Cultivation and genomic analysis of Candidatus Nitrosocaldus islandicus, a novel obligately thermophilic ammonia-oxidizing Thaumarchaeon

Ammonia-oxidizing archaea (AOA) within the phylum Thaumarchaea are the only known aerobic ammonia oxidizers in geothermal environments. Although molecular data indicate the presence of phylogenetically diverse AOA from the Nitrosocaldus clade, group 1.1b and group 1.1a Thaumarchaea in terrestrial high-temperature habitats, only one enrichment culture of an AOA thriving above 50 °C has been reported and functionally analyzed. In this study, we physiologically and genomically characterized a novel Thaumarchaeon from the deep-branching Nitrosocaldaceae family of which we have obtained a high (∼85 %) enrichment from biofilm of an Icelandic hot spring (73 °C). This AOA, which we provisionally refer to as “Candidatus Nitrosocaldus islandicus”, is an obligately thermophilic, aerobic chemolithoautotrophic ammonia oxidizer, which stoichiometrically converts ammonia to nitrite at temperatures between 50 °C and 70 °C. Ca. N. islandicus encodes the expected repertoire of enzymes proposed to be required for archaeal ammonia oxidation, but unexpectedly lacks a nirK gene and also possesses no identifiable other enzyme for nitric oxide (NO) generation. Nevertheless, ammonia oxidation by this AOA appears to be NO-dependent as Ca. N. islandicus is, like all other tested AOA, inhibited by the addition of an NO scavenger. Furthermore, comparative genomics revealed that Ca. N. islandicus has the potential for aromatic amino acid fermentation as its genome encodes an indolepyruvate oxidoreductase (iorAB) as well as a type 3b hydrogenase, which are not present in any other sequenced AOA. A further surprising genomic feature of this thermophilic ammonia oxidizer is the absence of DNA polymerase D genes - one of the predominant replicative DNA polymerases in all other ammonia-oxidizing Thaumarchaea. Collectively, our findings suggest that metabolic versatility and DNA replication might differ substantially between obligately thermophilic and other AOA.


Introduction
Thaumarchaea (Brochier-Armanet et al., 2008) are among the most abundant archaeal organisms on Earth, and thrive in most oxic environments (Francis et al., 2007;Erguder et al., 2009;Schleper and Nicol, 2010;Bouskill et al. 2012;Prosser and Nicol, 2012;Stahl and de la Torre, 2012; a pH of 6.5 and a temperature of 73 °C. The culture was initially incubated without agitation in 100 ml glass bottles in the dark at 60 °C and checked weekly for ammonium and nitrite content of the medium by using Nessler´s reagent (K 2 HgI 4 -KOH solution; Sigma-Aldrich) and nitrite/nitrate test stripes (Merkoquant; Merck). Ammonium (1 mM NH 4 Cl) was replenished when completely consumed. At the same time pH was monitored by using pH test stripes (Machery-Nagel) and kept at pH 7-8 by titration with NaHCO 3 . When the pH dropped below 6 the enrichment culture ceased to oxidize ammonia, but activity was restored by readjusting the pH to between 7 and 8. The ammonium and nitrite concentrations were quantified photometrically (Kandeler and Gerber, 1988;Miranda et al., 2001) using an Infinite 200 Pro spectrophotometer (Tecan Group AG). The microbial community composition of the enrichment was regularly monitored by fluorescence in situ hybridization (FISH) with 16S rRNA-targeted probes labeled with dyes Cy3, Cy5, or Fluos as described elsewhere (Daims et al., 2005). Probes targeting most bacteria (EUB338 probe mix; Amann et al., 1990;Daims et al., 1999), most archaea (Arch915, Stahl and Amann, 1991) and most Thaumarchaea (Thaum726, Beam, 2015) were applied. All positive results were verified using the nonsense probe nonEUB338 (Wallner et al. 1993) labeled with the same dyes. Treatments with the macrolide antibiotic spiramycin (15 mg l -1 ), which partly retains its antibacterial activity at 60 °C (Zorraquinio et al., 2011), were performed as described in Zhang et al. (2015) together with serial dilutions ranging from 10 -5 to 10 -8 to obtain a highly enriched (~ 85 %) AOA culture that was used for further characterization. Growth rates of Ca. N. islandicus were determined across a range of incubation temperatures (50 °C to 70 °C). Triplicate cultures (25 ml) and negative controls (cultures not supplied with ammonium or inoculated with autoclaved biomass) were incubated for ten days in 100 ml Schott bottles without agitation in the dark at the respective temperature. Samples from these experiments were either stored at -20 °C for subsequent qPCR analyses (150 µl) or centrifuged (21,000 x g, 15min, 18 °C) to remove cells and the supernatant was stored at -20 °C for