Sequence conservation of apolipoprotein A-I affords novel insights into HDL structure-function.

We performed alignment of apolipoprotein A-I (apoA-I) sequences from 31 species of animals. We found there is specific conservation of salt bridge-forming residues in the first 30 residues of apoA-I and general conservation of a variety of residue types in the central domain, helix 2/3 to helix 7/8. In the lipid-associating domain, helix 7 and helix 10 are the most and least conserved helixes, respectively. Furthermore, eight residues are completely conserved: P66, R83, P121, E191, and P220, and three of seven Tyr residues in human apoA-I, Y18, Y115, and Y192, are conserved. Residue Y18 appears to be important for assembly of HDL. E191-Y192 represents the only completely conserved pair of adjacent residues in apoA-I; Y192 is a preferred target for site-specific oxidative modification within atheroma, and molecular dynamic simulations suggest that the conserved pair E191-Y192 is in a solvent-exposed loop-helix-loop. Molecular dynamics testing of human apoA-I showed that M112 and M148 interact with Y115, a microenvironment unique to human apoA-I. Finally, conservation of Arg residues in the α11/3 helical wheel position 7 supports several possibilities: interactions with adjacent phospholipid molecules and/or oxidized lipids and/or binding of antioxidant enzymes through cation-π orbital interactions. We conclude that sequence alignment of apoA-I provides unique insights into apoA-I structure-function relationship.


Molecular dynamics simulations
One PL-rich particle containing 160 POPC, 24 unesterifi ed cholesterol (UC), and 2 apoA-I molecules was created helix ( 10,11 ). Exon 3 encodes residues 1-43 in human apoA-I. This domain, commonly referred to as the globular domain ( 12 ), contains an N-terminal segment of 8-10 residues (10 in human apoA-I), followed by three 11 mer amino acid tandem repeats, designated G1, G2, and G3, respectively (see Fig. 3). Barker and Dayhoff ( 13 ), Fitch ( 14 ), and McLachlan ( 15 ) independently observed that apoA-I contains multiple repeats of 22 amino acids (22 mer), each of which consists of two 11 mer tandem arrays. DNA sequencing work has confi rmed the existence of a 22 mer periodicity in apoA-I ( 16 ). The portion of apoA-I that follows the globular domain, residues 44-241 in human apoA-I, referred to as the lipid-associating domain ( 12 ), contains eight 22 mer and two 11 mer tandem repeats, each of which has the helical wheel signature of an amphipathic ␣ helix. The fundamental unit of repeat in exon 4 is not an 11 mer but a 22 mer repeat made up of two 11 mer repeats designated a and b; most 11 mer repeats are more similar to the 11-mer repeat one unit removed (e.g., helix 1a vs. helix 2a) than to the adjacent 11 mer unit (e.g., helix 1a vs. helix 1b) ( 14,16,17 ). This is indicated by the fact that almost all Pro residues appear in the fi rst column when the alignment is made using 22 amino acids as the basic unit of repeat, i.e., at position 1 in repeat a. It has therefore been suggested that the different apolipoprotein genes arose from a common ancestral gene by gene duplications of 11 codons ( 18 ).

RATIONALE FOR THIS REVIEW
It has been approximately 10 years since the last fundamental review articles pertaining to phylogenetic comparisons of apoA-I sequences were published ( 36,37 ). Each of these reviews compared the sequences of apoA-I from 11 species that were only partially identical, for a total of 14 species: 10 species of mammals, 2 of birds, and 2 of fi sh.
Based upon similarities among teleosts (a major infraclass of ray-fi nned fi sh) and human apoA-I sequences and the absence of an apoA-I ortholog in lamprey eels (a cyclostomatous, jawless fi sh), it is generally accepted that the repeat patterns of the apoA-I protein family arose between the teleost fi sh-mammal split, approximately 400 million years ago ( 38 ), and the cyclostomatous fi sh origin, approximately 430 million years ago. Sequences of apoA-I from sharks and other cartilaginous fi sh (animals that arose approximately 420 million years ago) have, as yet, not been published.

CORRELATION OF ALIGNMENT RESULTS
WITH CURRENT STRUCTURAL AND FUNCTIONAL PARADIGMS WHEEL motif of apoA-I In our double-belt model for discoidal HDL ( 19 ), apoA-I monomers form a curved, fl at amphipathic ␣ helical ring with 11/3 (approximately 3.67) residues per turn, termed an ␣ 11/3 helix, which has since been reported in other lipid-associating proteins such as ␣ -synuclein ( 53,54 ), so that the hydrophobic surface faces inward toward the lipid disc. Figure 2 shows a continuous ␣ 11/3 helical wheel diagram of tandem helixes 2-9 of apoA-I. Assigning the six Pro residues to ␣ 11/3 helical wheel position 1 ( wp 1), wp 6 is occupied entirely; wp 10 and wp 3 are occupied mostly by hydrophobic residues; wp 9 and wp 7 are occupied mostly by positively charged residues; and wp 5, wp 8, and wp 4 are occupied mostly by negatively charged residues, the class A amphipathic ␣ helix pattern ( 12 ). Only helical wp 2 is equally divided between positively and negatively charged residues. We have suggested that the antiparallel helical rings are held together by 20 interhelical salt bridges ( 19,55 ), 6 bridges between solvent-shielded ␣ 11/3 helical wp 2 -wp 2 and 14 bridges between the more-solvent-exposed by generating an all-atom particle containing 2 apoA-I molecules in a 100% ␣ 11/3 helical wheel conformation wrapped in an antiparallel double belt around a POPC/UC bilayer and then subjected to molecular dynamics (MD) simulated annealing (MDSA) 16 times by the methods described previously ( 52 ).
