Introduction

Over the last century, chemolithoautotrophic nitrification was believed to be a two-step process carried out by a consortium of microorganisms. However, this paradigm has been radically changed by epoch-making discoveries of single organisms able to oxidize ammonia to nitrate [1, 2]. These chemolithoautotrophic complete ammonia-oxidizing (comammox) bacteria belong to sublineage II of the deep-branching genus Nitrospira previously considered to only perform nitrite oxidation [3]. While attempts have been made to fully characterize comammox Nitrospira, currently only one isolated strain and a limited number of enrichment cultures have been obtained [4,5,6,7]. From two of these comammox Nitrospira cultures, kinetic characterization suggests that they have a high affinity for ammonia and prefer to inhabit oligotrophic environments [4, 5]. Thus, ecosystems with limited nutrient flow favor the presence of comammox Nitrospira, which may even dominate over canonical ammonia oxidizers [e.g., ammonia-oxidizing bacteria (AOB)]. However, field studies have suggested that factors other than ammonia availability will also play a role in the niche differentiation of ammonia oxidizers [8, 9].

The ecological significance of comammox Nitrospira has only been partially deciphered, although they have been found to be widely distributed in non-marine natural and engineered ecosystems. Environments, where they have been detected, include soils, natural freshwater settings, wastewater treatment installations, and drinking water production and distribution systems [10,11,12,13,14,15,16,17,18]. Except for the first and so far only isolated comammox strain Nitrospira inopinata, which was originally sampled from a hot water pipe (56°C) and cultured under moderately thermal conditions at 46 to 50 °C [1, 4], all other studies were associated with mesothermal conditions. This raises the question about the role comammox Nitrospira play in nitrogen cycling in thermal and hyperthermal habitats.

Geothermal habitats like submarine hydrothermal vents and terrestrial hot springs support a variety of chemolithoautotrophic microorganisms, possibly since the origin of cellular life [19,20,21]. Investigations of hot spring microbial communities can be traced back to the 19th century [22], while the understanding of their diversity and metabolic capacity has been significantly improved in the last two decades due to the advances in sequencing technologies. One compound in hot springs that is of particular interest is ammonium, which is the major nitrogenous compound observed in geothermal habitats [23]. Consequently, thermophilic aerobic ammonia-oxidizing archaea (AOA) have been widely found in thermal ecosystems [24,25,26,27]. Additionally, nitrite-oxidizing bacteria (NOB) affiliated with the genus Nitrospira or the phylum Chloroflexi have been identified in geothermal hot springs [28,29,30]. The potential for members of the genus Nitrospira, including the comammox species N. inopinata, to tolerate thermally induced stress indicates that also comammox Nitrospira may proliferate in high-temperature ecosystems, but no such report has been made to date for natural geothermal springs. This leads to questions of how widely distributed thermophilic comammox Nitrospira are, what strategies these species employ to protect themselves from heat stress, and if there are any genomic differences between comammox Nitrospira in thermal and non-thermal habitats.

Here, we report the occurrence and activity of comammox Nitrospira in terrestrial hot spring ecosystems, explore their potential survival mechanisms under (hyper)thermal conditions, and compare genomic features of comammox Nitrospira obtained from thermal and non-thermal habitats. These novel insights into the occurrence, activity, and adaptation of comammox Nitrospira in thermal habitats further expand our understanding of the global distribution of these unique microorganisms and provide important insights into microbial heat stress defense strategies.

Materials and Methods

Sample collection, DNA extraction, and metagenomic sequencing

A total of 10 biomass samples were collected for DNA extraction from the surface sediments (~top 5 cm) of five hot springs JinZe (JZ), ShuiReBaoZha (SRBZ), and ShiZiTouDuiMian (SZTDM) located at Tengchong county in Yunnan province, and DaGeJia (DGJ) and QuZhuoMu (QZM) located near Rikaze City in Tibet, China. All hot springs samples are at altitudes of ≥ 4000 m above sea level. At the time of sampling, these hot springs spanned a wide range of physicochemical parameters with temperatures and pH ranging from 36.7 to 80.0 °C and 3.0 to 8.2, respectively. Sample locations, sampling dates, and physicochemical parameters are summarized in Supplementary Table S1. The detailed methods for sample collection, chemical analyses, DNA extraction, and metagenomic sequencing are described in a previous study [31].

