Metagenomic Evidence for the Presence of Comammox Nitrospira-Like Bacteria in a Drinking Water System

Nitrification plays an important role in regulating the concentrations of inorganic nitrogen species in a range of environments, from drinking water and wastewater treatment plants to the oceans. Until recently, aerobic nitrification was considered to be a two-step process involving ammonia-oxidizing bacteria or archaea and nitrite-oxidizing bacteria. This process requires close cooperation between these two functional guilds for complete conversion of ammonia to nitrate, without the accumulation of nitrite or other intermediates, such as nitrous oxide, a potent greenhouse gas. The discovery of a single organism with the potential to oxidize both ammonia and nitrite adds a new dimension to the current understanding of aerobic nitrification, while presenting opportunities to rethink nitrogen management in engineered systems.


IMPORTANCE
Nitrification plays an important role in regulating the concentrations of inorganic nitrogen species in a range of environments, from drinking water and wastewater treatment plants to the oceans. Until recently, aerobic nitrification was considered to be a two-step process involving ammonia-oxidizing bacteria or archaea and nitrite-oxidizing bacteria. This process requires close cooperation between these two functional guilds for complete conversion of ammonia to nitrate, without the accumulation of nitrite or other intermediates, such as nitrous oxide, a potent greenhouse gas. The discovery of a single organism with the potential to oxidize both ammonia and nitrite adds a new dimension to the current understanding of aerobic nitrification, while presenting opportunities to rethink nitrogen management in engineered systems.
KEYWORDS: Nitrospira, comammox, drinking water systems U ntil recently, aerobic nitrification was considered to be a two-step process involving two functional guilds. The process of nitrification was considered split between ammonia-oxidizing bacteria (AOB) (1) and ammonia-oxidizing archaea (AOA) (2), which oxidize ammonia to nitrite, and strict nitrite-oxidizing bacteria (NOB) (3), which oxidize nitrite to nitrate. The phylogenetic distribution of AOB is limited to the Betaproteobac-teria and Gammaproteobacteria; AOA fall within the Thaumarchaea, and NOB span the Proteobacteria, Chloroflexi, Nitrospirae, and Nitrospinae. However, the recent discovery of complete ammonia oxidizer (comammox) organisms, i.e., bacteria that completely oxidize ammonia to nitrate (4), within the genus Nitrospira has significantly changed our understanding of the aerobic nitrification process (5,6). In the current study, we report metagenomic evidence of a Nitrospira-like organism that has the potential to perform both steps of the aerobic nitrification process and thus is likely to be a comammox bacterium. Specifically, it has genes to oxidize nitrite to nitrate (nitrite oxidoreductase) and possesses all genes required for ammonia oxidation, i.e., ammonia monooxygenase (amoA, amoB, amoC) and hydroxylamine dehydrogenase (also known as hydroxylamine oxidoreductase) (hao). This metagenome bin was discovered through shotgun DNA sequencing of samples from biologically active filters at a drinking water treatment plant (Ann Arbor, MI). IDBA-UD assembly (7), CONCOCT genome binning (8), and manual curation (see Text S1 in the supplemental material) resulted in 51 highquality draft genomes with 2 to 755 scaffolds (median ϭ 73), N 50 ranging from 7.4 to 114 kbp (median ϭ 86.9 kbp), and levels of completeness ranging from 77 to 100% (9). All genome bins were annotated by the Integrated Microbial Genomes Expert Review (IMG ER) system (10).
One of the genomes was phylogenetically assigned to the Nitrospira genus (JGI GOLD identification number Ga0074138) by Amphora2 (11). This 4.1-Mbp genome consists of 61 scaffolds (N 50 ϭ 150.74 kbp) with a GC content of 55%, 4,196 coding sequences, a complete 5S rRNA gene, and partial 16S and 23S rRNA genes. The genome bin was 88% complete, with 2.8% likely contamination based on 182 markers (9). The Amphora2 phylogenetic assignment was confirmed by Ribosomal Database Project (RDP) classification (12) of the 574-bp partial 16S rRNA gene in this genome (100% confidence). Maximum-likelihood phylogenetic analyses (13) of 16 syntenic ribosomal proteins (14) (Fig. 1A; see also Fig. S1 in the supplemental material) and the 16S rRNA gene (Fig. 1B) and Bayesian phylogenetic analyses (15) of the nxrA gene ( Fig. 1C; Text S1) indicated that this metagenome bin is related to Nitrospira lineage II bacteria. The metagenome bin contains two copies of nxrA, both of which have high levels of similarity to Nitrospira lineage II bacterial genes (Fig. 1C). We were unable to recover the nxrB gene in the metagenome bin and suspect that this was due to assembly issues, as both copies of nxrA were found at the ends of their respective scaffolds. Indeed, mapping of the reads extracted for reassembly indicated that 391 paired-end reads mapped to the nxrB genes of organisms in Nitrospira lineage II, confirming that this gene was present but not assembled. As with Nitrospira moscoviensis, the new metagenome bin also contained a full suite of genes for urea transport and its degradation to ammonia (16).