The fraction of frames in which each residue was determined to be ␣ helical was measured over the last 10 ns of the ensemble of 16 MDSA 160:24:2 simulations by using the visual molecular dynamics (VMD) implementation of the secondary structure determination software, STRIDE (52a). Fractions for each MDSA were averaged and the SEM was calculated for each residue.
General phylogenetic analysis Figure 1 shows an unrooted phylogenetic tree of apoA-I sequence orthologs from 31 species of fi sh, amphibians, birds, and mammals (for sequence alignments of individual orthologs, see Fig. 3). Unrooted phylogenetic trees visualize the level of mutual similarity among homologous sequences without direct referral to chronological relationships among members of the set. Note the separation of the two largest branches of the sequences, fi sh from nonfi sh. Fig. 1. An unrooted phylogenetic tree shows apoA-I sequence orthologs from 31 species of fi sh, amphibians, birds, and mammals. Version numbers used are shown next to species names. drogen exchange-mass spectrometry in the monomeric lipid-free form ( 56 ) to be in a highly helical domain ( Fig.  4 ). Residue Y18 has been shown by Tanaka et al. ( 57 ) to be important for assembly of HDL complexes through the interactions of apoA-I with phospholipid, "the ␣ helix around Tyr18 [conceals] a potential lipid-binding region in the N-terminal domain, which [is] exposed by the disruption of the helix-bundle structure." Results of that study indicate that Y18 in the ␣ helical segment of the N terminus of apoA-I modulates lipid interactions of the lipid-free structure in concert with the C-terminal domain ( 58 ).
Residue Y115 in helix 4b, wp 3 ( Fig. 2 ), also predicted to be in a highly helical domain in the R2-2 lipid-associated form of apoA-I ( Fig. 4 ), is completely conserved from fi sh to mammals. Furthermore, Y115 and R116 (helix 4b, wp 4) are conserved as a pair in all but 3 of the 31 species and form, after the E191-Y192 pair, the second most conserved sequence pair in apoA-I ( Fig. 3B ). Furthermore, the completely conserved residues R83 and Y115 bracket the longest helical domain, residues 81-115 ( Fig. 4 ), predicted for monomeric lipid-free apoA-I by hydrogen exchangemass spectrometry ( 56 ). Thus, R83, Y115, and R166 could play a major role in the folding of monomeric apoA-I, perhaps required for proper membrane insertion and HDL assembly.
Residues Y192 (helix 8a, wp 6) and E191 (helix 8a, wp 5), predicted to be in an apparent loop-helix-loop domain in the lipid-associated form of apoA-I ( Fig. 4 ), form a 100% conserved sequence motif, suggesting strongly that this site in apoA-I has an important biological function. It is therefore of interest that Shao et al. ( 59 ) suggest that Y192 acts as a preferred atherogenic oxidation target in apoA-I.
The fi nal fully conserved residue, R83 (helix 2b, wp 7), at the edge of a predicted helix-nonhelix junction in both lipid-associated dimers and lipid-free monomers ( Fig. 4 ), has no known or suggested function. Perhaps it is important in HDL particle assembly, or it might be important in interactions with antioxidant proteins, such as PON-1 (see Discussion).
As shown in Fig. 2 , the 22 mer/11 mer tandem repeats are often punctuated by Pro residues in the ␣ 11/3 wp 1a position. In humans, three Pro residues are also located in the 10 residues of the N-terminal G0 domain of mature apoA-I. As shown in Fig. 3 , these three residues are not well conserved between fi sh and mammals. In human apoA-I, Pro occurs at wp 1 in helixes 2a, 4a, 5a, 6a, 7a, 9a, and 10a; in addition to helix 9, the other 11 mer repeat in the lipid-associating domain is helix 3, classifi ed as helix 3b (see Fig. 6). These seven Pro residues are well conserved from fi sh to mammals; P66 (helix 2), P121 (helix 5), and P220 (helix 10) are conserved in 100% of the species shown in Fig. 3 . The most variable Pro in the lipid-associating domain is at wp 1 of helix 8a; in humans and many mammals, this Pro, located at human sequence position 187, is replaced by a fl exible G185-G186-A187 loop and in several mammals by G184-G185-G186-A187. This variability suggests that structural fl exibility is important for the helix7-helix 8 junction. As noted later, in MD simulations of lipid-associated apoA-I ( 52 ), helix 8 is the ␣ 11/3 helical wp 5 -wp 9 and wp 9 -wp 5 ( Fig. 2 , connector lines ).
