Site-specific 5-hydroxytryptophan incorporation into apolipoprotein A-I impairs cholesterol efflux activity and high-density lipoprotein biogenesis

Apolipoprotein A-I (apoA-I) is the major protein constituent of high-density lipoprotein (HDL) and a target of myeloperoxidase-dependent oxidation in the artery wall. In atherosclerotic lesions, apoA-I exhibits marked oxidative modifications at multiple sites, including Trp72. Site-specific mutagenesis studies have suggested, but have not conclusively shown, that oxidative modification of Trp72 of apoA-I impairs many atheroprotective properties of this lipoprotein. Herein, we used genetic code expansion technology with an engineered Saccharomyces cerevisiae tryptophanyl tRNA-synthetase (Trp-RS):suppressor tRNA pair to insert the noncanonical amino acid 5-hydroxytryptophan (5-OHTrp) at position 72 in recombinant human apoA-I and confirmed site-specific incorporation utilizing MS. In functional characterization studies, 5-OHTrp72 apoA-I (compared with WT apoA-I) exhibited reduced ABC subfamily A member 1 (ABCA1)-dependent cholesterol acceptor activity in vitro (41.73 ± 6.57% inhibition; p < 0.01). Additionally, 5-OHTrp72 apoA-I displayed increased activation and stabilization of paraoxonase 1 (PON1) activity (μmol/min/mg) when compared with WT apoA-I and comparable PON1 activation/stabilization compared with reconstituted HDL (WT apoA-I, 1.92 ± 0.04; 5-OHTrp72 apoA-I, 2.35 ± 0.0; and HDL, 2.33 ± 0.1; p < 0.001, p < 0.001, and p < 0.001, respectively). Following injection into apoA-I–deficient mice, 5-OHTrp72 apoA-I reached plasma levels comparable with those of native apoA-I yet exhibited significantly reduced (48%; p < 0.01) lipidation and evidence of HDL biogenesis. Collectively, these findings unequivocally reveal that site-specific oxidative modification of apoA-I via 5-OHTrp at Trp72 impairs cholesterol efflux and the rate-limiting step of HDL biogenesis both in vitro and in vivo.

Proteomic analyses of apoA-I recovered from human atheroma revealed that the protein is heavily oxidized, bearing the oxidative footprint of myeloperoxidase (MPO), an enzyme secreted by activated leukocytes (29,30). Lesion apoA-I is dysfunctional, lacking cholesterol acceptor activity, and has disease-promoting pro-inflammatory activity (30,31). Understanding the functional consequences of apoA-I modification has been the subject of intense investigation with efforts focused on recapitulating the physiological modifications in vitro using MPO-catalyzed modifications with chlorinating and nitrating oxidation systems. These studies are limited by oxidation on all susceptible residues, thereby confounding interpretation of functional studies and assignment of impaired activities to specific sites. Although natural amino acid substitution studies confirmed the relevance of oxidation at specific residues, the approach was also limited by unknown consequences of incorporating a nonoxidizable amino acid on the structure/ function of the protein. Thus, the functional consequence of site-specific (single-site) oxidation of apoA-I remains poorly understood.
Site-directed natural amino acid mutagenesis studies support a crucial role for apoA-I Trp 72 in ABCA1-dependent efflux activity (31,32). ApoA-I oxidized on Trp 72 (oxTrp 72 -apoA-I) is present in plasma of patients with coronary artery disease (CAD), is highly enriched in atherosclerotic lesions, and is associated with increased cardiovascular disease risk (31). In this study, we sought to address the absolute effect of oxidation at Trp 72 in the absence of any other modification in apoA-I. Genetic code expansion (GCE) has emerged as a powerful method to site-specifically incorporate noncanonical amino acids (ncAAs) into proteins in living cells (33). Specifically, engineered aminoacyl-tRNA synthetase:tRNA pairs are used to deliver a desired ncAA in response to a nonsense or frameshift codon. Herein, we used an evolved tryptophanyl tRNA-synthetase (Trp-RS):suppressor tRNA pair from Saccharomyces cerevisiae (34) to insert oxTrp at position 72 in apoA-I in Escherichia coli. We now report a significant impairment of in vitro cholesterol acceptor activity and in vivo HDL biogenesis function observed with 5-OHTrp 72 apoA-I relative to WT apoA-I.

