Incomplete denitrification phenotypes in diverse Thermus species from diverse geothermal spring sediments and adjacent soils in southwest China

A few members of the bacterial genus Thermus have been shown to be incomplete denitrifiers, terminating with nitrite (NO2−) or nitrous oxide (N2O). However, the denitrification abilities of the genus as a whole remain poorly characterized. Here, we describe diverse denitrification phenotypes and genotypes of a collection of 24 strains representing ten species, all isolated from a variety of geothermal systems in China. Confirmed terminal products of nitrate reduction were nitrite or N2O, while nitric oxide (NO) was inferred as the terminal product in some strains. Most strains produced N2O; complete denitrification was not observed. Denitrification phenotypes were largely consistent with the presence of denitrification genes, and strains of the same species often had the same denitrification phenotypes and largely syntenous denitrification gene clusters. Genes for nirS and nirK coexisted in three Thermus brockianus and three Thermus oshimai genomes, which is a unique hallmark of some denitrifying Thermus strains and may be ecologically important. These results show that incomplete denitrification phenotypes are prominent, but variable, within and between Thermus species. The incomplete denitrification phenotypes described here suggest Thermus species may play important roles in consortial denitrification in high-temperature terrestrial biotopes where sufficient supply of oxidized inorganic nitrogen exists. Supplementary Information The online version contains supplementary material available at 10.1007/s00792-022-01272-1.


Introduction
Complete denitrification is the process by which microorganisms sequentially reduce nitrate (NO 3 − ) to dinitrogen (N 2 ) through anaerobic respiration using the following Chrisabelle C. Mefferd and Enmin Zhou contributed equally.
* Wen-Jun Li liwenjun3@mail.sysu.edu.cn * Brian P. Hedlund brian.hedlund@unlv.edu enzyme complexes: nitrate reductase, Nar and/or Nap (NO 3 − → NO 2 − ); nitrite reductase, NirS and/or NirK (NO 2 − → NO); nitric oxide reductase, Nor (NO → N 2 O); and nitrous oxide reductase, Nos (N 2 O → N 2 ). Traditionally, most denitrifiers were thought to be full denitrifiers that catalyze all steps of the pathway (Zumft 1997). However, microbes with truncated, incomplete, or alternative denitrification pathways are also known. For example, Sanford and colleagues (2012) described so-called atypical nos genes, comprising all known nos genes outside the Pseudomonadota (formerly informally Proteobacteria (Oren and Garrity, 2021)), and showed that they are frequently present in genomes that do not contain other denitrification genes. The atypical Nos is present in 13 bacterial phyla and both Thermoproteota and Euryarchaeota, and typically contains Sec, rather than Tat N-terminal signal sequences. Other denitrifiers appear to lack a cytochrome bc 1 complex needed for N 2 O reduction to N 2 , but have an alternative complex III (ACIII), which performs the same function as the cytochrome bc 1 complex by transferring electrons from the quinol pool to N 2 O (Refojo et al., 2012). Still others contain novel NO reductase systems, sNOR, eNOR, and gNOR, instead of the canonical qNOR or cNOR, which encode enzymes that reduce NO to N 2 O (Hemp & Gennis, 2008;Hemp et al., 2015;Stein et al., 2007). Finally, some microorganisms carry out incomplete denitrification ending with N 2 O (Hart et al., 1965), which can result from mutations in the nosZ gene (Zumft and Kroneck 2006) or the absence of nos genes (Murugapiran et al., 2013).
Terrestrial geothermal environments host an array of thermophiles with differing denitrification abilities. Various reductases have been found in sequenced genomes of thermophiles belonging to both Thermoproteota (e.g., Aeropyrum, Sulfolobus, and Pyrobaculum) and Euryarchaeota (e.g., Archaeoglobus) (Cabello et al. 2004); however, many of these have not been examined functionally. Nitrate reduction is better known in thermophilic members of several bacterial thermophiles, including members of the Aquificota (Hedlund et al. 2015a;Huber et al. 1992), Bacillota (formerly informally Firmicutes (Oren and Garrity 2021)) (Poli et al. 2009), and Deinococcota. Members of the bacterial genus Thermus are widely distributed in geothermal systems and have been studied as models of thermophilic nitrate reduction and denitrification (Cava et al. 2009). Some strains of Thermus thermophilus reduce nitrate to nitrite (Cava et al. 2008), while other isolates of T. thermophilus and Thermus oshimai are incomplete denitrifiers that terminate with N 2 O as a final product (Hedlund et al. 2011). However, most Thermus species have been shown to be positive for nitrate reduction using traditional biochemical tests (Albuquerque et al. 2018), which qualitatively measure nitrate and nitrite but not gaseous products. As a result, their capacity for denitrification is not well understood.
