Structural Determination of Lipid-bound ApoA-I Using Fluorescence Resonance Energy Transfer*

Based on the x-ray crystal structure of lipid-free Δ43 apoA-I, two monomers of apoA-I were suggested to bind to a phospholipid bilayer in an antiparallel paired dimer, or “belt orientation.” This hypothesis challenges the currently held model in which each of the two apoA-I monomers fold as antiparallel α-helices or “picket fence orientation.” When apoA-I is bound to a phospholipid disc, the first model predicts that the glutamine at position 132 on one apoA-I molecule lies within 16 Å of glutamine 132 in the second monomer, whereas, the second model predicts glutamines at position 132 to be 104 Å apart. To distinguish between these models, glutamine at position 132 was mutated to cysteine in wild-type apoA-I to produce Q132C apoA-I, which were labeled with thiol-reactive fluorescent probes. Q132C apoA-I was labeled with either fluorescein (donor probe) or tetramethylrhodamine (acceptor probe) and then used to make recombinant phospholipid discs (recombinant high density lipoprotein (rHDL)). The rHDL containing donor- and acceptor-labeled Q132C apoA-I were of similar size, composition, and lecithin:cholesterol acyltransferase reactivity when compared to rHDL-containing human plasma apoA-I. Analysis of donor probe fluorescence showed highly efficient quenching in rHDL containing one donor- and one acceptor-labeled Q132C apoA-I. rHDL containing only acceptor probe-labeled Q132C apoA-I showed rhodamine self-quenching. Both of these observations demonstrate that position 132 in two lipid-bound apoA-I monomers were in close proximity, supporting the “belt conformation” hypothesis for apoA-I on rHDL.

Among plasma apolipoproteins, apoA-I occupies a unique role in regulating lipoprotein cholesterol metabolism as demonstrated by recent studies in transgenic and knockout mice (1)(2)(3)(4)(5)(6)(7)(8). ApoA-I is an amphipathic protein with high ␣-helical content that avidly binds phospholipid molecules, organizing them into soluble bilayer structures or discs that readily accept cholesterol. Whether the initial steps of disc formation require interaction with an ABCA1 transporter (9 -11) is unclear. However, once apoA-I is assembled into discs, it becomes the most potent physiological activator of plasma lecithin:cholesterol acyltransferase (LCAT). 1 This enzyme catalyzes the conversion of cholesterol to cholesterol ester and in so doing transforms discoidal HDL to spherical HDL. To better understand this process, many structural studies have focused on describing the three-dimensional conformation of lipid-bound apoA-I. Central to these investigations is the inevitable question of what makes apoA-I so potent in its ability to activate LCAT when compared with other plasma apoproteins (AIV, AII, CI, CII, CIII, and E), because they are structurally very similar (for excellent reviews see Refs. [12][13][14][15]. Although crystallization and x-ray diffraction have proven difficult, renewed interest in solving the 3-dimenstional structure of phospholipid-bound apoA-I came about when the crystal coordinates for lipid-free ⌬43 apoA-I (16) were reported in 1997. Based on both biophysical studies (17) and x-ray crystal coordinates it was hypothesized that apoA-I orients all of its helical 11-and 22-mers perpendicularly to the phospholipid bilayer in the lipid-bound state, referred to as the "belt" conformation. Although plausible, the belt model contrasts sharply with the "picket fence" model, which until now has been generally regarded as correct. In the picket fence model each of the ten amphipathic helices are arranged parallel to the phospholipid acyl chains.
Historically, the picket fence model describing the lipidbound conformation of apoA-I evolved from studies published in 1977 suggesting that apoA-I was located at the edge of the phospholipid bilayer rather than intercalated among the phospholipid head groups of the disc (18,19). With time the picket fence model gained support from experimental evidence, including Fourier transform infrared spectroscopy of dried films (21,22), and a variety of other experimental and theoretical approaches (19 -26). However, since the publication of the lipid-free crystal structure, the belt model has been closely examined and experimental support for it includes Fourier transform infrared measurements of DMPC/dimyristoylphosphatidylserine discs in 1 mM Ca 2ϩ (27), detailed molecular modeling of rHDL (28), and nitroxide spin label quenching of tryptophan fluorescence within helix 4 (29).