Metagenomic assembly and genome binning

Raw metagenomic reads were generated using a Hiseq 4000 instrument (Illumina, USA) and high-quality reads were obtained according to the procedures described previously [32]. Filtered reads were assembled subsequently using SPAdes v3.10.1 [33] with the following options: −k 21,33,55,77,99,111 –meta. Contigs with length >2500 bp were included for genome binning based on tetranucleotide frequency and sequencing depth using MetaBAT v2.12.1 [34]. To calculate sequencing depth, filtered reads were mapped to assembled contigs using BBMap v38.85 using a minimal identity setting of 0.97. The “jgi_summarize_bam_contig_depths” command in MetaBAT was performed to generate the sequencing depth table. Genomic bins were visualized using emergent self-organizing maps (ESOM) [35]. Completeness, contamination, and strain heterogeneity of each metagenome-assembled genome (MAG) were estimated using CheckM v1.0.5 [36]. To reduce contamination and heterogeneity, all MAGs were further manually examined to remove contigs with abnormal coverage or discordant positions within the ESOM. Finally, clean reads of each MAG were recruited using BBMap and reassembled using SPAdes v3.9.0 with parameters: −k 21,33,55,77,99,127 –careful.

Taxonomy and abundance of MAGs

Taxonomic assignment of the obtained MAGs was performed using GTDB-tk v1.7.0 [37]. The pairwise average amino acid identity (AAI) between MAGs and selected reference genomes was calculated as previously described [38]. The relative abundance of the retrieved MAGs was calculated using coverage information by dividing the sum of coverages for all contigs belonging to the respective MAG by the sum of coverages for all contigs, multiplied by 100. To compare the relative abundance of comammox Nitrospira with other populations such as AOB, AOA, and anammox bacteria, MAGs with the possession of any of the following genes were taken into consideration: ammonia monooxygenase (amoCAB), hydroxylamine dehydrogenase (hao), NO-forming nitrite reductase (nirS or nirK), hydrazine synthase (hzs), hydrazine dehydrogenase (hdh), and nitrite oxidoreductase (nxrAB).

Genome annotation and metabolic reconstruction

Protein coding sequences (CDS) were predicted using Prodigal v2.6.3 [39]. Functional annotation and metabolic pathway reconstruction were conducted by searching the predicted CDS against databases including NCBI non-redundant (nr), evolutionary genealogy of genes: Non-supervised Orthologous Groups (eggNOG), Kyoto Encyclopedia of Genes and Genomes (KEGG), and Carbohydrate-Active enZYmes (CAZy) using DIAMOND v0.7.9 with an E-value cut-off of < 1e−5 [40]. Putative hydrogenases belonging to PF00374 and PF02906 were identified by searching against the Pfam database using “hmmscan” command in HMMer v3.3.2 [41]. The amino acid sequences of putative hydrogenases were uploaded to the web tool HydDB (https://services.birc.au.dk/hyddb/) to determine the subtype classification and to predict possible functions. Amino acid usage and the IVYWREL index for each MAG were computed using custom-made Perl scripts (https://doi.org/10.6084/m9.figshare.22220173). Transmembrane helices in proteins were predicted using the online tool TMHMM-2.0 (https://services.healthtech.dtu.dk/service.php?TMHMM-2.0). The isoelectric point (pI) of each protein was calculated using the online tool IPC v2.0 (http://isoelectric.org/index.html).

Phylogenomic and phylogenetic analysis

To understand the phylogenomic relationships of the hot spring comammox Nitrospira, a set of 120 conserved marker genes was retrieved from 131 representative Nitrospira genomes from the NCBI Refseq database (Supplementary Dataset 1). The extracted sequences were aligned using MUSCLE v3.8.31 [42], and poorly aligned regions were removed using TrimAl v1.4.rev22 [43]. The phylogenomic tree was generated using IQ-TREE v1.6.10 [44] with ultrafast bootstrapping (-bb 1000), as well as a Shimodaira–Hasegawa–like approximate likelihood-ratio test (SH-aLRT, -alrt 1000). The best model of the tree was LG + F + R6, as determined by ModelFinder [45], and was well supported by the Bayesian Information Criterion.

For functional proteins (AmoCAB, NxrAB, and the different hydrogenase types), amino acid sequences were aligned and filtered. Phylogenies were constructed with the same procedures as described above. The best models for the AmoCAB, NxrAB, and hydrogenase-based phylogenies (based on the large subunits) were LG + F + R4, LG + F + G4, and LG + R9, respectively. These models were all well supported by the Bayesian Information Criterion. The generated maximum-likelihood trees were visualized using iTOL v6 (https://itol.embl.de).

