Duplication of Genes in an ATP-binding Cassette Transport System Increases Dynamic Range While Maintaining Ligand Specificity

Background: The functional role of gene duplicates in a bacterial metabolite transport system is unknown.

Results: Duplicated genes have similar ligand specificities, but affinities differ by three orders of magnitude.

Conclusion: This expands the ability to respond to nutrient stress and maximizes transport of preferentially metabolized substrates.

Significance: Functionalization of gene duplicates in transport pathways allows bacteria to adapt to varying nutrient flux.

Abstract

Many bacteria exist in a state of feast or famine where high nutrient availability leads to periods of growth followed by nutrient scarcity and growth stagnation. To adapt to the constantly changing nutrient flux, metabolite acquisition systems must be able to function over a broad range. This, however, creates difficulties as nutrient concentrations vary over many orders of magnitude, requiring metabolite acquisition systems to simultaneously balance ligand specificity and the dynamic range in which a response to a metabolite is elicited. Here we present how a gene duplication of a periplasmic binding protein in a mannose ATP-binding cassette transport system potentially resolves this dilemma through gene functionalization. Determination of ligand binding affinities and specificities of the gene duplicates with fluorescence and circular dichroism demonstrates that although the binding specificity is maintained the K_d values for the same ligand differ over three orders of magnitude. These results suggest that this metabolite acquisition system can transport ligand at both low and high environmental concentrations while preventing saturation with related and less preferentially metabolized compounds. The x-ray crystal structures of the β-mannose-bound proteins help clarify the structural basis of gene functionalization and reveal that affinity and specificity are potentially encoded in different regions of the binding site. These studies suggest a possible functional role and adaptive advantage for the presence of two periplasmic-binding proteins in ATP-binding cassette transport systems and a way bacteria can adapt to varying nutrient flux through functionalization of gene duplicates.

Introduction

The microbial cellular milieu is a complex environment with many chemically related compounds, some of which are preferentially metabolized. For proteins that acquire environmentally derived metabolites, several factors must be simultaneously balanced to optimally bind ligands and to efficiently deal with this chemical complexity. It is a known phenomenon in biological systems that the recognition of ligands is not necessarily optimized to achieve the strongest interaction (1). Rather protein ligand affinities are often tuned to match physiological ligand concentrations and fluctuations. However, many bacteria exist in a state of feast or famine with nutrient availability varying over many orders of magnitude, complicating the ability to tune the dynamic range of affinity to the locally available metabolite concentration (2). Ligand-binding proteins need to be able to discriminate among related and preferentially metabolized solutes either by increasing the affinity for cognate ligands or through narrowing the ligand specificity (range of chemically similar ligands that can be bound). Binding cognate ligands very tightly allows for binding at low environmental nutrient levels, but at higher concentrations, proteins will be saturated with chemically similar and potentially less preferentially metabolized compounds.

ABC² transport systems are the primary metabolite acquisition systems in bacteria. These systems are typically found in operons containing either a homodimeric or a heterodimeric transmembrane transporter and a cytoplasmically located protein that couples the energy from ATP hydrolysis to ligand transport (3). Most ABC transport operons code for a single periplasmic solute-binding protein (PBP) that binds and delivers metabolites to the inner membrane-located transporters (4, 5). As the primary gatekeepers of metabolite transport, PBPs would ideally be able to operate at both high and low levels of nutrient availability. However, this is not the case as PBPs typically bind ligands in the low micromolar range (6). It might also be expected that the K_m of the transmembrane ABC transporters would be attuned to the PBP K_d, but in fact, PBP K_d values are typically lower than transporter K_m values and can differ more than a 1000-fold (7). This has led to the hypothesis that PBPs are the specificity-determining unit of ABC transport systems (4, 5).

Microorganisms have evolved differing strategies to deal with the potentially broad nutrient availability dynamic range. One way this can occur is through functionalization of gene duplicates. Gaining an adaptive advantage through functionalization allows for retention of the original gene function, whereas the new copy evolves new activity through gain of function mutations (8). Several mycobacteria have duplicate phosphate transport systems that function at either low or high levels of environmental phosphate and are controlled through differential transcriptional regulation of operons (9, 10). In another variation of phosphate uptake, Rhizobium tropici contains duplicate ABC phosphate transport operons that have transport K_m values that are known to differ by almost two orders of magnitude (11). In both of these systems, gene duplication appears to confer an adaptive advantage by allowing for simultaneous optimization of affinity and specificity, although the molecular mechanisms by which this is achieved have yet to be determined.