Site-specific incorporation of 5-OHTrp in apoA-I by GCE
To examine the specific functional consequences of a singlesite oxTrp at position 72 in apoA-I, we used GCE to cotranslationally incorporate a noncanonical amino acid in vivo (33). We used an engineered Trp-RS:suppressor tRNA pair (engineered machinery) from S. cerevisiae, which incorporates 5-OHTrp in E. coli in response to a genetically programmed amber nonsense codon (TAG) (34). We first explored the possibility that this machinery might exhibit substrate promiscuity, thereby allowing us the opportunity to insert different oxidized tryptophans. The experimental approach is schematically depicted in Fig. 1A, where amber suppression in superfolder GFP (sfGFP) with a 150TAG leads to fluorescence of intact bacterial cells due to expression of full-length protein. We first examined the specificity of this orthogonal pair for oxidized tryptophan by co-transforming the plasmid encoding the engineered machinery with either one of two reporters encoding sfGFP-150TAG (Fig. 1, A and B), or apoA-I-72TAG (Fig. 1C), as described under "Experimental procedures." Transformed E. coli BL21ai cells were grown on defined autoinduction media in the presence of the indicated oxidized tryptophan (oxTrp) forms without or with added lactose to induce Trp-RS R313 (see "Experimental procedures"). A robust amber suppression resulting in full-length sfGFP protein expression was observed only in the presence of 5-OHTrp, exhibiting an 8.8 Ϯ 3.8-fold increase over lactose-induced control culture (Fig. 1B). This suggests that Trp-RS R313 selectively aminoacylates its cognate tRNA (tRNA Trp CUA 40A) with 5-OHTrp and not with 2-OHTrp, 2,3-Dioxindolyl alanine (2,3-diOHTrp) or kynurenine. There was a modest induction of full-length sfGFP in lactose-induced cultures in the absence of added oxTrp, which suggested that Trp-RS R313 may weakly aminoacylate its cognate suppressor tRNA with an endogenous (nonoxidized) Trp, which was present in the media at low levels, because Trp is an essential amino A, schematic depiction of site-specific incorporation of oxTrp using an engineered S. cerevisiae tryptophanyl-tRNA synthetase (S.c.oxTrp-RS) and its cognate suppressor tRNA (S.c.tRNA Trp CUA ). In the presence of oxTrp, the synthetase charges its suppressor tRNA, culminating in the synthesis of full-length sfGFP with oxTrp at position 150 (amber stop codon). B, E. coli BL21Ai cells co-transformed with plasmids encoding the oxTrp-RS/tRNA Trp CUA machinery and either sfGFP or sfGFP-150TAG reporter plasmids were grown in defined autoinduction media in the presence of different oxTrp with or without lactose to induce oxTrp-RS. Fluorescence (in intact cells) was measured after 64 h of growth at room temperature and normalized to cell density as described under "Experimental procedures". C, the amber codon in apoA-I-72TAG is suppressed only in response to 5-OHTrp. ClearColi cells co-transformed with the machinery and a plasmid encoding apoA-I with amber codon at position 72 were grown in LB and induced with IPTG in the presence of different oxTrp at room temperature. Crude extracts were examined for the presence of full-length Histagged apoA-I by Western blotting, as described under "Experimental procedures." Anti-apoA-I (top) and anti-His (bottom) antibodies were used. n.s., nonspecific.

5-OHTrp 72 apoA-I has impaired atheroprotective activities
acid and needed to be present for incorporation into other sites in sfGFP. Notably, yield of the mutant sfGFP in response to 5-OHTrp was 50% of the WT sfGFP reporter (no stop codon) (Fig. 1B). This suggests that the Trp-RS R313:tRNA Trp CUA 40A orthogonal pair is quite efficient at suppressing the nonsense codon, at least with respect to position 150 in sfGFP under the experimental conditions used.
We next proceeded to test whether we could insert 5-OHTrp at position 72 in apoA-I. Because WT apoA-I has an innate ability to bind lipopolysaccharide (35)(36)(37), we used the host strain ClearColi BL21 (DE3) (Lucigen, Madison, WI), which has been genetically engineered to express a modified lipopolysaccharide that does not trigger the endotoxic response (38), to express our recombinant apoA-I proteins, as described under "Experimental procedures." Analysis of E. coli crude extracts by SDS-PAGE followed by Western blotting with antibodies specific for apoA-I and His tag showed that full-length His-tagged apoA-I protein was expressed only in the presence of 5-OHTrp (Fig. 1C).
To confirm that the expressed full-length recombinant apoA-I contained 5-OHTrp at position 72, His-tagged WT (lacking amber (TAG) codon but otherwise having identical DNA sequence to apoA-I-72TAG) and the mutant protein were expressed in ClearColi, purified by nickel-nitrilotriacetic acid chromatography, and processed for proteomic analysis (see "Experimental procedures" for details). Equivalent amounts of each protein were fractionated by gradient (4 -20%) reducing SDS-PAGE and visualized by Coomassie Blue (Fig.  2A). The mutant protein exhibited similar purity and equivalent gel fractionation properties compared with WT His 10 -apoA-I ( Fig. 2A). Native peptides from WT and 5-OHTrp 72 apoA-I variant, generated by trypsin digestion, were identified by nano-LC-MS/MS. Whereas detectable oxidation at Trp 50 and Trp 108 was minimal, the degree of oxidation at Trp 72 in the mutant protein was 85% with the remaining 15% being WT tryptophan (Fig. 2B). This suggests that the incorporation fidelity of 5-OHTrp by the engineered orthogonal pair is 85% in 5-OHTrp-supplemented media in response to amber codon. Analysis of the peptide spanning residues 62-77 revealed a mass increase of 16 Da on all y ions present after the oxTrp 72 in the noncanonical amino acid incorporated peptide ( 62 EQLG-PVTQEFW(ox)DNLEK 77 ) spectra compared with the same y ions in WT apoA-I (Fig. 2C). This mass differential is consistent with single incorporation of an oxygen (5-OH moiety) on Trp 72 of apoA-I, corresponding to the location of the amber codon (TAG).

ApoA-I with site-specific incorporation of 5-OHTrp into position 72 is dysfunctional for cholesterol efflux activity
ABCA1-dependent cholesterol efflux activity from peripheral tissue and atherosclerotic lesions is an important atheroprotective function of apoA-I and is impaired through oxidation by MPO in vivo (30 -32, 40, 41). We previously reported that tryptophan residues (4 residues in total; Trp 8 , Trp 50 , Trp 72 , and Trp 108 ) are a target of MPO-mediated oxidation and inhibition of efflux activity (32). Furthermore, an apoA-I in which Trp 72 was site-specifically mutated to the relatively nonoxidiz- Recombinant (His 10 ) 5-OHTrp 72 apoA-I and His 10 -WT apoA-I were expressed in ClearColi BL21 (DE3) and purified by nickel-nitrilotriacetic acid affinity chromatography as described under "Experimental procedures." A, Coomassie-stained denaturing gel. B, an aliquot of the purified proteins was digested in solution with trypsin and analyzed by nano-LC/ MS/MS as described under "Experimental procedures." 5-OHTrp was incorporated at position 72 in apoA-I with high fidelity (ϳ85%). The remaining 15% had nonoxidized Trp at position 72. C, the collision-induced dissociation spectrum of WT (top) versus 5OHTrp 72 apoA-I peptide-spanning Trp 72 . Peptide y ions whose m/z is ϩ16 atomic mass units (due to the addition of an oxygen atom) relative to the native sequence are indicated in red.

