The physiological potential of anammox bacteria as revealed by their core genome structure

Abstract

We present here the second complete genome of anaerobic ammonium oxidation (anammox) bacterium, Candidatus (Ca.) Brocadia pituitae, along with those of a nitrite oxidizer and two incomplete denitrifiers from the anammox bacterial community (ABC) metagenome. Although ${\text{NO}}_2^{-}$ reduction to NO is considered to be the first step in anammox, Ca. B. pituitae lacks nitrite reductase genes (nirK and nirS) responsible for this reaction. Comparative genomics of Ca. B. pituitae with Ca. Kuenenia stuttgartiensis and six other anammox bacteria with nearly complete genomes revealed that their core genome structure contains 1,152 syntenic orthologues. But nitrite reductase genes were absent from the core, whereas two other Brocadia species possess nirK and these genes were horizontally acquired from multiple lineages. In contrast, at least five paralogous hydroxylamine oxidoreductase genes containing candidate ones (hao2 and hao3) encoding another nitrite reductase were observed in the core. Indeed, these two genes were also significantly expressed in Ca. B. pituitae as in other anammox bacteria. Because many nirS and nirK genes have been detected in the ABC metagenome, Ca. B. pituitae presumably utilises not only NO supplied by the ABC members but also NO and/or NH₂OH by self-production for anammox metabolism.

1. Introduction

Anaerobic ammonium oxidation (anammox) bacteria couple nitrite reduction to ammonium oxidation, with nitric oxide (NO) and hydrazine as intermediates, ultimately producing dinitrogen gas and nitrate.^1–3 NO, a key intermediate in anammox bacteria, is commonly produced by the reduction of nitrite to NO. This reaction is catalysed by two different types of enzymes—copper-containing (NirK) and cd₁ (NirS) nitrite reductases.⁴ However, anammox bacteria do not necessarily possess the genes encoding these enzymes. Interestingly, the nirK gene has been detected in the Candidatus (Ca.) Brocadia sp. UTAMX2⁵ and Ca. Brocadia caroliniensis⁶ genomes despite their low completeness; however, neither nirK nor nirS genes have been detected in other nearly complete Brocadia genomes. Ca. Brocadia sinica, which is one of the species with nearly complete genomes, lacks genes encoding NO-forming nitrite reductase (i.e. nirK and nirS).⁷ But, ¹⁵N-tracer experiments have demonstrated that Ca. B. sinica could reduce ${\text{NO}}_2^{-}$ to NH₂OH, instead of NO, with as yet unidentified nitrite reductase. On the other hand, anammox genomes contain 10 or 11 different types of paralogous genes encoding octaheme hydroxylamine oxidoreductase (Hao).^8,9 Hao-like proteins are thought to be the most likely candidate enzymes catalysing nitrite reduction to NO or NH₂OH in anammox bacteria. Indeed, among the Hao-like protein genes of Ca. Kuenenia stuttgartiensis,¹⁰KSMBR1_2163 (formerly kustc0458) and KSMBR1_3792 (formerly kuste4574) have been postulated to function in nitrite reduction based on their gene expression profile and protein sequence analyses, making them candidates for the elusive nitrite reductases producing NO.^8,11–13 Nevertheless, the physiological functions of Hao-like proteins remain enigmatic. Considering the published literature, a number of questions arise. First, what is the significance of the presence or absence of the nirS or nirK genes in anammox bacteria, and do any other members of an anammox bacterial community (ABC) have such genes? Second, do all anammox bacteria have the potential to reduce nitrite with Hao-like proteins? To answer these questions, we attempted to reconstruct the genomes of anammox and other predominant bacteria, from the ABC metagenome in the anammox bioreactor fed with nitrite and ammonium.¹⁴ We then performed comparative genomics on the anammox bacteria to determine their core genome structure, which was vertically inherited from a common ancestor, and examined the expression profiles of the key genes related to nitrogen metabolism. We also characterized the physiological and metabolic potential of not only anammox bacteria but also the predominant ABC members, especially for nitrite oxidizers, to determine their lifestyle in the anammox bioreactor.

2. Materials and methods

2.1. Anammox bacterial consortium

Anammox population was enriched in a small up-flow reactor, in which nonwoven fabric supports were placed in^15,16 and a synthetic medium was continuously supplied. A plastic cylinder (38-mm inner diameter, 400-mm length), of which bottom end was plugged with a rubber stopper, was set up in vertical position. Two sheets of porous polyester nonwoven fabric support (3-mm thick, 30 × 200 mm) were submerged to activated sludge, obtained from a municipal sewage treatment plant in central Japan, and placed in a plastic cylinder as a fixed-bed for the microbial population. Void volume of the reactor was 410 ml excluding the void volume of fabric supports, because supplied medium may not freely exchange inside and outside of support. The synthetic medium consisted of (per litre of tap water): NH₄Cl, 103–480 mg; NaNO₂, 104–570 mg; NaHCO₃, 500 mg; KH₂PO₄, 27 mg; MgSO₄·7H₂O, 300 mg; and CaCl₂·7H₂O, 18 mg. Dissolved oxygen in the medium was not eliminated and pH was controlled at 7.2–7.5. Concentrations of NH₄Cl and NaNO₂ were increased stepwise after anammox activity was detected. Samples were obtained from the medium reservoir and the effluent two to three times a week, and ${\text{NH}}_4^{+}$, ${\text{NO}}_2^{-}$, and ${\text{NO}}_3^{-}$ were determined using ion chromatograph (IC-2010, Tosoh Corp. Ltd., Tokyo, Japan).¹⁴ In the anammox enrichment culture, neither nitrite nor ammonium was consumed for 106 days after starting to supply the medium to the reactor, but consumption of nitrite and ammonium as well as production of nitrate was found on 107 days of operation, indicating that anammox organisms were growing in the reactor. Temperature of the reactor was maintained at 32.1 ± 1.0°C during 254 days of operation. Because the temperature unexpectedly raised and sometimes reached to around 35°C, the temperature was lowered to 30°C at the day 327, which was stably controlled until the 421 days, when the biomass was sampled. In addition to the measurement of anammox activity, we confirmed the 16S rDNA sequence of anammox bacteria amplified by PCR.