Here we show how the Thermotoga maritima ABC mannose transport system appears to code for yet another variation that permits a broad dynamic range of ligand binding affinity while at the same time maintaining ligand specificity (12). This gene cluster contains single copies of most typical ABC transport components but has two copies of mannose-binding PBPs with similar ligand specificities and two sugar-binding transcriptional regulators (13–15) (Fig. 1). The presence of two mannose-binding proteins with overlapping specificities in the same transport system is counterintuitive as bacteria minimize, diverge, or discard genes with duplicate functions (8). In the only other well studied case where a PBP is duplicated in an ABC transport system, the histidine transport operon of Escherichia coli, the two proteins (HisJ/ArgT) functionally diverged from one another and have differing specificities (16). By contrast, our biophysical characterization of the T. maritima PBP gene products demonstrate that although the amino acid sequence identity in the ligand binding site is 48%, remarkably the two proteins retain an almost identical ligand specificity, but they have an ∼1000-fold difference in K_d. This large difference in affinities permits a PBP-mediated response to ligand that ranges from nanomolar to low millimolar, effectively doubling the functional dynamic range of this transport system if only a single PBP was present. As these proteins are semispecific for their ligands, this is also a mechanism to minimize saturation of the PBPs by binding to less preferentially metabolized metabolites. These studies demonstrate how functionalization of duplicated PBPs provides an adaptive advantage by greatly increasing the dynamic range of ligand binding affinities and further demonstrate the remarkable adaptability of the PBP superfamily.

FIGURE 1.

T. maritima mannose utilization gene cluster. Although all of the genes involved in mannose metabolism in T. maritima are unknown, a gene cluster surrounding an ABC transport system contains all of the necessary components for ligand breakdown and transport. Polysaccharides are broken down into oligosaccharides by an extracellular β-mannase (TM1227). Two PBPs (TM1223/tmMnBP3 and TM1226/tmMnBP6) bind oligosaccharides in the periplasmic space. The T. maritima mannose ABC transporter is composed of a heterodimeric transmembrane transporter (TM1221 and TM1222) in addition to two ATP-hydrolyzing subunits (TM1219 and TM1220). Binding of ligand by the PBPs stimulates an interaction with the ABC transporter, which couples ATP hydrolysis to transmembrane ligand transport. This gene cluster also encodes two potential transcriptional regulators (TM1218 and TM1224) that may be involved in controlling expression levels of these gene products.

EXPERIMENTAL PROCEDURES

Protein Expression and Purification

Target genes from T. maritima (tm1223/tmMnBP3 and tm1226/tmMnBP6) were cloned in NdeI/XhoI sites of a pETDuet vector (Novagen). The predicted N-terminal signal sequence was removed, and a C-terminal His₆ tag was attached for affinity purification. Cells were grown in Terrific Broth medium at 37 °C for 16–18 h following induction with 1 mm isopropyl 1-thio-β-d-galactopyranoside. For ligand-free protein, cells were grown in Enfor's minimal medium. The cell pellet was resuspended in 20 mm imidazole, 300 mm NaCl and sonicated. The lysate was heated at 85 °C for 15 min and clarified by centrifugation at 35,000 × g. The supernatant was loaded onto chelating resin charged with NiSO₄, washed with resuspension buffer, and eluted in resuspension buffer supplemented with 300 mm imidazole. Eluted samples were loaded onto a Superdex75 26/60 gel filtration column (GE Healthcare). Fractions containing protein were pooled, concentrated, and dialyzed against 20 mm Tris (pH 8.0), 40 mm NaCl.

Thermal Denaturation by Circular Dichroism

Thermal melting profiles of the tmMnBP3 and tmMnBP6 in the presence of ligands were measured using a Jasco circular dichroism (CD) spectrophotometer. In all CD experiments, 2 m guanidinium chloride was used to bring temperature-induced denaturation within the measurable range of the instrument. CD signal was measured at 225 nm. Protein concentration was kept at 1 μm, and ligand concentrations were kept at either 33 μm (tmMnBP3) or 1 mm (tmMnBP6) throughout the experiment. Mannobiose, cellobiose, laminaribiose, xylobiose, galactobiose, mannopentaose, cellopentaose, xylopentaose, laminaripentaose, and mannohexaose were utilized in CD experiments (Fig. 2). To determine T_m values, the curves were fit to a two-state model that accounts for native and denatured baseline slopes (17, 18).

FIGURE 2.

Carbohydrates used in this study. Carbohydrates are colored as in Fig. 4 and Table 2.

Fluorescence Spectroscopy

Cysteine mutations were introduced in the plasmids containing tmMnBP3/tmMnBP6 genes by site-directed mutagenesis. Proteins were expressed and purified as described above. Mutant proteins were incubated with thiol-reactive acrylodan at 37 °C for 16–18 h for labeling and clarified by passing over Superdex 75 equilibrated with 20 mm Tris (pH 8.0), 150 mm NaCl to remove unbound fluorophore.