5-OHTrp 72 apoA-I has impaired atheroprotective activities
able residue phenylalanine retained 50% of efflux activity after exposure to the MPO-H 2 O 2 -Cl Ϫ oxidation system, suggesting that Trp 72 was responsible for 50% of the ABCA1-dependent efflux activity in lipid-poor apoA-I (32). We therefore sought to evaluate the functional effect of selective Trp 72 oxidation (5-OH-Trp incorporation) on ABCA1-specific cholesterol acceptor function of apoA-I, relative to native (WT) recombinant apoA-I as control. Efflux activity with increasing concentrations of each recombinant protein were examined using [ 3 H]cholesterol-loaded RAW264.7 macrophage cells in the presence and absence of 8-bromo-cyclic AMP (8-Br-cAMP) pretreatment to induce ABCA1 expression, as described under "Experimental procedures." At each concentration tested, a significant impairment (up to 40%) in cholesterol efflux activity was observed with 5-OHTrp 72 apoA-I relative to WT protein (Fig. 4). Notably, the measured degree to which 5-OHTrp 72 apoA-I is "dysfunctional" (40%) with respect to cholesterol efflux activity is likely an underestimate, because proteomic analyses of the 5-OHTrp 72 apoA-I preparation revealed that 15% of the residues at position 72 were native Trp (Fig. 2B).

ApoA-I with site-specific incorporation of 5-OHTrp at amino acid position 72 is not proinflammatory
We have previously reported that apoA-I recovered from human plasma or atherosclerotic plaque using a mAb specific for oxTrp 72 displayed potent pro-inflammatory function, activating NF-B and vascular cell adhesion molecule 1 (VCAM1) surface protein expression in endothelial cells (31). However, a direct link between oxidation at Trp 72 and observed proinflammatory activity could not be established, because the immunoprecipitated apoA-I recovered harbored numerous other oxidative post-translational modifications in addition to oxTrp 72 (31). The GCE-mediated expression of 5-OHTrp 72 apoA-I in the present study resulted in incorporation of a single residue harboring an oxidative modification at position 72, enabling us to interrogate the role of 5-OHTrp at this position of apoA-I with respect to a potential gain of proinflammatory function. WT apoA-I exposed to the complete MPO-H 2 O 2 -Cl Ϫ system is proinflammatory (31,42) and was used as a positive control to elicit surface expression of VCAM1 on bovine aortic endothelial cells (BAECs). Equivalent amounts of apoA-I proteins used in the VCAM1 induction assay on BAECs were fractionated by gradient (4 -20%), reducing SDS-PAGE and visualized by Coomassie Blue. Fig. 5A illustrates changes in migration of WT apoA-I after exposure to the MPO-H 2 O 2 -Cl Ϫ system, and Fig.  5B shows MS-based quantification of apoA-I site-specific Trp oxidation detected (see "Experimental procedures"). In addition to Trp oxidation, MPO-dependent tyrosine and methionine residue oxidation were also detected (data not shown). Consistent with previous observations, MPO-catalyzed oxidation of recombinant apoA-I elicited a 2-3-fold induction in VCAM1 expression (Fig. 5C, left, MPO-exposed WT apoA-I batch 1 and 2 relative to respective native proteins). This proinflammatory activity likely is not attributable to oxidation at position 72, because 5-OHTrp 72 apoA-I, which bears a single site-specific oxidation at this position, was not proinflammatory under the same experimental conditions (Fig. 5C, right). It must be noted that the MS MS2 method does not distinguish apoA-I proteins were assessed for their ability to enhance PON1 arylesterase activity and compared with reconstituted HDL as the positive control. The assay was performed as described under "Experimental procedures" with BSA serving as negative control. The median value for each group is indicated. Individual p values represent comparisons between PON1 alone and PON1 with the indicated proteins or reconstituted HDL prepared with WT apoA-I and were determined using two-tailed Student's t test.

5-OHTrp 72 apoA-I has impaired atheroprotective activities
between 2-oxindolyl alanine (2-OHTrp) and 5-OHTrp, as both display an increase in mass of 16 Da, and it is conceivable that the oxidation at position 72 produced by MPO is 2-OHTrp rather than 5-OHTrp and that 2-OHTrp may, per se, be pro-inflammatory.

