Insight into Lipid Surface Recognition and Reversible Conformational Adaptations of an Exchangeable Apolipoprotein by Multidimensional Heteronuclear NMR Techniques*

Apolipophorin III (apoLp-III) from the insectManduca sexta is a 166-residue (M r18,340) member of the exchangeable apolipoprotein class that functions to stabilize lipid-enriched plasma lipoproteins. In the present study, we present the secondary structure and global fold of recombinant apoLp-III derived from three-dimensional heteronuclear NMR spectroscopy experiments. Five discrete α-helical segments (21–30 residues in length) with well defined boundaries were characterized by four NMR parameters: medium range nuclear Overhauser enhancement contacts between proton pairs, chemical shift index, coupling constants, and amide proton exchange rates. An antiparallel arrangement of helical segments has been obtained based on the long range interhelical nuclear Overhauser enhancement contacts. The NMR solution structure reveals a globular, up and down helix bundle organization similar to that of Locusta migratoriaapoLp-III (Breiter, D. R., Kanost, M. R., Benning, M. M., Wesenberg, G., Law, J. H., Wells, M. A., Rayment, I., and Holden, H. M. (1991)Biochemistry 30, 603–608). However, a short helix (comprised of 5 amino acids) has been identified in the region between helix 3 and helix 4. This helix is postulated to play a role in lipid surface recognition and/or initiation of binding. Our results also indicate the existence of buried polar and charged residues in the helix bundle, providing a structural basis for the relatively low stability of apoLp-III in its lipid-free state. It is suggested that the intrinsic low stability of lipid-free apoLp-III may be important in terms of its ability to undergo a reversible, lipid binding-induced, conformational change. This study underscores the striking resemblance in molecular architecture between insect apoLp-III and the N-terminal domain of human apolipoprotein E. The potential for application of NMR techniques to studies of the exchangeable apolipoproteins, possibly in their biologically active, lipid-associated state, has broad implications in terms of our understanding of the molecular basis of their physiological functions.

Plasma lipoproteins are stable noncovalent assemblies of lipids and specialized proteins termed apolipoproteins. The apolipoproteins can be broadly categorized as exchangeable or nonexchangeable. Whereas nonexchangeable apolipoproteins are integral, nontransferable components, the exchangeable apolipoproteins stabilize lipoproteins through reversible binding as a function of particle surface defects or exposed hydrophobic regions. Specific exchangeable apolipoproteins are known to regulate lipoprotein metabolism through modulation of lipid metabolic enzymes or by functioning as ligands for cell surface receptors. To date, approximately 11 distinct human plasma-exchangeable apolipoproteins have been isolated and characterized (1). The capacity to form a stable interaction with lipid surfaces represents a common function shared by all exchangeable apolipoproteins and is essential for the expression of their alternate functions. Plasma levels of certain exchangeable apolipoproteins are correlated with several human diseases including dyslipidemia and cardiovascular disease (2). For example, overexpression of apolipoprotein E (apoE) 1 in transgenic mice results in a decrease in plasma cholesterol (3), whereas apoE knockout mice exhibit massive accumulation of remnant lipoproteins, leading to premature atherosclerosis (4,5).
Structural studies on exchangeable apolipoproteins indicate that the amphipathic ␣ helix provides an essential structural motif for lipoprotein binding (6). A major advance in our understanding of exchangeable apolipoprotein structure comes from studies using x-ray crystallography to determine the three-dimensional structures of the 22-kDa N-terminal domain of human apoE (7) and apolipophorin III (apoLp-III) from the insect Locusta migratoria (8). These structures, which were determined in the lipid-free state, share a similar molecular architecture comprised of a bundle of elongated amphipathic ␣ helices. Whereas the N-terminal domain of human apoE exists as a four-helix bundle, L. migratoria apoLp-III is organized as a five-helix bundle. The helices orient their hydrophobic faces toward the center of the bundle while their hydrophilic faces interact with the solvent. It has been postulated that association with a lipid surface triggers a dramatic conformational change wherein the loop regions between ␣-helical segments function as hinges about which the molecule opens to expose its hydrophobic interior, which then becomes available for lipid binding (8,9). To date, no lipid-bound structure of an exchangeable apolipoprotein is available.
