Apolipoprotein A-I Modulates Regulatory T Cells in Autoimmune LDLr−/−, ApoA-I−/− Mice*

The immune system is complex, with multiple layers of regulation that serve to prevent the production of self-antigens. One layer of regulation involves regulatory T cells (Tregs) that play an essential role in maintaining peripheral self-tolerance. Patients with autoimmune diseases such as systemic lupus erythematosus and rheumatoid arthritis have decreased levels of HDL, suggesting that apoA-I concentrations may be important in preventing autoimmunity and the loss of self-tolerance. In published studies, hypercholesterolemic mice lacking HDL apoA-I or LDLr−/−, apoA-I−/− (DKO), exhibit characteristics of autoimmunity in response to an atherogenic diet. This phenotype is characterized by enlarged cholesterol-enriched lymph nodes (LNs), as well as increased T cell activation, proliferation, and the production of autoantibodies in plasma. In this study, we investigated whether treatment of mice with lipid-free apoA-I could attenuate the autoimmune phenotype. To do this, DKO mice were first fed an atherogenic diet containing 0.1% cholesterol, 10% fat for 6 weeks, after which treatment with apoA-I was begun. Subcutaneous injections of 500 μg of lipid-free apoA-I was administered every 48 h during the treatment phase. These and control mice were maintained for an additional 6 weeks on the diet. At the end of the 12-week study, DKO mice showed decreased numbers of LN immune cells, whereas Tregs were proportionately increased. Accompanying this increase in Tregs was a decrease in the percentage of effector/effector memory T cells. Furthermore, lipid accumulation in LN and skin was reduced. These results suggest that treatment with apoA-I reduces inflammation in DKO mice by augmenting the effectiveness of the LN Treg response.

Increased concentrations of high density lipoprotein (HDL) apolipoprotein A-I (apoA-I) in plasma are a well established negative risk factor for coronary heart disease (1). ApoA-I is the major protein contained in HDL, and it is essential for its formation and function (2), as well as associated anti-inflammatory properties (3)(4)(5). Decreased concentrations of HDL apoA-I are associated not only with increased atherosclerosis but also with autoimmune disorders such as systemic lupus erythematosus and rheumatoid arthritis (6,7). Although the effects of HDL on atherosclerosis have been widely reported, its effects on autoimmunity are not well understood. Recent studies from our laboratory indicate that in hypercholesterolemic mice lacking plasma apoA-I, an increased susceptibility to autoimmunity exists. The autoimmune phenotype is characterized by enlarged lymph nodes (LN), 7 cellular activation, proliferation, and lipid accumulation (8 -10), Interestingly, the autoimmune phenotype appears to be triggered by a cholesterol-containing diet, likely because of the absence of HDL and the lack of apoA-I-facilitated cholesterol efflux (8,10) via ABCA1. This raises the possibility that HDL apoA-I may be a common link between autoimmunity, inflammation, and atherosclerosis. Therefore, to test whether treatment with lipid-free apoA-I could attenuate ongoing inflammation, DKO mice were first fed an atherogenic diet containing 0.1% cholesterol, 10% fat for 6 weeks, after which treatment with apoA-I was begun. Subcutaneous injections of 500 g of lipid-free apoA-I was administered every 48 h during the treatment phase. These and control mice were maintained for an additional 6 weeks on diet. At the end of the 12-week study, DKO mice showed decreased numbers of LN immune cells, whereas Tregs were proportionately increased. Accompanying this increase in Treg cell number was a decrease in the percentage of effector/effector memory T cells as well as a reduction in the amount of lipid accumulation in LN and skin. Overall, these results suggest that treatment with apoA-I reduced inflammation and the accompanying autoimmune phenotype by restoring the effectiveness of the LN Treg response.

EXPERIMENTAL PROCEDURES
Mice, Diet, and Study Design-LDLr Ϫ/Ϫ , apoA-I Ϫ/Ϫ (DKO), and LDLr Ϫ/Ϫ (SKO) mice were fully backcrossed into the C57BL/6 background (8 -10). Mice were housed at the Wake Forest University Medical Center, and all procedures were approved by the Wake Forest University Medical Center Animal Care and Use Committee. Male mice from both genotypes were fed an atherogenic diet consisting of 0.1% cholesterol with 10% palm oil beginning at 6 weeks of age for a period of 12 weeks (8 -10). As controls, mice of each genotype were fed a Purina chow diet for the same length of time. All mice were maintained in a temperature-controlled room with a 12-h light/ 12-h dark cycle. Mice were fasted for 3 h prior to being anesthetized for blood collection by cardiac puncture and subsequent blood processing (8 -10). All mice used were males unless otherwise noted.
In studies designed to test the effects of apoA-I treatment on established diet-induced inflammation, age-matched male DKO and SKO mice were fed an atherogenic diet for 6 weeks before beginning 6 weeks of treatment with lipid-free human apoA-I while still consuming the atherogenic diet. During the treatment period, mice were subcutaneously injected with 500 g of lipid-free human apoA-I every 48 h. As controls, DKO and SKO mice were fed as described above but were subcutaneously injected with 500 g of bovine serum albumin every 48 h. Some mice were sacrificed after 6 weeks of diet consumption before beginning the treatment phase of the study for comparison.