The newly described metagenome bin contained genes involved in ammonia oxidation (amoA, amoB, amoC, and hao) on a single 92.7-kbp scaffold (scaffold 158; scaffold range, 101 to 161) in close succession on a region beginning at 69 kbp (Fig. 1D). This observation is consistent with reports of recently described comammox bacteria that have radically altered our understanding of the nitrogen cycle (5, 6) and provides further indication of the metabolic versatility of Nitrospira bacteria (16)(17)(18). In addition to finding the amo and hao genes, we found several genes on scaffold 158 and a second scaffold (scaffold 123) within this metagenome bin with a gene arrangement similar to that of the recently published comammox bacteria. First, the hao gene cluster was preceded by several genes associated with heme export and cytochrome c biogenesis (ccmA to -H), a feature not seen in any published AOB genomes and thus suggested as a diagnostic feature for comammox Nitrospira (5) (Fig. 1D). Second, we found a gene that likely codes for membrane-bound protein OrfM based on its similarity to the corresponding gene in "Candidatus Nitrospira inopinata" (identity ϭ 69%, E value ϭ 5eϪ132, bit score ϭ 368) (5). However, IMG annotation and subsequent independent annotation efforts suggested that it was a hypothetical protein. Third, we found homologues of amoD (n ϭ 2) and amoE arranged in close succession on scaffold 123, genes that were originally annotated as encoding hypothetical proteins. Their levels of identity to the corresponding genes in "Ca. Nitrospira inopinata" were 66% (E value ϭ 1eϪ92) and 66.5% (E value ϭ 4eϪ97) for amoD and 65% (E value ϭ 2eϪ86) for amoE. These three genes were preceded by two genes annotated as encoding bacterioferritin and succeeded by the copper binding/resistance genes copC and copD, an arrangement consistent with that of the recently described comammox Nitrospira organism (5,6). The copD gene was followed by a partial amoC gene at the 5= end of scaffold 123, which is potentially contiguous to the partial amoC gene found at the 3= end of scaffold 158. Despite these highly conserved features between our Nitrospira metagenome bin and the three other genome bins described recently, it is important to note that our observation is purely metagenomic in nature. Further experimental analyses need to be performed to (i) estimate the abundance of comammox Nitrospira organisms in the sampled drinking water filters and (ii) confirm the expression of these genes and link them to ammonia/nitrite oxidation.
To assess whether the presence of ammonia oxidation genes in the Nitrospira-like metagenome bin was an artifact, we considered the potential for (i) misannotation, (ii) misassembly, and (iii) incorrect genome binning. First, IMG ER annotation indicated strong support for amoA, amoB, and hao against sequences in the KEGG, InterPro, TIGR, and Pfam databases. amoC was confirmed only against the Pfam database (E value ϭ 1.5eϪ25), potentially because it was a partial gene at the end of the scaffold. Second, we considered the likelihood of misassembly of scaffold 158 using coverage-per-base information ( Fig. S2A and S2B) and also checked the phylogenetic signal along the length of the scaffold (Text S1). Our analyses confirmed that the scaffold not only was properly assembled (there is strong support for this conclusion from properly mapped reads) but also had regions of similarity to Nitrospira-like bacteria across nearly the entire scaffold length, including ribosomal protein L31P found on this scaffold, whose best hit was Nitrospira sp. ENR4 (NCBI accession number CUQ65102.1, total score ϭ 132, query coverage ϭ 100%, E value ϭ 3eϪ38, percent identity ϭ 88%), an enrichment of "Ca. Nitrospira inopinata." Finally, we checked whether the scaffold was correctly binned into the Nitrospira-like metagenome bin. To do this, we compared the coverage per sample of scaffold 158 (Fig. S2C) and the k-mer frequency distribution (Fig. S2D) to those of other scaffolds in this metagenome bin. Both analyses suggested that this scaffold was not an outlier with respect to the other scaffolds in the bin. We also performed an Emergent Self-Organizing Map (ESOM) analysis (19) to test whether an independent binning approach would place the scaffold within this metagenome bin ( Fig. 2A). ESOM analysis identified three outlier scaffolds unrelated to ammonia oxidation ( Fig. 2B; Text S1), while the scaffold with ammonia oxidation genes binned with the Nitrospira-like metagenome bin (Fig. 2C). Based on these three lines of evidence (annotation, assembly, and binning), we conclude that the scaffold with ammonia oxidation genes belongs to the Nitrospira-like metagenome bin and thus suggests the presence of a comammox bacterium in the drinking water system of Ann Arbor, MI. The amoA gene of the newly reported Nitrospira-like metagenome bin clustered closely with the novel amoA gene of comammox Nitrospira bacteria. Phylogenetic analyses (15,   16) indicated that it branches from betaproteobacterial amoA but clusters closely with pmoA found in the metagenome of Crenothrix polyspora (Fig. 2C), a gammaproteobacterial methane oxidizer detected in a drinking water treatment plant in Germany (20), and "Ca. Nitrospira nitrificans" (6). It is interesting to note that one of the eight copies of 16S rRNA found in the Crenothrix polyspora metagenome on IMG belongs to a Nitrospira-like organism, suggesting that the gene annotated as pmoA might in fact be comammox amoA.