ApoA-I with site-specific incorporation of 5-OHTrp into position 72 is dysfunctional for HDL biogenesis in vivo
Our in vitro observations in Fig. 4 support a crucial role for Trp 72 in ABCA1-dependent efflux activity by apoA-I. Because this step initiates HDL biogenesis, we hypothesized that 5-OHTrp 72 apoA-I may be dysfunctional at assembling HDL particles in vivo. To test this, we used a mouse model, apoA-I knockout (AIKO), which we previously utilized for this purpose (31). Specifically, we subcutaneously injected female AIKO mice with equivalent amounts (250 mg/kg) of recombinant endotoxin-free His-tagged WT or 5-OHTrp 72 apoA-I proteins. Blood was collected over 4 h, and the prepared plasma was fractionated by sucrose/D 2 O density ultracentrifugation to per-mit separation of a top lipoprotein-enriched fraction (HDLcontaining fraction) from a lower lipoprotein-deficient fraction (LPDF) (Fig. 6A). We initially assessed the appearance of human apoA-I in plasma by Western blotting using a biotinylated mAb (Biotin-2D10.5) that we developed against human total apoA-I (29,31). We observed a comparable time-dependent appearance of total apoA-I in the circulation, suggesting that 5-OHTrp 72 apoA-I was as efficient as the WT protein at entering the vascular compartment (Fig. 6B).
We next examined whether the injected protein was associated with HDL in the circulation by probing the fractionated plasma and quantitating the total mass of apoA-I in each fraction using Western blot analysis, as described under "Experimental procedures." Human WT and 5-OHTrp 72 apoA-I were detectable in both fractions, but importantly, the levels of apoA-I in fractions derived from animals injected with the 5-OHTrp 72 apoA-I were 48% less in the HDL-containing fraction with a concomitant 95% increase in mutant protein levels in the LPDF fractions, with respect to WT protein (Fig. 6C). We previously demonstrated that the appearance of injected WT apoA-I in the HDL-enriched fraction is primarily ABCA1-dependent, because levels of injected human apoA-I were decreased by more than 70% in hepatic-specific ABCA1 knockout mice relative to corresponding controls (31). Our present results thus indicate that 5-OHTrp 72 apoA-I is inherently dysfunctional at assembling HDL particles in vivo.

Discussion
To the best of our knowledge, this is the first report investigating the functional consequences of inserting only a single oxidized Trp specifically into a protein and assessing its function both in vitro and in vivo. GCE technology was used to successfully incorporate 5-OHTrp site-specifically into human apoA-I with 85% fidelity. Although the remaining 15% of protein had WT Trp at position 72, we nonetheless observed striking functional effects relative to WT protein. Namely, 5-OHTrp 72 apoA-I exhibited a statistically significant impairment in ABCA1-dependent cholesterol efflux activity in vitro and a nearly 50% inhibition in capacity to assemble HDL particles in vivo. Furthermore, unlike MPO-H 2 O 2 -Cl Ϫ -modified apoA-I, the site-specific 5-OHTrp 72 -containing apoA-I does not induce surface expression of the pro-atherogenic adhesion protein VCAM1 on primary bovine aortic endothelial cells.
Proteomic studies have shown that apoA-I recovered from human atherosclerotic lesions is extensively oxidized, a phenomenon attributed in large part to MPO activity given the enhancement in oxidative post-translational modifications that can only be generated by MPO, such as chlorotyrosine (30,41,43). Although in vitro MPO-mediated oxidation assays in conjunction with natural amino acid substitution studies have confirmed the relevance of specific residues, the functional consequences of single-site oxidation/post-translational modification of residues are not fully understood. Loss of ABCA1-dependent cholesterol efflux activity, a presumed key atheroprotective function of apoA-I, has historically been associated with increased levels of chlorotyrosine-and nitrotyrosine-containing apoA-I recovered from subjects with cardiovascular disease (30, 40, 41). Specifically, Tyr 192 chlorination and methionine 148 oxidation were reported to impair apoA-I

5-OHTrp 72 apoA-I has impaired atheroprotective activities
efflux activity (44 -46). Whereas halogenated or nitrated tyrosines have been used as a molecular fingerprint for the types of oxidative pathways present in human atheroma, and methionine sulfoxide detection is commonly observed with numerous oxidative processes, site-directed mutagenesis studies by the Smith and Hazen groups have instead suggested tryptophan as the key residue(s) involved in oxidative inactivation of cholesterol efflux function of apoA-I in human atheroma (31,32), but these studies were limited by the unknown effects of incorporating a nonoxidizable amino acid at these sites on the structure/function of the protein. Beyond tracking with cardiovascular disease risk, immunoaffinity isolation studies with antibodies generated specifically to apoA-I forms harboring a 2-OH-Trp at residue 72 indicated that the recovered apoA-I were largely dysfunctional in promoting cholesterol efflux activity in vitro, HDL biogenesis in vivo, and possessed proinflammatory gain of function activity in vitro (31). Here, we provide unequivocal evidence for a causal link between oxidation of apoA-I at residue Trp 72 and impairment of cholesterol acceptor activity and capacity to assemble HDL particles in vivo.
GCE-mediated insertion of 5-OHTrp has been described for various reporter proteins, dihydrofolate reductase, and the bacteriophage T4 fibritin (foldon) domain in E. coli and mammalian cells, primarily to demonstrate the legitimacy of respective engineered Trp-RS, including Trp-RS R313 used in our study (34,(47)(48)(49). To our knowledge, the present study is the first to investigate the consequences of genetic 5-OHTrp insertion for known functions of a mammalian protein. MPO-catalyzed reactive oxidant species, such as hypochlorous acid (HOCl), are produced within the human atheroma, where it fosters posttranslational modifications of proteins, with the presumed effect of loss of apoA-I cholesterol efflux activity and other apoA-I functions (30,32,(50)(51)(52). Mono-and dihydroxytryptophans have been identified by MS in apoA-I exposed to HOCl or the MPO-H 2 O 2 -Cl Ϫ system (52,53) and in apoA-I recovered from lesions (32). Although a mass increase of 16 Da is consistent with monohydroxytryptophan, the current MS methods employed cannot permit one to distinguish the indole ring position of oxygen incorporation. ApoA-I has both structural flex- Heparin plasma was fractionated into HDL-containing lipoprotein fraction (d Յ 1.21 g/ml) and LPDFs (d Ͼ 1.21 g/ml) by ultracentrifugation on a sucrose D 2 O gradient as described under "Experimental procedures." Quantitation of apoA-I mass in each fraction was performed by Western blotting using a biotinylated antihuman apoA-I mAb as described under "Experimental procedures." B, total apoA-I levels in plasma. Each symbol shown represents the average of duplicate measurements from an individual mouse at the indicated times. The line denotes the median of measurements made in five animals. C, apoA-I distribution within the HDL fraction (d ϭ 1.063-1.21 g/ml) and LPDFs (d Ͼ 1.21 g/ml) of plasma at the indicated times was quantified by Western blotting as described under "Experimental procedures." Western blotting was performed on 3 and 12 l of HDL-containing and LPDF fractions, respectively. ApoA-I mass in each fraction (HDL fraction (left) and LPDF (right)) is shown at the indicated times post-injection. Each symbol represents mean of duplicate measurements from 5 animals/ group with fractions from each animal measured in duplicate at the indicated times. The median value for each group is indicated. Overall p value was determined by one-way ANOVA analysis. p values were determined by Student's t test.

