial community assembly (Table S3). The RDA analysis revealed that the bacterial communities in the root zone and the 0–0.5 cm soil compartment were positively correlated with soil pH under NO3− treatment, and those in the 2–9 cm soil compartments were negatively correlated with soil pH under NH4+ treatment (Fig. 4).
We identified the bacterial genera showing significant differences (Kruskal-Wallis test, p < 0.05) in abundance between CK and NH4+, CK and NO3−, and the NH4+ and NO3− treatments (Fig. 5). Compared with CK, the NH4+ treatment resulted in specific enrichment of four genera, Asticcacaulis, Acidibacter, Granulicella and Sinomonas within 0.5 cm of the maize root zone (Fig. 5a). The abundance of Acidibacter was negatively correlated with soil pH. In the NO3− treatment, four genera were specifically enriched within 0.5 cm of the maize root zone: Sphingomonas, Sphingobium, Pseudolabrys, and Azospirillum (Fig. 5b). The abundance of Sphingobium was positively correlated with soil pH. Comparing the NH4+ and NO3− treatments, the specifically enriched genera within 0.5 cm of the root zone under NH4+ treatment were Burkholderia, Mucilaginibacter, Acidibacter, Leifsonia, Catenulispora, and Asticcacaulis, and their abundance was negatively correlated with soil pH (Fig. 5c). The specifically enriched genera within 0.5 cm of the root zone under NO3− treatment were Sphingomonas, Sphingobium, Pseudolabrys, Azospirillum, and Novosphingobium, and their abundance was positively correlated with soil pH.
3.5 Bacterial co-occurrence patterns
Multiple network topological features consistently showed that bacterial co-occurrence pattern was greatly affected by the N form and distance from the roots (Table 1 and Table S4). Compared with CK, the NH4+ treatment resulted in higher values for edge density, complexity, average clustering coefficient (agvCC), and degree of centralization of bacterial association networks in all soil compartments, and the NO3− treatment resulted in lower values for all of these topological features (Table 1). This indicated that the NH4+ treatment resulted in the formation of more complex bacterial association networks, regardless of the distance from the maize roots. Under NH4+ treatment, the edge density, degree of centralization, average clustering coefficient, and complexity had similar values in the root zone, 0–0.5 cm, and 0.5–1 cm soil compartments, and the values of these topological features were lower in outer soil compartments (Table 1). Under NO3− treatment, the values of these topological features were similar between the root zone and its adjacent 0.5 cm soil compartment, and were lower in the compartments further from the roots. Under CK, the topological features showed no significant differences among the different soil compartments. This indicated that the bacterial association networks were more complex in the rhizosphere than in the bulk soil under both of the N-fertilization treatments, but not in CK.
Table 1
Co-occurrence topological features of bacterial networks in different soil compartments in CK and under NH4+ and NO3− treatments
Index | Soil compartment | Average |
Root zone | 0–0.5 cm | 0.5–1 cm | 1–2 cm | 2–4 cm | 4–9 cm |
| CK |
Nodes | 453a | 462a | 462a | 461a | 456a | 449a | 457 |
Edges | 5365a | 5576a | 5496a | 5438a | 5279a | 5157a | 5385 |
Edge density | 0.052a | 0.052a | 0.052a | 0.052a | 0.051a | 0.051a | 0.052 |
Degree centralization | 0.188a | 0.194a | 0.194a | 0.193a | 0.196a | 0.203a | 0.195 |
Betweenness centralization | 0.019a | 0.018a | 0.018a | 0.018a | 0.019a | 0.019a | 0.019 |
Average clustering coefficient | 0.349a | 0.349a | 0.347a | 0.348a | 0.343a | 0.342a | 0.346 |
Complexity | 11.82a | 12.01a | 11.90a | 11.78a | 11.53a | 11.47a | 11.76a |
Modularity | 0.338a | 0.346a | 0.339a | 0.330a | 0.336a | 0.314a | 0.334 |
| NH4+ |
Nodes | 402b | 432ab | 432ab | 439a | 432ab | 430ab | 428 |
Edges | 8850a | 9624a | 9496a | 9075a | 8649a | 8369a | 9011 |
Edge density | 0.110a | 0.104a | 0.102a | 0.094b | 0.093b | 0.091b | 0.099 |
Degree centralization | 0.362a | 0.361a | 0.364a | 0.356ab | 0.350b | 0.346b | 0.356 |
Betweenness centralization | 0.018a | 0.018a | 0.020a | 0.018a | 0.020a | 0.018a | 0.019 |
Average clustering coefficient | 0.437a | 0.428a | 0.425a | 0.413b | 0.420b | 0.409b | 0.422 |
Complexity | 21.97ab | 22.28a | 21.97ab | 20.68bc | 20.03bc | 19.45c | 21.06 |
Modularity | 0.224b | 0.223b | 0.239b | 0.252ab | 0.273a | 0.245ab | 0.243 |
| NO3− |
Nodes | 432c | 464a | 455ab | 447bc | 451ab | 446bc | 449 |
Edges | 4166ab | 4616a | 4121b | 3898b | 3913b | 3879b | 4099 |