Purification of Human Plasma ApoA-I and ELISA-Wildtype human apoA-I was purified by adjusting the equivalent of 1 unit of human plasma with KBr to d ϭ 1.225 g/ml and isolating the total lipoprotein fraction via ultracentrifugation for 18 h at 50,000 rpm, 15°C, in a Ti 50.1 rotor. To prevent methionine oxidation, hydrosulfite was maintained at a 10 M concentration at all centrifugation steps. The d Ͻ1.225 fraction was then sequentially ultracentrifuged (Ti 50.1 rotor, 18 h, 15°C, 50,000 rpm) to remove VLDL and finally LDL to obtain the d Ͼ1.080 g/ml fraction. The d Ͼ1.080 g/ml fraction was centrifuged again at d ϭ 1.225 g/ml to remove any traces of albumin. The d Ͻ1.225 g/ml fraction was then dialyzed again against 10 mM ammonium bicarbonate, pH 7.4, and then lyophilized. The dried pellet consisting of both apoA-I and apoA-II was solubilized with ϳ50 ml of 6 M guanidine hydrochloride and then exhaustively dialyzed against 10 mM ammonium bicarbonate, pH 7.4. The denatured HDL was ultracentrifuged at d ϭ 1.225 g/ml for 18 h, 15°C, in a Ti 50.1 rotor (11). The d Ͼ1.225 g/ml fraction was taken and exhaustively dialyzed against 10 mM ammonium bicarbonate, pH 7.4, and then lyophilized to dryness. The pellet was taken up in ϳ40 ml of 6 M guanidine hydrochloride dialyzed against 10 mM ammonium bicarbonate, pH 7.4, and the apoprotein was allowed to "refold." Following dialysis, protein concentration was determined, and the purity was verified by mass spectrometry (12). Both the concentrated BSA and apoA-I (4 -5 mg/ml) were filtered through a 0.45-m SpinX column (Costar) prior to subcutaneous injection. The human apoA-I concentration in plasma samples from mice bled at timed intervals following injection of apoA-I or BSA was determined by ELISA, as described previously (13,14).
Lipoprotein Separation and Characterization-The lipoprotein cholesterol distribution was determined following fast pro-tein liquid chromatography (FPLC) as described liter EDTA, pH 7.4, and 1.5 mmol/liter NaN 3 . Approximately 75-500-l fractions were collected, and 100 l from each fraction was used for cholesterol determination (13). To assess the human apoA-I distribution, 20 l from each fraction was separated on a 12% SDS-PAGE Western blot performed for human apoA-I, as described previously (13).
To assess the rate of apoA-I lipidation following subcutaneous injection, timed plasma samples were taken, and Western blot was performed. Plasma samples were separated by 4 -30% nondenaturing gradient gel electrophoresis (15)(16)(17) by running at 2800 V/h. Following electrophoresis, the gel was blotted onto a PVDF membrane (Whatman) at 23 V, overnight at 4°C. Following transfer, the membrane was fixed in 10% acetic acid for 15 min and then air-dried for 15 min. The membrane was then incubated in 50 mM Tris-HCl, pH 8.0, 80 mM NaCl 2 , 2 mM CaCl 2 , containing 5% nonfat dry milk and 0.2% Nonidet P-40 for 30 min. After the blocking step, the membrane was incubated with a 1:1000 dilution of goat anti-human apoA-I (Chemicon) for 1 h and then washed and incubated with a 1:3000 dilution of HRP-conjugated donkey anti-goat IgG for 1 h. The membrane was finally washed with Tris-saline, MgCl 2 , pH 9.6, before incubating with the Pierce SuperSignal West Pico chemiluminescent substrate. The chemiluminescence was visualized by an LSA-3000 imaging system (Fujifilm Life Science). Particle size was determined by comparison to protein standards of known Stokes radius (16,17).
Mass Spectrometry of Cellular Lipids-For the determination of free and esterified cholesterol, tissue and cells were extracted using chloroform/methanol (2:1) as described previously (8 -10). Cholesterol, together with other neutral lipids, was isolated from the lipid extract using a modification of the method of Kaluzny et al. (18) and analyzed by mass spectrometry as described previously (10).
Cholesterol Efflux Studies-Lymph nodes were collected and placed in serum-free RPMI 1640 medium, and cells were isolated as described previously (10,19,20). Once a single cell suspension was obtained, 3 ϫ 10 6 cells were added to each well of a 6-well plate, and the volume was adjusted to 1 ml with OpTmizer T cell expansion SFM (Invitrogen). 125 I-Labeled human apoA-I was added as an acceptor for the efflux of lymphocyte lipids to a final concentration of 150 g/ml and incubated overnight at 37°C in a CO 2 incubator. Approximately 24 h later, the medium was collected, dialyzed, and then concentrated before injection onto FPLC, as described previously for the isolation of ABCA1-derived nascent HDL (21). Following separation, each fraction was subjected to scintillation counting for 3 H content and run on 4 -30% nondenaturing gradient gels for particle sizing (15)(16)(17).
[ 3 H]Thymidine Proliferation Assay-Lymph nodes and spleens were collected and placed in serum-free RPMI 1640 medium, and cells were isolated as described previously (10,19,20). Once a single cell suspension was obtained, 1 ϫ 10 6 cells were added to each well of a 24-well plate, and the volume was adjusted to 0.5 ml with OpTmizer T cell expansion SFM (Invitrogen). Half the wells were stimulated with a mixture of anti-CD3 and anti-CD28 (BD Biosciences) containing 1 g/l of each at the final concentration. The cells were incubated at 37°C in a CO 2 incubator for 24 h after which time 1 Ci of [ 3 H]thymidine was added to each well, and the cells were incubated for an additional 24 h. To determine the amount of incorporated [ 3 H]thymidine, cold TCA was added to each well to adjust to a final concentration of 3%, and the mixture was filtered through glass fiber filters. The filters were thoroughly washed and then dried and subjected to scintillation counting.