Each of the published experimental approaches designed to reveal the lipid-bound conformation of apoA-I has one or more limitations. Therefore, we sought to utilize a technique that would allow an unambiguous distinction between the two prevailing models of lipid-bound apoA-I conformation. To do this we used fluorescence resonance energy transfer of acceptorand donor-labeled apoA-I rHDL. These studies show that residues delineated by repeat 5 (#121-142) within two apoA-I molecules are in close proximity, some 6 -30 Å from each other when bound to a phospholipid bilayer. Thus, these data are consistent with and support the belt conformation for lipidbound apoA-I as predicted from x-ray crystallography.
EXPERIMENTAL PROCEDURES Dithiothreitol (DTT), 5-iodoacetamido-fluorescein (5-IAF or donor probe), tetramethylrhodamine-5-iodoacetamide (5-TMRIA or acceptor probe) were purchased from Molecular Probes (Eugene, OR). High performance liquid chromatography-grade organic solvents were purchased from Fisher Scientific. [ 3 H]Cholesterol was purchased from PerkinElmer Life Sciences. Tissue culture reagents, restriction endonucleases, and other DNA modifying enzymes were purchased from Life Technologies, Inc. Q-Sepharose Fast Flow and NAP-25 columns were obtained from Amersham Pharmacia Biotech. Cholesterol Ͼ99% was from Nu-Chek Prep. Sodium cholate was from Calbiochem, and all other chemical reagents were purchased from Sigma Chemical Co. unless otherwise noted.
Q132C ApoA-I: Mutagenesis, Expression and Purification-The mutant Q132C apoA-I was constructed using the polymerase chain reaction primer (5Ј-CTCGTGCAGCTTACAGCGCGCGCCCTCTTG-3Ј) containing the mutant codon (underlined), which changed a glutamine at position 132 in wild-type apoA-I to a cysteine. The 5Ј-and 3Ј-end-most apoA-I cDNA primers, used to obtain the complete Q132C apoA-I cDNA polymerase chain reaction product, were the same as those described previously (30 -32). The mutant Q132C apoA-I was cloned into a baculovirus expression plasmid as described previously (31) and expressed in Sf-9 cells. The expressed mutant apoA-I was purified to homogeneity, and its molecular weight was determined by electrospray mass spectrometry on a Quattro II mass spectrometer as previously reported (30 -32). To obtain the correct experimental molecular weight, DTT treatment (final concentration of 50 mM DTT for 10 min on ice) of the purified Q132C apoA-I was employed to release the cysteinyl adduct formed during expression of the mutant protein in the Sf-9 baculovirus system.
Preparation and Purification of Donor-and Acceptor-labeled Q132C ApoA-I-The purified mutant Q132C apoA-I was treated with either 5-IAF (donor probe) or 5-TMRIA (acceptor probe) by mixing 2 mg/ml Q132C apoA-I with a 5-fold molar excess of 5-IAF or 5-TMRIA over total thiols, at a final concentration of 50 mM DTT and 10 mM sodium phosphate, pH 7.4 (33). This mixture was incubated on ice overnight in the dark and the next morning applied to a G-25 Sephadex column pre-equilibrated with 10 mM sodium phosphate, pH 7.4. The labeled Q132C apoA-I-containing fractions were pooled and dialyzed against 10 mM ammonium bicarbonate, pH 7.4, 15 M sodium azide, and 3 M EDTA. The extent and purity of both the 5-IAF-and 5-TMRIA-labeled Q132C apoA-I was determined by electrospray mass spectrometry.
