L-PGDS [lipocalin-type PG (prostaglandin) D synthase] is a multi-functional protein, acting as a PGD2-producing enzyme and a lipid-transporter. In the present study, we focus on the function of L-PGDS as an extracellular transporter for small lipophilic molecules. We characterize the binding mechanism of human L-PGDS for the molecules, especially binding affinity stoichiometry and driving force, using tryptophan fluorescence quenching, ICD (induced circular dichroism) and ITC (isothermal titration calorimetry). The tryptophan fluorescence quenching measurements revealed that haem metabolites such as haemin, biliverdin and bilirubin bind to L-PGDS with significantly higher affinities than the other small lipophilic ligands examined, showing dissociation constant (Kd) values from 17.0 to 20.9 nM. We focused particularly on the extra-specificities of haem metabolites and L-PGDS. The ITC and ICD data revealed that two molecules of the haem metabolites bind to L-PGDS with high and low affinities, showing Kd values from 2.8 to 18.1 nM and from 0.209 to 1.63 μM respectively. The thermodynamic parameters for the interactions revealed that the contributions of enthalpy and entropy change were considerably different for each haem metabolite even when the Gibbs energy change was the same. Thus we believe that the binding energy of haem metabolites to L-PGDS is optimized by balancing enthalpy and entropy change.
- haem metabolite
- lipid-transporter protein
- lipocalin family
- lipocalin-type prostaglandin D synthase (L-PGDS)
- protein–ligand interaction
L-PGDS [lipocalin-type PG (prostaglandin) D synthase] (also called prostaglandin-H2 D-isomerase; EC 18.104.22.168) is involved in the biosynthesis of PGD2, acting as an endogenous somnogen and allergy response . L-PGDS, also called β-trace protein, is the second most abundant protein (after serum albumin) in human CSF (cerebrospinal fluid) [2–4]. In clinical cases, the concentration of L-PGDS in CSF is closely associated with neurological diseases such as normal pressure hydrocephalus , spinal canal stenosis , and SAH (subarachnoid haemorrhage) [7,8]. Thus L-PGDS has been proposed as a clinical biomarker in the CSF.
L-PGDS belongs to the lipocalin superfamily which includes β-lactoglobulin , retinoic acid-binding protein , tear lipocalin , apolipoprotein M  and major urinary protein . Lipocalins are small globular proteins of approximately 200 amino acid residues that bind to lipophilic molecules such as retinoids and fatty acids [14,15]. We have previously reported that the tertiary structure of L-PGDS shows a typical lipocalin fold consisting of an eight-stranded antiparallel β-barrel and a long α-helix, and that L-PGDS holds small lipophilic ligands within a hydrophobic cavity of the β-barrel [16–18]. A series of in vitro binding studies using tryptophan fluorescence quenching indicated that rat and mouse L-PGDS could bind to various small lipophilic ligands including retinoids, biliverdin and bilirubin, and thyroids, just as non-mammalian L-PGDS does [14,18–23]. We call this feature of L-PGDS ‘broad ligand selectivity’ . In addition, L-PGDS becomes compact upon binding a small lipophilic ligand, indicating that such compact packing leads to high affinity for its ligands [14,18]. However, the endogenous or exogenous ligand for L-PGDS in vivo has not been identified.
All binding reactions of macromolecules in a closed system are thermodynamically governed. In addition, binding experiments using a system at equilibrium are necessary to determine the precise binding affinity between molecules. Therefore, systematic thermodynamic surveys such as calorimetry are essential for understanding the origin of binding reactions, because calorimetry directly measures reaction heat and provides the thermodynamic parameters Gibbs energy change (ΔG), enthalpy change (ΔH) and entropy change (ΔS) . Although binding studies for a wide variety of L-PGDSs have been carried out so far [14,18–23], the thermodynamic origins of binding affinities of L-PGDS for small lipophilic molecules remain unclear.
In the present study, we focus on the function of human L-PGDS as a lipid transporter and investigate the binding mechanism of L-PGDS for small lipophilic molecules using tryptophan fluorescence quenching, ICD (induced circular dichroism) and ITC (isothermal titration calorimetry). The combination of fluorescence, ICD and ITC measurements was very useful to obtain detailed information on the interactions between LPGDS and the ligands, especially driving force, stoichiometry, affinity and mechanism for binding. We found that two molecules of haem metabolites such as haemin, biliverdin and bilirubin bind to the L-PGDS molecule with high and low affinities.