2.2. DNA and RNA isolation

The biomass of the anammox bacterial consortium was filtered through a 53-μm mesh to make it uniform (final wet weight, 3.6 g). DNA was extracted using the ISOIL for Bead-Beating kit (Nippon Gene) according to the manufacturer’s protocol, except for an additional treatment with a lytic enzyme.¹⁷ Briefly, a biomass sample was suspended in 30 ml of TE buffer and vortexed vigorously. The biomass suspension was evenly divided among 20 tubes containing zirconia/silica beads (0.1 mm, 0.47 g; 0.5 mm, 0.23 g). After the addition of lysozyme solution (0.71 mg/ml as a final concentration) to the tube, the suspension was incubated for 15 min at 37 °C with gentle shaking, and then incubated for 15 min at 37 °C after the addition of 20 µl of proteinase K solution (12 U). The lytic enzyme–treated biomass was bead-beaten by vortexing at power range 5 for 7 min. The solution was then centrifuged at 12,000 × g for 5 min at room temperature and the DNA in the supernatant was purified.

To concentrate DNA from Ca. Brocadia species used for genome reconstruction, we tested the beating conditions with 0.5-mm zirconia/silica beads alone by changing the beating power and time. The DNA concentration was evaluated based on the amount of PCR product using a primer set for the 16S rRNA gene (Bro5F, GGGTATGATCTTGGCTCAGAACGA; Bro232R, CGCACGTTTGACTATCATCACC) of Ca. B. pituitae. For RNA isolation, the biomass of the anammox bacterial consortium (2.5 g wet weight) was filtered as described above for DNA isolation and transferred to 50-ml tubes containing 5 ml TE buffer (pH 8.0). Ten millilitres of Bacterial RNA protect reagent (Qiagen) was added to the biomass suspension and vortexed vigorously. After the biomass suspension was allowed to stand for 5 min at room temperature, it was centrifuged at 5,000 × g for 10 min at 4°C to collect the biomass, and the pellet was washed with TE buffer. Lysozyme (1.36 mg/ml as a final concentration) was added to biomass suspended in 5 ml TE buffer. After a 10-min incubation at 37°C, 12 U of Proteinase K was added for another 10-min incubation. To increase the disruption efficiency, the lytic enzyme-treated biomass was transferred to a new tube containing 2.7 g zirconia silica beads (0.1 mm, 1.8 g; 0.5 mm, 0.9 g) and then vortexed at maximum power for 10 min. RNA was extracted and purified using the RNeasy Protect Bacteria Mini kit and QIAzol Lysis Reagent (Qiagen). DNase I-treated RNA was re-purified using the RNeasy Mini Clean-up Kit (Qiagen), and rRNA was removed using the Ribo-Zero rRNA Removal Kit (Gram-negative bacteria; Epicentre).

2.3. Sequencing and raw data treatment

Genomic DNA was sheared using a Focused ultrasonicator (Covaris). A paired-end library with an insert size of 450 bp was prepared using a TruSeq DNA PCR-free LT Sample Prep Kit (Illumina) and sequenced on an Illumina MiSeq platform, generating 6 million reads with read lengths of 300 bp. The PEAR software (v0.9.6) was used to merge these paired-end reads after filtering the low-quality scores of the sequence using the FASTX toolkit and eliminating duplicates using PRINSEQ. Gene prediction from the merged sequences was performed using MetaGeneAnnotator.¹⁸ Until the final creation of a multi-FASTA file of amino acid sequences including these multiple processes, it was pipelined as MAPLE Submission Data Maker (MSDM).¹⁹ MSDM is available from https://maple.jamstec.go.jp/maple/maple-2.3.1/softdownload/MSDM.html. The multi-FASTA file consisting of 3 million amino acid sequences was subjected to metagenomic analysis using Genomaple^TM (formerly MAPLE) ver. 2.3.2. RNA-seq libraries were prepared from the anammox bacterial consortium according to the standard Illumina protocol, and cDNA libraries were checked for quality and quantity using the DNA-100 kit (Agilent Technologies) and a 2100 Bioanalyzer. Each library was sequenced with the Illumina Sequencing Kit v2 on one lane of a MiSeq desktop sequencer (Illumina) to obtain 150-bp average paired-end reads. Reads Per Kilobase per Million mapped reads (RPKM) values were calculated according to the standard method. The RPKM ratio, calculated by dividing the RPKM of each gene by the mean RPKM of all ribosomal proteins, was used to determine relative gene expression levels. Amino acid sequences used for Genomaple^TM analysis are available at https://zenodo.org/record/3491404#.XigFVhP7RTY. BioSample accession numbers: SAMD00057694 for metagenomic sequence, AP021856 for Ca. B. pituitae, AP021857 for Ca. Desulfobacillus denitrificans, AP021858 for Ca. Nitrosymbiomonas proteolyticus, BLAA01000001- BLAA01000004 for 4 contigs of Ca. D. symbiosum, and DRA009157 for RNA sequence RNA-seq, of the ABC.

