A role for tungsten in the biology of Campylobacter jejuni: tungstate stimulates formate dehydrogenase activity and is transported via an ultra-high affinity ABC system distinct from the molybdate transporter

Summary

The food-borne pathogen Campylobacter jejuni possesses no known tungstoenzymes, yet encodes two ABC transporters (Cj0300–0303 and Cj1538–1540) homologous to bacterial molybdate (ModABC) uptake systems and the tungstate transporter (TupABC) of Eubacterium acidaminophilum respectively. The actual substrates and physiological role of these transporters were investigated. Tryptophan fluorescence spectroscopy and isothermal titration calorimetry of the purified periplasmic binding proteins of each system revealed that while Cj0303 is unable to discriminate between molybdate and tungstate (K_D values for both ligands of 4–8 nM), Cj1540 binds tungstate with a K_D of 1.0 ± 0.2 pM; 50 000-fold more tightly than molybdate. Induction-coupled plasma mass spectroscopy of single and double mutants showed that this large difference in affinity is reflected in a lower cellular tungsten content in a cj1540 (tupA) mutant compared with a cj0303c (modA) mutant. Surprisingly, formate dehydrogenase (FDH) activity was decreased ∼50% in the tupA strain, and supplementation of the growth medium with tungstate significantly increased FDH activity in the wild type, while inhibiting known molybdoenzymes. Our data suggest that C. jejuni possesses a specific, ultra-high affinity tungstate transporter that supplies tungsten for incorporation into FDH. Furthermore, possession of two MoeA paralogues may explain the formation of both molybdopterin and tungstopterin in this bacterium.

Introduction

The Gram-negative microaerophile Campylobacter jejuni is the most commonly isolated bacterial cause of human gastrointestinal infection in the world (Skirrow, 1994). It causes diseases ranging from self-limiting gastro-enteritis to serious systemic infections (Wassenaar and Blaser, 1999) and has also been implicated in the serious neurodegenerative complication, Guillain–Barré syndrome (Nachamkin, 2002). C. jejuni is a commensal in a wide range of animal hosts, but is particularly prevalent in the gut microbiota of many bird species, and ingestion of contaminated poultry is the most common route for human infection. The pathogenic mechanisms of C. jejuni after infection of the human intestinal tract are relatively poorly understood but include mucosal adherence, host cell invasion and toxin production (Hu and Kopecko, 2008).

Understanding the complexity of the factors important in avian and human colonization will require increased knowledge about the physiology of C. jejuni. The bacterium is microaerophilic and cannot grow under strictly anaerobic conditions (Sellars et al., 2002). The electron transport chains of C. jejuni have been deduced from both genome sequence information and experimental data (Smith et al., 2000; Sellars et al., 2002; Myers and Kelly, 2005a,b; Pittman et al., 2007; Kelly, 2008; Weingarten et al., 2008) and show remarkable complexity for a small genome pathogen. A variety of organic and inorganic electron donors, including formate (Myers and Kelly, 2005a; Weerakoon et al., 2009) and sulphite (Myers and Kelly, 2005a) can be utilized, and pathways to several alternative electron acceptors including fumarate, nitrate, nitrite, trimethylamine N-oxide (TMAO) and dimethylsulphoxide (DMSO) have been identified, which allow energy conservation and growth under severely oxygen-limited conditions (Sellars et al., 2002; Pittman et al., 2007). The formate dehydrogenase (FDH) is membrane-bound but periplasmic facing and the sulphite oxidase (sulphite: cytochrome c oxidoreductase), nitrate reductase (Nap-type rather than Nar-type) and TMAO/DMSO reductases in C. jejuni are all thought to be periplasmic molybdoenzymes exported via the twin-arginine translocase system (Myers and Kelly, 2005a,b; Kelly, 2008). In addition, cj0379 encodes a homologue of the E. coli YedY periplasmic molybdoenzyme (Loschi et al., 2004), of unknown function.

Molybdenum and tungsten are chemically analogous elements that are found in the environment as highly soluble oxoanions with almost identical co-ordination chemistry (Johnson et al., 1996). Molybdenum is relatively abundant in the environment in comparison with tungsten (Kletzin and Adams, 1996), and tungstoenzymes have been generally considered to be restricted to prokaryotic obligate anaerobes (Hille, 2002). With the exception of the nitrogenase family of molybdoenzymes, both metals are found in a mononuclear form incorporated into the same pterin cofactor (‘Moco’) in a diverse group of enzymes (Hille, 1996; Johnson et al., 1996; Zhang and Gladyshev, 2008). There are four families of Moco containing enzymes: aldehyde:ferredoxin oxidoreductase (AOR), dimethylsulphoxide reductase (DMSOR), sulphite oxidase and xanthine oxidase (Kisker et al., 1997). The DMSOR family is the most diverse and includes the only examples of molybdoenzymes that can also function with tungsten; FDHs (de Bok et al., 2003; Brondino et al., 2004) and DMSO/TMAO reductases (Buc et al., 1999; Stewart et al., 2000). The AOR family are exclusively tungstoenzymes (Johnson et al., 1996). With the exception of the few examples of FDHs and TMAO/DMSO reductases, molybdoenzymes from organisms grown in the presence of tungstate are either inactive and lack any metal or are tungsten substituted with little or no catalytic activity (Kletzin and Adams, 1996).

All of the strains of C. jejuni that have been sequenced encode homologues of ModABC, the molybdate ABC-type transporter first characterized in Escherichia coli (Rech et al., 1996). Purified E. coli ModA (the periplasmic solute binding protein) has been shown to bind molybdate and tungstate oxoanions in a 1:1 ratio, each with a K_D of ∼20 nM (Imperial et al., 1998). However, the C. jejuni genome sequences also show the presence of an additional ABC-type transporter (Cj1538–1540 in NCTC 11168) that is homologous to the TupABC tungstate uptake system first characterized in the anaerobe Eubacterium acidaminophilum (Makdessi et al., 2001; Andreesen and Makdessi, 2008). The binding of tungstate by E. acidaminophilum TupA is highly specific as it was not influenced by a 1000-fold molar excess of molybdate.

