A biochemical fluorometric method for assessing the oxidative properties of HDL.

Most current assays of HDL functional properties are cell-based. We have developed a fluorometric biochemical assay based on the oxidation of dihydrorhodamine 123 (DHR) by HDL. This cell-free assay assesses the intrinsic ability of HDL to be oxidized by measuring increasing fluorescence due to DHR oxidation over time. The assay distinguishes the oxidative potential of HDL taken from different persons, and the results are reproducible. Direct comparison of this measurement correlated well with results obtained using a validated cell-based assay (r2 = 0.62, P < 0.001). The assay can be scaled from a 96-well format to a 384-well format and, therefore, is suitable for high-throughput implementation. This new fluorometric method offers an inexpensive, accurate, and rapid means for determining the oxidative properties of HDL that is applicable to large-scale clinical studies.

Animal Medicine at the David Geffen School of Medicine at UCLA as previously described ( 15 ). The mice were maintained on a Western diet (Teklad, Harlan, catalog #TD88137) for two weeks; a group of mice was also treated with pravastatin at 12 g/ ml drinking water (or approximately 50 g per day) for two weeks. All experiments were performed using protocols approved by the Animal Research Committee at UCLA.

HDL and LDL purifi cation
HDL and LDL were isolated from cryopreserved human plasma (with or without added sucrose) by ultracentrifugation (UC), fast-performance liquid chromatography (FPLC), or dextran sulfate and precipitation with polyethylene glycol. These were aliquoted and stored as previously described ( 4,(17)(18)(19)(20). HDL cholesterol was quantifi ed using a standard colorimetric assay (Thermo DMA Co., San Jose, CA) as previously described ( 8 ). HDL cholesterol was oxidized by use of copper ions (OxHDL) as previously described ( 21 ).

HDL infl ammatory index
The HDL infl ammatory index was determined for each subject's HDL as described previously ( 11,12 ). In this assay, a value >1.0 is considered proinfl ammatory, and a value <1.0 is considered anti-infl ammatory.

DCF-based cell-free assay of HDL function
The DCF-based cell-free assay was performed as previously described ( 8,16 ).

DHR-based cell-free assay of HDL function
Quadruplicates of HDL (5 µg of cholesterol, unless otherwise specifi ed) were added to 96-well plates (polypropylene, fl at bottom, black, Fisher Scientifi c). HBS was added to each well to a final volume of 150 µl, followed by addition of 25 µl of the working solution of 50 µM DHR, for a total volume of 175 µl (fi nal DHR concentration of 7 µM). Immediately following DHR addition, the plate was protected from light and placed in a fl uorescence plate reader. The fl uorescence of each well was assessed at two minute intervals for an hour with either a DTX 800/880 Multimode Detector (Beckman Coulter, CA) or Synergy 2 Multi-Mode Microplate Reader (Biotek, VT), using a 485/538 nm excitation/emission fi lter pair with the photomultiplier sensitivity set at medium. Both readers gave equivalent readings (r 2 = 0.99). Using Microsoft Excel software, the oxidation rate was calculated for each well as the slope for the linear regression of fl uorescence intensity between 10 and 50 min, expressed as FU minute Ϫ 1 (fl uorescence units per minute). HDL oxidative function was calculated as the mean of quadruplicates for the wells containing the HDL sample.
The assay was also run in a 384-well format (polypropylene, fl at bottom, black, Fisher Scientifi c), using 1.5-3 µg of HDL produced from oxidation of lipoproteins) are signifi cantly associated with the monocyte chemotactic activity and functional properties of HDL that are measured by a cellbased assay ( 8 ). Thus this measurement of reactive oxygen species after HDL exposure as refl ected by the increased DCF fl uorescence refl ects the oxidative properties of different types of HDL that vary in their capacity to engage in redox cycling ( 1 ).
This approach promised to allow a direct biochemical assessment of functional properties of HDL without relying on a biologic readout. Although this assay yielded results that correlated to the cell-based assay, it did not achieve widespread usage due to i ) the oxidative instability of DCF-DA and the resulting tendency for minor variations in experimental conditions to cause inconsistency of the assay and ii ) the sensitivity of the assay to any degree of hemolysis.
Here we describe an approach to quantify the oxidative activity of HDL in a cell-free, biochemical assay. The products of redox cycling are detected as time-dependent oxidation of the fl uorogenic probe dihydrorhodamine 123 (DHR) to fl uorescent rhodamine 123 ( 13 ). The rate of DHR oxidation in the presence of HDL refl ects the oxidative/antioxidative activity of HDL. This assay provides a readout that is highly correlated to a validated cell-based assay ( 5,11,12 ), and it is highly reproducible and amenable to high-throughput implementation.

