Role of the helical domain in fatty acid transfer from adipocyte and heart fatty acid-binding proteins to membranes: analysis of chimeric proteins.

The adipocyte and heart fatty acid-binding proteins (A- and HFABP) are members of a lipid-binding protein family with a beta-barrel body capped by a small helix-turn-helix motif. Both proteins are hypothesized to transport fatty acid (FA) to phospholipid membranes through a collisional process. Previously, we suggested that the helical domain is particularly important for the electrostatic interactions involved in this transfer mechanism (Herr, F. M., Aronson, J., and Storch, J. (1996) Biochemistry 35, 1296-1303; and Liou, H.-L., and Storch, J. (2001) Biochemistry 40, 6475-6485). Despite their using qualitatively similar FA transfer mechanisms, differences in absolute transfer rates as well as regulation of transfer from AFABP versus HFABP, prompted us to consider the structural determinants that underlie these functional disparities. To determine the specific elements underlying the functional differences between AFABP and HFABP in FA transfer, two pairs of chimeric proteins were generated. The first and second pairs had the entire helical domain and the first alpha-helix exchanged between A- and HFABP, respectively. The transfer rates of anthroyloxy-labeled fatty acid from proteins to small unilamellar vesicles were compared with the wild type AFABP and HFABP. The results suggest that the alphaII-helix is important in determining the absolute FA transfer rates. Furthermore, the alphaI-helix appears to be particularly important in regulating protein sensitivity to the negative charge of membranes. The alphaI-helix of HFABP and the alphaII-helix of AFABP increased the sensitivity to anionic vesicles; the alphaI-helix of AFABP and alphaII-helix of HFABP decreased the sensitivity. The differential sensitivities to negative charge, as well as differential absolute rates of FA transfer, may help these two proteins to function uniquely in their respective cell types.

Fatty acids (FA) 1 are major substrates for the synthesis of complex lipids and for energy production. Due to their limited solubility, specific carriers, known as fatty acid-binding proteins (FABP), are expressed in various tissues that use FA. Structural analyses of several FABPs have revealed markedly similar three-dimensional folds consisting of ten antiparallel ␤-strands that form a ␤-barrel, which is capped by two short ␣-helices arranged as a helix-turn-helix segment (1)(2)(3)(4). It is believed this ␣-helical domain, along with the ␤ C-D and D-E turns, functions as a "dynamic portal" that regulates FA entry and exit from the internal ligand binding cavity (5)(6)(7).
Heart FABP (HFABP) and adipocyte FABP (AFABP) are homologous proteins sharing Ͼ60% amino acid sequence identity (8). Both proteins bind one FA in the binding pocket (1,2), although x-ray crystallographic studies show that the FA adopts an entirely different conformation in their respective binding sites (9,10). AFABP has generally lower binding affinities for FA than HFABP (11), and it appears that HFABP shows greater affinity for saturated versus unsaturated fatty acids, whereas AFABP does not show such a preference (9). Aand HFABP also exhibit unique patterns of tissue distribution. AFABP is found in adipose tissue and monocytes/macrophages (12,13), whereas HFABP has a wider distribution, being found in muscle and numerous other tissues (14,15). Furthermore, HFABP has been suggested to be associated with FA ␤-oxidation in heart and skeletal muscle (16 -18), whereas AFABP may be involved in triacylglycerol storage and lipolysis (19). Thus, these two proteins are hypothesized to have distinct physiological functions. A-and HFABP were found in cardiac muscle cell in Antarctic Teleost fish and in different cell types of bovine mammary gland, suggesting further that the two proteins have specialized functions in the metabolism of fatty acids (20,21).
The potential role of FABPs in FA trafficking has been investigated in vitro using fluorescent anthroyloxy-labeled fatty acid (AOFA) and a resonance energy transfer assay (22)(23)(24)(25). AOFA transfer from both A-and HFABP to membranes appears to occur through a collisional process. The rates of FA transfer were increased by the incorporation of anionic phospholipids, decreased by chemical neutralization of lysines of FABP, and decreased by site-directed mutagenesis of specific lysines (23, 26 -28). Thus, electrostatic interactions between surface lysine residues and negatively charged membrane lipids were suggested to be involved in the collisional FA transfer process (26 -28). Recently, Fourier transform infrared spectroscopy, equilibrium binding studies, and competitive binding assays have also demonstrated direct interactions of AFABP with anionic phospholipid membranes (29,30). Despite their sharing the same transfer mechanism, however, AFABP and HFABP have several distinct transfer properties. For example, AFABP consistently has approximately 10-fold faster AOFA transfer rates (23), and AFABP also displays greater sensitivity to the negative charge of acceptor membranes, relative to HFABP. These two proteins also have a different response to ionic strength: An increase in ionic strength increases the rate of AOFA transfer from HFABP to vesicles, whereas no change in rate from AFABP is observed (31).