Campylobacter jejuni has no known tungstoenzymes and therefore no obvious requirement for tungsten, which makes the presence of a putative high-affinity tungstate transporter unusual and, given the inhibitory nature of tungsten to most molybdoenzymes, possibly even detrimental to the organism. In order to determine the actual physiological roles of the two potential ABC transporters for molybdate and tungstate, we have purified their respective solute-binding proteins and carried out a comparative thermodynamic investigation of ligand binding using isothermal titration calorimetry (ITC). In addition we have constructed and characterized single and double mutants in the cognate genes and determined in detail the effects on intracellular molybdenum and tungsten levels and on the activity of each of the assayable putative molybdoenzymes. The results show that the Cj1538–1540 system functions as a TupABC-like ultra-high affinity tungstate transporter, the likely role of which is to supply tungstate for incorporation into the periplasmic FDH.

Results

Overexpression and purification of Cj0303 and Cj1540

The cj0303c and cj1540 genes encode the putative ModA and TupA proteins, respectively, in the NCTC 11168 strain of C. jejuni. Initially, the full-length genes were PCR amplified and cloned into the pET21a expression vector such that the recombinant proteins were expressed with C-terminal six-histidine tags. The proteins were purified to homogeneity from cell-free extracts by nickel affinity and gel filtration chromatography. The N-terminal signal sequence of Cj0303 is predicted to contain both signal peptidase I and II recognition sites, suggesting that the protein may be lipid modified at a conserved cysteine in the lipobox sequence LFVSC. When the full-length Cj0303 protein was overexpressed and purified from E. coli, it was found to be highly susceptible to aggregation and precipitation. Expression of a truncated version of Cj0303 that lacked the N-terminal 19 amino acids, corresponding to cleavage at the signal peptidase I recognition site, resulted in a much greater proportion of soluble protein and the purified protein did not precipitate upon storage. Cj1540 is predicted to contain a simpler N-terminal signal sequence with a single signal peptidase I cleavage site. When this protein was overexpressed with its native N-terminal signal sequence, only the mature periplasmic form accumulated, which was found to be highly soluble and stable. Mass spectrometry confirmed the identities of the truncated Cj0303 and mature Cj1540 proteins after purification, and these were used in the experiments below.

Characterization of purified binding proteins using fluorescence spectroscopy

Changes in intrinsic tryptophan fluorescence of periplasmic binding proteins have been widely exploited as a useful reporter of ligand-induced conformational changes (Thomas et al., 2006). Cj0303 contains no tryptophans but 12 tyrosine residues whereas Cj1540 contains four tryptophans and seven tyrosines in the mature protein. Excitation at 280 nm resulted in an emission maximum at 340 nm for both proteins (data not shown). Addition of molybdate or tungstate (1 μM final concentration) to Cj0303 (0.1 μM) resulted in no changes in the fluorescence emission. This might be expected because the quantum yield of tyrosine is much lower than that of tryptophan and less sensitive to the protein environment. Thus, fluorescence spectroscopy cannot be used to determine ligand interactions with Cj0303. The same experiment repeated with 0.14 μM Cj1540, however, resulted in a ∼10% quench upon addition of 1 μM tungstate and ∼35% quench upon addition of 1 μM molybdate (data not shown). Titration of the fluorescence signal from Cj1540 with molybdate (Fig. 1A) revealed a tight binding interaction with K_D of ∼40 nM and an approximate 1:1 stoichiometry. Repetition of the experiments showed some variation in the stoichiometry of binding, although the relative size of the quench and the measured dissociation constants remained similar in all experiments (K_D of molybdate of 40 ± 0.8 nM, n = 3 repeats). The variation in stoichiometry was ascribed to a slow aggregation of the protein in the stock solution or the experiments, thus reducing the number of binding sites available. In contrast to molybdate, a direct titration of tungstate against Cj1540 could not be performed due to the small magnitude of the fluorescence quench observed upon tungstate binding. In order to determine a K_D value for tungstate, a binding competition strategy was adopted. In these experiments, titrations of Cj1540 with tungstate were repeated in the presence of differing concentrations of molybdate (10 μM, Fig. 1B; 50 μM, Fig. 1C; 100 μM; Fig. 1D). The data were fitted globally to a tight binding relationship (see Eq. 1, Experimental procedures), where the apparent dissociation constants for tungstate depend on the molybdate concentration and the previously determined K_D for molybdate of 40 nM, yielding an underlying K_D for tungstate of 8 ± 5 pM.

Figure 1. Characterization of ligand binding to Cj1540 using fluorescence spectroscopy. A. Direct titration of sodium molybdate against 140 nM Cj1540. B–D. Displacement titrations of sodium tungstate against 140 nM Cj1540 in the presence of either 10 μM (B), 50 μM (C) or 100 μM (D) sodium molybdate. In each titration, the quench in protein fluorescence was monitored (excitation 280 nm; emission 340 nm) as molybdate or tungstate were added in 10 nM increments until a ligand concentration of ∼3 times the protein concentration was reached. The data were fitted, and dissociation constants determined, as described in Experimental procedures using Eq. 1 (solid lines). The data shown are representative of three independent titrations that were used to calculate K_D values.

Isothermal titration calorimetry

The lack of fluorescence changes upon addition of ligands to Cj0303 required that another method be used to define the binding behaviour of this protein. ITC has the advantage that nearly all interactions give rise to a heat change, which can be monitored with a high-sensitivity calorimeter, and the binding enthalpy (ΔH_obs) and dissociation constant can be derived from the heats of binding. Example thermograms and isotherms for the binding of molybdate and tungstate to Cj0303 are shown in Fig. 2A and B respectively. The observed behaviour is consistent with an exothermic process at this temperature (37°C), with a single binding site model of binding. The binding behaviour is extremely similar with the two ligands, with closely comparable ΔH_obs and K_D values (Table 1). The tight binding nature of these isotherms means that the K_D values are not very well determined, but are within error of each other. It is thus clear from Fig. 2 and Table 1 that Cj0303 does not distinguish between molybdate and tungstate.

Figure 2. Isothermal titration calorimetry of ligand binding to Cj0303 and Cj1540. Purified Cj0303 (A, B) or Cj1540 (C, D) each at 10 μM in reaction buffer was titrated with sodium molybdate (A, C) or sodium tungstate (B, D) as described in Experimental procedures. Data were fit using the ORIGIN software associated with the calorimeter, and the resulting thermodynamic constants are listed in Table 1. Molar ratio refers to moles of injectant per mole of protein.

Table 1.  Thermodynamic data derived from ITC of Cj0303 and Cj1540 proteins.