2.4. Community structure analysis

We calculated the proportional representation of bacteria in the metagenome based on the mapping pattern of module M00178 for bacterial ribosomes.^20,21 We calculated the proportion of bacteria at the individual taxonomic rank as defined by the Kyoto Encyclopaedia of Genes and Genomes (KEGG; i.e. phylum, order, and class level). For a more detailed analysis of bacterial communities, we searched all sequences assigned to ribosomal proteins by Genomaple^TM using the NCBI non-redundant protein database.

2.5. Reconstruction of genomes and data analyses

In addition to the Illumina platform, a PacBio RS II DNA sequencer was also used for metagenomic sequencing. Paired-end reads (400-bp insertion size on average) and mate-pair reads (3- and 5-kb insertion sizes on average) generated on a HiSeq 2500 (Illumina) were assembled using Platanus ver. 1.2.1²² to reconstruct the genome of predominant species in Ca. genus Brocadia. Reads from PacBio RS II were assembled using HGAP3. In the finishing step, a fosmid library was constructed according to the previously published method²³ and sequenced by Sanger and Illumina Technologies to fill gaps and increase sequence quality. In addition to Ca. Brocadia species, the genomes of three other predominant species in the phylum Armatimonadetes, class Betaproteobacteria (β-proteobacteria), and family Anaerolineae were reconstructed in the same manner. Gene identification and annotation were initially performed using DFAST.²⁴ We then manually curated the output results by carefully comparing the BLAST search results with the NCBI protein database as the genome database of DFAST is too small for annotating the genomes of uncultured microbes. A circular map of each reconstructed genome was constructed using the CGView Server.²⁵

2.6. Phylogenetic and genomic similarity analyses

To determine the phylogenetic position of Ca. Brocadia species whose genome was reconstructed in this study among anammox bacteria, we retrieved ribosomal proteins from each complete or draft genome registered in the NCBI database. Subsequently, we selected 20 commonly conserved ribosomal proteins (rplA, rplA, rplN, rplQ, rplV, rplW, rplX, rplC, rplE, rplF, rplI, rpsK, rpsM, rpsP, rpsS, rpsB, rpsD, rpsE, rpsH, and rpsI) among the anammox bacteria. These concatenated protein sequences were aligned with those of Rhodopirellula baltica, used as an outgroup. We used the LG+G model in MEGA 6.06 to construct a phylogenetic tree using the maximum likelihood method.²⁶ We tested the average nucleotide identity among the genomes used for phylogenetic analysis to identify species in the genus Ca. Brocadia using JSpeciesWS.²⁷ We also determined the phylogenetic positions of the nitrite reductases (NirK and NirS) horizontally acquired in the anammox bacterial genomes using the LG+G + I model in MEGA 6.06.

2.7. Orthologous analysis and estimation of the core genome structure

Orthologous groups (OGs) for Ca. Brocadia species, Ca. K. stuttgartiensis, and six other anammox bacteria were constructed using the rapid classification programme DomClust²⁸ on the Microbial Genome Database (MBGD) server.²⁹ A core genome is defined as a set of genes (OGs) syntenically conserved in at least half of the compared strains. In our study, a set of genes in the syntenic regions shared by at least five species was defined as the core genome for the eight analysed anammox species. A set of syntenic regions and the consensus order of the OGs in these regions, designated as the core genome structure, were created using the CoreAligner programme³⁰ based on conserved linkages between orthologous genes in each chromosome. Comparative analysis of gene organization around the nirK gene was performed using GenomeMatcher.³¹

2.8. Evaluation of the metabolic and physiological potential

The patterns of metabolic and physiological potential of Ca. Brocadia species and three other predominant species were investigated using Genomaple^TM (formerly MAPLE) ver. 3.2.^21,32 Genomaple^TM is available through a web interface (https://maple.jamstec.go.jp/maple/maple-2.3.1/) and as a stand-alone package from Docker Hub (https://hub.docker.com/r/genomaple/genomaple). Genes were mapped to 795 functional modules defined by the KEGG (pathways, 305; complexes, 294; functional sets, 157; and signatures, 40), and the module completion ratio (MCR) was calculated according to a Boolean algebra–like equation described previously.²⁰ To evaluate MCR, Q-value suggesting working probability of the modules was also calculated by Genomaple^TM. The Q-value near zero indicates high working probability of the module.²¹ The MCR and Q-value patterns of Ca. Brocadia species were compared with those of three other anammox species with genomes reconstructed to fewer than five contigs.