Lysine (K) to isoleucine (I) mutagenesis of specific lysines in the portal region have shown distinct effects for each protein on FA transfer (27,28). A unique lysine (Lys-32) found on the ␣II-helix of AFABP but not present in HFABP was shown to be an important determinant of the rate of FA transfer. Furthermore, a K59I mutation on the ␤ C-D turn of each protein resulted in opposite effects: A decrease in rates from AFABP to neutral phospholipid vesicles whereas a 2-fold increase was found in AOFA transfer from HFABP to membranes. In addition, neutralization of lysines 10, 97, and 113, located in the non-portal ␤-A, G, and I strands of AFABP, respectively, suggested that these residues may play a role in FA transfer; however, the corresponding lysine mutations on HFABP were found to have virtually no effect. Based on these unique properties, we hypothesized that the divergence in primary amino acid sequence and/or certain structural elements may underlie the functional differences between these two closely related proteins.
The aim of this study was to examine and compare the structural components or regions that are responsible for the differences in FA transfer properties between A-and HFABP. We generated two pairs of chimeric proteins. In one pair, the complete ␣-helical region, including the ␤-A strand, ␣I-helix, and ␣II-helix (residues 1-36) of the proteins were switched. In the other pair, the ␤-A strand plus the first ␣-helix (amino acids  were switched between A-and HFABP. Due to the somewhat different sensitivity to membrane-negative charge of the ␣I-helix lysines of A-and HFABP (27,28), and the different amino acid sequence of the ␣I-helices in the A-and HFABP, we also constructed point mutant E23S-AFABP. This particular mutation makes the hydrophilic side of the amphipathic ␣Ihelix of AFABP resemble that of HFABP. The results demonstrate that the entire helix-turn-helix motif is important in determining the absolute rate of FA transfer from protein to membranes. The ␣II-helix, in particular, may be the major structural element involved in the rate of FABP interactions with acceptor membranes. Furthermore, the ␣I-helices of the two proteins were shown to have different sensitivities to membrane-negative charge, which may help these proteins function uniquely in FA transfer.
Construction of Mutant Genes-Site-directed and chimeric mutant cDNAs were constructed. The numbering of amino acid residues was with the first methionine as Met-1. Oligonucleotide primers bearing mismatches, indicated by underlining, are listed in Table I. The resultant mutants with various structural elements are listed in Table II. The incorporation of the mutation was confirmed by sequence analysis (32). The AFABP point mutant having a glutamic acid 23 to serine mutation (E23S-AFABP) was generated using specific and outermost primers by overlapping PCR (33). A unique BalI site exists in the cDNA of AFABP at residue 37 next to the end of ␣II-helix (residue 36). A BalI site introduced to HFABP by overlapping PCR resulted in a threonine to alanine mutation (T37A-HFABP) to match the alanine 37 on AFABP. The mutated T37A-cDNA was subcloned into the pET-11d vector and digested by XbaI and BamHI. The digested fragment was further ligated into the pBS vector for chimeric DNA construction. Two vectors, pBS-AFABP and pBS-T37A-HFABP, were digested by BalI and BamHI to obtain four resultant fragments. By ligating different pieces, the ␣-helical region chimeric mutants ␣A/␤H-FABP and ␣H/␤A-FABP were constructed to contain residues 1-37 of AFABP plus 38 -133 of HFABP, and 1-36 of HFABP plus 37-132 of AFABP, respectively. To examine the effect of each helix on the transfer of FA to acceptor membranes, we generated a second pair of chimeric constructs. In this pair, the first 25 amino acids were exchanged between A-and HFABP. The helix-turn-

TABLE I
Oligonucleotide primers for mutagenesis Oligonucleotide primers carrying mutations or specific restriction sites were synthesized and further purified by anion-exchange high pressure liquid chromatography (HPLC) by Midland Certified Reagents (Midland, TX). All primers are listed from the 5Ј to the 3Ј end. HL06 and -09 were used for AFABP mutants, which contain XbaI and BamHI sites, respectively. HL19 and -22 were utilized for HFABP constructs bearing BamHI and NcoI sites individually. Antisense primer HL17 carries the silent mutation from TGT to GGT to generate a fragment that has overlap to amino acid residue 30 (Thr) of HFABP. Sense primer HL18 has a silent mismatch GGT to GGA to produce a fragment carrying Gly-25 of AFABP. Primers HL20 and -21 were used to produce the E23S mutant. A threonine to alanine mutation was introduced into HFABP using oligos HL23 and HL24. Primers FH16 and -17 were used to introduce DraIII and XhoI sites, respectively, into AFABP. helix region consists of helix I, from residues 17 to 24, followed by a short tripeptide, followed by the second helix (␣II) from amino acid residues 28 to 36. Thus, the first 25 amino acids from the N-terminal end essentially include the ␤-A strand and ␣I-helix. Chimeric ␣I-A/␤H-FABP containing residues 1-25 of AFABP and 26 -133 of HFABP was generated via overlapping PCR using a natural overlap region from amino acid residues 25 to 31 of A-and HFABP.