Metatranscriptomic analysis

The in situ metabolic activity of comammox Nitrospira in hot springs was explored using metatranscriptomic sequencing to determine which of their genes were actively expressed. Due to difficulties with preserving RNA of sufficient quality from these extreme thermal environments, RNA was successfully extracted only from only one sample from QiaoQuan (QQ) in July 2022 and two samples from JinZe (JZ-2 and JZ-4) in Jan 2021 (Supplementary Table S1). Total RNA was extracted using the RNeasy PowerSoil Total RNA kit (QIAGEN) and 20–30 Gbps of raw sequence data was generated using a HiSeq 4000 instrument at Guangdong Magigene Biotechnology Co., Ltd. (Guangzhou, China). Raw reads were quality controlled using Sickle v1.33 (https://github.com/najoshi/sickle) with the parameters -t sanger –quiet -l 50. Seven of the 11 hot spring comammox MAGs were manually selected as reference genomes according to AAI ≥ 95% at the whole genome level (Supplementary Table S2). Clean reads were mapped onto reference genomes using BBMap v35.85 with the same parameters as above. The interconversions between the.bam and.bed file formats were carried out using SAMTools v1.3.1 [46] and BEDTools v2.30.0 [47]. Gene expression levels in fragments per kilobase of transcript per million fragments mapped (FPKM) were calculated using Cufflinks v2.2.1 [48] for each sample.

Results and discussion

Assembly of eleven comammox Nitrospira genomes from hot springs

Ten metagenomes were generated from sediment samples collected from five hot springs located in Tengchong (Yunnan Province) and Angren (Tibet Autonomous Region), China (Supplementary Table S1). Temperatures at the sampling locations ranged from 36.7 to 80.0 °C, while the pH varied from 3.0 to 8.2. Ammonium concentrations in examined sediments were mostly below 10 mg N/kg (0.7 mmol/kg), except for the sample taken from ShiZiTouDuiMian hot spring (16.3 mg N/kg, 1.2 mmol/kg).

A total of 11 comammox Nitrospira MAGs were successfully reconstructed (Table 1 and Supplementary Table S2). Specifically, four comammox Nitrospira MAGs were assembled from two different locations in JinZe hot spring (named JZ-3 and JZ-4), two from DaGeJia (DGJ01_4 and DGJ02_3), two from ShuiReBaoZha (SRBZ-2), two from QuZhuoMu (QZM_B4_2 and QZM_B4_3), and one from ShiZiTouDuiMian (SZTDM-2). These 11 high-quality comammox Nitrospira MAGs range in size from 2.46 to 2.95 Mbp with completeness estimates > 88% and contamination values < 3% (Table 1). For the DGJ01_4_Bin137 and DGJ02_3_Bin65 comammox Nitrospira, these MAGs were highly similar to each other (99.8% AAI), but stemmed from hot springs with disparate environments in terms of pH and temperature parameters (Supplementary Table S1). The gene expression patterns for DGJ02_3_Bin65 in three samples with pH of 6.8 to 8.6 and temperature of 27 to 70 °C (see below) indicated that these two strains were active in neutral to slightly alkaline environments at a broad temperature range. Whether they can also survive in acidic conditions remains unclear and could only be verified by cultivation-based experiments.

Table 1 Characteristics of 11 comammox Nitrospira reconstructed from the hot-spring metagenomes.
Full size table

In these hot spring environments, comammox Nitrospira MAGs represented the dominant ammonia-oxidizing organisms (0.01 to 1.28% relative abundance compared to all bacteria and archaea detected in metagenomes) in eight out of ten samples, except for the samples JZ-4_201709 (38.0 °C) and SZTDM-2_201709 (54.0 °C), in which AOB and AOA showed a higher relative abundance, respectively (Fig. 1a). Additionally, NOB were also found present in seven samples at relative abundances of 0.24–0.96%. The phylogenomic analysis revealed that the 11 comammox MAGs formed two sub-clades within the comammox Nitrospira clade A (Fig. 1b), with eight MAGs in one cluster with N. inopinata and the remaining three forming an adjacent sister lineage. This clustering with N. inopinata showed that all currently known thermophilic strains within the clade A comammox Nitrospira are closely related to each other. These results indicate that the thermal tolerance of hot spring comammox Nitrospira is likely inherited from their last common ancestor within clade A, which acquired the thermal tolerance after moving into thermal habitats. However, one cannot rule out the existence of other thermophilic comammox strains that are divergent from this cluster.