chemical analysis (600 µl). qPCR analysis with primers CrenamoA19F (Leininger et al, 2006) and CrenamoA616R (Tourna et al., 2008) targeting the archaeal amoA gene was otherwise performed as described in Pjevac et al. (2017) before the genome sequence of Ca. N. islandicus was available. However, subsequent analysis demonstrated that the employed qPCR primers contain mismatches to the amoA sequence of this AOA in the middle of the forward and reverse primer. The specific growth rate was calculated from log-linear plots of amoA gene abundance in individual cultures. In this analysis, three out of seven time points were interpolated through linear regression. To test whether the NO-scavenger 4,5,5,TCI,Germany) inhibits ammonia oxidation by Ca. N. islandicus, 40 ml aliquots of mineral medium containing 1 mM ammonium were inoculated with 10 % (v/v) of an exponential-phase culture and incubated in duplicates in the presence of 0, 33, and 100 µM PTIO, respectively. PTIO was dissolved in sterile mineral medium before addition to the cultures. The cultures not exposed to PTIO were supplemented with the same volume of sterile medium. The cultures were sampled (2 ml) at the beginning of the experiment and after 15 days of incubation. Nitrite and ammonium concentrations were measured as described above.
DNA extraction, genome sequencing, and annotation DNA from three replicate enrichment cultures containing Ca. N. islandicus as the only detectable ammonia oxidizer was extracted as described by Angel and Conrad (2013) and sequenced by Illumina HiSeq next generation sequencing (250 bp paired end reads). Since we did not obtain a complete genome with this approach we re-extracted genomic DNA from the enrichment at a later stage according to Zhou et al. (1996) yielding high molecular weight DNA. Genomic DNA was then sheared in a Covaris g-TUBE (Covaris, USA) at 9000 RPM for 2x 1 min. in an Eppendorf mini spin plus centrifuge (Eppendorf, DE). The DNA was run on a E-Gel™ EX 1 % agarose gel (ThemoFisher, USA) and small DNA fragments were removed by excising a band with a length of ~8 kb. The DNA was purified from the gel cut using the ultraClean 15 DNA Purification Kit (Qiagen, USA). The DNA was prepared for sequencing using the "1D Low Input gDNA with PCR SQK-LSK108" protocol (Oxford Nanopore Technologies, UK) and sequenced on a FLO-MIN106 flowcell using the MinION MK1b (Oxford Nanopore Technologies, UK) following the manufacturers protocol using MinKNOW (v. 1.7.14). The nanopore reads were basecalled using Albacore (V. 2.0.1) (Oxford Nanopore Technologies, UK). The complete genome was assembled using a hybrid approach combining the data from the Illumina and nanopore sequencing with the hybrid assembler Unicycler (v. 0.4.1, Wick et al., 2017). The genome bins of the two contaminating organisms were assembled from the Nanopore reads using Miniasm  and polished twice with the Nanopore reads using Racon (Vaser et al., 2017). No other microbe encoding genes indicative for ammonia-oxidation was identified in either of the two the metagenomes. The complete genome of Ca. N. islandicus was uploaded to the MicroScope platform (Vallenet et al., 2013) for automatic annotation, which was amended manually where necessary. The full genome sequence of Ca. N. islandicus has been deposited in GenBank (accession CP024014) and associated annotations are publicly available in MicroScope (Candidatus Nitrosocaldus islandicus strain 3F). Protein-coding genes from the novel Thaumarchaeon were compared to those from 30 Thaumarchaea with available genomic data (Table S1) downloaded from NCBI. The coding sequences (CDS) with accession numbers from each genome, as downloaded from NCBI, were combined with additional CDS predictions made by Prodigal (Hyatt et al., 2010) to account for variability in CDS predictions from different primary data providers and platforms. Predicted CDS from the novel Thaumarchaeon were aligned to CDS from reference genomes using blastp (Word_size=2, substitution matrix BLOSUM45). Genes were considered homologous only if the blastp alignment exceeded 50 % of the length of both query and subject sequences. CDS of Ca. N. islandicus that lacked any homologs in other Thaumarchaea were considered "unique". Unique CDS of unknown function were searched for secretion signals and for predicted membranespanning domains of the encoded proteins using the Phobius web server (Käll et al., 2007) and putative structures were determined using the Phyre2 web server (Kelley et al., 2015). Homology to "Thaumarchaea-core" proteins was assessed by cross-referencing the blastp homology search to the proteins defined for Ca. Nitrosotalea devanaterra by .

