Peroxisome Proliferator-activated Receptor γ Ligands Regulate Myeloperoxidase Expression in Macrophages by an Estrogen-dependent Mechanism Involving the -463GA Promoter Polymorphism*

A functional myeloperoxidase (MPO) promoter polymorphism, -463GA, has been associated with incidence or severity of inflammatory diseases, including atherosclerosis and Alzheimer's disease, and some cancers. The polymorphism is within an Alu element encoding four hexamer repeats recognized by nuclear receptors (AluRRE). Here we show that peroxisome proliferator-activated receptor γ (PPARγ) agonists strongly regulate MPO gene expression through the AluRRE. Opposite effects were observed in granulocyte/macrophage colony-stimulating factor (GMCSF)- versus macrophage colony-stimulating factor (MCSF)-derived macrophages (Mφ): Expression was markedly up-regulated (mean 26-fold) in MCSF-Mφ and down-regulated (34-fold) in GMCSF-Mφ. This was observed with rosiglitazone and three other PPARγ ligands of the thiazolidinedione class, as well as the natural prostaglandin metabolite 15-deoxy-Δ12,14 prostaglandin J2. The selective PPARγ antagonist, GW9662, blocked both the positive and negative effects on MPO expression. Gel retardation assays showed PPARγ bound hexamers 3/4, and estrogen receptor-α bound hexamers 1/2, with -463A in hexamer 1 enhancing binding. Estrogen blocked PPARγ effects on MPO expression, especially for the A allele. Charcoal filtration of fetal calf serum eliminated the block of PPARγ, whereas replenishing the medium with 17β-estradiol reinstated the block. These findings suggest a model in which estrogen receptor binds the AluRRE, preventing PPARγ binding to the adjacent site. The positive and negative regulation by PPARγ ligands, and the block by estrogen, was also observed in transgenic mice expressing the G and A alleles. The mouse MPO gene, which lacks the primate-specific AluRRE, was unresponsive to PPARγ ligands, suggesting the human MPO transgenes will enhance the utility of mouse models for diseases involving MPO, such as atherosclerosis and Alzheimer's.

macrophages. MPO plays a key role in the innate immune system as a microbicidal agent (1). Upon ingestion of microbes by neutrophils or monocyte-macrophages, azurophilic granules containing MPO fuse with the phagosome, releasing the enzyme, which reacts with chloride and superoxide-generated hydrogen peroxide to produce hypochlorous acid (HOCl), a toxic oxidant (2). The MPO-HOCl pathway also generates reactive nitrogen species (3) and tyrosyl radicals (4). The inadvertent release of MPO and its reactive byproducts at inflammatory sites can damage bystander cells, implicating MPO in oxidative damage at atherosclerotic lesions, Alzheimer's plaques, and some cancers (5).
Several lines of evidence link MPO to oxidative damage in atherosclerosis. First, MPO and the end products of MPOgenerated oxidants are detected in lesions, colocalizing with foam cell macrophages (6 -9). Second, MPO levels in circulating leukocytes and serum are higher in individuals with coronary artery disease (CAD) (10,11). Third, individuals with inherited MPO deficiencies have less cardiovascular disease (12). Fourth, a functional MPO promoter polymorphism has been associated with increased incidence of CAD (13,14) and severity of atherosclerosis (15,16). MPO may contribute to atherosclerosis through its ability to oxidize LDL (7), causing aggregation that enhances uptake through macrophage scavenger receptors (17), leading to lipid-laden foam cell macrophages, a hallmark of atherosclerotic lesions (18).
MPO is highly expressed in promyelocytes (19), with expression sharply decreasing as these precursors mature along the granulocyte or monocyte lineages. MPO protein is retained, stored in cytoplasmic vesicles. When monocytes differentiate to tissue macrophages, MPO protein is no longer present, yet the gene can be reactivated in subsets of reactive macrophages. For example, quiescent brain macrophages, microglia, lack MPO, yet ␤-amyloid, a component of Alzheimer's plaques, is able to induce MPO gene expression, and MPO is abundant in reactive microglia surrounding plaques (20), and in foam cell macrophages at atherosclerotic lesions (21). It is important to understand the transcription factors and regulatory pathways that reinitiate MPO gene expression in reactive macrophages (22). Peroxisome proliferator-activated receptor ␥ (PPAR␥) is a transcription factor abundant in foam cell macrophages (23) that regulates some genes involved in inflammatory responses, such as scavenger receptor CD36 (24) and inducible nitric-oxide synthase (25). In this study, we examine the role of PPAR␥ in the regulation of the human MPO gene.
PPAR␥ is a member of the nuclear receptor superfamily of ligand-activated transcription factors. It forms heterodimers with retinoid X receptor (RXR) to bind two direct repeat hexamers, related to the AGGTCA consensus half-site, with preferred spacing of one base pair (26). PPAR␥ is a key regulator of adipogenesis and lipid metabolism (27), abundant in fat cells and macrophages (23,28), with suppressive effects in some cancers (29).
Naturally occurring PPAR␥ ligands include derivatives of fatty acids and prostaglandin metabolites such as 15-deoxy-⌬ 12,14 prostaglandin J 2 (15d-PGJ 2 ) (30,31), which is present at significant levels in macrophages (32). Synthetic ligands with higher affinity for PPAR␥ include the insulin-sensitizing drugs of the thiazolidinedione class (TZD) (33). Diabetic patients are often treated with TZDs, and are also at heightened risk for heart disease, which has raised concerns as to the potential impact of TZDs on PPAR␥-mediated expression of CD36 or other genes involved in atherosclerosis.
The promoter elements that regulate the myeloid-specific expression of the MPO gene are only partially understood (34). A functional upstream promoter polymorphism, Ϫ463GA, has been associated with expression levels (35,36), as well as incidence or severity of atherosclerosis (13,14), Alzheimer's (20, 38 -40), MPO-anti-neutrophil cytoplasmic antibodies vasculitis (41), hepatitis C virus-induced fibrosis (42), multiple sclerosis (43), periodontal disease (44), lung cancer (44,46), and myeloid leukemia (35). A number of studies report gender differences in association of Ϫ463GA with risk (20,38,40,41,44,46). This polymorphism is in an Alu element, within a cluster of four hexamer half-sites recognized by various nuclear receptors (36,47), termed an Alu receptor response element (AluRRE). These hexamers are organized as direct repeats with spacing of 2-4-2 base pairs. The Ϫ463A is in hexamer 1, within an estrogen receptor binding site (38), whereas Ϫ463G promotes binding by SP1 transcription factor (36). The hexamer 3/4 pair is recognized by retinoic acid receptor as a heterodimer with retinoid X receptor, and the middle DR-4 pair (hexamers 2/3) is recognized by thyroid hormone receptor (36). The Ϫ463GG genotype has been linked to higher MPO mRNA and protein expression than GA/AA genotypes in primary myeloid leukemia cells (35). In transfection assays, the Ϫ463G promoter element supported higher expression of a reporter gene than Ϫ463A (36).
In the present study, we investigated the role of PPAR␥ in the regulation of MPO gene expression in macrophages. The findings indicate both positive and negative effects of PPAR␥ ligands on MPO expression, influenced by MCSF and GMCSF, the Ϫ463GA polymorphism, and estrogen.