Fig. 1: Overview of nitrifying communities in analyzed hot springs and phylogeny of hot spring comammox Nitrospira.
figure 1

a Inferred relative abundances of comammox Nitrospira and AOB MAGs in analyzed hot springs. b Phylogenomic affiliations of comammox Nitrospira MAGs in relation to selected reference organisms. The maximum likelihood tree was constructed using 120 marker proteins from in total 142 genomes (see Methods). Genomes from Thermotogae were used as the outgroup. Nitrospira lineage I and lineage II are highlighted by blue and purple boxes, respectively. Numbers in brackets at collapsed lineages indicate the numbers of genomes/MAGs per group. Comammox Nitrospira clade A and clade B are indicated by green and yellow backgrounds, respectively. The identifiers of comammox Nitrospira MAGs obtained in the present study are shown in red. Bootstrap values (based on 1000 iterations) ≥70 and ≥90 are indicated by gray and black circles, respectively. Genome size and GC content of each Nitrospira genome are given on the left side with MAGs from this study indicated by filled bars. The source habitats of genomes/MAGs are summarized on the right side of the tree: DWTP, drinking water treatment plant; WWTP, wastewater treatment plant; RAS, recirculating aquaculture system.

Full size image

Predicted metabolic capacity of hot spring comammox Nitrospira

Consistent with their phylogenetic position within comammox Nitrospira clade A and confirming a comammox-type lifestyle, the 11 hot spring MAGs were found to contain genes of ammonia monooxygenase (amoCAB), hydroxylamine dehydrogenase (hao), and nitrite oxidoreductase (nxrAB), which are required for complete oxidation of ammonia to nitrate (Fig. 2). AmoCAB and nxrAB in each MAG co-localized on contiguous scaffolds and formed monophyletic clades with N. inopinata in phylogenetic analyses (Supplementary Fig. S1). While acknowledging the limitation given by incomplete MAGs, there is a notable absence of ammonia transporters (Rh50 and amt) in nearly all hot spring comammox Nitrospira MAGs, except for JZ-3_Bin7, despite their presumable dependency on ammonia for assimilatory processes. Despite undetected, urea is a general intracellular metabolite of prokaryotes and eukaryotes and is also released during the remineralization of organic matter [49]. A full complement of genes for urea utilization was observed in all 11 hot spring comammox Nitrospira MAGs (Fig. 2, Supplementary Fig. S2) [1, 4, 6], and the ammonia generated from intracellular urea hydrolysis may potentially support the growth of hot spring comammox Nitrospira in ammonium-limited environments. Analogous to other reported comammox genomes, hot spring comammox Nitrospira also lack an assimilatory nitrite reductase, suggesting their inability to grow on nitrite or nitrate as the only nitrogen source.

Fig. 2: Cell cartoon based on the genome annotations of 11 hot-spring comammox Nitrospira MAGs.
figure 2

Shown are pathways and genes involved in metabolisms of carbon, hydrogen, nitrogen and sulfur, as well as respiration, electron transport, and stress defense. Metabolic modules are indicated by distinct color schemes. The presence of metabolism types across 11 hot spring comammox Nitrospira MAGs from this study is indicated as follows: dark blue solid circles represent all 11 MAGs; light blue solid circles represent 5 to 10 MAGs; light blue hollow circles represent 1 to 4 MAGs; and gray solid circles represent 0 MAGs. The numbering scheme associated with the shown pathways is given in Supplementary Dataset 2. Respiratory chain complexes are labelled by Roman numerals. Abbreviations: 2-Oxo 2-Oxo-glutarate, APS adenosine 5′-phosphosulfate, Arg arginine, Car-P carbamoyl-P, Cit citrate, Citr citrulline, Cys cysteine, EhrL energy-converting hydrogenase-related large subunit, Fum fumarate, HydG [NiFe]-Group 3b hydrogenases, Iso isocitrate, L-Arg L-Arginosuccinate, LCFA long-chain fatty acids, Mal malate, Orn ornithine, Oxa oxaloacetate, PAPS 3′-phosphoadenosine 5′ -phosphosulfate, PHA polyhydroxyalkanoate, Pit inorganic phosphate transporter, Poly-P polyphosphate, Pst phosphate transporter, rTCA cycle, reductive tricarboxylic acid cycle, sHSP small heat-shock protein, SOD superoxide dismutase, Suc succinate, Suc-CoA succinyl-CoA.