5-OHTrp 72 apoA-I has impaired atheroprotective activities
ibility and an inherent tendency to self-associate. It is remarkable that a single oxTrp in 5-OHTrp 72 apoA-I would induce such a profound inhibitory effect on cholesterol efflux and HDL biogenesis capacity. Intriguingly, Trp 72 resides in a short helix (residues 70 -76) in the highly organized N-terminal domain that is rich in amphipathic lipid-binding ␣-helices. In contrast, the C terminus has only one ␣-helix (residues 227-232) and is, for the most part, a random coil domain that closely interacts with the N terminus to influence both stability and lipidation of apoA-I (54 -57).
The exact mechanism of ABCA1-mediated apoA-I lipidation (and concurrent HDL biogenesis) is not fully understood, and it is unknown whether apoA-I dimerizes before or after lipid binding. It is plausible that oxidation on a large aromatic amino acid in a highly organized region of the protein might perturb the stability of the protein, thus affecting the mechanisms of lipidation. Interestingly, naturally occurring single-amino acid amyloidogenic mutations in apoA-I, which are known to induce conformational changes and lead to altered lipid-binding profile, reside primarily in the N terminus (residues 1-100) and within residues 170 -178 (58 -60). It remains to be seen whether structural interrogation of 5-OHTrp 72 apoA-I, as well as ABCA1-binding studies, may reveal mechanistic aspects of loss of efflux function in this post-translationally modified form of apoA-I.
Epidemiological studies have shown an inverse association between circulating HDL-cholesterol (HDL-c) or apoA-I levels and CAD (61,62). Infusions of HDL or apoA-I into animals, and experiments with genetic models for increased apoA-I, have consistently demonstrated decreased atherosclerosis burden (63)(64)(65)(66)(67). The HDL hypothesis has shifted away from HDL-c levels to HDL function as the critical parameter correlating with development of CAD. This paradigm shift was primarily driven by two major clinical observations: randomized Mendelian studies failed to establish a link between SNPs that increase HDL-c levels and CAD (68), and HDL-c-elevating drugs did not demonstrate significant therapeutic benefit in human studies (69 -71).
In conclusion, the present studies show site-specific incorporation of monohydroxy-Trp 72 into apoA-I, a post-translational modification present in apoA-I recovered from human atheroma, impaired cholesterol efflux activity in vitro, and HDL biogenesis activities in vivo. Given the accumulation of oxidized apoA-I forms in human atheroma, 5-OHTrp apoA-I and other possible ox-apoA-I forms are likely candidates to contribute to cholesterol accumulation during atheroma development. The current study also validates the use of noncanonical amino acid incorporation as a powerful tool for identifying functionally critical residues in apoA-I or other potential proteins that are post-translationally modified in human disease.
The RAW264.7 cell line was from the American Type Culture Collection (Manassas, VA). BAECs were from Lonza (Walkersville, MD) (catalogue no. AC-6001T25). All other chemical reagents were purchased from Sigma-Aldrich and Fisher unless otherwise noted. SDS-PAGE was performed as described by Laemmli (74) using precast Criterion Bio-Rad 4 -20% TGX gradient gels (Bio-Rad). Fractionated proteins were transferred (100 V, 1 h at 4°C) onto polyvinylidene difluoride (Immobilon-FL, Millipore, Burlington, MA) or nitrocellulose membrane (LI-COR Biosciences, Lincoln, NE) and developed for detection by Chemiluminescence (GE Healthcare) or the Odyssey detection system (LI-COR Biosciences) as detailed below. Animal protocols were approved by the institutional review board of the Cleveland Clinic. Breeders of apoA-I Ϫ/Ϫ (AIKO) mice (B6.129P2-Apoa1tm1Unc, catalogue no. 002055) were obtained from Jackson Laboratories (Bar Harbor, ME).

Plasmids
The plasmid pRST11D.R313.40A encoding the engineered 5-OHTrp-specific S. cerevisiae tryptophanyl tRNA-synthetase clone 313 (Trp-RS R313 (34)) and its cognate tRNA version 40A (tRNA Trp CUA 40A (34)) was generated from the parental pRST.11B.R313.40A (34) (a kind gift from Randall Hughes and Andrew Ellington, University of Texas (Austin, TX)) by subcloning the cDNA for both synthetase and tRNA as an AfeI-HindIII fragment into respective restriction sites in pRST.11D (a kind gift from Randall Hughes) plasmid backbone comprising LacI, ColA ori, and Kan R genes. E. coli codon-optimized human apoA-I cDNA sequence encoding the mature apoA-I polypeptide, amino acids 25-267, or the variant with amber stop codon (TAG) replacing Trp 72 (amino acids preceded by an in-frame N-terminal His 10 tag; gene synthesized by Genescript (Piscataway, NJ)) was cloned into the EcoRV site of pUC57. The His 10 apoA-I DNA fragment was excised after digestion with NcoI and BamHI restriction enzymes, and the gene fragment was cloned into pET20b to generate pET20bW72TAGAI and pET20bWTAI plasmids.