Completely conserved residues Figure 3 shows alignments and conservation of sequences of 31 individual orthologs. There are eight sequence positions that are conserved in all 31 of the orthologs ( Fig. 3, open arrowheads ).
Of the 10 Pro residues in human apoA-I, 3 residues are conserved in all 31 of the orthologs (P66, P121, and P220). Of the 7 Tyr residues in human apoA-I, 3 residues are conserved in all 31 of the orthologs (Y18, Y115, and Y192). The remaining two fully conserved residues are basic (R83) and acidic (E191).
While Pro residues almost certainly have a predominantly structural role in apoA-I, that is less likely to be true for Tyr. Thus, full conservation of 43% of the Tyr residues in apoA-I is highly signifi cant and suggests a major functional role for this amino acid residue. Two of these conserved Tyr residues are located in the lipid-associating domain, and one is located in the N-terminal domain.
Residue Y18 in helix G2 is predicted by MD simulations in the lipid-associated dimeric form of apoA-I and by hy- Fig. 2. Conservation of helical repeats and wheel positions (for helixes 2-9) in apoA-I. Helical wheel analysis is shown of tandem amphipathic ␣ helices 1-10 (residues 44-241) of human apoA-I, oriented with the N-terminal up, of tandem helices 2-9 of apoA-I plotted with a pitch of 11/3 residues/turn using the WHEEL program. Basic residues are represented by sky-blue circles with dark blue centers (Arg), sky-blue centers (Lys), and yellow centers (His); acidic residues are represented by solid red circles; hydrophobic residues are represented by solid black circles with magenta centers (Tyr), gold centers (Met), gray centers (Phe), and light-blue centers (Trp); and Pro is represented by solid green circles. Left and right docking interfaces are shown by arcs labeled L and R, respectively. Completely conserved residues are indicated by closed double arrowheads and largely conserved residues by open double arrowheads. Solvent-accessible salt bridges in human apoA-I are denoted by straight lines between the involved basic and acidic residues in wp 9 and wp 5, respectively, while the solvent-inaccessible salt bridges are shown by brackets linking the involved basic and acidic residues in wp 2.
Based upon MD simulations, we propose that the two N-terminal domains of apoA-I are "sticky" (M. K. Jones, et al., unpublished observations): that is, they tend to be locked together, sort of like molecular Velcro, in the discoidal double-belt model. Figure 5 illustrates one example least stable of the last 10 helical repeats in human apoA-I ( Fig. 4 ).

Conservation of residues in the N-terminal (globular) domain
G0 domain. Based upon its degree of conservation, the N-terminal G0 domain appears to be unusually important in apoA-I structure-function relationships. All apoA-I orthologs, except for those of eel, begin their N termini (G0-1) with Asp. Mammals, amphibians, and birds begin G0 with two N-terminal acidic residues, while fi sh (except eel) begin G0 with a single Asp residue. As noted earlier, G0 domains in all orthologs (except hedgehog) contain 1-3 Pro residues; in all but two orthologs, the third residue is Pro. Sequence position G0-6 relative to that of human apoA-I in all but one ortholog is either Ser or Thr, perhaps representing a site of phosphorylation. Sequence position G0-8 is almost always a hydrophobic residue (either Leu or Trp), G0-9 is generally an acidic residue, and G0-10 is generally basic, His in fi sh and Arg in mammals.

G1 domain.
Residue position G1-1 (residue 11) is almost always hydrophobic, while G1-2 (residue 12) is always of this spatial arrangement in a 160:24:2 particle simulated by MDSA. An illustration of this spatial arrangement in the average ensemble of four MDSAs was published previously ( 52 ). This "stickiness" is derived at least partially through a complex series of salt bridges, until, at a critical threshold of particle size, the interacting domains separate to expose extensive hydrocarbon regions of the phospholipid bilayer to solvent, putatively resulting in particle growth by fusion. Relevant to the discussion of the G0, G1, and G2 domains above, we have noted the presence of a variable cluster of nine interhelical salt bridges among the G0, G1, and G2 domains: D1-R10, D1-K12, E2-R10, E2-K12, D1-K23, D1-R27, D9-R10, K12-D13, and R10-D13. The fi rst four salt bridge pairs are those most commonly seen. Importantly, all seven of the charged residues that form these interhelical salt bridges are generally well conserved.

Well-conserved residues in the lipid-associating domain (helixes 1 -10)
A total of 13 ( Fig. 3 ) hydrophobic residues are well conserved in the lipid-associating domain. These conserved hydrophobic residues are located in hydrophobic wheel positions 2, 3, 6, and 10; fi ve residues are at wp 6, four are at wp 10, four are at wp 3, and one is at wp 2 ( Fig. 2 ).