Changes in the fluorescence signal induced by ligand binding were measured by using a Flurolog-3 fluorometer. Fluorescence emission scans were collected using a λ_ex of 389 nm and a λ_em of 520 nm at 25 °C. Protein concentrations were kept at 5 nm for tighter binding ligands and 80 nm for weaker binding ligands. Binding isotherms were fit to a quadratic velocity equation.

Isothermal Calorimetry Titrations

A TA Instruments Nano ITC was used for isothermal calorimetric titrations of tmMnBP6 with cellobiose at 25 °C. For these experiments, 1 ml of tmMnBP6 at a concentration of 30 μm in 20 mm Tris (pH 8.0), 150 mm NaCl was titrated with 5-μl injections of a 600 μm solution of cellobiose in a matched buffer. After correction for heats of injections, integrated heat effects were analyzed using a single site binding model to determine the association constant, number of binding sites, and enthalpy of binding (19).

Crystallization, Structure Determination, Model Building, and Refinement

Mannobiose and mannohexaose (Megazyme) were added to a final concentration of 1 mm. tmMnBP3 crystals were obtained in hanging drops containing 2 μl of protein (20 mg/ml) mixed with 2 μl of 0.1 m Mg(NO₃)₂, 20–30% PEG 3350 at 20 °C. tmMnBP6 crystals were obtained by batch method containing 2 μl of protein (20 mg/ml) mixed with 2 μl of 2.2 m AmSO₄, 0.2 m trilithium citrate covered with paraffin oil. Data were collected at 100 K using a Rigaku 007HFmicromax x-ray generator.

The diffraction data were scaled and indexed using HKL3000 (20). The data collection statistics are listed in Table 1. Initial phases were derived by molecular replacement using Phaser (21). The crystal structure of the apo form of tmMnBP3 (Protein Data Bank code 1VR5) was used as the initial model. Manual model building was carried out in Coot (22), and refinement was performed in REFMAC5 (21) and PHENIX (23). The models exhibit good stereochemistry as validated by MolProbity (24); final refinement statistics are listed in Table 1. In all four structures, 99.8% of residues are in the allowed Ramachandran region. Atomic coordinates and structure factors have been deposited in the Protein Data Bank (25) under the accession codes 4PFT, 4PFY, 4PFU, and 4PFW for the tmMnBP3 mannobiose, tmMnBP3 mannohexaose, tmMnBP6 mannobiose, and tmMnBP6 mannohexaose complexes, respectively.

TABLE 1.

Data collection and refinement statistics

	tmMnBP3		tmMnBP6
	Mannobiose	Mannohexaose	Mannobiose	Mannohexaose
Data collection
Resolution range (Å)	50.0–1.75	50.0–1.5	50.0–2.05	50.0–2.2
Unique reflections	103,164	163,023	96,558	76,389
Redundancya	3.4 (3.2)	3.5 (3.3)	5.7 (5.3)	3.9 (3.6)
Mean I/σa	15.8 (2.4)	15.7 (2.5)	18.2 (3.5)	15.7 (2.4)
Completeness (%)a	99.8 (99.6)	99.9 (99.4)	98.5 (96.1)	97.2 (97.2)
Rmerged (%)a	7.4 (58.0)	5.3 (60.4)	6.2 (58.4)	6.6 (56.8)
Unit cell
Dimensions (Å)	a = 96.5, b = 53.5, c = 106.3	a = 96.6, b = 53.6, c = 106.3	a = 100.9, b = 193.9, c = 159.5	a = 99.5, b = 194.2, c = 159.4
Space group	P21	P21	C2221	C2221
Angles (°)	β = 110.5	β = 110.5
Refinement
No. of reflections (working set/test set)	97,961/5,162	146,635/8,181	91,671/4,824	72,446/3,843
Rcryst/Rfree (%)	15.4/19.9	16.3/18.7	17.2/20.8	16.3/21.5
Non-hydrogen atoms in refinement
Protein	8,864	8,891	9,024	9,036
Water	903	1,147	741	449
Metal ion	2	2	2	2
Carbohydrate	46	134	46	134
r.m.s.d.b from ideal
Bond lengths (Å)	0.013	0.009	0.008	0.008
Bond angles (°)	1.4	1.2	1.1	1.1
B-factors (Å2)
Protein (Chain A, B)	20.6, 20.7	20.6, 20.7	40.5, 36.5	50.3, 45.0
Ligand	12.9	12.9	27.7	53.1
Metal ion	30.4	30.4	55.9	63.7
Solvent	32.8	32.8	49.1	53.6

^a Numbers in parentheses represent values in the highest resolution shell.

^b Root mean square deviation.