5-OHTrp 72 apoA-I has impaired atheroprotective activities
Comparison of oxTrp incorporation using the sfGFP reporter GFP protein expressions were performed in E. coli BL21ai cells co-transformed with the evolved orthogonal S. cerevisiae pRST11D.R313.40A (kanamycin-resistant) machinery and either pBad-sfGFP (ampicillin-resistant) or pBad-sfGFP150TAG (ampicillin-resistant). The arabinose autoinduction media and method were used (78) except that leucine was omitted from the media (but was present in the 18-amino acid mix) and lactose (0.03%) was included to induce expression of the evolved Trp-RS in pRST11D.R313.40A. ncAA stocks (100 mM) were prepared just before use (5-OH-L-tryptophan (M r ϭ 220.08) in N,NЈ-dimethyl formamide (Sigma, D119-50); 2,3-diOHTrp (M r ϭ 236.23) in DMSO; 2-oxindolyl alanine (2-OHTrp) (M r ϭ 220.23); and kynurenine (M r ϭ 208.22) were solubilized with NaOH in sterile water) and then added to the media to a final concentration of 1 mM. Aliquots of autoinduction media supplemented with ampicillin and kanamycin (100 g/ml each) and ncAAs (1 mM) were inoculated with 1:100 dilutions of saturated noninducing overnight cultures. The negative control experiments excluded the ncAAs from the autoinduction media. Protein expressions were performed in 3 ml of media in 17 ϫ 100-mm (14-ml) culture tubes with 250-rpm shaking, in subdued light, at room temperature.
Fluorescence measurements of the cultures were collected 64 h after inoculation using a SpectraMax M2 (Molecular Devices, Sunnyvale, CA) spectrophotometer. Cultures were diluted (1:3) in PBS for OD 600 measurements or in water for fluorescence measurements in a 96-well format (black Whatman Uniplates 7701-2350 were used for fluorescence measurements). Diluted cultures were excited at 485 nm, and emission was monitored at 510 nm with 495-nm cutoff to monitor fluorescence.

Expression and purification of 5-OHTrp 72 apoA-I
Recombinant oxTrp 72 apoA-I protein was expressed in ClearColi BL21 (DE3) (Lucigen) cells harboring the evolved orthogonal machinery (pRST11D.R313.40A) and pET20bW72TAGAI. Host cells were initially assessed for Trp-RS oxTrp specificity (Fig. 1C: cells were grown (37°C, 270 rpm) in 14-ml glass tubes in 3 ml of Luria broth (LB) media supplemented with ampicillin and kanamycin (100 g/ml each) to OD 600 of 0.6. Oxidized tryptophan (5-OH-L-tryptophan, 2,3-diOHTrp, 2-OHTrp, and kynurenine; stock solutions prepared as described above) were added (to 1 mM each), and protein expression was induced at room temperature with aeration (270 rpm) in the presence of isopropyl ␤-D-1-thiogalactopyranoside (IPTG) (0.4 mM). Full-length apoA-I in crude extract prepared from overnight-induced cultures was detected by Chemiluminescent Western blotting (polyvinylidene difluoride membranes) using anti-total apoA-I mAb developed in our laboratory (10G1.5 (79)) or anti-His (Abcam, 18184) antibody with goat anti-mouse IgG horseradish peroxidase (Bio-Rad, 170-6516) as detection antibody. Large-scale expression of 5-OHTrp 72 apoA-I was performed in 2-liter flasks in subdued light. LB media supplemented with antibiotics (ampicillin and kanamycin, 100 g/ml each) and 1 mM 5-OH-L-tryptophan dissolved in N,NЈ-dimethyl formamide was inoculated (1%, v/v) with freshly transformed overnight starter cultures (LB with antibiotics and glucose at 1% (w/v)) and grown at 37°C with 250 rpm shaking, to an OD 600 of 0.6. Cultures were cooled down and induced in the presence of IPTG (0.4 mM) for 24 h at 18°C. Cultures were harvested by centrifugation at 4°C for 20 min at 5,000 ϫ g, and cells were snap-frozen on dry ice and stored at Ϫ80°C. Cell pellets were resuspended in lysis buffer (100 mM Tris/HCl, pH 8.0, 800 mM NaCl, protease inhibitor (Sigma, P-8849), 1 mM phenylmethanesulfonyl fluoride) and treated (60 min at 4°C) with lysozyme (Sigma, L6876) before disruption under nitrogen by a microfluidizer (Microfluidics) and clarification (30 min at 29,000 ϫ g, 4°C). Clarified extracts were routinely stored in 3 M guanidine chloride (GnHCl), pH 8.0, at Ϫ80°C. Expressed His-tagged proteins were purified in a onestep immobilized metal affinity chromatography procedure using His60 Ni Superflow Resin (Clontech, 635662) in 50 mM Tris/HCl, pH 8.0, 400 mM NaCl, 3 M GnHCl, and 20 mM imidazole (binding buffer, with binding at 4°C with constant rotation, 2 h) followed by extensive washing with binding buffer containing 40 mM imidazole (wash buffer) before eluting with binding buffer containing 250 mM imidazole (elution buffer). All buffers were sparged with argon, and proteins were eluted into 100 M diethylenetriaminepentaacetic acid (DTPA) to remove any trace levels of redox-active transition metals. All glassware was baked at 500°C prior to use. Recombinant proteins were stored in 3 M GnHCl at Ϫ80°C. Just before use, proteins were extensively dialyzed against PBS or saline supplemented with 100 M DTPA, except prior to the final round of dialysis where DTPA was omitted. The concentration of apoA-I was determined by measuring absorbance at 280 nm and applying its predicted molar extinction coefficient (31,720 M Ϫ1 cm Ϫ1 ) at 280 nm and calculated molecular weight of 29,870 (His 10 WT apoA-I) or 29,884 (His 10 5-OHTrp 72 apoA-I).