Another ␣I-helix chimeric, ␣I-H/␤A-FABP, was constructed earlier by a different approach. Without changing the amino acid sequence, a DraIII restriction site was introduced into the AFABP cDNA at residue 25. An XhoI site was also introduced after the cDNA stop codon for subsequent subcloning. The DraIII and XhoI-digested fragment consisting of residues 26 -132 of AFABP was ligated to the predigested pBS-HFABP fragment containing residues 1-25 of HFABP. Having confirmed the mutations by sequence analysis, chimeric ␣I-H/␤A-FABP cDNA was subcloned into the pET-11d expression vector using XbaI and BamHI sites.
Energy Minimization and Homology Modeling-The template structures upon which the models of the chimeric proteins were built were human FABP (PDB file 1HMS) (34), and mouse adipocyte FABP (1LID) (1). Both files were in the holoprotein form, each containing one molecule of oleic acid in the binding pocket. The oleate was removed to obtain the apoprotein form prior to modeling. Because the biochemical work was done with rat heart FABP, those amino acids in the human protein that differ from the rat sequence were replaced by the appropriate rat residues, and the conformational energy of the resulting molecule was minimized to obtain a new protein model in PDB format. This was done because there is no three-dimensional structure for rat protein, and the human FABP was the closest high resolution starting point.
All molecular modeling was done with the package Insight II (Molecular Simulations, Inc.) on Silicon Graphics workstations. Initial construction of the four chimeric proteins entailed appending the appropriate parts of the muscle and adipocyte structures to one another. This, of course, produces some close steric contacts and other conformational properties not characteristic of native globular proteins such as packing defects in the molecular interiors. These were removed by minimization of the conformational energy using the default parameters of Insight II. Titratable groups carried the charges, which the amino acids containing them would have at pH 7 as a reasonable approximation to the electrostatic environment of the native proteins. Solvent was not included. Several thousand iterations were required to achieve convergence for each protein. To enable the comparisons of the models with the native forms, the adipocyte protein, file 1LID, and the original human muscle file, 1HMS, were also subjected to full energy minimization after the oleate molecules had been removed. The minimized structures of the adipocyte and muscle proteins and those of the four chimeras were written out in PDB format for further processing.
To evaluate whether the models show structural characteristics of normal globular proteins, two properties were calculated, the solventaccessible surface areas and the packing volumes (35). The solventaccessible surface areas were calculated by the method of Lee and Richards (36) using the program ACCESS (35), and each atom was classified as aliphatic, aromatic, polar-uncharged, or polar-charged. The detailed classification of the atom has been described by Kajander et al. (37). The volumes occupied by the protein atoms were computed by the Voronoi algorithm as implemented by Richards (35,38).
Protein Purification and Control Analysis of Structural Integrity-Wild type A-and HFABP and mutant proteins were purified as previously detailed (27,30), and protein concentrations were estimated as described (11,27).
Several methods were used to examine the integrity of mutant FABP structure and binding properties, to ensure that alterations introduced did not change the overall protein conformation and/or its fatty acid binding properties. Circular dichroism spectra were measured to verify that there were no overall folding modifications in mutant FABP structures as previously described (28). To check the hydrophobic properties of the internal ligand binding pocket, fluorescence quantum yields (Q f ) of the fluorescent fatty acid analog 2-AP bound to wild type or mutant FABP were determined with an SLM 8000C fluorescence spectrophotometer as previously described (28). Binding of oleate to wild type Aand HFABP and mutant FABPs was analyzed using the fluorescent probe ADIFAB to obtain the FA binding affinity and stoichiometry (11).
Vesicle Preparation-Small unilamellar vesicles (SUV) were prepared by sonication and ultra-centrifugation, according to the method of Huang and Thompson (39). The standard vesicles were prepared to contain 90 mol % of EPC and 10 mol % of NBDPC. Two anionic vesicles were also used; 25 mol % of phosphatidylserine (PS) or cardiolipin (CL) was incorporated into the neutral SUVs in place of an equimolar amount of EPC. The phospholipid concentrations were determined as previously described (30).