Bilirubin, all-trans-retinoic acid, 9-cis-retinoic acid, palmitic acid, stearic acid and arachidic acid were purchased from Wako. Haemin, T4 (L-thyroxine), T3 (3,3′,5-triiodo-L-thyronine), genistein, daidzein and progesterone were obtained from Sigma. Biliverdin, testosterone, naringenin, and TNS [2-(p-toluidinil) naphthalene-6-sulfonic acid] were purchased from MP Biomedicals, Fluka, Chroma Dex and Invitrogen respectively. The ligands were dissolved in DMSO or ethanol. The concentrations of ligands in DMSO were determined spectroscopically with a molar absorption coefficient of ∊406 for haemin=126200 M−1·cm−1, ∊382 for biliverdin=49400 M−1·cm−1, ∊454 for bilirubin=59000 M−1·cm−1, ∊358 for T4=6070 M−1·cm−1, ∊301 for T3=4690 M−1·cm−1, ∊360 for all-trans and 9-cis retinoic acid=27000 M−1·cm−1, and ∊318 for TNS=27800 M−1·cm−1. The concentrations of ligands in ethanol were determined spectroscopically with ∊377 for biliverdin=52300 M−1·cm−1, ∊300 for T4=3400 M−1·cm−1 and ∊345 for retinoic acid=27900 M−1·cm−1. The concentrations of the other ligands were calculated based on the molecular mass of each compound. All of the chemicals were analytical grade.
Purification of recombinant human L-PGDS
The open reading frame for human L-PGDS, which is composed of 190 amino acid residues (GenBank® accession number M61900), was ligated into the BamHI and EcoRI sites of the expression vector pGEX-2T (GE Healthcare). The N-terminal 22-amino acid residues corresponding to the putative secretion signal peptide of L-PGDS were truncated. The C65A/C167A (∊280=25900 M−1·cm−1)- and W54F/C65A/W112F/C167A (∊280=14000 M−1·cm−1)-substituted human L-PGDS, which preserve a disulfide bond between Cys89 and Cys186, were expressed in Escherichia coli BL21 (DE3) (Toyobo). Site-directed mutagenesis was performed using the QuikChange™ site-directed mutagenesis kit (Stratagene). When Trp43 in L-PGDS was replaced by a phenyalanine residue, W43F/C65A/C167A-L-PGDS could not be purified owing to the inclusion body formation of protein. The mutated L-PGDS was expressed as a glutathione transferase fusion protein. The fusion protein was bound to glutathione–Sepharose 4B (GE Healthcare) and incubated overnight with thrombin (100 units/100 μl) to release L-PGDS. The recombinant proteins were further purified by gel filtration chromatography with HiLoad 16/600 Superdex 75 (GE Healthcare) in 5 mM Tris/HCl (pH 8.0) and dialysed against each buffer. The disulfide bond formation of the proteins between Cys89 and Cys186 was confirmed by the thiol modification of cysteine residues of L-PGDS with 5,5′-dithiobis-2-nitrobenzoic acid .
Tryptophan fluorescence quenching measurements
Various concentrations of each small lipophilic ligand were added to C65A/C167A- and W54F/C65A/W112F/C167A-L-PGDS solutions, and the final concentration of L-PGDS in both was adjusted to 1.5 μM in 5 mM Tris/HCl (pH 8.0) containing 5% (v/v) DMSO. After incubation at 25°C for 30 min, the intrinsic tryptophan fluorescence was measured using an F-7000 fluorescence spectrophotometer (Hitachi) with an excitation wavelength of 290 nm and an emission wavelength of 334 nm. The quenching of tryptophan fluorescence, caused by non-specific interactions with each ligand, and the absorption of each ligand were corrected with 5 μM N-acetyl-L-tryptophanamide. The mass law equation was used to evaluate apparent dissociation constant (Kd) values for a single binding between the ligands and L-PGDS. The fitting equation of this model (eqn 1) was: (1) where [L0] and [P0] are the total concentrations of ligand and protein respectively. The value of n is the number of independent binding sites for ligands, and each binding site is assumed to show the same affinity and fluorescence quenching ratios. F is the relative fluorescence intensity at a particular L0 value, F0 is the initial fluorescence intensity in the absence of ligand, and Fmin is the minimum fluorescence intensity after the saturation of all protein-binding sites.
TNS competition assay
Various concentrations of TNS as a fluorescence probe were added to L-PGDS solutions, and the final concentration of L-PGDS was adjusted to 1.5 μM in 5 mM Tris/HCl (pH 8.0) containing 5% (v/v) DMSO. After incubation at 25°C for 30 min, the fluorescence intensities of TNS bound to L-PGDS were measured using a F-7000 fluorescence spectrophotometer (Hitachi) with an excitation wavelength of 350 nm and an emission wavelength of 440 nm. The dissociation constant for binding L-PGDS to TNS (Kd-TNS) was calculated from the fluorescence enhancement curve using a single-binding-site model (eqn 1) to be 4.92 μM (Supplementary Figure S1 at http://www.BiochemJ.org/bj/446/bj4460279add.htm). Next, the inhibition constants (Ki) of saturated fatty acids were assessed by observing the reduction of TNS fluorescence intensity in the various concentrations of each fatty acid as a competitor. The final concentrations of L-PGDS and TNS were adjusted to 1.5 μM and 50 μM respectively, under the conditions described above. After incubation at 25°C for 30 min, the fluorescence intensity of TNS was measured at 440 nm after excitation at 350 nm in the presence of various concentrations of fatty acids. The Ki values of fatty acids were calculated by the following equation (eqn 2): (2) where [TNS] and [FA] are the total concentration of TNS and fatty acid respectively. F is the relative fluorescence intensity at a particular [FA] value. Fquench is the reduction in TNS fluorescence induced by the competitor binding. Fluorescence intensities by non-specific binding are denoted NSF.
