Function analysis

Of the 2843 proteins, 65% were grouped to an orthologous cluster (COG), and specific and general functions could be assigned for 45% and 12% of the predicted proteome, respectively. Proteins with specific function belong to the functional categories metabolism (17%), transport and cellular processes (9%), genetic information processing (7%), environmental information processing (5%), and miscellaneous (7%) (see also Supplemental Table S2). For the remaining proteins, only general functions (12%), partly derived through gene-context analyses, or no functions (43%) were assigned. One-fifth of the N. pharaonis proteins have no homologs in other species (singletons). Among them are two probable Natronomonas-specific membrane complexes, each encoded by nine adjacent genes (NP2336A–NP2352A, NP5354A–NP5338A), which are highly similar and arranged in the same gene order. All 15 previously published N. pharaonis genes could be identified within the genome, although minor sequence variations were detected usually owing to strain differences. However, a partially sequenced protease from N. pharaonis, which is extremely similar to vertebrate chymotrypsin (Stan-Lotter et al. 1999), could not be found.

Central metabolism and transport

Metabolic enzymes comprise a large number of fatty-acid degradation genes, and the complete set of enzymes involved in biosynthetic pathways leading to amino acids and coenzymes. Thus, the chemo-organotrophic N. pharaonis, which is usually grown on media with amino acids as the carbon source, has a high degree of nutritional self-sufficiency. In agreement with this, we were able to simplify the synthetic medium for this species omitting all amino acids except leucine. Requirement of this amino acid might be caused by disruption of the isopropylmalate synthase gene (leuA_1, NP2206A) in the 5′-region. As H. salinarum, N. pharaonis is likely not capable of sugar utilization because of the lack of genes encoding key enzymes of glycolytic pathways. A gene cluster (NP4962A–NP4944A) encoding a set of probable anaerobic dehydrogenase subunits might indicate growth of N. pharaonis cells under anaerobic or micro-aerophilic conditions. Genes encoding mevalonate pathway enzymes are present, as are other genes required for the synthesis of membrane lipids and other components such as menaquinones (Soliman and Truper 1982) and retinal pigments (Seidel et al. 1995), which are derived from prenyl-precursors. A large set of predicted transporter genes was found, including homologs of transporters for metal ions such as iron, manganese, copper, and cobalt, which are scarce in a highly alkaline environment. A more extensive description of the predicted metabolism and transport of N. pharaonis is given as Supplemental material.

Nitrogen metabolism

Natronomonas grows under highly alkaline conditions in brines of pH around 11 (Soliman and Truper 1982; Tindall et al. 1984). These extreme pH conditions cause not only low availability of metal ions, but also reduced levels of ammonium ions. For the uptake of ammonium, which can be assimilated in the central metabolite glutamate, N. pharaonis possesses transporters for several exogenous nitrogen sources, ammonium (AmtB, NP0922A), nitrate/nitrite (NarK, NP4228A), and urea (ABC transporter Urt-ABCDE, NP1996A–NP2004A) (Fig. 1). Genes involved in nitrate reduction (narB_1, NP4226A; nirA_1, NP4224A) and urea conversion (ure cluster, NP2008A–NP2020A) to ammonium were found to be clustered with their respective transporter genes. The observed nitrate assimilation pathway has also been described for cyanobacteria (Hirasawa et al. 2004), and the enzymes involved show 34%–51% sequence identity to those present in N. pharaonis. It is likely that Natronomonas uses ferredoxin and not NADH as the electron donor for all three reductive conversions. This view is supported by the occurrence of conserved ferredoxin-binding residues within the N. pharaonis NirA protein (Hirasawa et al. 1998) and ferredoxin dependence of nitrate and nitrite reductases in the halophile Haloferax mediterranei (Martinez-Espinosa et al. 2001). Nine ferredoxins of four orthologous groups (COG0633, COG1141, COG1146, COG3411) are present in N. pharaonis, and ferredoxin appears to be the common proteinaceous electron carrier for functional N-assimilation as well as the conversion of 2-oxoacids and aldehydes.

Open in a separate window

Figure 1.