Subjects
Blood samples were collected from patients with coronary artery disease (CAD) or equivalent as defi ned by National Cholesterol Education Program Adult Treatment Panel III criteria ( 14 ) and were collected from patients referred to the cardiac catheterization laboratory at the Center for Health Sciences at the University of California, Los Angeles (UCLA). After signing a consent form approved by the Human Research Subject Protection Committee of UCLA, patients donated a fasting blood sample collected in a heparinized tube. Plasma samples were also randomly selected from pretreatment samples remaining from a previously described study in which all patients had coronary artery disease or equivalent ( 12 ). All of these patients were on a statin ( 12 ). Rheumatoid arthritis (RA) patients were recruited from the rheumatology offi ces at UCLA via fl yers posted in the offi ces and in the UCLA Medical Center. All RA patients met the American College of Rheumatology criteria for RA, which was verifi ed by chart review. Human immunodefi ciency virus (HIV)-infected subjects had HIV-1 RNA у 10,000 copies/ml ; were not receiving antiretroviral therapy; had no documented coronary atherosclerosis; had normal total cholesterol (200 mg/dl), LDL cholesterol (130 mg/dl), HDL cholesterol (males, >45 mg/dl; females, >50 mg/dl), and triglycerides (<150 mg/dl); were not receiving hypolipidemic medications; and were not diabetic. Normal volunteers were recruited under a protocol approved by the Human Research Subject Protection Committee of UCLA.

Mice
ApoE-and LDLR-null mice originally purchased from the Jackson Laboratories on a C57BL/6J background were obtained from the breeding colony of the Department of Laboratory and linear rate between 10 and 50 min ( Fig. 1 ). The rate of oxidation of DHR was signifi cantly less with added HDL, and the two HDL samples (an example of anti-infl ammatory HDL (aHDL) compared with proinfl ammatory HDL (pHDL)] showed clearly different effects in this regard ( Fig. 1A ). Furthermore, when the amount of added HDL was varied, there was a clear dose dependence in the oxidative effects on DHR that was linear in the range 2.5-15 µg (cholesterol) of added HDL per well in the assay ( Fig. 1C ).

Lipid-probe interactions explain the reduction in the oxidation rate of DHR after addition of HDL
We observed a signifi cant reduction in fl uorescence signal and oxidation rate of DHR after addition of HDL ( Fig. 1 ).
To determine whether the reduction in fl uorescence signal of DHR after addition of HDL is caused by antioxidant effect of HDL or nonspecifi c lipid-probe interactions, we tested the effect of addition of different lipids with known oxidative properties on the oxidation rate of DHR. We found that addition of LDL, HDL ( Fig. 1 , supplementary Fig. I), and other lipids that are known not to be antioxidant, such as oxidized L -␣ -1-palmitoyl-2-arachidonoyl-sn -glycero-3phosphorylcholine (oxPAPC) (data not shown), leads to signifi cant decrease in oxidation signal of DHR. In addition, coincubation of a stable amount of DHR with different concentrations of two lipids simultaneously (HDL with LDL, supplementary Fig. I; HDL with oxPAPC, data not shown) caused signifi cant decrease in the oxidation signal of DHR. Similar lipid-probe interactions were confi rmed with DCFH (data not shown). The lipid-probe interaction between DHR and HDL was confi rmed with an independent experiment (supplementary Fig. II).
To determine whether the lipid-probe interactions were dose dependent, increasing doses of DHR (1-50 M) were added to specifi c amount of HDL or LDL (2.5 g) (supplementary Fig. III). We found that decreasing concentrations of DHR resulted in decreasing the oxidation rate of both LDL and HDL and that the oxidation rate of LDL was higher than HDL (supplementary Fig. III). To increase the sensitivity of the assay and to detect smaller differences in oxidative properties of lipoproteins, we used a concentration of 50 M of DHR for all our experiments; this concentration has also been previously used to quantify redox activity ( 13 ).
To further investigate the dose-dependent interaction between DHR and lipoproteins, increasing concentrations of lipoproteins (HDL and LDL) were incubated with 50 M of DHR. Fig. 1C and supplementary Fig. I show that the fl uorescent signal generated by HDL was dose dependent but that increasing the concentrations of lipoproteins resulted in a statistically signifi cant decrease in fl uorescence for each concentration of lipoprotein used. The correlation observed between the quantity of lipoprotein added and DHR fl uorescence and oxidation rate was inverse and highly signifi cant (