Phylogenetic analysis and habitat preference
For 16S rRNA and amoA gene-based phylogenetic analysis, the full-length 16S rRNA and amoA gene sequences of Ca. N. islandicus retrieved from the genome assembly were imported into the ARB software package (Ludwig et al., 2004) together with other full length 16S rRNA or amoA gene sequences from cultivated AOA strains and aligned with the integrated ARB aligner with manual curation. 171 sequences from the Aigarchaea were included in the alignment and used as outgroup in the 16S rRNA gene phylogenetic analyses. For the amoA gene phylogenetic analyses no outgroup was selected. The 16S rRNA and amoA gene consensus trees were reconstructed using Maximum-Likelihood (ML; using the GTRGAMMA evolution model), Neighbour Joining (NJ) and Maximum Parsimony (MP) methods. For all calculations, a sequence filter considering only positions conserved in ≥50 % of all thaumarchaeal and aigarchaeal sequences was used, resulting in 2444 and 488 alignment positions for the 16S rRNA and amoA genes, respectively. A Bayesian-inference phylogenomic tree was obtained using the automatically generated alignment of 34 concatenated universal marker genes (Table S2), which were identified by CheckM in Parks et al. (2015). This alignment was used as input for PhyloBayes (Lartillot et al., 2009) with ten independent chains of 4,000 generations using the CAT-GTR model; 2,000 generations of each chain were discarded as burn-in, the remainder were subsampled every second tree (bpcomp -x 2000 2 4000) and pooled together for calculation of posterior probabilities. Whole-genome based average nucleotide identity (gANI, Varghese et al., 2015) and average amino acid identity values (AAI, Konstantinidis and Tiedje, 2005) were calculated between the genomes of Ca. N. islandicus and Ca. N. yellowstonensis using sets of annotated genes supplemented with additional gene calls predicted by Prodigal (Hyatt et al., 2010). gANI was calculated using the Microbial Species Identifier (MiSI) method (Varghese et al., 2015). For AAI, bidirectional best hits were identified using blastp, requiring that query genes aligned over at least 70 % of their length to target genes (in each unidirectional blastp search). Query gene length was used to calculate a weighted average % identity over all best hit pairs and the calculations were repeated using each genome as query and target. The occurrence of organisms closely related to Ca. N. islandicus and Ca. N. yellowstonensis in publicly deposited amplicon sequencing data sets was assessed using IMNGS (Lagkouvardos et al., 2016) with the full-length 16S rRNA gene sequences of both organisms as query and a nucleotide identity threshold of 97 %. PolB amino acid sequences were extracted from the arCOG database (arCOG14 ftp://ftp.ncbi.nih.gov/pub/wolf/COGs/arCOG/ (arCOG15272, arCOG00329, arCOG00328, arCOG04926, arCOG15270). Additional thaumarchaeal polB sequences were identified using Ca. N. islandicus as a query in a blastp search against the nr protein database. These additional thaumarchaeal sequences, the polB sequence from Ca. N. islandicus and arCOG database sequences were de-replicated using usearch (Edgar, 2010) with -sortbylength and -cluster_smallmem (-id 0.99 -query_cov 0.9), aligned using default settings in mafft (Katoh and Standley, 2013) and a phylogenetic tree was calculated using FastTree (Price et al., 2010). Nitrilase superfamily amino acid sequences were obtained from . Alignment and phylogenetic reconstruction was carried out with Bali-Phy (Suchard and Redelings, 2006; randomize alignment, iterations=11000, burnin=6000). Posterior tree pools from 10 independent runs were combined to generate a 50 % PP consensus tree and to assess bipartition support. A dataset for assessing the phylogenetic relationship of the alpha subunit of 2-oxoacid:ferredoxin oxidoreductases (OFORs) was based on  and supplemented with additional indolepyruvate oxidoreductase (ior) sequences. Genomes available (as of October 30, 2017) from the NCBI genomes database were downloaded, genes were predicted using Prodigal V2.6.3 (Hyatt et al., 2010) and predicted genes were screened for iorA (TIGRFAM03336) using hmmsearch v3.1b2 (hmmer.org) with an e-value cutoff of 0.001. Genes meeting the search criteria were used as queries against the complete TIGRFAM database to ensure that the extracted iorA sequences matched the iorA model as the best-hit model with an e-value cutoff of 0.001. Reciprocal best-hit genes were required to align to the hmm over at least 500 contiguous bases. Amino acid sequences were then clustered into centroids using usearch v8.0.1517 (sortbylength and cluster_smallmem -id 0. 8 -query_cov 0.9;Edgar, 2010). Centroids were aligned using mafft v7.245 (Katoh and Standley, 2013) and trees were constructed using FastTree 2.1.4 (Price et al., 2010). The initial phylogenetic placement of Ca. N. islandicus iorA in the resulting large tree (3,179 sequences) was used to choose a small set of bacterial iorA sequences to include in the final tree. The final dataset was aligned using mafft v7.245 (Katoh and Standley, 2013) and trees were constructed using FastTree 2.1.4 (Price et al., 2010).