Schematic representation of the nitrogen metabolism of N. pharaonis. The figure illustrates proposed transport (green arrows) and metabolic processes (blue arrows) for nitrogen compounds and the associated gene clusters (nark–narB–nirA, urtABCDE, ureBACGDEF). Ammonia can be supplied by (1) direct uptake, (2) uptake and reduction of nitrate () via nitrite (), and (3) uptake and hydrolysis of urea. It is utilized by a two-step reductive conversion of 2-oxoglutarate (2-OG) to glutamate (Glu) involving ferredoxin (fdx). The two glnA paralogs, gltB, and amtB were found distributed over the genome.

While H. marismortui also possesses all necessary genes for urea conversion and nitrate assimilation but also for nitrate respiration, H. salinarum lacks these genes. Urea-cycle enzymes for the conversion of ornithine to arginine are encoded in all three halophilic strains. With respect to arginine degradation, halophiles adopted different strategies. H. marismortui splits arginine into ornithine and urea by arginase (EC 3.5.3.1; rrnAC0383, rrnAC0453), while H. salinarum uses plasmid-encoded enzymes of the arginine deiminase pathway for the fermentation of arginine. In this pathway, arginine is converted to ornithine and carbamoylphosphate, which is further degraded to CO₂ and NH₃ with concomitant ATP generation (Ruepp and Soppa 1996). N. pharaonis lacks both pathways for arginine utilization.

Respiratory chain

The available biochemical data on electron transport chain components for N. pharaonis (Scharf et al. 1997) and other respiratory archaea (Schafer et al. 1996), as well as the genomic data from Halolex (http://www.halolex.mpg.de) and STRING databases (von Mering et al. 2003), were used to generate an archaeal profile of the electron transport chain (Fig. 2A). The profile revealed a high degree of plasticity in the composition of the respiratory chain, often differing greatly from the “classical” five complex systems found in mitochondria. Notably, type, number, and composition of terminal oxidases vary widely in the archaeal domain of life. Furthermore, NADH is not oxidized by proton-pumping type I NADH dehydrogenase complexes because the NADH acceptor module (nuoEFG) is absent. In N. pharaonis, NADH dehydrogenation is likely to occur via the non-proton-pumping type II NADH dehydrogenase encoded by a homolog (NP3508A) of the ndh gene characterized in Acidianus ambivalens (Gomes et al. 2001). All subunits of the succinate dehydrogenase (sdhCDBA, NP4264A-NP4270A) are present, and menaquinone biosynthesis genes were found, supporting its proposed function as a mobile carrier (Scharf et al. 1997). However, no complex III subunits could be identified in the genome in spite of searching with the respective H. salinarum genes (petABC). The crenarchaeote A. ambivalens is able to channel electrons from complex I/II directly into a terminal quinole oxidase using a mobile quinol carrier (Gomes et al. 2001). None of the three terminal oxidases in N. pharaonis, though, shows sufficient sequence similarity to support the idea that one of them is a quinole oxidase. The N. pharaonis cytochrome ba₃-type oxidase (cba cluster, NP2960A–NP2968A) (Mattar and Engelhard 1997) interacts with the blue copper protein halocyanin (hcp_1, NP3954A), and the cbaD subunit of the orthologous complex in H. salinarum occurs as a fusion protein with halocyanin. The probable electron transport from halocyanin to the terminal oxidase complex in haloarchaea indicates a novel yet unknown type of complex III, which mediates the electron transfer step between menaquinone and halocyanin.

Open in a separate window

Figure 2.