EXPERIMENTAL PROCEDURES
Isolation and Culture of Human Peripheral Blood Mononuclear Cells (PBMC)-PBMC were isolated from 500 ml of whole blood from healthy donors. A blood bank provided the cells without personal identifiers as a 50-ml concentrate obtained by low speed centrifugation. A single AA genotype donor was identified, and provided 20 ml of blood. Collection of samples from human donors and anonymous samples from the blood bank was approved by our affiliated institutional review committee. The concentrated leukocytes in 20-ml volume were layered over 20 ml of Ficoll Hypaque (Lymphoprep, Axis Shield) and centrifuged for 30 min at 900 ϫ g. The interphase cell layer was again centrifuged over Lymphoprep, and the interphase layer was diluted with three volumes of RPMI medium 1640 (Invitrogen) and collected by centrifugation for 30 min at 900 ϫ g. The cell isolate includes monocytes, macrophages, and lymphocytes, and is largely depleted of neutrophil/granulocytes. The cells were resuspended in 40 ml of RPMI medium supplemented with 200 mM L-glutamine, 10,000 units/ml penicillin G, 10,000 g/ml streptomycin, nonessential amino acids (Irvine Scientific), 1ϫ Fungizone (Invitrogen), and 10% fetal calf serum (HyClone). The human serum for each donor was heat-inactivated (55°C, 30 min) and added to the culture medium at 10% volume. The cells were plated in 24-well tissue culture plates (10 6 cells/well) in 400-l volume with human GMCSF (Sigma) (10 ng/ml) or human MCSF (R&D Systems) (10 ng/ml) in a humidified CO 2 incubator at 37°C. The cells were incubated for 24 h prior to addition of PPAR␥ ligands for an additional 24 h, followed by harvesting.
Culture of Monocyte-derived Macrophages-To obtain monocyte-derived macrophages, PBMC were seeded at 10 6 cells/400 l/well in 24well plates in RPMI with 10% FCS and 10% autologous human serum, along with recombinant human GMCSF (10 ng/ml) or MCSF (10 ng/ml). Where indicated in the figures, 17␤-estradiol (Sigma) (10 Ϫ7 M) was added. On days 2 and 3, the medium was supplemented with 100 l of fresh medium containing GMCSF or MCSF (10 ng/ml). On days 4 -6, medium was exchanged daily, without human serum, with GMCSF at 10 ng/ml or MCSF at 10 ng/ml, and where indicated, 17␤-estradiol (10 Ϫ7 M). On day 7, fresh medium with rosiglitazone or other ligands was added for an additional 24 h prior to harvesting. The resultant cell population was uniformly adherent, with macrophage morphology, and 98% were positive for CD68 by immunostaining.
RNA Isolation and Quantitation by TaqMan Real-time PCR-Total RNA was isolated from cells by TRIzol reagent (Invitrogen). Medium was aspirated and 400 l of TRIzol added directly to the adherent macrophages in 24-well plates. For PBMC, TRIzol was added to combined nonadherent and adherent cells. The RNA was reverse transcribed with the Omniscript RT kit (Qiagen) and random hexamer primers. Five l of the 20-l cDNA reaction volume was used in realtime quantitative PCR using ABI PRISM 7900 (PerkinElmer Applied Biosystems) using TaqMan Master mix and primers. Normalization was to glyceraldehyde 3-phosphate dehydrogenase (GAPDH) for human RNA and ␤ 2 -microglobulin for mouse RNA. Fluorescence was measured with the Sequence Detection Systems 2.0 software.
Measurement of MPO Enzyme Activity-Bone marrow cells (10 6 ) were harvested by centrifugation (10,000 ϫ g, 5 min), washed with PBS (137 mM NaCl, 2.7 mM KCl, and 10 mM phosphate buffer), and resuspended in 50 l of PBS containing Triton X-100 (0.2%) and phenylmethylsulfonyl fluoride (1 mM). Protein concentrations were determined using Bio-Rad Protein Assay. MPO activity was measured colorimetrically by oxidation of substrate guaiacol, monitored by absorbance at 470 nm. A reaction volume of 200 l contained 5 l of cell extract, sodium phosphate buffer, pH 7.0 (100 mM), 13 mM guaiacol, 2 mM 3-aminotriazole, and 670 M H 2 O 2 . The change in optical density at 470 nm was measured over 5 min using a 96-well plate reader (Spectramax Plus, Molecular Devices). The rate in milliunits/ml/min was calculated from the slope for the initial 2 min.
To measure MPO activity in 7-day macrophages, adherent macrophages were cultured in 6-well plates, washed twice with PBS, and harvested in 50 l of PBS containing Triton X-100 (0.2%) and phenylmethylsulfonyl fluoride (1 mM).
Western Blot Analysis-Cells were lysed in SDS sample buffer, and equal amounts of total protein were electrophoresed on a 4 -20% gradient denaturing SDS-PAGE gels and transferred to polyvinylidene difluoride membrane. Blots were blocked in 5% nonfat dry milk in Tris-buffered saline containing 1% Tween 20 (TBS-T) for 1 h, and incubated with primary antibodies for 1 h at room temperature. Blots were then washed in TBS-T twice at 15-min intervals, and incubated with horseradish peroxidase-conjugated secondary antibody for 30 min at room temperature. After two subsequent washes in TBS-T, the blots were developed with ECL reagent (Amersham Biosciences).
Determination of MPO Genotype by Allelic Discrimination Assay-DNA was isolated from leukocytes by proteinase K digestion followed by phenol extraction and ethanol precipitation. DNA (100 ng) was used in the ABI7900 allelic discrimination assay using dual fluorophore probes, which discriminate between alleles based on the single base mismatch. Primers and probes, designed using the Primer Express 2.0 software (ABI), are as follows: forward primer, 5Ј-AATCTTGGGCTGGTAGT-GCTAAA-3Ј; reverse primer, 5Ј-GCCAGGCTGGTCTTGAACTC-3Ј; Ϫ463A-specific probe, 5Ј-FAM TCCACCTGCCTCAG MGB; Ϫ463Gspecific probe, 5Ј-VIC TCCACCCGCCTCA MGB. Probes are labeled at the 5Ј end with the fluorophores FAM or VIC, and are stabilized by a minor groove binding moiety (MGB). End point allelic specific fluorescence was measured on the ABI Prism 7900 using Sequence Detection Systems 2.0 software for allelic discrimination.
Generation of MPO G and A Transgenic Mice-Transgenic mice carrying the human G or A alleles were created by microinjection of C57BL6/J eggs with a 32-kb BST11071 restriction fragment including the gene with 7 kb of upstream and 11 kb of downstream sequence. The G allele was isolated from a sequenced BAC clone (AC004687). The A allele was obtained by hybridization screening of a genomic BAC library from Research Genetics. The transgenic facility at the University of California (Irvine, CA) performed the microinjection of the purified 32-kb DNA fragment into mouse eggs. One founder was obtained for each allele.
Isolation of Mouse Bone Marrow Cells and Cell Culture-Bone marrow cells were plated at a density of 10 6 cells/well in 24-well plates in RPMI with mouse GMCSF (R&D Systems) (10 ng/ml) or mouse MCSF (R&D Systems) (10 ng/ml). Cells were incubated for 24 h prior to addition of PPAR␥ ligands for another 24 h in the continued presence of GMCSF or MCSF.
To obtain bone marrow-derived macrophages, bone marrow cells were incubated for 24 h in 25-cm 2 flasks containing 5 ml of RPMI with GMCSF at 5 ng/ml or MCSF at 5 ng/ml. Non-adherent cells were seeded at 10 6 cells/400 l/well in 24-well plates and incubated for 7 days. The medium was supplemented on days 2 and 3 with 100 l of fresh medium with GMCSF or MCSF (10 ng/ml). The medium was exchanged on day 4, 5, and 6 to remove nonadherent cells. Where indicated, ␤-estradiol was present throughout at 10 Ϫ7 M. Adherent cells at day 7 were homogeneous in morphology, consisting of mononuclear cells with abundant cytoplasm.