Electron microscopy
For scanning electron microscopy, Ca. N. islandicus cells were harvested by centrifugation (4,500 x g, 15 min, 25 °C) and fixed on poly-L-lysine coated slides with a filter-sterilized 2. in a 30 to 100 % (v/v) ethanol series, washed in acetone, and critical point dried with a CPD 300 unit (Leica). Samples were mounted on stubs, sputter coated with gold using a sputter coater JFC-2300HR (JEOL), and images were obtained with a JSM-IT300 scanning electron microscope (JEOL).

Results and Discussion
Enrichment and basic physiology of Ca. N. islandicus An ammonia-oxidizing enrichment culture was established from biofilm material sampled from a hot spring located in the geothermal valley Graendalur of South-Western Iceland. Temperature tests for optimal activity and growth were performed at different time points during the enrichment period and showed varying results, but below 50 °C and above 75 °C activity and growth was never observed. Only during the initial enrichment phase did ammonia oxidation occur at 75 °C. At 65 °C the highest ammonia oxidation rates and the shortest lag phases were usually measured (data not shown), however in a single experiment the optimal temperature was 70 °C (Fig. S1). Likely, these variations reflect varying abundance ratios of Ca. N. islandicus and accompanying bacteria over time as described in Lebedeva et al. (2008). A high enrichment level of a single AOA phylotype (see below) was achieved by applying the antibiotic spiramycin (15 mg l -1 ) followed by biomass transfers into fresh medium using serial dilutions. This enrichment culture showed near stoichiometric conversion of ammonium to nitrite when incubated at 65 °C ( Fig. 1). This was accompanied by growth of the AOA with a specific growth rate of 0.128 ± 0.011 d -1 (mean generation time of 2.32 ± 0.24 d), which is substantially slower than those reported for Ca. Nitrosocaldus yellowstonensis HL72, N. viennensis EN76, or N. maritimus SCM1 (de al Tore et al., 2008;Könneke et al., 2005;Martens-Habbena et al., 2009;Stieglmeier et al., 2014b; Table 1), but faster than a marine enrichment culture (Berg et al., 2015).