Several genetic and genomic studies have shed light on nitrate reduction and denitrification pathways in Thermus. Genes for nitrate reduction are found within the nar gene cluster and neighboring genes code for other denitrification genes (Gounder et al. 2011;Jiao et al. 2022;Mefferd et al. 2016;Murugapiran et al. 2013;Zhou et al. 2016). Thermus denitrification genes are sometimes present on plasmids (Ramirez-Arcos et al. 1998;Brüggemann et al. 2006). For example, genes encoding the ability to reduce nitrate to nitrite in T. thermophilus HB8 and NAR1 comprise three adjacent gene clusters, nar, nrc, and dnr, which are located on a megaplasmid termed the nitrate conjugative element (NCE). The megaplasmid carrying the NCE can be transferred among T. thermophilus strains by conjugation (Ramirez-Arcos et al. 1998), suggesting horizontal gene transfer (HGT) of denitrification genes is possible in natural populations of Thermus. Other denitrification genes can also be found on megaplasmids. Such is the case with T. oshimai JL-2 and T. thermophilus JL-18, whose megaplasmids are approximately 0.27 Mb and 0.26 Mb, respectively, and contain a gene cluster for the reduction of nitrate to N 2 O (Murugapiran et al., 2013). In some Thermus strains, the megaplasmid has been shown to be less stable and fast evolving compared to the chromosome (Brüggemann et al. 2006;Murugapiran et al. 2013). The possibility of HGT of denitrification genes is further supported by research done using Thermus species as models for thermophilic HGT. For instance, whole-genome studies of Thermus scotoductus SA-01 (Gounder et al. 2011) and T. thermophilus strains HB8 and HB27 (Kumwenda et al. 2014) have uncovered evidence for large-scale genetic loss, acquisition, and rearrangement. T. thermophilus HB27 is naturally competent, containing many proteins associated with competence and conjugation (Averhoff 2009), and can take up DNA at any stage during growth (Hidaka et al. 1994;César et al. 2011). Given the importance of HGT in the genus Thermus and the variable presence of denitrification genes on mobile genetic elements, denitrification genotypes and phenotypes might be expected to respond to selective pressures imparted by the geochemistry of individual springs and spring systems.
Our previous research on geothermal springs in the U.S. Great Basin has revealed the coexistence of T. thermophilus, T. oshimai, T. aquaticus and Thermus sediminis in sediments near the upper temperature limit for the genus (> 80 °C) (Dodsworth et al. 2011a;Hedlund et al. 2011;Zhou et al. 2020). The T. thermophilus and T. oshimai strains produced nitrous oxide as the terminal denitrification product and their presence coincided with high rates of nitrous oxide production in situ. Comparative analysis of representative genomes from these strains revealed a full complement of genes for the respiration of nitrate to nitrous oxide in each species, with genomes encoding the isofunctional tetraheme cytochrome cd1-containing nitrite reductase (NirS) and T. oshimai JL-2 additionally encoding the copper-containing nitrite reductase (NirK) (Murugapiran et al., 2013). Finally, in situ electron donor stimulation experiments showed that only yeast extract and peptone stimulated denitrification, whereas inorganic electron donors or defined organic electron donors (glucose and organic acid mixtures) had no effect, suggesting a key role for heterotrophs like Thermus in high-temperature denitrification (Dodsworth et al. 2011b).
These studies have shed some light on denitrification pathways in Thermus and inferred an important role for them in heterotrophic denitrification; however, they only encompass a few species and a limited geographic range, and do not provide a comprehensive characterization of the denitrification phenotypes across the genus. A recent paper describes the presence and evolution of denitrification gene clusters in representatives of most species of Thermus, but did not examine phenotypes (Jiao et al. 2022). The goal of this study was to characterize denitrification phenotypes in a diversity of Thermus species to gain insight into their potential roles in the nitrogen cycle in high-temperature environments. To address this goal, 24 strains representing ten Thermus species from various geothermal areas in China were grown under denitrifying conditions and nitrogenous products were measured. Additionally, denitrification genes were recovered from available genomes and by PCR using Thermus-specific primers designed in this study. These experiments show incomplete denitrification pathways to be common across the genus, further supporting their purported role as important incomplete denitrifiers in terrestrial geothermal systems.