2.9. Determination of anammox activity using a ¹⁵N-tracer

The potential anammox and denitrification activities of biomass samples were determined using a ¹⁵N-tracer technique based on a method described previously.³³ To determine the activities of the samples, reactive substrates for anammox were added to the vials under the following three combinations: (i) 1 mM unlabelled NH₄Cl and 1 mM Na¹⁵NO₂ or 1 mM ¹⁵NH₂OH (SI Science, Tokyo, Japan; 99.9 atom% ¹⁵N), (ii) 1 mM ¹⁵NH₄Cl (SI Science; 99.9 atom% ¹⁵N) and 1 mM unlabelled NaNO₂ or NH₂OH, and (iii) 0.4 mM ¹⁵NH₄Cl without nitrite. During anaerobic incubation with substrate combinations (i) and (ii), ${\text{NO}}_2^{-}$- or NH₂OH-dependent anammox could be detected by the production of ¹⁴N¹⁵N (²⁹N₂), because anammox uses 1 mol each of ${\text{NH}}_4^{+}$ and ${\text{NO}}_2^{-}$ or NH₂OH. Substrate combination (iii) was a negative control to determine whether anammox occurs without nitrite or NH₂OH. The biomass granule of the anammox bacterial consortium, which was sieved through a 53-μm mesh, was used for measurement of anammox activity. When NO was used as a reactive substrate instead of ${\text{NO}}_2^{-}$, we added 2500 ppm of ¹⁵NO gas (SI Science; 99.6 atom% ¹⁵N, +98%) to the headspace of the vial containing the ABC biomass suspended in 5-ml substrate solution. The concentration of dissolved NO was ∼7 µM. The following standard gas was used for GC/MS analysis. A small amount of ¹⁵N-labelled N₂ gas (53.4 ¹⁵N atom%; SI science, Co. Ltd., Tokyo, Japan) was added in a vial filled with ultrapure He (>99.99995%) as background to prepare standard to demine ²⁹N², and ³⁰N₂. Concentration of ²⁹N₂ and ³⁰N₂ in the preparation was, 386 and 221 ppmv, respectively. Other GC/MS analytical conditions and subsequent calculations were performed as described previously,^14,33 except that the ABC biomass was not sieved. After the GC/MS measurement, the biomass was recovered by filtration and dried at 100 °C for 12 h to calculate the dry weight. To disperse the ABC granules, the sample was agitated using an SW-M6000 stirrer at 130 rpm before measurement of anammox activity.

3. Results and discussion

3.1. Metagenomic analysis of anammox bioreactor

The ABC comprised members of four major taxa, Planctomycetes (48.5%), Chloroflexi (14.3%), Armatimonadetes (10.8%), and β-proteobacteria (12.6%); Ca. Brocadia was the most predominant member in Planctomycetes, with a relative abundance of 38.1% (Fig. 1a). Accordingly, we performed genome reconstruction of Ca. Brocadia species and representative species of three other major taxa using sequenced fosmid clones and shotgun reads. We successfully reconstructed the genomes and determined the whole-genome sequences of three currently non-isolatable species in genus Ca. Brocadia, phylum Armatimonadetes, and class β-proteobacteria. Similarly, we were able to assemble four contigs of Anaerolineaceae species in phylum Chloroflexi. We designated this anammox bacteria as Ca. Brocadia pituitae (‘pituitae’ is ‘sludge’ in Latin). Ca. B. pituitae was shown to be a new species by phylogenetic analysis based on the sequences of concatenated ribosomal proteins and genomic identity analysis among 12 Brocadia strains including Ca. B. pituitae (Supplementary Fig. S1 and Supplementary Table S1). We also designated 3 other predominant species as Ca. N. proteolyticus (Armatimonadetes), Ca. D. denitrificans (β-proteobacteria), and Ca. D. symbiosum (Anaerolineaceae) based on their genomic features and metabolic and physiological potential deduced by the Genomaple. The description of Ca. N. symbiomonas is as follows: Nitrosymbiomonas (ni.tro.sym.bi.o’mo.nas. L. n. nitrum nitrate; Gr. n. symbios a companion; Gr. n. monas a unit, monad; M.L. fem. n. Nitrosymbiomonas, nitrate producing symbiotic monad.), proteolyticus (pro.te.o.ly’ti.cus. Ger. protein from Gr. protos first; adj. lyticus dissolving; M.L. adj. proteolyticus protein-dissolving). We also designated β-proteobacteria and Anaerolineae phylotypes as Ca. D. denitrificans and Ca. D. symbiosum. The description of Ca. D. denitrificans and Ca. D. symbiosum are as follows: Desulfobacillus (de.sul.fo.ba.cil’lus. L. pref. de from; L. n. sulphur; L. dim. n. bacillus, a small rod; M.L. masc. Desulfobacillus a bacillus that reduces sulphur compound), denitrificans (de.ni.tri’fi.cans. L. prep. de away from; L. n. nitrum soda; M.L., nitrum nitrate; M.L. v. denitrifico. denitrify; and M.L. part. adj. denitrificans denitrifying). Denitrolinea (de.ni.tro.li.ne’a. L. pref. de from; L. n. nitrum nitrate; L. fem. n. linea line; N.L. fem. n. Denitrolinea line-shaped nitrate reducing), symbiosum (sym.bi.o’sum. Gr. n. symbios a companion; N.L. neut. adj. symbiosum living together with, symbiotic).

Figure 1.

Metagenomic analysis of anammox reactor. (a) Community structure analysis based on ribosomal proteins identified in the metagenome. The genomes of unisolable Ca. B. pituitae, Ca. N. proteolyticus, and Ca. D. denitrificans were reconstructed in this study, although the genome of Ca. D. symbiosum is still divided into four contigs. Taxa representing <2% of the total were classified as ‘Other’. (b) Schematic representation of transformation steps of nitrogen compounds and the genes associated with each reaction step in the ABC. Dashed line indicates the predicted reaction.