Fluorescent FA Transfer Assay-The rate of 2-AP transfer from wild type and mutant FABPs to acceptor vesicles was determined using a fluorescence resonance energy transfer assay as detailed previously (28). AOFA transfer was monitored at 25°C. The conditions were developed prior to each experiment such that no photobleaching of FA was observed. Final concentrations in a typical transfer assay were 10 M FABP, 1 M 2-AP, and 100 -500 M acceptor phospholipid vesicles. AOFA transfer from protein to membranes was monitored and analyzed as a time-dependent decrease in fluorescence, as described (28). The rates of AOFA transfer are represented as means Ϯ S.E. from at least three or more separate sets of experiments.
Statistical Analysis-Two-tailed paired t tests were used to analyze the differences among rates of FA transfer from mutant proteins versus their respective wild type FABPs, to various concentrations and compositions of acceptor phospholipid membranes. Results were considered significant at p Ͻ 0.05.

RESULTS
Homology Modeling of Chimeric Proteins-The solvent-accessible surface areas and volumes are shown in Fig. 1. It is seen that the differences between the chimeras and the wild type proteins are small and comparable to those between the two proteins for which crystallographic structures are known.
All the values are well within the range seen for high resolution globular proteins of this size (37).
By way of further illustration, we superimposed the backbone atoms of the human muscle protein (PDB file 1HMS) after removing its oleate and minimizing its energy, onto those of the chimera whose ␤-body was derived from residue 38 -133 of the rat protein (␣A/␤H). The superimposition is shown in Fig. 2. The two molecules tracked one another quite closely. The backbone atoms used for the superimposition were the peptide N, C␣, peptide C, and peptide O for all residues. The root mean square difference (r.m.s.d.) for these 1048 atoms is 1.42 Å. Table III shows the r.m.s.d. values for all four chimeras after superimposition onto the protein from which the chimera's ␤-body was taken. They are all reasonably small given the numbers of atoms involved and given the close tracking shown in Fig. 2. For all chimeras, subtle differences from wild type FABP were found in many parts of the tertiary structure, rather than a single domain exhibiting large alterations.
Controls for Structural Integrity-It is critical to ensure that any significant experimental results were due solely to the specific alterations introduced. Several methods were used to examine the physical-chemical properties of mutant FABPs, circular dichroic spectroscopy, AOFA fluorescence quantum yield, and equilibrium binding of native FA to FABP.
Circular dichroic spectra showed that the mean residue ellipticity for ␣-helical content at 222 nm ( 222 ) for wild type Aand HFABP were approximately Ϫ8127 and Ϫ8623 deg cm 2 / dmol, respectively, representing 19 and 21% of the amino acids in ␣ helices. Furthermore, these two proteins displayed ϳ71 and 75% ␤-sheet secondary structure, in close agreement with their known tertiary structures (1, 2). Each mutant essentially   (Table IV). Overall, the two pairs of chimeric proteins as well as point mutant E23S-AFABP were found to maintain similar secondary structures in comparison with native A-and HFABP (Table IV).
The quantum yield (Q f ) of the fluorescent fatty acid analog 2-AP was determined to assess the hydrophobicity of the ligand binding cavity in wild type and mutant proteins, and the values are listed in Table IV. The observed quantum yields provide evidence that the binding site of HFABP is more hydrophobic than that of AFABP, in agreement with previous findings (42).   indicating that the FA binding affinity of these three proteins was slightly higher but not statistically different from wild type AFABP (110 Ϯ 11 nM). The calculated number of binding sites (n) was consistent with a 1:1 stoichiometry, in agreement with the x-ray crystal structure (1) and other ADIFAB studies (11). However, mutant ␣I-H/␤A-FABP, which has the first 25 amino acids from HFABP ligated to the ␤-body of AFABP, seemed to have "n" close to two (1.8 Ϯ 0.1). The oleate binding stoichiometries for native rat HFABP, the ␣A/␤H-, and the ␣I-A/␤H-FABP were also consistent with one binding site. In addition, the K d values obtained for mutants possessing the ␤-body of HFABP reflected a similar oleate binding affinity (31 Ϯ 6 and 14 Ϯ 4 nM) relative to the native HFABP (17 Ϯ 1). These results are consistent with the HFABP tertiary structure determined by Zanotti et al. (2) and with earlier reports of FA equilibrium binding affinities (11). Overall, the results indicate that mutants largely retain the properties of their "parent" FABP, i.e. their ␤-barrel domain. (28) and HFABP (27) strongly suggested that the helical domain is the primary region involved in the effective collisional mechanism of FA transfer from FABPs to model membranes. Therefore, the present studies, in which we exchanged the entire helical cap between the two proteins, were designed to examine the structure-function relationships in this domain. The rates of 2-AP transfer from the A-and HFABP and the first pair of chimeric proteins (␣A/␤H-and ␣H/␤A-FABP) to SUVs are shown in Fig. 3. As previously found, A-and HFABP transferred FA via a collisional mechanism, as seen by the increase in the absolute rates of 2-AP transfer as a function of acceptor vesicle concentration, as well as the modulation of 2-AP transfer to SUVs containing negatively charged phospholipid relative to neutral SUVs. Based on their qualitatively similar regulation of AOFA transfer rates, the results indicate that the two chimeric proteins retain the collisional FA transfer mechanism. Notably, as previously observed, 2-AP transfer from AFABP is markedly faster than from HFABP (23).