Genome reconstruction, phylogeny, and environmental distribution
Metagenomic sequencing of the enrichment culture with Illumina and Nanopore demonstrated that the current culture contained an AOA as the only taxon encoding the repertoire genes required for ammonia oxidation. Hybrid assembly allowed reconstruction of the complete genome of this AOA as one circular contiguous sequence of 1.62 Mbps length ( Table 1). The 16S rRNA gene and amoA gene of the newly enriched AOA are 96 and 85 % identical respectively to the genes of Ca. Nitrosocaldus yellowstonensis, the only other cultured obligately thermophilic AOA. The average amino acid sequence identity (AAI) and the genomic average nucleotide identity (gANI) between the genome and the one of Ca. N. yellowstonensis are 65.4 % (alignment fraction: 0.86) and 75.8 % (alignment fraction: 0.59), respectively, which is above the proposed genus and below the proposed species boundary thresholds (Qin et al., 2014;Varghese et al., 2015). Consequently, the enriched obligately thermophilic AOA was assigned to the same genus and referred to as Ca. Nitrosocaldus islandicus. According to 16S rRNA gene-and amoA gene-based phylogenies, Ca. N. islandicus is a novel member of the Nitrosocaldales clade, which seems to predominantly encompass AOA from thermal environments (Fig. 2). An extended phylogenomic analysis using a concatenated alignment of 34 proteins (Table S2) identified by CheckM (Parks et al., 2015) confirmed that Ca. N. islandicus represents a basal lineage within the known ammonia-oxidizing Thaumarchaea (Fig. 3). This result lends strong support to the earlier notion, which was based on single-gene 16S rRNA and amoA phylogenies (de la Torre et al., 2008), that the thermophilic Nitrosocaldales clade is an early diverging group of the ammonia-oxidizing Thaumarchaea. It would also be compatible with the possibility that archaeal ammonia oxidation originated in thermal environments (de la Torre et al., 2008;Hatzenpichler et al., 2008;Groussin et al., 2011;Brochier-Armanet et al., 2012). Metagenomic sequencing revealed that in addition to Ca. N. islandicus the culture also contained two heterotrophic bacterial contaminants, which were identified as a Thermus sp. and a member of the Chloroflexi phylum (Fig. 4). The enrichment level of Ca. N. islandicus was approximately 85 % based on read counts from the Nanopore sequencing, whereas the Thermus sp. and Chloroflexi accounted for 12 % and 3 %, respectively. FISH-analysis of the enrichment culture confirmed the dominance of Ca. N. islandicus and showed that the AOA grew mainly in aggregates, whereas the bacterial cells grew either co-localized with the archaeal flocs or planktonic (Fig. 5A). Electron microscopy demonstrated that the cells of Ca. N. islandicus are small (with a diameter of approximately 0.5 to 0.7 μm) and have an irregular coccoid shape (Fig. 5B). Morphologically they resemble the cells of Ca. N. yellowstonensis (Qin et al., 2017b). The environmental distribution of the two cultured Nitrosocaldales members and closely related AOA was assessed by screening all publicly available 16S rRNA gene amplicon datasets (n=93,045) for sequences highly similar (97 %) to the 16S rRNA genes of Ca. N. islandicus and Ca. N. yellowstonensis using the pipeline described by Lagkouvardos et al. (2016). This analysis revealed that these taxa are highly confined in their distribution and occur predominantly in terrestrial hot springs where they can reach high relative abundances between 11.4 % and 86 % (Ca. N. islandicus and Ca. N. yellowstonensis, respectively) of the total microbial community (Fig. 6). Interestingly, Ca. N. yellowstonensis-related organisms seem to occur mainly in hot springs described as alkaline with a pH of around 8.5, but were also detected in a sample from a Tibetan wastewater treatment plant (Niu et al., 2017). The unexpected detection of members of the Nitrosocaldales in the latter sample was confirmed by 16S rRNA-based phylogenetic analyses (data not shown) and it would be interesting to know whether this wastewater treatment plant is in some way connected to water from a close-by hot spring.

Genome features
Addition of the complete genome of Ca. N. islandicus to the set of available thaumarchaeal genome sequences (n=30) reduced the number of gene families identified as representing the "Thaumarchaea-core"  from 743 to 669 (reduction by 9.96 %; Table S3). In a few cases, genes with low sequence homology to apparently absent core gene families are actually present in the genome of Ca. N. islandicus, but were not scored as they did not match the alignment length criterion. For example, Ca. N. islandicus, like all other AOA sequenced to date, has a gene encoding the K-subunit of RNA polymerase class I, but with a low sequence similarity to the respective orthologous genes in other AOA. In a few other cases, enzymes found in all other AOA genomes are absent but functionally replaced by members of different enzyme families. For example, all other genome-sequenced AOA contain a cobalamin-dependent methionine synthase. In contrast, Ca. N. islandicus possesses only an unrelated cobalamin-independent methionine synthase, which is also found in some other AOAs.