Genomic and experimental studies of the N. pharaonis electron transport chain. (A) The electron transport chain profile for several respiratory archaea displays subunits (squares) of respiratory complexes (boxes I–IV). These subunits are often encoded adjacently in the archaeal genomes (straight connections) and can be fused to each other (curved connections). The proposed electron flow between respiratory complexes is indicated by arrows. Complexes, which have been characterized experimentally, are indicated in blue, and protein-sequenced subunits of isolated complexes are yellow. The profile for Sulfolobus acidocaldarius (Sa) was established by complementation with genetic data from the completely sequenced Sulfolobus solfataricus (Ss) (asterisks). Menaquinone (MQ), caldariellaquinone (CQ), halocyanin (Hcy), and sulfocyanin (Scy) are shown or predicted to function as mobile carriers in archaeal respiratory chains. Homologs to complex I subunits have been found in archaeal genomes, but genes encoding the NADH acceptor module are missing. A functional NADH-dehydrogenating complex I has been experimentally excluded for three species (red). Instead, it is replaced by NADH dehydrogenase type II, which is not capable of proton translocation. (B,C) Oxidative and photo-phosphorylation processes in N. pharaonis cells were investigated through measurements of ATP levels (upper curves, mean values with error bars from triplicates) and extracellular pH (lower curves, continuous recording at a rate of 10/sec) in the (B) absence or (C) presence of the protonophore CCCP (0.2 mM). The effects (ON: above curve arrow, OFF: below curve arrow) of light (hν, λ > 515 nm, 32 mW/cm²) and aeration (O₂) were determined. All experiments were performed at pH 8.1. Vertical scaling bars indicate ATP level and amount of proton uptake, whereas horizontal scaling bars indicate a 10-min time interval.

Through physiological experiments, we showed that N. pharaonis is able to eject protons during respiration (oxygen-induced acidification) with a subsequent increase in ATP levels (oxidative phosphorylation) (Fig. 2B,C). Since both effects are sensitive to the protonophore CCCP, N. pharaonis needs to possess a functional respiratory chain including a component responsible for proton pumping upon electron transfer (either one of the oxidases or the postulated complex III analog). Furthermore, CCCP did not prevent light-induced alkalinization² by the light-driven chloride pump halorhodopsin, but did prevent the subsequent increase in ATP levels. Thus, the N. pharaonis ATP synthase also operates with protons, and completes a full proton circuit. This finding shows that N. pharaonis does not replace protons with sodium ions as the coupling ion between respiratory chain and ATP synthase. In contrast, the alkaliphilic bacterium Bacillus halodurans FTU switches from proton to sodium energetics in case of alkaline conditions or the addition of a protonophore (Skulachev 1992). Induction of respiratory complexes and ATPases with altered ion-specificity permits alkaliphiles to cope with inverted proton gradients requiring no increase of ΔΨ and avoiding reduced ATP yield as well as the risk of membrane electric breakdown (Skulachev et al. 1999). However, dependent on the external pH, varying internal pH values were observed for alkaliphilic Bacillus strains, and the internal pH reaches values greater than pH 9 (Horikoshi 1999). For N. pharaonis we measured internal pH values up to 9.3. By accepting high pH values in the cytoplasm, the difference between intra- and extracellular pH remains moderate, and consequently protons are permissible as the coupling ion. It should be noted, that our data are not in agreement with the previous suggestion that chloride is used as an alternative coupling ion by N. pharaonis (Avetisyan et al. 1998).

Secretion and membrane anchoring

N. pharaonis encodes the same probable components of the Sec and Tat (twin-arginine) protein translocation pathways that were found in the Halobacterium genome (Pohlschroder et al. 2004). For the transport of archaeal flagellins, involvement of flaI and flaJ genes was proposed (Thomas et al. 2001), since those show similarity to components of type II/IV secretion systems and were described to influence flagella formation (Patenge et al. 2001). To estimate the contribution of different systems for protein secretion in N. pharaonis, the complete secretome was predicted as described in Methods (Fig. 3A; see also Supplemental Table S3). This and previous genome surveys for H. salinarum strain NRC-1 (Bolhuis 2002; Rose et al. 2002) found that the Tat system, which secretes folded proteins, might be extensively used in haloarchaea compared to nonhalophilic archaea. Haloarchaea probably use the Tat pathway not only for coenzyme-containing redox components like halocyanins (e.g., hcp_4, NP0050A) and thioredoxins (e.g., trx_1, NP3914A), but also for the export of non-redox proteins such as substrate-binding proteins of ABC transport systems (e.g., dppA_2, NP3578A) and chemotactic signal transduction (e.g., H. salinarum cosB, OE3476R). Thus, an adaptation to the high-salt and alkaline environment may involve the avoidance of protein folding in the extracellular space where chaperones are absent (Rose et al. 2002). However, there are also several proteins in Natronomonas that are likely cotranslationally delivered to and exported through the Sec system, among them three extracellular subtilisin-like proteases (NP1682A, NP2654A, NP4628A).