Effect of the Helix-Turn-Helix Motif on FA Transfer-Studies of lysine to isoleucine point mutants of AFABP
The rates of FA transfer from ␣H/␤A-FABP were compared with those from wild type AFABP, to both neutral and acidic vesicles (Fig. 3). For all vesicle compositions, 3-to 4-fold reductions in rates were found from the chimeric protein relative to AFABP. Likewise, replacement of the ␣H 3 ␣A helical domain on HFABP with ␣A (resultant chimeric ␣A/␤H-FABP) led to 3to 4-fold increases in 2-AP transfer rates relative to those of wild type HFABP. These results indicate that the helical segment plus the ␤-A strand (first 36 amino acids) of AFABP is an important determinant of the more rapid rate of FA transfer from AFABP compared with HFABP.
The fact that the helix-domain chimera did not entirely convert the rate of 2-AP transfer to that of its respective FABP indicated that the ␤-body of AFABP also aided in determining the absolute rates of ligand transfer to membranes. When comparing chimeric ␣A/␤H-FABP with native AFABP, i.e. proteins sharing the same cap region, a 2-to 3-fold reduction was found in the rate of FA transfer to both neutral and acidic vesicles. Similarly, a comparison of ␣H/␤A-FABP with wild type HFABP reveals a 3-to 4-fold increase in AOFA transfer rates from the chimeric relative to wild type HFABP. This agreed with our earlier Lys 3 Ile point mutagenesis studies, which suggested that several ␤-strands of AFABP were involved to a modest extent in the collisional transfer process (28). Although the changes in absolute rates of FA transfer were relatively small (20 -40%) from individual site-directed mutant proteins (28), the severalfold changes observed in the present chimeric studies likely reflect the collective effects of several residues in the ␤-body.
Effect of Single Helix Domains on FA Transfer-The helical lid consists of two short ␣ helices, from residues 17-24 and 27-35, and a small turn (Gly-Val-Gly) that links them together. The first helix of A-and HFABP is an amphipathic helix whereas the second one of both proteins is not. Two differences in amino acid residues are found between A-and HFABP in each helix. In the ␣I-helix, glutamic acid 23 and valine 24 are found in AFABP, whereas serine 23 and leucine 24 are present in HFABP. In the second helix, lysine 32 and glycine 35 exist in AFABP, but HFABP contains the corresponding glutamine 32 and serine 35. Notably, lysine (Lys-32), which displayed a unique effect on FA transfer, exists in the ␣II-helix of AFABP, whereas no lysine is found in the same region of HFABP. Neutralization of Lys-32 resulted in increased rates to neutral vesicles but decreased rates to acidic membranes relative to the rates of 2-AP transfer from wild type AFABP (28). Hence, we sought to distinguish the different contributions of each helix to the FA transfer process. Thus, two additional chimeric mutants were constructed in which the first 25 amino acids were exchanged (␤-A strand and ␣I-helix) between the two proteins.
The rates of 2-AP transfer from this pair of mutants relative to the two native proteins are shown in Fig. 4. As seen by the increases in transfer rates with acceptor membrane concentration, a collisional process was maintained for these two chimeric proteins (results for ␣I-A/␤H-FABP shown in the inset). Qualitatively, the pattern of FA transfer from these two mutants to neutral vesicles was similar to the ␣A/␤Hand ␣H/␤A-FABP chimera, but the effects were of lesser magnitude. Adding the ␣I-helix of HFABP to the remaining structure of AFABP (␣II-helix and ␤ B-J strands), to obtain mutant ␣I-H/ ␤A-FABP, resulted in a 2-fold reduction in rates relative to wild type AFABP. Changing the ␣I-helix from HFABP to AFABP, to obtain mutant ␣I-A/␤H-FABP, showed small but statistically significant increases in the transfer rates (13-25%) to 250 and 500 M neutral vesicles relative to native HFABP (p Ͻ 0.05 and 0.01, respectively) (Fig. 4, inset).