Histology-Freshly isolated tissue was collected and embedded into OCT from which 10 -15-m-thick sections were prepared using a Leica cryostat at Ϫ20°C. Sections were stained with Oil Red O and counterstained with Mayer's hematoxylin as described previously (8,10) or stained with hematoxylin and eosin (9).
Immunofluorescence Microscopy-Fresh tissue was embedded into OCT and then 10 -15-m sections were obtained. Dried acetone-fixed sections were first incubated in a solution consisting of 1% normal goat serum and 1% BSA in PBS for 1 h at room temperature. Then sections were incubated with purified rat anti-mouse CD68 (AbD Serotec) or CD11c hamster anti-mouse (Pharmingen) at a 1:100 dilution or goat antimouse IgG (Jackson ImmunoResearch) at a 1:200 dilution for 30 min at room temperature and then briefly washed in PBS. Next, sections were incubated in their respective secondary antibody goat anti-rat IgG CD68 Cy3 (Rockland) at 1:300 or biotin-conjugated CD3e, CD11c mouse anti-hamster IgG (Pharmingen) for 45 min at room temperature. After washing in PBS, slides used for the detection of CD11c were incubated in Vectastain ABC-AP reagent from the Vectastain ABC-AP standard kit (Vector Laboratories) for 30 min. Vector Red (Vector Laboratories) was used as a substrate according to the manufacturer's protocol.
Finally, sections were washed in PBS and then counterstained with DAPI for 5 min before being mounted with Fluorogel mounting media (Electron Microscopy Sciences). Immunofluorescence was visualized using a Nikon Eclipse TE2000-S microscope. For all staining, at least four sections per animal, with a minimum of five mice per group, were analyzed.
Tregs were stained using the mouse regulatory T cell staining kit (eBioscience) according to the manufacturer's instructions. Briefly, 1 ϫ 10 6 cells were stained with rat anti-mouse CD4 and CD25 antibodies for 30 min at 4°C to measure these surface molecules. Cells were then washed and resuspended in a permeabilization solution for 30 min at 4°C. Cells were again washed and incubated in 1 g of Fc block for 15 min. Next, 0.5 g of rat anti-mouse FoxP3, an intracellular stain, was added to each sample and incubated for 30 min at 4°C. A mouse IgG isotype control (eBioscience) was used in each assay to subtract background fluorescence. Finally, samples were washed and resuspended in 2% paraformaldehyde in PBS before acquisition of samples using a FACSCalibur (BD Biosciences).
Analysis of Immune Cell Population Data-Data were analyzed using FlowJo software (TreeStar, San Francisco, CA). Cell populations measured included effector memory phenotype T cells (CD4 ϩ CD62L low , CD8 ϩ CD62L low , CD4 ϩ CD44 high , and CD8 ϩ CD44 high ), activated T cells (CD4 ϩ CD69 high and CD8 ϩ -CD69HIGH), Treg (CD4 ϩ CD25 ϩ FoxP3 ϩ ), DCs (CD3 Ϫ CD11c ϩ and CD3 Ϫ CD11b ϩ CD11c ϩ ), and macrophages (CD11c Ϫ -CD11b ϩ F4/80 ϩ ). The total number of cells staining positive for a particular set of stains was calculated based on the total number of LN cells in four subsets of skin draining LNs and the percent of those cells staining positive for each condition, as described previously (10). Populations were also expressed as the percent of the total number of LN cells such as DCs and macrophages or as the percent of total CD4 ϩ or CD8 ϩ cells as in the case of effector/ effector memory T cells, activated T cells, and Tregs.
Measurement of Plasma Antibodies-Plasma anti-dsDNA and oxLDL antibodies were measured by ELISA, as described previously (22)(23)(24)(25). Briefly, plates were coated with specific antigen and blocked in 1% BSA/PBS for 2 h at room temperature. Mouse serum was added at a dilution between 1:500 and 1:5000 and incubated overnight at 4°C. Plates were washed with 0.5% Tween 20/PBS (PBS-T) and incubated with biotinconjugated goat anti-mouse Ig(HϩL) (Southern Biotech, Birmingham, AL) for 45 min at room temperature and then incubated with avidin-peroxidase for 30 min at room temperature. Plates were then washed with PBS-T and developed using tetramethylbenzidine substrate (BD Biosciences). Anti-oxidized immunoglobulin isotype ELISAs were performed as described above using a biotin-conjugated goat anti-mouse IgG1, IgG2c, or IgM (Southern Biotech) secondary antibody.