RESULTS

Sequence Analysis of tmMnBP Duplications

ABC transport operons typically contain a single copy of a PBP. The T. maritima MSB8 mannose transport gene cluster contains two genes encoding PBPs, tm1223 (tmMnBP3) and tm1226 (tmMnBP6) (13, 14) (Fig. 1). We were interested in whether the presence of two MnBPs was restricted to T. maritima MSB8 or whether they were also present in closely related species. A BLAST search identified close homologs (>90% amino acid identity) of tmMnBP3 in six Thermotogales members, whereas tmMnBP6 was found in only four (26). A single tmMnBP3 homolog was found in Thermotoga petrophila (99% identity) and Thermotoga neapolitana (94% identity to tmMnBP3) (Fig. 3). In all cases where tmMnBP6 homologs were found, a tmMnBP3 homolog was found in close proximity in the genome. These organisms were T. maritima MSB8, Thermotoga naphthophila, Thermotoga sp. RQ2, and Thermotoga sp. EMP (Fig. 3). The close genomic proximity and sequence similarity to the tmMnBPs suggest that other closely related Thermotoga species also utilize two PBPs. In this sampling, the presence of tmMnBP6 is always coupled to tmMnBP3 in the genome, whereas the converse is not true, suggesting that tmMnBP6 duplicated from tmMnBP3.

FIGURE 3.

Phylogenetic tree of close tmMnBP homologs. The two tmMnBP3 homologs (lower branch) that do not have a tmMnBP6 homolog in the genomic sequence are highlighted in red. The number of substitutions per site and the locus tag are indicated.

Specificity of tmMnBP: Circular Dichroism

Previous characterization of tmMnBP3 and tmMnBP6 proteins demonstrated that they bind to mannose with essentially an identical K_d and have a similar ligand specificity profile (15). Unless this was a recent gene duplication event, these results are counterintuitive as nature tends to minimize genes with duplicate functions (12). We therefore sought to further test the ligand binding specificity and affinities of these gene products by monitoring ligand-induced shifts in the thermal melting midpoints (T_m) with CD.

The T_m values of the tmMnBP3 and tmMnBP6 apoproteins differ significantly in 2 m guanidinium chloride with the 11.5 °C difference suggesting that tmMnBP3 is more thermally stable than tmMnBP6 (Table 2 and Fig. 4). The T_m values of the apoproteins were used to compare the T_m values determined in the presence of a total of 10 carbohydrates including mannobiose, cellobiose, laminaribiose, xylobiose, galactobiose, mannopentaose, cellopentaose, xylopentaose, laminaripentaose, and mannohexaose (Figs. 2 and 4 and Table 2). To bring the data into a measurable range, different concentrations of sugar were used for tmMnBP3 (33 μm) and tmMnBP6 (1 mm), suggesting a substantial difference in affinity for these ligands (Fig. 4). Although the CD experiments were carried out under different concentrations of sugar, the apparent specificity of these two proteins is largely conserved (cellopentaose > mannopentaose/mannohexaose > cellobiose > mannobiose) for the ligands that produce the largest shift in T_m (Table 2).

TABLE 2.

Protein T_m and K_d values

^a Colored spheres correspond to the plots in Fig. 2.

^b Underlined K_d values are believed to be stoichiometric and represent an upper K_d value estimate.

^c Estimated.

FIGURE 4.

Carbohydrate specificities of tmMnBP3 and tmMnBP6. Thermal melting studies of tmMnBP3 (A) and tmMnBP6 (B) in 2 m guanidinium chloride in the presence of 33 μm and 1 mm sugars, respectively, are shown. Apoprotein, light blue; galactobiose, dark blue; xylopentaose, dark pink; xylobiose, light pink; laminaribiose, light green; laminaripentaose, dark green; cellobiose, light gray; cellopentaose, dark gray; mannobiose, orange; mannopentaose, red; and mannohexaose, dark red. Solid lines represent a fit to a two-state thermal denaturation model. Fluorescence titrations of D453C tmMnBP3 (C) and tmMnBP6 (D) mutants labeled with cysteine-reactive acrylodan in the presence of 10 different sugars are shown. Titration curves are colored as in A. The inset shows the fluorescence spectra of the apo and the fully saturated forms of the proteins. deg, degrees.