3.2. Genomic features of predominant ABC members

The genome of Ca. B. pituitae consists of a single circular chromosome (4,075,302 bp) with mean G + C content of 43.4% (Table 1 and Supplementary Fig. S2). We identified 3,593 protein coding sequences (CDSs), 47 tRNA genes and 3 rRNA genes. Out of the 3,593 CDSs in the Ca. B. pituitae genome, 1,138 were included in the core genome structure, comprising 1,152 OGs (Supplementary Fig. S3 and Table S2). The genomes of two other predominant species, Ca. N. proteolyticus and Ca. D. denitrificans, are also single circular chromosomes consisting of 2,809,316 and 3,145,360 bp with a higher G + C content of 61.1% and 66.7%, respectively. Although the genome of Ca. D. symbiosum is still divided into 4 contigs, this genome is thought to be nearly completed because 50 ribosomal protein genes were identified in the 4 contigs. In general, a bacterial ribosome is composed of 52 ribosomal proteins with 1 or 2 of them missing depending on the bacterial strain. Total length of the contigs is 3,705,798 bp and their mean G + C content is also as high (59.5%) as the genomes from two other predominant species (Table 1).

Table 1.

General features of the reconstructed genomes of four major members in the anammox reactor

General features	Ca. B. pituitae	Ca. N. proteolyticus	Ca. D. denitrificans	Ca. D. symbiosum
Size (bp)	4,075,302	2,809,316	3,145,360	3,705,798
contig	1	1	1	4
G + C (%)	43.4	61.1	66.7	59.5
protein coding genes	3,593	2,561	3,104	3,442
function assigned	2,015	1,626	2,378	1,811
conserved hypothetical	1,418	869	684	1,520
hypothetical	160	66	42	111
rRNA	3	3	4	3
tRNA	47	47	51	49
ribosomal protein (KO-assigned)	51	54	53	50

Genomic analysis of the reconstructed genomes revealed that Ca. B. pituitae possesses no nitrite reductase genes, such as nirS and nirK, and these genes were not included in the core structure of the anammox bacterial genome. In the anammox bioreactor, however, those genes were complemented by those from other ABC members (i.e. NO producers from nitrite) in the bioreactor. Of all nitrite reductases detected in the ABC bioreactor, the nirS genes derived from Ca. D. denitrificans and Ca. D. symbiosum occupied 25% and 7%, respectively (Fig. 1b). These two ABC members also possessed nitrate reductase genes (narGHI and/or napAB) but not NO reductase genes (norBC), indicating that they are imperfect denitrifiers, although norBC genes are present in other ABC members. Indeed, when a small portion of biomass granules was anaerobically incubated with ammonium and nitrite as reactive substrates, 84% of the total nitrogen emission was due to the anammox reaction, and the remaining 16% was due to denitrification. Thus, it is likely that ABC members capable of producing NO were supplying it to Ca. B. pituitae, which is able to use nitrite. Indeed, in the case of NO, 69% of the total nitrogen emission was derived from the anammox reaction (Supplementary Table S3). Expression of the nirS genes was confirmed by metatranscriptomic analysis, which showed that the gene expression level from Ca. D. denitrificans and Ca. D. symbiosum was 5- to 10-fold lower than the average level for ribosomal proteins (Supplementary Table S4). Thus, although predominant ABC members including imperfect denitrifiers such as Ca. D. denitrificans and Ca. D. symbiosum have the ability to produce and supply NO to Ca. B. pituitae, Ca. B. pituitae does not necessarily depend solely on other ABC members for NO production. If an anaerobic incubation experiment, in which ${\text{NH}}_4^{+}$is not added but only ¹⁵NO₂⁻ is added to anammox population, is performed, it is expected that ¹⁵${\text{NO}}_2^{-}$ reduction (to NO) rate would be similar to anammox rate, when anammox bacteria depend on NO supplied by cross feeding. In this experimental system, because anammox reaction consuming NO is suppressed, NO would be accumulated in the incubation vial and it has been reported that NO could enhance ${\text{NO}}_2^{-}$ reduction activity.^34,35 Thus, the fair comparison of reduction rate between only ¹⁵${\text{NO}}_2^{-}$ and ¹⁵${\text{NO}}_2^{-}$ + ${\text{NH}}_4^{+}$ is thought to be difficult.

On the other hand, the genome of Ca. K. stuttgartiensis possesses the nirS gene, but this gene is hardly expressed at the transcriptional and protein levels compared with those encoding other key catabolic genes.^12,13 In contrast, nirS is one of the highest expressed genes in the marine anammox species, Ca. Scalindua brodae.³⁶ On the other hand, Ca. Jettenia caeni KSU1, as well as Ca. B. fulgida and Ca. B. caroliniensis, possess nirK instead of nirS, although other Brocadia species possess no such nitrite reductase (Supplementary Fig. S1). Ca. B. pituitae does not possess these genes although this species utilises nitrite, NO, and NH₂OH as the substrate for anammox metabolism.