Protein	Ligand	n	KA (M^−1)	ΔH (kJ mol^−1)	KD (app) (nM)
In each case 10 μM protein was used for the titrations. n = measured stoichiometry of binding.
Cj0303	WO4^2–	0.806 ± 0.002	2.3 × 10^8 ± 0.9 × 10^8	−29 ± 2	4 ± 2
Cj0303	MoO4^2–	0.718 ± 0.004	1.3 × 10^8 ± 0.8 × 10^8	−31.2 ± 0.3	8 ± 5
Cj1540	WO4^2–	0.851 ± 0.001	> 10^9	−46.8 ± 0.1	< 0.1
Cj1540	MoO4^2–	0.979 ± 0.008	2.2 × 10^7 ± 0.6 × 10^7	−16.8 ± 0.3	50 ± 10
Cj1540 + 1 mM molybdate	WO4^2–	0.726 ± 0.002	1.1 × 10^8 ± 0.2 × 10^8	−32.2 ± 0.1	9 ± 1
Cj1540 + 2 mM molybdate	WO4^2–	0.910 ± 0.001	4.0 × 10^7 ± 0.8 × 10^7	−29.9 ± 0.3	25 ± 5
Cj1540 + 3.5 mM molybdate	WO4^2–	0.631 ± 0.011	9 × 10^6 ± 3 × 10^6	−38.5 ± 0.8	105 ± 30
Cj1540 + 6 mM molybdate	WO4^2–	0.863 ± 0.005	1.5 × 10^7 ± 0.2 × 10^7	−30.3 ± 0.3	67 ± 7
Cj1540 + 10 mM molybdate	WO4^2–	0.681 ± 0.001	4.5 × 10^6 ± 0.9 × 10^6	−40.5 ± 0.8	220 ± 40

The ITC experiments were repeated with Cj1540, to confirm the results of the fluorescence titrations and to provide a comparison with the Cj0303 data. Direct titration of sodium molybdate against Cj1540 produced an exothermic binding isotherm with a K_D value of 50 ± 10 nM (Fig. 2C, Table 1), which is very comparable to the K_D value obtained by fluorescence, and significantly weaker than the binding of molybdate to Cj0303. The value of ΔH_obs (∼ 17 kJ mol⁻¹) is also significantly less favourable. In contrast, the binding of tungstate to Cj1540 is much more exothermic (Fig. 2D; Table 1), with ΔH_obs being increased to ∼47 kJ mol⁻¹ (Table 1). As in the fluorescence experiments, the dissociation constant is too low to be determined accurately. However, the large difference in ΔH_obs, allowed competition experiments to be performed, by repeating the tungstate titrations at a range of molybdate concentrations. Exothermic binding isotherms with single site binding were recorded at 1 mM (Fig. 3A), 3.5 mM (Fig. 3B) and 10 mM sodium molybdate (Fig. 3C), with ΔH_obs consistent with the difference between the heats recorded for each ligand. Tungstate binding clearly becomes progressively tighter, as the molybdate concentration decreases. The slope of the relationship between the apparent K_D for tungstate and a range of molybdate concentrations (Fig. 3D) indicates that tungstate binds 50 000 ± 7000-fold more tightly than molybdate to Cj1540, according to the relationship described by Eq. 2, which using the K_D established by ITC for molybdate (50 nM) yields a K_D of Cj1540 for tungstate of 1.0 ± 0.2 pM. This agrees well with the value obtained above by fluorescence titration. The evidence from the displacement titration and the exceptionally low K_D value for tungstate compared with molybdate indicates that unlike Cj0303, Cj1540 strongly differentiates between these ligands with a marked preference for tungstate.

Figure 3. Determination of the Cj1540 tungstate K_D by isothermal titration calorimetry in the presence of molybdate. Cj1540 (10 μM) pre-incubated with 1 mM (A), 3.5 mM (B) or 10 mM (C) sodium molybdate in reaction buffer, was titrated with tungstate as described in Experimental procedures. Data were fit using the ORIGIN software associated with the calorimeter and the apparent tungstate K_D at each molybdate concentration (K_D′) was determined (see Table 1). In D, these K_D′ values plus additional values determined at 2 mM and 6 mM molybdate are plotted against the molybdate concentration. Equation 2 indicates a Cj1540 K_D for tungstate of 1.0 ± 0.2 pM based on a K_D for molybdate of 50 nM. Molar ratio refers to moles of injectant per mole of protein.

Determination of cellular Mo and W concentrations in wild-type and mutant strains

In order to investigate the individual roles of the two transporters in cells of C. jejuni, the cj0303c and cj1540 genes were insertionally inactivated using chloramphenicol and kanamycin-resistance cassettes respectively. A double mutant strain in which both genes were insertionally inactivated was also created (cj1540::kan cj0303c::cat). The cassettes used carry their own promoter and were cloned in the same transcriptional orientation as the target genes in order to minimize polar effects. That this strategy was successful is shown by the complementation data described below.

Inductively coupled plasma mass spectrometry (ICP-MS) was used to analyse the cellular concentration of molybdenum and tungsten in each of the above strains (Fig. 4) and in the growth media. Müller–Hinton plus serine broth (MH-S) made with laboratory-distilled water was found to contain approximately 5 nM tungsten and 300 nM molybdenum. Cells of the wild-type and each mutant strain were grown to early stationary phase in this medium, and processed as described in Experimental procedures for ICP-MS. Wild-type cells contained approximately seven times more molybdenum than tungsten (125 pmol Mo mg cell protein⁻¹ compared with 18 pmol W mg cell protein⁻¹), but neither metal was detectable in cells of the double mutant, showing that the two transport systems account for all of the cellular uptake of molybdenum and tungsten (Fig. 4). cj1540::kan cells contained the same concentration of molybdenum on a per mg of cell protein basis as that of wild-type cells but they showed a ∼75% reduction in tungsten concentration. cj0303c::cat cells showed a ∼60% reduction in molybdenum compared with wild type and a 25% reduction in tungsten. These results show a clear distinction between cellular levels of molybdenum and tungsten in each transporter mutant strain and the data are consistent with the differing substrate specificity of the cognate binding proteins of the two transport systems noted above. Overall, the data from the fluorescence, ITC and ICP-MS experiments show that Cj0303 is a typical ModA protein and is a component of a ModABC transporter that can accumulate both molybdate and tungstate, while Cj1540 is a TupA-like protein that has a marked preference for tungstate and is a component of a TupABC-like tungstate transporter. The ModA and TupA designations for Cj0303 and Cj1540 will be used henceforward.