Study Design and Treatment with Lipid-free Human ApoA-I-
Studies were carried out to determine whether treatment with lipid-free human apoA-I could attenuate the autoimmune phenotype in diet-fed DKO (LDLr Ϫ/Ϫ , apoA-I Ϫ/Ϫ ) mice. The design of the study is shown in Fig. 1, panel A. All male DKO mice were first fed an atherogenic diet containing 0.1% cholesterol, 10% fat for 6 weeks, after which one-half of the mice received subcutaneous injections of 500 g of apoA-I every 48 h and are designated DKO ϩ A-I. The second half of the DKO mice received 500 g of BSA every 48 h, designated DKO ϩ BSA. All mice were maintained for an additional 6 weeks on the diet during the treatment period for a total of 12 weeks of diet. Additional control mice, SKO (LDLr Ϫ/Ϫ ) were also treated in a similar fashion with either apoA-I or BSA treatment. Fig. 1, panel B, shows that 7-8 h post-lipid-free apoA-I injection, plasma levels reached a peak concentration of 4 mg/dl, suggesting that injection every 48 h would maintain low levels of plasma apoA-I. Separation of terminal DKO plasma into lipoprotein fractions, Fig. 1, panel C, using FPLC suggests that treatment with either apoA-I or BSA had little effect on the overall HDL cholesterol levels because both DKO ϩ AI (closed circle) and DKO ϩ BSA (open diamond) showed no discernable HDL-sized cholesterol peak, which was easily observed after separation of SKO mouse plasma (Fig. 1, panel C, shaded squares). However, Western blot analysis on individual FPLC fractions corresponding to HDL-sized particles indicated the presence of human apoA-I in fractions 66 -84 from DKO ϩ A-I mice but not from DKO ϩ BSA mice.
Further studies were conducted to determine the rate of apoA-I lipidation following injection. Western blot analysis was performed on post-injection plasma samples following 4 -30% nondenaturing gradient gel electrophoresis, as shown in Fig. 1, panel D. Results confirmed the presence of small HDL-sized particles in DKO ϩ A-I mice but not DKO ϩ BSA. Particle enlargement appeared to begin as early as 4 h post-injection, with particle size increasing up to 24 h post-injection.
ApoA-I Treatment and Cellular LN Lipid Content-In response to the atherogenic diet, both DKO and SKO mice display an expansion in the number of LN cells. Although DKO mice typically show a 6-fold expansion when compared with chow-fed DKO mice, SKO mice show only a 2-fold increase over chow-fed counterparts (10). Fig. 2, panel A, shows that the diet-induced LN cellular expansion appears to be progressive, because at 6 weeks after initiating the diet both genotypes of mice show a significant increase over their chow-fed counterparts. However, when the cellular content of DKO ϩ A-I LNs was examined, they showed a significant reduction compared with 12-week DKO mice but similar to the 6-week DKO mice. These results suggest that apoA-I treatment prevented any further increase in LN cell expansion but was not sufficient to bring LN cellularity back to levels seen in chow-fed mice during the course of this study.
In addition to increases in LN cells, previous studies have shown that cellular expansion is also accompanied by an increase in LN cholesterol content, specifically for DCs, B, and T cells (10), which make up the majority of LN cells. Similar to previous studies, DKO mice had ϳ2.5-fold more cellular cholesterol than SKO mice. Consistent with our hypothesis, the LN cells isolated from DKO ϩ A-I mice showed a significant decrease in their total cholesterol, as shown in Fig. 2, panel B, which was not statistically different from SKO levels. The decrease in total LN cell cholesterol in DKO ϩ AI mice appears to occur mainly through a reduction in free cholesterol, shown in Fig. 2, panel C, and although this appeared to be slightly higher, it was not statistically different from SKO levels. The ester cholesterol, Fig. 2, panel D, appeared to be slightly lower than DKO levels but did not reach statistical significance, again suggesting that apoA-I treatment attenuated cellular cholesterol accumulation but was not sufficient to bring sterol content completely back to SKO levels.
To demonstrate that cholesterol-loaded immune cells can efflux lipids, 125 I-labeled apoA-I was added to the serum-free culture medium of freshly isolated LN cells obtained from 12-week DKO and SKO mice. Shown in supplemental Fig. 1, immune cells isolated from 12-week diet-fed DKO mice produced a greater amount of lipidated apoA-I particles that Panel A, study design shows 6 -7-week-old mice were fed an atherogenic diet for 6 weeks before beginning 6 weeks of subcutaneous injections of 500 g of lipid-free human apoA-I or BSA every 48 h while maintained on the atherogenic diet for a total of 12 weeks; panel B, timed plasma samples were assayed for human apoA-I following injection of lipid-free apoA-I; panel C, lipoprotein cholesterol distribution following FPLC of mouse plasma; panel D, Western blot of 4 -30% nondenaturing gradient gel of timed plasma samples following injection of lipid-free human apoA-I. LDLr Ϫ/Ϫ , apoA-I Ϫ/Ϫ ϭ DKO mice and LDLr Ϫ/Ϫ ϭ SKO mice. All data represent a minimum of 5-8 male mice per group. ranged between 8.5 and 7.5 nm in diameter when compared with the amount of particles generated from SKO LN immune cells. Therefore, because lymphocytes from diet-fed DKO mice contain a greater mass of cholesterol than SKO lymphocytes (Fig. 2), it follows that they are more likely to efflux more cho-lesterol creating more lipidated particles once an acceptor, such as apoA-I, is available.