Ca. N. proteolyticus was shown to be a nitrite oxidizer through the detection of nxrAB genes encoding nitrite oxidoreductase. Ca. N. proteolyticus possesses 78 genes encoding peptidases and 27 of them were predicted to be extracellular enzymes by the SOSUI program (http://harrier.nagahama-i-bio.ac.jp/sosui/). Although the expression level varied depending on the individual gene, almost all peptidase genes were expressed (Supplementary Table S5). Such a large number of peptidase genes were not observed in other reconstructed genomes. In addition, this species possesses 32 genes encoding cell wall-associated enzymes such as glycosyl transferase, lytic transglycosidase, and cell wall-associated hydrolase assisting in adherence, autoaggregation, turnover of cell wall and autolysis.^37–39 Out of those genes, 22 genes containing 6 secretory soluble enzyme genes were expressed with various expression levels (Supplementary Table S6). Since Ca. N. proteolyticus also possesses multiple expressed genes related to Type II secretory pathway (Supplementary Table S7), many of the peptidase and cell wall-associated enzymes are presumably secreted via Type II secretory pathway.⁴⁰ It is not yet clear how ABC members are selected and maintained, but it seems that the members of anammox bioreactors using non-woven fabric as a carrier share some common features. In fact, because 16S rRNA genes with more than 99% identity to that of Ca. N. proteolyticus have been detected in many other anammox bioreactors,^41,42 this microbe was inevitably suggested to be an ABC member responsible for nitrite oxidation via consumption of O₂ in the anammox bioreactor. Ca. N. proteolyticus is one of nitrite oxidizing bacteria (NOBs), but it has no potential for carbon fixation unlike the other NOBs (Supplementary Table S10) although the bioreactor is maintained through feeding with a synthetic inorganic medium containing sodium bicarbonate as the sole carbon source. Given such operating conditions, the immediate question of how Ca. N. proteolyticus cells acquire nutrients and are keeping their population in the bioreactor arises.

Since Ca. N. proteolyticus possesses multiple secretory peptidases and lytic transglycosidases (which are expected to be attractive new targets for the development of broad-spectrum antibiotics³⁸) and also Type II secretion systems,⁴⁰ it is possible that proteolysis of biomass from autolysed old cells and also the lysis of active ABC members sensitive to these enzymes may supply nutrients to Ca. N. proteolyticus itself and to other heterotrophic predominant species such as Ca. D. denitrificans and Ca. D. symbiosum, for maintaining their populations. In fact, ABC members in the anammox bioreactor repeat cell metabolism, but cannot maintain an active cell population in the reactor unless biomass from old cells is properly reused. In addition, keeping the population balance of ABC members is considered to be an important factor in maintaining stable anammox activity. In that sense, Ca. N. proteolyticus possessing high proteolysis potential must be one of the important key species necessary for keeping balanced population of ABC members and the stable anammox activity in the bioreactor.

On the other hand, imperfect denitrifiers, Ca. D. denitrificans and Ca. D. symbiosum, commonly appeared even in the ABC of anaerobically controlled anammox bioreactor. In fact, the genes with 98–99% identities to Ca. D. denitrificans (bin ID: PRO2, β-proteobacteria) and Ca. D. symbiosum (bin ID: CFX3, Chloroflexi) were detected in the metagenomic sequences from this bioreactor although their abundance is very low (1%>).⁵ Therefore, the involvement of these common ABC members in anammox metabolism may be related to the difficulty of isolating anammox bacteria.

3.3. Placement of nitrite reductase gene on the anammox bacterial genome and its phylogeny

As mentioned earlier, anammox bacteria do not necessarily possess NirK and NirS-type nitrite reductase genes but where and how their genes are located on the anammox genome is still a question. In fact, Ca. Brocadia sp. UTAMX2, (identified as Ca. B. fulgida based on the results of phylogenetic and genomic identity analyses; Supplementary Fig. S2 and Table S1), and J. caeni KSU1, possess the nirK gene unlike Ca. B. pituitae. In contrast, Ca. K. stuttgartiensis possesses the nirS gene, and Ca. Scalindua rubra BSI-1 possesses both genes. A comparison of the genome structure near nirK with those of other anammox bacteria that lack the gene revealed that the nirK in the KSU1 genome is positioned between the core gene set together with an insertion sequence (IS) belonging to the IS630 family⁴³ (Fig. 2). In addition, the nirK gene was incidentally inserted at the same position of the UTAMX2 genome, even though there was no IS element nearby. This mobile gene-like behaviour makes it clear that the nirK gene was acquired by horizontal gene transfer. Different types of ISs have also been found in this region of Ca. K. stuttgartiensis and of Ca. B. sinica,⁴⁴ which lacks the nirK gene. Indeed, Ca. K. stuttgartiensis and Ca. B. pituitae genomes contain 74 and 68 transposase genes for ISs, respectively, belonging to IS families IS630, IS4, and ISL3, which are widely disseminated throughout various bacterial genomes.^43,45 These foreign nirK genes exhibit 67% identity with each other, but their ancestral host organism is unknown. The nirK gene has also been detected in Ca. Scalindua draft genomes with low completeness, but whether it was acquired by the same route as the other two genes remains unknown.

Figure 2.

Comparison of gene organization around the nirK gene among anammox bacteria. Gene organization was compared among six anammox bacteria with and without the nirK gene. Genomic fragments containing nirK or the corresponding region were extracted from four nearly completed draft genomes and two completed genomes, and then aligned. Purple arrows show the core gene set among anammox bacteria, as shown in Supplementary Fig. S3. K. stuttg: Kuenenia stuttgartiensis, B: Brocadia, J: Jettenia.