Figure 4. Cellular molybdenum (A) and tungsten (B) contents measured by ICP-MS. Cells grown for ∼18 h under standard microaerobic conditions in MH-S media were harvested, treated and analysed for Mo and W content by ICP-MS as described in Experimental procedures. The strains used were the NCTC 11168 wild-type and isogenic mutants constructed as described in Experimental procedures; cj1540::kan (tupA), cj0303c::cat (modA) and the double mutant cj0303c::cat cj1540::kan (modA tupA). The data represent averages and standard error of the mean of three biological replicate cultures subjected to ICP-MS.

Sodium tungstate has differential effects on C. jejuni molybdoenzyme activity

The existence of a distinct tungstate transporter in C. jejuni indicates a requirement for this metal in some aspect of cellular physiology, most likely as a component of the cofactor of one or more tungstoenzymes. Of the five predicted molybdoenzymes in C. jejuni NCTC 11168 (Kelly, 2008), four have known functions that can be assayed. These are FDH, TMAO/DMSO reductase, sulphite oxidase and nitrate reductase. The putative molybdoenzyme Cj0379 has an unknown function, although the crystal structure of the E. coli YedY homologue shows that it has a molybdopterin cofactor (Loschi et al., 2004). Sodium tungstate acts as a potent inhibitor of molybdoenzymes due to its misincorporation into the pterin cofactor that is essential for catalytic activity (Kletzin and Adams, 1996). The addition of 1 mM sodium tungstate to the growth medium had no effect on the microaerobic growth rate of wild-type C. jejuni cells (data not shown), but it affected the activity of each of the above enzymes in a different manner. Sulphite oxidase and nitrate reductase activity were both completely inhibited in wild-type cells grown in the presence of 1 mM sodium tungstate, TMAO reductase activity was reduced by ∼40%, whereas FDH activity was actually increased by ∼50% (Fig. 5A–D, middle columns). The activity of these enzymes in cells grown in the presence of an excess (1 mM) sodium molybdate was also determined (Fig. 5A–D, right hand columns). As might be expected given the presence of some molybdenum in the control growth media, the activities of TMAO reductase, nitrate reductase and sulphite oxidase were not affected. However, FDH activity was reduced by ∼65% (Fig. 5B). Taken together, these data suggest that the C. jejuni FDH functions either exclusively with tungsten, or with a preference for tungsten rather than molybdenum for catalytic activity. The only partial inhibition of the C. jejuni TMAO reductase by tungstate could suggest that it functions with either molybdenum or tungsten, similar to that seen in E. coli (Buc et al., 1999).

Figure 5. Effect of tungstate and molybdate on the activities of enzymes in wild-type C. jejuni cells. Cells were grown for ∼18 h under microaerobic conditions in MH-S media in the absence of any additions (control) or in the presence of 1 mM sodium tungstate or sodium molybdate. Cells were harvested and assayed as described in Experimental procedures. TMAO reductase (A), nitrate reductase (D) and formate dehydrogenase (B) were assayed spectrophotometrically using viologen dyes, while sulphite oxidase (C) was assayed as sulphite dependent oxygen uptake (Myers and Kelly, 2005a). The data represent the averages and standard error of the mean of three biological replicate cultures.

Evidence that the Tup system supplies tungsten for FDH

Molybdoenzyme activities were next assayed in each of the mutant strains described above, that were deficient in one or both of the Mod and Tup transport systems (Fig. 6). Activity in the modA tupA double mutant strain was zero or very low for all of the assayed enzymes. These data show that in the absence of either binding protein, molybdenum and tungsten are not taken into the cells and thus fail to be incorporated into any enzymes, consistent with the ICP-MS data above. The tupA strain displayed increased sulphite oxidase and TMAO reductase activities compared with wild type, a similar nitrate reductase activity and an FDH activity that was decreased by ∼50% (Fig. 6). In independent experiments, a very similar decrease in FDH activity was also seen when formate-dependent oxygen respiration in intact cells was measured using an oxygen electrode (117 ± 6 nmol⁻¹ O₂ min⁻¹ mg protein⁻¹ in the tupA strain compared with 310 ± 30 nmol⁻¹ O₂ min⁻¹ mg protein⁻¹ in the wild-type parent strain). In contrast, the modA mutant strain had similar sulphite oxidase, TMAO reductase and FDH activities to wild type but a moderately decreased (∼30%) nitrate reductase activity (Fig. 6A). These results again show that there is a degree of redundancy between the two transport systems, as measurable activities of several molybdoenzymes are present in both of the single mutants. Most significant, however, is the observation that FDH activity is markedly decreased in the tupA strain whereas all other molybdoenzymes have a similar or increased activity relative to wild type. This suggests that transport of tungstate by the Tup system serves to increase FDH activity at the expense of the other molybdoenzymes.

Figure 6. Activities of C. jejuni molybdoenzymes in wild-type, mutant and complemented strains. Cells were grown for ∼18 h under microaerobic conditions in MH-S media, harvested and assayed as described in Experimental procedures and the legend to Fig. 5. In the upper panel the activities of the indicated enzymes (A–D) in the wild-type and isogenic single and double mutants were measured. In the lower panel, the activities of nitrate reductase (E) and formate dehydrogenase (F) in the tupA modA double mutant alone were compared with activities in double mutant cells that had been complemented with either a chromosomal tupA⁺ or a modA⁺ allele. The data represent averages and standard error of the mean of three biological replicate cultures.

Complementation of the modA tupA double mutant strain with either of the wild-type genes was carried out, using the pRR plasmid integration system described by Karlyshev and Wren (2005). Nitrate reductase and FDH activity were assayed in these strains (Fig. 6E and F respectively) as indicators of restoration of transporter function. Nitrate reductase activity was restored in mutant strains complemented with either modA or tupA to ∼75% of that of wild-type cells. FDH activity was also restored in both strains, but to markedly differing extents. The FDH activity in the modA tupA/tupA⁺ merodiploid was restored to 100% of that seen in the wild type, whereas in the modA tupA/modA⁺ merodiploid, FDH activity was only ∼20% of that seen in the wild type. These data confirm the pattern of FDH activity seen in the individual modA and tupA mutant strains (Fig. 6D) and are consistent with a key role for the Tup system in the supply of tungsten for incorporation into FDH.