Redistribution of Lymph Node Cell Populations-We next examined whether apoA-I treatment would alter the distribution of specific subsets of LN immune cells. As we have shown in Fig. 2, and in previous studies (10), that DKO LNs increase in size, as well as in the total number of cells in response to diet, we next examined how apoA-I administration might alter the distribution of LN cell subsets. Fig. 3, panels A and C, shows the total number of LN DCs and macrophages, and Fig. 3, panels B and D, shows their percent distribution among all LN cells, respectively. As expected, both DKO and SKO mice have significantly increased numbers of DC after 6 and 12 weeks in response to the cholesterol-containing diet, shown in Fig. 3, panel A. However, the DKO ϩ AI group showed significantly more LN DCs when compared with 12-week DKO mice. An increase in DC cell number translated to an enrichment in the percent of DCs in LN cells, with apoA-I-treated DKO mice showing a slight enhancement over 6-and 12-week DKO mice, as shown in Fig.  3, panel B. Also expected, the number of LN macrophages was increased in response to both 6 and 12 weeks of diet, as seen previously (10). A small but not statistically significant decrease in the number of macrophages was seen between DKO mice that received apoA-I treatment and DKO mice that did not, as shown in Fig. 3, panel C. Despite the increase in total LN macrophages, the percent of macrophages in total LN cells was actually lower than in chow-fed mice, as shown in Fig. 3, panel D. Interestingly, apoA-Itreated DKO mice showed a modest but significant increase in the percent of LN macrophage content over nontreated DKO mice, as shown in Fig. 3, panel D, but did not reach levels seen in chow-fed mice. Overall, these data suggest that diet increases LN macrophage as well as DC cell numbers regardless of the genotype. In addition, apoA-I treatment appears to have increased both DC and macrophage cell num-   A and B show the number and percent of DCs in total LN cells, respectively. Panels C and D show the total number and percent of macrophages in total LN cells, respectively. LN cells were stained with CD3, CD11c CD11b, and F4/80 surface markers, and the CD3 Ϫ CD11c ϩ population was measured for DCs and CD11c Ϫ CD11b ϩ F4/80 ϩ population was measured for macrophages. All cell populations were determined by FACS. Total LN cells were isolated from four sets of skin draining LNs that include the brachial, inguinal, axillary, and superficial cervical subsets per mouse. Data represent mean Ϯ S.D. for a minimum of n ϭ 5-14 mice per group. Different lowercase letters indicate significant differences at p Ͻ 0.05. bers in LN as well as their relative proportion among immune cells.
Reduction in Activated T Cell Populations-We next examined levels of T cells, which are one of the largest population of cells found in the LN. Striking effects of apoA-I treatment were seen on lymph node T cell isotypes, as shown in supplemental Fig. 2. The supplemental Fig. 2, panel A, shows that a large significant effect was observed in the number of CD4 ϩ CD69 high T cells in mice treated with apoA-I. This was accompanied by a significant reduction in the percent of activated T cells as total CD4 ϩ cells, as shown in supplemental Fig. 2, panel B when compared with untreated diet-fed DKO mice. Other subtypes of activated T cells examined, such as CD8 ϩ CD69 high and CD8 ϩ CD62L low , were also reduced in DKO ϩ A-I treated mice when compared with untreated DKO mice (data not shown). However, despite the dramatic changes in activated T cells populations in response to apoA-I treatment, there were minimal to no reductions in autoantibody titers to dsDNA, as shown in supplemental Fig. 3, panel A, when compared with untreated DKO mice. This outcome is not unexpected because autoantibody titers have been shown to turn over slowly in animal models of autoimmunity (26). Additionally, titers to anti-oxLDL total IgM isotype were unchanged, as shown in supplemental Fig. 3, panel B, among all treatment groups. The total IgG isotype of anti-oxLDL was increased significantly in mice receiving apoA-I injections as compared with mice injected with BSA, which were similar to mice that did not receive any injections, shown in supplemental Fig. 3, panel C. Although there was a change in the total IgG isotype titer for anti-oxLDL, there was no significant difference in the ratio of IgG1 to IgG2c isotypes for oxLDL, among the different groups, as shown in supplemental Fig. 3, panel D.
Induction of T-regulatory Cells-Within the T cell subtypes, we also examined regulatory T cells or Tregs that are CD4 ϩ CD25 ϩ Foxp3. Both the number and percent distribution of CD4 ϩ CD25 ϩ FoxP3 ϩ cells are shown in Fig. 4, panels A-C, respectively. Interestingly, after treatment with apoA-I, the number of LN CD4 ϩ CD25 ϩ FoxP3 ϩ cells was similar to levels seen in 6-week DKO mice (Fig. 4, panel A), whereas 12-week diet DKO mice showed a large increase compared with 12-week SKO mice. However, the distributions of Tregs expressed as the percentage of total CD4 ϩ cells, shown in Fig. 4, panels B and C, were significantly increased in DKO ϩ AI mice when compared with either 6-or 12-week DKO, suggesting that apoA-I treatment induced conditions favoring Treg cell expansion within the CD4 ϩ T cell pool.
As shown in previous studies (10), diet-fed DKO mice displayed a significant increase in different types of activated T cells. Activation markers such as CD69 and CD44 and low expression of CD62L typify the DKO response to dietary cholesterol, which is not observed in diet-fed SKO mice (10). It then seems logical to determine whether the increase in Tregs was associated with changes in the activation state of T cells in apoA-I-treated DKO mice. In Fig. 4, panel D, the number of CD4 ϩ CD62L low effector/effector-memory T cells was reduced 2-fold in DKO ϩ A-I mice compared with DKO 12-week mice, although not completely back to levels seen in SKO mice.