In contrast to the pattern observed for the ␣-helical (␣I ϩ ␣II) domain chimera, a completely opposite trend was found for 2-AP transfer from the ␣I chimera to anionic vesicles (Fig. 4). Transfer rates from these mutant proteins were not between the values for wild type A-and HFABP. Instead, the rates were either 2-fold faster for the ␣I-H/␤A-FABP, or 2-fold slower for the ␣I-A/␤H-FABP, compared with their respective wild type proteins sharing the same ␣II-helix plus ␤-barrel structure (p Ͻ 0.05). For instance, mutant ␣I-H/␤A-FABP transferred 2-AP to 500 M PS-vesicles at a rate of 94 Ϯ 3.8 s Ϫ1 , whereas the rate of 2-AP transfer from native AFABP ("␣I-A/␤A-FABP") was 53.4 Ϯ 3.9 s Ϫ1 . Similarly, when compared with native HFABP, ␣I-A/␤H-FABP transferred FA at 2-fold slower rates to PS-incorporated membranes. For example, the rates of FA transfer to 100 M PS vesicles were 0.45 Ϯ 0.02 s Ϫ1 and 0.90 Ϯ 0.04 s Ϫ1 from ␣I-A/␤H-FABP and HFABP, respectively (inset of Fig. 4). The same pattern was found in transfer to CL-containing membranes, although, as for all proteins examined, faster absolute rates were observed (Fig. 4). The transfer rates from mutant ␣I-H/␤A-FABP were faster when compared with those from wild type AFABP, however, the differences were statistically significant only when FA transferred to 100 M CL-SUVs. This may be due to the very rapid rates of FA transfer to acceptor CL membranes, which approach the limit of detection range by the stopped flow mixing chamber and therefore may lead to large variability. Collectively, these results suggest that the ␣Iand ␣II-helices play distinct roles in the regulation of collision-mediated fatty acid transfer from Aand HFABP and further suggest that helix ␣I of HFABP has a greater sensitivity than that of AFABP to the negative charge of acceptor phospholipid membranes.
Comparison of the rates of 2-AP transfer from native AFABP with those from chimeric ␣I-A/␤H-FABP, as well as native HFABP with mutant ␣I-H/␤A-FABP, allows us to examine the impact of the ␣II-helix plus the ␤-body of the two proteins in the transfer process. Rate 3-to 6-fold faster were found for proteins that have the ␣II/␤-body from AFABP, implying that ␣-II helix plus ␤ B-J strands are the primary determinants of the absolute rates of FA transfer. Moreover, these results suggest that ␣I-helix is important in determining the sensitivity to negative charge for ligand transfer from A-and HFABP to membranes.
Effect of Glutamic Acid 23 to Serine Mutation on FA Transport-A higher sequence identity (77.5%) between A-and HFABP is found in the first 40 amino acid residues relative to the overall protein (62%). Among those residues, the first helix from both proteins is amphipathic, and such helices are thought to be involved in protein-membrane interactions (43)(44)(45). Six out of eight residues in the ␣I-helix are identical between adipocyte and heart FABP, with one different residue on the hydrophobic side (valine and leucine 24 on A-and HFABP, respectively) and one on the hydrophilic side (glutamic acid and serine 23 on A-and HFABP) of the helix. To mimic the hydrophilic side of the ␣I-helix of HFABP, an E 3 S mutation was introduced into AFABP at residue 23, changing the negatively charged glutamic acid to the neutral polar serine. Having verified that this mutation did not alter the conformational integrity and binding properties of the protein relative to native AFABP (Table IV), 2-AP transfer to acceptor membranes was examined.
Like other mutants studied so far, FA transfer from E23S-AFABP to model membranes occurs via a collisional process. Unlike the dramatic effects of lysine to isoleucine mutations in the helical regions (28), the E 3 S alteration did not result in large differences in rates of FA transfer. On average, 10 -15% decreases in transfer rates, which did not reach statistical significance, were found relative to native AFABP, both to neutral and acidic vesicles (data not shown). Furthermore, little change (5-10% decrease) was found in the sensitivity of mutant E23S-AFABP to negatively charged vesicles compared with wild type AFABP. These results indicate that changing of negatively charged glutamic acid to the uncharged but polar serine did not convert the transfer characteristics from those of AFABP to those of HFABP. This suggests that the charge of the glutamic acid may not be its significant property, but, rather, the polarity of the residue may be necessary for maintaining the amphipathic character of the helix for protein-membrane interactions.