Overall Three-dimensional Structure of tmMnBP3 and tmMnBP6

We sought to determine the potential structural determinants that lead to the similar ligand specificities but different affinities of tmMnBP3 and tmMnBP6. X-ray crystallography was used to determine the molecular structures of tmMnBP3 and tmMnBP6 bound to the β(1,4)-linked disaccharide mannobiose and hexasaccharide mannohexaose (Fig. 5). The structures of the mannobiose-bound tmMnBP3 and tmMnBP6 were determined to a resolution of 1.75 and 2.05 Å and refined to an R_cryst/R_free of 15.4/19.9 and 17.2/20.8%, respectively (Table 1). The structures of the mannobiose and mannohexaose-bound tmMnBP3 and tmMnBP6 were determined to a resolution of 1.5 and 2.2 Å and refined to an R_cryst/R_free of 16.3/18.7 and 16.3/21.5%, respectively (Table 1). All structures were solved by molecular replacement with the initial structure solved using the individual domains of the unpublished apo form of tmMnBP3 (Protein Data Bank code 1VR5). Unless otherwise noted, all subsequent discussions will focus on the description of the mannohexaose-bound structure as the conformation of the mannobiose and the protein-ligand interactions are essentially identical to the first two sugar rings of mannohexaose (Fig. 5).

FIGURE 5.

Crystal structures of the mannobiose and mannohexaose-bound tmMnBP3 and tmMnBP6. Overall structures of mannobiose-bound tmMnBP3 (A), mannohexaose-bound tmMnBP3 (C), mannobiose-bound tmMnBP6 (E), and mannohexaose-bound tmMnBP6 (G) are shown. The numbering of the sugar ring position used in the text is indicated in C. Mannohexaose ligand is colored with yellow carbons. The putative magnesium (green sphere) binding site is highlighted with a red circle. Close-up views of the mannobiose-bound tmMnBP3 (B), mannohexaose-bound tmMnBP3 (D), mannobiose-bound tmMnBP6 (F), and mannohexaose-bound tmMnBP6 (H) binding sites are shown. Mannohexaose ligands are in ball and stick representation with yellow carbons, water molecules that directly bind to the mannohexaose are shown as red spheres, and water molecules that are found in the adjacent cavity are shown as dark red spheres. Direct hydrogen bonds and specifically bound water molecule-mediated hydrogen bonds are represented as black dashed lines. Hydrogen bonds between the protein and specifically bound water molecules are shown as red dashed lines.

The overall structures of tmMnBP3 and tmMnBP6 are very similar to one another with a backbone atom Cα root mean square deviation of 0.75 Å. Interestingly, the amino acids comprising the surface of the proteins that potentially interact with the transmembrane transporters are also highly conserved (Fig. 6) (27, 28). Like other PBPs with the oligopeptide binding fold, the tmMnBP proteins are composed of three domains with domains I and II forming the N-terminal half of the ligand binding site and domain III forming the C-terminal half. The ordering of β-strands in domains I and III places tmMnBP3 and tmMnBP6 in the Group II PBP subfamily (29) or in the C cluster based on a more recent, updated classification of PBPs (30). In both of these classification systems, the closest structural homolog, the T. maritima cellobiose-binding protein (tmCBP) is also found. Like tmCBP, the minimal disaccharide binding site of tmMnBP3 and tmMnBP6 is located at one extreme of the interdomain interface next to a large solvent-filled cavity (Fig. 5). The four additional sugar rings of the mannohexaose fill the remainder of the interface, displacing water molecules in the adjacent cavity of tmMnBP3 and tmMnBP6 (Fig. 5). The positioning of the terminal sugar ring suggests that an additional sugar ring could be specifically bound in the tmMnBP3 and tmMnBP6 binding sites. The steric restrictions of the tmCBP binding site impose an upper limit of binding to pentasaccharides (29). Another difference among the tmMnBPs and tmCBP is the presence of an endogenously bound metal in both tmMnBPs. The metal binding site is a canonical DXDXDG coordination site in a loop encompassing residues 362–370 in the C-terminal domain (Fig. 5) (31). The 2.1-Å average bond lengths of the hexacoordinate geometry are consistent with the identity of the metal being Mg²⁺ (32). However, the biological significance of the metal, if any, is unknown at present.

FIGURE 6.

Conservation of the tmMnBP surface residues. Two views of tmMnBP3 separated by a 180° rotation are shown. The surface representation of the protein is colored based on the conservation of amino acids with tmMnBP6. Identical amino acids are colored light gray, conserved amino acids are colored dark gray, and non-conserved amino acids are colored red. Based on the known interactions of PBP with ABC transporters, the potential interaction region is circled to highlight the conservation at this interface between tmMnBP3 and tmMnBP6.

tmMnBP3 and tmMnBP6 Oligosaccharide Binding Site

Close examination of the ligand binding sites in the mannohexaose-bound tmMnBP3 and tmMnBP6 structures reveals an extensive interaction network between the sugars and amino acid residues from both the N- and C-terminal halves of the protein (Fig. 5). Almost all of the ligand atoms that can form hydrogen bonds have at least a single hydrogen bond to either the protein, specifically bound water molecules, or bulk water molecules (33). The largest interaction surface and the number of hydrogens bonds are formed with the first and sixth sugar rings (Table 3). In both cases, the fourth and fifth sugar rings form the fewest interactions with the protein. Only a single direct hydrogen bond is formed with the fourth and fifth sugar rings in tmMnBP3, whereas no hydrogen bond is found in tmMnBP6 (Table 3). The second sugar ring has a differential hydrogen bonding pattern in tmMnBP3 and tmMnBP6. In tmMnBP6, all are water-mediated, whereas in tmMnBP3, one is a direct hydrogen bond with the protein, and the other three are water-mediated.