There are 10 well-conserved basic residues ( Fig. 3 ) in the lipid-associating domain. Six residues are located in wp 7 ( Fig. 2, right docking interface), and of these, it has been suggested that R149 and R160 in helix 6a and 6b ( wp7 ) play a role as recognition sites for the binding of the LCAT enzyme ( 60 ). Two well-conserved basic residues are located in wp 9 (a superfi cial salt bridge position), and two are in wp 2 (a solvent-shielded salt bridge position).
Eight acidic residues ( Fig. 3, red arrowheads) are well conserved in the lipid-associating domain, three in the wp 4 position, three in the wp 5 (superfi cial salt bridge) position, and one in the solvent-shielded salt bridge in position wp 2 ( Fig. 2 ), and one in helix 1 at position 8a (we do not consider helix 1 to be an α11/3 helix). Other than Pro, there is only one sequence position in the lipid-associating domain where neutral residues ( Fig. 3, green arrowheads) are well conserved: T78 (helix 2b, wp 3). This Thr residue is conserved in all species, except one, where it has been mutated to an Ala residue. Met residues are generally not well conserved in apoA-I; M86 is conserved in most mammals but not in other animal classes.
Generally, the conserved residues of the lipid-associating domain are concentrated in its central domain, from the last half of helix 2 to the fi rst portion of helix 8 ( Figs. 3 and 4 ). This domain is both implicated in LCAT activation and the location of the solvent-shielded interhelical salt bridges ( Fig. 2 ).

Additional conserved motifs in the lipid-associating domain
The motif in helix 1a, N49-W50 ( wp 6 and wp 7) is completely conserved in mammals but not at all in fi sh, birds, or amphibians ( Fig. 3, small gold arrowheads,). Similarly, the motif in helix 2a, E70-F71-W72 ( wp 5, wp 6 and wp 7), is conserved in mammals but not in the other classes. Residues K77 and L159 in helix 2b ( wp 1) and helix 6b ( wp 6), respectively, are conserved in mammals, birds, and amphibians; in fi sh, K77 and L159 are replaced with an acidic residue and Thr, respectively. Finally, R153 in helix 6a ( wp 11) is completely conserved in mammals, birds, and amphibians but not in fi sh. These differences, especially those between fi sh and mammals, suggest certain signifi cant local structurefunction relationship differences in the apoA-I sequences between the two classes. For example, R153 has been suggested to be important for LCAT association with lipidbound apoA-I ( 60 ), and the lack of conservation in fi sh suggests the possibility of divergent evolutionary changes in LCAT structure between fi sh and land animals.

Location of conserved residues on the ␣ 11/3 helical wheel plot
Conserved residues, completely conserved residues, and well-conserved residues are shown in Fig. 2 . The most conserved wheel positions are wp 6, wp 7, and wp 1. Three of the seven conserved residues in the always-hydrophobic wp 6 are aromatics, and two, the Tyr residues, are completely conserved. Five of the six conserved residues in wp 7 are Arg, and one of these is completely conserved. All of the conserved residues in wp 1 are Pro. The next most conserved wheel position is wp 2, with four conserved residues. The remainder of the wheel positions have either three ( wp 5, wp 10, wp 3, and wp 4) or two ( wp 9) conserved residues, except for wp 8, which has none.
The pattern of conservation of amino acid residues shown in Fig. 2 strongly supports our model for apoA-I in which the lipid-associating domain forms an ␣ 11/3 helix (3 helical turns in 11 residues), rather than the classic tremely well conserved in mammals and poorly conserved in fi sh. In helix 4a, Y100, L101, and F104 ( wp 2, wp 3, wp 6, respectively) and helix 4b ( wp 2) are well conserved in mammals, and, while these residues are conserved in fi sh, they are conserved as different residues. In helix 4a mammals, N102 and W108 ( wp 3 and wp 10) are also well conserved, but in fi sh, residue 102 is not well conserved, and residue 108, while hydrophobic, is not aromatic.
In helixes 5a and 5b, A130 and K133 ( wp 10 and wp 2, respectively) are well conserved in mammals but are basic and acidic, respectively, in fi sh. On the basis of MD simulations, we have suggested that these two residues are critical to the formation of an amphipathic presentation tunnel located between the antiparallel helix 5-helix 5 pair of the double belt, for migration of hydrophobic acyl chains and amphipathic UC from the bilayer to the phospholipase A2-like and esterifi cation active sites of LCAT, respectively ( 62 ). This model is similar to one proposed by Martin et al. ( 26 ) based upon structure studies suggesting a looped belt located between residues 133 and 146, centered at residue 129. The major differences in the two models are: i ) our presentation tunnel is centered at residue 133 rather than 139, and ii ) the looped-belt model requires a change in the registration of the antiparallel helix 5-helix 5 central domain. Because MD simulations suggest that K133 is a key residue in stabilizing the presentation tunnel ( 62 ), it is entirely possible that any mutation at this position that incorporates a spin label or a fl uorescent molecule would signifi cantly change the dynamics of the presentation tunnel structure.