Isolation of Thermus strains
The sources of Thermus strains are described in Table 1. All strains were isolated by standard serial dilution plating technique onto R2A and T5 media (Ming et al. 2014) from geothermal spring sediments or mats or nearby geothermal soils in Yunnan Province, China, with the exception of T. arciformis JCM15153 T , a type strain isolated from Guangxi Autonomous Region, China, which was included to increase taxonomic coverage. The physical and chemical properties of most springs that were sources for isolation of Thermus strains have been described previously (Hedlund et al. 2012;Hou et al. 2013;Song et al. 2013).

Screen for nitrate reduction phenotype
Thermus strains were revived from frozen stocks on Castenholz Medium D (CMD) agar plates amended with 9 mM nitrate and supplemented with 0.1% yeast extract and 0.1% tryptone and adjusted to pH 8.2 (Castenholz 1969;Hedlund et al. 2011). Thermus strains were screened for nitrate reduction phenotype by testing for the ability to grow in anaerobic liquid CMD modified as described above, but with 4.5 mM nitrate. The anaerobic medium was sparged with N 2 for 45 min to remove oxygen and distributed into glass serum bottles, or Balch tubes with Durham vials, in an anaerobic chamber (Coy Type B, Coy Laboratory Products Inc., Grass Lake, MI, USA) containing an atmosphere of N 2 (~ 90%), CO 2 (~ 5%), and H 2 (~ 5%). The culture bottles were sealed with butyl rubber stoppers and aluminum crimps and the headspace was exchanged prior to autoclaving by 5 cycles of evacuation (30 s) and filling to 1 atm with 99.999% He. For all nitrate reduction experiments, a pure colony of each strain was suspended and grown in 10 mL of anaerobic medium in 25 mL glass serum bottles to serve as a starter culture. To dilute contaminating N 2 and ensure denitrification pathways were active, cells were grown to early stationary phase and passed using He-rinsed syringes with a 1:50 inoculum into pre-warmed medium twice before a final transfer into experimental bottles. 160 mL serum bottles containing 40 mL of liquid medium described above, or Balch tubes with Durham vials with 10 mL of liquid medium, were used for the final transfer. Serum bottles were incubated in the dark at 60 °C with rotary shaking at 100 rpm with serum bottles in a horizontal position to maximize gas equilibration. Balch tubes were incubated in the dark at 60 °C in a static incubator. Unless otherwise noted, cell density was measured using a Petroff-Hausser counting chamber on an Olympus BX-51 phase-contrast microscope with brightness and contrast optimized using PictureFrame software (Optronics, Goleta, CA, USA). Nitrate, nitrite, N 2 O, and N 2 were then assayed after 96 h using the methods described below. All strains were capable of nitrate reduction and robust growth in the medium used for this work. A few strains initially included in the study, T. brockianus YIM 77709, T. scotoductus YIM 77398, and T. tengchongensis YIM 77427, did not grow well under the denitrification conditions tested here and were subsequently dropped from the study, leaving 24 strains. T. calditerrae YIM 77777 failed to grow under denitrification conditions, consistent with the absence of annotated denitrification genes in its genome ). All strains were tested in triplicate and data presented are from replicates with a final cell concentration of ≥ 1 × 10 6 cells/mL.