Open in a separate window

Figure 3.

Schematic representation of protein secretion, anchoring, and glycosylation in N. pharaonis. (A) Substrates of the Tat, Sec and flagellin-specific protein translocation systems (blue boxes) are cleaved by signal peptidases (flash signs), and partly remain C- or N-terminally anchored to the cell membrane. Secreted proteins are cleaved by signal peptidase type I (blue), whereas lipobox-containing proteins are cleaved by signal peptidase type II (orange) and N-terminally attached to a lipid anchor (orange box). Lipoproteins are frequently transported via the Tat pathway (substrate numbers indicated in light-blue arrows). For three lipobox-containing proteins, the export pathway remains as yet unassigned. Furthermore, six proteins are likely to be modified by a C-terminally attached lipid anchor (yellow box). After cleavage by membrane-bound preflagellin peptidase (green), the substrates of the flagellin-specific export pathway reveal an N-terminal hydrophobic stretch possibly involved in membrane retention. (B) Signal sequence and peptide repeat modules (indicated by colored boxes) for a representative gene cluster (white arrows) are presented diagrammatically in models of the cell surface proteins. A Thr-rich tetrapeptide repeat (red box in genes, red line in proteins), likely to be O-glycosylated, occurs in several cell surface proteins adjacent to the C-terminal or N-terminal lipid anchor. An Asn-Gln dipeptide repeat (gray box, oval) follows directly after the lipobox-containing Tat-related signal sequence (orange box) of several membrane components. Other indicated features are Sec-related signal sequences (blue box in genes) and N-glycosylation sites (red hexagons in proteins).

For the N. pharaonis halocyanin (hcp_1, NP3954A), attachment of a diphytanyl chain to an N-terminal cysteine residue has been shown previously (Mattar et al. 1994). Prokaryotic lipoproteins are cleaved by signal peptidase II upstream of a cysteine residue that is part of a conserved motif (lipobox), and subsequently modified by lipid-anchor attachment to the cysteine residue of the processed N-terminus (Hayashi and Wu 1990). N. pharaonis halocyanin exhibits a probable lipobox motif (LAGC), indicating signal peptidase II-like processing before the observed lipid attachment. However, a signal peptidase II homolog is absent as in other archaeal genomes, and an alternative signal peptidase II-like protease remains to be identified. The N. pharaonis genome encodes a large number of proteins (91) containing lipobox motifs as adapted from PROSITE (PS00013) (see Methods). Probable N-terminal-anchored lipoproteins comprise a third of the predicted secretome, and seem to be commonly translocated via the Tat pathway (85 proteins) (Fig. 3A). More than 20 transporter subunits, six of the halocyanin homologs, and many Natronomonas-specific proteins of unknown function are found among predicted lipoproteins. Halophilic archaea have a higher fraction of lipobox-containing proteins than nonhalophilic archaea, the highest fraction being found in N. pharaonis (N. pharaonis: 91 [3.20% of all proteins], H. marismortui: 116 [2.74%], H. salinarum: 49 [1.74%] vs. Archaeoglobus fulgidus: 16 [0.66%], Methanosarcina mazei: 23 [0.68%], Pyrococcus furiosus: 9 [0.42%]). Lipoproteins seem to be more common in bacteria (Bacillus subtilis: 57 [1.39%], Escherichia coli: 90 [2.12%], Corynebacterium glutamicum: 77 [2.57%]). The retention of 34% of secreted N. pharaonis proteins by lipid anchors may reflect a protection mechanism against alkaline extraction of proteins from the cell membrane, an effect commonly exploited to deplete membrane preparations of peripheral membrane proteins (Klein et al. 2004).