In helix 6b, residue 155 ( wp 2, a solvent-shielded salt bridge) is well conserved as His in mammals and completely conserved as Asp in fi sh. In helix 6b, residue 159 ( wp 6, generally a hydrophobic residue) is completely conserved as Leu in mammals and as Thr in fi sh. Finally, in helix 7a, residue 166 ( wp 2) is conserved as Tyr in all but one mammal but is uniformly a nonaromatic hydrophobic residue in fi sh.
Pauling-Corey-Branson ␣ helix of 5 turns in 18 residues ( 61 ). Parenthetically, in the initial Pauling-Corey-Branson publication ( 61 ), 3 turns per 11 residues was mentioned as one possible solution to the protein helix problem.
Comparison of similarities and differences between apoA-I sequences of fi sh and mammals in the lipid-associating domain Figure 6 shows a comparison of the differences between mammalian and fi sh sequences for each of the tandem amphipathic helical repeats 2 -9 in the lipid-associating domain (residues 65 -220 for humans). In addition to the completely conserved residues mentioned earlier, there is one charged motif of six residues in helix 7 that stands out as being well conserved in both fi sh and mammals ( Fig. 6 ; see also Fig. 3C), namely, R177-L178-E179-E180-L181-K/ R182 helix 7b ( wp 2 -wp 7, respectively). Residue R177 forms a solvent-shielded salt bridge with E89 in all PL-rich reconstituted particle sizes (R2-2, R2-1, and R2-0) subjected to MDSA simulations ( 52 ), and both are well conserved. A number of publications have suggested that the central domain of the apoA-I double-belt, helixes 3 -7, is critical for LCAT activation; perhaps the conserved 177 -182 motif is involved.
In addition to similarities, there are a number of dramatic differences between the two classes ( Fig. 6 ). In helix 2a, residue 72 ( wp 7, is usually a basic residue in most helical repeats) is always a Trp in mammals but is quite variable in fi sh (see Fig. 2). At the junction of helix 2a and 2b, the E76-K77-E78 cluster ( wp 11 -wp 1 -wp 2, respectively) is ex-  In fi sh, central helixes 4 -8 are signifi cantly more conserved than terminal helixes 1 -3 and 9 -10. Except for helix 10, this type of dichotomy between central and terminal helixes is not discernible in mammals, perhaps because of the greater phylogenic distances between different species of fi sh compared with that of mammals.
The six residues involved in forming the three pairs of solvent-inaccessible salt bridges in human apoA-I are E78 (helix 2)-R188 (helix 8), D89 (helix 3)-R177 (helix 7), and E111 (helix 4)-H155 (helix 6). Five of the putative solventinaccessible salt bridge residues appear in mammal consensus wheels; the lack of conservation of the sixth residue, R188, in mammals is compatible with our observation that the most terminal pair of solvent-inaccessible salt bridges, E78 (helix 2)-R188 (helix 8), unlike the two more central ones, are easily broken (and are thus unstable) during temperature-jump MD simulations ( 52,55 ). In the fi sh consensus wheels, only the putative solvent-inaccessible salt bridge residues D89 (helix 3)-R177 (helix 7) are conserved; the other four putative solvent-inaccessible salt bridge residues are not conserved.
The conservation of one of the three pairs of putative solvent-inaccessible salt bridge residues in fi sh suggests that the LL5/5 registration of apoA-I dimers might persist in fi sh, a possibility briefl y discussed in our original detailed double-belt model for discoidal HDL ( 19 ). To further test the phylogenetic persistence of the LL5/5 Finally, none of the 11 species of fi sh has even a single Trp residue; Trp fi rst appears in apoA-I at position W108 in land animals. Four Trp residues appear in humans, W7 (G0), W50 (H1), W72 (H2, wp 7), and W108 (H4, wp 10). The fi rst three Trp residues ( Fig. 3A ) are completely conserved in mammals but are present in no other classes of animals. The fourth Trp, W108, is conserved in all mammals but hedgehog and in all other land animals ( Fig. 3B ). Figure 7 shows helical wheel diagrams of consensus sequences of lipid-associating domain helical repeats 1 -10 of fi sh and mammals. Sequence positions with less than 50% identity are denoted by X. In the fi gure, the fraction of residues in each helical wheel that is at least 50% conserved is shown. With that as a measure of helical repeat conservation, helix 9 is the most conserved, and helix 10 is the least conserved consensus tandem repeat in mammals. Using a cutoff higher than 50% identity, helix 10 is also the least conserved helical repeat in mammals (data not shown). In fi sh, helix 7 is the most conserved, and helix 9 is the least conserved. As helix 7 is well conserved and helix 10 is poorly conserved in both mammals and fi sh, these appear to represent the most conserved and least conserved consensus helixes, respectively, across the phylogenetic tree.  6. Comparison of differences between mammalian and fi sh sequences for each of the tandem amphipathic helical repeats 2-9 in the lipid-associating domain (residues 65-220 for humans) is shown. Top and bottom scales indicate conservation in mammals and fi sh, respectively. If the residue is completely conserved across all species, no residue is indicated on the graph (e.g., see residue 66). If there are differences between the species, the height of the letter indicates conservation of given residues within mammals and within fi sh (e.g., residue 77 is 100% Trp in mammals but is a mixture of residue types in fi sh). Black diamonds, 100% conserved residues; yellow diamonds, comments; red diamonds, conserved residue clusters across both fi sh and mammals. lated with current concepts of apoA-I structure and function. The general overall conservation of Pro residue punctuations between most of the 11/22 mer tandem helical repeats is not surprising as Pro almost certainly plays a mostly structural role in apoA-I.