Given recent advances in the application of NMR spectroscopy to protein solution structure determination, this technique should be applicable to studies of exchangeable apolipoproteins. It has been shown by others that NMR techniques can be used to determine three-dimensional structures of proteins up to 30 kDa (10). Whereas most exchangeable apolipoproteins meet this size limitation, their tendency to selfassociate in the absence of lipids hinders the application of NMR methods. Several investigators have used two-dimensional 1 H NMR to study apolipoprotein peptide fragments in the range of 15-36 amino acids (11)(12)(13)(14). To date, however, no NMR solution structures of intact exchangeable apolipoproteins have been reported.
NMR structure determination is attractive because it can be used to study the conformations as well as the dynamic changes of exchangeable apolipoproteins in both lipid-free and lipid-bound states. In considering possible candidate apolipoproteins for solution structure determination by NMR, we found that apoLp-III (M r 18,300) meets several essential criteria. Although the x-ray crystal structure of L. migratoria apoLp-III has been solved (8), there are no reports of expression of this protein in bacteria, possibly due to the presence of complex carbohydrate moieties (15), which may be important for protein stability in the lipid-free state (16). On the other hand, the homologous apoLp-III from the sphinx moth, Manduca sexta, lacks covalent oligosaccharide moieties (17), is monomeric over a broad concentration range (18), and has been overexpressed in bacteria (19). Although the sequence identity between these two apoLp-IIIs is Ͻ30%, the two proteins are functionally indistinguishable (20).
The goal of our research is to understand the molecular basis of reversible conformational adaptations of the exchangeable apolipoproteins upon lipid binding. As the first step, we report here the solution secondary structure and global fold of lipidfree apoLp-III derived from three-dimensional heteronuclear NMR experiments. To our knowledge, this represents the first solution structure determination of an intact exchangeable apolipoprotein. Furthermore, it represents one of the largest ␣-helical proteins whose structure has been solved using NMR techniques (21). The NMR structure of M. sexta apoLp-III provides new insight into the mechanism of apoLp-III lipid surface recognition/binding and provides a structural rationale for the reversibility of its interactions with lipoprotein surfaces.

Expression, Purification, and Isotope Labeling of Recombinant
ApoLp-III-Bacterial expression of recombinant apoLp-III was carried out as described by Ryan et al. (19). For isotopic labeling experiments, a saturated overnight cell culture (grown in 2 ϫ yeast tryptone medium at 37°C) was diluted 1:100 (v/v) in M9 minimal medium containing 0.1% 15 NH 4 Cl (Isotec, Miamisburg, OH) and/or 0.15% D-[6-13 C]glucose (Isotec) as the sole nitrogen and/or carbon source with 50 g of ampicillin/ml and cultured at 30°C. The initial A 600 values for cell cultures upon dilution into minimal media were typically 0.06 -0.09. After 4 -5 h, when the A 600 reached 0.7-0.9, protein expression was induced by the addition of isopropyl-␤-D-thiogalactopyranoside (1 mM final concentration). As reported by Ryan et al. (19), significant amounts of recombinant apoLp-III specifically accumulate in the culture medium following induction. Recombinant apoLp-III is recovered from the cell culture by pelleting the cells and collecting the supernatant. Upon concentration to Ͻ50 ml, the culture supernatant was dialyzed against deionized H 2 O for 48 h and lyophilized. Protein was dissolved in water and subjected to high pressure liquid chromatography on a preparative Zorbax C 8 reversed phase column. The purity of the sample was assessed by SDS-polyacrylamide gel electrophoresis, and the relative isotope enrichment was characterized by electrospray ionization mass spectrometry. Using the protocol described above, 70 -100 mg of 13 C, 15 N double-labeled or 120 -150 mg of 15 N single-labeled apoLp-III was obtained per liter of cell culture with Ͼ99% purity and an isotope enrichment of Ͼ95%.