Immunohistochemistry-Macrophages were briefly fixed in methanol and 4% paraformaldehyde, blocked with 10% goat normal serum for 12 h, followed by incubation with rabbit polyclonal antibody (Ab) to human MPO (Biodesign, Inc., 1:1000) or monoclonal Ab to CD68 (clone EBM11; Dako, 1:1000), then detected with fluorescent anti-rabbit or anti-mouse Ab (AlexaFluor 488 or 594, Molecular Probes,1:1000), and visualized by confocal microscopy. For nonfluorescent detection, cells were treated with 10% hydrogen peroxide for 10 min prior to incubation with primary antibody as above, followed by biotinylated secondary Ab using the ABC method (avidin and biotinylated horseradish peroxidase macromolecular complex; Vectastain ABC kit, Vector Laboratories, Burlingame, CA).
Low density lipoprotein receptor-deficient mice (Jackson Laboratories) were crossed to the MPO G transgenic mice and fed a high fat diet (16% fat, 1.25% cholesterol, Harlan Teklad Laboratory, Winfield, IA) for 16 weeks. Mice were sacrificed and perfused through the left ventricle with PBS. The heart and proximal aorta were embedded in OCT compound and sectioned. Sections were incubated with polyclonal Ab to MPO (Biodesign Int.) and PPAR␥ (Biomol), followed by biotinylated anti-rabbit secondary Ab, and detected with peroxidase chromogen kits (DAB nickel, or 3-amino-9-ethylcarbazole, Vector Laboratories).
Electrophoretic Mobility Shift Assay-Human PPAR␥ and RXR␣ proteins were synthesized in vitro from the corresponding expression plasmids in rabbit reticulocyte lysate by using the TNT transcription/translation system (Promega). To obtain an unprogrammed lysate as a negative control for electrophoretic mobility shift assay, a reaction was performed with the empty vector, pCDNA3 (Invitrogen). Purified human estrogen receptor ␣ (80% purity) was purchased from PanVera. Double-stranded oligonucleotides were radiolabeled at overhanging ends using Klenow polymerase and gel-purified. Proteins were incubated with 1 g of poly(dI-dC) in binding buffer containing 20 mM Hepes, pH 7.9, 50 mM KCl, 1 mM MgCl 2 , 1 mM dithiothreitol, 10% glycerol, 3 g/20 l bovine serum albumin for 10 min at 4°C. 32 P-Labeled probe (100,000 cpm) was then added and incubation continued for 30 min at 4°C. Samples were electrophoresed at 4°C in a 5% polyacrylamide gel in 0.5ϫ TBE buffer (45 mM Tris, 45 mM boric acid, 1 mM EDTA, pH 8.0). Oligonucleotide sequences for probes used in this assay are shown in Fig. 1. The control ERE sequence was an inverted repeat hexamer with 3-bp spacing, sense strand 5Ј-GATCCGTC-AGGTCACAGTGACCTGATC-3Ј.
Statistical Analysis-The real time PCR ⌬C T scores were compared by analysis of variance using Statview software. A value of p Ͻ 0.05 is considered significant.

PPAR␥-RXR Heterodimer Binds to the AluRRE in the MPO
Promoter, Adjacent to the Ϫ463GA Polymorphism and an Estrogen Receptor Binding Site-The Alu upstream of the MPO gene is typical of the major class of Alu elements (Sx) (47) with hexamer spacing of 2-4-2 bp (36) (Fig. 1A). To test for binding by PPAR␥ to the MPO AluRRE, gel retardation assays were carried out with PPAR␥ produced in rabbit reticulocyte lysates programmed with in vitro synthesized mRNA. Synthesized receptors were incubated with 32 P-labeled oligonucleotides (45 bp) including the four hexamers with 6 -7 bp of flanking sequences and either Ϫ463G or Ϫ463A at position 5 of hexamer 1 (Fig. 1A). PPAR␥ bound the AluRRE with either Ϫ463G or Ϫ463A, requiring co-presence of RXR (Fig. 1B). To further define the binding site, we tested oligonucleotides including the individual hexamer pairs and found that PPAR␥ binds to hexamers 3/4 ( Fig. 1C), but not hexamers 1/2 or 2/3 (data not shown). The presence of PPAR␥ in the retarded complex was demonstrated by incubating the proteins with anti-PPAR␥ antibodies, which blocked complex formation (Fig. 1D). Binding by PPAR␥ to direct repeat hexamers with 2-bp spacing, though not optimal, has been reported previously (48).
In an earlier study, we found that estrogen receptor ␣ (ER␣) binds selectively to an oligonucleotide including Ϫ463A (38). In the present study, the 45-bp oligonucleotide included 6 -7 bp of flanking sequences. Under these conditions, purified baculovirus-generated recombinant ER␣ (Panvera) bound the AluRRE with either Ϫ463G or Ϫ463A, in the presence or absence of 17␤-estradiol (Fig. 1E). However, when the DNA sequence was restricted to hexamers 1/2 with 7 bp of 5Ј-and 3Ј-flanking sequence, ER␣ bound preferentially to the A allele sequence (Fig. 1F). When the sequence was further restricted to include one or two bases of flanking sequence, ER␣ bound exclusively to the Ϫ463A sequence (Fig. 1G). This indicates that ER␣ was able to bind either the G or A allele sequence in the context of the longer oligonucleotide, but preferentially bound to the A allele when the available sequence was restricted. The hexamer 1/2 region lacks the consensus ER␣ binding site, which is a palindromic repeat with 3-bp spacing (PR3); however, ER has been previously found to bind to half-sites with atypical spacings (49). Consistent with these binding assays, an earlier study showed that Alu elements can function as estrogen response elements (50).

PPAR␥ Ligands Have Either Positive or Negative Impact on MPO Expression in Human Peripheral Blood Leukocytes-To
investigate the potential impact of the upstream PPAR␥ binding site, we assayed the effects of PPAR␥ ligands on MPO expression in PBMC. PBMC were isolated by Ficoll-Hypaque density gradient. The mononuclear cells were placed in culture medium with 10% autologous serum and 10% FCS for 24 h with RS, a representative of the TZD class of antidiabetic agents. RNA was isolated, and cDNA generated with reverse transcriptase using random primers. Quantitation of MPO cDNA was by real-time PCR. MPO cDNA levels were normalized to GAPDH as endogenous control, co-amplified in a dual label multiplex reaction.
Surprisingly, rosiglitazone had strong yet opposite effects on MPO expression in various donor PBMC, increasing expression by 25-70-fold in some cases, and decreasing expression by 35-70-fold in other cases ( Fig. 2A). There was no correlation of MPO genotype or gender with directionality of this response (data not shown). The proximity of ER␣ and PPAR␥ binding sites in the AluRRE raised the possibility that ER␣ may be competing with PPAR␥ for binding. To test this hypothesis, estrogen (17␤-estradiol, 10 Ϫ7 M) was added to the culture medium. In the absence of its ligand, estrogen receptor is restricted to the cytoplasm. Ligand-bound ER is transported to the nucleus where it binds estrogen response elements. The addition of 17␤-estradiol was found to block the effects of RS, both positive and negative, in most cases.
PPAR␥ Ligands Down-regulate MPO in Cells Treated with GMCSF, and Up-regulate MPO in Cells Treated with MCSF-To investigate the reason for the variable response to RS, PBMC were incubated for 24 h with either GMCSF or MCSF, growth factors that promote monocyte survival and differentiation to macrophages. Five different PPAR␥ ligands were then added for an additional 24 h, in the continued presence of the growth factors. The ligands included the natural prostaglandin metabolite, 15d-PGJ 2 , and four representatives of the TZD class, rosiglitazone, troglitazone, ciglitazone, and MCC-555. RNA was isolated from adherent and nonadherent cells, and cDNA prepared for real time quantitation. In GMCSF-treated PBMC, all five PPAR␥ ligands down-regulated MPO strongly, by 25-250-fold (Fig. 2B). Conversely, in MCSFtreated PBMC, all five ligands up-regulated MPO by 20 -80fold. This suggested that GMCSF induces conditions, such as co-repressors, which allow PPAR␥ ligands to down-regulate MPO, whereas MCSF induces conditions, such as co-activators, allowing PPAR␥ ligands to up-regulate MPO.