Interestingly, when the ratio of Treg cells to total CD4 ϩ cells, or Teff cells was examined, as shown in Fig. 4, panel E, a small decrease was seen in both diet-fed DKO and SKO mice in response to diet. However, DKO ϩ A-I mice showed an increase in the Treg/Teff ratio, suggesting that apoA-I treatment may modulate T cell activation through an increase in Treg populations. T Cell Stimulation and Proliferation-In view of the increase in the population of regulatory T cells within LNs of apoA-Itreated DKO mice, we sought to determine whether the T cell response had been restored by apoA-I treatment. Ideally, if Treg function was restored, then T cell activation would be expected to be suppressed following apoA-I treatment, and the magnitude of [ 3 H]thymidine incorporation was lower than the chronically stimulated T cells from untreated DKO mice. Fig. 5 shows the percent [ 3 H]thymidine incorporation for LN-derived immune cells shown in panels A-C and for spleen cells in panels D-F under the following conditions: Fig. 5, panels A and D, unstimulated cells; panels B and E, stimulated cells, and panels C and F, the relative fold increase upon stimulation. Therefore, when T cells are stimulated with anti-CD3/CD28, the extent of [ 3 H]thymidine incorporation monitoring proliferation is greater in DKO ϩ apoA-I-treated cells when compared with T cells that have been chronically activated in the absence of apoA-I (Fig. 5, panel C).
Skin Lipid Levels Normalized with ApoA-I Treatment-The hallmark of the autoimmune phenotype in diet-fed DKO mice is the accumulation of cholesterol and inflammatory cells within the skin. Therefore, we sought to determine the effects of apoA-I treatment on this tissue. Both the skin total cholesterol (Fig. 6, panel A) and triglyceride content (Fig. 6, panel B) were measured as a function of time on the diet. As seen in Fig.  6, panel A, DKO skin cholesterol increased with time on the diet but not in SKO mice. After 6 weeks on the diet, DKO mice had ϳ10 mg/g cholesterol in the skin, and by 12 weeks this doubled to ϳ20 mg/g. Not only did apoA-I treatment prevent any further accumulation of cholesterol in the skin, but it actu-ally reduced skin cholesterol levels to those seen in 4-week dietfed DKO mice, suggesting that further apoA-I treatment might have reduced it to levels seen in chow-fed mice.
Unlike cholesterol, the triglyceride content of DKO and SKO mice skin differed ϳ2-fold after only 2 weeks on the diet, as shown in Fig. 6, panel B. Interestingly, the diet-fed DKO skin triglycerides actually dropped with time on diet, whereas SKO mice showed a modest increase after 6 weeks on the diet. By 12-week DKO, skin triglyceride had dropped to only 25% of its original value. Unexpectedly mice treated with apoA-I showed skin triglyceride levels similar to 12-week SKO mice.
Lipid changes were also seen at the histological level. Fig. 6, panel C, shows the results of Oil Red O staining of skin sections from each of the four different study groups. No differences were seen in skin morphology between DKO and SKO mice before 6 weeks of diet consumption (data not shown). By 6 weeks, thickening of the skin and small lipid droplets were apparent, with a reduction in the size and appearance of the skin fat layer. After 12 weeks of diet, DKO skin showed increased thickening and intense lipid staining in the dermis with an apparent loss of its fat layer. Quite dramatically, the apoA-I treatment group showed normalization in skin morphology and appeared similar to that seen in 6-week diet-fed DKO mice.
Because tissue damage in certain autoimmune disorders has been associated with the presence of autoantibodies (27), we examined skin sections from each of the diet-fed groups for the presence of IgG. The supplemental Fig. 4, panels A-C, shows that skin sections from diet-fed DKO mice (supplemental Fig. 4,  panel B) contain the greatest deposit of IgG when compared with either SKO (supplemental Fig. 4, panel A) or DKO mice treated with apoA-I (supplemental Fig. 4, panel C). These results strongly suggest that IgG or IgG-containing complexes represent a major contributor to skin inflammation in diet-fed DKO mice, because tissue damage in autoimmune disorders have been associated with the presence of autoantibody (27).
Infiltration of Immune Cells Is Reversed after ApoA-I Treatment-Further investigations into DKO skin morphology were carried out to characterize immune cell infiltration before and after apoA-I treatment. Fig. 7, panels A-D, show H&E staining of representative skin sections from each of the four study groups. Fig. 7, panels E-L, shows immunofluorescence detection of CD68 ϩ cells (panels E-H) and CD11c ϩ cells (panels I-L) for mice from each of the four study groups. From these data, we see that both CD68 ϩ as well as CD11c ϩ cells are present in the skin of diet-fed DKO after 6 weeks of diet, and their infiltration into the skin only increased with time on diet. Treatment of 6-week diet-fed DKO mice with apoA-I appears to have attenuated the infiltration of these immune cells and restored normal skin morphology.

DISCUSSION
This study demonstrates that the administration of lipid-free human apoA-I reduces inflammation and cholesterol accumulation in the skin and skin-draining LN of diet-fed male DKO mice. These conclusions were based on the findings that subcutaneous injections of apoA-I decreased cholesterol accumulation as well as the activation state of immune cells in LNs. In addition, in response to apoA-I treatment, there was a large decrease in skin cholesterol and an increase in triglyceride content accompanied by decreased cellularity and the restoration of skin architecture, morphology, and composition. Most importantly, a significant increase in the LN Tregs population was associated with a decrease in LN effector/effector-memory cells and an improved Treg/Teff ratio as well as a reduction in the activation state and proliferative response of the T cells. . ApoA-I treatment restores skin neutral lipid content. Panel A shows total skin cholesterol content expressed as mg/g wet weight. Panel B shows the total skin triglyceride content expressed as mg/g wet weight. Data represent mean Ϯ S.D. for a minimum of n ϭ 5-8 mice per group. Different lowercase letters indicate significant differences at p Ͻ 0.05. Panel C shows Oil Red O staining of 10-m-thick skin sections from DKO 12-week chow-fed mice; DKO 6-week diet-fed mice; DKO 12-week diet-fed mice; DKO 12-week diet-fed mice ϩ 6-week apoA-I treatment. Sections shown are representative of at least four sections per animal with at least 4 -6 different animals per genotype and treatment group. There were no regional skin staining differences observed among any of the genotypes or treatment groups.