In addition to updating the thaumarchaeal core genome we also specifically looked for genes that are present in Ca. N. islandicus, but were not reported for other AOA before. In the following sections, the most interesting findings from these analyses are reported and put in context. Like all other AOA, the Ca. N. islandicus genome encodes the typical repertoire for CO 2 fixation via the modified 3-hydroxypropionate/4-hydroxybutyrate (3HP/4HB) cycle and for archaeal ammonia oxidation ( Fig. 7; Fig. S2; Table 1) Spang et al., 2012;Könneke et al., 2014;Otte et al., 2015;Kerou et al., 2016). Unexpectedly however, the gene nirK encoding an NO-forming nitrite reductase is absent. NirK has been suggested to play an essential role for ammonia oxidation in AOA by providing NO for the NO-dependent dehydrogenation of hydroxylamine to nitrite (Kozlowski et al., 2016). Interestingly, ammonia oxidation by Ca. N. islandicus was completely inhibited after the addition of ≥33 µM of the NO-scavenger PTIO (Fig.  S3), a concentration that is lower or in the same range as previously reported to be inhibitory for other AOA (Shen et al., 2013;Jung et al., 2014;Martens-Habbena et al, 2015;Sauder et al., 2016). This finding suggests that NO is required for ammonia oxidation in Ca. N. islandicus despite the absence of NirK. The only other known AOA without a nirK gene are the sponge symbiont Cenarchaeum symbiosum Bartossek et al., 2010) and Ca. N. yellowstonensis (Stahl and de la Torre, 2012). For the uncultured C. symbiosum ammonia-oxidizing activity has not been demonstrated and the absence of nirK might have resulted from gene loss during adaptation to a life-style as symbiont. Ca. N. yellowstonensis is the closest cultured representative of Ca. N. islandicus, and the lack of nirK may thus be a common feature of the Nitrosocaldales. These AOA might produce NO by a yet unknown mechanism. In this context it is interesting to note that the hydroxylamine dehydrogenase of AOB, of which the functional homolog in archaea has not been identified yet, has recently been reported to produce NO instead of nitrite (Caranto and Lancaster, 2017). Alternatively, NO could be provided by accompanying organisms such as the Thermus and Chloroflexi-like bacteria that remain in the enrichment. Indeed, the genome bins obtained for these organisms both encode a nirK gene. The Thermus sp. genome bin further contains a norBC and narGH genes, in line with described denitrification capabilities for the genus Thermus (Alvarez et al., 2014). A dependence of Nitrosocaldales on NO production by other microorganisms could explain why no pure culture from this lineage has been obtained yet.
Ca. N. islandicus possesses genes coding for urease that are present in some but not all AOA Spang et al., 2012;Kerou et al., 2016;Lehtovirta-Morley et al., 2016;Sauder et al., 2017) (Fig. 7; Table 1), but lacks a cyanase that is used by Nitrososphaera gargensis for cyanate-based growth (Palatinszky et al., 2015). Additionally, the genome encodes an enzyme that either belongs to a novel class of the nitrilase superfamily or to the cyanide hydratase family ( Fig. 7;  Fig. S4). Nitrilases catalyze the direct cleavage of a nitrile to the corresponding acid while forming ammonia  and cyanide hydratases convert HCN to formamide. Both substrates are relatively thermostable (Miyakawa et al., 2002;Isidorov et al., 1992). Nitriles occur as intermediates of microbial metabolism (Kobayashi et al., 1993) and nitrile hydratases have previously been isolated from several thermophiles (Cramp et al., 1997;Almatawah et al., 1999;Kabaivanova et al., 2008). Furthermore, both compounds are intermediates of the proposed abiotic synthesis of organics at hydrothermal sites (Miller and Urey, 1959;Schulte and Shock, 2005) and could thus be available in the hot spring habitat of Ca. N. islandicus. Similar genes have been found in the genomes of several other AOA from the Nitrosopumilus and Nitrosotenuis genera (Walker et al., 2010;Lebedeva et al., 2013;Park et al., 2014;Bayer et al., 2016) (Table 1) and it will be interesting to find out for which metabolism they may be used in AOA. Intriguingly, Ca. N. islandicus might be able to ferment amino acids under anaerobic conditions as it contains the entire pathway used by some hyperthermophilic archaea for ATP generation from aromatic amino acids (Mai and Adams, 1994;Adams et al., 2001;Ozawa et al., 2012) (Fig. 7). In this pathway arylpyruvates are formed from aromatic amino acids by the activity of amino acid aminotransferases using 2-oxoglutarate as the amine group acceptor. The glutamate produced by this transamination can be recycled back to 2-oxoglutarate via glutamate dehydrogenase (gdhA) with the concomitant reduction of NADP + . With ilvE and aspC genes present, Ca. N. islandicus encodes at least two enzymes for which an