Growth characteristics of mutants and complemented strains

Under microaerobic conditions the modA, tupA and modA tupA mutants all grew at the same rate as wild-type cells and reached similar final cell densities in MH-S medium (data not shown). However, under oxygen-limited conditions, C. jejuni cells require an alternative terminal electron acceptor for significant growth to occur (Sellars et al., 2002). TMAO or nitrate-dependent growth under oxygen-limited conditions is conditional on a functional TMAO reductase (Sellars et al., 2002) or nitrate reductase (Pittman et al., 2007) respectively. When 20 mM TMAO was present as an alternative electron acceptor under oxygen-limited conditions (Fig. 7A), the growth of each mutant strain reflected the activity of TMAO reductase seen in Fig. 6B, with the tupA mutant showing the best growth, followed by wild type and then the modA mutant. The modA tupA double mutant hardly grew at all under these conditions (Fig. 7A). Growth experiments were also performed with the complemented strains described above. In the presence of 20 mM TMAO the wild-type and modA tupA/modA⁺ merodiploid strains grew best, whereas the modA tupA/tupA⁺ merodiploid grew only slightly better than its modA tupA mutant parent (Fig. 7C). These results are consistent with the ModABC system being the key transporter that supplies molybdenum for incorporation into TMAO reductase. When 20 mM sodium nitrate was present as electron acceptor no difference was seen between the growth of each mutant or complemented strain except for the double modA tupA mutant, which again showed insignificant growth (Fig. 7B and D respectively).

Figure 7. Oxygen-limited growth of wild-type, mutants and complemented strains with TMAO (A, C) or nitrate (B, D) as electron acceptor. In A and B, the growth of wild-type (filled triangles), tupA (open diamonds), modA (open squares) and tupA modA (open circles) strains is shown in BHI-FCS with 20 mM electron acceptor. In C and D, the growth of the wild-type and tupA modA mutant is compared with the complemented strains tupA modA/tupA⁺ (open diamonds) and tupA modA/modA⁺ (open squares). The cells were grown in filled conical flasks to generate oxygen-limited conditions as described previously (Sellars et al., 2002). No growth occurred in the absence of added electron acceptor.

Discussion

The properties of the transport systems characterized in this study show that C. jejuni possesses both a high-affinity ModABC type molybdate transporter and an extremely specific ultra-high affinity TupABC-like tungstate transport system capable of binding and transporting tungstate when present even at low picomolar concentrations. The presence in C. jejuni of a functional TupABC system is surprising, given that the current genome sequences have given no clues that this bacterium might have a tungsten requirement. Nevertheless, our data show that FDH activity is specifically enhanced by tungstate and is closely linked with the activity of the Tup system, implying that this enzyme is a novel tungstoenzyme in C. jejuni.

Molybdate and tungstate are extremely similar in size and geometry, and it is not entirely clear how a solute binding protein might distinguish between these oxoanions sufficiently so that a high specificity for tungstate transport can be achieved. The early work of Rech et al. (1996) and Imperial et al. (1998) showed that the E. coli ModA protein is unable to discriminate between molybdate and tungstate, as evidenced by similar K_D values. Our data with the C. jejuni ModA protein using ITC suggested K_Ds of 4–8 nM for both ligands, showing that this protein is also unable to distinguish between the two ligands, although it binds both of them with high affinity. Thus, given the much higher concentration of molybdate versus tungstate in most terrestrial environments, a ModABC type system alone is unlikely to be able to satisfy the tungsten requirement of a tungstoenzyme-containing prokaryote. In recent years, however, it has become clear that tungstate-specific ABC systems exist. The TupABC system was originally identified in the obligately anaerobic bacterium E. acidaminophilum (Makdessi et al., 2001), but homologues appear to be present in a range of bacteria (Andreesen and Makdessi, 2008). Apart from the current study, the E. acidaminophilum TupA protein is the only other ligand binding protein from this family to have been biochemically characterized. Makdessi et al. (2001) originally reported a K_D of 0.5 μM for tungstate using native polyacrylamide gel retardation assays. This crude method will be inaccurate, and much higher affinities for tungstate binding by this TupA protein have apparently been determined using fluorescence (K_D < 10 nM) and ITC (K_D of 0.2 nM; D. Rauh et al., unpublished data cited in Andreesen and Makdessi, 2008). The K_D value of 1 pM determined here for the interaction of the C. jejuni TupA and tungstate is the lowest value yet reported for any periplasmic molybdate/tungstate binding protein (Rech et al., 1996; Imperial et al., 1998; Makdessi et al., 2001; Bevers et al., 2006), and the measured binding enthalpies indicate that bond making in the interaction with tungstate is exceedingly good.

The crystal structures of the E. coli (Hu et al., 1997), Azotobacter vinelandii (Lawson et al., 1998) and Xanthomonas axonopodis pv. citri (Balan et al., 2008) ModA proteins have all shown that the ligand is bound in a tetracoordinate fashion with the oxygen atoms tetrahedrally arranged around the metal centre. However, recent crystallographic and EXAFS studies with a set of molybdate and tungstate binding proteins from archaea (the ModA/WtpA family) unequivocally show a hexacoordinate arrangement of tungstate with a distorted octahedral geometry (Hollenstein et al., 2007; 2009), suggesting distinct binding modes in the bacterial compared with archaeal proteins. It has been proposed that the addition of two extra oxygen ligands that are donated by acidic amino-acid side-chains of the archaeal proteins contribute to a vastly increased affinity (Hollenstein et al., 2009). The Pyrococcus furiosus WtpA protein has a K_D of 17 pM for tungstate and 11 nM for molybdate, as determined by ITC (Bevers et al., 2006). Nevertheless, the values reported here for the bacterial TupA protein from C. jejuni show an even higher affinity for tungstate and a larger difference with molybdate. Crystallographic definition of the mode of ligand binding in this bacterial protein is needed to explain how this achieved.