TABLE 3.

Interactions of tmMnBPs with carbohydrates

Sugar ring	Buried surface area (Å2)		Direct hydrogen bonds		Water-mediated hydrogen bonds
	tmMnBP3	tmMnBP6	tmMnBP3	tmMnBP6	tmMnBP3	tmMnBP6
1	238	233	8	6	1	0
2	195	190	1	0	3	3
3	187	188	1	3	1	2
4	177	175	1	0	0	0
5	170	162	1	0	0	0
6	246	232	4	4	1	2

Specificity of tmMnBP: Fluorescence

The circular dichroism characterization of tmMnBP3 and tmMnBP6 produced a qualitative measure of ligand binding affinities and specificities. We sought to quantitatively determine the ligand dissociation constants (K_d) and specificities of tmMnBP3 and tmMnBP6 by labeling these proteins with an environmentally sensitive fluorophore that reports on ligand binding. Based on the x-ray crystal structures, a peristerically located D453C mutation was chosen (6). Labeling tmMnBP3 and tmMnBP6 with acrylodan at this position led to a 6-nm red shift in the emission maximum coupled to a 50 and 40% increase in emission intensity upon addition of saturating amounts of ligand, respectively (Fig. 4). The acrylodan-labeled D453C tmMnBP3 and tmMnBP6 proteins were used for subsequent K_d determinations. For all titrations, an attempt was made to utilize a concentration of labeled protein below the K_d; however, a lower limit of ∼5 nm labeled protein was imposed because of fluorophore extinction coefficient/quantum yield and fluorometer sensitivity. Although we are interested in the relative differences among K_d values for the different carbohydrates, the localization of this labeling site is in the vicinity of the ligand binding site and may perturb the actual K_d. An isothermal calorimetry titration was used to determine an actual K_d value for binding of cellobiose to a wild-type tmMnBP6, resulting in determination of a ∼6-fold weaker K_d value than determined with fluorescence titrations (Fig. 7). Based on the high sequence and structural homology of tmMnBP3 and tmMnBP6, it is anticipated this difference is consistent between them.

FIGURE 7.

Isothermal calorimetric titration of tmMnBP6 with cellobiose. A, raw isothermal calorimetric data of tmMnBP6 titrated with cellobiose. B, the integrated intensities of the raw data (solid circles) were fit to an independent binding site model (solid line) to obtain binding parameters. W, watts.

Stoichiometric binding of ligands to tmMnBP3 was observed for cellopentaose, cellobiose, mannobiose, laminaribiose, mannopentaose, and mannohexaose at a protein concentration of 5 nm, and the determined K_d values represent an upper maximum value (Table 2 and Fig. 4). Additionally stoichiometric binding was also observed for binding of cellopentaose, mannopentaose, and mannohexaose to tmMnBP6 at a protein concentration of 5 nm (Table 2 and Fig. 4). The determined K_d values largely match the trend observed in the circular dichroism experiments with cellopentaose, mannohexaose, mannopentaose, mannobiose, and cellobiose exhibiting the tightest binding to tmMnBP3 and tmMnBP6. Among the weaker binding ligands, direct comparisons of tmMnBP3 and tmMnBP6 K_d values can be made among xylobiose, xylopentaose, and laminaripentaose, which range from a 4- to a 435-fold difference in K_d (Table 2). Interestingly with laminaribiose for which only stoichiometric binding to tmMnBP3 was observed the K_d differs from that of tmMnBP6 by 2867-fold. Combined with the trends in the CD data, we postulate that this ∼1000-fold difference in K_d continues for the ligands that bind the tightest to these proteins.