All CD measurements were carried out using a J-820 spectropolarimeter (Jasco). In the case of ICD measurements, the temperature of the solution in the cuvette was controlled at 25±0.5°C by circulating water. The pathlength of the optical quartz cuvette was 10 mm for near-UV to visible range CD measurements at 300–800 nm. Various concentrations of each small lipophilic ligand were added to L-PGDS solutions, and the final concentrations of L-PGDS were 20 μM for haemin and 5.2 μM for biliverdin and bilirubin in 20 mM Tris/HCl (pH 8.0) containing 5% (v/v) DMSO. The obtained changes in ICD intensities (IICD) at the various wavelengths were globally fitted to (eqn 3) based on the two sets of an independent binding-sites model using the Global Fitting Package of IGOR Pro, version 6.22. The fitting equation of this binding model (eqn 3) was: (3) where (eqn 4): (4) where: IICD is the CD intensity of various wavelengths at a certain total ligand concentration (L0). ICD1 and ICD2 are the maximum CD intensities of ligands bound to the protein for the high and low affinity sites respectively. [L] is the concentration of free ligand. The values of n1 and Kd1, and n2 and Kd2, represent the binding stoichiometry and the dissociation constant for the high and low affinity sites respectively and are defined as global parameters for the data fitting.
For thermal-induced unfolding of L-PGDS, the pathlength of the optical quartz cuvette was 1.0 mm for far-UV CD measurements. The final L-PGDS concentration of all solutions was 5 μM in 20 mM sodium phosphate (pH 7.0) containing 5% (v/v) ethanol. The solutions of L-PGDS–ligand complexes contained 25 μM biliverdin, 40 μM all-trans-retinoic acid, 45 μM T4, 25 μM progesterone and 25 μM genistein. Changes in the CD intensity of the sample solution in the absence or presence of ligands were measured at 200 nm over a temperature range of 25–90°C at a scan rate of 1°C/min. The data were expressed as molar residue ellipticity (θ) per residue. The thermal unfolding curves were analysed using a two-state unfolding equilibrium model described previously .
Isothermal titration calorimetry
Calorimetric experiments were performed with a NanoITC instrument (TA Instruments) and a VP-ITC instrument (MicroCal) in 50 mM Tris/HCl (pH 8.0) containing 5% (v/v) DMSO at 25°C. L-PGDS (500 μM) in the injection syringe was reverse-titrated into 100 μM haemin, 41.7 μM biliverdin or 31.4 μM bilirubin in the cell respectively. Titration experiments consisted of 25–40 injections spaced at intervals of 300–360 s. The injection volume was 2 μl for haemin and 5 μl for biliverdin and bilirubin, and the cell was continuously stirred at 350 rev./min for haemin and 270 rev./min for biliverdin and bilirubin. The corresponding heat of dilution of L-PGDS titrated into the buffer was used to correct the data. The thermodynamic parameters were evaluated using the two sets of an independent binding-sites model supplied by MicroCal Origin 5.0 software.
Binding affinities of L-PGDS for various small lipophilic ligands
Human L-PGDS contains three tryptophan residues at positions 43, 54 and 112 (Figure 1A), and shows intrinsic fluorescence around 340 nm by an excitation wavelength of 290 nm. Ligand binding often induces quenching of intrinsic tryptophan fluorescence of L-PGDS . We used fluorescence quenching measurements to determine the binding affinities of C65A/C167A-substituted human L-PGDS for various small lipophilic ligands such as haem metabolites (haemin, biliverdin and bilirubin), retinoids (all-trans-retinoic acid and 9-cis-retinoic acid), thyroids (T4 and T3), steroids (progesterone and testosterone) and flavonoids (genistein and naringenin) (Figure 1B). The fluorescence quenching curves were obtained by binding L-PGDS with small lipophilic ligands (Figures 2A–2E). After adding biliverdin, L-PGDS showed fluorescence quenching in a concentration-dependent manner (Figure 2A). Other ligands behaved similarly. We plotted relative fluorescence intensity against the molar ratio of ligand concentration to L-PGDS concentration (Figures 2A–2E). The binding stoichiometry (n) and Kd value for ligand binding were calculated using a theoretical equation (eqn 1) and are summarized in Supplementary Table S1 (at http://www.BiochemJ.org/bj/446/bj4460279add.htm). In all cases, the n values of 0.8–1.1 were applicable for fitting, indicating that L-PGDS possesses a single binding site for these small lipophilic ligands. Interestingly, the haem metabolites showed high binding affinity for L-PGDS, with Kd values of 20.9 nM for haemin, 19.5 nM for biliverdin and 17.0 nM for bilirubin. Other small lipophilic ligands showed lower binding affinity, with Kd values ranging from 308 nM to 11.3 μM. The Gibbs energy change for binding (ΔG°bind) ranged from −28.2 to −44.4 kJ/mol (Supplementary Table S1).