NMR Sample Preparation-Previous denaturation studies of lipid-free apoLp-III, monitored by circular dichroism spectroscopy, revealed a midpoint of guanidine HCl-induced denaturation of 0.35 M, corresponding to a ⌬G D H2O of 1.29 kcal/mol (22). Consistent with these results, a time course study of apoLp-III stability monitored by 1 H/ 15 N HSQC experiments showed that apoLp-III in H 2 O at pH 7.0 was stable for Ͻ16 h. New peaks that correspond to an unfolded state of the protein appeared at longer times (data not shown). Empirical stability optimization studies monitored by NMR resulted in the finding that apoLp-III was stable for Ͼ2 weeks in 250 mM potassium phosphate, 0.5 mM sodium azide, pH 6.4 -6.5. In this buffer, the chemical shift dispersion in both the 1 H and 15 15 N]apoLp-III was employed for three-dimensional CBCA-(CO)NNH, HNCACB, HCCH-TOCSY, and simultaneous 15 N-and 13 Cedited NOESY experiments. All samples contained a small amount of 2,2-dimethyl-2-silapentanesulfonic acid as an internal standard for proton chemical shifts.
NMR Spectroscopy-All NMR experiments were carried out at 30°C on a Varian Unity 600 NMR spectrometer equipped with three channels, a pulsed field gradient triple resonance probe with a z gradient, and a gradient amplifier unit. Carrier positions used in the various three-dimensional NMR spectra were as follows: 15  Triple resonance three-dimensional NMR spectra correlating backbone amide with ␣ and ␤ carbons were acquired on a uniformly 15 15 N-Edited NOESY (23,30) and 15 N-edited TOCSY (23, 31) experiments were recorded on a uniformly 15 N-labeled sample of apoLp-III in H 2 O using the enhanced sensitivity pulsed field gradient method (32,33). 15  For backbone assignment of apoLp-III, 15 N-edited NOESY and 15 Nedited TOCSY spectra were used mainly due to the ␣-helical nature of the protein. HNCACB and CBCA(CO)NNH spectra were used to confirm the backbone atom assignment obtained from 15 N-edited NOESY and 15 N-edited TOCSY spectra as well as to assign those residues that could not be assigned due to missing NH-NH connectivities in 15 (34). Side-chain atoms were assigned using HCCH-TOCSY, 15 N-edited NOESY, 15 N-edited TOCSY, and 13 Cedited NOESY spectra.
All spectra were processed on SUN workstations using NMRPipe software from Delaglio et al. (35). The spectral assignment was achieved using Pipp software (36). Postacquisition solvent subtraction was employed in spectra where NH protons were detected in the acquisition dimension (37). Typically, spectra were processed in the acquisition and indirect dimensions with 90°-shifted sine bell-squared apodization. For constant time 15 N/ 13 C evolution periods, mirror image linear prediction was used to double the time domain signals (38). A time domain deconvolution procedure (33) was used to minimize the signal from residual water for simultaneous 15 N-and 13 C-edited NOESY experiments.
torsion angles were obtained from either HNHA and/or HMQC-J experiments. In HMQC-J experiments, few cross-peaks were split (Ͻ20), indicating small coupling constants for most residues. Coupling constants were obtained from the ratio of the peak intensity from the diagonal to the ␣H-NH cross-peak in HNHA spectra for a given residue (39). Amide proton exchange rates were estimated by the intensity of the water peak in each residue strip of 15 N-edited NOESY spectra. For residues with no water peak in the strip, slow exchange with solvent is concluded. In cases of weak intensity and medium or strong intensity water peaks in the strips, medium and fast proton exchange, respectively, is concluded.