Clearly, extremely high transport affinities are essential for effective tungstate uptake against invariably higher environmental molybdate concentrations. The concentration of tungstate in fresh water is less than 100 nM, and tungstate in soils only accounts for 0.1–3.0 mg kg⁻¹ (Kletzin and Adams, 1996). C. jejuni normally lives as a commensal in the caecum of birds, particularly poultry. Relevant to understanding the function of the Mod and Tup systems is the relative concentration of molybdenum and tungsten in the avian caecum. Our own ICP-MS analysis recorded ∼14 nM tungsten in a sample of chicken caecal material, in contrast to ∼1 μM molybdenum in the same sample (J.P. Smart and D.J. Kelly, unpubl. data). Thus, C. jejuni clearly has access to both tungsten and molybdenum in its major niche at concentrations well above the respective K_D values for the cognate transporters. Analysis of cellular molybdenum and tungsten concentrations by ICP-MS in the transporter mutant strains shows that the identified Mod and Tup systems account for all of the uptake of molybdenum and tungsten in C. jejuni, and reveal a degree of redundancy between the two transport systems, as the cells are not totally starved of either molybdate or tungstate when the individual binding protein genes are inactivated. Nevertheless, the remarkable difference in affinity of the Tup system for tungstate and molybdate is reflected in the phenotype of the tupA mutant, which showed a large reduction in cellular tungsten but no change in molybdenum content. In contrast the modA mutant showed a large reduction in molybdenum and only a modest reduction in tungsten content. These data point to a differential physiological role for the Mod and Tup systems in supplying molybdenum and tungsten, respectively, in C. jejuni. Once transported, the initial fate of molybdate in other bacteria is by binding to a Mop-family protein with a ‘molbindin’ domain (Andreesen and Makdessi, 2008). The cj0302c gene, located within the mod operon in C. jejuni NCTC 11168, encodes a protein with homology to Mop proteins that may fulfil this role. An analogous gene is not present in the tup operon however, so the specificity of Cj0302 is unclear.

The existence of a high-affinity tungstate transporter in a small genome host-adapted pathogen like C. jejuni indicates a functional role for tungsten, and we hypothesized that one or more of the putative molybdoenzymes in this bacterium may in fact be a tungsten enzyme. Using the well-known inhibition of molybdoenzyme activity after growth with tungstate, we suggest that nitrate reductase and sulphite oxidase are typical tungsten-sensitive molybdoenzymes, while the TMAO reductase is clearly much less sensitive to inhibition. It has previously been reported that TMAO/DMSO reductases from several bacteria can retain catalytic activity with a tungstopterin cofactor (Buc et al., 1999; Stewart et al., 2000), albeit with altered kinetics, so the minimal tungsten sensitivity observed here may be a reflection of this. The only enzyme that was positively affected by tungstate and negatively affected by excess molybdate during growth was FDH (Fig. 5B), a pattern consistent with FDH being a tungstoenzyme. We sought further evidence for this by examination of the effect on FDH activity of the inactivation by mutation of either the Tup or the Mod transporters. The results clearly showed a specific reduction in FDH activity in the tupA mutant, while the activities of TMAO reductase and sulphite oxidase were actually enhanced. Moreover, wild-type levels of FDH activity in the modA tupA double mutant were only restored when a tupA⁺ allele (and not a modA⁺ allele) was supplied by complementation. Taken together, the data further support the view that the activity of FDH is tungsten dependent and provide evidence for a specific role of the Tup system in tungsten provision for FDH. The data also suggest a potential ‘trade-off’ in wild-type cells between the necessity for tungsten uptake via TupABC and a degree of inhibition of the molybdoenzymes TMAO reductase and sulphite oxidase.

Prokaryotic tungsten-dependent FDHs have been identified and characterized in E. acidaminophilum (Graentzdoerffer et al., 2003), Clostridium thermoaceticum (Yamamoto et al., 1983), the sulphate-reducing organisms Desulfovibrio gigas (Almendra et al., 1999) and Desulfovibrio alaskensis (Brondino et al., 2004), and the syntrophic propionate-oxidizing bacterium Syntrophobacter fumaroxidans (de Bok et al., 2003). A number of other organisms are likely to contain tungsto-FDHs, such as Acetobacterium woodii, Clostridium formicoaceticum and Clostridium acidiurici (Andreesen and Makdessi, 2008). The common characteristic of all these bacteria is that they are obligate anaerobes. Only a few aerobes seem to contain tungsten FDH enzymes, such as the α-proteobacterium Methylobacterium extorquens AM1 (Laukel et al., 2003) and the β-proteobacterium Ralstonia eutropha (Burgdorf et al., 2001).

Campylobacter jejuni is a microaerophile but has a number of metabolic characteristics found in obligate anaerobes, particularly the use of oxygen-sensitive enzymes such as the pyruvate and 2-oxoglutarate: acceptor oxidoreductases and an oxygen-labile serine dehydratase (Kelly, 2008). In autotrophic anaerobes, tungsto-FDHs often function in carbon assimilation via the reductive acetyl-CoA pathway as they more efficiently catalyse the reduction of carbon dioxide to formate compared with molybdenum-containing FDHs (Andreesen and Makdessi, 2008). However, there is no evidence that C. jejuni operates a reductive acetyl-CoA pathway, or that it could fix the carbon dioxide resulting from formate oxidation. On the contrary, the role of FDH in C. jejuni as a formate oxidising respiratory electron donor has been shown in previous studies (Myers and Kelly, 2005a; Weerakoon et al., 2009) and is consistent with its periplasmic facing localization. It has also been implicated in host colonization (Weerakoon et al., 2009), implying a key role for tungsten during growth in vivo. A detailed characterization of the purified enzyme will be needed to determine whether C. jejuni FDH functions exclusively with tungsten, why tungsten rather than molybdenum is favoured for formate oxidation by this enzyme, and whether the enzyme also works with molybdenum, as found with some FDHs such as D. alaskensis (Brondino et al., 2004).

A further important implication of our results is that the pterin biosynthesis pathway in C. jejuni must be branched, to allow for the synthesis of both molybdopterin and tungstopterin cofactors. MoeA (along with MogA) catalyses the final step of metal oxoanion ligation onto the dithiolene moiety of the pterin cofactor (Nichols and Rajagopalan, 2002), and most bacteria possess a single moeA gene. However, in C. jejuni there exist two moeA paralogues (moeA; cj0857c and moeA2; cj1519) that are 33% identical at the amino-acid level and it is possible that these could provide the necessary differential oxoanion specificity. Finally, our results do not exclude the possibility that additional tungstoenzymes exist in C. jejuni that have yet to be identified.

