Oligomeric States of the Detergent-solubilized Human Serum Paraoxonase (PON1)*

Human plasma paraoxonase (HuPON1) is a high density lipoprotein (HDL)-bound enzyme exhibiting antiatherogenic properties. The molecular basis for the binding specificity of HuPON1 to HDL has not been established. Isolation of HuPON1 from HDL requires the use of detergents. We have determined the activity, dispersity, and oligomeric states of HuPON1 in solutions containing mild detergents using nondenaturing electrophoresis, size exclusion chromatography, and cross-linking. HuPON1 was active whatever its oligomeric state. In nonmicellar solutions, HuPON1 was polydisperse. In contrast, HuPON1 exhibited apparent homogeneity in micellar solutions, except with CHAPS. The enzyme apparent hydrodynamic radius varied with the type of detergent and protein concentration. In C12E8 micellar solutions, from sedimentation velocity, equilibrium analytical ultracentrifugation, and radioactive detergent binding, HuPON1 was described as monomers and dimers in equilibrium. A decrease of the detergent concentration shifted this equilibrium toward the formation of dimers. About 100 detergent molecules were associated per monomer and dimer. The assembly of amphiphilic molecules, phospholipids in vivo, in sufficiently large aggregates could be a prerequisite for anchoring of HuPON1 and then allowing stabilization of the enzyme activity. Changes of HDL size and shape could strongly affect the binding affinity and stability of HuPON1 and result in reduced antioxidative capacity of the lipoprotein.

Human plasma paraoxonase (HuPON1) is a high density lipoprotein (HDL)-bound enzyme exhibiting antiatherogenic properties. The molecular basis for the binding specificity of HuPON1 to HDL has not been established. Isolation of HuPON1 from HDL requires the use of detergents. We have determined the activity, dispersity, and oligomeric states of HuPON1 in solutions containing mild detergents using nondenaturing electrophoresis, size exclusion chromatography, and crosslinking. HuPON1 was active whatever its oligomeric state. In nonmicellar solutions, HuPON1 was polydisperse. In contrast, HuPON1 exhibited apparent homogeneity in micellar solutions, except with CHAPS. The enzyme apparent hydrodynamic radius varied with the type of detergent and protein concentration. In C 12 E 8 micellar solutions, from sedimentation velocity, equilibrium analytical ultracentrifugation, and radioactive detergent binding, HuPON1 was described as monomers and dimers in equilibrium. A decrease of the detergent concentration shifted this equilibrium toward the formation of dimers. About 100 detergent molecules were associated per monomer and dimer. The assembly of amphiphilic molecules, phospholipids in vivo, in sufficiently large aggregates could be a prerequisite for anchoring of HuPON1 and then allowing stabilization of the enzyme activity. Changes of HDL size and shape could strongly affect the binding affinity and stability of HuPON1 and result in reduced antioxidative capacity of the lipoprotein.
Human serum paraoxonase (HuPON1) 1 is a high density lipoprotein (HDL)-bound protein exhibiting antioxidative prop-erties (1,2). It has been shown to hydrolyze phospholipid oxidation products (3), the platelet-activating factor (4), and the L-homocysteine thiolactone (5), thereby contributing to the prevention of atherogenesis and inflammation in blood vessel walls. In addition, HuPON1 has large substrate specificity toward different unnatural compounds: arylesters, organophosphates (OP), including the nerve agents sarin and soman (6), and lactones (7). The physiological importance of PON1 was demonstrated through studies on PON1 knockout mice, which were more sensitive to OP poisoning and more susceptible to atherogenesis relative to wild-type mice (8). In serum, PON1 is exclusively associated with HDL-type complexes, which stabilize the enzyme activity (9,10) and, presumably, provide the enzyme an optimal environment for interacting with its physiological substrates. The binding of HuPON1 to HDL has been shown to occur at the cell surface, where the enzyme is anchored prior to secretion (10). The molecular basis for the binding selectivity of HuPON1 to HDL-type complexes has not been established. The three-dimensional structure of HuPON1 has not yet been determined. However, as demonstrated by Sorenson et al. (9), the enzyme hydrophobic N terminus, which corresponds to a retained signal peptide, is essential for binding to HDL and phospholipid micelles. Besides interacting with HDL-type complexes, HuPON1 has the capability to associate with other amphiphilic complexes such as phospholipid micelles (9,10) and, although not demonstrated, detergent molecules. For instance, the nonionic detergent Triton X-100 was used to isolate HuPON1 from other HDL-bound proteins and, similarly to HDL in serum, allowed stabilization of the enzyme activity in vitro (11).
The objectives of this study were to structurally and functionally characterize HuPON1 in detergent solutions and determine whether the detergent type and concentration modulate the activity and interactions of HuPON1. Using detergent molecules as models of amphiphiles, we intended to establish the molecular basis for the interaction selectivity between amphiphilic molecules and HuPON1. Our long term goal is to design HuPON1 mutants with an increased OP hydrolase ac-tivity for use as a therapeutic in OP poisoning. A rational design of such mutants ultimately will benefit from the knowledge of the catalytic mechanism for the hydrolysis of OP, based on the HuPON1 three-dimensional structure. Previously, sitedirected mutagenesis has been used to identify residues essential for the HuPON1 OP hydrolase activity (12). However, in the absence of a structural model, it is difficult to confirm whether these residues are located in the enzyme active site and what their functional role is. One of our aims was to find a rationale for crystallization properties of HuPON1, a prerequisite for the determination of its three-dimensional structure through x-ray diffraction. Crystallization of HuPON1 first requires the preparation of a pure and homogeneous sample. The purified plasma HuPON1 is a 354-amino acid protein that exhibits 1-3 bands at 37-44 kDa following SDS-PAGE (13). These bands have been shown to correspond to protein glycoforms (13). Therefore, homogeneity of HuPON1 pure samples a priori depends on two main factors: the protein glycosylation and oligomeric states. Earlier attempts have been made to determine the homogeneity and apparent molecular mass of the native PON1 isolated from human, rabbit, sheep, and bovine sera (see Ref. 14 for a review). The apparent molecular mass of the native PON1 was found to be variable and ranged within 70 -500 kDa. This polydisperse behavior suggests that PON1 exhibits multiple oligomeric states or that the apparent molecular mass is greatly affected by the amount of bound lipids or detergents that were used to dissociate the enzyme from HDL. Since then, the issue of homogeneity and oligomeric states of the native HuPON1 and the influence of detergent in the stoichiometry of the enzyme had not been addressed. In addition, although it was shown that the rabbit PON1 (RaPON1) and HuPON1 monomers retained some activity after SDS-PAGE (15), it had not been clearly established whether the HuPON1 native monomer was active.
In the present work, nondenaturing polyacrylamide gradient gel electrophoresis (PAGGE), size exclusion chromatography, and sedimentation velocity in the presence of mild nonionic or zwitterionic detergents were used to measure the apparent hydrodynamic radius (Rh) and sedimentation coefficient (s) of HuPON1 and establish whether the HuPON1 dispersity was affected by the detergent type and concentration. The specific arylesterase activity of HuPON1 was determined on fractions eluted from size exclusion chromatography, or semiquantitatively estimated on gel following nondenaturing electrophoresis. Chemical cross-linking and analytical ultracentrifugation equilibrium experiments were performed at different detergent concentrations in order to characterize the HuPON1 oligomeric states and determine the effects of the detergent concentration on the enzyme oligomerization. Gel filtration chromatography was used to determine the amount of radioactive detergent molecules bound to HuPON1 in micellar solutions.