Detection of terminal denitrification nitrogen products
To measure aqueous nitrate and nitrite concentrations, approximately 6 mL of liquid medium was sampled after a 96 h endpoint, filtered through a 0.2 μm filter (28145-501 VWR, Radnor, PA, USA), stored at 4 °C, and analyzed by colorimetric methods and confirmed using ion chromatography (IC). Nitrite concentrations were measured by diazotization with sulfanilamide, followed by coupling with N-(1-naphthyl)-ethylenediamine dihydrochloride (LaMotte, Chestertown, MD, USA). For nitrate measurements, powdered cadmium was used to reduce nitrate to nitrite prior to diazotization (LaMotte, Chestertown, MD, USA). To confirm colorimetric measurements, IC analysis was performed on samples from a subset of experiments as previously described ) using a Dionex DX-500 Chromatograph with an AS22 anion exchange column with a 4.5 mM Na 2 CO 3 /0.8 mM NaHCO 3 eluent. To measure N 2 O, headspace gas samples were collected from culture bottles at a 96 h endpoint for gas chromatography (C2014 Shimadzu GC) analysis. Headspace N 2 O concentration was measured by injecting a 2 mL headspace gas sample into a GC-2014 Nitrous Oxide Analyzer (Shimadzu, Moorpark, CA, USA) operated as described previously (Dodsworth et al. 2011a, b). To detect N 2 gas, cultures were screened for the ability produce N 2 gas in Durham vials in 25 mL Balch tubes after 96 h and by gas chromatography using a thermal conductivity detector (TCD) (Dodsworth et al. 2011a). Statistical significance for all measurements was calculated by comparing data to an uninoculated control using a Student's t test in R (p < 0.1).
DNA extraction, PCR amplification of 16S rRNA genes, and DNA sequencing DNA was extracted from Thermus cell pellets using the FastDNA Spin Kit for Soil (MP Biomedicals) according to the manufacturer's protocol. 16S rRNA genes were amplified with PCR using primers specific for bacteria: 9bF (Eder et al. 1999) and 1512uR (Eder and Huber 2002). The 25 μL PCR mixture contained 10-125 ng of DNA, 200 nM of each primer, 200 μM each dNTP (Promega, Madison, WI, USA), 1.5 mM MgCl 2 , 0.625 U of GoTaq DNA polymerase (Promega), and 1 × GoTaq buffer (Promega). Cycling conditions were as follows: denaturation at 95 °C for 4 min followed by 30 cycles of denaturation (30 s at 95 °C), annealing (1 min at 55 °C), and elongation (2 min at 72 °C), with a final elongation step (7 min at 72 °C). PCR products were sequenced using the Sanger method at Functional Biosciences, Madison, WI, using the forward and reverse PCR primer.

Design of PCR primers and amplification of nitrogen-oxide reductase genes
Conserved regions for narG, nirK, nirS, and norB in Thermus species were chosen to design the primers used in this study. Existing primers for nitrogen-oxide reductases (Braker et al. 1998;Phillippot et al. 2002;Throbäck et al. 2004) were shown or predicted to be ineffective with Thermus genes (data not shown). Thermus denitrification gene sequences were harvested from genomes (Gounder et al. 2011;Mefferd et al. 2016;Murugapiran et al. 2013;Zhou et al. 2016) available at the Joint Genome Institute's Integrated Microbial Genomes (IMG) website (Markowitz et al. 2014) and RAST (Aziz et al. 2008;Overbeek et al. 2014) and used for alignment. For each gene, the available sequences were aligned using default parameters using MUSCLE (Edgar 2004;Dereeper et al. 2008). Conserved regions used for primer design were chosen manually and were predicted to give PCR products of ~ 1000 bp and include conserved functional domains.
To optimize annealing temperatures for each primer set, gradient PCR amplification over a range of ± 5 °C from the mean of the calculated melting temperature for each primer set. Primer combinations and optimal annealing temperatures can be found in Table S2. The sequences and locations of the binding sites of the primers are shown in Table 1.
Hot-start PCR for amplification of narG, nirK, nirS, and norB was performed with DNA from each Thermus strain as template. The 25 μL PCR mixture contained 10-125 ng DNA, 200 nM of each primer, 1.5 mM MgCl 2 , 200 μM each dNTP, 0.625 U of GoTaq DNA Polymerase (Promega, Madison, WI), and 1 × Green GoTaq Reaction Buffer. Cycling conditions were as follows: denaturation at 95 °C for 4 min followed by 35 cycles of denaturation (2.5 min at 95 °C), annealing (1 min, see Table S2 for temperatures), and elongation (2.5 min at 72 °C), with a final elongation step at 72 °C for 7 min. PCR products were sequenced using the Sanger method at Functional Biosciences, Madison, WI, using the forward and reverse PCR primer.