mAb 2D10.5 generation, specificity, and labeling
Hybridoma cell lines were generated by immunizing apoA-I Ϫ/Ϫ mice with purified delipidated human apoA-I isolated from HDL recovered from healthy donors. Among the positive clones, subclones were screened until a mAb with the desired binding specificity for equal recognition of all forms of apoA-I (see below) was identified. The subclone, mAb 2D10.5, was selected by screening for equal recognition of lipid-free and lipidated (in reconstituted HDL) apoA-I under native conditions, as well as following oxidation by exposure to multiple different systems, including MPO- Ϫ , and CuSO 4 , as described (29). Two equivalent apoA-Ireactive monoclonal antibodies were isolated, one being 10G1.5 (29) and the other 2D10.5, with the only noticeable difference being that 2D10.5 is slightly better at apoA-I immunoprecipitation. mAb 2D10.5 was scaled up for production in serum-free media in the Lerner Research Institute Hybridoma core and purified using protein A/G (Thermo Scientific Pierce) as described (29). mAb 2D10.5 was biotinylated as described for 10G1.5 (31).

PON1 protein expression and arylesterase enzyme assays
Recombinant C-terminally His-tagged PON1 variant G3C9, originally evolved from human, mouse, rabbit, and rat PON1 5-OHTrp 72 apoA-I has impaired atheroprotective activities (77), was expressed in BL21 (DE3) pLysS cells. LB media (in 2-liter flasks) supplemented with ampicillin (100 g/ml), chloramphenicol (50 g/ml), and CaCl 2 ⅐2H 2 O (1 mM) were inoculated with 1.9% (v/v) overnight saturated starter culture and allowed to grow at 37°C with 250 rpm shaking, until an OD 600 of 0.7 was reached. Cultures were cooled and induced with IPTG (0.4 mM) at 18°C for 20 h with shaking. Cultures were harvested by centrifugation at 4°C for 20 min at 5000 ϫ g, and cells were snap-frozen on dry ice and stored at Ϫ80°C. Cells were resuspended in base activity buffer (50 mM Tris/HCl, pH 8.0, 50 mM NaCl, 1 mM CaCl 2 ⅐2H 2 O, 10% (v/v) glycerol) supplemented with the following: protease inhibitor (Sigma, P-8849), 1 mM phenylmethanesulfonyl fluoride, 10 mM benzamadine (Sigma, B6506), and 0.1 mM DTT (Sigma, D0632). Cells were treated (60 min at 4°C) with lysozyme (Sigma, L6876) before disruption at 18,000 p.s.i. by a microfluidizer (Microfluidics) and clarification (30 min at 29,000 ϫ g, 4°C). PON1 was purified immediately by adding imidazole (to 10 mM) and His-60 resin and allowing binding to proceed over 2 h at 4°C with constant rotation at 4°C before extensive washing with activity buffer with increased NaCl concentration (350 mM) and 30 mM imidazole. Bound protein was eluted with 250 mM imidazole in base activity buffer (50 mM Tris/HCl, pH 8.0, 50 mM NaCl, 1 mM CaCl 2 ⅐2H 2 O, 10% (v/v) glycerol). The eluted fraction was dialyzed against activity buffer without glycerol, and protein purity and concentration was assessed by SDS-PAGE and absorbance at 280 nm (molar extinction coefficient ϭ 1.097), respectively. PON1 was concentrated prior to use. PON1 arylesterase activity was measured by spectrophotometry in a 96-well plate format (SpectraMax M2, Molecular Devices). The initial phenylacetate hydrolysis rates by PON1 were determined at 270 nm in reaction mixtures composed of 3.4 mM phenylacetate (Sigma-Aldrich), 9 mM Tris hydrochloride, pH 8, and 0.9 mM calcium chloride at 24°C. An extinction coefficient (at 270 nm) of 1310 M Ϫ1 cm Ϫ1 was used for calculating units of arylesterase activity, which are expressed as micromoles of phenyl acetate hydrolyzed per minute per milligram of PON1 (80).

ABCA1-dependent cholesterol efflux assays
The cholesterol efflux (ABCA1-dependent and total) activity of apoA-I was determined as described (30) Fig. 4) in 0.25 ml of DMEM ϩ antibiotics with or without 0.3 mM 8-Br-cAMP. Cholesterol efflux to apoA-I was calculated after 16 h as described previously (30).

Cell-surface VCAM1 protein expression
Cell-surface VCAM1 protein levels were determined in BAECs (Lonza (Walkersville, MD), catalogue no. AC-6001T25). BAECs were seeded in a 96-well plate at 20,000 cells/100 l of DMEM supplemented with 10% FBS and antibiotics per well and grown to near confluence (23 h) before being starved in DMEM with antibiotics for 24 h. Cells were incubated with 80 g of protein per ml of HDL or 80 g of protein per ml of apoA-I exposed to the MPO-H 2 O 2 -Cl Ϫ system (Cl.HDL) or, where indicated, individual protein for 16 h. After one wash with PBS, the cells were fixed in 4% paraformaldehyde for 20 min and blocked in 0.5% casein in PBS with 0.02% NaN 3 for 1 h at room temperature. Surface VCAM1 protein was determined using goat polyclonal anti-VCAM1 primary antibody ( (81) immediately prior to use. These reaction conditions included physiologically relevant amounts of MPO, chloride, and hydrogen peroxide concentrations.