LC/MS/MS analysis
LC/MS/MS was performed using a 4000 QTRAP quadruple mass spectrometer (Applied Biosystems) equipped with an electrospray ionization source. The HPLC system utilized an Agilent 1200 series LC pump equipped with a thermostatted autosampler (Agilent Technologies). Chromatography was performed using a Luna C-18(2) column (3 µm particle, 150 × 3.0 mm, Phenomenex) with a security guard cartridge (C-18; Phenomenex) at 40°C. Mobile phase A consisted of 0.1% formic acid in water, and mobile phase B consisted of 0.1% formic acid in acetonitrile. The autosampler was set at 4°C. The injection volume was 20 µl, and the fl ow rate was controlled at 0.4 ml/min. The gradient program was as follows: 0-8 min, linear gradient 0-95% B; 8-9 min, 95% B; 9-9.15, 95-0% B; 9-12 min, 0% B. Data acquisition and instrument control were accomplished using Analyst 1.4.2 software (Applied Biosystems). Detection was accomplished by using the multiple reaction monitoring (MRM) mode with positive ion detection. The parameter settings used were the following: ion spray voltage = 5500 V; curtain gas = 22 (nitrogen); ion source gas 1 = 34; ion source gas 2 = 25; ion source gas 2 temperature = 250°C. Collision energy, declustering potential, and collision cell exit potential were optimized to obtain optimum sensitivity. The transition monitored was m/z 347.05 to m/z 315.1 for DHR.

LC/MS/MS experiments
HDL from control and CAD-patient donors was combined with DHR, and the amount of DHR remaining in the samples after 2 h incubation was determined by LC/MS/MS. First, UC-and FPLC-isolated HDLs were combined in triplicate with DHR. Each sample contained 2.5 g HDL and 50 M DHR in a fi nal volume of 175 l HEPES-buffered saline. The samples were incubated in the dark for 2 h. Urate was then added (fi nal concentration 25 M) to slow further oxidation of DHR ( 22 ). The samples were transferred in minimal light to autosampler vials (Fisher Scientifi c) for LC/MS/MS analysis. The samples were ordered so that HDL-control samples were measured last, to ensure that any additional air oxidation of DHR did not result in a false positive as regards the difference between HDL-CAD and HDL-Ct with respect to DHR. In a separate experiment, UC-isolated HDL from a healthy donor was serially diluted, in triplicate, from 10 to 0.625 g cholesterol. The dilutions were combined with DHR and incubated in the dark for 2 h as above. After incubation, urate was added, the samples were transferred to autosampler vials, and the amount of DHR remaining was determined by LC/MS/MS.

Statistical analysis
Statistical analysis was performed using Excel (Microsoft Corp., Seattle, WA). Group means were compared using Student t -test for unpaired variates with P < 0.05 considered statistically significant. Correlation coeffi cients between variables were calculated using least-squares linear regression.