RESULTS
Assignment Strategy-For ␣-helical proteins in the 150-amino acid size range, the overall chemical shift dispersion is generally poor, especially in the 1 H dimension. In 1 H/ 15 N HSQC spectra of apoLp-III, the amide proton chemical shift dispersion is Ͻ2.5 ppm, with significant spectral overlap observed in the range of 7.5-8.5 ppm. By contrast, the chemical shift dispersion in the 15 N dimension is relatively large, improving the overall cross-peak separation. Two independent methods were used to obtain the assignment of apoLp-III. One is through the amide-amide proton connectivities in the threedimensional 15 N-edited NOESY together with the information obtained from three-dimensional 15 N-edited TOCSY; the other is through the C ␤ (i Ϫ 1)-C ␣ (i Ϫ 1)-N(i)-NH(i) and C ␤ (i)-C ␣ (i)-N(i)-NH(i) connectivities in CBCA(CO)NNH and HNCACB experiments. In NOESY spectra of an ␣-helical protein, strong/ medium intensity d NN (i, i Ϯ 1) NOEs can usually be found (40). These d NN NOEs, which provide sequential connectivities, were used for backbone atom assignment. In cases of overlap in the NH-NH region, [ 15 N]leucine-, [ 15 N]lysine-, and [ 15 N]valine-specific labeled apoLp-III 1 H/ 15 N HSQC spectra provided useful additional information (34). For example, the [ 15 N]leucine-specific labeled apoLp-III 1 H/ 15 N HSQC spectrum gave 11 distinct cross-peaks that served as the standard for the 11 leucine residues in the three-dimensional 15 N-edited NOESY spectra. Fig. 1, which shows strip plots extracted from the 150-ms mixing time 15 N-edited NOESY spectrum of apoLp-III (residues Lys 135 -Val 149 ), illustrates the general quality of the spectra. The amide proton region (6.5-9.5 ppm) clearly shows three amide-amide cross-peak patterns along the apoLp-III sequence that involve d NN (i Ϫ 1, i), diagonal peak and d NN (i, i ϩ 1) cross-peaks for residue i, indicating these residues adopt an ␣-helical secondary structure. For example, in the strip of residue Ile 141 , two amide-amide cross-peaks can be observed other than the diagonal peak. Whereas the upfield peak has the same chemical shift as the diagonal peak of residue Lys 140 , the downfield peak has the same chemical shift as the diagonal peak of residue Lys 142 . We assigned the upfield peak as the d NN (Lys 140 , Ile 141 ) and the downfield peak as the d NN (Ile 141 , Lys 142 ), respectively. It is worth noting that severe overlap was observed in several planes of 15 N-edited NOESY spectra due to chemical shift degeneracy of backbone nitrogen atoms. For example, the nitrogen chemical shift of residue Ala 138 is degenerate with several other residues (Fig. 1). Despite these problems, we managed to assign ϳ80% of the backbone NH and N atoms on the basis of sequential NH-NH connectivities in 15 N-edited NOESY spectra. Fig. 2 shows C ␤ (i Ϫ 1)-C ␣ (i Ϫ 1)-N(i)-NH(i) connectivities from a CBCA(CO)NNH spectrum together with the corresponding C ␤ (i)-C ␣ (i)-N(i)-NH(i) and C ␤ (i Ϫ 1)-C ␣ (i Ϫ 1)-N(i)-NH(i) (in different sign) connectivities identified by an HNCACB experiment for residues Lys 73 -Thr 86 of apoLp-III. An independent assignment, obtained on the basis of these two spectra, was used to confirm the assignment derived from 15 N-edited NOESY and 15 N-edited TOCSY spectra. For residues Ala 4 -Gly 5 -Gly 6 -Asn 7 -Ala-8 , where no d NN connectivities were found, and residues Ser 33 -Lys 34 -Asn 35 -Thr 36 -Gln 37 -Asp 38 -Phe 39 , where significant overlap was observed for NH protons, assignment could not be made on the basis of the 15 N-edited NOESY spectrum alone. In this case, HNCACB and CBCA(CO)NNH spectra were used to obtain the assignment. Side-chain assignment was obtained by analysis of HCCH-TOCSY, 15 N-edited TOCSY, 15 N-edited NOESY, and 13 C-edited NOESY spectra. The 15 N-edited TOCSY spectrum was not very useful in this regard, however, since it generally showed the connectivities only up to ␣ protons. Three-dimensional HCCH-TOCSY provided a powerful spectrum for side-chain assignment, although, in some cases, chemical shift overlap was observed in some planes, creating ambiguity for some side-chain assignments. In these cases, three-dimensional 15 N-edited NOESY and 13 C-edited NOESY spectra were used to confirm the assignment. Fig. 3 shows a complete assignment of the 1 H/ 15 N HSQC spectrum of apoLp-III in 50 mM potassium phosphate, pH 7.0. The 1 H/ 15 N HSQC spectra of apoLp-III obtained in our standard buffer condition (250 mM potassium phosphate, pH 6.4 -6.5) showed considerably more overlap in the center of the spectrum, which makes it difficult to label the cross-peaks.