The results of this study allowed us to determine the quaternary structure and activity of HuPON1 in solutions as a function of detergent type and concentration and establish how the transitions between HuPON1 oligomers are modulated by the detergent concentration. In addition, a model of interaction between HuPON1 and detergent molecules was designed. Finally, given the structural and functional similarities between the detergent-and HDL-bound HuPON1, the biological significance of these findings was addressed.

EXPERIMENTAL PROCEDURES
Chemicals-The detergents Triton X-100, n-dodecyl-␤-D-maltopyranoside (C 12 -maltoside), and CHAPS were from Sigma; n-dodecyl-N,Ndimethylamine-N-oxide (LDAO) and octaethylene glycol monododecyl ether (C 12 E 8 ) were from Fluka; cyclohexyl-hexyl-␤-D-maltoside (CY-MAL-6) and N-dodecylphosphocholine (FOS-CHOLINE-12) were from Anatrace. Radioactive C 12 E 8 at 54.2 mCi/mmol was from CEA Saclay, Service des Molécules Marquées. High and low molecular weight standard proteins were purchased from Amersham Biosciences or Sigma. Human apolipoprotein A-I (apoA-I) and apolipoprotein A-II (apoA-II) were provided by Calbiochem. Phenylacetate, ␤-naphtyl acetate, and Fast Blue RR were from Sigma. PON1 Purification, Activity, and Concentration-Human and rabbit PON1 were purified according to the method of Gan et al. (11). Briefly, an affinity chromatography on blue Cibacron allowed the isolation of hydrophobic particles from plasma, mainly lipoproteins. Then PON1 was separated from other HDL-bound proteins, mainly apoA-I, using Triton X-100 and DEAE anion exchange chromatography. Enzyme assays were performed, at 25°C, in 50 mM Tris/HCl buffer, pH 8.0, containing 1 mM CaCl 2 (buffer A) as described by Gan et al. (11). One unit of enzyme activity corresponded to the hydrolysis of 1 mol of phenylacetate/min. The protein concentration of PON1 samples was determined using the BCA or micro-BCA kit (Pierce).
SDS-PAGE and Western Blot Analysis-PAGE under denaturing conditions (0.1% SDS) was carried out using 4 -20% polyacrylamide gradient separating gels according to the method of Laemmli (16). Prior to loading on a 4% polyacrylamide stacking gel, samples were denatured in a loading buffer containing SDS 1% (w/v) and glycerol 10% (v/v). Prestained molecular weight markers (Kaleidoscope; Bio-Rad) were loaded on each gel. SDS-PAGE was performed at constant voltage (100 V) for 8 h with tap water cooling. Electrotransfer (0.5 A, 1 h) onto a polyvinylidene difluoride membrane (Bio-Rad) was carried out at 4°C. The membrane was first incubated in Tris-buffered saline containing 0.2% Tween and 5% dry milk at 25°C for 1 h and then incubated with the F41F2-K monoclonal anti-human PON1 antibody (17) at 25°C for 2 h. The membrane was then washed in Tris-buffered saline buffer containing 5% milk and incubated at 25°C for 2 h with an anti-mouse Ig, horseradish peroxidase-linked antibody from sheep (Amersham Biosciences). After a final step washing in Tris-buffered saline buffer, the peroxidase activity was revealed with a chemiluminescent detection kit (ECL kit, Amersham Biosciences).
Staining of Gels for Protein Detection and Arylesterase Activity-Gels were stained for arylesterase activity according to Furlong et al. (15). Gels were placed in a 100-ml solution of 50 mM Tris/HCl buffer, pH 8.0, containing 1 mM CaCl 2 and 50 mg of ␤-naphtyl acetate dissolved in 1 ml of ethanol. Reaction between ␤-naphtol, generated upon enzymatic hydrolysis of ␤-naphtyl acetate, and Fast Blue RR (50 mg) yielded red bands. Gel fixing and staining for proteins were performed in 60 mg/ liter Coomassie Brilliant Blue, 5% methanol, and 7.5% acetic acid, with destaining in 10% acetic acid. Detection limit of the activity staining was ϳ0.1 g of active PON1 per band; it was ϳ1 g for the protein staining using Coomassie Blue.
Detergent Exchange-Prior to size exclusion chromatography and analytical ultracentrifugation, Triton X-100 in the purified enzyme was exchanged for C 12 E 8 or C 12 -maltoside using anion exchange chromatography on a Q-Sepharose fast flow gel (Amersham Biosciences). Five milliliters of gel were placed on an 8-mm-diameter column (Amersham Biosciences). The gel was first equilibrated with 25 ml of buffer A containing an appropriate concentration of C 12 E 8 (buffer B). Purified HuPON1, solubilized in 1.5 mM Triton X-100, was first diluted by a 2-fold factor in buffer B and then injected on the gel at a 2 ml/min flow rate. The gel was then washed with 25 ml of buffer B. Elution of HuPON1 was performed with buffer B containing an appropriate concentration of NaCl.
Size Exclusion Chromatography-Size exclusion chromatography was performed at 25°C using a 30-ml Superdex 75 HR 10/30 column (Amersham Biosciences). The gel was equilibrated in buffer A containing 0.36 mM C 12 E 8 or 0.68 mM C 12 -maltoside. Samples of HuPON1 (100 l; ϳ0.5 mg/ml) were loaded on the gel after Triton X-100 exchange with C 12 E 8 or C 12 -maltoside, as already described. The flow rate was 0.5 ml/min, and 0.5-ml fractions were collected. The absorbance at 280 nm and arylesterase activity toward phenylacetate were measured on each fraction collected. Water-soluble globular proteins of known Rh (20) were used as standards: thyroglobulin (8.6 nm), ferritin (6.3 nm), catalase (5.2 nm), aldolase (4.6 nm), bovine serum albumin (3.5 nm), ovalbumin (2.8 nm), and ␤-lactoglobulin (2.75 nm). The Rh of these standard proteins has already been shown to be unchanged in buffers containing various detergents, including C 12 E 8 (21). The linear calibration curve representing the logarithm of Rh as a function of the elution volume was used to calculate the Rh of HuPON1. For HuPON1 in the presence of C 12 E 8 , an independent set of measurements was performed at 4°C and two protein concentrations on Superose 12 HR10 -30, using thyroglobulin (8.6 nm), apoferritin (6.3 nm), aldolase (4.6 nm), transferrin (3.6 nm), peroxidase (3.0 nm), ovalbumin (2.8 nm), trypsin (2.2 nm), myoglobin (1.9 nm), and cytochrome c (1.7 nm), in parallel to the measurement of bound detergent (see below).