Phylogenetic analysis and tree construction
To determine the evolutionary relationships among Thermus species, a phylogenetic tree was generated from the 16S rRNA gene sequences of the isolates. 16S rRNA genes of the new isolates and all type strains of the genus Thermus were aligned using the mothur-provided SILVA alignment (Quast et al. 2012) in mothur v.1.39.5 (Schloss et al. 2009). The alignment was checked manually and filtered using the mothur-provided, SILVA-compatible 1349-position Lane mask (Lane 1991). Phylogenetic trees were constructed using a maximum-likelihood algorithms within RAxML (Stamatakis et al. 2014) using the TIM3-F + 1 + G4 model, chosen based on the Bayesian Information Criterion (BIC). Neighbour-Joining and Maximum-parsimony trees were also implemented within MEGA 7.0 (Kumar et al. 2016). Branch support values were based on bootstrap analysis performed using 1000 replicates.

Nucleotide accession numbers
Nucleotide accession numbers for all gene fragments amplified by PCR are available in GenBank under the following accession numbers: 16S rRNA genes (ON429762-ON429782) and denitrification genes (ON456559-ON456600) are available in GenBank. Accession numbers for genome sequences are provided in Table 2.

Isolation and identification of a diverse collection of Thermus strains from geothermal springs in Yunnan, China
Thermus strains were isolated by serial dilution and plating from samples collected from sediments and mats from springs and heated soils at Rehai National Park, Gongxiaoshe Spring, Hehua Spring, and Nuijiea Ancient Hot Spring, all in Yunnan Province, China (Table 1). 16S rRNA gene sequencing and phylogenetic analysis revealed that the strains belonged to nine different species: Thermus amyloliquefaciens, Thermus antranikianii, Thermus brockianus, Thermus calditerrae, Thermus igniterrae, T. oshimai, T. thermophilus, T. scotoductus, and Thermus tengchongensis, spanning about half of the Thermus species with validly published names (Fig. 1) and most species known to be capable of nitrate reduction or denitrification.

Determination of nitrate reduction intermediates and terminal products
The newly isolated Thermus strains, plus Thermus arciformis JCM 15153 T , were screened for growth under anaerobic conditions with nitrate as the sole terminal electron acceptor. Almost all strains grew well in anaerobic media with nitrate as the sole terminal electron acceptor and were inferred denitrifiers since fermentation has not been described in Thermus (Albuquerque et al. 2018). T. caliditerrae YIM 77,777 did not reduce nitrate, as confirmed by the absence of denitrification genes in its genome , and was dropped from the study.
The 24 strains reduced nitrate but failed to produce N 2 . Terminal denitrification products varied among the isolates (Table 2); however, most species were coherent with regard to denitrification phenotype. Most strains (n = 11) produced N 2 O as the final denitrification product, including all strains of T. antranikianii (n = 3) and T. tengchongensis (n = 7) and the single strain of T. arciformis. By comparison, several strains only reduced nitrate to nitrite, including the single tested strains of T. caliditerrae, T. igniterrae, T. scotoductus, and T. thermophilus. However, consistent denitrification phenotypes were not observed in T. amyloliquefaciens, T. brockianus, or T. oshimai, where different strains terminated with different intermediates of denitrification (Table 2). In T. amyloliquefaciens, one strain reduced nitrate to N 2 O, whereas the other only reduced nitrate to nitrite. In three of the four T. brockianus strains, nitrate was removed during anaerobic growth, yet neither nitrite, N 2 O, nor N 2 were detected at the 96 h endpoint that was assayed. This phenotype points to NO as the most likely denitrification product in these strains, which was also supported by the detection of narGHI, nirS, and nirK genes in the genome of T. brockianus YIM 77927, which lacks norCBH genes (Fig. 3). A similar genotype and phenotype was observed for T. oshimai YIM 77923-2, and similar genotypes (nar and nir genes but lacking nor genes) have been observed previously in genomes from T. antranikianii DSM 12462 T , Thermus parrtaviensis RL T , Thermus thermamylovorans CFH72773 T , Caldithermus terrae DSM 26712 T , and Caldithermus chliarophilus DSM 9957 T (Jiao et al. 2022). These genomes also lack nrf systems that could support dissimilatory nitrite reduction to ammonium as an alternative pathway. The possible accumulation of NO as the main denitrification product by many Table 2 Summary of denitrification phenotype and genotype + Statistically significant decrease in substrate and/or increase in product compared to an uninoculated negative control (p ≤ 0.1) (+)Inferred nitrite reduction activity due to a significant decrease in nitrite or significant increase in N 2 O compared to negative control (p ≤ 0.1) [+]Incomplete denitrifier that produces ≥ 1 mM of N 2 O in at least two replicates in multiple assays − Phenotype not detected a Phenotype detected, but no genome was available and PCR for the corresponding gene was negative b Denitrification genes present, but phenotype not detected c Unless a genome accession is included, genes were amplified by PCR and sequenced To examine a nitrate reduction phenotype in more detail, T. arciformis JCM15153 T was selected for further analysis based on its consistent growth phenotype (Fig. 2). There was a long lag phase until exponential growth was detected after 36 h after which a near-stoichiometric conversion of nitrate to N 2 O was observed, with transient production of nitrite and no production of N 2 .