Comparison of consensus sequences of tandem amphipathic helical repeats in lipid-associating domains of apoA-I between fi sh and mammals
The complete conservation of three of seven Tyr residues in apoA-I, however, is functionally more signifi cant. It has been suggested that certain Tyr residues, in particular Y166 ( 34 ) and Y192 ( 59 ), are preferred targets for sitespecifi c oxidative modifi cation within atheroma. However, this concept alone provides little in the way of biological insight into the function of Tyr residues in apoA-I.
Although it is a somewhat controversial idea, Tyr residues have been suggested to act as antioxidants inside lipid bilayers and protect cells from oxidative destruction by interfering with lipid-oxidizing free radical chain reactions ( 63 ) through conversion to relatively stable and nonreactive phenoxyl radicals ( 64,65 ). The re-reduction of oxidized, radicalized membrane phase antioxidants, such as vitamin E, is known to require aqueous reducing components, such as vitamin C ( 66,67 ). As in the amphipathic helical domains of apoA-I in general, Tyr residues have a strong tendency to partition into the boundary between the polar and nonpolar regions of bilayers. Considering registration of apoA-I dimers from fi sh to mammals, we performed analyses, using ALIGN software ( 19 ), of apoA-I sequences for four mammals (human, baboon, rabbit, and dog), one bird (chicken), one amphibian (frog), and two fi sh species (zebrafi sh and seabream); results are shown in Fig. 8 . While there are differences in seven of the eight species in the relatively weighted salt bridge scores of the three registrations emphasized previously, LL5/4, LL5/5, and LL5/6, each of the seven, including zebrafi sh, a tropical freshwater fi sh, has a pronounced 11/22 mer periodicity. However, seabream, a saltwater fi sh, shows no obvious 11/22 mer periodicity at all along the LL docking interface, suggesting that interhelical salt bridge formation between left docking interfaces of apoA-I monomers may be a recent evolutionary adaptation that occurred since freshwater fi sh split off from saltwater fi sh.

Structure and function roles of completely conserved Pro and Tyr residues
The detailed sequence alignment and conservation analyses of apoA-I amino acid described here provide a phylogenetic window into the structure, assembly, function, and dynamics of apoA-I. These results are useful as stand-alone information but are more valuable if corre-  ( 51 ), the helical wheel for helix 1 (residues 44-65) is displayed as an idealized ␣ 18/5 helix, while the remainder are displayed as ␣ 11/3 helixes. The helix number is shown in the center of each wheel (e.g., H1, H2, etc.). The fraction of residues in each helical wheel that is at least 50% conserved (c) is shown in red. The position of the residues involved in the putative solventinaccessible salt bridges in the human sequence are shown as stars: red stars denote acidic residues that are conserved, blue stars denote basic residues that are conserved, and black stars denote residues that are not conserved.
double-belt structure of the lipid-associating domain (residues 44 -243) of apoA-I on discoidal HDL ( 52 ). In the majority of our simulations, the region of helix 8 containing the conserved E191-Y192 motif formed a solvent-exposed loophelix-loop on the edge of the bilayer disc [ Fig. 4 and ( 52)] . Thus, the fully conserved E191-Y192 motif appears well situated to perform an antioxidant function within HDL.
Interestingly, Y100 and Y166 (both at wp 2 in helix 4a and 6a, respectively) create a unique microenvironment in most mammals, being positioned to form -ring stacks ( Fig. 9 ), perhaps stabilizing interhelical interactions in the extensive dynamics of lipid bilayers and apoA-I in HDL particles ( 52,55,62 ), Tyr seems ideally suited to form a partition at the polar and nonpolar interface of the lipid bilayer to scavenge lipid peroxyl radicals and in the aqueous phase to be re-reduced ( 65 ). With this in mind, the location of the completely conserved residues Y115 and Y192 in the most-solvent-shielded ␣ 11/3 wheel position, wp 6 ( Fig. 2 ), is remarkable.