Interestingly, eight of the putative lipoproteins have Asn-Gly dipeptide repeats directly following the lipobox (Fig. 3B), among them two halocyanins (hcp_1, NP3954A; hcp_2, NP4744AA) and three substrate-binding proteins (dppA_1, NP0758A; livJ_1, NP4140A; sfuA, NP5000A). These proteins are likely involved in interactions with membrane protein complexes (respiratory complexes, ABC transporters); thus, the repeat regions (length up to 24 amino acids) might function as flexible hinges promoting protein interactions.

Apart from the putative lipobox-containing proteins, 12 proteins with probable cleavage sites (RGQ) for pre-flagellin-peptidase (flaK, NP1276A), as previously proposed for halobacterial flagellins (Thomas et al. 2001), were identified. These proteins include not only the three flagellins (NP2086A–NP2090A), but also proteins with unknown functions, most of them encoded adjacent to each other and to genes similar to components of type II/IV secretion systems.

Cell envelope

Thr-rich tetrapeptide repeats varying in length and amino acid composition were detected in nine Natronomonas-specific proteins, three of whose genes (NP4616A, NP4620A, NP4622A) clustered within one genomic region (Fig. 3B). Four of the repeat-containing proteins revealed high regional similarity to the N- and C-termini of halophilic cell surface glycoproteins (csg), which form regular S-layer cell envelopes. Within these similar regions, three topological features have been described for the H. salinarum Csg; an N-terminal signal sequence, a C-terminal lipid-anchor region (Kikuchi et al. 1999), and—directly in front of this membrane anchor—a pattern of O-glycosylated threonines (Lechner and Sumper 1987). In all N. pharaonis proteins with Thr-rich tetrapeptide repeats, secretion, lipid-retention, and glycosylation signals were present (Fig. 3B). Interestingly, the Thr-rich repeating units are located not only next to the C-terminal lipid anchor regions as in H. salinarum, but also adjacent to N-terminal lipid anchors. This finding suggests that the likely function of observed Thr-rich tetrapeptide repeat patterns is as a glycosylated spacing region that forms a periplasma-like reaction space between cell envelope and membrane. Because N. pharaonis encodes several proteins with Csg-like features, it might not possess a typical S-layer, but instead form a more complex cell envelope consisting of various glycoprotein species with distinct saccharides.

Gene prediction and annotation

For gene prediction, REGANOR (McHardy et al. 2004) from the annotation package GENDB (Meyer et al. 2003) was used, which integrates results from CRITICA (Badger and Olsen 1999) and GLIMMER (Delcher et al. 1999). In addition, sixframe translation (>100 codons) was performed. From the resulting raw set of 11,874 distinct ORFs, a set of 2843 validated genes was selected by the following procedure: Using BLAST (Altschul et al. 1997), predicted amino acid sequences were bidirectionally compared to the H. salinarum strain R1 ORF set (http://www.halolex.mpg.de) containing 1060 genes experimentally verified by proteomic analysis (Klein et al. 2004; Tebbe et al. 2004). This allowed identification of undetected small genes, the discrimination between real proteins and spurious ORFs, the improvement of start codon selection, and the initial assignment of protein function. The ORF set was also analyzed by BLAST against itself, and against the NR database. Overlapping ORFs were adjusted based on gene context as well as characteristic halophilic pI and amino acid distribution patterns. tRNAs and other RNAs were predicted using tRNAscan (Lowe and Eddy 1997) and BLAST against H. salinarum, respectively. In the final validated gene set, 90.9% of the start codons were ATG, 7.9% GTG, and the residual 1.2% are pseudogenes, for example, due to interruption by ISH elements.

The generated validated gene set was used to assess performance of the three gene predictors, REGANOR, CRITICA, and GLIMMER (trained by CRITICA ORF set for optimal results) (McHardy et al. 2004). For the chromosome, GLIMMER predicted 12.9% false positives (FP) and 2.4% false negatives (FN), while CRITICA predicted no FP and 9.3% FN ORFs. The REGANOR gene finder performed best with 1.2% FP and 4.2% FN. One-third of the GLIMMER starts were reassigned (31.3% of the genes were shortened, 6.5% extended), while 13.2% (13.9%) of the genes predicted by CRITICA (REGANOR) were shortened and 0.9% (1.1%) were extended.