Full size image

In comparison to mesophilic comammox Nitrospira, the pentaheme cytochrome c nitrite reductase (nrfAH) appears to be exclusively present in hot spring comammox MAGs (Fig. 2 and Supplementary Dataset 2) and the moderately thermophilic N. inopinata [1]. This suggests the potential of thermophilic comammox organisms to perform dissimilatory nitrate reduction to ammonium (DNRA), using the potentially reversible NXR for the initial reduction of nitrate to nitrite. This unique trait of comammox genomes recovered from high-temperature environments indicates a possible role of DNRA in their survival as previously suggested for other thermophilic bacteria and archaea [50, 51]. Specifically, the capability to perform DNRA provides an alternative mechanism for comammox Nitrospira to conserve energy under anoxic conditions, potentially allowing them also to survive in anoxic environments. Alternatively, using DNRA to produce ammonia could be a strategy for hot spring comammox Nitrospira to survive with nitrite as the only nitrogen source [1], and thus under the ammonium-limited conditions observed in hot springs. Even then, however, a low-potential electron donor for quinone reduction would be required to initiate the DNRA process and to provide electrons for biosynthetic processes, including CO2 fixation. In these pristine hot springs, organic metabolites (released by other organisms) or H2 (a potential product of the geochemical reduction of H2O at high temperatures and biomass fermentation at moderately thermal conditions) could be potential electron donors for DNRA. This is supported by the presence of key metabolic genes in the hot spring comammox Nitrospira MAGs that are potentially associated with these processes, i.e., fadD, crt, and acd for long-chain fatty acid catabolism, and the group 3b [NiFe] hydrogenase that might facilitate H2 oxidation (Fig. 2 and Supplementary Dataset 2). Together, these results suggest that hot spring comammox Nitrospira have versatile ecological functions as both ammonia sinks and sources.

In high-temperature environments, metabolic strategies that utilize or produce H2 represent an essential approach to energy conservation by chemolithoautotrophic microorganisms [23, 52]. The bidirectional group 3b [NiFe] hydrogenases and accessory proteins for hydrogenase maturation, which widely occur in canonical NOB and comammox Nitrospira [1, 53, 54], have been also observed in nearly all hot spring comammox Nitrospira (Fig. 3a, b). As observed in a hyperthermophilic archaeon (Pyrococcus furiosus) [55], the group 3b [NiFe] hydrogenases might also participate in sulfur cycling, which is another key element commonly found in terrestrial hot springs. Despite these speculated roles, the bona fide function of the group 3b [NiFe] hydrogenases in Nitrospira has not been characterized thus far. Moreover, we also observed the presence of energy-converting hydrogenase-related complexes (Ehr) in seven hot spring comammox Nitrospira MAGs presented here (Fig. 3c and Supplementary Fig. S3), which are phylogenetically close to the group 4 f [NiFe] hydrogenases but lack the typical CxxC motifs required for binding of the [NiFe] center. Thus, these Ehr complexes unlikely perform as [NiFe] hydrogenases and their function remains unknown. However, it is interesting to note their similarity to subunits of respiratory Complex I and of energy-converting hydrogenases (Ech), which are involved in electron transport and energy transduction [56].

Fig. 3: Classification and phylogeny of hydrogenases and energy-converting hydrogenase-related (Ehr) complexes.
figure 3

a The maximum likelihood tree shows the phylogenetic affiliation of hydrogenases and Ehr complexes identified in comammox Nitrospira in relation to 305 reference sequences. The tree is color-coded by hydrogenase subgroup. The red stars represent hydrogenases and Ehr complexes from hot spring comammox Nitrospira. The maximum likelihood trees show the phylogenetic affiliation of the b Group 3 [NiFe] hydrogenases and c Group 4 [NiFe] hydrogenases and Ehr complexes identified in comammox Nitrospira and reference hydrogenases. Bootstrap values (based on 1000 iterations) ≥ 70 and ≥ 90 are indicated by gray and black circles, respectively.