HDL samples reduce the oxidation rate of DHR consistently by differing amounts
To assess whether the oxidation rate of DHR is affected by the properties of HDL, HDL samples were tested for their effect on DHR oxidation rate. Alone, DHR exposed to air became oxidized (and therefore fl uorescent) at a Differences in fl uorescence after addition of the same amount of different HDL samples correspond to real differences in the degree of oxidation of DHR To ensure that the differences we observed with respect to fl uorescence corresponded to real differences in the degree of oxidation of DHR, we subjected samples of HDL-CAD and HDL-non-CAD that had been combined with DHR in the manner of the fl uorometric assay to LC/ MS/MS analysis to determine the amount of DHR remaining in each ( Fig. 2 ). Regardless of whether the HDLs were isolated by sequential UC ( Fig. 2A ) or by FPLC ( Fig. 2B ), HDL from healthy non-CAD donors contained significantly more DHR than HDL from CAD-patient donors after an initial 2 h incubation with 50 M DHR ( P = 0.001, n = 3 for each pair). As the amount of DHR remaining in the samples corresponds in this setting to the degree of oxidation of DHR, it follows that HDL from non-CAD  saline diluent (150 mM NaCl and HEPES 20 mM, pH 7.4) used for the results in Fig. 1 , added sodium azide reduced the oxidation rate of DHR (supplementary Fig. VI-A). Standard phosphate-buffered saline (PBS) also reduced the oxidation rate, which was further reduced by the addition of sodium azide (supplementary Fig. VI-A). Oxidation of DHR was also pH dependent (supplementary Fig. VI-B). Testing a range of pH values from 3 to 9, the oxidation rate was highest at pH 3, lowest at pH 9, and similar at pH 5 and 7. These results suggested that sensitivity of the assay is best using saline without sodium azide and that it is less prone to pH effects near the physiologic range.

Different methods of HDL isolation affect the readout of the assay
Different methods of HDL purifi cation may affect the functional properties of HDL, so this DHR-based assay was tested for effects of common methods of HDL isolation such as UC, FPLC, and use of dextran sulfate and PEG ( Fig. 4 ). The rate of oxidation of DHR with added HDL was signifi cantly higher in patients compared with controls regardless of whether the HDL was isolated by FPLC or UC (supplementary Fig. VII). In addition, the DHR assay gave consistent results using all different methods of HDL isolation (supplementary Fig. VIII). However, the oxidation rate of DHR was signifi cantly higher when HDL isolated by UC was added compared with when HDL isolated by other methods was added ( P < 0.001). The results obtained by all the methods correlated strongly (r 2 > 0.8, P < 0.0001; supplementary Fig. VIII).

Results from the DHR assay correlate with a validated cell-based assay
To compare the results of this DHR assay to those obtained using a validated cell-based assay ( 11,12 ), 30 HDL donors signifi cantly inhibited the oxidation of DHR compared with HDL from CAD patients. In separate experiments, these same pairs of HDL were shown to exhibit signifi cant differences in both the rate of increase of fl uorescent signal and in fl uorescent signal after 2 h incubation during the standard fl uorometric assay. Both UC-and FPLC-isolated HDL-CAD exhibited a signifi cantly greater rate of increase in fl uorescent signal and signifi cantly more fi nal fl uorescence than the respective HDL-non-CAD samples (data not shown). Finally, the dose-dependent interaction between DHR and HDL was confi rmed by LC/MS/MS (supplementary Fig. IV).

The concentration-dependent effect of addition of HDL on oxidation of DHR is not secondary to lipid-free apoAI
To determine whether lipid-free apoAI has a concentration-dependent effect on oxidation of DHR at concentrations that correspond to the physiologic range of concentration of apoAI in normal subjects (90-130 mg/dl) ( 1, 23 ), commercially available apoAI from human plasma was obtained, and the effect of different concentrations of apoAI on oxidation of DHR was determined. We found that although addition of apoAI caused reduction in the oxidation rate of DHR, the reduction was not concentration dependent and not specifi c to the addition of increasing concentrations of apoAI, consistent with nonspecifi c probe-apoAI interactions (supplementary Fig. V).