PPAR␥ expression increases significantly as monocytes mature to macrophages (51); thus, we next examined the effects of PPAR␥ ligands on monocyte-derived macrophages. PBMC were cultured with GMCSF or MCSF for 7 days to generate macrophages. The adherent macrophages were then treated for 24 h with optimal concentrations of PPAR␥ ligands, 15d-PGJ 2 , rosiglitazone, troglitazone, ciglitazone, and MCC-555 (Fig. 2C). The results were similar to findings with PBMC treated for 48 h; MPO expression was uniformly up-regulated in MC-M, and uniformly down-regulated in GM-M.
Estrogen Blocks Some Effects of PPAR␥ Ligands by a Mechanism Involving the Ϫ463GA Polymorphism-In untreated PBMC, estrogen appeared to block effects of PPAR␥ ligands in most cases ( Fig. 2A). The proximity of the Ϫ463GA polymorphism to the PPAR␥ binding site, and the favored binding by ER␣ to Ϫ463A (Fig. 1), suggested that ER␣ might preferentially interfere with PPAR␥ binding on the MPO A allele. To Ϫ463A or Ϫ463G were incubated with rabbit reticulocyte lysate programmed with in vitro synthesized mRNA encoding PPAR␥ and/or RXR␣ as indicated. The arrow indicates the position of a retarded complex for both A and G allele sequences. C, an oligonucleotide encoding only hexamers 3/4 were incubated with lysates programmed with PPAR␥ expression vector or control vector. Arrows indicate PPAR␥-specific complexes. D, antibodies against PPAR␥ block complex formation with PPAR␥/RXR␣ on hexamers 3/4. Non-immune serum (nis) does not block. E, estrogen receptor ␣ generates a complex on a DNA oligonucleotide encoding the four hexamers with either Ϫ463A or Ϫ463G. ERE is a canonical estrogen response element with palindromic repeat AGGTCA with 3-bp spacing. F, ER␣ forms a complex preferentially with Ϫ463A when using a shorter oligonucleotide consisting of hexamers 1/2 with 7 bp of flanking nucleotides. G, ER␣ forms a complex exclusively with Ϫ463A when using an oligonucleotide with hexamers 1/2 with one or two flanking nucleotides. Nucleotides in parentheses are not present in the MPO Alu sequence. investigate this possibility, we assayed the effects of PPAR␥ ligands on PBMC and macrophages for genotypes GG, GA, and AA, in the presence or absence of supplemental 17␤-estradiol. The representative experiments shown in Fig. 3 assayed varying concentrations of RS. PBMC were initially incubated for 24 h with GMCSF or MCSF, followed by 24 h with RS. The effects were similar for the GG and GA genotypes; RS and PG markedly decreased MPO expression in GM-PBMC (30 -60fold) and increased expression in MC-PBMC (30 -35-fold) (Fig.  3A). AA genotype differed, being down-regulated 40-fold in GMCSF, but unaffected in MCSF. Estrogen blocked all PPAR␥ ligand effects except the down-regulation of GG in GMCSF.
The identical experiments performed with the natural ligand 15d-PGJ 2 produced the same results, with MPO expression increasing 30 -45-fold in MCSF, and decreasing 25-65-fold in GMCSF, with estrogen blocking all effects except the downregulation of GG genotype in GMCSF (data not shown).
Similar findings were obtained with GM-M and MC-M (Fig. 3B). GG and GA genotypes were markedly up-regulated (18 -30-fold) in MC-M and down-regulated (25 to 60-fold) in GM-M. The AA genotype was mildly down-regulated in GMCSF M (7-fold), and did not respond to RS in MCSF-M. Estrogen again blocked all effects except the down-regulation of GG in GMCSF. Equivalent results were obtained using 15d-PGJ 2 (data not shown).
The regulation of the homozygous GG and AA genotypes suggested an explanation for the expression pattern of heterozygote GA genotype, which resembled GG genotype in MCSF, and AA genotype in GMCSF. The A allele was not up-regulated by RS in MCSF, suggesting the up-regulation of GA genotype was the result of up-regulation of G allele alone. Estrogen blocks the down-regulation of AA genotype in GMCSF, suggesting the sustained basal expression of A allele could be masking the down-regulation of G allele in the heterozygote GA (Fig. 3, A and B).
Consistent with the data on mRNA levels, MPO enzyme activity increased in MC-M treated with RS, and decreased in GM-M treated with RS (Fig. 3C). The maximal level of MPO activity in macrophages was observed in MCSF-cultured cells treated with RS, and this maximal level was still severalfold lower than that observed in adherent PBMC (Fig. 3C). The activity measured in GMCSF-or MCSF-cultured macrophages may reflect not only newly synthesized enzyme, but could also include some enzyme retained from monocytes, based on a prior study showing that GMCSF prevents the loss of MPO protein during the differentiation of monocytes to macrophages (21). Table I summarizes the effects of rosiglitazone on PBMC and macrophages from multiple donors. In all cases, MPO expression was strongly up-regulated by RS in MCSF-treated cells, and down-regulated in GMCSF. In GM-PBMC, RS down-regulated GG genotype by an average of 32-fold (p ϭ 0.001), GA by 45-fold (p ϭ 0.0001), and AA by 32-fold (panel I: GG, GA, AA, rows D, column 2). In MC-PBMC, RS up-regulated GG genotype by 20-fold (p ϭ 0.001), GA by 39-fold (p ϭ 0.0001), but had no affect on AA expression levels (column 6). Estrogen blocked up-regulation of GG and GA in MCSF (column 8). In GMCSF, estrogen blocked the down-regulation of GA and AA, but reduced GG down-regulation by less than 2-fold (column 4).
In GMCSF macrophages, RS down-regulated GG genotype by 34-fold (p ϭ 0.0001), GA genotype by 32-fold (p ϭ 0.0001), and AA by 7-fold (panel II: GG, GA, AA, rows D, column 2). Thus, the ability of A allele to respond to PPAR␥ ligands diminishes during differentiation of PBMC to GMCSF-macrophages. Estrogen blocked down-regulation of GA and AA, but not GG genotype in GM-M (column 4). In MCSF macro-phages, GG genotype was up-regulated by 19-fold (p ϭ 0.0001), and GA genotype by 26-fold (p ϭ 0.0001), and both effects were completely blocked by estrogen (columns 6 and 8). AA genotype did not respond to RS in MCSF. Table I lists not only the -fold changes in MPO expression with RS and estrogen (D), but also the average ⌬ threshold cycle (⌬C T ) (rows A and B), which is the difference between MPO C T and the C T of the endogenous control gene. In realtime PCR, fluorescence values are recorded at each cycle and represent the amount of product at that point in the amplification process. The threshold cycle (C T ) refers to the number of cycles required to reach a fluorescence level significantly above background. If more template is present at the beginning of the reaction, fewer cycles will be required to reach C T . C T is always attained during the exponential phase of amplification, such that each cycle represents a doubling of reaction product. The ⌬C T is the difference between the cycles required to attain threshold levels for MPO cDNA versus the control GAPDH cDNA. A difference of 1 C T between target and control cDNA represents a 2-fold difference in RNA levels, whereas a difference of 3 C T indicates an 8-fold difference, and so on. For example, in MCSF-PBMC, GA genotypes had an average ⌬C T of 7.5 (Table I, Table I allow direct comparison of relative MPO mRNA levels in PBMC, macrophages, and in GG, GA, and AA genotypes, all normalized to GAPDH. The GA Genotype Is Higher Expressing than GG Genotype in the Presence of GMCSF, RS, and Estrogen-The GG genotype was 4.6-fold higher expressing than GA in freshly isolated, untreated PBMC (Table I,  Conversely, GA is significantly higher expressing than GG genotype in GMCSFtreated cells in the presence of rosiglitazone and 17␤-estradiol. This is the result of the down-regulation of GG genotype by RS, whereas estrogen blocks this effect on the A allele, which then maintains basal expression. In GM-PBMC treated with RS and 17␤-estradiol, GA was 52-fold higher expressing than GG (panel I, row E, column 4), and in GM-M with RS and 17␤estradiol, GA was 5.6-fold higher expressing than GG (panel II, row E, column 4). Interestingly, the GA genotype was higher expressing than GG in most circumstances in 48-h treated GMCSF-or MCSF-PBMC (panel I, row E), but not in untreated PBMC or GM-or MC-M (panel II, row E).