Other than the crowded central region, however, the overall chemical shift dispersion is similar under both conditions. Five glycine residues in apoLp-III only give rise to three cross-peaks in Fig. 3 due to fast exchange of the amide protons of two glycines (Gly 5 and Gly 6 ) at pH 7.0. At pH 6.5, however, five distinct cross-peaks were observed for those five glycine residues. Cross-peaks connected by dotted lines in the upper right corner of Fig. 3 were derived from the side-chain amine groups of Asn and Gln residues. Interestingly, several cross-peaks derived from backbone N-H groups (such as Ser 24 , Ser 128 , Ser 66 , and Ser 29 ) were located in this region of the spectrum. We have recently reported a strategy that results in specific labeling of peptide backbone nitrogen atoms without side-chain nitrogen atom labeling (34). The 1 H/ 15 N HSQC spectrum of the backbone nitrogen specifically labeled apoLp-III sample confirmed the assignment of these four Ser residues (data not shown).
The Secondary Structure of ApoLp-III- Fig. 4 summarizes short and medium range NOE connectivities (͉i Ϫ j͉ Յ 4), amide exchange rates, 3 J ␣N coupling constants, and chemical shift index data for apoLp-III. The medium range (2 Յ ͉i Ϫ j͉ Յ 4) NOEs, especially d ␣N (i, i ϩ 3) NOEs, together with strong d NN (i, i ϩ 1) and weak d ␣N (i, i ϩ 1) NOEs, indicate that the peptide bonds adopt a folded structure (i.e. ␣ helix), due to the 3.6-residue repeats in the helical structure. The observed slow amide proton exchange rates for specific residues suggest these protons are hydrogen-bonded, which prevents proton exchange with the solvent, again consistent with an ␣-helical structure. The chemical shift index is a consensus of ␣ proton and ␣ and ␤ carbon atom chemical shift deviation from random coil values, which correlates with specific secondary structure elements (41). In the case of apoLp-III, the observed negative CSI for most residues in the protein is consistent with ␣-helical secondary structure. Likewise, the observed lack of consensus positive CSI for any residue indicates an absence of ␤ structure in apoLp-III. Finally, the presence of 3 J ␣N coupling constants Ͻ6.0 Hz (shown by filled circles in Fig. 4) provides further evidence of ␣ helix structure for given residues. Whereas reliance on a single parameter may cause ambiguity, the use of four independent NMR measurements permits conclusive identification of secondary structure elements within the protein sequence. The consensus NMR parameters for apoLp-III shown in Fig. 4 clearly identify five discrete ␣-helical regions (21-30 residues in length) with well defined boundaries. These data are in accord with structural information on apoLp-III from another insect, L. migratoria (8), despite the fact that these proteins share Ͻ30% sequence identity. In M. sexta apoLp-III, the NMR data also reveal a short helix, termed helix 3Ј, located between helix 3 and helix 4 just beyond Pro 95 . This helix includes residues Asp 96 -Glu 100 . The existence of this short helix is confirmed not only by medium range NOEs d ␣N (Glu 99 , Asp 96 ), d ␣N (Glu 100 , Val 97 ), d NN (Lys 99 , Val 97 ), and d NN (Glu 100 , Glu 98 ), but also by slower amide proton exchange rates, smaller 3 J ␣N coupling constants, and chemical shift index criteria.