The DHR-based assay of HDL oxidation function yields highly reproducible measurements despite fl uorescence quenching
To counteract the effect of fl uorescence quenching between HDL and DHR, relative differences in the oxidation rate of DHR for a specifi c amount of HDL (2.5 g or 5 g) were studied between different samples. To assess whether the assay is reproducible so that results of different experiments are comparable, six HDL samples were assessed in four independent experiments ( Fig. 3 ). Again, all HDL samples reduced the oxidation rate of DHR with low intraassay variability between the quadruplicates, which ranged 5.8-7.6% (mean 6.6%). Between independent replicates of the experiment, inter-assay variability for each of the samples ranged 5.0-11.9% (mean 8.6%).
To determine the effect of the variance from HDL isolation method, six samples were isolated with dextran/polyethylene glycol (PEG) method four different times, as described in "Materials and Methods." Using dextran and PEG, the mean inter-assay variability for these six samples was 9.1% (range 6.0-12.1%) and 8.6% (range 5.4-11.8%), respectively. The mean intra-assay variability for these six samples was 8.3% (range 4.9-11.1%) and 7.1% (range 4.7-10.3%), respectively. These results indicated that the assay could reliably measure HDL oxidative function.

Different diluents affect the readout of the assay
Different methods of HDL purifi cation and storage utilize varying compositions of diluents, so this DHR-based assay was tested for effects of common types of diluents utilized for HDL suspension and storage in preparation for HDL functional assays. Compared with the buffered Fig. 3. Low inter-assay variability between measurements of HDL effects. Oxidation of DHR in the presence of six different samples of HDL was assessed as described in Fig. 1 , using 2.5 µg (cholesterol) of added HDL. The data (means of quadruplicates) from four independent experiments are plotted. The mean inter-assay variability for these six samples was 8.6% (range 5.0-11.9%), and the mean intra-assay variability was 6.6% (range 5.8-7.6%).

The DHR assay can be scaled to a high-throughput, 384-well format
Although inter-assay variability is low (as shown above for the 96-well format), the capability to compare numerous samples in the same experiment could be advantageous for clinical studies of HDL function. Scaling down samples were assessed using both assays ( Fig. 5 ). Comparing DHR oxidation rate with the HDL infl ammatory index, there was a strong positive correlation (r 2 = 0.62, P < 0.001).

Oxidized HDL is proinfl ammatory HDL
To determine whether there is an association between oxidative and infl ammatory properties of HDL, HDL was oxidized as described in "Materials and Methods," and infl ammatory properties were determined by a validated cell-based assay ( 11,12 ) ( Fig. 6 ). Anti-infl ammatory HDL from healthy volunteers became proinfl ammatory after oxidation ( Fig. 6 ). Thus, the DHR assay determines functional, oxidative properties of HDL that are associated with infl ammatory properties of HDL.

The DHR assay can detect dysfunctional HDL in different conditions
To further validate the new method, we used the DHR assay to detect dysfunctional HDL in conditions in which dysfunctional HDL is known to be present, such as established animal models of atherosclerosis ( 1, 23 ), RA ( 16 ), and HIV infection ( 24 ). The DHR assay detected the established effect of statins on functional properties of HDL ( 1,23 ) in animal models of atherosclerosis such as LDLR Ϫ / Ϫ ( Fig. 7A )