We also investigated the binding affinity of L-PGDS for saturated fatty acids such as palmitic acid, stearic acid and arachidic acid, but when we added these fatty acids to L-PGDS, we observed no fluorescence quenching. Thus we performed TNS competition assays (see the Experimental section and Supplementary Figure S1) and calculated the Ki value of these fatty acids to the L-PGDS–TNS complex using a theoretical equation based on a competitive inhibition model (eqn 2) (Figure 2F). The Ki values were 1.4 μM for palmitic acid, 0.77 μM for stearic acid and 2.4 μM for arachidic acid (Supplementary Table S2 at http://www.BiochemJ.org/bj/446/bj4460279add.htm). Taken together, these results demonstrate that L-PGDS shows broad ligand selectivity for small lipophilic molecules and that the haem metabolites bind to L-PGDS with significantly higher affinity than do the other small lipophilic ligands.
Contribution of Trp43 to fluorescence quenching
The quenching efficiency of tryptophan fluorescence caused by energy transfer is a function of the distance between tryptophan residues and ligand molecules [27,28]. We observed the complete or partial quenching behaviour of L-PGDS tryptophan fluorescence by adding small lipophilic ligands (Figures 2A–2E). The ligands varied in their quenching efficiency. L-PGDS contains three tryptophan residues: Trp43 is located at the bottom of the cavity of L-PGDS, and Trp54 and Trp112 are located on the H2-helix and EF-loop respectively (Figure 1A). In order to determine the contributions of intrinsic Trp43 by the small lipophilic ligand binding and the binding modes of L-PGDS for the ligands, we measured the ligand-induced fluorescence quenching of the W54F/C65A/W112F/C167A-substituted human L-PGDS mutant, which contains the Trp43 residue. The far-UV CD spectrum of the mutant was almost identical with that of C65A/C167A-L-PGDS, indicating no significant difference in secondary structure (results not shown). In the absence of ligands, the fluorescence intensity of the tryptophan mutant was approximately 60% of that of C65A/C167A-L-PGDS (Supplementary Figure S2 at http://www.BiochemJ.org/bj/446/bj4460279add.htm). For biliverdin, all-trans-retinoic acid, T4, progesterone and genistein, the fluorescence quenching of the tryptophan mutant was concentration-dependent. The quenching efficiency of the tryptophan mutant in the presence of excess ligands was similar to that of C65A/C167A-L-PGDS (Supplementary Figure S2). The n and Kd values for ligand binding were calculated using a theoretical equation (eqn 1) and are summarized in Supplementary Table S3 at http://www.BiochemJ.org/bj/446/bj4460279add.htm. In all cases, n values of 0.7–1.2 were suitable for fitting, indicating that the tryptophan mutant also possesses a single binding site for these small lipophilic ligands. The Kd values of the tryptophan mutant for the ligands were in good agreement with those of C65A/C167A-L-PGDS, indicating that substituting the tryptophan residue for phenylalanine at positions 54 and 112 did not affect its binding affinity for the ligands, and that Trp43 located at the bottom of the cavity, not Trp54 and Trp112, is mainly responsible for the fluorescence quenching of L-PGDS. Thus these results demonstrate that the small lipophilic ligands used in the present study do bind to L-PGDS in the inside of its β-barrel.