Human PON1 Cross-linking-The HuPON1 oligomeric states were probed using homobifunctional cross-linking reagents selective for primary amines, mainly Lys residues, or for the sulfhydryl group of Cys residues. The former reagents, 3,3Ј-dithiobis(sulfosuccinimidylpropionate) (DTSSP) and dithiobis(succinimidylpropionate) (DSP), are thiol-cleavable analogs that exhibit a 12-Å-long spacer arm and preferentially react with Lys present in hydrophilic or hydrophobic environments, respectively (see product description numbers 22585 and 21578 from Pierce). HuPON1 has one free Cys residue in position 284 and two disulfide-bonded Cys residues in positions 42 and 353 (13). Proximity of Cys-284 in HuPON1 oligomers was examined by using two hydrophobic, thiol-specific cross-linking reagents provided by Aldrich, N,NЈ-o-phenylenedimaleimide (o-PDM) and N,NЈ-p-phenylenedimaleimide (p-PDM), exhibiting reactive maleimido groups at fixed distances of about 6 and 10 Å, respectively. This approach has been particularly useful to study spatial relationships between residues that belong to adjacent ␣Ϫhelices in transmembrane proteins (22,23). Five l of purified HuPON1 samples (100 mg/liter), solubilized in buffer A containing 0.18 mM C 12 E 8 , were incubated overnight, at room temperature, with 5 l of the cross-linking reagents and 90 l of the following buffers: CHES (50 mM)/NaOH, CaCl 2 (1 mM), pH 9.0 (for the reactions with DTSSP or DSP) and Tris (25 mM)/HCl, CaCl 2 (1 mM), pH 8.0 (for the reactions with o-PDM or p-PDM), containing or not containing 0.45 mM C 12 E 8 . Residual arylesterase activity was measured in 5-l aliquots, using 5 mM phenylacetate as substrate. Samples were denatured using 20 l of 5% SDS dissolved in Tris (0.25 M)/HCl, pH 6.8. Western blot analysis of samples was performed as described above.
Analytical Ultracentrifugation-Analytical ultracentrifugation was carried out using a Beckman Optima XL-I analytical ultracentrifuge equipped with an An-60 Ti four-hole rotor. Measurements were made in the presence of C 12 E 8 . In a first set of experiments, Triton X-100 in the pure samples was exchanged for C 12 E 8 by ultrafiltration, using 10-kDa cut-off filters (Macrosep, Pall Filtron) and, in a second set as described above, providing four samples in buffer B containing 0.36 mM C 12 E 8 and 280 mM NaCl, 0.81 mM C 12 E 8 and 280 mM NaCl, 1.54 mM C 12 E 8 and 280 mM NaCl, or 7.29 mM C 12 E 8 and 140 mM NaCl. They were used directly or after precise dilution in buffers of appropriate detergent and salt concentrations. Prior to the experiments, the samples were filtered through a 0.45-m disposable sterile syringe filter (Gelman Sciences).
We estimated the molar mass and partial specific volume of the polypeptide chain from the amino acid composition to be 39,618 Da and 0.7265 ml/g, respectively, and calculated the solvent density, , and viscosity, , at the appropriate temperature, neglecting the presence of C 12 E 8 , using Sednterp software (version 1.01; developed by D. B. Haynes, T. Laue, and J. Philo; available on the World Wide Web at www.bbri.org/RASMB/rasmb.html). These values were used to normalize the experimental sedimentation scale as s 20,w values and to derive the buoyancy factors. We checked at 20°C by measuring solvent densities and viscosities (with a DMA5000 and an AMVn from PAAR) that the presence of C 12 E 8 did not significantly affect the density and viscosity of the solvents.
Sedimentation Velocity-Sedimentation velocity experiments were carried out at 40,000 rpm, at 6 or 8°C, for 7 h, using 400 l of HuPON1 samples in the two-channel 12-mm path length centerpieces. Absorbance scans were taken at about 5-min intervals at 277 nm. The sedimentation profiles were analyzed by the size distribution analysis of Sedfit (available on the World Wide Web at www. AnalyticalUltracentrifugation.com), providing a distribution of ap-parent sedimentation coefficients. Finite element solutions of the Lamm equation for a large number of discrete, independent species, for which a relationship between mass and sedimentation and diffusion coefficients s and D is assumed, were combined with a maximum entropy regularization to represent a continuous size distribution (24). It has to be mentioned that an inadequate relationship between mass, s, and D only decreases the resolution of the distribution. Sedfit also takes advantage of a radial and time-independent noise subtraction procedure (25). We modeled 20 or 80 experimental profiles with 100 -200 generated sets of data for sedimentation coefficients comprised between 1 and 15 S on a grid of 100 -200 radial points, calculated with a frictional ratio of 1.2-1.25, a partial specific volume of 0.726 -0.74 ml/g, and a confidence level of 0.68 for the regularization procedure, and we checked the relevance of the results by Monte Carlo statistical analysis.
Sedimentation Equilibrium-Sedimentation equilibrium experiments were performed at four rotor speeds (10,000, 12,000, 15,000, and 20,000 rpm) in six-channel centerpieces. Radial scans of the absorbance at 277 nm were taken every 3 h for a total of 27 h, and equilibrium conditions were checked by comparing the scans. An experimental base-line offset was determined from meniscus depletion at 40,000 rpm at the end of the run. Data of the absorbance at 277 nm were considered as either noninteracting species or a reversible self-associating system between monomers and dimers (and eventually trimers or tetramers) and analyzed using the Multifit analysis program (version 4.01; Beckman Instruments). Association constants, obtained in absorbance units, were converted to molar units by using a molar extinction coefficient of 75,000 M Ϫ1 cm Ϫ1 for the C 12 E 8 -solubilized HuPON1.
Evaluation of the Amount of Detergent Released upon Protein Association-Wyman (26) demonstrated the influence of ligand binding on protein equilibrium processes through the theory of linked functions. The dissociation of dimer is characterized by the apparent dissociation constant, K d . K d is an apparent value, since it is determined in view of the concentrations of monomers and dimers only. If the dissociation process is accompanied by a change in the number of bound molecules of detergent (⌬N det ), K d evolves with the activity a det of detergent as follows, The effects of hydration, emphasized by Tanford (27), were ignored here in view of the low detergent concentration relative to that of water. We related the activity of the detergent to that of the "free" monomer detergent, a monomer detergent , which is linked to that of the micelle species, a micelle , since, for a detergent undergoing autoassociation to micelles with an association number n (28), the following is true: °m onomer detergent Ϫ °d etergent in micelle )/RT Ϸ ln(a micelle 1/n / a monomer detergent ) where R is the gas constant (8.315 J/mol⅐K) and I is the absolute temperature.
Thus, equation (1) can be written in the following form.
We considered for a micelle the micelle concentration calculated from the total detergent concentration of the solvent, the CMC, and n as follows.
The total detergent concentration of the solvent was accurately known for HuPON1 samples after the chromatography on Q-Sepharose as described above. We neglected the variation of the bulk detergent related to HuPON1 dissociation upon dilution. For C 12 E 8 , we used n ϭ 89, and CMC ϭ 0.09 mM (29).