and ApoE
Ϫ / Ϫ mice ( Fig. 7B ). In addition, the functional properties of HDL in patients with RA as determined by a previously validated cell-free assay using DCF correlated signifi cantly with the functional properties of HDL as determined by the oxidation slope of DHR ( Fig. 8 ). Finally, the DHR assay detected dysfunctional HDL in HIV infection, even when the HDL from these patients was mixed with HDL from non-HIV patients at various ratios (supplementary Fig. IX). Fig. 4. Infl uence of method of HDL isolation on measurements of HDL oxidative activity. Infl uence of method of HDL isolation on DHR oxidation was assessed using 100 µl in a 384-well fl at-bottom plate, and the rate of change in fl uorescence was measured as in Fig. 1 . The HDL was isolated from 10 healthy volunteers (non-CAD). An amount of 15 µM DHR was exposed to 2.5 µg (cholesterol) of FPLC-purifi ed (FPLC), ultracentrifugation-purifi ed HDL (UC), or HDL that was isolated using dextran sulfate or polyethylene glycol as described in "Materials and Methods." Rates of oxidation of DHR are plotted as means of quadruplicates for each sample. diffi cult for large-scale studies. The function of HDL has been determined in previous assays based on the capacity of HDL to prevent the formation of or to inactivate oxidized phospholipids produced by LDL ( 8 ). Recent interest has focused on the functional consequences of HDL oxidation. Oxidation could conceivably contribute to the formation of dysfunctional HDL, proposed to be present in humans with cardiovascular disease ( 27 ). One potentially important pathway for generating dysfunctional HDL via oxidation involves myeloperoxidase that mediates conversion of protein tyrosine residues to 3-chlorotyrosine and methionine residues to methionine sulfoxide (MetO) ( 28 ). MetO can also be formed from exposure of HDL's major protein, apoA-I, to H 2 O 2 ( 29 ) or lipid hydroperoxide ( 30,31 ), the latter generated during the oxidation of HDL lipids. ROS, such as 1 e -oxidants (i.e., hydroxyl radical and metal ions) and 2 e -oxidants (i.e., HOCl, H 2 O 2 , and peroxynitrite), have previously been shown to oxidize tyrosine and methionine residues ( 32 ), which can have dramatic consequences on the functions of apoA-I/HDL, including reverse cholesterol transport ( 33 ).
Dihydrorhodamine 123 (DHR) may be oxidized by a variety of oxidants, including hypochlorous acid generated by myeloperoxidases, peroxynitrite anion formed by oxidation of nitric oxide and hydrogen peroxide in the presence of peroxidases ( 13,34 ). Although DHR has mostly been used to determine oxidation of various molecules in intracellular matrix ( 34 ) previous studies have demonstrated that oxidation rate of DHR can be used to quantify redox activity in extracellular matrix such as plasma ( 13 ). The oxidation of HDL is the result of oxidation of both the lipid and protein component. During air oxidation of the HEPES-saline-lipoprotein-DHR solution both 1 e -oxidants and 2 e -oxidants are generated ( 13 ). The 1 e -oxidants preferentially react with the lipoprotein's lipids, and this causes lipid peroxidation with the resulting accumulation of hydroperoxides of phospholipids and cholesteryl-esters ( 35 ) that then oxidize apoA-I's methionine to MetO ( 30,31,36 ). We demonstrated that oxidation of lipid-free apoAI did not cause a concentration-dependent effect on oxidation of DHR. The measurement of the slope of changes in fl uorescence of DHR over time in response to addition of different types of HDL corresponds to the rate of formation of lipid hydroperoxide products and ROS. Thus, the oxidation rate of DHR can be used to quantify the intrinsic ability of HDL to be oxidized, which has been shown to affect the function of apoA-I/HDL ( 33 ). Thus, in the current study, we demonstrated a novel cell-free assay that biochemically examines the oxidative ability of HDL, providing a functional measurement that correlates well with a validated cell-based assay. We utilized DHR as a substrate similar to a previously reported assay to measure the redox potential of iron in plasma ( 13 ). Using the new method, we demonstrated that addition of normal HDL led to a signifi cantly smaller increase in the fl uorescent signal of DHR compared with pHDL.
In the previous cell-free assay, we demonstrated lipidprobe interactions between the fl uorochrome DCFH and various lipids ( 8 ). To our knowledge, the interactions of Fig. 6. Oxidized HDL has infl ammatory properties. The HDL infl ammatory index was determined for three samples of FPLCpurifi ed HDL from healthy volunteers in a cell-based assay as described in "Materials and Methods." The same HDL samples were oxidized as described in "Materials and Methods" (oxHDL), and the HDL infl ammatory index was determined. The oxidized HDL had signifi cantly higher HDL infl ammatory index (** P = 0.002, paired t -test). the assay to the 384-well format yielded results that were comparable to those for the 96-well format when HDL samples were tested in both formats (supplementary Fig. X, r 2 = 0.96, P < 0.0001). The intra-assay variability averaged 7.8% and the inter-assay variability was about 10.2% in the 384-well format (data not shown), and thus it was slightly higher than for the 96-well format. The increased variability was not due to the longer time to load the plate manually, because the oxidation rate was similar when the plate was allowed to incubate for 2 h at room temperature (data not shown). Overall, these data suggested that a high-throughput, 384-well format is feasible for assessing large numbers of HDL samples simultaneously.