The Selective PPAR␥ Antagonist GW9662 Blocks the Effects of PPAR␥ Ligands on MPO Expression-TZDs and 15d-PGJ 2 function as PPAR␥ ligands but have also been demonstrated to inhibit gene expression by means independent of the receptor PPAR␥ (24,52). For example, 15d-PGJ 2 interferes with NFBmediated gene regulation by generating adducts with cysteine residues in NFB and IB kinase (53). One means to test whether a given effect is mediated by the receptor PPAR␥ is to use the PPAR␥ antagonist, GW9662, which covalently modifies cysteine 286 in the ligand binding domain (54). Although GW9662 reacts with all three PPAR subtypes ␣, ␦, and ␥, the antagonist activity is at least 100 -1000-fold more potent for PPAR␥ than ␣ or ␦ (54); therefore, it is considered a selective PPAR␥ antagonist.
In Fig. 4A, PBMC were treated with GMCSF or MCSF for 24 h, followed by 24 h with RS (5 or 7 m) in the presence or absence of GW9662 (5 M). GW9662 completely blocked the ability of RS to down-regulate MPO in GMCSF or up-regulate MPO in MCSF. Similarly, in Fig. 4B, macrophages were treated with 9 M 15d-PGJ 2 in the presence or absence of increasing amounts of GW9662. A concentration of 5 M GW9662 completely blocked the 15d-PGJ 2 -dependent increase in expression in MCSF, and the decrease in GMCSF. These findings provide evidence that the receptor PPAR␥ mediates both the positive and negative effects of PPAR␥ ligands on MPO gene expression.
Unlike MPO, Expression of CD36 and Liver X Receptor (LXR) Is Up-regulated by PPAR␥ Ligands in Both GMCSF and MCSF Macrophages-We investigated the possibility that other genes regulated through PPAR␥ might be subject to down-regulation in GMCSF and up-regulation in MCSF. Scavenger receptor CD36 and LXR are known to be up-regulated by PPAR␥ in macrophages (24,52,55). We therefore assayed the effects of RS on the expression of these genes in GM-M and MC-M. Consistent with prior reports, rosiglitazone (10 M) was found to up-regulate CD36 expression by 8-fold in both GMCSF and MCSF macrophages, and LXR was up-regulated by 2-fold in both cases (data not shown). Thus, the opposing regulation observed for MPO appears to be promoter-specific, and may be AluRRE-specific.
One possible explanation for the differences in RS response for the MPO gene in GMCSF-versus MCSF-treated cells is that key transcription factors are expressed at different levels in these conditions. To investigate this possibility we assayed the expression levels of ER␣, PPAR␥, and SP1 in GMCSF-versus MCSF-treated cells. The differences in expression levels for PPAR␥ and SP1 were less than 2-fold in MC-M and GM-M. However, ER␣ expression levels were 4-fold higher in MC-M than GM-M (data not shown). This may contribute to the greater ability of estrogen to block PPAR␥ effects on the A allele in MCSF cultured cells (discussed below).
The Regulation by PPAR␥ Ligands and Estrogen Is Replicated in Transgenic Mice Expressing the Human G and A Alleles-Transgenic mice expressing the human MPO G and A alleles allow independent analysis of the two alleles. In human cells, this requires comparison of homozygous GG genotype to the rare AA genotype (3-5% of population). Analysis of allelic expression in heterozygote cells is not possible because of the apparent absence of sequence differences in the mRNA, including exons or 5Ј-and 3Ј-noncoding transcribed regions, based on GenBank TM entries and our sequencing efforts. To facilitate the analysis of the Ϫ463GA polymorphism, we generated transgenic mice expressing the G and A alleles under control of extensive native human promoter sequences. The transgenics were created by microinjection of C57BL6/J eggs with a 32-kb restriction fragment including the MPO gene. It is worth noting that the injected alleles did not differ at another reported MPO promoter polymorphism, Ϫ129GA (56). PCR analysis confirmed the presence of at least 6 kb of upstream and 4 kb of downstream sequences for both the G and A transgenes. The human MPO gene is appropriately expressed in bone marrow cells, with MPO protein restricted to bone marrow, circulating leukocytes, and subsets of reactive macrophages. Human MPO is not detected in quiescent tissue macrophages, such as brain microglia or liver Kupffer cells (data not shown).
To confirm that the human MPO transgenes were functional, reverse transcription PCR was carried out with cDNA prepared from bone marrow cells. The MPO G transgene was expressed at a high level, equivalent to ␤ 2 -microglobulin used as endogenous control (Table I, panel III, G, row A). Expression of G allele was 7-fold higher than A allele ( Fig. 5A and Table I, panel III, rows A), and the mouse MPO gene was expressed at an intermediate level (Fig. 5A). To demonstrate that the transgene mRNAs coded for functional MPO protein, enzyme activity was monitored by guaiacol oxidation assay. To subtract the contri-  Asterisks, PBMC or M were treated with GMCSF or MCSF, with or without 17␤-estradiol (E2) or rosiglitazone (RS) (10 M) for 24 h. A and B, mean ⌬C T is the difference in PCR threshold cycle (⌬C T ) between MPO and the endogenous control, either human GAPDH or mouse ␤ 2 -microglobulin. A shows the threshold cycle for untreated cells. B shows the threshold cycle for cells treated with GMCSF, MCSF, with or without estrogen. C, ⌬C T RS is the change in threshold cycle as a result of RS treatment (difference between ⌬C T minus or plus RS). D, 2 (⌬C T RS) is the -fold change in MPO cDNA levels because of RS. Boxes indicate conditions where RS had significant effect on MPO expression. P value is for the RS-induced change in threshold cycle (C T ) values. E, -fold transcriptional advantage for GG or GA genotypes, or G or A genotype in transgenic mouse macrophages. bution of native mouse MPO enzyme, the G and A transgenics were crossed onto the MPO-knockout background (57,58). Consistent with a previous report (57), bone marrow cells from the MPO-knockout (KO) strain lacked enzyme activity (Fig. 5B). Bone marrow cells from the MPO G transgenic on the MPO KO background exhibited 3-fold higher enzyme activity than the MPO A transgenic, whereas wild-type C57BL6/J bone marrow cells had intermediate enzyme activity. These findings confirm that the human MPO G and A transgenes produce functional mRNA and protein, and that the G transgene is higher expressing than the A transgene.
As an added assay, Western analysis was carried out on bone marrow cells from MPO G and A transgenics crossed onto the mouse MPO knockout background. The polyclonal antibodies to human MPO (Biodesign) detect the 60-kDa subunit for both the G and A transgene. Higher levels of MPO protein were detected in the MPO G transgenic mouse than the A transgenic mouse (Fig. 5C).