Evaluation of the Amount of Bound Detergent-HuPON1 was loaded on a 6-ml Q-Sepharose gel (see "Detergent Exchange") equilibrated in 0.36 mM C 12 E 8 in buffer A, washed by 60 ml of buffer A* (0.36 mM radiolabeled 14 C 12 E 8 at 0.031 Ci/ml in buffer A), and eluted by buffer B* (0.36 mM 14 C 12 E 8 at 0.031 Ci/ml in buffer A with NaCl 280 mM). 200 l of the more concentrated fraction (0.3 mg/ml) was loaded on a gel filtration column (Superose 12 HR 10 -30 from Amersham Biosciences) previously equilibrated in B* and eluted in the same buffer. A second gel filtration column was performed with the following fractions of HuPON1 from Q-Sepharose, after concentration at 1.2 mg/ml on microconcentrator Amicon 50 (loaded volume, 100 l). The protein concentration in the resulting fractions was evaluated by micro-BCA assays. Aliquots of 50 and 100 l, for the base line and the fractions eluted from ion exchange and gel filtration columns, respectively, were counted for radioactivity (providing cpm bl and cpm peak , respectively).
The amount of bound detergent, ⌬ det , was calculated using the equation, where C det and C HuPON1 correspond to the concentration of the detergent and HuPON1 in the solvent, respectively.

Influence of Detergent Type and Concentration on the Dispersity and Apparent
Rh of HuPON1-The size dispersity and Rh of HuPON1 were determined by nondenaturing PAGGE and size exclusion chromatography. Upon nondenaturing electrophoresis in a gel made of a continuous concentration gradient of polyacrylamide, the proteins are forced to migrate through progressively smaller pores, the sizes of which depend on the polyacrylamide concentration (30). The proteins stop migrating when they reach pores narrower than their diameter. Similarly to size exclusion chromatography, this allows resolution of proteins according to their size. However, by contrast to size exclusion chromatography, which leads to dilution of resolved proteins in the eluate, nondenaturing PAGGE results in concentration of resolved proteins in thin bands.
As illustrated in Fig. 1, the dispersity of HuPON1 from the electrophoresis experiments mainly depended on the presence or absence of detergent in the polyacrylamide gel. With the exception of CHAPS, a unique red band corresponding to active HuPON1 was observed when the detergent concentration incorporated in the gel was slightly higher than the CMC (Fig.  1A). Electrophoresis in gels that did not include detergent molecules (Fig. 1B) or that incorporated a Triton X-100 concentration 10 times lower than the CMC (not shown), or 8.1 mM CHAPS (not shown) led to streaking of HuPON1 and resolution of multiple oligomeric species. As shown in Table I, the main HuPON1 species had Rh of 3.8, 4.2, and 4.6 nm. This indicated enzyme heterogeneity in media containing nonmicellar concentrations of detergent. Identical electrophoretic profiles were obtained for HuPON1 initially solubilized in buffer A containing 0.18 mM C 12 E 8 , 0.34 mM C 12 -maltoside, or 1.5 mM Triton X-100 (Fig. 1B, lanes 6 -8). This suggested that the HuPON1 oligomers resolved in the gel, if associated with detergent molecules, had the same general structure in the presence of C 12 E 8 , C 12 -maltoside and Triton X-100 molecules or that the HuPON1 self-associated species were no longer bound to nonionic detergent molecules as a result of dissociation during electrophoresis. In addition, the electrophoretic pattern of HuPON1 did not change when the amount of protein loaded on the gel was reduced from 10 to 1 g (Fig. 1B, lanes 6 and 5), suggesting that the dispersity of HuPON1 in nonmicellar media was not dependent of the protein concentration.
Rh of HuPON1 could be determined using the linear relationship between the logarithm of Rh and Rf of protein standards. We first observed that incorporation in polyacrylamide gel of a variety of nonionic or zwitterionic detergents, uncharged in the alkaline conditions of electrophoresis, did not highly affect the migration of water-soluble protein standards when compared with the HDL-binding proteins apoA-I, apoA-II, and PON1. With a detergent concentration higher than the CMC, the Rh of HuPON1 varied between 3.8 and 4.8 Ϯ 0.1 nm, depending on the type of detergent incorporated in the gel (Table I). This is indicative of interaction between HuPON1 and detergent molecules incorporated in the gels. Additionally, as shown in Table I, there was a positive correlation between Rh of the HuPON1 and the size of the detergent micelles, suggesting either that HuPON1 could be associated with micelles in gels or that HuPON1 could autoassociate in a detergent type-dependent way (see "Discussion").
At detergent concentrations below the CMC, the Rh values can result from several scenarios. The lowest value determined for the HuPON1 Rh after PAGGE without detergent in gel was found to be 3.8 nm, corresponding to a globular protein of ϳ90 kDa. The 4.2-and 4.6-nm Rh species, of 130-and 170-kDa apparent molecular mass, respectively, could correspond to HuPON1 globular trimers and tetramers. Alternatively, this could be the result of species of lower stoichiometry with associated detergent or elongated shapes. Comparison of band intensity indicated that the 3.8-nm and, to a lesser extent, the 4.6-nm species were the predominant forms when compared with the 4.2-nm ones.
In gels incorporating micellar concentrations of detergents (see Fig. 1A for Triton X-100), RaPON1 was slightly larger than HuPON1. This could result from the 4-amino acid differences between the polypeptide chain size of HuPON1 and RaPON1. It could also be due to a different shape of detergent-enzyme complexes or to distinct protein-protein or protein-detergent interactions for RaPON1 and HuPON1.
Human plasma apoA-I (28 kDa, 243 residues) and A-II (17.4 kDa, a disulfide-linked dimer of two 77-residue molecules) are the major proteins of HDL (31). Similarly to HuPON1, the electrophoretic mobility of apoA-I (Fig. 1, A and B, lane 4) and apoA-II (Fig. 1B, lane 3) depended on the type and concentration of detergent incorporated in gels. However, in contrast to HuPON1, there was no correlation between Rh of apoA-I or apoA-II and Rh of the detergent micelles (not shown). This suggested that interactions of HuPON1 in the presence of detergent micelles were distinct from that of apoA-I and apoA-II.