Preparation of rHDL Containing 5-IAF-and 5-TMRIA-labeled Q132C ApoA-I-Labeled Q132C apoA-I rHDL were prepared using the sodium cholate dialysis method (31,32). L-␣-Dimyristoyl-phosphatidylcholine (DMPC) was added at a starting molar concentration of 100 mol of DMPC to 5 mol of cholesterol to 1 mol of Q132C donor or acceptor probe-labeled apoA-I or at selected ratios of donor-to acceptor-labeled Q132C apoA-I. After extensive dialysis to remove sodium cholate, the rHDL were purified by fast protein liquid chromatography using three Superdex 200 HR 10/30 columns linked in tandem and run at a flow rate of 0.5 ml/min. Individual fractions or pooled fractions corresponding to the eluted rHDL peak were then assayed for phospholipid, cholesterol, and protein content. The final molar ratio of all DMPC rHDL made using this procedure was approximately 90 mol of DMPC to 4.1 mol of cholesterol to 1 mol of apoprotein, as described previously (31,32). The diameter of DMPC rHDL containing 5-IAF-and 5-TMRIAlabeled apoA-I was determined by 4 -30% non-denaturing gradient gel electrophoresis and was compared with proteins of known Stokes diameter (31,32). Cross-linking studies (34) were performed to confirm that 104-Å rHDL contained only two molecules of apoA-I per particle (30).
To measure fluorescence resonance energy transfer between apoA-I monomers bound to a single rHDL bilayer, discs were prepared at a starting acceptor probe-labeled Q132C apoA-I to donor probe-labeled Q132C apoA-I molar ratio of 3:1. Control rHDL contained either accep-tor probe-labeled Q132C apoA-I and unlabeled Q132C apoA-I at a molar ratio of 3:1 or unlabeled Q132C apoA-I and donor probe-labeled Q132C apoA-I at a molar ratio of 3:1. Additional controls included rHDL prepared using either pure donor-labeled Q132C apoA-I or pure acceptor probe-labeled Q132C apoA-I. rHDL were prepared from acceptor probe-labeled Q132C apoA-I to unlabeled Q132C apoA-I at a 1:3 molar ratio.
Determination of LCAT Reactivity-LCAT reactivity was measured on purified DMPC rHDL containing either probe-labeled Q132C apoA-I or human plasma apoA-I. The rHDL were prepared and purified as described in the preceding section. Measurements of the LCAT reactivity of each preparation were carried out as described (31,32)  Fluorescence and Absorbance Measurements-Steady-state fluorescence measurements were made at 25°C using either an ISS K2 (ISS, Inc., Champaign, IL) or SLM AB2 (SLM Instruments, Urbana, IL) fluorometers equipped with a xenon arc lamps in a 1-cm ϫ 5-mm quartz fluorometer cell. The protein concentrations were 1-2 g of rHDL complexes in 10 mM Tris-HCl, 140 mM NaCl, 150 M sodium azide, 25 M EDTA, and 50 mM Hepes, pH 7.4. The fluorescence resonance energy transfer experiments were performed on three different preparations of rHDL prepared at a starting molar ratio of 3:1 acceptor-to donorlabeled Q132C apoA-I. Excitation wavelengths of 490 and 550 nm and emission wavelengths of 530 and 580 nm were used (4-nm slit width) for donor and acceptor probes, respectively. Fluorescence spectra were also obtained after disruption of rHDL by the addition of NaOH and SDS to give a final concentration of 0.1 N NaOH and 0.1% SDS. Absorption spectra were measured using a Hewlett-Packard diode array spectrophotometer with a resolution of Ϯ1 nm.