DISCUSSION
Growing evidence suggests that HDL varies signifi cantly in its phenotype and infl uence on cardiovascular disease risk ( 6,25 ). Atherosclerosis begins when LDL is trapped in the arterial wall and seeded with ROS ( 26 ), resulting in oxidized LDL that attracts monocytes and induces localized infl ammation ( 3,4 ). Properly functioning HDL removes ROS from LDL, antagonizing this process. HDL, however, can be dysfunctional under certain clinical conditions, such as acute phase responses ( 2 ), and various pathogenic states, such as diabetes. Measuring the functional status of HDL, rather than HDL cholesterol concentration, may be more informative in predicting cardiovascular disease. Unfortunately, the best assays for HDL function have been cell-based assays (9)(10)(11)(12), which are technically prohibitive for many researchers and logistically lipids with DHR have not been previously tested. We found that addition of lipids, including pro-oxidant lipids contained in lipoproteins, led to a signifi cant decrease in the oxidation signal of DHR; thus the decrease in the oxidation rate of DHR after addition of HDL was caused by lipid-probe interactions. In addition, changes in fl uorescence that were observed with different buffers are likely due to an effect of the buffer on the HDL complex; hence, the interaction between the buffer and the fl uorescent indicator were affected by both the temperature of the reaction and the concentration of buffer (data not shown). Moreover, addition of detergent (SDS) in the DHR + HDL reaction caused an increase in fl uorescence (data not shown), which could be explained by possible dissociation of the HDL/dye complex and reduction in quenching (personal communication with Molecular Probes). We demonstrated with two different independent methods, including LC/MS/MS ( Fig. 2 , supplementary Fig. II), that the interaction of DHR-HDL is responsible for the fl uorescence quenching in the DHR assay.
However, we show that our DHR-based assay of HDL oxidation yields highly reproducible measurements despite fl uorescence quenching. In addition, the inverse correlation between HDL concentration and oxidation rate of DHR allows quantifi cation of functional properties of HDL using very low concentrations of HDL (1.25 g per sample). Thus, using this assay we were able to quantify relative differences in the oxidative properties of different HDL samples that correlate with their functional properties.
This DHR-based cell-free assay improves upon the prior dichlorofl uorescein-diacetate (DCFH-DA)-based cell-free assay ( 8 ). Although measuring the ability of HDL to inhibit oxidation, the earlier assay had several technical challenges that limited its utility. It was designed to mimic the cell-based assay and test the effect of HDL in combination with LDL. This introduced additional variability because different preparations of LDL can vary in their oxidative activity and lipid-probe interactions become less consistent with multiple lipids. Mixing LDL and HDL with DHR in our assay, we observed inconsistent results that appeared to be caused by fl uorescence quenching (data not shown) compared with the highly consistent results seen with HDL alone. Another technical challenge was the instability of DCFH-DA, which needed to be dissolved in methanol for activation to the active molecule DCFH and protected from room air because DCFH is unstable and prone to auto-oxidation ( 22,34,(37)(38)(39). In contrast, DHR is relatively stable ( 22,34,(37)(38)(39) and oxidizes at a predictable rate when exposed to room air. Furthermore, conversion of DCFH-DA to DCFH can be mediated by esterases Ϫ / Ϫ mice on Western diet for two weeks that were also treated with pravastatin 12.5 g/ml for two weeks. Each plasma sample was pooled from 4 mice (12 mice in total). Oxidation of DHR was assessed as in Fig. 1  for two weeks that were also treated with pravastatin 12.5 g/ml for two weeks. Each plasma sample was pooled from 4 mice (12 mice in total). Oxidation of DHR was assessed as in Fig. 1  more precise. The inter-assay variability of ‫ف‬ 8-10% compares favorably with the cell-based assay, which has a variability of >15% ( 12 ). The correlation of HDL effects on DHR oxidation rate to the biological readout of HDL in a cell-based assay ( Fig. 5 ) is consistent with a proposed mechanism whereby HDL exerts its effects through modulating ROS ( 8 ), and it suggests that the assay accurately refl ects the balance of aHDL and