As illustrated in Fig. 2, even if the elution profiles were not strictly symmetrical, size exclusion chromatography confirmed the rather good apparent homogeneity of C 12 E 8 -and C 12 -maltoside-solubilized HuPON1 observed using nondenaturing PAGGE in gels containing micellar concentrations of C 12 E 8 or C 12 -maltoside, respectively. All of the fractions shared the same specific arylesterase activity toward phenylacetate. The derived Rh values for the C 12 -maltoside-and C 12 E 8 -solubilized HuPON1 (4.6 and 4.4 Ϯ 0.1 nm, respectively, for loading concentrations of 0.5 mg/ml) were similar although not identical to those determined by nondenaturing PAGGE (Table I,  Cross-linking of HuPON1-The homobifunctional Lys-selective DSP and DTSSP and Cys-selective o-PDM and p-PDM were used as cross-linking reagents of HuPON1. In the control without cross-linking agent, the ϳ40-kDa SDS-denatured HuPON1 monomeric glycoforms were the only species detected (Fig. 3B, lane 4). By contrast, following cross-linking with DTSSP, DSP, o-PDM, or p-PDM, multiple SDS-resistant oligomeric proteins were detected by Western blot analysis. Crosslinking of HuPON1 with o-PDM and p-PDM suggested that the free Cys residues of HuPON1 monomers were accessible, correctly oriented within an adequate distance for reacting with neighboring subunits of a complex. As shown in Fig. 3, the proportion of HuPON1 SDS-resistant oligomers depended on type and concentration of cross-linking reagents and concentration of detergent relative to the CMC. The proportion of HuPON1 oligomers increased with the concentration of crosslinking reagents. For instance, monomers were no longer detected after treatment of HuPON1 with 1 mM DTSSP (Fig. 3B, lane 2), suggesting total conversion of HuPON1 monomers into multimers. We did not observe a significantly different pattern of cross-linking species following reaction of HuPON1 with DTSSP and DSP, whereas they are supposed to react with Lys residues in a preferentially hydrophilic or hydrophobic environment, respectively. After dithiothreitol treatment, the DTSSPand DSP-cross-linked HuPON1 were converted into monomers, demonstrating reversibility of Lys-selective cross-linking reactions (not shown).
When the detergent concentration was lower than the CMC, the HuPON1-cross-linked oligomers had ϳ80 -180-kDa apparent molecular mass. The broadness of bands corresponding to HuPON1 oligomers suggested heterogeneity of reactions between HuPON1 and the reagents used, or it possibly came from enzyme aggregation (Fig. 3B). The proportion of HuPON1cross-linked oligomers was much lower in micellar media as opposed to that in nonmicellar media. Reaction of HuPON1 with DTSSP and DSP led to the formation of mainly 80-kDa oligomers and, at a lower extent, to 120-kDa species (Fig. 3A,  lanes 3 and 4). Given that the HuPON1 monomer has a molecular mass of ϳ40 kDa, the 80-and 120-kDa cross-linked oligomers could correspond to HuPON1 dimers and trimers, respectively. Reaction of HuPON1 with o-PDM and p-PDM led to  2. Size exclusion chromatography of C 12 -maltoside-solubilized HuPON1. C 12 -maltoside-solubilized HuPON1 (100 l, 0.5 mg/ ml) was injected onto a Superdex 75 size exclusion column equilibrated in Tris (50 mM)/HCl, pH 8.0, buffer containing CaCl 2 (1 mM) and C 12 -maltoside (0.68 mM). The elution profile was monitored by UV absorbance at 280 nm (Ⅺ), and arylesterase activity was measured on 0.5-ml fractions collected using phenylacetate as substrate (E). the formation of two cross-linked species. They had apparent molecular masses of 100 and 120 kDa, respectively, higher than expected for SDS-denatured HuPON1 covalent dimers. However, since we did not get evidence for HuPON1 trimers by sedimentation equilibrium (see below), the bands of 100 -120-kDa apparent molecular mass could be related to chemical modification of multiple Lys residues by DTSSP or DSP, resulting in formation of intra-and interchain covalent bonds or different oxidation states of the disulfide bond between the Cys-42 and -353 residues following treatment with o-PDM or p-PDM, or to partial denaturation of the cross-linked dimers leading to abnormal migration in SDS gels. Alternatively, HuPON1 trimers could result from dynamic collisions between monomers and dimers. Additionally, we observed that, in micellar media, reactivity of HuPON1 with p-PDM was apparently much more important than with o-PDM, the latter giving rise to hardly visible cross-linked species by Western blot analysis (not shown). This suggests that, in micellar media, the orientation of paired Cys residues is relatively specific and that the distance between them is in the range of 6 -10 Å.
We observed a reduction of the arylesterase activity of HuPON1 cross-linked oligomers when compared with control (not shown). The extent of inactivation was proportional to the concentration of cross-linking reagents. This suggested that HuPON1 active site residues might have been chemically modified by the Lys-and Cys-selective reagents used in this work.

Effects of the Protein and Detergent Concentration on the Sedimentation Coefficient of C 12 E 8 -solubilized HuPON1-Sed-
imentation velocity experiments were performed at various HuPON1 and C 12 E 8 concentrations. They were analyzed in terms of distribution of sedimentation coefficients, which allowed a qualitative evaluation of the enzyme homogeneity and autoassociation capability. Fig. 4A presents the distribution obtained for HuPON1 at 18 and 6 M in 50 mM Tris/HCl, pH 8.0, containing CaCl 2 1 mM and C 12 E 8 0.36 mM. In this experiment, Triton X-100 was exchanged for C 12 E 8 by membrane ultrafiltration. We observed a complex pattern, with two main peaks, the position of the main one being shifted from s 20,w ϭ 4.9 S to s 20,w ϭ 4.6 S when the protein concentration was decreased, which suggested an autoassociation equilibrium process. The light species at 3.5 S was not related to remaining micelles of Triton X-100; considering a molar mass of 90 kDa and a partial specific volume of 0.908 cm 3 /g (28), we calculated a s 20,w value of 1.9 S. Complexity of the HuPON1 sedimentation profile could result from only partial Triton X-100 substitution for C 12 E 8 when using membrane ultrafiltration for the detergent exchange. Consequently, part of HuPON1 could still be associated with Triton X-100, another part with C 12 E 8 or with a mixture of both detergents. Accordingly, when the detergent was exchanged by using Q-Sepharose chromatography, the sedimentation profile of HuPON1 suggested apparent homogeneity of the sample, possibly as a result of complete Triton X-100 substitution for C 12 E 8 . We thus determined the effect of the detergent concentration on the sedimentation pattern of HuPON1, by using Q-Sepharose chromatography for the detergent exchange. From Fig. 4B, it is clear that increasing the C 12 E 8 concentration from 0.81 to 7.29 mM, while maintaining the HuPON1 concentration in the 5-8 M range, decreased the sedimentation coefficients, s 20,w , from 3.5 to 3 S. This suggested a dissociation of a multimer when the detergent concentration increased. An equilibrium between monomers and dimers was suggested from the 2-5 S range of experimental s 20,w , since for two noninteracting globular proteins of molar masses 40 and 80 kDa, corresponding to the HuPON1 monomer and dimer, we calculated theoretical values for s 20,w of 3.4 and 5.4 S, respectively, and the presence of the detergent or the form asymmetry would decrease these values. However, the analysis of the distribution of sedimentation coefficients in the case of interconverting system is complex, even in the absence of interactions with detergents, and we used sedimentation equilibrium to characterize the stoichiometry of HuPON1 in solution.
Molecular Mass of C 12 E 8 -solubilized HuPON1 by Sedimentation Equilibrium-For sedimentation equilibrium experiments, we used samples obtained after detergent exchange through a chromatographic step prior to eventual precise dilution in order to control the detergent concentration. Table II indicates the range of protein and C 12 E 8 concentrations investigated. The detergent C 12 E 8 was chosen because, as opposed to Triton X-100, it does not absorb light at 277 nm, and the reported values of its partial specific volume are close to 1 (0.973 cm 3 /g, according to Ref. 29), and thus its density (approximately the inverse of the partial specific volume) is close to that of the solvent. Thus, we can consider that free or bound C 12 E 8 are not detected in the sedimentation equilibrium profiles, which essentially address the protein stoichiometry. This strategy was used previously to address by equilibrium sedimentation the oligomerization of transmembrane fragments (32) and membrane proteins (33,34).