Experimental procedures

Bacterial strains, media and culture conditions

Cultures of C. jejuni strain NCTC 11168 were grown at 37°C in microaerobic conditions (5% [v/v] O₂, 10% [v/v] CO₂ and 85% [v/v] N₂) in a MACS-VA500 Incubator (Don Whitley Scientific Ltd, UK) on Columbia agar containing 5% (v/v) lysed horse blood and 10 μg ml⁻¹ each of amphotericin B and vancomycin. To select C. jejuni mutants, kanamycin or chloramphenicol were added to a final concentration of 30 μg ml⁻¹ and erythromycin to 15 μg ml⁻¹. Cells were subcultured onto fresh medium every 2–3 days to maintain actively dividing cells. Liquid cultures of C. jejuni for metal content and enzyme assays were grown in Müller–Hinton broth supplemented with 20 mM L-Serine (MH-S) under standard microaerobic conditions (gas concentrations as above) with 100 ml of medium contained in 250 ml conical flasks. Liquid cultures of C. jejuni for growth curves under oxygen-limited conditions were grown in Brain Heart Infusion (BHI) broth supplemented with 5% (v/v) foetal calf serum (BHI-FCS) at 37°C (gas concentrations as above) but diffusion of oxygen was severely restricted by using 500 ml medium contained in a 500 ml conical flask as described previously (Sellars et al., 2002). Electron acceptors were added from filter-sterilized stock solutions to a final concentration of 20 mM. Cultures were maintained in the MACS-VA500 Incubator with continuous orbital shaking at 200 r.p.m. Growth curves shown are representative single experiments, but all growth experiments were repeated at least once with similar results. Media for all growth experiments was pre-incubated in the microaerobic gas-atmosphere for 24 h at 37°C before inoculation with a microaerobically grown MH-S or BHI-FCS starter culture. Growth was monitored by measuring optical density at 600 nm in an Ultrospec 2000 spectrophotometer (Amersham Pharmacia Biotech, UK). E. coli DH5α, BL21(DE3) and Origami B strains were cultured in Luria–Bertani (LB) medium supplemented with appropriate antibiotics at 37°C.

Assay of enzyme activities

Methyl or benzyl viologen-linked assays were carried out with intact cells in a 1 ml assay volume using a Shimadzu UV-240PC recording spectrophotometer. The assay mixture contained 25 mM sodium phosphate buffer (pH 7), 1 mM methyl viologen (TMAO reductase) or 1 mM benzyl viologen (FDH and nitrate reductase) and 5 mM electron acceptor (sodium nitrate or TMAO) or 10 mM sodium formate in a screw-topped glass cuvette fitted with a silicone seal. For the reductase assays, cells were added to the buffer plus viologen mixture and aliquots of a freshly prepared sodium dithionite solution were injected into the cuvette until the absorbance at either 578 nm (benzyl viologen) or 585 nm (methyl viologen) stabilized at ∼2.5. The assay was started by the injection of an anaerobic solution of the appropriate electron acceptor and the absorbance decrease recorded. For FDH, the cell/buffer/viologen mixture was sparged for 10 min with oxygen-free nitrogen before addition of an anaerobic solution of sodium formate and measurement of the absorbance increase. Extinction coefficients used were 11 800 M⁻¹ cm⁻¹ at 585 nm (methyl viologen) and 8600 M⁻¹ cm⁻¹ at 578 nm (benzyl viologen). Cell protein was determined according to Markwell et al. (1978).

Measurement of respiration rates by oxygen uptake

Sulphite oxidase and formate respiration rates were determined as described previously (Myers and Kelly, 2005a), by measuring the change in dissolved oxygen concentration of cell suspensions in a Clark-type oxygen electrode linked to a chart recorder, calibrated using air-saturated 25 mM phosphate buffer (pH 7) (220 nmol dissolved O₂ ml⁻¹ at 37°C). A zero-oxygen baseline was determined by the addition of sodium dithionite. The cell suspension was maintained at 37°C and stirred at a constant rate. Substrates (0.5 mM sodium sulphite or 10 mM sodium formate) were added by injection through a fine central pore in the airtight plug. Rates were expressed in nmol O₂ utilized min⁻¹ mg cell protein⁻¹.

DNA isolation and manipulation

Plasmid DNA was isolated using Qiagen Miniprep kits (Qiagen, UK). C. jejuni genomic DNA was extracted using a MicroLysis kit (Web Scientific, UK). Standard techniques were employed for the cloning, transformation, preparation and restriction analysis of plasmid DNA from E. coli (Sambrook et al., 1989).

Overexpression and purification of proteins

A set of PCR primers were designed to amplify the entire coding region of the cj1540 gene and a truncated version of the cj0303c gene lacking the N-terminal signal sequence, for cloning into the pET-21a(+) vector (Merck Chemicals Ltd, UK) under the control of the isopropyl-β-D-thiogalactopyranoside (IPTG)-inducible T7 promoter. The primers used were cj0303-F (5′-TAG CGA ATT CCA TAT GCA AAA TTT AAG TAT TTT TGT AGC T-3′), cj0303-R (5′-CGA TGA ATT CCT CGA GTG GTG TGC TAA ATC CGT A-3′), cj1540-F (5′-TAG CGA ATT CCA TAT GAA AAA AAT CAT TTC TTT AGC C-3′) and cj1540-R (5′-CGA TGA ATT CCT CGA GGT CTT TTC TTG TTT TTG CAT C-3′). The genes were amplified from C. jejuni NCTC 11168 chromosomal DNA using Pwo DNA polymerase (Roche Applied Science, USA). These generated plasmids pJS0303 (Cj0303) and pJS1540 (Cj1540) that produced recombinant proteins with in-frame 6 × His tags at the C-termini. Automated DNA sequencing (Lark Technologies, UK) showed that the sequence of each of the cloned genes was correct. pJS1540 was transformed into E. coli BL21 (DE3) and pJS0303 was transformed into E. coli Origami B. Both strains were grown at 37°C in LB medium containing carbenicillin (100 μg ml⁻¹). At an OD 600 nm of 0.6, 1 mM IPTG was added, and the induced cells were then grown for a further 4 h at 25°C before harvesting by centrifugation (5000 g, 20 min, 4°C). Cell pellets containing overexpressed proteins were disrupted in a French press (1050 psi) and the soluble fraction isolated as the supernatant after centrifugation (15 000 g, 20 min, 4°C). Cell extracts were fractionated on a HisTrap FF crude 1 ml column (Pharmacia Biotech AB, Sweden) by affinity chromatography. The proteins were bound to the column in 25 mM phosphate buffer pH 7.4 containing 0.5 M NaCl and 20 mM imidazole and eluted from the resin by a linear gradient from 20 to 500 mM imidazole in the same buffer. The proteins were then dialysed overnight against 4 l 20 mM Tris-HCl pH 7 to remove salts, imidazole and any bound molybdate and tungstate oxoanions. Cj1540 and Cj0303 were judged to be pure by Coomassie blue staining on overloaded SDS-PAGE gels and their identities were confirmed by mass spectrometry performed by Dr A.J.G. Moir, Molecular Biology and Biotechnology Department, University of Sheffield.