pHDL. Further support for the biological relevance of this measurement is the fi nding that oxidation of DHR was higher with LDL than with any of the HDL samples tested (supplementary Fig. I) and that it was significantly reduced after addition of HDL from statin-treated mice with atherosclerosis compared with addition of HDL from nonstatin-treated mice ( Fig. 7 ). Indeed, treatment of these mice with statins has been shown to reduce the infl ammatory properties of HDL ( 15 ), and we demonstrated that the DHR assay could detect the favorable effect of statins on the functional properties of HDL. Finally, the DHR assay can detect dysfunctional HDL in patients with RA ( Fig. 8 ) and HIV infection (supplementary Fig. IX), confi rming our previous results ( 15,16 ).
The fact that our assay quantifi es oxidation of HDL by a range of different oxidants increases its applicability to biological samples, at least in the context of cardiovascular diseases. This is because various oxidants contribute to oxidative modifi cations taking place in the affected arterial wall during atherogenesis ( 42 ). Thus, while "oxidized HDL" is not a chemically defi ned term, the oxidation rate of DHR corresponds to the intrinsic oxidative features of modifi ed lipoproteins. However, we exclusively used in vitro-generated oxidized forms of HDL, whereas the oxidative modifi cations occurring to HDL in the diseased artery wall are conceivably more complex ( 42 ). An additional potential limitation of the present assay is that HDL is subject to continuous remodeling in vivo. This includes dissociation of apoA-I from the lipoprotein particle, a process that could be increased by oxidation ( 43 ). Clearly, future studies are required to assess these aspects, as well as the utility of the assay for clinical studies.
In conclusion, this new assay offers a rapid method for measuring the properties of HDL that inhibit oxidation. It yields results that correlate well with a validated cellbased assay. This new technical approach may offer a convenient tool for studies of the role of the HDL functional phenotype in the development of atherosclerosis in vivo.
that are carried over variably during the lipid purifi cation process, adding another variable that is diffi cult to control, whereas DHR requires no activation and is not prone to this effect ( 22,34,(37)(38)(39). Another point of improvement is the kinetic approach for measuring oxidation rate during a linear phase, which lends greater precision compared with a single endpoint measurement. Finally, we demonstrated that different diluents for the assay affect the oxidation rate of the indicator compound; this may also help explain the variability observed in HDL measurements using the DCFH-DA assay among research groups.
We also demonstrated that although different methods of HDL isolation affect the oxidation rate of DHR when HDL is added, UC, FPLC, and isolation of HDL using precipitation with dextran sulfate or PEG could be used to determine functional properties of HDL using our DHRbased cell-free assay. The use of dextran sulfate or PDG may allow HDL isolation and use of this method in largescale studies. The oxidation rate of HDL was higher in samples isolated using UC. Removal of much of the albumin bound to the HDL particle during the process of UC may alter the association of nonpolar and polar substances, including ROS, associated with lipoproteins ( 40,41 ). The process of UC is longer than other methods and may yield additional ROS, which may account for the higher oxidation rate of DHR using UC compared with other methods of HDL isolation when the same amount of HDL is added ( 40 ). Finally, we found that addition of sucrose in plasma during cryopreservation did not change the HDL oxidation rate signifi cantly and that sucrose-free samples can be used in our assay (data not shown).
This novel DHR-based assay offers an attractive alternative to current cell-based assays. Because it measures a biochemical rather than biologic phenomenon, it may be Fig. 8. Correlation of DHR method with previous cell-free method. Twenty samples (10 from healthy volunteers and 10 from patients with rheumatoid arthritis) of HDL isolated using precipitation with dextran sulfate were assessed for their ability to inhibit DHR oxidation as shown in Fig. 1 , and their HDL infl ammatory index was determined in a cell-free assay using DCFH as described in "Materials and Methods." The values from each assay are plotted against each other.