When the equilibrium sedimentation profiles were fit according to a one-species model, the molecular mass measured, 60 kDa, was intermediate between the molecular mass of a monomer (ϳ40 kDa) and of a dimer (ϳ80 kDa). This agreed with the sedimentation velocity experiments and suggested that PON1 monomers and dimers were in equilibrium in C 12 E 8 micellar solutions. Thus, the data corresponding to the same solvent composition at four rotor speeds were then modeled together assuming a monomer-multimer equilibrium. We first fitted the monomer molar mass, the association number, and the association constant. The derived molar masses for the monomer being always close to the theoretical value for the HuPON1 polypeptide chain and the association number being close to the value of 2, we then fixed this monomer molar mass value to the theoretical one and fitted only the association constant. Considering an equilibrium between monomers and dimers gave better results than monomer-trimer and monomer-tetramer equilibrium. We were unable to fit our data with a model consisting of three species in equilibrium (monomers, dimers, and trimers or tetramers). Using the monomer-dimer model, the quality of the fit was reasonably good, even if the residuals were not completely random, as can be seen for one example on Fig. 5. The reason could be that the detergent bound in different amounts by HuPON1 in its different oligomeric forms would not be completely masked. So, even if multimers of higher stoichiometry in small amounts are not definitively excluded from our data, we restricted our analysis to a monomer-dimer equilibrium.

Effect of the C 12 E 8 Concentration on the Equilibrium Constant (K d ) Corresponding to the HuPON1 Monomer-Dimer
Equilibrium-The values of the dissociation constants, K d , corresponding to the monomer-dimer equilibrium, measured at different concentrations of C 12 E 8 , are reported in Table II. It was clear that an increase of the detergent concentration was accompanied by an increase of K d . This was in qualitative agreement with the sedimentation velocity experiments described above. The fact that K d varied with the C 12 E 8 concentration means that the detergent had distinct interactions for the monomer and the dimer. Consequently, the detergent is a partner of the equilibrium. Increasing the C 12 E 8 concentration favored the species that interact the most efficiently with the detergent; as shown in Table II, in the case of HuPON1, it was clearly the monomeric forms. To quantify the difference in the bound detergent between the HuPON1 monomeric and dimeric forms, we plotted the logarithm of K d as a function of logarithm of the detergent concentration in the solvent. As seen in Fig. 6, the data can be fit with a straight line. The inferred slope of 1.1, expressed in micelle units, is the number of detergent mole-

TABLE II Equilibrium sedimentation of HuPON1 in the presence of C 12 E 8
Column I represents the total detergent concentration of the solvent, and column II shows the protein concentration in the samples. Column III was inferred from column I, considering for C 12 E 8 a critical micelle concentration and an aggregation number of 0.09 mM and 89, respectively (29). Column IV results from the simultaneous fit of four equilibrium profiles obtained at 10,000, 12,000, 15,000, and 20,000 rpm at 6°C, in the model of an equilibrium between HuPON1 monomers and dimers.  Fig. 7 shows the elution profiles of 200 and 100 l of HuPON1 injected at 0.3 and 1.2 mg/ml, respectively, in the presence of 0.36 mM radioactive 14 C 12 E 8 and the measurement of the radioactivity on the recovered fractions. From the measurement on three fractions on each of the two gel filtration columns eluted with radiolabeled C 12 E 8 , bound detergent ⌬ det was estimated to be 0.9 and 1.1 g of C 12 E 8 per g of HuPON1. The maximum concentrations of elution of HuPON1 were 61 and 31 g/ml, respectively. A significantly lower ratio of 0.45 g/g was measured for the fractions eluted from the Q-Sepharose. However, the measurement of low levels of radioactive bound detergent is generally reported for other proteins after ion exchange chromatography (29), and we will not consider this result here. From the gel filtrations, the value of ⌬ det is in the range usually found for membrane proteins (29). It corresponds to about 80 mol of C 12 E 8 /mol of polypeptide. A value of K d ϭ 0.24 M is interpolated from our results in 0.36 mM C 12 E 8 . It provides, for a HuPON1 concentration of 40 g/ml (i.e. close to that of the considered fractions), a value for the proportion f mono of the polypeptide chain in the monomer form of 0.5. Considering, from equilibrium sedimentation, the release of ⌬N det /2 ϭ 50 molecules of detergent/monomer upon HuPON1 dimerization, the number of detergent molecules bound to the monomer, ⌬ det(monomer) , can be estimated from the following.
This gives a value of 105 and 110 mol of C 12 E 8 that would be bound per mole of monomer and dimmer, respectively. This measurement indicates that the monomer of HuPON1 is associated with a number of C 12 E 8 molecules corresponding to nearly a micelle. Note, however, that, as first pointed out by Tanford (35), this result does not demonstrate that the C 12 E 8 molecules are actually arranged as in a protein-free micelle (see "Discussion").

DISCUSSION
In this work, we have characterized different oligomeric states of detergent-solubilized HuPON1 purified from plasma. They all exhibited arylesterase activity. We demonstrated that the detergent type and concentration could strongly affect both the HuPON1 dispersity and oligomeric state. In micellar solutions of C 12 E 8 , a monomer to dimer equilibrium, regulated by the detergent concentration, with the number of bound detergent molecules nearly 100 per monomer and per dimer was found to be the most reliable model. This number corresponds roughly to one micelle for the detergent in solution. In nonmicellar media, HuPON1 was shown to self-associate in multiple oligomeric forms and aggregates.

Detergent-dependent Interactions
Similarly to membrane proteins, we found that protein-protein and protein-detergent interactions specific to HuPON1 were strongly affected by the detergent concentration relative to the CMC and the detergent type.
Nonmicellar Solutions-In nonmicellar solutions, the sizes of the main oligomeric species observed by nondenaturing PAGGE could correspond to globular dimers, trimers, and tetramers of HuPON1. Accordingly, HuPON1 covalent dimers, trimers, and tetramers were formed upon cross-linking in nonmicellar solutions. Oligomerization in nonmicellar solutions is a usual trait for membrane proteins and has been demonstrated to occur for other HDL-bound proteins: apoA-I (36), cholesteryl ester transfer protein (37), and lecithin:cholesterol acyltransferase (38). It has been shown to come from the tendency of hydrophobic domains of distinct subunits to autoassociate as a result of unmasking by bound lipids or detergents. The existence of exposed hydrophobic regions for HuPON1 cannot be firmly established without knowledge of the enzyme three-dimensional structure. However, when combining the Kyte-Doolittle hydropathy profile (39) with predictions for locations of transmembrane regions using TMpred (available on the World Wide Web at www.isrec.isb-sib.ch/ftp-server/tmpred/ www/TMPRED_form.html), we found that HuPON1 could have one transmembrane ␣-helical region corresponding to the first 19 amino acids from its N terminus. Therefore, this putative transmembrane ␣-helix, which actually corresponds to the retained N-terminal signal peptide of HuPON1, could be involved in the protein autoassociation process in nonmicellar solutions. Whether or not the HuPON1 oligomers formed in these conditions interact with detergent molecules has not been established. This could actually depend on whether the protein-protein interactions in the HuPON1 oligomers lead to total or only partial shielding of the exposed hydrophobic domains.