Construction and complementation of C. jejuni mutants

For mutagenesis of cj0303c a chloramphenicol acetyltransferase (cat) cassette originating from C. coli (Wang and Taylor, 1990) was cloned into a unique AflII site in pJS0303 generating the plasmid pJS0303::cat. For mutagenesis of cj1540 the aphAIII (kan) kanamycin-resistance cassette from C. coli (Wang and Taylor, 1990; van Vliet et al., 1998) was cloned into a unique HindIII site in pJS1540 generating the plasmid pJS1540::kan. Transformation of C. jejuni NCTC 11168 with plasmids pJS0303::cat and pJS1540::kan was carried out by electroporation, and transformants were selected on CA plates supplemented with chloramphenicol or kanamycin at final concentrations of 30 μg ml⁻¹. Colonies were re-streaked onto CA plates and correct insertion of the antibiotic-resistance cassettes into the target genes was verified by extraction of chromosomal DNA by the MicroLYSIS kit (Web Scientific, UK) according to manufacturer's instructions. PCR using the gene-specific primers listed above confirmed allelic exchange by double cross-over, as demonstrated by an increase in amplicon size of approximately 0.9 or 1.4 kb for the chloramphenicol or kanamycin cassette insertions respectively. A cj0303c cj1540 (modA tupA) double mutant was constructed by electroporation of the C. jejuni cj0303c mutant with pJS1540::kan and selecting on CA plates containing both kanamycin and chloramphenicol.

For complementation of the mutations, complementation constructs were produced by cloning the cj1540 and cj0303c genes plus their upstream ribosome binding sites into the C. jejuni complementation vector pRR (Karlyshev and Wren, 2005) modified to include an erythromycin-resistance cassette, designated pRRE. The cj1540 gene was amplified by PCR using primers C-cj1540-F (5′- ATC GCA ATT GAA GGC TAG TAT CAA CTA T−3′) and C-cj1540-R (5′- GTC ACA ATT GAA AGT TAC AAT AAG TAA A−3′), while the cj0303c gene was amplified by PCR using primers C-cj0303c-F (5′– ATC GCA ATT GAA TTA CTC AAG AAC TCA T−3′) and C-cj0303c-R (5′– GTC ACA ATT GAA AGA TTT TAA TTC AAG C−3′). Both genes were cloned into the MfeI site of pRRE to produce plasmids pcj1540-COMP and pcj0303c-COMP. These were then transformed into the C. jejuni cj0303c cj1540 double mutant described above, by electroporation and selection on Columbia blood agar plates containing chloramphenicol, kanamycin and erythromycin. Confirmation of the insertion of the complementation cassette at one of the rRNA loci was carried out by PCR using primers C-cj1540-F or C-cj0303c-F and C-seq-R (5′-CTC TTG CAC ATT GCA GTC CTA C-3′).

Quantification of molybdenum and tungsten content by ICP-MS

Triplicate 50 ml early stationary phase cultures of C. jejuni NCTC 11168, cj1540::kan (tupA), cj0303c::cat (modA) and cj0303c::cat cj1540::kan (modA tupA) strains were harvested by centrifugation (10 000 g, 10 min, 4°C). Cell pellets were resuspended in 20 ml Milli-Q H₂O and the centrifugation and resuspension steps were repeated three times. Finally, the cell pellet was resuspended in 6 ml Milli-Q H₂O. A 5 ml sample of each suspension was analysed for molybdenum and tungsten by ICP-MS. This was performed using a Perkin-Elmer Elan 6100 instrument by Mr A.G. Cox, Centre for Analytical Sciences, Department of Chemistry, University of Sheffield. Results were normalized to cell protein using the Markwell et al. (1978) assay.

Analysis of binding proteins by fluorescence spectroscopy

Changes in the intrinsic UV fluorescence of Cj0303 and Cj1540 upon molybdate and tungstate binding were recorded at 37°C with a Cary Eclipse Fluorometer (Varian) (λ_ex 280 nm λ_em 300–400 nm using 5 nm excitation and 20 nm emission slit widths) in 20 mM Tris-HCl pH 7. Titration of intrinsic fluorescence of Cj0303 and Cj1540 with molybdate was performed at 37°C in 20 mM Tris-HCl, pH 7 at λ_ex 280 nm λ_em 340 nm with 5 nm excitation and 20 nm emission slit widths. Spectra were corrected for dilution effects. The data from competition experiments were fit globally using in-house Levenberg-Marquadt non-linear least squares routines, to the tight binding equation:

• (1)

where the apparent dissociation constant (K_D′) is defined as

• (2)

and [L]_TOT and [P]_TOT indicate the total ligand and protein concentrations at any particular point in the titration. F, F_bound and F_free are the observed, fully saturated and free fluorescence signals of the protein respectively. Global fitting allowed explicit propagation of errors. The direct titrations were fitted to Eq. 1 where K_D′ equals K_D(MoO4²⁻₎.

Isothermal titration calorimetry

Isothermal titration calorimetry experiments were performed using a VP-ITC calorimeter (MicroCal USA, now GE Healthcare). Prior to experiments, protein was dialysed extensively against the reaction buffer (20 mM Tris-HCl pH 7.0 made with Milli-Q dH₂O). Binding protein (10 μM) was equilibrated in reaction buffer at 37°C in the cell of the calorimeter and subsequently, 18 injections of 15 μl of a 100 μM sodium tungstate or molybdate solution were performed and the heat response recorded. After subtraction of the baseline, the integrated heats were fitted to the single binding site model using the ORIGIN software package supplied with the calorimeter. For competition experiments, the reaction buffer was supplemented with the stated concentrations of molybdate prior to the injections with sodium tungstate. The relationship between apparent dissociation constants and the underlying constants is as in Eq. 2.

Acknowledgements

This work was supported by a UK Biotechnology and Biological Sciences Research Council DTA studentship award to J.P.S. We thank Ed Guccione and Chat Phansopa for help with preparation of the Figures.