Full size image

Transcription of core metabolic pathways

The in situ activity of hot spring comammox Nitrospira was affirmed by metatranscriptomic sequencing (Fig. 4). To avoid biases in estimating gene expression levels of highly similar MAGs, we manually selected seven out of the 11 hot spring comammox MAGs as reference genomes based on an AAI ≥ 95% and genome quality (Supplementary Table S2). From this analysis, transcriptional activities of 976 out of 19142 (5.1%) encoded genes in these seven MAGs were detected, suggesting that these organisms have an active role in the three hot spring samples that were measured.

Fig. 4: Gene expression of seven comammox Nitrospira MAGs for nitrogen metabolism, thermal adaptation and oxidative stress across three hot springs derived metatranscriptomes.
figure 4

Colour intensity represents relative abundance of gene expression as measured by FPKM (Methods section). The white boxes indicate the gene expression is undetectable.

Full size image

The nitrification pathway was at least partially expressed across all seven hot spring comammox reference MAGs in all three metatranscriptomes (Fig. 4). For example, in JZ-2, JZ-4, and QQ, the MAGs JZ-4_Bin17, QZM-B4_2_Bin429, and SRBZ-2_Bin320, respectively, expressed the entire suite of genes related to nitrification including amoCAB, hao, and nxrAB. Moreover, the expression of nitrification-related genes accounted for the majority of transcripts assigned to the hot spring comammox Nitrospira MAGs. The expression of genes associated with nitrification constituted 4.4% and 4.0% of comammox transcripts for SBRZ-2_Bin320 in sample QQ and JZ-4_Bin17 in sample JZ-2, respectively (Supplementary Fig. S4). This indicates that nitrification was a key process catalyzed by comammox Nitrospira in the examined hot springs. Additionally, low expression levels of other nitrogen metabolism pathways were observed in at least one hot spring comammox Nitrospira MAG. These genes included NO-forming nitrite reductase (nirK), pentaheme cytochrome c nitrite reductase (nrfA), and urea transporters (urtABC). This suggests that hot spring comammox Nitrospira can be involved in additional nitrogen-cycling processes such as nitrite reduction to nitric oxide, DNRA, and urea utilization.

Moreover, the expression of complete nitrification pathway was not consistently observed for all comammox MAGs. For example, DGJ02_3_Bin65 expressed nxrAB in all three samples while no hao or amoCAB transcripts were detected, and amoC appeared to be the only transcribed nitrification-related genes by QZM_B4_3_Bin89. This lack of nitrification gene expression by some comammox strains might reflect limited metabolic activity of these organisms under extreme conditions in hot springs. Similar to our observations, partial expression of nitrification-related genes in comammox Nitrospira was also observed in a previous study [57], where the transcripts of amoCAB and nxrAB were inconsistently detected across different comammox species.

The transcription of genes associated with thermal adaptation also confirmed the activity of hot spring comammox Nitrospira and shed light on their potential survival mechanisms. Heat shock proteins (HSPs) are expressed by cells in response to external stress. They contribute to removing and refolding damaged proteins, protein assembly, secretion, and regulation of transcription factors, and may also play a protective role for other proteins by forming temperature-dependent complexes [58]. As expected, a variety of chaperones were encoded and transcribed by all seven selected hot spring comammox Nitrospira MAGs (Fig. 4 and Supplementary Dataset 2). For example, the gene for the highly conserved chaperone protein DnaK, which functions in protein folding and preventing the aggregation of unfolding proteins under stress [59], was expressed by nearly all hot spring comammox Nitrospira MAGs. Several copies of small HSPs (HSP20) were detected in SRBZ2-Bin320 and exhibited high expression levels in the metatranscriptomic dataset from the sample QQ (Supplementary Fig. S4). Chaperonins of the GroEL-GroES and caseinolytic protease (Clp) families were also highly transcribed by hot spring comammox Nitrospira. GroEL-GroES chaperonins can encapsulate and fold proteins, while Clp family members may regulate the disaggregation and degradation of distorted proteins and are also able to orchestrate DnaK to refold misfolded proteins [60]. Additionally, nucleotide-exchange factor (GrpE), ATP-dependent heat-shock locus genes (HslVU), and several genes associated with oxidative stress defense were highly expressed, indicating that diverse thermal stress mitigation mechanisms are adopted by comammox Nitrospira.