aminotransferase activity specific for tyrosine, phenylalanine and aspartate has been demonstrated (Gelfand and Steinberg, 1977). Subsequently, these 2-ketoacids could be oxidatively decarboxylated and converted to aryl-CoAs by the oxygensensitive enzyme indolepyruvate oxidoreductase (Ozawa et al., 2012) encoded by iorAB using oxidized ferredoxin as electron acceptor. IorAB is absent from all other genome-sequenced AOA and the ior genes present in Ca. N. islandicus have the highest similarity to and cluster together with iorAB-genes found in Kyrpidia tusciae and Dadabacteria (Fig. S5). Finally, transformation of aryl-CoAs to aryl acids catalyzed by ADP-dependent acetyl-CoA/acyl-CoA synthetase (Glasemacher et al., 1997) leads to ATP formation via substrate-level phosphorylation (Fig. 7). Ca. N. islandicus encodes four acetyl-CoA/ acyl-CoA synthetases, two of which are most similar to non-syntenous homologs of acetyl-CoA/ acyl-CoA synthetases found in other AOA. However, the third gene is absent in all other AOA to date and its most similar homologs are encoded by species of the peptidolysing thermophilic archaea Thermoproteus and Sulfolobus and the fourth is most similar to an acetyl-/ acyl-CoA synthetase found in members of the thermophilic Bathyarchaea and Hadesarchaea. The fermentation of aromatic amino acids also requires regeneration of oxidized ferredoxin (reduced by IorAB) and NADP + (reduced by glutamate dehydrogenase). However, no canonical ferredoxin:NADP + oxidoreductase, or other enzymes (Buckel and Thauer, 2013) described to regenerate oxidized ferredoxin, are encoded in the genome of Ca. N. islandicus. It seems unlikely that the amount of ferredoxin oxidized by an encoded ferredoxin-dependent assimilatory sulfite/nitrite reductase (Fig. 7) would be sufficient to compensate for all ferredoxin reduced in the dissimilatory fermentation pathway. However, Ca. N. islandicus can also oxidize reduced ferredoxin with a 2:oxoglutarate-ferredoxin oxidoreductase (Fig. S5). NAD(P)H can be re-oxidized by a cytosolic, bidirectional, NAD(P)-coupled type 3b [NiFe] -hydrogenase that is encoded by Ca. N. islandicus in contrast to all other genomically characterized AOA ( Fig. 7; Table 1). NAD(P)H oxidation by this hydrogenase could lead to hydrogen generation, or the enzyme could act as a sulfhydrogenase that reduces zero valent sulfur compounds (produced by other organisms or present in the environment) to hydrogen sulfide (Ma et al., 1993, Adams et al., 2001. The hydrogenase genes are clustered at a single locus and code for the four subunits of the holoenzyme and accessory proteins (Fig. S6). This hydrogenase might also allow Ca. N. islandicus to use hydrogen as energy source providing reduced NAD(P)H under oxic conditions as this type of hydrogenase has been shown to tolerate exposure to oxygen (Bryant and Adams, 1989;Berney et al., 2014;Kwan et al., 2015). Surprisingly, the genome of Ca. N. islandicus lacks genes for both subunits of the DNA polymerase D (PolD), which is present in all other AOA and most archaeal lineages (including thermophiles) with the exception of the Crenarchaea (Cann et al., 1998;, Saw et al., 2015 ( Table 1). It is assumed that either PolD alone or together with DNA polymerases of the B family (PolB) is required for DNA synthesis and elongation in these archaea (Čuboňová et al., 2013;Ishino and Ishino, 2013;. The Ca. N. islandicus genome encodes only one B-type DNA polymerase (PolB1, Fig. S7) and one DNA polymerase of the Y family (PolY), generally considered to be involved in the rescue of stalled replication forks and enhancement of cell survival upon DNA damage (Friedberg et al., 2002). Recently, it has been demonstrated for the PolD-lacking Crenarchaeon Sulfolobus acidocaldarius that both its PolB1 and PolY have polymerase activities in vitro (Peng et al., 2016). However, Ca. N. islandicus (like other AOA) does not encode the PolB1binding proteins PBP1 and PBP2, which are required to form a multisubunit DNA polymerase holoenzyme together with PolB in the Crenarchaeon Sulfolobus solfataricus P2 (Yan et al., 2017). We hypothesize that Ca. N. islandicus may utilize one or both of the present polymerases for DNA replication, possibly in combination with its heterodimer PriSL, which has been demonstrated to function as a primase, a terminal transferase and a polymerase capable of polymerizing RNA or DNA chains of up to 7,000 nucleotides (Lao-Sirieix and Bell, 2004). It is also interesting to note that the obligate thermophile Ca. N. islandicus like all genomesequenced Thaumarchaea  does not encode a reverse gyrase, which is widespread in hyperthermophilic microbes including other archaea of the TACK superphylum (Heine and Chandra, 2009;Makarova et al., 2007;López-García et al., 2015), but is not essential for growth under these conditions (Atomi et al., 2004).