The five major helices in apoLp-III range between 21 and 30 residues and, other than helix 3Ј, are connected to one another by short 5-7-residue flexible loop segments. For example, helix 4 and helix 5 are connected by a seven-residue loop comprising amino acids 129 -135. Two d ␣N (i, i ϩ 2) NOEs were found between residue Glu 131 and Lys 135 , indicating turn-type structure in the loop. Helix 5 contains a proline at position 139 which, interestingly, does not break the helix. The C terminus FIG. 4. Summary of short and medium range (ͦi ؊ jͦ 4) NOE connectivities, amide exchange rates, chemical shift index, and 3 J ␣N coupling constant data for apoLp-III. Below the one-letter code for amino acids, the relative strengths of d ␣N and d NN NOE crosspeak intensities are indicated by the bar thickness. Various medium range NOE connectivities (2 Յ ͉i Ϫ j͉ Յ 4) observed between specific proton pairs are indicated by a solid line between two residues. Amide proton exchange rate, determined as described under "Experimental Procedures," is designated as follows: filled box, slow exchange rate; open box, medium exchange rate; no box, fast exchange rate. CSI represents the consensus chemical shift index for ␣ proton and ␣ and ␤ carbon atom chemical shifts. For a given residue, a negative consensus CSI (which indicates that the residue is in an ␣-helical region) is depicted by a negative bar. Filled circles represent 3 J ␣N coupling constants Ͻ6 Hz, whereas open circles represent 3 J ␣N coupling constants larger than 8 Hz (note that none were observed in this protein). For those residues where no circle is present, the 3 J ␣N coupling constants are either in the range of 6 -8 Hz or not detectable due to resonance overlap. The rectangular boxes indicate regions of predicted secondary structure defined by the NMR parameters, with helices numbered as described in the text.  NOEs, other NMR parameters indicate these residues do not adopt a stable helical secondary structure. These residues appear to be more flexible, as judged by fast or medium amide proton exchange rates and the larger 3 J ␣N coupling constants for these residues. A summary of the secondary structure of M. sexta apoLp-III and comparison with L. migratoria apoLp-III is shown in Table I.
Interhelical Contacts in the Helix Bundle-Whereas the consensus NMR parameters shown in Fig. 4 identify secondary structure elements, long range NOEs were used to characterize helix topology and the tertiary structure of apoLp-III. These NOEs identify proton pairs Ͻ5.0 Å apart in space, despite the fact that they may reside on amino acids from distant regions of the primary sequence. Fig. 5 summarizes Ͼ300 interhelical long range NOEs derived from 15 N-edited NOESY spectra at 150-ms mixing time. The helical wheel diagrams shown are based on the secondary structure of apoLp-III identified by the consensus NMR parameters described above. This depiction clearly demonstrates that each helix is amphipathic, with well segregated hydrophobic and hydrophilic surfaces. Analysis of the pattern of long range NOEs indicates that the helices in apoLp-III adopt an up and down, antiparallel orientation. For example, helix 1 contacts with both helix 4 and helix 5. We determined 46 interhelical NOEs between helix 1 and helix 5 and another 46 between helix 1 and helix 4. Long range NOEs were observed between residues 10 and 127, 20 and 153, and 29 and 159. These NOEs indicate a parallel orientation between helix 1 and helix 5 and an antiparallel orientation between helix 1 and helix 4. Interestingly, many more NOEs were found between the N terminus of helix 1 and the C terminus of helix 4 than those between the C terminus of helix 1 and the N terminus of helix 4, suggesting stronger interhelical contacts in one end of the two helices. Helix 2 interacts with three helices, helix 3 (70 interhelical NOEs), helix 4 (26 interhelical NOEs), and helix 5 (42 interhelical NOEs). Helix 2 contacts both helix 3 and helix 5 in an antiparallel manner, with contacts through the entire length of these three helices. Helix 2 contacts helix 4 weakly in a parallel orientation since interhelical NOEs were identified only between the two N termini and a few residues in the center region of the two helices. Other than contacting helix 2, helix 3 interacts with helix 4 in an antiparallel arrangement. A total of 37 interhelical NOEs were found between helix 3 and helix 4, with NOEs observed along the entire length of the two helices.