As part of this study, we also assembled a draft metagenome bin of Nitrosomonaslike AOB (GOLD identification number Ga0074132). However, due to the highly fragmented nature of this metagenome bin (598 scaffolds, N 50 ϭ 7.4 kbp), we were unable to recover all genes associated with nitrification. Nonetheless, its close affiliation with bacteria within the genus Nitrosomonas suggests that other canonical AOB were also present in these filters. To check for the presence of bacterial and archaeal amoA genes in the drinking water filter samples, we annotated the master assembly against a custom database of bacterial and archaeal amoA genes (Text S1). Only four significant amoA hits were detected in the master assembly, with two of these mapping to Betaproteobacteria-like amoA. The amoA gene within the Nitrospira metagenome bin was also detected in the master assembly, and we found an additional comammox-like amoA gene that was not present in any of the genomes assembled from this data set. Differences in coverage of the scaffolds on which these amoA genes were present across samples suggest that the four amoA genes (two comammox, two Betaproteobacteria) belong to four different organisms. The phylogenetic affiliations of all of these amoA genes and coverages per sample can be seen in Fig. S4.
The presence of the complete ammonia oxidation capacity in Nitrospira organisms has significant implications for the nitrogen cycle, particularly if this organism is widespread. Using amoA sequence matches in the NCBI database and their associated environmental ontologies, we find support for previous detection of this Nitrospira-like organism in engineered systems (n ϭ 8), soil ecosystems (n ϭ 19), and groundwater (n ϭ 10). This suggests that comammox Nitrospira organisms may be of importance for both engineered and natural systems. Interestingly, most of the matches in the NCBI database were attributed to methane-oxidizing bacteria, as also highlighted by Daims et al. (5). The recently published discovery of a comammox Nitrospira organism (5,6) in combination with our finding strongly suggests that microbial contributions to the nitrogen cycle in engineered and natural environments will need to be reevaluated. The presence of a comammox is also congruent with previous observations of abundances of Nitrospira-like bacteria that were significantly higher than those of AOB/AOA based on 16S rRNA gene assays (21)(22)(23), indicating that comammox activity likely contributed substantially to nitrate formation in these environments. While direct evidence of the conversion of ammonia to nitrate by a Nitrospira organism was provided (5, 6), it will be critical to build on this initial work to understand the extent to which comammox organisms contribute to ammonia and nitrite oxidation in the wide range of environments where nitrogen cycling is important. This might be particularly critical for oxidation genes that were binned with the Nitrospira metagenome. The ESOM binning procedure and contents of the three outlier scaffolds/scaffold fragments are presented in Text S1 in the supplemental material. (D) RAxML-based maximum-likelihood tree constructed using amino acid sequences of the amoA gene in the Nitrospira metagenome bin and pmoA/amoA sequences from a range of ammonia-oxidizing bacteria/archaea and methane-oxidizing bacteria, including the Nitrospira comammox. The tree was built from a trimmed muscle alignment using the Dayhoff model for protein evolution, gamma distribution model, and 500 bootstraps using the archaeal amoA gene of Nitrosopumilus maritimus as the outgroup. Branches are colored according to phylogenetic affiliation, and node support of >70% is indicated. This placement of the amoA gene from the Nitrospira-like genome and overall tree topology were also confirmed by neighbor-joining analysis (500 bootstraps) and the unweighted pair group method with arithmetic mean (UPGMA) (500 bootstraps) in Geneious and Bayesian phylogeny inference (20,000 generations) (Fig. S3).
wastewater treatment systems that rely on partial nitrification followed by anammox processes (24) or short-cut nitrification-denitrification (25) for reducing energy costs of nitrogen removal. Similarly, nitrification in biofilms, a predicted ecological niche for comammox bacteria (4), is a considerable concern in drinking water distribution systems (26), and strategies devised to inhibit AOB/AOA may or may not yield optimal results if comammox activity primarily drives ammonia oxidation. On the other hand, the benefits of comammox bacteria can be exploited for ammonia removal from drinking water sources through promoting their activity in biofiltration systems, such as the system from which the currently described Nitrospira-like metagenome bin was obtained.
Raw reads and all draft genomes are available through NCBI BioProject PRJNA301005. The draft genomes and annotation information can be accessed through IMG ER using JGI GOLD identification numbers Ga0074129-141 and Ga0077522-560. JGI GOLD, IMG-ID, and NCBI accession numbers for the Nitrospira metagenome bin discussed in this paper are Ga0074138, 2619618852, and LNDU00000000, respectively.