Unique amino acid composition of hot spring comammox Nitrospira

Adjusting the amino acid compositions of proteins is an important adaptation strategy of microorganisms to overcome thermal stress [61, 62]. In agreement with this hypothesis, comparing the amino acid compositions of 30 clade A comammox Nitrospira MAGs (Supplementary Text S1) revealed that comammox Nitrospira recovered from thermal and non-thermal habitats have evolved distinct amino acid usage patterns (Supplementary Fig. S5). Specifically, thermophilic comammox Nitrospira utilize more non-polar and charged amino acids at the expense of polar amino acids (p < 0.01). This phenomenon may be attributed to the chemical characteristics of different amino acids: strong ionic bonds between oppositely charged amino acid residues increase the stability of a protein, while polar amino acids (e.g., Asn and Gln) generally result in a decrease in protein stability due to thermal deamidation [63, 64]. The specific amino acid usage pattern by thermophilic comammox Nitrospira is in line with a previous comparison between the amino acid composition of 60 thermophilic proteins and their mesophilic homologues [65], as well as general observations on protein composition in other thermophiles [64, 66]. These results strongly suggest that thermophilic phenotypes of comammox Nitrospira can be identified based on their predicted amino acid composition deduced from genomic materials. Consequently, we further tested a typical amino acid composition-based index, the IVYWREL index [67], to decern thermophilic comammox Nitrospira from non-thermophilic members. As expected, the IVYWREL index was significantly different (p < 0.01) for thermophilic and mesophilic comammox Nitrospira (Supplementary Fig. S6), providing further evidence for compositional differences based on their amino acid usage profiles. Notably, the IVYWREL index of Nitrospira sp. ST-BIN4, Nitrospira sp. SG-bin2, Nitrospira sp. UBA2082, and Nitrospira sp. UBA6909, which were obtained from non-thermal habitats but are phylogenetically close to thermal comammox Nitrospira (Fig. 1), was determined to be 40.0~40.3 and thus also significantly lower than that of thermal comammox Nitrospira (p < 0.01). This lends strong support to the hypothesis that the observed unique amino acid profiles are genuinely due to the thermophilic phenotype rather than phylogenetic affiliation.

Genomic properties shed light on the adaptation of hot spring comammox Nitrospira

In addition to the changes in amino acid utilization, we also compared the genomic properties of 30 clade A comammox Nitrospira MAGs (Supplementary Text S1) and observed several notable differences between thermal and non-thermal clade A comammox Nitrospira that may shed light on possible adaptation strategies in high-temperature environments. Specifically, the thermophilic comammox Nitrospira MAGs have a significantly smaller genome size (2.46–3.30 Mbp) than mesophilic clade A comammox Nitrospira (2.93–4.55 Mbp; p < 0.05) (Fig. 5a). This is not surprising, as genome size reduction is considered to be beneficial for microbial heat tolerance through the reduced energy requirements for nucleotide synthesis [68]. For example, Sabath et al. [68] compared more than 1000 prokaryotes genomes and observed significant differences in the genome sizes between mesophiles, thermophiles, and hyperthermophiles, and Sorensen et al. [69] reported a significant and negative correlation between the average genome size and the site temperature in 12 metagenomes obtained from fire-affected soils. Correspondingly, a strong correlation between the genome size and the predicted optimal growth temperature (OGT, see details in Supplementary Text S1) was observed here for the 30 clade A comammox Nitrospira genomes (Supplementary Fig. S7).

Fig. 5: Comparisons of general genomic properties between 12 thermal and 18 non-thermal clade A comammox Nitrospira (Supplementary Dataset 4).
figure 5

Statistical analysis was conducted using the Wilcoxon rank sum test. a genome size; b GC content; c predicted optimal growth temperature (OGT); d gene count; e average gene length; f coding density.

Full size image

Another important genomic feature is the GC content. With contradictory results observed in different studies [67, 70, 71], the correlation between GC content and microbial tolerance against heat stress is still elusive. Here, we found that the GC content was significantly higher in thermophilic comammox Nitrospira MAGs than in non-thermal clade A comammox Nitrospira (55.48–59.31% vs. 54.67–57.89%; p < 0.05) (Fig. 5b). As expected, the OGT was also found to be significantly higher for the genomes derived from thermal habitats (Fig. 5c, p < 0.05). Additionally, the observed decreased gene count, increased average gene length, and increased coding density (Fig. 5d-f) indicate that dispensable and low-stability proteins might have been preferentially lost from the genomes of these thermophilic comammox Nitrospira [72]. Moreover, thermophilic comammox Nitrospira generally possess fewer genes encoding for transporters and bacterial secretion systems in comparison to non-thermal clade A comammox Nitrospira (Supplementary Fig. S8). This observation is in agreement with the hypothesis that the reduction of functional complexity is likely a cost-minimizing and energy-saving mechanism for thermophiles to adapt to elevated temperatures [73]. Taken together, the observed genomic features of thermophilic comammox Nitrospira indicate that these organisms are well adapted to survive in high-temperature ecosystems.