Fluorescence lifetimes were measured by the phase and modulation technique using the ISS K2 multi-frequency fluorometer. Donor probelabeled Q132C apoA-I was excited at a wavelength of 488 nm using a Coherent INNOVA 307 argon ion laser. Excitation light was modulated over a frequency range of 2-250 MHz using a Pockel's cell. The fluorescence emission of donor-labeled protein was measured at a wavelength of 530 nm (16-nm slit width). Under these conditions, no fluorescence of the acceptor probe was detected. 6-Carboxyfluorescein (1 M) was used as the reference sample. A non-linear least-squares fitting program was used to analyze the phase and modulation data in terms of single-, double-, and triple-exponential decays. Phase and modulation data were fit by minimizing the chi square parameter ( 2 ). Improvement of 2 by a factor of 2 was considered justification for increasing the number of exponentials.

Hypothetical Models of Lipid-bound ApoA-I Conformation-
Two hypothetical models for the lipid-bound conformation of apoA-I are shown in Fig. 1. Panel A shows the recently proposed belt conformation, and panel B shows the picket fence conformation. In the belt conformation, residues at position 132 are approximately 16 Å apart. While in the picket fence conformation these residues are approximately 104 Å apart. Thus, probe attachment at the repeat 5 position 132 provided for the unambiguous distinction between belt and picket fence models.
Acceptor and Donor Probe-labeled Q132C ApoA-I-Purified Q132C apoA-I was labeled with either acceptor or donor probes, and then non-covalently bound probe was removed by chromatography on Sephadex G-25. The extent of labeling was determined by electrospray mass spectroscopy. Fig. 2A shows the spectrum from DTT-treated Q132C apoA-I. The predominant peak had a molecular weight of 28,939 within the experimental error of that calculated for proapoA-I of Q132C apoA-I (31). Figs. 2B and 2C show the spectrum of acceptor and donor probe-labeled Q132C apoA-I, respectively. In each case, the experimental molecular weight was close to the predicted molecular weight. These data suggested that our labeling conditions produced Ͼ98% of covalently labeled Q132C apoA-I.
To assess the presence of non-covalently bound probe, purified labeled apoA-I was dissolved in 6 M urea and then applied to a Sephadex G-25 column equilibrated with 6 M urea/10 mM Tris, pH 7.4. The fluorescence intensity of individual column fractions was measured (data not shown). These results also showed that both purified acceptor-and donor-labeled apoA-I contained Ͻ0.1% non-covalently bound probe.
Size, Composition, and LCAT Reactivity of Probe-labeled rHDL-To determine whether the mutation or labeling procedure employed altered the size, composition, or LCAT reactivity of Q132C apoA-I rHDL, DMPC complexes were prepared with acceptor and donor probe-labeled apoA-I and compared with rHDL prepared with purified human plasma apoA-I (30). These rHDL were analyzed by 4 -30% non-denaturing polyacrylamide gel electrophoresis with the equivalent of 6 g of total rHDL protein per lane, as shown in Fig. 3. By comparison to standards of known Stokes diameter, we found that each preparation of DMPC rHDL averaged approximately 104 Å in diameter. The final phospholipid:cholesterol:protein ratios of the labeled rHDL were approximately 98:4:1 (data not shown). Cross-linking studies confirmed that each preparation contained two monomers of apoA-I per 104-Å particle. The LCAT reactivities of the various DMPC rHDL containing probe-labeled apoA-I were essentially identical (V max and K m ) to rHDL containing human plasma apoA-I (data not shown).
Fluorescence Properties of Donor Probe-labeled rHDL-Overall, the fluorescence properties of donor and acceptor probelabeled apoA-I were similar to other published studies using fluorescein and rhodamine probes (35). There was substantial overlap of the donor emission spectrum with the acceptor ab-sorbance spectrum. Fluorescence polarization values of both probes were Ͻ0.1, consistent with rapid rotation of the probes with an R 0 (distance for 50% transfer efficiency) of approximately 45 nm (data not shown).
Changes in the excitation and emission spectra at alkaline pH showed that a substantial fraction of fluorescein probes underwent a transition from the less fluorescent monoanion to the more fluorescent dianion (data not shown). Comparison of these spectra with previously published spectra (36) indicated that 50% of the donor probes in DMPC rHDL were protonated at pH 7.4. This is typical of fluorescein probes that interact with the head group region of phospholipid bilayers, such as those attached to head groups of membrane phospholipids (37) or those attached close to the membrane anchor sequence of transmembrane proteins (36). The less polar environment of the phospholipid head groups raises the pK a of fluorescein, favoring the less charged monoanionic form.