Presence and arrangement of denitrification gene clusters
Denitrification genes were recovered from newly sequenced genomes of T. arciformis JCM15153 T , T. brockianus YIM 77927, T. caliditerrae YIM 77925 T , and T. tengchongensis YIM 77924 T , and previously available genomes of T. amyloliquefaciens YIM 77409 T ) and Fig. 1 16S rRNA gene phylogeny. Maximum-likelihood phylogeny based on 16S rRNA gene sequences of all type strains of Thermus plus the isolates described in this manuscript (in bold). Shading at nodes depict bootstrap support using maximum-likelihood, neighbor-joining, and maximum parsimony Thermus tengchongensis YIM 77392 / ON429780

Meiothermus luteus YIM 72257 T / KY608080
Meiothermus rufus DSM1279 T / AUHY01000095 T. tengchongensis YIM 77401 ). The genome of T. caliditerrae YIM 77777 did not contain denitrification genes , consistent with its inability to grow in nitrate reduction experiments. The other six strains contained a complete nitrate reductase operon (narGHJI) and two nitrate/nitrite transporters (narK1 and narK2). Genes encoding nitrite reductase (nirS) and cytochrome c-dependent nitric oxide reductase (norBCH) were also found in proximity to the nitrate reductase operon, except for strain T. brockianus YIM 77927, which contained two nitrite reductases (nirK and nirS) and was missing nor genes. Genes encoding NorR, the MarR-family transcription factor conserved in Thermus denitrification gene clusters (Sánchez-Costa et al. 2020), was conserved and syntenic in all genomes. As noted previously (Murugapiran et al. 2013;Zhou et al. 2016;Mefferd et al. 2016), nitrous oxide reductase (nos) genes were absent in all Thermus genomes examined. Denitrification genes in these six genomes, plus T. scotoductus SA-01 (Gounder et al. 2011), T. thermophilus JL-18, and T. oshimai JL-2 (Murugapiran et al. 2013), were co-localized and spanned over 25 kbp (Fig. 3). The genes were generally syntenous, but with some rearrangements among norCBH, nirK, and nirS. The (narGHJIK(K2)) operons were entirely syntenous and consistently divergent from the two component regulatory system, dnrST (Alvarez et al. 2017). Genes encoding the drpAB nitrate sensor protein system (Chahlafi et al. 2018) were consistently downstream of the nar operon. Genes encoding nrcD, a ferredoxin associated with denitrification gene clusters in Thermus (Cava et al. 2004) were conserved and syntenous. The denitrification gene cluster was chromosomally encoded on each of the six newly sequenced genomes.
PCR was used to obtain denitrification genes from the other strains. Since existing primers (Braker et al. 1998;Dodsworth et al. 2011a;Phillippot et al. 2002;Throbäck et al. 2004) were shown or predicted to be ineffective for all Thermus denitrification genes, a few sets of novel primers were designed and optimized based on existing Thermus denitrification genes (Tables S1, S2). All PCR amplicons were most closely related to those from other Thermus species, with nucleotide sequence identities of 82-99%. Putative narG fragments were amplified from DNA extracts of all Thermus strains tested, except T. thermophilus YIM 77430-2, despite its ability to grow under nitrate-reducing conditions and produce nitrite as a terminal product ( Table 2). The presence of nir and nor genes was variable. Including genome-derived genes and PCR products, genes encoding the isoenzymes NirK (n = 9) and NirS (n = 14) were both common in Thermus genomes. Two strains of T. antranikianii appeared to have only nirK. Others appeared to have only nirS (n = 9), including most members of T. tengchongensis (n = 5), one strain each of T. brockianus and T. amyloliquefaciens, and the only strain of T. caliditerrae. Other species harbored both nirK and nirS, including all three strains of T. oshimai and three of four strains of T. brockianus. Including both genome-and PCR-derived data, norB was present in at least 14 of the genomes, including all 8 isolates of T. tengchongensis, both strains of T. amyloliquefaciens, two of three strains of T. oshimai and the only strains of T. arciformis, T. calditerrae, T. thermophilus, and T. scotoductus.