Micellar Solutions-In micellar solutions, sedimentation velocity and equilibrium indicated that the oligomeric state of C 12 E 8 -solubilized HuPON1 could change upon varying the detergent concentration. The most reliable model that fit the sedimentation equilibrium data was that of a monomer-dimer reversible equilibrium. The tendency of HuPON1 to autoassociate was qualitatively confirmed in the cross-linking experiments demonstrating the existence of large species of 80 -120-kDa apparent molecular mass. The number of C 12 E 8 molecules bound to the HuPON1 monomer and dimer was shown to be close to the aggregation number for this detergent (i.e. 100). Additionally, from the relationship between K d , the dissociation constant corresponding to the monomer-dimer equilibrium, and the detergent concentration, we determined that the dimerization process led to the liberation of ϳ100 molecules of C 12 E 8 . The results of cross-linking studies indicated that HuPON1 had fewer amino acids accessible and that its paired Cys residues were more specifically distant and oriented in micellar solutions when compared with nonmicellar solutions. These results suggest that the HuPON1 dimers formed in nonmicellar and micellar solutions are structurally distinct. In micellar solutions, detergent molecules associated with HuPON1 dimers could prevent direct protein-protein interactions between protected hydrophobic domains.
Bovine cytochrome c oxidase, which is a dimer within most crystals and assumed to be a dimeric complex in the membrane, was characterized in a variety of detergent solutions as a mixture of monomers and dimers and/or aggregates (33). This system presented a lot of similarities with the HuPON1 one. In The activity of the micelles of C 12 E 8 a micelle was assumed to be equal to their concentration and inferred from the solvent detergent concentration (see Table II). The linkage relationship between HuPON1 dissociation constant K d and the detergent activity provides the number of detergent molecules released upon dissociation, ⌬N det , obtained in units of detergent micelle (⌬N det /n, where n is the aggregation number of the micelle): Ѩ ln K d /Ѩ ln a micelle ϭ ⌬N det /n. The linear fit of the plotted data provides for ⌬N det /n a value of 1.1. particular, the relative amount of monomers and dimers was found to depend on the type and concentration of the detergent, whereas the monomeric form was functional for all of the activities evaluated. In a recent paper, it was found that the addition of bile salts stabilized the dimer form and caused, in appropriate conditions, complete reassociation of the monomer (40). It would be interesting to see whether, as it is the case for cytochrome c oxidase, a low amount of cholate stabilizes the dimeric form of HuPON1.

Anchoring Versus Shielding of a Large Hydrophobic Area by Detergent Molecules
At least two different models of detergent binding by HuPON1 can be considered. One could simply result from anchoring of the hydrophobic N-terminal signal peptide of HuPON1 to detergent aggregates. Assuming that the C 12 E 8 molecules bound to HuPON1 are organized as micelles, a reasonable model for the association process of HuPON1 in micellar media would thus consist of two HuPON1 monomeric forms, each associated with one detergent micelle leading to the formation of a HuPON1 dimeric form associated with one detergent micelle. In this model, the dimerization process would be associated with the release of one detergent micelle. At the present stage of knowledge, reviewed by le Maire et al. (41), detergent micelles may provide a suitable membrane-like environment around small membrane peptides. This anchoring could require a minimal size and specific shape of detergent micelles. Accordingly, the autoassociation of HuPON1 in nonmicellar media or media containing 6-kDa CHAPS micelles could result from an incapacity of HuPON1 to bind free detergent molecules or very small micelles. However, the more general model for detergent binding consists of a monolayer on the hydrophobic surface of the protein. In this model, for large proteins, the number of bound detergent molecules was related to the exposed hydrophobic surface. Can our results on HuPON1 give insight concerning the surface that is exposed upon dimer dissociation when the detergent concentration is increased? We calculated that about 100 molecules of released detergent with a cross-sectional area of 0.5 nm 2 would correspond to an exposed surface of 50 nm 2 (41). Bacteriorhodopsin is formed of seven trans-membrane helices and described to bind nearly the same amount (119 molecules) of C 12 E 8 (29). These arguments suggest that the hydrophobic exposed surface is quite large and would correspond to contact surfaces in addition to that of an entirely extended N terminus for the monomer. Thus, this would not be consistent with the structural data predicted from the HuPON1 amino acid sequence.

Effect of the Detergent Type
Assuming that HuPON1 binds to detergent molecules organized as micelles, the positive correlation of apparent Rh of the HuPON1-detergent complex with that of the detergent micelle could result from a higher affinity of the HuPON1 dimer, when compared with that of the monomer, for the largest micelles. Alternatively, the size of the interacting species could change with the detergent type, as has already been observed for the monomer of the integral membrane protein bacteriorhodopsin (42). In that case, the dimension of detergent-solubilized HuPON1 would be determined by the size of the free micelle.

Structural and Functional Similarities between HDL-and Detergent-bound Forms of HuPON1
Structural Features-In serum, the apolipoprotein apoA-I is involved in the assembly of phospholipid micelles, thus generating nascent HDL. Whereas large reconstituted apoA-I-HDL of 8.5-nm Rh were very efficient to release HuPON1 from transfected Chinese hamster ovary cells (10), in human plasma, PON1 has been shown by electron microscopy to be associated with an HDL subspecies of about 4.8 Ϯ 1.6-nm Rh (17). The latter Rh values are of the same order as those of PON1-detergent complexes, 3.8 -4.8 nm, as determined using nondenaturing PAGGE and size exclusion chromatography, suggesting that the detergent aggregate can at least partially mimic the hydrophobic environment of the protein bound to HDL although the HuPON1-detergent complexes characterized in this work lack protein components, such as apoA-I, apoJ For clarity of presentation, the absorbance of the 0.3 mg/ml injection is shown increased by a factor of 2. Hydrodynamic radii of 4.4 and 4.6 nm correspond to the elution of 0.3 and 1.2 mg/ml samples, respectively. Bottom panel, radioactivity measured on the 400-l recovered fractions. Aliquots of 100 l were counted for radioactivity. The most concentrated fractions were at 0.031 and 0.061 mg/ml, from micro-BCA assays. A negative level of radioactivity corresponding to a depletion of radioactive detergent is seen after HuPON1 peak for the 0.3 mg/ml sample. It is probably related to the fact that this protein sample comes from the early elution of the anion-exchange column (see the Detergent Exchange section under "Experimental Procedures"). Only the first part of this peak was considered for the calculation of the bound detergent. (43), HuPON3 (44), and lecithin:cholesterol acyltransferase, that have been shown to be specifically associated with HDL.