Sensitivity of Chimeric Mutants to Negative Charge-As expected, 2-AP transfer rates were faster from all proteins examined to acceptor membranes containing 25 mol % of the negatively charged phospholipids phosphatidylserine and cardiolipin, relative to neutral vesicles. Transfer of 1 M 2-AP from 10 M protein to 100 M SUV is shown as a representative for the relative effect of membrane charge on FA transfer rate (Fig. 5). AFABP showed a large response to negative charge, with rates increasing ϳ5and 40-fold to PS and CL vesicles, respectively, relative to neutral membranes. HFABP demonstrated increases of about 2-and 20-fold to PS-and CL-SUV, respectively. Those proteins with the faster absolute transfer rates also showed more sensitivity to membrane-negative charge. For example, the greatest sensitivity to negative charge was found in mutant of ␣I-H/␤A-FABP, whereas ␣I-A/␤H-FABP was the least sensitive protein (Fig. 5). Mutant ␣I-H/␤A-FABP showed increases about 22.5-and 137-fold to PS-and CL-SUV, respectively, and displayed the highest 2-AP transfer rates observed (Fig. 4). These increases are ϳ3.5and 7-fold greater in sensitivity to anionic vesicles compared with those of wild type A-and HFABP. In contrast, mutant ␣I-A/␤H-FABP displayed only 1.5-and 8.5-fold increase in rates of FA transfer to PS and CL vesicles, respectively, and showed the slowest 2-AP transfer rates (Fig. 4). This protein had only 20 and 40% of the sensitivity to anionic phospholipids compared with native AFABP and HFABP, respectively.
These results suggested that mutant ␣I-H/␤A-FABP may contain the structural elements from each FABP protein that provide increased sensitivity to phospholipid membrane-negative charge, i.e. the ␣II-helix and ␤-barrel of AFABP, and the ␣I-helix of HFABP. In contrast, mutant ␣I-A/␤H-FABP may consist of structural components that are relatively less sensitive to phospholipid-negative charge. The chimeric proteins having their entire helical domains (residues 1-36) exchanged displayed sensitivities between those of AFABP and HFABP (left panel in Fig. 5), thus they appear to contain a mixture of sensitive and insensitive elements.

DISCUSSION
AFABP and HFABP each display a remarkable N-terminal sequence identity among different species (46,47). Indeed, AFABP isolated from five different mammals (murine, rat, human, porcine, and bovine) is identical from position 1 to 39 of the N-terminal end (46). Although the exact importance of the sequence identity at the N terminus is not yet understood, it may include the maintenance of specific collisional fatty acid transfer properties. It is well established that ␣-helical motifs are often involved in protein-membrane interactions (43). Hence, we focused on the N terminus helix-turn-helix motif, believed to serve as a portal for FA entering and exiting the binding pocket. The present results support a primary role for the ␣-helical domain in dictating both the overall rate of FA transfer, as well as the effective electrostatic interactions between A-and HFABP and membranes. This is in agreement with our recent studies in which deletion of the helix-turn-helix domain of intestinal FABP (IFABP) resulted in the loss of virtually all sensitivity to membrane charge and acceptor vesicle concentration (48).
The importance of the helical region of the FABP's in putative protein-membrane interactions is supported by a recent study by LiCata and Bernlohr (49), who used structure-based calculations to analyze the surface properties of a series of apoand holo-FABPs. The surface of AFABP was shown to have a strong ridge of positive electrostatic potential across the "top" of the molecule (across the helix-turn-helix motif), which extends to include the opening to the binding cavity. In addition, upon ligand binding, subtle but consistent changes were found only in the shape of the large positive potential contour across the top of the protein, and no significant changes were observed in the electrostatics outside this positive ridge. It is possible that this positive cap on the holo-protein directs the effective electrostatic collision between AFABP and negatively charged membranes. We hypothesize, in addition, that this domain may also direct the removal of fatty acids from membranes by AFABP, similar to what we have recently found for IFABP (50).
Neither of the two helical region mutants was able to convert the properties of wild type AFABP completely to HFABP, and vice versa, indicating that the helical region is not the only structural element associated with the FA transfer process. The impact provided by the barrel structure of AFABP was not negligible, and thus it may also be involved in FA transport.
Results here are in agreement with our previous site-directed mutagenesis studies of AFABP (28), which indicated that the lysine residues located in ␤-strands A, G, and I (␤-body) also participate to some extent in the electrostatic interactions between the protein and phospholipid membranes.
Each helix appears to have unique attributes in transporting FA to membranes. A comparison of the ␣I chimera and the "␣I ϩ ␣II" chimera suggests that the ␣II-helices are the primary determinants of absolute rates of AOFA transfer from A-and HFABP to anionic acceptor membranes (Figs. 3 and 4). To further dissect the contribution of each helix, we directly compare the impact of the helical domains on 2-AP transfer rates from the FABPs (Fig. 6). Proteins that share the body of AFABP or HFABP are shown in the left and right panels, respectively. Taking into consideration that (a) the entire helical region of AFABP has a greater sensitivity to negative charge than that of HFABP (␣A Ͼ ␣H), and (b) the ␣I-helix (plus ␤-A strand) of HFABP was more sensitive to negative charge than that of AFABP (␣I-H Ͼ ␣I-A), the ␣II-helix of AFABP must be very sensitive to anionic vesicles relative to that of HFABP to overcome the contribution by helix I (␣II-A Ͼ ␣II-H). Collectively, the ␣I-helix (plus ␤-A strand) of HFABP, and the ␣II-helix of AFABP, displayed the greatest ability in sensing negative charge of acceptor phospholipid membranes. The overall abilities of AFABP and HFABP to respond to acceptor membrane charge, therefore, appears to be an average function of the ␣Iand ␣II-helix sensitivities.