In general, denitrification genes appeared to be intact, but it should be noted that pseudogenes predicted to encode proteins with C-terminal truncations in NirS and NorB were annotated in T. caliditerrae YIM 77925 T . This genome is also missing norC, an essential periplasmic Nor subunit that shuttles electrons to the catalytic subunit NorB (Thorndycroft et al. 2007;Hino et al. 2010). Together, this is consistent with the observation that nitrite was the terminal product in T. caliditerrae YIM 77925 T cultures (Table 2). Interestingly, there is no indication that truncations in NirS and NorB include binding sites or amino acid residues of known functional importance (Rees et al. 1997;Watmough et al. 1999;Hemp and Gennis 2008).
The presence of denitrification genes in the Thermus strains studied here was generally consistent with the denitrification phenotypes; however, in several cases the detected phenotype and genotype did not agree (Table 2). nar genes were not found in T. thermophilus YIM 77430-2 using primers for this study, but nitrate reductase activity was detected. nir genes were not detected for T. antranikianii YIM 77311 and T. tengchongensis YIM 77392, though there was measured nitrite reductase activity. Finally, nor genes were not detected in any T. antranikianii strains or T. brockianus YIM 79134, though N 2 O was detected. In these cases, inconsistencies are likely the result of the limited sensitivity of the PCR primers used in this work. Future work will be needed to design primers with better coverage within the Thermales, perhaps using primer sequences similar to those described here but with additional degeneracies, or by incorporating alignments with denitrification gene sequences from other published genomes.

Discussion
Here, 24 strains representing nine different Thermus species isolated from diverse geothermal areas in China were shown to be capable of respiratory nitrate reduction or incomplete denitrification. Combined with previous studies of incomplete denitrification in Thermus from other locations, this work shows that incomplete denitrification pathways are common across species of the genus Thermus regardless of geographic location. Consistent with the incomplete denitrification phenotypes of other Thermus strains, nitrate reduction to either nitrite, NO, or N 2 O as terminal products indicates varying denitrification capabilities within Thermus, with most strains ending with N 2 O as a final product. These data support the inference that Thermus in geothermal areas may serve as a source of N 2 O, a strong greenhouse gas and stratospheric reactant, as has been measured in geothermal springs in the U.S. Great Basin (Hedlund et al. 2011). Incomplete denitrification pathways producing N 2 O may also promote N-cycling within hot spring systems because N 2 O is more soluble than N 2 and its reduction can be coupled to proton translocation and growth (Flock et al. 2005;Zumft and Kroneck 2006). In contrast, N 2 has a lower aqueous solubility and Annotated catalytic and structural proteins: nar nitrate reductase; nir nitrite reductase; nor nitric oxide reductase; nrcD ferredoxin associated with denitrification gene cluster in Thermus (Cava et al. 2004). Annotated regulatory proteins: dnr denitrification regulator; nsr NOdependent regulator of nir/nor (Alvarez et al., 2017); drp nitrate sensor proteins (Chahlafi et al. 2018); norR MarR-family transcription factor conserved in Thermus denitrification gene clusters (Sánchez-Costa et al. 2020). Other annotated genes: Feo ferrous iron transport system is relatively inert, especially in systems with sufficient dissolved inorganic nitrogen supply where selection for or expression of nitrogen fixation may be weak. However, single-cell genomes and metagenome-assembled genomes from members of the Caldarchaeales (synonym: 'Aigarchaeota') show several members of this often-abundant, but not isolated group to contain putative N 2 O reductase genes (Rinke et al. 2013;Hedlund et al. 2015b) and no other denitrification genes and are, therefore, possible metabolic partners of Thermus. Similarly, multiple species of Thermoflexus that often cohabit with Thermus have putative N 2 O reductase genes, yet this function was not demonstrated in cultures of Thermoflexus hugenholtzii (Thomas et al. 2021). A multispecies coupled denitrification pathway is consistent with high rates of complete denitrification in sediments in Great Boiling Spring, NV, inferred from much higher N 2 O fluxes in acetylene block experiments compared to no-acetylene controls (Dodsworth et al. 2011b;Hedlund et al. 2011). In contrast, similar rates of nitrous oxide production were obtained with and without acetylene blocks in heated soils adjacent to springs in Yellowstone National Park, inferring high rates of N 2 O production and low rates of complete denitrification (Burr et al. 2005). Whether other high-temperature geothermal systems support both N 2 O emissions and complete denitrification has yet to be addressed. It is similarly unknown whether any Thermus strains are capable of complete denitrification or of the terminal step of denitrification at all. At least one report inferred complete denitrification in T. thermophilus HB27 despite the absence of an annotated nitrous oxide reductase, based on missing nitrogen balance during denitrification experiments ), though we recommend this result should be followed up more rigorously, for example, using stable isotopes of nitrogen followed by mass spectrometry.