Binding states of haem metabolites complexed with L-PGDS
To learn more about the high-affinity binding of L-PGDS for haem metabolites, we studied the binding states of L-PGDS with haem metabolites using ICD measurements. Within the molar ratio ([ligand]/[L-PGDS]) range 0.2–1.0, the ICD spectra of haemin bound to L-PGDS showed positive CD Cotton effects at a peak of 400 nm and negative CD Cotton effects at peaks of 326, 438 and 526 nm, giving an isosbestic point of 413 nm (Figure 3A). Above this range, the Cotton effects of the positive peak at 400 nm were enhanced and red-shifted to 405 nm. The negative Cotton effects at peaks of 326, 438 and 526 nm decreased in a concentration-dependent manner. Above the molar ratio of 2.0, these negative Cotton effects became positive effects (Figure 3A). We plotted the changes in ICD intensity of haemin bound to L-PGDS against the molar ratio at various wavelengths (Figure 3D). These transition curves suggest that the haemin bound to L-PGDS was involved in two distinct binding reactions. In the ICD spectra of biliverdin bound to L-PGDS, both negative (385 nm) and positive (688 nm) CD Cotton effects were observed at the molar ratio of 0.2 (Figure 3B, red line). These Cotton effects increased in a molar-ratio-dependent manner up to the molar ratio of 0.8 (Figure 3B). Above the molar ratio of 1.0, the negative Cotton effects at 385 nm decreased and a new positive peak appeared at 374 nm, whereas the positive Cotton effects at 688 nm decreased. Above the molar ratio of 1.8, the positive Cotton effects at 688 nm changed to negative Cotton effects (Figure 3B). These ICD spectra showed two isosbestic points at 319 and 456 nm. The changes in ICD intensity of biliverdin bound to L-PGDS as biliverdin concentration increases indicate two distinct binding reactions (Figure 3E). When bilirubin bound to L-PGDS within the molar ratio 0.2–1.0, positive and negative Cotton effects were observed at 412 and 466 nm respectively, with an isosbestic point at 435 nm (Figure 3C). Above the molar ratio of 1.0, these Cotton effects were red-shifted to 418 and 476 nm respectively. Like haemin and biliverdin, the changes in ICD intensity of bilirubin bound to L-PGDS indicate two distinct binding reactions (Figure 3F).
In order to investigate binding stoichiometry in detail, the changes in ICD intensity plotted in Figures 3(D)–3(F) were globally analysed using a theoretical equation based on a two sets of independent binding sites model (eqn 3). From these data, we found that L-PGDS possesses two binding sites for haem metabolites. The values of n and Kd obtained are summarized in Supplementary Table S4 at http://www.BiochemJ.org/bj/446/bj4460279add.htm. The Kd values of ligand binding for the high-affinity site were 9.5 nM for haemin, 2.8 nM for biliverdin and 8.1 nM for bilirubin, which is in good agreement with the results obtained by the fluorescence quenching measurements. The Kd values for the low-affinity site were 250 nM for haemin, 710 nM for biliverdin and 430 nM for bilirubin. The ΔG°bind values ranged from −35.1 to −48.8 kJ/mol (Supplementary Table S4). These ICD data indicate that two molecules of the haem metabolites bind to each L-PGDS molecule.
Thermodynamic characterization of interactions between L-PGDS and haem metabolites
Next, the thermodynamic parameters for binding L-PGDS with the haem metabolites were further determined by ITC. By using the sedimentation equilibrium of AUC (analytical ultracentrifugation) and SAXS (small-angle X-ray scattering) measurements, we found that L-PGDS in high concentration is monodispersed without showing any aggregation (results not shown). Figures 4(A)–4(C) show the thermograms and binding isotherms between L-PGDS and the haem metabolites. By the measurements of the heat during the reverse titration of L-PGDS to the ligands, we observed mixed thermal reactions with an exotherm and an endotherm. In the case of haemin, the ITC profile shows an exothermic reaction in the molar ratio range ([L-PGDS]/[ligand]) 0–0.45, followed by an endothermic reaction in the range 0.45–1.2 (Figure 4A). The binding isotherm was fitted with the two sets of an independent binding-sites model (Figure 4A). The ITC profiles of biliverdin and bilirubin were similar to that of haemin, and their binding isotherms were also fitted with the same model (Figures 4B and 4C). The calculated thermodynamic parameters are summarized in Table 1 and Figure 4(D). On the basis of the binding stoichiometry and Kd values obtained by ITC, we conclude that two molecules of the haem metabolite binds to each L-PGDS molecule and that one ligand molecule binds to the high-affinity site and the other to the low-affinity site.
All of the observed enthalpy changes for binding (ΔH°bind) were negative, indicating that ΔH°bind contributes to complex formation (Table 1). ΔH°bind for the high-affinity site was −10.1 kJ/mol for haemin, −12.6 kJ/mol for biliverdin and −29.0 kJ/mol for bilirubin, and for the low-affinity site was −14.8 kJ/mol for haemin, −23.9 kJ/mol for biliverdin and −52.0 kJ/mol for bilirubin. We noted that the absolute ΔH°bind values for the high-affinity site were lower than those for the low-affinity site. Therefore an increasing amount of L-PGDS was added to initially excess ligand, and the exothermic reactions in the molar ratio range up to approximately 0.5 were accompanied by binding reactions at the sites of both affinities. The endothermic reaction observed at a molar ratio above 0.5 is explained by the difference between the dissociation and association heat due to the change in binding equilibrium toward L-PGDS bound to one molecule of ligand.