MPO G Transgene, but Not MPO A, Is Up-regulated by Rosiglitazone in Bone Marrow
Cells-In human peripheral blood mononuclear cells untreated with growth factors, the response of the MPO gene to rosiglitazone was variable, in some cases positive and in other cases negative ( Fig. 2A), possibly the result of variation in cytokine or hormone levels in the donors. The MPO transgenics provide the advantage of less genetic and environmental variation than in human subjects. We tested the effects of PPAR␥ ligands on untreated bone marrow cells from the MPO G and A transgenics, along with the parental C57BL6/J strain. The cells were incubated with rosiglitazone for 24 h in medium with 10% FCS, prior to isolation of mRNA. MPO G expression was up-regulated by RS, suggesting that mouse bone marrow cells resemble MCSF-treated cells in their response to PPAR␥ ligands. RS had no effect on MPO A expression (Fig. 6A), again similar to findings for AA genotype in MCSF treated human cells (Fig. 3). The mouse MPO gene, which lacks the primate-specific Alu with PPAR␥ binding site, did not respond to RS.

The G and A Transgenes Are Down-regulated by PPAR␥ Ligands in GMCSF-treated Bone Marrow Cells, Whereas G, but
Not A, Is Up-regulated in MCSF-Bone marrow cells were treated with GMCSF or MCSF for 24 h, followed by 24 h with varying concentrations of RS or PG. In the representative experiment shown in Fig. 6B, MPO G expression decreased 30fold in GMCSF, and increased 20 -50-fold in MCSF. Estrogen blocked the up-regulation in MCSF, but had relatively little effect (0 -2-fold) on down-regulation of G allele in GMCSF. The A allele was down-regulated 30-fold in GMCSF, and this effect was blocked by 17␤-estradiol. The A allele was not affected by RS or PG in MCSF-treated cells. The mouse MPO gene was completely unresponsive to RS or PG in either GMCSF-or MCSF-treated cells.
The Transgenic A Allele Is Unresponsive to PPAR␥ Ligands in Macrophages-To obtain transgenic macrophages, bone marrow cells were cultured in GMCSF or MCSF for 7 days. With RS, MPO G was down-regulated 50 -60-fold in GM-M and up-regulated 12-55-fold in MC-M (Fig. 6C). 17␤-Estradiol reduced (0 -2-fold) but did not block the down-regulation of G allele in GMCSF, yet completely blocked the up-regulation in MCSF. In macrophages, the MPO A allele was no longer downregulated by RS or PG in GMCSF, and remained unresponsive in MCSF. Thus, the MPO A allele lost ability to respond to PPAR␥ ligands during differentiation to GM-macrophages. The mouse MPO gene remained unresponsive to PPAR␥ ligands in either GM-or MC-macrophages. Table I, part III, shows the mean ⌬C T for transgenic MPO G and A macrophages, treated with rosiglitazone, in the presence or absence of estrogen. In GM-M, the 48-fold suppression of G allele (p ϭ 0.0001) was reduced to 19.7-fold by estrogen (columns 2 and 4). In MCSF-M, the 38-fold up-regulation of G allele (p ϭ 0.0001) was completely blocked by estrogen (columns 6 and 8). The A allele failed to respond to RS in either GM-M or MC-M (row D).
In most circumstances, the transgenic G allele was higher expressing than A allele. The G allele was 7.5-fold higher expressing in untreated bone marrow cells (⌬C T 0.2 versus 3.1) (panel III, rows A), and 3.7-9-fold higher expressing in GM-M and MC-M, respectively (III, row E, column 1 and 5). However, as in human cells, in GM-M treated with RS and estrogen, the A allele was 37-fold higher expressing than G (column 4), because down-regulation of G allele, but not A allele, by RS. Conversely, in MC-M, G allele was 294-fold higher expressing than A allele (column 6), because of strong up-regulation of G allele, but not A allele.
Charcoal Stripping of Fetal Calf Serum Permits PPAR␥ to Regulate MPO A Allele; Replenishing with 17␤-Estradiol Reinstates the Block-We hypothesized that the inability of A allele to respond to PPAR␥ ligands in MCSF macrophages was because of trace amounts of estrogen present in 10% FCS, thus enabling transport of ER to the nucleus, where it can bind the preferred site at Ϫ463A. To test this possibility, fetal calf serum was treated with activated charcoal to deplete estrogen and other agents. The use of charcoal-filtered FCS (CF-FCS) allowed RS to strongly up-regulate MPO A in MCSF-treated bone marrow cells (20-fold), whereas unextracted 10% FCS blocked RS-mediated induction (Fig. 7A). To confirm that this rescue was the result of removal of trace amounts of estrogen, we added back 17␤-estradiol (10 Ϫ7 M) to the charcoal-filtered FCS, and thereby reinstated the complete block of RS-induced up-regulation.
Similar findings were obtained in transgenic macrophages. In 10% FCS, the A allele was resistant to RS in GM-M as well as MC-M. Charcoal stripping of the FCS restored the ability of RS to up-regulate MPO A 20-fold in MCSF, and downregulate 20-fold in GMCSF (Fig. 7A). Replenishing with 17␤estradiol reinstated the block to PPAR␥ ligand effects.
Charcoal-filtered FCS enhanced RS effects on the MPO G allele; down-regulation by RS in GMCSF bone marrow cells increased from 15-to 30-fold, whereas the up-regulation in MCSF bone marrow increased from 30-to 50-fold (Fig. 7B). In GM-M, down-regulation similarly increased from 30-to 50fold, and there was no change in MCSF macrophages. This indicates that trace estrogen levels in FCS blunts (2-fold) RS effects on the G allele, consistent with weaker binding by ER␣ to Ϫ463G.
Charcoal filtration of FCS had similar effects in human macrophages (data not shown). In CF-FCS with 10 M RS, AA genotype was down-regulated 18-fold in GM-M, and up-regulated 34-fold in MC-M. In contrast, in regular FCS, AA genotype had not been affected by RS in MCSF-treated cells (Fig.  3). GA genotype was down-regulated 14-fold by 10 M RS in regular FCS, and 32-fold in CF-FCS. In MCSF macrophages, GA was up-regulated 12-fold in regular FCS, and 28-fold in CF-FCS. The readdition of 17␤-estradiol completely blocked both effects. It is worth noting that readdition of estradiol to CF-FCS had no significant effect on MPO expression in the absence of PPAR␥ ligands. Estrogen receptor appears to bind, but has little direct impact on MPO expression, except to block effects of PPAR␥.
Immunodetection of MPO in Macrophages Treated with Rosiglitazone-Rosiglitazone significantly increases MPO mRNA levels in MC-M. To determine whether this results in detectable increases in MPO protein in RS-treated macrophages, we carried out immunocytochemical analysis. Human GMCSF and MCSF macrophages were incubated with polyclonal antibodies to human MPO (Biodesign, 1:1000), and monoclonal antibodies to the macrophage marker CD68 (R&D), followed by fluorescent secondary antibodies, and analyzed by confocal microscopy. The cells were uniformly of macrophage morphology, and 97% were immunopositive for the macrophage marker CD68 (Fig. 8, C, F, I, and L). MPO was detected in MCSF macrophages (B), with a peripheral staining pattern resembling that of CD68 (A). Rosiglitazone treatment increased the levels of MPO in apparent vesicles in more central regions (E, D).
In GMCSF macrophages, MPO immunostaining was abundant in the cytoplasm and perinuclear regions (H) and decreased in rosiglitazone treated macrophages (K). MPO was also detected by nonfluorescent means in MC-M treated with RS, using diaminobenzidine (DAB) as substrate for peroxidase conjugated secondary antibodies (N). MPO appears to be in vesicles at higher magnification (O).