The interhelical contacts among helices in apoLp-III are mainly from hydrophobic residues such as leucine, valine, isoleucine, phenylalanine, and alanine. These hydrophobic contacts likely exert a stabilizing force in terms of the globular fold of apoLp-III. However, Fig. 5 also indicates that charged residues, such as Lys 105 and Glu 13 , and several polar residues such as Gln 20 , Asn 40 , Ser 47 , Gln 53 , Ser 58 , Gln 109 , Gln 113 , Ser 119 , Gln 120 , and Gln 156 are located within the hydrophobic core of the helix bundle. These buried charged and polar residues, which are available for contact with the other residues, may contribute to the low intrinsic stability of this protein in the lipid-free state. Long range NOEs were identified between all residues in helix 3Ј and several residues in the N terminus of helix 2, indicating the existence of interactions between helix 2 and helix 3Ј. These NOEs demonstrate that helix 3Ј is in close proximity to the N terminus of helix 2, suggesting the possible topology of helix 3Ј. Long range NOEs were also identified in the loops that connect different helices. For example, NOE contacts were found between residues in loop 1 and residues located in the N termini of helix 2 and helix 4; from loop 2 to Tertiary Structure and Global Fold- Fig. 6 shows a cylinder diagram of the global fold and tertiary structure of apoLp-III. This diagram clearly demonstrates a five-helix bundle structure for lipid-free M. sexta apoLp-III that is similar to the x-ray crystal structure of L. migratoria apoLp-III (8) as well as the four-helix bundle architecture of the N-terminal domain of human apoE (7). The short helix, helix 3Ј, which connects helix 3 and helix 4, orients almost perpendicular to helix 3 and helix 4 and is positioned at the end of the elongated helix bundle. In terms of proposed helical repositioning upon lipid association (8), helix 3Ј may function in initiating conformational opening of the protein. Interestingly, a similar short helix is present in the N-terminal domain of apoE (7). DISCUSSION Previous NMR studies on exchangeable apolipoproteins have been limited to small fragments ranging from 15 to 36 residues using two-dimensional NMR techniques. For example, Rozek et al. (42) and Buchko et al. (12) reported NMR structures of two synthetic fragments of apolipoprotein C-I, whereas Lycksell et al. (11) investigated the structural properties of a 30-residue synthetic fragment of human apolipoprotein C-II (43). In a recent study, Wang et al. (13) determined the structure of a peptide fragment (residues 166 -185) of apolipoprotein A-I and two fragments of apoE (14), which range from 14 to 18 residues, in complex with sodium dodecyl sulfate and dodecyl phosphocholine micelles. Using a similar approach, we have solved the structure of a 36-residue C-terminal fragment of M. sexta apoLp-III in the presence of SDS. 2 The present study is unique in that we report the first NMR structure of an intact exchangeable apolipoprotein. We suggest that extension of this approach to the lipid-associated state of apolipoproteins employing model lipid surfaces such as dodecyl phosphocholine micelles, points to a potentially fruitful research direction.
Following an extensive empirical search for optimal conditions for NMR experiments, we found that apoLp-III is stable for Ͼ2 weeks in 250 mM phosphate buffer at pH 6.4 -6.5. This is a critical step toward the practice of three-dimensional NMR studies of apoLp-III. It was known from previous denaturation studies that apoLp-III is relatively labile and susceptible to environmental perturbation (22). Indeed, the low stability of apoLp-III is a common property of most exchangeable apolipoproteins. For example, the midpoint of guanidine HCl-induced denaturation of the 10-kDa C-terminal domain of human apoE was found to be 0.7 M guanidine HCl (44); that for human apolipoprotein A-II was 0.6 M (45), and that for human apolipoprotein A-IV was found at 0.4 M (46). A potentially important factor contributing to the low stability of apoLp-III has been identified in the NMR structure of M. sexta apoLp-III. Although hydrophobic interactions are the major force contributing to stabilization of the helix bundle structure, Fig. 5 indicates that there are several charged residues, such as Lys 105 , Glu 13 , as well as several polar residues (e.g. Gln 20 , Asn 40 , Ser 47 , Gln 53 , Ser 58 , Gln 109 , Gln 113 , Ser 119 , Gln 120 , and Gln 156 ) located in the interior of the helix bundle. We suggest that charged and polar residues localized in otherwise nonpolar regions of the molecule contribute to the low stability of apoLp-III. This hypothesis can be tested by site-specific mutagenesis experiments coupled with denaturation studies.
It is conceivable that the low stability of apoLp-III is required for its known reversible lipoprotein binding activity. Based on the crystal structure of L. migratoria apoLp-III, it was proposed that the helix bundle opens about putative hinge regions, exposing its hydrophobic interior, which interacts di- rectly with lipid surfaces (8). This open conformation model is supported by the NMR structure of apoLp-III presented in this study. Whereas conformational opening of apoLp-III may disrupt certain interhelical contacts within the protein, this is more than compensated by subsequent replacement of helixhelix contacts with helix-lipid interactions. It is recognized that association with lipoprotein surfaces and consequent conformational opening of apoLp-III is induced by the content of surface monolayer-localized diacylglycerol. On the other hand, dissociation from lipoprotein surfaces and return to the helix bundle conformation results from depletion of lipoprotein diacylglycerol content (47)(48)(49)(50). Thus, in essence, apoLp-III association/ dissociation with lipoproteins is dictated by the lipid composition of the particle. The intrinsic reversibility of apoLp-III conformational changes permits the protein to function in multiple association/dissociation events during its lifetime in plasma.