In the case of haemin, ΔS°bind was a major driving force in ligand binding, and the entropic terms for binding (−TΔS°bind) were −38.3 kJ/mol and −23.3 kJ/mol at the high and low affinities respectively (Table 1). In the case of biliverdin, ligand binding was an entropic-driven process (−TΔS°bind=−34.0 kJ/mol) at the high-affinity site and an enthalpic-driven process (ΔH°bind=−23.9 kJ/mol) at the low-affinity site (Table 1). In the case of bilirubin, however, ΔH°bind was a major driving force for ligand binding. The ΔH°bind values for bilirubin were −29.0 kJ/mol and −52.0 kJ/mol at the high- and low-affinity sites respectively (Table 1). Together with the gains in enthalpy for the complexation, we found positive entropy changes for binding (ΔS°bind), but not for the low-affinity site of bilirubin, indicating that ΔS°bind stabilizes the complex of haem metabolite and L-PGDS.
The values of Kd and ΔG°bind were estimated by using the thermodynamic relationships, and are summarized in Table 1. The Kd values of ligand binding for the high-affinity site were 3.29 nM for haemin, 6.86 nM for biliverdin and 18.1 nM for bilirubin, and for the low-affinity site were 209 nM for haemin, 1.2 μM for biliverdin and 1.63 μM for bilirubin.
Effects of small lipophilic ligands on thermal unfolding of L-PGDS
To study the effects of small lipophilic ligands on the thermal stability of L-PGDS, we measured the heat-induced unfolding process of L-PGDS in the presence or absence of small lipophilic ligands using far-UV CD spectroscopy. The thermal unfolding process of L-PGDS is reversible at pH 7.0 (results not shown). Changes in CD intensity in thermal unfolding of L-PGDS were monitored at 200 nm, and showed a typical two-state unfolding transition (Figure 5, black circles). The apparent melting temperature (Tm) was 67.7°C (Table 2). The farUV CD spectra of L-PGDS in the presence of various small lipophilic ligands such as biliverdin, all-trans-retinoic acid, T4, progesterone and genistein corresponded to that of L-PGDS in the absence of ligands (results not shown). The entire thermal unfolding curves monitored at 200 nm also exhibited a typical two-state unfolding transition (Figure 5). The apparent Tm values for the complexes were estimated to be 80.7°C for L-PGDS–biliverdin, 73.8°C for L-PGDS–retinoic acid, 71.0°C for L-PGDS–T4, 68.6°C for L-PGDS–progesterone and 69.2°C for L-PGDS–genistein (Table 2). The Tm values for the complexes were higher than that of L-PGDS in the absence of ligands. Surprisingly, the Tm value of the L-PGDS–biliverdin complex was 13.0°C higher than that of L-PGDS. These data suggest that the conformation of L-PGDS is significantly stabilized by the complex formation of L-PGDS and haem metabolite.
Specific binding of human L-PGDS for haem metabolites
In the present study, we demonstrate that human L-PGDS binds to fifteen variations of small lipophilic ligands including haem metabolites, retinoids, thyroids, steroids, flavonoids and saturated fatty acids. The measurements of tryptophan fluorescence quenching and the TNS competition assay revealed that human L-PGDS also shows broad selectivity for small lipophilic ligands as we have already reported in mouse L-PGDS  (Figure 2). Previously Breustedt et al.  reported that human L-PGDS binds to retinol and retinoic acid with Kd values of 170 and 290 nM respectively, indicating that the Kd value for retinoic acid was almost consistent with those shown in the present study (Supplementary Table S1). Interestingly, the binding affinities of human L-PGDS for the haem metabolites such as haemin, biliverdin and bilirubin were much higher than those for other small lipophilic ligands (Supplementary Tables S1 and S2). The same tendency was found in mouse L-PGDS: at pH 8.0, the Kd values were 70 nM for biliverdin and 103 nM for bilirubin, whereas they were 138 nM for all-trans-retinoic acid and 640 nM for T4 [14,21]. Surprisingly, the Kd values of biliverdin and bilirubin for human L-PGDS were approximately 4-fold and 6-fold lower respectively than those for mouse L-PGDS, showing that the haem metabolites bind more tightly to human L-PGDS. These findings support the previous suggestion that human L-PGDS plays an important role in quickly clearing harmful and unwanted lipophilic molecules . Furthermore, the lower Kd values might suggest that humans are more sensitive than mice to stress caused by haem metabolites.
We further characterized the tighter binding of haem metabolites to L-PGDS with respect to the thermal unfolding of proteins. It has been noted that the unfolding temperature of proteins changes in the presence of ligands , a result of equilibrium shifting toward folded species. The conformation of L-PGDS in the presence of haem metabolite was significantly stabler than in the absence of small lipophilic ligands (Figure 5). This stabilization can be simply explained by changes in a stability curve characterized by the Gibbs–Helmholtz equation, thereby increasing the Tm and the ΔG values of unfolding. Intriguingly, there was a positive correlation (correlation coefficient R2=0.795) between the Tm and the ΔG°bind values obtained by the fluorescence quenching measurement (Figure 6), which also suggests high binding affinities of L-PGDS for the haem metabolites.