Modeling and MD simulations ( 68,69 ) predict that the completely conserved residue Y115, in its lipid-associated form, is in a highly helical domain ( Fig. 4 ). M112 (helix 4b, wp 3), found only in primates and pigs, and M148 (helix 6a, wp 6, Fig. 2 ), found only in humans and a few fi sh, interact with Tyr115 during MD simulations of the LL5/5 rotamer ( Fig. 9 ); the driving force for this interaction is not clear. In no species, other than human, for which apoA-I sequences are available does this intriguing microenvironment occur. Oxidation of Met residues introduces hydrophilicity to the nonpolar face of the amphipathic helical structure of human apoA-I, altering its lipid affi nity and local secondary structure ( 70 ). Thus, the presence of M112 and M148 in a microenvironment unique to humans, combined with the possibility that they are easily oxidized, may have a deleterious effect on the local structural environment and function (perhaps as an antioxidant?) of what should otherwise be a solvent-inaccessible residue Y115 in humans.
We recently reported, based upon MD simulations, that helix 8 is the least stable tandem amphipathic helix in the  in fi sh or amphibians ( Fig. 3C ), suggesting that the YXXK motif arose at or before the branching of birds and mammals and after amphibians. This conservation suggests that the motif has a functional role, perhaps as an antioxidant.
Scanning spin label studies have suggested that an 18residue random coil loop (amino acids 188-205) near the C terminus becomes ␣ helical when the protein associates with lipid and energetically drives apoA-I's conformational adaptation to the presence of lipid ( 33 ). Our MD simulations suggest that the E191-Y192 motif when associated with lipid lies in the helical portion of a loophelix-loop motif ( Fig. 4 ). As Y192 is in the helical portion of this 18-residue region, its chlorination might alter the energy levels of protein remodeling to favor retention of this region. There are, however, no aromatic residues at either position in fi sh. As Y166 ( 34 ) has been suggested to be a preferred target for site-specifi c oxidative modifi cation within atheroma, the uniqueness of the Y100/Y166 microenvironment to mammals is intriguing.

Myeloperoxidase oxidative pathway for generating dysfunctional HDL
While the functional role of Tyr in apoA-I is not clearly understood, there does exist extensive literature on the role of oxidative damage to HDL directed toward Tyr (as well as Met and Trp) residues of apoA-I by myeloperoxidase (MPO), a heme enzyme secreted by human artery wall macrophages ( 71 ).
Macrophages use reactive oxygen species to kill invading pathogens ( 72 ). One oxidative pathway involves MPO that is expressed in macrophages in human atherosclerotic lesions ( 73 ). The enzyme uses hydrogen peroxide, and the major end product is generally thought to be hypochlorous acid (HOCl), a potent antimicrobial agent.
A single tyrosine residue, the fully conserved Y192, has been identifi ed as the major chlorination site when HOCl oxidizes apoA-I ( 74 ), and a strong linear association has been shown to exist between the extent of Y192 chlorination and the loss of ABCA1 transport activity ( 75 ). Chlorination of Y192 has been shown to be directed by the YXXK motif ( 74 ). The K in this motif, K195, is conserved in all but 2 of 14 mammals and in all 3 birds but is not conserved  Fig. 11. Three examples of Arg cation-interactions with aromatic residues in basic fi broblast growth factor 1. This stereo image of PDB coordinates taken from the Protein Database was produced with RasMol. Aromatic residues (green) and Arg residues (CPK color) forming cation-pairs are shown in space-fi lling mode.
sults, it was proposed that oxidation of M148 disrupts the LCAT activation domain of lipid-associated dimeric apoA-I ( 79 ) by disrupting the central loop.
We would add that M148 also appears by MD simulations to interact strongly with the completely conserved Y115 residue. As noted later, Y115 appears to form interhelical cation-orbital interactions with R151, thus helping to stabilize the edge of the putative presentation tunnel for LCAT activation. Stabilization of the tunnel edge would almost certainly be seriously affected by oxidation of Met to create a negatively changed residue.
Role of the prevalence of Arg residues on the right docking interface Figure 10 shows a histogram plot of the fraction of frames involved in salt bridge formation, between all basic residues and POPC during the last 10 ns of MD simulations of 16 160:24:2 particle replicas. All Arg residues located in wp 7 and two of the three located in wp 11 on the right docking interface ( Fig. 2 ) are involved in salt bridge the conformation of lipid-free apoA-I and thus inhibit ABCA1 transport activity ( 71 ). The complete conservation of the E191-Y192 motif may be related to the putative role of residues 188 -205 in the assembly of lipidated apoA-I.
HOCl can also oxidize tryptophan residues in apoA-I ( 76,77 ). However, oxidation failed to prevent an apoA-I mutant in which all four Trp residues were replaced with Phe from activating ABCA1 ( 78 ).