Donor probes did not self-quench when both monomers of DMPC rHDL were labeled with donor probes. This was shown by comparing the fluorescence intensities of DMPC rHDL prepared with pure donor probe-labeled apoA-I to rHDL prepared from donor probe-labeled apoA-I and unlabeled Q132C apoA-I at a 1:3 mole ratio, respectively. In the latter case, most of the donor probe-labeled apoA-I monomers should be paired with unlabeled monomers. The fluorescence intensities of these two preparations were proportional to their content of donor probelabeled apoA-I, i.e. their fluorescence quantum yields were the same, and their excitation and emission spectra were superimposable (data not shown).
Fluorescence Properties of Acceptor Probe-labeled rHDL-In marked contrast to donor probe-labeled apoA-I, there was a pronounced self-quenching of probes in DMPC rHDL prepared from pure acceptor probe-labeled apoA-I. Self-quenching was caused by the formation of non-covalent rhodamine dimers that were much less fluorescent than rhodamine monomers (38). Because dimer formation requires contact between rhodamine probes, these data showed that position 132 of the individual apoA-I monomers of DMPC rHDL were in close proximity. Quenching was most easily observed by comparing the fluorescence excitation spectrum (dashed line) to the absorbance spectrum (solid line) of DMPC rHDL containing pure acceptor probe-labeled apoA-I (Fig. 4A). Both spectra had maxima near 550 nm, which is typical of tetramethylrhodamine. However, the presence of a shoulder at 510 nm in the absorption spectrum indicates the presence of the relatively non-fluorescent, non-covalent rhodamine dimer (38). After disrupting rHDL with SDS/NaOH, the excitation and absorption spectra were nearly superimposable (Fig. 4B), confirming the presence of non-covalent rhodamine dimers. The spectral changes of the acceptor probe after disruption cannot be attributed to differences in protonation, because tetramethylrhodamine lacks titratable groups in this pH range.
In another series of experiments we quantified the extent of rhodamine dimer formation by dividing the rHDL fluorescence intensity by the intensity after SDS/NaOH disruption for a series of rHDL prepared at different ratios of acceptor probelabeled apoA-I to unlabeled Q132C apoA-I. It was anticipated that decreasing the ratio of acceptor apoA-I to unlabeled Q132C apoA-I would decrease the formation of rhodamine dimers. As shown in Fig. 5 the relative fluorescence yield from rHDL increased as the amount of unlabeled Q132C apoA-I increased. However, the decrease in rhodamine dimers by addition of unlabeled Q132C apoA-I (stippled bars) was less than expected if the monomers were randomly distributed among the rHDL (solid bars). Prolonged incubation of labeled and unlabeled Q132C apoA-I in the presence of detergent prior to formation of  (40). Both models show the relative position of an acceptor and donor probe labeled at position 132 (repeat 5) in the mutant Q132C apoA-I. A, a 104-Å diameter discoidal rHDL particle composed of approximately 185 molecules of DMPC surrounded by two apoA-I monomers, which have been arranged as a pair of continuous amphipathic ␣-helices forming a belt conformation. B, a discoidal rHDL particle of the same size and composition as that shown in the top panel, but where each of the two apoA-I monomers have been arranged to form antiparallel ␣-helices along the edge of the particle forming a picket fence conformation. rHDL did not change these results. This indicates that rhodamine dimer formation, although not of particularly high affinity (39), drives the preferential pairing of acceptor apoA-I monomers during the formation of rHDL. The conclusions that can be drawn from analysis of rhodamine self-quenching in DMPC rHDL are first, that the residues at position 132 in the two monomers of rHDL are sufficiently close together to allow formation of non-covalent rhodamine dimers. Because dimer formation requires contact between rhodamine residues, this result provides strong support for the belt model. The second conclusion is that formation of non-covalent rhodamine dimers lead to the preferential association of acceptor apoA-I monomers with each other, so that the distribution of probe-labeled and unlabeled apoA-I monomers among labeled rHDL was not random.

