Microbes play important roles in biogeochemical cycling of nitrogen (N), which is often a limiting nutrient for agricultural production. Fertilization shapes soil microbial diversity and community structure, while microbes in turn affect the fate of applied fertilizers [1]. Microorganisms have been recognized as the main drivers for nitrogen loss and soil N2O emission, which constitutes the dominant N2O source, emitting annually over 4 Tg N2O-N into the atmosphere [2]. Various microbial groups involved in N cycling have been studied (e.g. [3]), typically using different activity-based or marker gene-based approaches. Well-established primer systems and PCR assays are available for many marker genes of microbial N cycling, and have been extensively utilized in next-generation sequencing studies (e.g. [4]).

Oxygenic denitrification is a recently proposed N-transforming process, where nitric oxide (NO) is directly disproportionated into N2 and O2, thus avoiding the powerful greenhouse gas, N2O emissions. The reaction is catalyzed by a NO dismutase (Nod) [5]. The use of nod gene as a marker for oxygenic denitrifiers was recently established [6, 7], with which, evidence was obtained suggesting that oxygenic denitrifiers could be widespread and phylogenetically diverse [6, 8]. However, the high diversity inferred from nod genes provided no clue to the actual identities of these novel oxygenic denitrifiers [6]. So far, the identification of oxygenic denitrifiers and analysis of their community structures solely depend on the length and quality of obtained nod sequences. In previous studies, long nod sequences were obtained via laborious and low-throughput Sanger sequencing, since next-generation sequencing often generates short reads. Recently, PacBio SMRT sequencing provides access to high-throughput long-read environmental sequence data, but has been applied mostly to metagenomics and full-length 16S rRNA gene analysis (e.g. [9, 10]). Here, we provide a first proof-of-principle for the use of nod-targeted PacBio sequencing to study the distribution and community structure of oxygenic denitrifiers in terrestrial systems. Soils, especially under N fertilization, are known as hotspots of microbial N cycling; thus, agricultural soils undergoing different long-term fertilization regimes were chosen as example environments. To demonstrate the wide applicability of this approach, several sediment samples from lakes with different trophic status were also queried (Table 1).

Table 1 Soil samples and lake sediments analyzed in this study and their key geochemical characteristics
Full size table

Soil from the top layer (0–10 cm) of several agricultural fields with and without long-term fertilizations in Ning-Xiang, China [13], and sediments from a meso-oligotrophic, an eutrophic and a dystrophic lake in Germany [11, 12, 14], were sampled (Table 1). The abundance of nod genes and bacterial 16S rRNA was quantified by qPCR, representing respective measure of putative oxygenic denitrifiers and total bacteria in these samples. In all soils, the respective counts of oxygenic denitrifiers (1.12 × 107–1.54 × 107 copy nod g−1 soil) and total bacteria (from 9.75 × 108 to 1.37 × 109 copies 16S rRNA g−1 soil) were similar (Fig. 1b), resulting in a rather comparable relative abundance of nod genes at 1–1.3%. nod gene abundance was in a similar range as that of nirK and nirS genes observed in other soil ecosystems [15, 16], indicating that oxygenic denitrifier abundance can be similar to that of conventional denitrifiers. In contrast to conventional nitrifying and denitrifying microbes [16, 17], oxygenic denitrifier abundance examined here seemed not significantly influenced by nitrogen fertilization. While nod gene abundances varied more widely in lake sediments, from 2.17 × 106 g−1 sediment in Lake Grosse Fuchskuhle to 3.2 × 107 g−1 sediment in Lake Stechlin, accounting for about 0.4–3.0% of total bacterial number (Fig. 1b).