HFABP and AFABP both contain six hydrophilic residues in the helical cap region, however, differences in these residues could result in different properties of FA transfer to neutral versus anionic membranes. In particular, the hydrophilic side of helix I of AFABP has more available negatively charged side chains than that of HFABP, and may therefore interact with positive charges on the zwitterionic phosphatidylcholine trimethylamino groups, resulting in faster AOFA transfer rates to zwitterionic SUVs. In contrast, for transfer to negatively charged vesicles, the greater number of negative charges on the ␣I-helix of AFABP may result in a more repulsive interaction, resulting in slower FA transfer from AFABP relative to ␣I-H/ ␤A-FABP. This speculation does not contradict the proposal that the helix-turn-helix cap of AFABP has a higher positive electrostatic potential than that of HFABP (49), because a unique lysine (Lys-32) located on the ␣II-helix of AFABP, re- sults in an overall greater positive potential for the entire cap relative to HFABP.
We have hypothesized that the FA collisional transfer mechanism is likely to be a multistep process that includes electrostatic interactions between the entire helix motif with target phospholipid membranes, a resultant conformational change in the putative cap from an ordered closed state to a more disordered open one, long chain fatty acid release from the internal ligand binding cavity, and finally, the association of FA with the acceptor membrane (48). The present investigations suggest a modified model, in which the first step may be directed by the ␣II-helix alone (or perhaps with some contributions of the ␤-body) to affect a close proximity to acceptor phospholipid membranes. Second, the entire helical domain may interact with phospholipid membranes by electrostatic interactions, causing a conformational change in this domain from the ordered closed state to a disordered open state (6,7,51). FABP then releases its bound ligand from the binding cavity, and finally the FA associates with the acceptor membrane. Cistola and Hodsdon (6) have suggested that the binding of fatty acid shifts the order-disorder equilibrium of FABP toward the ordered state by stabilizing a series of cooperative interactions resembling a C-terminal helix capping box (C cap), in the region that links the ␣II-helix and the ␤-B strand of intestinal FABP (6). We hypothesize that, upon interacting with membranes at the ␣-helical region, FABP may shift from an ordered state to a disordered state, which allows the FA to be released from FABP. Unlike the typical C cap, with glycine or proline in the location next to the last residue of the ␣-helix (52, 53), Aand HFABP do not have such a motif to stabilize their ␣IIhelices. Therefore, as proposed for IFABP (6), A-and HFABP may also use side-chain and/or backbone amide hydrogen bonds and hydrophobic interactions to stabilize the helical lid. Our results suggest that the modified helices result in altered interactions between the ␣II-helix and the ␤ C-D turn, leading to differences in FA transfer rates. However, these presumably subtle conformational changes cannot yet be defined, due to the absence of actual three-dimensional structures for the chimeric A/HFABPs. Equilibrium binding studies have shown that fatty acid binding affinities are higher for HFABP than for AFABP (11). Thus, the release of FA from the binding cavity may play a role in the absolute rates of FA transfer to membranes. Woolf and Tychko (54) suggested that the extended form of ligand bound in the AFABP cavity, compared with the bent form in the HFABP pocket, contributed to a more rapid release from AFABP. In our assay system, the fluorescence signal is lost when the AOFA leaves the relatively hydrophobic FABP binding pocket (31). Although this is suggested to occur during the final step of the proposed multistep process, the rate at which the fluorescence quenching occurs will reflect the slowest step that occurs prior to or including the desorption/membrane association. Because the present mutations resulted in little or no alterations in binding site properties or equilibrium binding affinities, it is likely that the FA release from the binding pocket is not the rate-limiting step in the FA transfer process, although proportional changes in off-and on-rates cannot be ruled out.
In summary, these studies show that the FABP ␣II-helix is important in determining the absolute FA transfer rates to membranes; the ␣I-helix appears to be central in regulating protein sensitivity to the negative charge of membranes. Within the cell, A-and HFABP may interact with specific domains on phospholipid membranes, and/or specific domains on target proteins. Several studies have, indeed, provided experimental support for this hypothesis (55,56). Based on our in vitro studies, it is suggested that electrostatic interactions, between cationic FABP surface residues and anionic functional groups on membranes or proteins, are involved in targeted intracellular movement of FA. The differential sensitivities of A-and HFABP to negative charge appear to be largely a combinatory effect from both their ␣Iand ␣II-helices. The differential FA transfer properties may result in unique physiological functions of these proteins in their respective cell types, resulting, for example, in FA targeting to sites of FA storage or oxidation.