Edge density | 0.045a | 0.043a | 0.040b | 0.039b | 0.039b | 0.039b | 0.041 |
Degree centralization | 0.175a | 0.160a | 0.138b | 0.131b | 0.128b | 0.126b | 0.143 |
Betweenness centralization | 0.022a | 0.018a | 0.019a | 0.021a | 0.020a | 0.021a | 0.020 |
Average clustering coefficient | 0.327a | 0.318a | 0.309b | 0.306b | 0.306b | 0.309b | 0.313 |
Complexity | 9.64ab | 9.94a | 9.04b | 8.72c | 8.67c | 8.70c | 9.12 |
Modularity | 0.349c | 0.368bc | 0.393a | 0.383ab | 0.385ab | 0.384ab | 0.377 |
Different lowercase letters in each row indicate significant differences among various soil compartments (p < 0.05). |
3.6 Nitrogen fertilizer-correlated OTUs and their co-occurrence patterns
According to the results described above, we defined the root zone and its adjacent 1-cm soil compartment as the maize rhizosphere in the NH4+ treatment, and the root zone and its adjacent 0.5-cm soil compartment as the maize rhizosphere in the NO3− treatment. Four new bacterial networks were constructed for the NH4+ rhizosphere, NH4+ bulk soil, NO3− rhizosphere, and NO3− bulk soil (Fig. S1). Compared with the NO3− treatment, the NH4+ treatment resulted in bacterial networks with more associations in both the rhizosphere and bulk soil (Fig. S1). To further explore the potential links between bacterial networks and soil residual fertilizer N in maize rhizosphere and bulk soil, we identified the “soil residual fertilizer N-correlated microbes” showing strong (r > 0.8, p < 0.01) correlations with soil residual fertilizer N in the networks (Table S5, Fig. 6). In the maize rhizosphere in the NH4+ treatment, soil N derived from NH4+ fertilizer was positively correlated with five bacterial OTUs (in Micrococcales, Burkholderiales, Sphingomonadales, and Rhodospirillales) and negatively correlated with 20 OTUs (mainly in Gemmatimonadales and Myxococcales) (Table S5). In the bulk soil, soil N derived from NH4+ fertilizer was positively correlated with one OTU (in Bacillales) and negatively correlated with four OTUs (in Sphingomonadales, Rhizobiales, and Legionellales). In the maize rhizosphere in the NO3− treatment, soil N derived from NO3− fertilizer was positively correlated with one OTU (in Bacillales) and negatively correlated with nine OTUs (mainly in Gemmatimonadales, Streptomycetales, Burkholderiales, and Solirubrobacterales). In the bulk soil, soil N derived from NO3− fertilizer was positively correlated with one OTU (in Enterobacteriales) and negatively correlated with two OTUs (in Rhizobiales and Gaiellales) (Table S5). These results indicated that N fertilizer, regardless of the N form, exerted more negative than positive effects on the growth of bacterial taxa. Bacterial networks in the maize rhizosphere were more sensitive than those in bulk soil, and were more sensitive to NH4+ than to NO3−.
The “N fertilizer-correlated microbes” showed different patterns of co-occurrence (Fig. 6). The OTUs that were positively correlated with soil N derived from N fertilizers had more negative edges than positive edges in the bacterial networks in the maize rhizosphere and bulk soil. In contrast, the OTUs that were negatively correlated with soil N derived from N fertilizers had more positive edges than negative edges in the bacterial networks in maize rhizosphere and bulk soil (Fig. 6a and 6c). In the NH4+ treatment, the rhizosphere OTUs that were positively correlated with soil N derived from NH4+ fertilizer showed 95 positive correlations with other microbes mainly in the Frankiales, Xanthomonadales, Burkholderiales, Micrococcales, and Rhodospirillales orders, and 130 negative correlations with other microbes mainly in the Gemmatimonadales and Myxococcales orders (Fig. 6a). In bulk soil in the NH4+ treatment, the OTUs that were positively connected with soil N derived from NH4+ fertilizer showed three positive correlations with OTUs (in the Burkholderiales, Rhizobiales and an unknown order), and three negative correlations with the OTUs (in the Sphingomonadales, Xanthomonadales, and an unknown order) in the bacterial network (Fig. 6b). In the NO3− treatment, the rhizosphere OTUs that were positively correlated with soil N derived from NO3− fertilizer showed seven negative correlations with other microbes in the Chlorobiales, Rhizobiales, Solirubrobacterales, Rhodospirillales, Gemmatimonadales, and Sphingobacteriales (Fig. 6c). In the bulk soil in the NO3− treatments, the 123 OTUs that were positively correlated with soil N derived from NO3− fertilizer showed only one negative correlation (with an uncultured bacteria) in the network (Fig. 6d).