Binding mode of two haem metabolite molecules to L-PGDS
The analysis of ITC and ICD data demonstrated that two molecules of the haem metabolites bind to each L-PGDS with high and low affinities (Figures 3 and 4). It should be noted that there is a clear discrepancy between the binding stoichiometry obtained from the ICD and ITC measurements and that from the fluorescence quenching measurements. This discrepancy can be explained as follows: in the case of the fluorescence quenching measurements, the Trp43 residue at the bottom of the hydrophobic cavity of L-PGDS is responsible for most of the fluorescence quenching by the ligand binding (Supplementary Figure S2). The ligand-induced fluorescence quenching of the W54F/C65A/W112F/C167A-L-PGDS mutant strongly indicated that L-PGDS holds small lipophilic ligands within the hydrophobic cavity of its β-barrel (Supplementary Figure S2). The secondary binding reaction of haem metabolites could not be observed because the primary binding of haem metabolites in the cavity of L-PGDS induces complete quenching of the Trp43 residue as well as the other tryptophan residues. Thus only a single binding site was detected by fluorescence quenching, whereas for the ICD and ITC measurements we could detect the binding of the two molecules by measuring overall changes in binding equilibrium. In addition, in the case of ligand binding for the high-affinity site, the values of Kd and ΔG°bind obtained by ITC and ICD measurements corresponded well to those obtained by the fluorescence quenching measurements (Table 1, and Supplementary Tables S1 and S4). In the case of ligand binding for the low-affinity site, the Kd and ΔG°bind values are in good agreement with the ITC and ICD measurements (Table 1 and Supplementary Table S4).
Enthalpy–entropy balance for interaction between L-PGDS and haem metabolites
The relative contributions of ΔH°bind and ΔS°bind differ considerably for each haem metabolite, despite the similar values of ΔG°bind (Figure 4D). It has been generally thought that the extensive contacts of lipophilic molecules with the hydrophobic core were dominated by the classical hydrophobic interaction [9,31]. At the high- and low-affinity sites for haemin and the high-affinity site for biliverdin, binding reactions are mainly driven by the highly positive ΔS°bind value (Figure 4D), suggesting the dehydration of the ordered waters surrounding the hydrophobic cavity and surfaces of L-PGDS and small lipophilic ligands into bulk. In addition to the hydrophobic effect, favourable contributions of conformational entropy changes in the backbone flexibility of proteins upon binding hydrophobic ligands are considered to stabilize the protein–ligand complex [32,33]. Frederick et al.  demonstrated that thermodynamically favourable changes in the conformational entropy of calmodulin upon binding with various ligands are linearly correlated with the changes in overall ΔS°bind values obtained by ITC. We recently reported that biliverdin binding increases the internal mobility of ligand-contacting residues in mouse L-PGDS . On the basis of these results, the conformational entropic gains of human L-PGDS may also work to achieve high binding affinities for haemin and biliverdin by increasing protein flexibility.
For the low-affinity site of biliverdin and both binding sites of bilirubin, the favourable ΔH°bind value mainly contributed to the binding reactions (Table 1). The gains in enthalpy may come from hydrogen bonds and electrostatic interactions formed by internal and external contacts between proteins and ligands . The entrance of the hydrophobic cavity of L-PGDS is surrounded by a polar residue cluster consisting of Ser45, Asn51, Ser52, Lys59, Ser81, Glu90, Arg92, Tyr107 and Ser133 . The contacts between these residues and the ligands would contribute to the ΔG°bind value as a favourable gain in enthalpy. In addition, bilirubin has a more flexible backbone than the other haem metabolites, so it can conform structurally well to the L-PGDS molecule, thereby generating more enthalpy contributions and fewer entropic gains than the other haem metabolites (Table 1).
In general, binding reactions in biological mimic systems are in practice driven more by gains in enthalpy than entropy . Although unfavourable energetic terms (i.e. positive ΔH and negative ΔS) can often be neglected, these unfavourable thermodynamic properties are also critical for tuning binding affinities [33,38,39]. As we described above, the interactions between L-PGDS and haem metabolites are not controlled by the single contribution of enthalpy or entropy. The complex formations of haem metabolites with L-PGDS are thermodynamically optimized by the delicate balance of gains in both enthalpy and entropy.