Ecological implications

The discovery and characterization of comammox Nitrospira significantly changed our understanding of the microbially mediated nitrogen cycle, as it overturned a century of accumulated knowledge by demonstrating that complete nitrification can be carried out by a single rather than multiple microorganisms in concert. Here, we report the detection of comammox Nitrospira in hot spring environments at temperatures up to 80 °C and their transcriptional activities at temperatures up to 70 °C. While numerous efforts have been made to show the ubiquitous existence of comammox Nitrospira in different environmental settings [12, 13, 53], the present study adds a key element to research into biogeochemical nitrogen cycling in high-temperature environments.

In comparison to human-affected (e.g., contaminated, fertilized, or engineered) soils and aquatic systems, the ammonium concentrations in the examined hot springs were much lower. This could theoretically explain the dominance of comammox Nitrospira over the other ammonia-oxidizing guilds in eight out of ten samples because they likely are well adapted to nitrogen-limited conditions [4, 5]. Additionally, the comparison of genomic properties showed relatively smaller genome sizes of hot spring comammox bacteria, which generally reflects a smaller cell size [74, 75] and tends to be advantageous for optimizing the surface-to-volume ratio for the optimal uptake of low-abundance nutrients [76,77,78]. However, limited ammonia availability may not be the only factor favoring their occurrence, as comammox Nitrospira were found to dominate over other ammonia oxidizers in many soil systems where the ammonium concentration varied from 0.1 to 20 mg kg−1 soil [79,80,81,82]. Other environmental factors such as pH, oxygen level, temperature, and salinity also likely contribute to the niche differentiation among ammonia oxidizers [79, 81], and their differential responses to these factors are often used to explain their co-occurrence, or to compare interspecies competitiveness of different ammonia-oxidizing guilds. Temperature and pH directly influence the kinetics of ammonia oxidizers [78, 83, 84], and a recent study demonstrated that the ammonia affinity constants of N. inopinata and the AOA ‘Candidatus Nitrosocosmicus oleophilus MY3’ increased 3–4 times when the pH was elevated from ~6.5 to ~8.5 [78].

We also searched the 16S rRNA and amoA gene sequences retrieved from SRBZ-2_Bin320 against the public database and found highly similar sequences in other thermal habitats in Russia, India, Argentina, and Australia (Supplementary Text S2 and Dataset 3), suggesting a potentially wide distribution of comammox Nitrospira in thermal habitats. It should be noted that these detected strains identified based on short sequences are only potential comammox bacteria. Future studies, including cultivation efforts in combination with metagenomic, metatranscriptomic, and metaproteomic analyses, are required to confirm their roles in ammonia oxidation in these systems.

Conclusions

Collectively, this study confirms the frequent distribution and transcriptional activity of comammox Nitrospira in high-temperature ecosystems and highlights their contributions to nitrogen conversions therein. By comparing thermal and non-thermal comammox bacteria, the results provide insights into potential adaptation strategies that hot spring comammox bacteria use against heat stress, including adapted amino acid usage profiles, modified genomic features, and expression of a large number of HSPs. This study is also of particular importance as terrestrial hot springs represent an important niche supporting a variety of chemolithoautotrophic extremophiles, and understanding the microbial nitrogen conversions therein is highly relevant from the ecological point of view [19,20,21]. While the contributions of AOB and AOA to nitrogen cycling in hot springs have been documented previously [24,25,26,27, 29, 85], the present findings fill an important knowledge gap by providing evidence for the occurrence and activities of comammox bacteria in the examined hot springs.

While the in situ transcriptomic evidence presented here confirms the activity of comammox Nitrospira in hot springs for the first time, follow-up research will be needed to further validate the results. Once enrichment cultures of these comammox organisms become available, targeted physiological experiments can provide more direct evidence of their activities, thermal adaptations, and further ecophysiological characteristics. Currently, environmental data, for example the concentrations of H2, organic compounds, urea, and dissolved oxygen are lacking for most hot springs we analyzed. Such information will be needed to better understand how thermophilic nitrifiers, including comammox bacteria, cope with extreme conditions and regulate their metabolism in a variable environment.