Each N. pharaonis protein was assigned to a cluster of orthologous groups (COG) and to a functional category (Tatusov et al. 1997), with a minimal BLAST e-value of e^-05. BLAST results against H. salinarum and other databases were carefully evaluated for annotation of gene functions or descriptions. Intergene distances and configurations between gene pairs were analyzed, and genes <35 bases apart were considered to be cotranscribed. Regions were defined as GC-poor when the GC content of 30 adjacent coding regions was 5% below the replicon average.

Motif searches

The PROSITE pattern PS00013 (Hulo et al. 2004) was used to search for lipid attachment sites (lipobox) within the first 50 amino acids of predicted proteins of N. pharaonis and other prokaryotes. The lipobox is highly variable, resulting in a rather unspecific sequence motif, but with organism-specific preferences (e.g., LAGC in E. coli and N. pharaonis). We counted only those lipobox variations that occur frequently in a given species (i.e., at least three times) to reduce the number of false positives. Proteins with a predicted lipobox that contain an additional twin-arginine motif and proteins identified through predictions from TATFIND2.2 (Rose et al. 2002) were used to determine the total number of twin-arginine translocation pathway (Tat) substrates. The number of general secretion pathway (Sec) substrates results from SIGNALP3.0 (Nielsen et al. 1997), excluding specific Tat substrates and proteins with TMHMM-based transmembrane helix predictions past the first 50 residues (Krogh et al. 2001). Proteins with signal peptide predictions were searched for the proposed pre-flagellin peptidase cleavage site in haloalkaliphilic archaea (RGQ) (Thomas et al. 2001). N-/O-Glycosylation sites were predicted using NetNGlyc and NetOGlyc tools (http://www.cbs.dtu.dk/services/). NG-dipeptide repeats/T-rich tetrapeptide repeats ([VP][TE][ED]T) with 4/2 repeating units were detected by pattern searches within a sequence window of 30/50 amino acids.

ATP and pH measurements

N. pharaonis cells grown to late logarithmic phase in DSM medium 205 were harvested and resuspended in basal salt (medium without casamino acids) to result in an OD₆₀₀ of 4. Cells were kept under a continuous nitrogen flow in a thermostated (20°C) glass vessel. For illumination, a 100-W mercury lamp (HBO 100 W-2; Oriel) was used, fitted with a heat protection filter (Calflex 3000; Balzers), and a yellow cutoff filter (OG 515; Schott) resulting in an irradiance of 32 mW/cm². Air was flushed through the medium for oxygenation. The pH traces were recorded with a standard glass electrode. For ATP determination by a luciferin/luciferase assay, 0.1 mL of cells were lysed in 5 mL of ice-cold buffer (10 mM MgCl₂, 0.02% NaN₃, 0.1 mM EDTA, 25 mM HEPES at pH 7.5). Luminescence was measured (Lumac Biocounter) in triplicate by adding 0.1 mL of a 1:2 (w/w) mixture of D-Luciferin (0.1 mg/mL) and Photinus pyralis luciferase (0.2 mg/mL; Sigma-Aldrich) to 0.5 mL of lysed cells. The protonophore CCCP (carbonyl cyanide m-chloro-phenylhydrazone) was added to a final concentration of 0.2 mM.

For measurement of the internal pH, freshly harvested cells were disrupted by sonification after washing twice with unbuffered basal salt without citrate (Koch and Oesterhelt 2005). When grown between pH 9.0 and pH 9.5, the internal pH was similar to that of the medium (external/internal: 9.0/9.3 and 9.5/9.2).

We acknowledge Hermann Lederer, Markus Rampp, and Reinhard Tisma for their help with the MIGENAS system (http://www.migenas.mpg.de); Burkhard Linke and Folker Meyer for successful cooperation regarding GenDB (Meyer et al. 2003); Günter Raddatz and Stephan Schuster for their expert advise on using genome assembly and annotation tools; Douglas Griffith and Martin Engelhard for careful manuscript reading and critical discussions; and Jan Wolfertz for continuous maintenance and improvements of the HaloLex system. We further thank the reviewers for valuable advice on the manuscript.