To identify the ancestral host organism, we constructed an unrooted phylogenetic tree based on the amino acid sequences of NirK protein. Ca. Brocadia^5,6 and Ca. Jettenia^46,47 species formed a cluster together with Methylomarinum vadi in γ-proteobacteria, and two Ca. Scalindua species independently formed their own cluster, although all anammox bacteria were classified into the family Brocadiaceae (Fig. 3a). Proteobacteria formed a large cluster comprising four subdivisions (α, β, γ, and δ), but some of the species in β- and γ-proteobacteria formed another small cluster with other taxa. Given the complicated phylogenetic relationships among NirK protein clusters, except for the major Proteobacteria cluster, it appears that anammox bacteria did not necessarily acquire the nirK gene from monophyly. In addition, the nirS genes of Ca. Kuenenia and Ca. Scalindua species did not form a cluster, but Ca. Kuenenia formed a cluster with subsets of β-proteobacteria and δ-proteobacteria, whereas Ca. Scalindua formed a cluster with the Parcubacteria within the unclassified bacteria group (Fig. 3b). Because the nirS gene, like the nirK gene, was also not horizontally acquired from monophyly (Fig. 3b), it was not possible to determine the original taxonomic affiliation of the nitrite reductase gene acquired in the anammox bacteria genome.

Figure 3.

Phylogenetic position of NirK and NirS proteins identified in anammox bacteria. (a) A maximum likelihood tree based on NirK proteins. (b) A maximum likelihood tree based on NirS proteins. The numbers indicate the percentages of bootstrap support. Bootstrap values <50% were omitted from this figure.

3.4. Metabolic and physiological potential harboured in the reconstructed genomes

To examine metabolic and physiological potential harboured in the anammox bacterial genomes, we applied Ca. J. caeni and Ca. B. sinica genomes reconstructed to less than five contigs together with the two complete genomes (Ca. B. pituitae and Ca. K. stuttgartiensis) to Genomaple^TM and compared the patterns of MCR and Q-value among the four anammox bacteria (Supplementary Table S9). We found that their pattern is quite similar among the four species through all KEGG modules. In the pathway modules, when completed modules or modules with a Q-value of < 0.5 (i.e. suggesting high working probability of the module) were picked up in each species, only Brocadia species had the module for undecaprenyl phosphate alpha-L-Ara4N biosynthesis (M00761) involved in amino and nucleotide sugar metabolism despite the fact that this is γ-proteobacteria-specific rare module (Supplementary Fig. S4). In contrast, Euryarchaeota/Planctomyces-specific rare module for haem biosynthesis (M00847) was incomplete only in Ca. B. sinica. Since there was no noticeable difference regarding the other pathway modules, we inferred that there is no difference in the basic anabolic potential of the four major anammox bacteria.

In general, NOB have an ability to fix carbon through Calvin cycle,⁴⁸ but Ca. N. proteolyticus has only enzymes involved in 5 reaction steps of the module for Calvin cycle (M00165), which comprises 11 reaction steps (i.e. MCR: 54.5%). On the other hand, a complete ammonia oxidizer (Comammox) can use reductive citrate cycle for carbon fixation,⁴⁹ but Ca. N. proteolyticus has no ATP citrate lyase and citryl-CoA synthetase, which are key enzymes for this cycle. So far, seven carbon fixation pathways including these two are known, but Ca. N. proteolyticus does not have any of them and thus it seems to lack carbon fixation ability unlike the other NOBs (Supplementary Table S10). On the other hand, one of the incomplete denitrifiers, Ca. D. denitrificans, was found to complete the module for dissimilatory sulphur reduction (M00596) unlike Ca. D. symbiosum.

3.5. Orthologous analysis of Hao-like protein genes in anammox bacteria

Hao catalyses oxidation of NH₂OH to NO₂⁻ or NO in the nitrification process of ammonia oxidizing bacteria (AOB).⁹ On the other hand, Hao-like proteins are predicted to be the most likely candidate enzymes catalysing ${\text{NO}}_2^{-}$ reduction to NO in Ca. K. stuttgartiensis^10,12 and Ca. Scalindua profunda⁴⁰ and to NH₂OH in Ca. B. sinica.⁷Ca. B. pituitae possesses nine paralogous genes encoding Hao-like proteins, similar to other anammox bacteria. Thus, we first classified hao- and hao-like genes identified in typical AOB and all available anammox bacterial genomes through clustering analysis. We designated them as hao1–hao10 and one of the large independent clusters, hao5, was divided into two groups, hao5A and hao5B and the hao8 cluster constructed a large one with the AOB hao cluster (Fig. 4). As a result, Ca. B. pituitae was found to possess all genes categorized into hao1–hao 9 except for hao10, which is specific to Ca. Scalindua. Among those genes, at least 5 (hao1–hao4 and hao5B) were found in the core genome whereas Ca. B. sinica is missing hao5B gene (Fig. 5a and b). The genes encoding hydrazine synthase subunit A, B, and C (hzsABC) and hydrazine oxidoreductase (hzo), which are responsible for the main anammox reaction (Fig. 1), and nitrite oxidase (nxrAB) are also included in the core genome (Fig. 5a and Supplementary Table S10). Ca. B. fulgida UTMX2 and Ca. J. caeni KSU1 possessed the same hao-like gene repertoire as Ca. B. pituitae, but Ca. K. stuttgartiensis were missing hao8 and hao9 genes.

Figure 4.

Phylogenetic tree of Hao-like proteins. KSU1 C0486, which was identified as Hao5B by MBGD 27, is not shown in this figure because of its short partial sequence. Hao proteins from AOB were also included in this figure as the reference proteins. BPIT: Ca. B. pituitae, BROSI: Ca. B. sinica JPN1, EX330: Ca. Brocadia sp. BROELEC01, B6D34; Ca. Brocadia sp. UTAMX1, B6D35: Ca. Brocadia sp. UTAMX2, KSU1: Ca. J. caeni KSU1, KSMBR1: Ca. K. stuttgartiensis MBR1, SCALIN: Ca. Scalindua japonica husup-a2, Noc: Nitrosococcus oceani ATCC19707, NE: Nitrosomonas europaea ATCC 19718, Nmul: Nitrosospira multiformis ATCC 25196. Amino acid sequences of hao-like gene products were aligned by MUSCLE. Phylorgenetic relationships were inferred by Maximum Likelihood method with LG+G + I model using MEGA6.06 24.