Biological implications of high and low binding affinity features of L-PGDS for small lipophilic ligands
L-PGDS is the second most abundant protein in human CSF [1–4]. We previously showed that the concentration of L-PGDS transiently increases in the CSF of patients after SAH and reaches a peak on day 5 [7,8]. Purified L-PGDS from the CSF binds to haem metabolites such as biliverdin and its derivatives . Biliverdin is well known as the degradation product of haem, a haemin analogue, catalysed by haem oxygenase-1 whose expression level is immediately up-regulated by SAH [40–45]. Subsequently, biliverdin is reduced by biliverdin reductase to form bilirubin. The bilirubin concentrations in the CSF are closely related to the time course of cerebral vasospasms , and the oxidative degradation products of bilirubin are involved in the delayed cerebral vasospasms of patients with SAH [46–50]. We suspect that the high-affinity binding properties of L-PGDS protect haemin, biliverdin and bilirubin from enzyme catalyses. L-PGDS acts as a scavenger of haem metabolites whose degradation products are involved in SAH-induced vasospasms. In this context, when the haem concentration is rapidly up-regulated by SAH, the simultaneous-capture abilities of L-PGDS for two molecules of the haem metabolites may play an important role in inhibiting the onset of vasospasms.
On the basis of our results, we propose a high- and low-affinity binding model of L-PGDS to haem metabolites. In the CSF after SAH, the extracellular concentrations of haem metabolites are rapidly increased [40,51]. In this situation, free L-PGDS traps the two molecules of haem metabolites due to the high up-regulation of ligands by SAH. Such binding modes were clearly demonstrated by the ITC measurements (Figures 4A–4C), and they may make possible a high capacity for toxic molecules such as haemin and bilirubin. Most importantly, the ITC data also demonstrated the changes in binding equilibrium from the low-affinity site to the high-affinity site after an increase with L-PGDS concentrations (Figures 4A–4C). The same changes in binding equilibrium were also confirmed by ICD measurements. Indeed, the ICD spectra of haem metabolites such as biliverdin and bilirubin in the presence of excess LPGDS almost correspond to those at the molar ratio ([ligand]/[L-PGDS]) of 1.0 (Figure 3 and Supplementary Figure S3 at http://www.BiochemJ.org/bj/446/bj4460279add.htm). Therefore, as L-PGDS concentration increases, the ligand molecule bound to the low-affinity site is transferred to the high-affinity site to yield a thermodynamically more stable state due to the energetic gains of ΔΔG°bind of approximately −11 kJ/mol (Table 1). We consider that the suggested binding system underlying the ligand transfer may provide a plausible haem-scavenging mechanism to counteract vasospasms after SAH.
Satoshi Kume and Takashi Inui designed the research; Satoshi Kume and Young-Ho Lee performed the biochemical and biophysical experiments; Satoshi Kume, Yuya Miyamoto and Takashi Inui analysed the results; Harumi Fukada provided helpful discussions for all of the data analyses; Yuji Goto provided the VP-ITC instrument and helpful discussions; and Satoshi Kume, Young-Ho Lee and Takashi Inui wrote the paper.
This work was supported by Grants-in-Aid for Scientific Researches (B) and (C) [grant numbers 17300165 and 21500428 respectively (to T.I.)], a Grant-in-Aid for Scientific Research on Innovative Areas [grant number 02120076 (to T.I.)], a Grant-in-Aid for JSPS (Japan Society for the Promotion of Science) Fellows [grant number 09J10176 (to Y.M.)], and a Special Research Grant from Osaka Prefecture University (to T.I.). Part of this work was performed under the Cooperative Research Program of the Institute for Protein Research, Osaka University.
We thank Ms M. Sakai (Institute for Protein Research, Osaka University, Suita, Osaka, Japan) for help with the AUC (analytical ultracentrifugation) measurements, Dr J. Rodrigue (Osaka Prefecture University, Sakai, Osaka, Japan) for critical reading of the paper prior to submission, Dr N. Yagi (Japan Synchrotron Radiation Research Institute, Sayo, Hyogo, Japan), Dr S. Kidokoro (Nagaoka University of Technology, Nagaoka, Nigata, Japan), and Mr S. Nishimura (Osaka Prefecture University, Sakai, Osaka, Japan) for helpful discussions. The synchrotron radiation experiments were performed at the beamline BL40B2 at SPring-8 with the approval of JASRI (proposal numbers 2008A1657, 2008B1784, 2009A1695 and 2010A1603). We thank Mr R. Takahashi, Mr M. Tabata, Mr M. Kohno and Mr T. Shinohara (Osaka Prefecture University, Sakai, Osaka, Japan) for their technical assistance with the SAXS experiments.
Abbreviations: CSF, cerebrospinal fluid; ICD, induced circular dichroism; ITC, isothermal titration calorimetry; PG, prostaglandin; L-PGDS, lipocalin-type PGD synthase; SAH, subarachnoid haemorrhage; SAXS, small-angle X-ray scattering; T3, 3,3′,5-triiodo-L-thyronine; T4, L-thyroxine; TNS, 2-(p-toluidinil) naphthalene-6-sulfonic acid
- © The Authors Journal compilation © 2012 Biochemical Society