Met residues, which are not well conserved, are also extremely susceptible to oxidation by HOCl. A single methionine residue in apoA-I, M148, resides in the general vicinity of the LCAT activation domain of apoA-I (H4 -H6, residues 99 -164). Shao et al. ( 79 ) demonstrated that oxidation of M148 to the negatively charged methionine sulfoxide produced a loss of LCAT activity in HDL. Reversing oxidation with methionine sulfoxide reductase restored the ability of HDL to activate LCAT. Discoidal HDL prepared with apoA-I containing a mutation in which Met was replaced with Leu at position 148 (M148L) was significantly resistant to inactivation by MPO. Based on these re- exists for lipid-free monomeric apoA-I, we will confi ne our discussion to the lipid-associated form.
Inspection of the four MDSAs of the 160:24:2 doublebelt particle reveals that of the three completely conserved Tyr residues, Y115 and Y192 frequently form cation-orbital interactions with basic residues within apoA-I itself.
Y115, in the majority of the eight possible pairings, forms interhelical cation-orbital interactions with R151, suggesting that this pairing might be important in stabilizing the double belt. However, unlike Y115, R151 is only moderately well conserved ( Fig. 3 ).
Residue Y192, on the other hand, forms intramolecular cation-orbital interactions in the majority of the eight possible pairings with R188 (an n+4 intrahelical pairing). Perhaps this intramolecular interaction might be related to the tendency of this region of helix 8 to form a helical domain surrounded by coils ( Fig. 4 ). Interestingly, however, R188 is conserved only in primates (and one frog species); this residue position is completely conserved as Tyr in fi sh. One could speculate that the uniqueness of the Y192-R188 intramolecular pairing in humans might play a role in the tendency for Y192 to act as a preferred atherogenic oxidation target in apoA-I ( 59 ).
Finally, examination of the six Phe residues in human apoA-I, F33 (G3), F57 (H1), F71 (H2, wp 6), F104 (H4, wp 6), and F225 and F229 (both H10), revealed that F71 forms several interhelical cation-interactions with R188, K195, and H199, while F104 forms a few intramolecular cation-interactions with K107. Interestingly, these two are the only Phe residues conserved in apoA-I: F71 is completely conserved in mammals and in a few fi sh, while F104 is almost completely conserved in land animals ( Fig. 3A and B ) but not in fi sh.
formation with the negatively charged PO 4 Ϫ moieties of the annular POPC molecules over 70% of the time. This is one possible explanation for why fi ve of the six Arg residues in wp 7 are well conserved ( Fig. 2 ). One additional possible function of Arg in the right docking interface in general and in wp 7 in particular is based on recent research from this laboratory showing that Arg-containing amphipathic helical peptides sequester oxidized lipids more avidly than corresponding Lyscontaining peptides ( 80 ), suggesting that Arg may associate more avidly with oxidized phospholipids than Lys (see Fig.  12E for the proposed mechanism). Since the other major HDL protein, apoA-II, contains no Arg residues, perhaps the presence of Arg residues in the right docking interface of apoA-I partially explains why apoA-I-containing particles are more anti-atherogenic than apoA-II-containing particles ( 81 ).
In a related possibility, it is known that the fl at face of an aromatic ring has a partial negative charge due to orbitals. It has been suggested that cationic amino acid residues, Arg or Lys, which are within 6.0 Å of the face of an aromatic ring (Phe, Tyr, or Trp), may engage in polar interactions called cation-orbital interactions by centering themselves over the faces of aromatic rings; Arg participates in nearly twice as many cation-interactions as Lys does; and the numbers of cation-interactions involving Trp, Tyr, and Phe are roughly similar ( 82 ). For three examples of Arg-interactions in a single protein, basic fibroblast growth factor, see Fig. 11 . In addition, because of the conjugation between the double bond and the nitrogen lone pairs in Arg, the positive charge is delocalized, enabling Arg to form multiple H bonds. Thus, despite interactions with PO 4 Ϫ and/or oxidized lipids ( Fig. 12E ) , Arg would be available for interactions with certain critical proteins, such as antioxidants.
Based upon X-rays of the crystal structure of PON-1 ( 83 ), not only is the HDL binding domain of PON-1 enriched in -electron-containing aromatic residue clusters ( Fig. 12 ) but even more remarkably, a complete ring of aromatic residues encircles the opening of the cleft that leads to the active site of PON-1 ( Fig. 12A-C ). Binding of PON-1 to HDL, per se, appears to happen through the signal sequence that is still attached to PON-1 ( 84 ). It is therefore tempting to speculate that the Arg residues in wp 7 on the right docking interface orient PON-1 to an oxidized lipid binding site of apoA-I by interacting withelectron-rich clusters of the HDL binding domains of this antioxidant enzyme ( Fig. 12D ). One potential driving force in the creation of an oxidized lipid binding site on apoA-I is the planarity of the quanidino group of Arg that causes delocalized planar positive charges to interact with the planar delocalized negative charges of the SN-2 chain of oxidized PC ( Fig. 12E ).

Possible role of cation-orbital interactions within the double-belt structure
Another possibility involves considering whether cation-orbital interactions could play a role in the structure of apoA-I itself. As no detailed structural information