The presence of immunodetectable MPO in macrophages untreated with RS is consistent with a previous study showing that culture of monocytes in GMCSF for 7 days prevents the loss of stored vesicular MPO protein (21). Findings here suggest that RS treatment during the final 24 h increases MPO immunostaining in MC-M, and decreases MPO immunostaining in GM-M, consistent with the observed changes in mRNA levels.
Immunostaining of Human MPO in Early Atherosclerotic Lesions in the LDL Receptor Knockout Model-These findings indicate that the human MPO transgenes are regulated by PPAR␥ ligands in primary macrophages in culture. To obtain evidence that these transgenes are expressed in vivo in tissue macrophages, we immunostained a section of the aortic valve region from MPO G transgenics crossed onto the LDL receptor null background, and fed a high fat diet. These LDL null mice are a model for atherosclerosis and develop fatty streak lesions containing foam cell macrophages, analogous to early human atherosclerotic lesions. A previous study found no mouse MPO in aortic valve lesions in the LDL receptor null model (57), suggesting that this mouse model was unlike human atherosclerosis as regards MPO expression. Findings here suggest that the absence of mouse MPO in vascular lesions may be the result of the lack of response of the mouse MPO gene to PPAR␥. Immunostaining of the MPO G/LDL receptor null mice showed that human MPO is present in lesions in aortic valve lesions (Fig. 9A), colocalizing with PPAR␥ (Fig. 9〉). This observation provides evidence that the human MPO G transgene is appropriately expressed in vivo in tissue macrophages at sites containing PPAR␥. DISCUSSION Results reported here show that PPAR␥ ligands strongly regulate MPO gene expression. An unusual aspect of this regulation is the opposite effects of PPAR␥ ligands in macrophages cultured in GMCSF versus MCSF. MPO expression is markedly up-regulated in MCSF-macrophages, and down-regulated in GMCSF macrophages. As evidence that these effects are me- diated by the receptor PPAR␥, the selective PPAR␥ antagonist, GW9662 blocks both the positive and negative effects on MPO expression. This regulation is not specific to particular ligands, having been obtained with four distinct PPAR␥ ligands of the TZD class, as well as the natural prostanoid 15d-PGJ 2 .
A second intriguing aspect to this regulation is the influence of estrogen and the Ϫ463GA polymorphism. This polymorphism is within an estrogen receptor binding site in hexamer 1/2 of the AluRRE. Estrogen receptor ␣ binds to hexamer 1/2 in both the G and A promoters, binding more effectively to Ϫ463A in more stringent conditions. This ER␣ site is immediately adjacent to the PPAR␥ binding site in hexamers 3/4. Addition of 17␤-estradiol has strong impact on regulation by PPAR␥, blocking up-regulation of both alleles in MC-M, and blocking down-regulation of A allele, but not G allele, in GM-M.
One interpretation of these findings is illustrated in the schematic model in Fig. 10. This model suggests that estrogen receptor and PPAR␥ compete for binding to the AluRRE. Estrogen is required for translocation of ER from the cytoplasm to the nucleus, where ER can bind to the AluRRE, thereby blocking binding by PPAR␥. Because ER␣ binds with greater avidity to Ϫ463A than G (Ref. 38 and Fig. 1), the A allele appears to be more readily blocked in low levels of estrogen or ER. In this model, the G allele is down-regulated by PPAR␥ in GM-M, and up-regulated in MC-M (Fig. 10A). Estrogen blocks the up-regulation in MCSF, but not the down-regulation in GMCSF, perhaps because of the lower ER␣ levels in GM-M.
The A allele is down-regulated by PPAR␥ in GM-PBMC, and this is blocked by estrogen/ER (Fig. 10B). In MC-PBMC, supplemental estradiol is not required; trace estrogen levels in 10% FCS appear to be sufficient for ER translocation and blockage of PPAR␥ binding. In macrophages, trace estrogen levels in 10% FCS appear sufficient to block PPAR␥ binding to the A allele in both GM-M and MC-M (Fig. 10C). Charcoal filtering of FCS depletes estrogen, preventing ER translocation and binding, such that PPAR␥ is able to down-regulate the A allele in GMCSF and up-regulate in MCSF (Fig. 10D). Both effects are blocked by the readdition of 17␤-estradiol. Accordingly, charcoal stripping of serum resulted in comparable regulation of the G and A alleles by PPAR␥; the only difference remaining was that the down-regulation of G allele was not blocked by estrogen/ER in GM-M (Fig. 10A). The differential expression of the G and A alleles in physiologically low levels of estrogen (10% FCS) is likely to be biologically significant, as suggested by the gender differences in Ϫ463GA genotype association with disease risk in some studies. Indeed, the mutation at Ϫ463A may have been evolutionarily selected because of its ability to dampen effects of PPAR␥ on MPO expression.
Three lines of evidence suggest the AluRRE mediates the effects of PPAR␥ ligands on MPO gene expression. First, there is the PPAR␥ binding site in the AluRRE (Fig. 1). Second, the Ϫ463GA polymorphism in the AluRRE strongly affects PPAR␥ regulation, probably through competitive binding by estrogen receptor (Figs. 3, 6, and 7). Third, the mouse MPO gene lacks . ER␣ is unable to bind the low avidity Ϫ463G site; thus, the addition of estrogen (plus 17␤-estradiol) does not block PPAR␥ binding or MPO down-regulation. In MCSF-treated cells, PPAR␥ binds and up-regulates MPO G (arrow up). With added 17␤-estradiol, ER␣ is able to enter the nucleus and bind the G promoter, blocking PPAR␥ binding and MPO up-regulation, such that basal level expression (horizontal arrow) is maintained. The ability of 17␤-estradiol to block PPAR␥ effects in MCSF, but not in GMCSF, may have to do with the severalfold higher levels of ER␣ expression in MCSF cells. B, MPO A promoter. In GMCSF-treated human PBMC or mouse bone marrow cells, PPAR␥ binds and down-regulates the A allele. With added 17␤-estradiol, ER␣ is able to enter the nucleus, bind and block PPAR␥ binding, thereby blocking MPO down-regulation, maintaining basal MPO expression levels. In MCSF-treated cells, supplemental 17␤-estradiol was not required to block PPAR␥. The low levels of estrogen in 10% FCS were sufficient, suggesting the higher levels of ER␣ in MCSF, coupled with the higher avidity Ϫ463A site, allows sufficient ER␣ to transport to the nucleus, blocking PPAR␥ binding. C, MPO A promoter in macrophages. The ability of 17␤-estradiol to block PPAR␥ effects was enhanced in M. The low levels of estrogen in 10% FCS were sufficient to enable ER␣ to block PPAR␥ in GMCSF as well as MCSF. Supplemental estrogen was not required. D, CF-FCS provided evidence that estrogen is responsible for the block of PPAR␥ effects on the A allele. Charcoal stripping of serum restored the ability of PPAR␥ ligands to down-regulate the A allele in GMCSF and up-regulate in MCSF, such that the MPO A response is like that of MPO G. The readdition of estrogen reinstated the complete block of PPAR␥ effects on the A allele. the primate-specific AluRRE, and is completely unresponsive to PPAR␥ ligands (Fig. 6).
The mechanisms underlying the opposite effects of PPAR␥ ligands in GMCSF versus MCSF macrophages are unknown. Both factors have been used in various studies to induce cultured monocytes to differentiate to macrophages (59,60,61). Interestingly, catalase is expressed at high levels in GM-M and is induced by H 2 O 2 whereas in MC-M, catalase is expressed at low levels and is not inducible (62). Catalase consumes H 2 O 2 , thereby removing the substrate for MPO. It seems consistent that, in GM-M, MPO is down-regulated whereas catalase is up-regulated, consistent with an antioxidant phenotype. Conversely, MPO is up-regulated and catalase downregulated in MC-M, consistent with a pro-oxidant phenotype.