Modulating Affinity in tmMnBP

The biophysical analysis of the tmMnBPs presented here allows for the identification of the structural elements that are potentially involved in the reduction of the ligand binding affinity of tmMnBP6 relative to tmMnBP3. tmMnBP3 and tmMnBP6 have an overall amino acid identity of 61%. However, if only the 46 amino acids comprising the binding site are compared, the identity is reduced to 48% (Fig. 8). Overall, the combined effect of amino acid substitutions and small differences in the localization of ligands in the tmMnBP binding pockets leads to a net loss of three direct hydrogen bonds, three water-mediated hydrogen bonds, and 32 Ų of buried surface area in tmMnBP6 compared with tmMnBP3 (Table 3) (34). The largest differences in recognition occur in the first and second sugar ring subsites. This disaccharide binding site has been postulated to be where specific recognition of the sugars occurs (29). In the tmMnBPs, a total of 17 amino acids can potentially interact with the first two sugar rings (Fig. 8). Of the 17 positions, nine residues are identical, another three are conservative substitutions, and five are non-conservative. The non-conservative substitution of N512G and W515A in tmMnBP6 results in a loss of loss of 5 Ų of buried surface area and two direct hydrogen bonds to the first sugar ring, one of which is replaced by a hydrogen bond to a directly bound water molecule (Table 3). One direct hydrogen bond is formed with the second sugar ring in tmMnBP3, and the substitution of Asn-433 for Ala-433 results in the replacement of this direct hydrogen with a water-mediated hydrogen bond and a loss of 5 Ų of buried surface area. Interestingly, the third sugar ring forms two more direct hydrogen bonds and one more water-mediated hydrogen bond in tmMnBP6 than tmMnBP3. In the fourth and fifth sugar ring subsites of tmMnBP3, two direct hydrogen bonds are formed along with a gain of 4 Ų of buried surface area. Although only an additional water-mediated hydrogen bond is found in the sixth sugar ring subsite of tmMnBP3, a slight difference in the localization of the ligand leads to an additional 14 Ų of buried surface area compared with tmMnBP6. In addition to any potential entropic effects due to differential displacement of water molecules upon binding of ligands, we postulate that the losses of buried surface area and hydrogen bonds are the major contributing factors to the reduced binding affinities observed in tmMnBP6.

FIGURE 8.

Comparison of tmMnBP3, tmMnBP6, and tmCBP. A, ClustalW sequence alignment of tmMnBP3, tmMnBP6, and tmCBP. Residues forming the DXDXDG motif and metal binding site are highlighted in green. The laminaripentaose specificity-determining loop and insertion are in black boldface type and underlined. Amino acids that interact with the first, second, third, fourth, fifth, and six sugar rings are in red-, orange-, green-, cyan-, purple-, and magenta-faced type, respectively. When these residues form hydrogen bonds to the sugar they are also underlined. When a single residue interacts with multiple sugar rings, only the first sugar ring color is shown. Asterisks are where amino acids are identical among all three sequences, colons are strongly conserved substitutions, and periods are weakly conserved substitutions. B, close-up view of the binding site of tmCBP (black loop representation) (laminaripentaose (LP)- and cellopentaose (CP)-bound forms) superimposed with the mannohexaose (MH)-bound tmMnBP6 (gray loop representation). The laminaripentaose specificity-determining loop and insertion are colored in red.

DISCUSSION

Bacteria cannot easily escape the environment in which they live. Dramatic cycling of environmentally derived nutrients requires microbial life to genomically encode adaptation mechanisms that facilitate or maintain biological processes in times of feast and famine. The duplication of mannose-binding proteins in T. maritima represents an efficient strategy to deal with varying nutrient availabilities while minimizing gene duplications. In this metabolite transport system, the two tmMnBPs have a similar ligand specificity profile, although the ligand binding sites share only 48% amino acid identity. The overall net effect of the differences in the binding site leads to K_d values for ligands that differ ∼1000-fold. Based upon fluorescence binding data to laminaribiose, this two protein system allows for a response range to ligands from ∼1E⁻¹⁰ to 1E⁻⁴ m, representing six orders of magnitude and a doubling of the range of a single protein (Fig. 3).

Based upon the conservation of amino acids that comprise the known interaction surface of PBPs with ABC transporters, it is likely that the two tmMnBPs are able to interact with the same transport system (Fig. 6) (27, 28). Although the PBP K_d and transporter K_m are often matched, large differences among these two values are known to exist (7). Coupling of duplicated PBPs to a transporter with a high K_m would allow for a functional acquisition system at both high and low environmental metabolite levels while minimizing saturation with less preferentially metabolized/transported molecules.

The strategy to duplicate and functionalize only the PBP component differs from other metabolite transport systems identified thus far where entire operons are duplicated and represents a genomically efficient strategy to accommodate varying nutrient flux (9–11). Interestingly, this duplication was not identified in all Thermotoga family members. Four close homologs of the low affinity PBP could be found, and these were always coupled to a high affinity PBP. However, the converse is not true: in two instances, only the high affinity PBP is present in the genome. Based on the coupled presence of tmMnBP6 and tmMnBP3 and the increased K_d and reduced T_m of tmMnBP6, we speculate that tmMnBP6 functionalized from tmMnBP3 following a gene duplication event rather than horizontal gene transfer of tmMnBP6 into a transport system with an existing tmMnBP3 gene. Functional divergence of substrate specificities following a tandem duplication event has been suggested in the only other well studied case of PBP duplication in a single operon (16, 35).