Interaction Selectivity of HuPON1 with Amphiphilic Aggregates-In serum, HuPON1 was exclusively found associated with HDL-type complexes, which differ in size and composition from the much larger low density and very low density lipoproteins of 20-nm and 30 -80-nm diameter, respectively (45). As hypothesized by Oda et al. (46), since the HuPON1-HDL lipoparticles characterized in plasma were mainly found associated with apoA-I (9, 17) and HDL-like complexes containing apoA-I were found to be necessary for optimal release and stabilization of HuPON1 (10), the association specificity of HuPON1 with HDL might rely on transient interaction of HuPON1 with apoA-I. However, interaction with apoA-I was not an absolute requirement for binding of HuPON1 to HDLlike complexes in apoA-I-deficient sera (9,47,48), to phospholipid micelles (9,10), and, as suggested by the present work, to detergent micelles.
Alternatively, results of a recent study from Deakin et al. (10) and this work suggest that the size of detergent or phospholipid complexes might be the predominant factor that affects the binding affinity and stabilization of HuPON1 in serum and in vitro. Thus, subpopulations of apoA-I-containing lipoparticles of specific size and shape could provide optimal hydrophobic environment for anchoring HuPON1 and stabilizing the enzyme activity. This interpretation was suggested for cholesteryl ester transfer protein, which also specifically binds to apoA-I-containing HDL (49).
From the already mentioned predictive program of transmembrane regions, TMpred, the HuPON1 homologue, HuPON3, is predicted to have one transmembrane ␣-helix at its N-terminal extremities. This suggests that, as predicted for HuPON1, HuPON3 could interact specifically with HDL through anchoring of an N-terminally located hydrophobic ␣-helix. Accordingly, the HuPON1 homologous HuPON2, which was expressed exclusively intracellularly (50), is not predicted to exhibit an N-terminal transmembrane domain. Apo A-I consists of several amphipathic ␣-helices that interact with the acyl chains of the lipid bilayers on the HDL side (51). The HDL-binding mode of HuPON1 and HuPON3 would thus be distinct from that of apoA-I and other apolipoproteins involved in recruiting phospholipids and generating nascent lipoparticles.
Functional Features-In apoA-I-deficient sera (9,47,48) or cell culture media (9,10,46), PON1 was found to be associated with relatively small and dense HDL-type complexes. The enzyme was active, but it displayed a specific activity toward phenylacetate significantly lower than that in apoA-I-containing media probably as a result of enzyme partial inactivation. Binding of HuPON1 to the larger apoA-I-containing HDL was thus assumed to be required for optimal specific activity and stability of the enzyme (9,10,46). This is consistent with preliminary results indicating that, in vitro, HuPON1 storage stability was much less important in nonmicellar solutions than that in micellar solutions and had very low stability in solutions containing small detergent micelles (i.e. CHAPS) compared with relatively large detergent micelles (i.e. Triton X-100) (current work in progress). As shown using nondenaturing PAGGE, enzyme inactivation in nonmicellar media could result from protein aggregation. This could also be related to aggregation and inactivation of recombinant HuPON1 when secreted in culture media by Hi-5 insect cells or when expressed in Escherichia coli (52). A practical consequence of this is that formulation of HuPON1 for a use as therapeutics in OP poisoning should include adequate concentration of biocompatible amphiphilic compounds that will preserve the enzyme stability on storage and activity in vivo. In addition, the low specific activity of HuPON1 in serum associated with certain pathologies, such as that reported for diabetes (53)(54)(55), might in fact be due to changes in HDL size and/or shape, which affect the binding affinity and, as a result, stability of the HuPON1 enzyme activity.

Choice of the Detergent Type and Concentration to Obtain a Homogeneous and Active HuPON1 Sample
Activity of HuPON1 Samples-In this work, we showed that the arylesterase activity of HuPON1 toward ␤-naphtyl acetate was preserved after electrophoresis in gels containing or not containing nondenaturing concentrations of the nonionic C 12 E 8 , C 12 -maltoside, Triton X-100, FOS-CHOLINE-12, or CY-MAL-6 detergent or the zwitterionic LDAO or CHAPS detergent. In addition, size exclusion chromatography strongly suggested that the HuPON1 monomers and dimers, in equilibrium in C 12 E 8 micellar solutions, exhibited the same arylesterasespecific activity. Thus, in the time scale of the experiments performed in this work, the activity of HuPON1 was preserved in a variety of detergent-solubilized forms, independently of the detergent concentration. However, as already mentioned, the stability of the enzyme activity could be strongly affected by the detergent type and concentration relative to the CMC.
Dispersity of HuPON1 Samples: Importance of the Protein and the Detergent Concentration-Nondenaturing PAGGE and cross-linking clearly showed a highly polydisperse HuPON1 when the detergent concentration was lower than the CMC. In micellar solutions, the apparent homogeneity of C 12 E 8 -solubilized HuPON1 observed by nondenaturing PAGGE and size exclusion chromatography is probably related to the fast interconversion between the monomeric and dimeric forms. Given the apparent homogeneity of HuPON1 in gels containing Triton X-100, C 12 -maltoside, FOS-CHOLINE-12, CYMAL-6, or LDAO, it is reasonable to assume that HuPON1 exhibits a similar monomer-dimer equilibrium model in micellar solutions of other nondenaturing detergents. In the presence of CHAPS above the CMC, HuPON1 was highly heterogeneous. This probably extends to media containing micelles of detergents not tested in the present work.
According to our data, HuPON1 dimers would be the main oligomeric species in solutions containing a low concentration of C 12 E 8 micelles in the solvent. For instance, from the linear relationship between K d and the micelle concentration (Fig. 6), we determined that extremely low concentrations of C 12 E 8 micelle of ϳ1 or ϳ0.1 M would be required to obtain 95% of HuPON1 dimers in solutions containing 20 or 2 M HuPON1, respectively. Conversely, the proportion of HuPON1 monomers increases with the C 12 E 8 concentration. We calculated that a very high C 12 E 8 micelle concentration of ϳ1 mM would be required to obtain 95% of HuPON1 monomers in a medium containing 20 M of HuPON1. However, there might be a critical C 12 E 8 concentration above which HuPON1 undergoes denaturation. Monomers and dimers coexist in most of the C 12 E 8 solutions, above the CMC. Thus, results of this study clearly showed that choice of the type and concentration of detergent relative to the CMC is crucial to obtain homogeneous HuPON1 samples suitable for crystallization trials. CONCLUSIONS Altogether, the results of this study indicate that the nonionic detergent-solubilized forms of HuPON1 in vitro are reasonable models of its in vivo lipid-bound forms. In this regard, nonionic detergent aggregates probably mimic the HDL environment for proteins such as PON1, which exhibits hydrophobic anchors, in a better way than the phospholipid bilayers for integral membrane proteins. However, to what extent the de-tergent can mimic the hydrophobic environment of the protein is a question that remains open. Consequently, the effective stoichiometry of HuPON1 in vivo cannot be definitively established from the present study. The results of the present work strongly suggest that assembly of amphiphilic molecules in aggregates of specific size and shape could be a prerequisite for anchoring of HuPON1. Consequently, the lower antioxidative capacity of HDL associated with certain diseases could be related to subtle changes of HDL size and/or shape, which directly affect the binding affinity of enzymes such as HuPON1, lecithin:cholesterol acyltransferase, cholesteryl ester transfer protein, and HuPON3, involved in the lipoprotein metabolism.