Fluorescence Resonance Energy Transfer in Acceptor and
Donor Probe-labeled rHDL-The proximity of residues at position 132 in two apoA-I monomers bound to rHDL was readily apparent by the appearance of acceptor probe fluorescence following excitation of the donor probe. DMPC rHDL were prepared from donor and acceptor probe-labeled apoA-I at a 1:3 mole ratio. Control rHDL were prepared from unlabeled Q132C apoA-I mixed with donor or acceptor probe-labeled apoA-I at 3:1 and 1:3 mole ratios, respectively. Emission spectra obtained by excitation at the donor maximum (490 nm) are shown in Fig.  6A. Donor emission (red line) had a maximum at 520 nm, whereas the acceptor emission maximum (blue line) was at 575 nm. DMPC rHDL containing donor and acceptor probe-labeled apoA-I had a less intense donor fluorescence but higher acceptor fluorescence intensity than the corresponding controls. This can be seen by comparing the emission spectrum of donor-and acceptor-containing rHDL (green line) with the sum of the FIG. 2. Electrospray mass spectra. A, 20 ng of purified Q132C apoA-I treated with 50 mM DTT showing a molecular ion at molecular weight 28,939, which corresponds to the predicted molecular weight of the proapoA-I form of Q132C apoA-I (31). It should be noted that to obtain the correct experimental molecular weight, DTT treatment of the purified Q132C apoA-I was employed to release the cysteinyl adduct formed during expression of the mutant protein in the Sf-9 baculovirus system. B, 20 ng of Q132C apoA-I labeled with acceptor probe (5-TMRIA) with the predicted molecular ion at molecular weight 29,380. C, 2 g of Q132C apoA-I labeled with donor (5-IAF) with a molecular ion at molecular weight 29,327. emission spectra of the two control rHDLs (black line). From the extent of donor fluorescence quenching, the efficiency of energy transfer was determined to be 40.5 Ϯ 4% (n ϭ 3). The emission spectrum of donor and acceptor probe-labeled apoA-I rHDL treated with NaOH/SDS was similar to the sum of the spectra of the two controls (Fig. 6B) showing that there was no energy transfer when rHDL was disrupted with NaOH/SDS. These data indicate that residues at position 132 in two monomers of apoA-I bound to rHDL were close enough to give substantial energy transfer between donor and acceptor probes. This supports the belt model of rHDL in which residues at position 132 are close together (Fig. 1). However, these data by themselves do not rule out models intermediate between the belt and picket fence, because the observed energy transfer efficiency of 40.5% could result from an intermediate spacing between the probes. Such intermediate models were ruled out by analysis of the donor fluorescence lifetime distribution. These data showed that the residual fluorescence of the donor probe in Fig. 6A was due almost entirely to rHDL in which both monomers were labeled with the donor probe. Little donor fluorescence arose from rHDL containing a mixture of donor and acceptor probe-labeled apoA-I.
Donor fluorescence lifetimes were determined by the multifrequency phase and modulation technique. Excitation intensity was sinusoidally modulated at various frequencies, and the extent of emission phase shift and demodulation was measured. The delay between excitation and emission introduced by the excited state lifetime caused both a phase shift and demodulation. Data were fitted to multiple lifetime component models by a non-linear least squares method. The data obtained for control DMPC rHDL containing a 3:1 molar ratio of unlabeled and donor probe-labeled apoA-I were well-described by two lifetime components (Table I). Most of the fluorescence was attributable to a component with a lifetime of 4.3 ns. There was a small amount of fluorescence from a short lifetime component (0.44 ns), probably due to conversion of the less fluorescent monoanionic form by the protonation of some fluorescein residues. The data were fit well by assuming two fluorescence components with lifetimes and fractional intensities similar to those obtained in the absence of acceptor probe-labeled apoA-I as shown in Table I. Thus, nearly all of the donor fluorescence was attributable to rHDL with no energy transfer. The fit was not improved by assuming three lifetime components. No species with an intermediate lifetime corresponding to rHDL with an intermediate efficiency of energy transfer could be detected.