In vivo HDL biogenesis
To examine the de novo HDL biogenesis in AIKO mice following apoA-I (in saline) injection (with the indicated form of endotoxin-free recombinant His 10 -apoA-1 subcutaneously at 250 mg/kg body weight), female mice (5-6 months old; 5 animals/group) were used. Blood was collected by saphenous vein puncture just before protein injection (baseline) and at 2 and 4 h after apoA-1 injection, and the appearance and distribution of total human apoA-I and 5-OHTrp 72 -apoA-1 levels in the blood were monitored as follows. The HDL-containing lipoprotein fraction (d Յ 1.21 g/ml) and LPDF fractions (d Ͼ 1.21 g/ml) were recovered from plasma by sequential buoyant density ultracentrifugation using sucrose (Sigma, 84097) and D 2 O to avoid the high ionic strength-associated alterations to the protein composition of lipoprotein particles observed with KBr use (82). Mouse plasma (20 l) was diluted to 500 l with 1ϫ PBS buffer (pH 7.0) and 1155 l of sucrose D 2 O buffer (d ϭ 1.3 g/ml; 0.1742 g of K 2 HPO 4 , 0.1361 g of KH 2 PO 4 , 0.8182 g of NaCl, 110.06 g of sucrose, and 100 ml of D 2 O). The fraction with density less than 1.21 g/ml was obtained after a 48-h ultracentrifugation at 40,000 rpm (172,301 ϫ g avg ) at 20°C (Beckman rotor 50.4Ti; 8 ϫ 49-mm ultraclear centrifuge tubes). The HDL-containing fraction (d Յ 1.21 g/ml) was recovered by slicing the upper 0.28 ml of the tube. To quantify human apoA-I in whole and fractionated plasma, the following sample volumes

Mass spectrometry analyses
All buffers were prepared fresh and flushed with argon. WT apoA-I and apoA-I containing oxidized W 72 samples were prepared in triplicate as follows: the sample (ϳ10 g) was evaporated under vacuum and resuspended with 25 l of 6 M urea and 100 mM Tris buffer and then flushed with argon and left 10 min at room temperature. 100 l of 100 mM Tris buffer and 100 l of trypsin (0.5 g diluted in 100 l of 100 mM Tris buffer) were then added to a final concentration of 0.6 M urea. Next, the sample was flushed with argon and incubated overnight at 37°C. The reaction was stopped with acetic acid glacial (5 l), and the sample was spun at 17,000 ϫ g for 15 min. 30 -40 l of the sample was transferred into an MS vial and flushed with argon. The vials were loaded into the autoinjector of a Proxeon Easy-nLC coupled to an LTQ Orbitrap Velos mass spectrometer (Thermo Scientific) with a nano-LC electrospray ionization source on positive mode. The MS analysis was performed with the Xcalibur data system. Samples were first loaded onto an Acclaim PepMap 100 sample trap (inner diameter 100 m, length 2 cm, nanoViper, packing material: C18, 5 m, 100 A˚) and separated by an Acclaim PepMap 100 column (inner diameter 75 m, length 15 cm, nanoViper, packing material: C18, 3 m, 100 A˚from Thermo). Mobile phase was as follows: buffer A, water with 0.1% formic acid; buffer B, acetonitrile with 0.1% formic acid. The HPLC gradient started with 5% B and increased to 35% B over 40 min then increased to 70% B over 10 min and was held at 70% B for another 5 min. For repeat injections, the column was re-equilibrated for the next sample by decreasing to 5% B over 1 min and held at this mobile phase composition for 12 min. The flow rate was 0.3 l/min. The applied spray voltage of the nano-LC electrospray ionization source was 2.0 kV. For MS2 data collection, one high-resolution full scan by Orbitrap was followed by 10 data-dependent scans of the most intense ions by ion trap. Dynamic exclusion was also applied so that any ions that had been repeatedly scanned two times in 10 s were excluded from scan for 30 s. Mass spectra of digested peptides were processed by the SEQUEST algorithm of Proteome Discoverer 1.1 (Thermo Scientific) and searched against a fasta file containing the sequence of His-tagged apoA-I. Static modifications of ϩ16 and ϩ32 were introduced for single and double oxidation of tryptophan and ϩ4 for kynurenine. The SEQUEST output files were filtered by the crosscorrelation coefficient (XCorr) with minimal thresholds of 2.5 for double-charged and 3.5 for triple-charged peptides. All spectra for targeted peptides were inspected manually. The percentage of tryptophan oxidation in the sample was calculated as follows. The chromatogram peaks corresponding to peptides containing native tryptophan or one of its oxidized forms were integrated with the Xcalibur program, and their areas were summed up. Then the percentage of each species was calculated as the peak area fraction (peak area of peptides containing oxidized tryptophan/total peak area of peptides containing oxidized and native tryptophan) multiplied by 100. A list of all native or oxidatively modified apoA-I peptides identified in the proteomics analyses given in Figs. 2 (B and C) and 5B are given in Tables S2 and S3. For all modified peptides identified, annotated, mass-labeled MS/MS spectra are provided in Figs. S1-S38. The information given in these tables about the identified peptides in protein tryptic digests includes the peptide identification score (X score), the peptide m/z, and the mass difference relative to anticipated from unmodified native sequence (⌬ppm). In addition, annotated MS/MS spectra and a full list of the b and y peptide fragmentation ions (identified by MS/MS and used for peptide identification) are provided for peptides identified in the following samples: A11(WT apoA-I) and B11(5-OHTrp 72 apoA-I) (Table S2) and WT batch 1, WT ϩ MPO batch 1, and 5OHTrp batch 1 (Table S3). Table S1 lists the peptide search parameters used by the Proteome Discoverer version 1.1 program.

Statistical analysis
Statistical significance of differences was determined by Student's t test or analysis of variance where more than two comparisons were made. p values for statistical significance are reported for p Յ 0.05.

Data availability
The mass spectrometry raw data were deposited to the Pro-teomeXchange Consortium (PRIDE repository) (83) and can be accessed with the data set identifier PXD017226.