Taken together, these results suggest that treatment with human apoA-I reduced inflammation within the skin and LNs, possibly through an increased Treg response.
The importance of Tregs in modulating T cell responses is of great significance because the immune system is complex, with multiple levels of regulation in place to prevent reaction against self. Tregs maintain peripheral self-tolerance by suppressing activation of T cells (28). In addition, Tregs can act on dendritic cells, macrophages, and B cells by inducing apoptosis or by inhibiting their function (29). Tregs exert their inhibitory effects by either a contact-dependent process or by secreting anti-inflammatory cytokines such as IL-10 and TGF-␤ (29 -31).
Although Tregs have the ability to suppress other cells, it is known that under chronic inflammatory conditions both effector cells and antigen-presenting cells can become resistant to Treg suppression and produce cytokines that inhibit Treg func-tion (29,32). Thus, targeting of pathways to inhibit the inflammatory response is a potential therapeutic approach that may result in improved Treg function (29). It is important, however, to maintain a balance between Treg and effector cells because, although a deficiency of Tregs can result in severe autoimmune disease (33)(34)(35), it is possible that increased numbers of Tregs could negatively affect the immune response to infection (33).
Previous studies from our laboratory indicate that the lack of cholesterol efflux in DKO mice was responsible for inflammation in the skin and LNs and the autoimmune phenotype (8 -10). Importantly, inflammation and autoimmunity are induced by dietary cholesterol and can be fatal for as early as 12 weeks after beginning the diet (8 -10). It is well accepted that HDL apoA-I has beneficial effects on atherosclerosis (1, 36 -39), with both human and animal studies demonstrating that infusion of apoA-I or reconstituted HDL reduces atherosclerosis and vascular inflammation (40 -44). D-4F, an apoA-I mimetic peptide, when administered in combination with pravastatin (45), was shown to inhibit collagen-induced arthritis, an animal model of rheumatoid arthritis. While further studies with the 4F peptide showed significant reduction in lupus-like manifestations in mice (63). Thus, we posed the following question. Could treatment with human lipidfree apoA-I reverse inflammation and autoimmunity in DKO mice? Given that cholesterol accumulation in DKO mice begins as early as 4 weeks on the diet (8,9), we chose to start apoA-I injections after mice had been on the diet for 6 weeks, a time point where the cascade of inflammation and cholesterol accumulation has already begun. As expected, injection of human apoA-I resulted in the formation of HDL.
With HDL apoA-I injections, male DKO mice exhibited reduced infiltrates and inflammation in skin and LNs, including a decrease in LN-activated T cells. These effects have not been demonstrated in female diet-fed DKO mice. We speculate that this was due to both the anti-inflammatory properties of HDL apoA-I and the restoration of cellular cholesterol efflux. These beneficial effects may be due in part to direct effects of apoA-I on cellular function as apoA-I has been shown to inhibit the contact-mediated activation of monocytes by stimulated T cells FIGURE 7. ApoA-I treatment reduces immune cell infiltrate into the skin. Panels A-D show H&E staining of skin sections from each of the four study groups; panels E-H show immunofluorescence staining for CD68 ϩ (macrophages); panels I-L show immunofluorescence staining for CD11c ϩ (dendritic cells). Panels A, E, and I are skin sections from 12-week diet-fed SKO mice; panels B, F, and J are skin sections from 6-week diet-fed DKO mice; panels C, G, and K are skin sections from 12-week diet-fed DKO mice; panels D, H, and L are skin sections from 12-week diet-fed DKO mice treated for 6 weeks with apoA-I. Sections shown are representative of at least four sections per animal with at least 4 -6 different animals per genotype and treatment group. (46,47). ApoA-I of HDL has been shown to reduce the monocyte inflammatory response (48). Interestingly, when ABCA1 was blocked, the beneficial effect of apoA-I was also blocked, suggesting that cholesterol efflux is essential for this mechanism (48). Thus, cholesterol efflux to apoA-I plays an important role in maintaining both cholesterol and immune system homeostasis.
With cholesterol efflux restored in DKO ϩ A-I mice, cholesterol along with macrophage and DC cell infiltrates were removed from the skin, resulting in an overall reversal of inflammation. This is significant given that skin cholesterol accumulation can prove to be fatal for DKO mice (8,9). We speculate that cholesterol efflux and the presence of apoA-I reduced cellular activation and inflammation in the skin, allowing for this restoration of skin structure. Importantly, two potentially inflammatory cell types, macrophages and DCs, were reduced in the skin. The decrease in DCs and macrophages in the skin was accompanied by a trend toward increases in both of these cell populations in the LNs, suggesting that these cells may have migrated from the skin to the LNs. This would agree with results from previous studies that indicated that HDL restored migration of DCs from the skin to LNs in hypercholesterolemic mice (49). In hypercholesterolemic apoE Ϫ/Ϫ mice, the DCs that are in the skin are activated, and their inability to migrate back to the LNs results in poor immunologic priming and contributes to local inflammation in the skin (49). It is possible that impaired migration of activated DCs from the periphery may aid in exacerbating memory responses to the periphery (49). By helping maintain normal DC migration, HDL supports a healthy protective immune response.