Compared with the crystal structure of L. migratoria apoLp-III, the NMR structure of M. sexta apoLp-III reveals an overall similar molecular architecture. One apparent difference, however, is the presence of a short helix, helix 3Ј, in M. sexta apoLp-III. Interestingly, a similar short helix was also found in the x-ray crystal structure of human apoE N-terminal domain (7). In human apoE, this short helix, which involves nine residues, connects helix 1 and helix 2 and orients perpendicular to helix 1 and helix 2. The short helix found in the NMR structure of M. sexta apoLp-III is in a similar orientation. It is worth noting that helix 3Ј is a flexible helix displaying amide exchange rates that are faster than those from the other helices.
The helix-short helix-helix organization in this molecule may comprise an important structural element in apoLp-III in terms of initiating interactions with lipoprotein surfaces. It is noteworthy that helix 3Ј is positioned where, in the open conformation model, it could initiate contact with lipid surfaces. We have proposed a structural model to describe the interactions between apoLp-III and the phospholipid monolayer of insect lipophorins based on 31 P NMR studies designed to evaluate the mobility of phospholipids in different lipophorin subspecies (48). In this model, we suggested that diacylglycerol partitioning into the particle surface monolayer creates a binding site for apoLp-III interaction. It is recognized that a strong correlation exists between lipophorin diacylglycerol content and apoLp-III binding, and recently we have provided direct experimental evidence for surface localization of diacylglycerol in lipophorin particles known to bind apoLp-III (49). We speculate that surface diacylglycerol intercalates between phospholipids in the monolayer surface, creating a gap between phospholipid head groups. This defect is effectively repaired by apoLp-III binding.
An important question, however, is whether hydrophobic interactions are responsible for initiation of binding or whether ionic interactions localize apoLp-III at the particle surface, positioning the protein to "respond" to surface defects created by diacylglycerol partitioning into the surface monolayer. Soulages and Wells (51) have presented a "hydrophobic sensor" hypothesis to describe the initiation of interactions between exchangeable apolipoproteins and lipoproteins. According to this hypothesis, a two-step sequential mechanism for binding of apoLp-III to lipoprotein surfaces has been proposed. The first step involves a recognition process (through exposed hydrophobic amino acids) consisting of the adsorption of apoLp-III to a nascent hydrophobic defect in the phospholipid bilayer caused by the presence of diacylglycerol. This is followed by a conformational opening to expose the protein interior. Soulages and Wells suggest (51) that in L. migratoria apoLp-III, conserved leucines located in the loops between helices are responsible for the initial association process. An alternative proposal is that helix 3Ј may function in initiation of stable lipoprotein binding. This concept is supported by disulfide bond engineering studies, which indicate that apoLp-III interactions with lipoproteins is oriented at the end of the molecule containing helix 3Ј (52). Helix 3Ј contains five residues (Asp 96 , Val 97 , Glu 98 , Lys 99 , and Glu 100 ), four of which possess charged hydrophilic side chains. The suggestion that charge-charge interactions between the phospholipid head groups and helix 3Ј provide a relatively long range attraction, which localizes apoLp-III in close proximity to the lipoprotein surface, is supported by studies with model phospholipids (53). Once the recognition process is complete, surface-exposed diacylglycerol could trigger opening of the helix bundle, resulting in formation of a stable binding interaction. Judging by the fact that the N-terminal domain of apoE contains a similar short helix, an analogous binding mechanism may be suggested for human exchangeable apolipoprotein-lipoprotein interactions. Although further experiments are required to determine the precise mode of lipid association of apoLp-III, it is likely that concepts developed with this model protein may have important implications for amphipathic-exchangeable apolipoproteins in general.