The synteny of denitrification gene clusters in the diverse Thermus isolates suggests these gene clusters evolve largely together, whether vertically or horizontally. Several of the Thermus strains studied here contain a nirK homolog (two T. antranikianii strains, three T. brockianus strains, and three T. oshimai strains) and were capable of nitrite reduction, inferring activity for the NO-forming, Cu-containing nitrite reductase in denitrification. Alternatively, or in addition, several strains contained homologs of the isofunctional tetraheme cytochrome cd1-containing nitrite reductase, nirS (T. amyloliquefaciens YIM 77409 T , T. arciformis JCM 15153 T , three T. brockianus strains, three T. oshimai strains, and five T. tengchongensis strains). Six of the strains belonging to T. brockianus and T. oshimai encoded both NirK and NirS homologs. NirK and NirS were previously considered incompatible (Zumft 1997) but several species of Thermus have been previously noted to encode both enzymes, including strains of T. scotoductus (Gounder et al. 2011), T. oshimai (Murugapiran et al. 2013) and T. antranikianii (Liu et al. 2020. The current study adds T. brockianus to that list. Studies in T. antranikianii DSM 12462 T (Liu et al., 2020) showed that both genes are co-transcribed during denitrification, with higher expression of nirK under a wider range of nitrite concentrations and only nirS expressed in the presence of oxygen. Thermus strains encoding both nitrite reductases have also been shown to lower the transient and final concentrations of nitrite during the denitrification with nitrate as the substrate, inferring more efficient use of available nitrite under both high and low substrate concentrations (Hedlund et al. 2011). The co-occurrence of these nitrite reductases in several species and the lack of nitrous oxide reductase across the genus appear to be unique features of the genus Thermus and further underscores the likely ecological importance of Thermus in denitrification.
This study did not address the distribution of Thermus genotypes and phenotypes across physicochemical space. Thus, although it is reasonable to expect physicochemical conditions and ecological relationships to constrain the presence and composition of denitrification pathways, it is currently unknown how Thermus denitrification pathways are distributed in physicochemical space in geothermal systems. Temperature influences both steps of the oxidative nitrogen cycle in terrestrial geothermal systems (De la Torre et al. 2008;Edwards et al. 2013;Dodsworth et al. 2011a;Spieck et al. 2020), which in turn influences the supply of oxidized nitrogen in systems sourced by ammonium, which is commonly the dominant form of dissolved inorganic nitrogen in geothermal systems (Holloway et al. 2011). Similarly, as these enzymes use cofactors containing different trace metals such as molybdenum, iron, and copper, the availability of these metals may constrain these organisms and pathways. For example, Nar and NirK use molybdenum and copper as cofactors, which are both less soluble in the presence of hydrogen sulfide (Anbar and Knoll 2002). Given the likely importance of Thermus to the high-temperature nitrogen cycle, the distribution of Thermus species and denitrification genes within the greater context of denitrification genotypes is ripe for further study and could be addressed using coupled metagenomics and physicochemical data.

Conclusions
Incomplete denitrification pathways are common in the genus Thermus. More specifically, Thermus denitrification is characterized by nitrate reduction to nitrite, NO, or N 2 O as terminal products, evident by the varying phenotypes and genotypes. Despite a few inconsistencies between detected reductase activity and amplified genes, the work done here significantly expands on the current state of knowledge of denitrification in Thermus and further suggests and important role for the genus in thermophilic denitrification in terrestrial geothermal environments.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.