Following treatment with apoA-I, the total cholesterol in LNs was reduced compared with DKO mice. This reduction in cholesterol is significant given that an environment with increased cholesterol can lead to increased T cell proliferation (50). Consistent with this idea, a reduction in the total LN cell number was seen in DKO ϩ A-I mice compared with DKO mice; however, this was not different from 6-week diet-fed DKO. This suggests that apoA-I treatment did not reverse the cellular accumulation in LNs but rather served to arrest the accumulation and proliferation. It is also possible, however, that the LN cells are involved in cholesterol removal from the skin, with lymphocytes constantly moving from the skin to the LNs, and that the number would be further reduced if the treatment period had been extended to the point where the skin cholesterol accumulation was completely reversed. Nevertheless, the reduction in cells, especially activated T cells, in the LNs demonstrates an important effect of apoA-I treatment on the LNs, where immune responses are stimulated.
Tregs are an important component of the immune response, acting to maintain self-tolerance. The absence of or defects in Treg can lead to severe autoimmune disorders (29,(33)(34)(35)51). Despite the autoimmune phenotype in diet-fed DKO mice, Tregs were increased in LNs. This is similar to findings from previous studies in patients with rheumatoid arthritis, where Tregs were increased in synovial fluid of inflamed joints (52,53). The increase of Treg in tissue suggests that Tregs are homing to sites of inflammation. The correlation between Treg levels in peripheral blood and disease activity is not clear, as some studies have reported increased Tregs in blood, and others show no difference or decreased Tregs compared with healthy patients (33, 54 -57). Thus, the level of circulating Treg does not always reflect the levels at the site of inflammation because Treg are typically increased at sites of inflammation (33).
When Tregs are increased at sites of inflammation, why does the autoimmunity still exist? Studies suggest that the suppressive activity of Treg may be maintained but that Teffs or chronically activated antigen-presenting cells in some autoimmune models may be able to resist this suppression (29, 32, 33, 58 -60). Both the MAPK and NFB pathways have been implicated in the chronic activation of antigen-presenting cells (29). Thus, the MAPK pathway, which is affected in cholesterolloaded T cells from ABCG1 Ϫ/Ϫ mice (50), may also play a role in the ability of cells to overcome suppression by Treg. In addition, the cytokine environment may have an effect on the ability of Teff to overcome suppression by Treg, and resistance to Treg suppression by Teff typically occurs in the setting of chronic inflammation (29,32,33). Although Tregs produce anti-inflammatory cytokines that can act to suppress Teff, Teffs produce inflammatory cytokines that can render Treg ineffective. Thus, it is important to maintain a balance between Treg and Teff.
Although the increase in Treg in DKO mice may seem paradoxical considering the autoimmune phenotype, it may be that the function of LN Treg in DKO mice is impaired or that the Teffs are simply more effective at overcoming suppression. We have previously shown increased proliferation in CD4 ϩ cells in DKO mice. The increase in LN Treg (which are CD4 ϩ ) in DKO mice may be a direct result of increased CD4 ϩ cell proliferation. In other words, Treg may be proliferating along with other CD4 ϩ cells. It appears, however, that the Tregs are not proliferating to the same extent as other CD4 ϩ cells because the ratio of Treg/Teff in DKO mice is reduced compared with both SKO and DKO ϩ A-I. Nevertheless, treatment with apoA-I reduces inflammation in DKO mice, resulting in an increased LN Treg population. In this case, T cell activation is also reduced, suggesting that the Treg function is improved or that Teff cells are less resistant to Treg suppression. Overall, the ratio of Treg/ Teff is increased in DKO ϩ A-I mice compared with diet-fed mice. In DKO ϩ A-I mice, apoA-I reduces inflammation, improving the ability of Treg to respond to Teff. In this case, the increased Treg/Teff ratio may be a result of selective proliferation of Treg. In other words, Tregs are now proliferating in increased proportion compared with other CD4 ϩ cells. Taken together, these data demonstrate that DKO mice have a defective Treg response caused by impaired Treg function or by improved ability of Teff to resist suppression and that the reduction of systemic inflammation in DKO ϩ A-I mice allows for an improved Treg response.
Interestingly, studies have suggested that Treg can prime DCs in sites of inflammation, leading to a reduction in co-stimulatory signaling with T cells (61,62). These DCs gain the ability to migrate to secondary lymphoid organs where they present self-antigen to Teff. The reduction in co-stimulatory signaling results in reduced T cell activation (29). This is an interesting point of reference considering the trend toward increased DCs in the LNs. It is possible that Tregs are also present in increased numbers in the skin, a major site of inflammation in DKO dietfed mice (8,9), but that DCs are unable to migrate back to the LNs after interaction with Treg. As seen in previous studies that demonstrated improved migration of DCs from the skin to the LNs with the addition of HDL (49), apoA-I treatment may restore migration of Treg-primed DCs from the skin to the LNs in DKO mice, thus contributing to the reduction in T cell activation. Although both the number and percentage of CD4 ϩ CD62L low cells in DKO LNs was not decreased to SKO levels, the observed reduction in these cells is significant in that it indicates that the immune system in apoA-I-treated mice has less potential for self-reactivity when compared with controls.
In conclusion, we have demonstrated that treatment of dietfed DKO mice with apoA-I results in reduced inflammation in the skin and LNs and an increase in LN Treg, which leads to reduced T cell activation. Thus, administration of apoA-I in diet-fed DKO mice reduces the potential for self-reactivity. These findings have important implications for the treatment of diseases involving chronic inflammation and autoimmunity such as rheumatoid arthritis and systemic lupus erythematosus, including the use of apoA-I to improve the Treg response.