Fig. 1
figure 1

a Principal component analysis (PCA) of the oxygenic denitrifier community in all samples, based on the occurrence and relative abundance of all nod OTUs. b The abundance of total bacteria and oxygenic denitrifiers in different samples as quantified by qPCR targeting 16S rRNA and nod genes, respectively. cnod community in different environments. OTU was categorized based on a 90% similarity of ca. 1.0 kb nod gene fragments. The sample designations are the same as in Table 1

Full size image

With PacBio SMRT sequencing, long nod amplicons (ca. 1.0 kb) from all samples were sequenced. Details on the procedure and data analysis are provided in the supplementary information (SI). In total, 61 nod OTUs with at least 50 sequences were classified based on 90% similarity on nucleotide level [7]. Representative sequences were used for phylogenetic analysis, revealing that almost all OTUs detected in this study were related to previously reported nod lineages [6], with most of them belonging to the nod “Aquifer cluster,” 5 OTUs in the “NC10 cluster,” and 1 OTU in the “Reactor cluster 1” (Fig. 2). All representative sequences of these OTUs that related to nod clusters possessed previously identified Nod-characteristic residual substitutions (Fig. 3).

Fig. 2
figure 2

Bootstrapped neighbor-joining phylogeny of nod OTU representative nucleotide sequences. Bootstrap support (1000 replicates) greater than 50% is indicated at the nodes. Nod clusters (same as proposed in [6]) containing soil and lake nod OTUs are shown in bold. OTUs only detected in soil are shown in blue, lake-specific OTUs in red, and OTUs comprising sequences from both soil and lake sediment are shown in green. The scale bar represents 10% nucleotide sequence divergence. The phylogeny was calculated in MEGA-X

Full size image
Fig. 3
figure 3

Multiple-sequence alignment of Nod representatives (deduced from nod gene sequences) and qNor enzymes around the quinol-binding site and the catalytic site of qNor. The conserved residues for quinol binding and catalytic functioning in qNor are highlighted in orange, whereas substitutions at these sites in Nod and unknown Nor-related sequences are shown in green. Due to the length limit of the recovered Nod sequences, the functional sites shown here are incomplete. Signature residual mutations are observed in soil and lake Nod sequences as well. The residual numbering was according to the Geobacillus stearothermophilus qNor (AB450501). The alignment was generated with ClustalW in MEGA-X

Full size image

Overall nod gene pools appeared similar in the agriculture soils subjected to different fertilization regimes (Fig. 1a, c). However, the soil receiving both chemical fertilizer (CF) and manure harbored the highest nod diversity across all samples (Fig. 1c). Lake Stechlin and Lake Dagow sediments hosted a higher nod diversity compared with Lake Grosse Fuchskuhle (Fig. 1c), which was mostly dominated by OTU748. Possibly, this was related to their different pH and trophic status (Table 1). Principal component analysis (PCA) of nod composition clearly supported that distinct oxygenic denitrifier communities were present in agriculture soils and lake sediments (Fig. 1a). Only a limited number of OTUs were shared between soils and sediments, yet habitat-specific OTUs were evident in both soil and lakes, and they tended to be closely placed within the “Aquifer cluster” on the phylogenetic tree (Fig. 2), such as OUT74 was highly abundant in all soils but was not detected in sediments, suggesting that the environment largely determines the composition of nod gene pools.

Nevertheless, three OTUs, OTU462, 748, and 945, clustered more closely to the “unknown-qNor-related” sequences [6] in phylogenetic analysis (Fig. 2). Although OTU462 sequences also showed the presumed characteristic residual substitutions for Nod (Fig. 3), the functional connotation of this gene lineage must be interpreted with caution. This suggests that both phylogenetic and residual substitution analysis are necessary for identifying environmental nod sequences, a result clearly facilitated by the long sequences obtained from PacBio SMRT sequencing.

Taken all together, the results indicated that agriculture soils harbor diverse and abundant oxygenic denitrifiers, which have been overlooked in the past. Low pH in Lake Grosse Fuchskuhle seemed to disfavor oxygenic denitrifiers. Moreover, functional-gene-targeted long-read sequencing was proven to be a powerful tool for analyzing novel microbial guilds in the environment, and it will allow us to gain better insights into the influencing environmental factors that shape the distribution and community structure of oxygenic denitrifiers and uncover their ecological roles in N cycling in natural habitats.