If rHDL containing both donor-and acceptor-labeled apoA-I monomers had an intermediate efficiency of energy transfer, this would have been easily detected. This can be demonstrated by modeling phase and modulation data for the hypothetical intermediate lifetime component. For example, if the distribution of labeled monomers in rHDL was completely random, then 75% of the donor-labeled apoA-I monomers would be paired with acceptor-labeled apoA-I monomers, whereas 25% of the donor-labeled apoA-I monomers would be paired with another donor apoA-I monomer (a starting molar ratio of acceptor to donor probe-labeled monomer of 3:1). Because rHDL containing two donor apoA-I monomers do not exchange energy, the efficiency of energy transfer in rHDL containing both donor and acceptor probe apoA-I would have to be 54% for the average energy transfer to be 40.5%. An efficiency of 54% would correspond to a reduction in donor lifetime from 4.3 to 2.0 ns. Thus, the experimental data clearly rule out this model.
In summary, two conclusions can be drawn from the analysis of the lifetime distribution of donor probe-labeled apoA-I rHDL. First, donor apoA-I monomers were distributed between two populations of rHDL. One population had little if any energy transfer. This population corresponds to rHDL with two donor apoA-I monomers. The second population had virtually complete energy transfer and corresponds to rHDL with both donor and acceptor apoA-I monomers. If the monomers were randomly distributed, then 75% of donor apoA-I monomers should have been paired with acceptor apoA-I monomers. However, only 40% of donor apoA-I monomers were paired with acceptor apoA-I monomers. Thus the second conclusion to be drawn from the analysis of donor fluorescence lifetimes is that the distribution of monomers was not random. Both of these conclusions are consistent with the analysis of non-covalent rhodamine dimer formation in rHDL containing acceptor apoA-I. DISCUSSION Elucidation of the lipid-bound conformation of apoA-I has proven to be difficult leading to a recent debate. To distinguish between the two popular models describing the conformation of apoA-I (illustrated in Fig. 1) we have used fluorescence resonance energy transfer. In these studies, glutamine 132 within repeat 5 was mutated to cysteine so we could attach thiolreactive fluorescent donor or acceptor probes at a single defined helical position within the protein. This position was picked for several reasons, first because of the extreme separation distance in donor and acceptor probes predicted from each of the two models (Fig. 1). Residues at position 132 would be approximately 16 Å from each other in the belt conformation, whereas residues at position 132 would be approximately 104 Å apart in the picket fence conformation. Thus, probe placement at this site allowed each model to serve as its own control. Finally, we chose to attach probes at position 132 because of its location on the amphipathic helix polar face of repeat 5. At this location we would expect the probes to be directed toward the aqueous environment and possess optimal flexibility and rotation, even when bound to phospholipid.
Acceptor and donor probe-labeled Q132C of apoA-I DMPC rHDL were found to be similar in size, lipid composition, and FIG. 5. Self-quenching of rHDL containing acceptor probe-labeled Q132C apoA-I. Fluorescence of rHDL containing acceptor probe-labeled Q132C apoA-I to unlabeled Q132C apoA-I at the indicated molar ratio was measured at an emission wavelength of 580 nm. Relative fluorescence yields shown by the dotted bars were experimentally determined by dividing the fluorescence intensity obtained at pH 7.4 by the intensity obtained at a final concentration of 0.1 N NaOH/ 0.1% SDS and represent the mean Ϯ S.D., from three different rHDL preparations. Data shown by the solid bars represent the calculated theoretical yields assuming that labeled and unlabeled Q132C apoA-I monomers are randomly distributed among the rHDL.