The structures of the two tmMnBPs further demonstrate the remarkable adaptability of the PBP superfamily. Based upon the presented binding and structural data, we divide the binding pocket into two parts: the subsite for the first two sugar rings and the water-filled cavity where the remaining sugar rings are placed when longer carbohydrates are bound. We postulate that each half of the bipartite tmMnBP binding pocket plays different roles in the adaptation of this binding site. The binding site of the first two sugar rings is largely what determines the affinity of the receptor. It is in this region that the largest differences in direct molecular interactions with ligands appear, and these differences are likely the major determinants of ligand affinities. However, the adjacent cavity where the remaining sugar rings are placed in the tmMnBPs largely functions in determining specificity. In the first two sugar ring subsites, all disaccharide ligands appear to bind albeit with different affinities. Like tmCBP, it would be expected that an increase in the number of sugar rings and affinity would concomitantly occur and does for β(1,4)-mannose and β(1,4)-glucose ligands. Indeed, the affinity increased (at a minimum due to stoichiometric binding of pentasaccharides) at least 180- and 50-fold when the glucose and mannose pentasaccharides bind, respectively, compared with the disaccharides (Table 1). However, with the β(1,3)-glucose-based and β(1,4)-xylose-based ligands, this trend is not observed. Both xylobiose and laminaribiose bind tighter than the corresponding pentasaccharides, suggesting that the water-filled pocket adjacent to the disaccharide binding site plays a role in discriminating these ligands. Some insights into the potential structural features allowing for this phenomenon can be gained by superposition of the tmMnBPs and tmCBP, a close tmMnBP homolog (Fig. 8) (29). The localization of disaccharide binding sites in the tmMnBPs is identical to that in tmCBP with conservation of many of the residues comprising this site (Fig. 8). The trajectory of the cellopentaose ligand essentially follows the mannohexaose ligand, and both would be placed in essentially the same recognition site in the tmMnBPs. However, due to the β(1,3) linkage of the laminaripentaose ligand, it is placed into a different region of the binding site. In tmCBP, this allows for specific recognition of laminaripentaose through the loop following the first N-terminal β-strand (29). In the tmMnBPs, not only is this loop in a different conformation that sterically restricts laminaripentaose, an insertion of a tryptophan residue (Trp-47) is found when the amino acids sequences of the tmMnBPs are aligned with tmCBP (Fig. 8). This residue would clash with the fourth and fifth laminaribiose sugar rings and likely restricts binding to only trisaccharide laminarins (Fig. 8).

It is less intuitive as to how xylobiose and xylopentaose are discriminated. It would be expected that xylose ligands would bind in a manner essentially the same as the cellulose ligands as xylose lacks the C6 carbon and hydroxyl. The trend of increasing T_m/affinity with the number of sugar rings is not observed with xylose-based ligands. Considering that few if any direct hydrogen bonds are made with these sugar rings, we postulate that the discrimination of xylose from mannose and cellulose ligands is an entropic effect of the large solvent-filled cavity. Having fewer atoms, xylose ligands will displace fewer waters than cellulose and mannose ligands, thereby affecting the entropic contribution of ligand binding. This suggests that the bulk trapped water molecules may be playing a role in determining ligand specificity in the PBPs that utilize these binding site water pockets.

Although the presented data demonstrate the potential molecular determinants allowing for adaptation to varying nutrient flux, further detailed studies of the transcriptional control of these systems need to be carried out. Microarray studies of the T. maritima mannose transport operon have shown differential expression of the tmMnBP3 and tmMnBP6 gene products in response to the nature of the carbohydrate in the growth medium (14, 36). The results of these genetic studies essentially mirror the data presented here with tmMnBP3 transcript levels up-regulated in response to a broad range of carbohydrates, consistent with the tight binding observed by fluorescence. tmMnBP6 was up-regulated in response to only those carbohydrates to which it binds the tightest (mannose, β(1,4)-mannose, or mixed β(1,4)-linked mannose and glucose). β(1,4)-linked glucans were not tested in these studies, but both tmMnBPs were up-regulated in response to mixed β(1,4)-linked mannose and glucose, and our studies demonstrate that they bind the tightest to β(1,4)-linked glucans. Interestingly, the mannose transport gene cluster not only contains two PBPs but also two potential transcriptional regulators. One of these transcriptional regulators, TM1218, mirrors the differential carbohydrate transcriptional up-regulation of tmMnBP3/TM1223, whereas the other, TM1224, has a transcriptional up-regulation pattern identical to that of tmMnBP6/TM1226 (14). This suggests that the control of the expression of the two tmMnBPs may lie at the transcriptional level whereby environmental carbohydrate levels are sensed and expression of the tmMnBPs is adjusted accordingly.