Figure 5.

Phylogenetic relationships and core gene set among anammox bacteria. (a) Phylogenetic tree based on amino acid sequences of the concatenated ribosomal proteins of anammox bacteria and retention patterns of hao-like and nitrite reductase nirS and nirK, genes. Two complete genomes Ca. B. pituitae and Ca. K. stuttgartiensis, and six nearly completed draft genomes were used for this study. (b) Core gene alignment of anammox bacteria. Rows and columns represent genomes and OGs, respectively. Black lines represent direct adjacency, green lines represent non-adjacent neighbourhoods indicating the existence of insertions, and red lines represent inversions. Closed circles and squares indicate single and paralogous genes, respectively. Dashed lines show the connection to the divided gene. Red characters show genes associated with anammox reactions. ANI: average nucleotide identity. The ANI values show identity % of Ca. B. pituitae genome to other anammox bacterial with nearly complete genome.

The expression of KSMBR1_2670 gene (formerly kustc1061) corresponding to hao1 (Fig. 5), and its gene products have been experimentally confirmed as hydroxylamine oxidase in Ca. K. stuttgartiensis. In addition, the expression of KSMBR1_2163 (formerly kustc0458) and KSMBR1_3792 (formerly kustc4574) genes corresponding to hao2 and hao3, respectively, has also been confirmed along with their gene products.⁸ Based on protein sequence analyses, it has been postulated that the KSMBR1_2163 (hao2) and KSMBR1_3792 (hao3) genes encode the elusive nitrite reductases reducing nitrite to NO in Ca. K. stuttgartiensis. However, Ca. B. sinica JPN1 as well as Ca. B. pituitae lack the genes encoding canonical NO-producing nitrite reductases (i.e. nirS and nirK; Fig. 5b). Reduction of nitrite to NH₂OH and not NO was confirmed by ¹⁵N-tracer experiments in Ca. B. sinica although the enzyme responsible for this reaction is still unidentified. It was also confirmed that this species utilises NH₂OH and ${\text{NH}}_4^{+}$, but not NO and NH₄⁺ for N₂H₄ synthesis, which is downstream of nitrite reduction, demonstrating that the anammox metabolism of Ca. B. sinica is NH₂OH-dependent.⁷ We also confirmed that Ca. B. pituitae can utilise NH₂OH and ${\text{NH}}_4^{+}$ for anammox metabolism similar to Ca. B. sinica (Supplementary Fig. S5). Oshiki et al.⁷ speculated that Hao-like proteins are the most likely candidate enzymes catalysing nitrite reduction to NH₂OH because ${\text{NH}}_4^{+}$-forming pentahaem cytochrome c nitrite reductase has evolved to octahaem cytochrome c protein (Hao)^50–52 and this Hao protein is capable of reduction of nitrite to NH₂OH using electrons shuffled from quinone pools by a membrane-anchored cytochrome c protein that appeared during the evolutionary process.^53,54 High expression of genes corresponding to hao1–hao3 and hao5A (BROSI_A2677, BROSI_A0131, BROSI_A0501, and BROSI_A3864) was observed in Ca. B. sinica JPN1, and the expression levels of those genes were lower under conditions unfavourable for anammox metabolism.⁵⁴ Through a series of experimental results, they arrived at BROSI_A0501 (hao3) as the most plausible candidate gene responsible for nitrite reduction to NH₂OH. In addition, it was suggested that NH₂OH-forming nitrite reductase is involved in nitrite reduction by Ca. B. sinica by difference in nitrogen isotope fractionation in comparison with anammox bacteria possessing NO-forming enzymes.⁵⁵ Out of the five Brocadia species, only Ca. B. sinica lacks hao5B gene included in the core genome. Ca. B. fulgida lacks hao5A, whereas hao5B is duplicated in its genome (Fig. 4 and Supplementary Fig. S2). Although the function of hao5B is still unknown, since both Ca. K. stuttgartiensis and Ca. B. pituitae possess hao5B together with hao2 and hao3 and can utilize NO for N₂H₄ synthesis, there could be some relationship between the presence of hao5B and the availability of NO.

On the other hand, all hao-like genes in Ca. B. pituitae genome are expressed at various expression levels under regular operational conditions of an anammox bioreactor. High RPKM ratio of more than 10 was observed in several core genes, hao1, hao2, and hao5B, and the hao3 gene was also significantly expressed with the RPKM ratio of 3.09 (Supplementary Table S4). As predicted in Ca. K. stuttgartiensis and Ca. B. sinica, assuming hao2 and hao3 encode NO- and NH₂OH-forming nitrite reductase, respectively, it is possible that Ca. B. pituitae reduces nitrite to NO and NH₂OH using these enzymes and utilize them for N₂H₄ biosynthesis. Although details are unclear as yet, the expression of these two genes may be regulated in response to the environment in the anammox bioreactor. Therefore, Ca. B. pituitae presumably utilizes not only NO supplied by the ABC members, but also NO and NH₂OH by self-production for anammox metabolism.

Supplementary data

Supplementary data are available at DNARES online.