FIG. 6. Emission spectra of donor and acceptor probe-labeled Q132C apoA-I rHDL. Fluorescence was measured at an excitation wavelength of 490 nm for DMPC rHDL containing acceptor probelabeled Q132C apoA-I to donor probe-labeled apoA-I at a molar ratio of 3:1 (green line), DMPC rHDL containing unlabeled Q132C apoA-I to donor probe-labeled apoA-I at a molar ratio of 3:1 (red line), DMPC rHDL containing acceptor probe-labeled Q132C apoA-I to unlabeled apoA-I at a molar ratio of 3:1 (blue line). The black spectrum is the sum of the control spectra for donor (red line) and acceptor (blue line) probe-labeled Q132C apoA-I. A, spectra obtained at pH 7.4; B, spectra obtained at final concentrations of 0.1 N NaOH/and 0.1% SDS. LCAT reactivity to DMPC rHDL containing human plasma apoA-I. These data strongly suggest that neither the Q132C mutation nor the presence of donor and or acceptor probes altered apoA-I conformation when bound to phospholipid. The choice of the fluorescein/rhodamine probe pair was made because their distance-dependent transfer efficiency falls intermediate between the two different models. Thus for this acceptor/donor probe pair, one would predict approximately 50% transfer efficiency (R 0 ) between 45 and 50Å (35)(36)(37)(38)(39). Therefore, when the donor and acceptor probes are configured in the belt orientation, nearly 100% transfer efficiency would be observed, whereas, if the donor and acceptor probes are configured in the picket fence, nearly 0% transfer efficiency would be observed.
In our studies the energy transfer efficiency between acceptor and donor probes is nearly 100% in DMPC rHDL containing both acceptor and donor probe-labeled apoA-I. Thus, these data unambiguously confirm the proximity of residues at position 132 in 2 lipid-bound apoA-I monomers and further suggest that the maximum distance between acceptor and donor probes is 30 -35 Å (35-39). These studies also provide a second line of evidence confirming the belt conformation, namely the demonstration of non-covalent rhodamine dimer formation in rHDL. From analysis of the absorbance spectra (Fig. 4A) it has been estimated that the distance between probes as rhodamine dimers is approximately 6 Å. Therefore, in addition to supporting the belt conformation for apoA-I repeat 5, our data further support the antiparallel arrangement of apoA-I monomers, described by the x-ray crystal structure of ⌬43 apoA-I (29) and molecular modeling studies of lipid-bound apoA-I (28,41). From these studies it has been suggested that certain residues within repeat 5 of one apoA-I monomer form intermolecular salt bridges with residues within repeat 5 in the second apoA-I monomer stabilizing the entire disc structure.
Although these data are consistent with the majority of apoA-I in the belt conformation, we have calculated from the data presented in Fig. 5 that less than 7% of the rHDL contain apoA-I in the picket fence conformation. Although our studies are consistent with the belt conformation for lipid-bound apoA-I, additional studies will need to be carried out to distinguish between two alternative forms of the belt conformation. Fig. 7 shows that both the hairpin (A) and extended (B) belt conformations are equally possible given the results presented in this report. Additional studies are currently underway using mutant apoA-I proteins in which cysteine substitutions have been introduced into different apoA-I repeats. In this way we will be able to distinguish between alternate forms of the belt as well as determine whether the entire apoA-I monomer conformation is uniformly of belt arrangement or if selected regions of the protein form a picket fence conformation. In summary, this report shows for the first time that the use of fluorescence resonance energy transfer provides a reliable and sensitive method for measuring the molecular distance between lipid-bound apoA-I monomers.
FIG. 7. Illustration of two potential configurations of the belt conformation. A, hairpin conformation; B; extended conformation. In each case, repeat 5 of one monomer is in close proximity to repeat 5 in the second monomer bound to a phospholipid bilayer.