In conclusion, we have obtained a highly enriched (~ 85 %) culture of an obligately thermophilic AOA from a hot spring in Iceland. Despite the impressive diversity of AOA in high temperature environments as revealed by molecular tools (Zhang et al., 2008;Wang et al., 2009;Zhao et al., 2011;Nishizawa et al., 2013;Li et al., 2015;, cultivation of only a single obligately thermophilic AOA species -Ca. Nitrosocaldus yellowstonensis -was reported before (de la Torre et al., 2008). The newly enriched AOA represents a new species of the genus Nitrosocaldus and was named Ca. N. islandicus. Comparative analysis of its closed genome revealed several surprising features like the absence of DNA polymerase D and the lack of canonical NO-generating enzymes, although physiological experiments with a NO-scavenger demonstrated NO-dependent ammonia-oxidation, as described for other AOA (Shen et al., 2013;Jung et al., 2014;Martens-Habbena et al., 2015;Sauder et al., 2016). Furthermore, Ca. N. islandicus encodes the enzymatic repertoire for fermentation of aromatic amino acids that is, so far, unique among sequenced AOA. A pure culture of Ca. N. islandicus will be required to physiologically verify this genome-based hypothesis. Peptide or aromatic amino acid fermentation would enable an anaerobic lifestyle of Ca. N. islandicus and, if more widespread among Thaumarchaea not yet characterized (including mesophiles), might help explain their sometimes surprisingly high abundance in anaerobic ecosystems (Molina et al., 2010;Bouskill et al., 2012;Buckles et al., 2013;Beam et al., 2014;Lin et al., 2015).
Based on the data presented here, we propose the following provisional taxonomic assignment for the novel Thaumarchaeon in our enrichment culture.

Tables and figures
23  26 Figure 4. Sequence composition-independent binning of the metagenome scaffolds from two ammonia-oxidizing enrichment cultures. Circles represent scaffolds, scaled by the square root of their length. Clusters of similarly coloured circles represent potential genome bins. The x-axis shows binning of the scaffolds from an early enrichment culture, which still included other genera as well (not shown). The y-axis shows binning of the scaffolds from the latest enrichment culture containing only Ca. N. islandicus and the two remaining accompanying organisms. Genome bins for the Thermus (34 % complete) and the Chloroflexi (56 % complete) organism were obtained. The genome bin of the Chloroflexi organism contains genes that cluster within a clade of Nitrobacter/ Nitrolancea nitrite oxidoreductase (nxrAB) genes (data not shown). Since we did not observe nitrate production by the enrichment culture, the function of these genes remains unknown.  Kozlowski et al. (2016). Locus tags of all genes discussed in the main text are given in table S4. Like some but not all AOA, Ca. N. islandicus encodes all genes required to assemble an archaeal flagellar apparatus that is composed of the flagellar filament, the motor, and its switch Lehtovirta-Morley et al., 2016;Lebedeva et al., 2013;Qin et al., 2017;, although no archaellum could be detected in our electron microscopic analysis. The fla gene cluster of Ca. N. islandicus shows a similar arrangement to Nitrososphaera gargensis and contains six genes including one gene for structural flagellin subunit FlaB/FlaA as well as the flagellar accessory genes flaG, flaF, flaH, flaJ and flaI .    Nitrosocaldus islandicus strain 3F in a novel branch consisting of Thaumarchaea (Novel Thaumarchaea branch). Sequences and branch labels are from (29). Grey labels indicate "nonfused outliers" as indicated by (29) which were not assigned to any of the 12 named branches. Black labels indicate sequences obtained from the thaumarchaeal genomes. Sequences assigned to branches 10 and 11 in (29) that do not clade with other members of those branches in this phylogenetic tree are labelled "orphan".