The mechanism by which MPO is up-regulated by PPAR␥ in MCSF and down-regulated in GMCSF is likely to involve the coactivators and corepressors that modulate nuclear receptor activities (63). Ligand binding to PPAR␥ or other nuclear receptors induces conformational changes resulting in the reorientation of a helical motif termed the AF2 domain, reorienting this motif to the ligand binding pocket (64). This conformational rearrangement promotes the association of coactivators and the dissociation of corepressors. A number of receptor coactivators have been identified including the p160/SRC family, as well as CREB-binding protein (CBP), PPAR␥-binding protein (PBP), PPAR-interacting protein (PRIP), PPAR␥ coactivator 1 (PGC-1), and others (63). Coactivators have associated histone acetyltransferase activity important for chromatin remodeling and interactions with the basal transcriptional machinery. In the absence of ligands, nuclear receptors assume a conformation that allows interaction with corepressors, such as nuclear receptor corepressor (N-CoR), silencing mediator for retinoid and thyroid receptor (SMRT), or SHARP (65). These corepressors recruit a histone deacetylase complex that modifies chromatin structure to suppress transcription. The opposing effects of PPAR␥ ligands may reflect the dominance of coactivators in MCSF and the dominance of corepressors in GMCSF macrophages.
The observation that PPAR␥ ligands increase MPO expression in MC-M is consistent with established mechanisms in which ligand binds PPAR␥, causing dissociation of corepressor complexes, and recruitment of coactivator complexes, enabling transcription. However, the repression of MPO in GM-M is not consistent with this model because this repression is PPAR␥ ligand dependent. Another means by which PPAR␥ can repress gene expression is termed transrepression, in which ligand-bound PPAR␥ competitively sequesters limiting amounts of coactivators SRC-1 and CBP (25), or transcription factors NFB, AP1, and STAT (28). One observation seems inconsistent with transrepression; estrogen blocked the downregulation by PPAR␥ suggesting that ER binding physically blocks PPAR␥ binding, and if so, PPAR␥ binding to the MPO AluRRE is necessary for repression. This may represent a novel mechanism for down-regulation through PPAR␥, perhaps using a corepressor that associates with PPAR␥ in the specific context of the AluRRE. It is important to note that the downregulation of MPO by PPAR␥ in GM-M was promoter specific; CD36 and LXR were up-regulated by PPAR␥ in both GM-M and MC-M.
For disease states in which MPO is implicated, such as CAD and AD, the relative presence of GMCSF or MCSF could significantly influence expression levels. Findings here suggest that MPO expression will be higher in monocyte/macrophages exposed to MCSF, and lower in GMCSF. Both MCSF (66) and GMCSF (67) are elevated in cerebrospinal fluid in Alzheimer's patients. MCSF receptor is elevated in microglia surrounding amyloid deposits in AD (68) and in mouse models for AD expressing human amyloid precursor protein (69). Both GMCSF and MCSF are present in human atheromata (21,70,71). Atherosclerotic vessels are associated with higher levels of MCSF mRNA than nonatherosclerotic vessels (71), whereas MPO-positive macrophages have been detected in GMCSFpositive atheroma (21). Considering the use of TZDs in diabetic patients with high risk for CAD, our findings suggest that TZDs should reduce MPO expression if GMCSF conditions predominate, and increase MPO expression in MCSF conditions. Induction of MPO by PPAR␥ could generate a positive feedback loop promoting accumulation of oxidized lipids in atherosclerotic plaques. Release of MPO by foam cell macrophages results in oxidation of LDL, producing ligands for PPAR␥, such as hexadecyl azelaoyl phosphatidylcholine (37). Ligand-bound PPAR␥ is able to induce MPO expression as well as CD36 scavenger receptor that would promote further uptake of LDL for oxidation by MPO, creating lipid-laden foam cell macrophages at the core of plaques.
An earlier report found that, in serum-free medium, GMCSF (but not MCSF) maintains levels of MPO protein, but not MPO mRNA, in monocyte-derived macrophages (21). Our culture protocol differs significantly in that we use 10% FCS in addition to 10% autologous human serum. Consistent with the earlier report, we find that, in the absence of FCS, MPO mRNA levels drop precipitously after 24 h in culture. In the absence of FCS, MPO mRNA levels were 120-fold lower in GM-M and 1000-fold lower in MC-M (data not shown).
Findings here confirm that the human MPO gene can be expressed in macrophages, with expression increasing markedly in response to MCSF and PPAR␥ ligands. Nonetheless, levels of MPO mRNA expression are very low in macrophages in the absence of PPAR␥ ligands, and thus these findings are consistent with the general consensus that MPO expression is low to undetectable in macrophages (19,21). MPO expression was ϳ400-fold lower in macrophages than in HL60 cells; therefore, methods less sensitive than real-time PCR are unlikely to detect MPO mRNA in quiescent macrophages. Expression in macrophages increases 20 -40-fold in the presence of MCSF and PPAR␥ ligands. This amount of MPO expression is considerably lower than in bone marrow precursors, yet is likely to be biologically significant, especially at chronic inflammatory sites such as atherosclerotic lesions or Alzheimer's plaques.
Consistent with our earlier findings (35,36), the Ϫ463G allele is in most circumstances higher expressing than the A allele. The GG genotype was 5-7-fold higher expressing than GA genotypes in macrophages and 5-fold higher expressing than GA in untreated PBMC (Table I). However, in PBMC treated for 48 h with GMCSF or MCSF, expression of A allele increases relative to G allele, such that GA genotype was higher expressing than GG genotype. This raises the possibility that GA/AA genotypes may be higher expressing than GG in circulating monocytes, depending on individual variation in serum levels of GMCSF or MCSF.
The A allele was significantly higher expressed than G allele in the presence of GMCSF, PPAR␥ ligands, and estrogen. These conditions markedly down-regulate G allele expression, whereas the A allele maintains basal expression levels. GA expression was 4.6-fold higher than GG in GM-M, and 52-fold higher in GMCSF PBMC (Table I). Conversely, the GG genotype had strong transcriptional advantage in MCSF, PPAR␥ ligands, and estrogen. These conditions markedly induce G allele expression, but do not affect the A allele, which again maintains basal expression. Thus, in disease states in which MPO is implicated, estrogen replacement therapy could be deleterious for A allele carriers in GMCSF conditions, by preventing the suppression of MPO A by PPAR␥. Conversely, estrogen replacement therapy may be beneficial in MCSF conditions, by suppressing the PPAR␥-mediated upsurge in MPO expression for both alleles.
The ability of estrogen to block PPAR␥ effects, especially on the A allele, may underlie gender differences in association of Ϫ463GA genotype with disease risk (20,38,40,41,44,46). The relative levels of estrogen, GMCSF, or MCSF may vary with age, gender, or genetic subpopulation, and could explain why the G allele is associated with AD risk in some studies (20,39) and the A allele in others (38,40). Because of conversion of testosterone to estrogen by aromatase, aged males may have higher estrogen levels than postmenopausal females. An earlier study found the A allele to be a male risk factor for AD in an aged Finnish population (38). Moreover, AA males were depleted from Finnish aged controls, suggesting selective mortality because of a common disease such as atherosclerosis.
A more recent study of Finnish males found the A allele (GA/AA genotypes) to be associated with increased severity of atherosclerosis, correlating with larger aortic lesions than GG genotypes (16). Conversely, an ultrasonic study of Finnish postmenopausal females (15) found that hormone replacement therapy decreased progression of atherosclerosis in abdominal aorta in GG, but not GA/AA genotypes.
The MPO G and A transgenic mice provided key evidence in this study, confirming observations obtained with the relatively rare AA genotype. The complex regulation of the G and A alleles by GMCSF, MCSF, estrogen, and PPAR␥ ligands in human cells was reproduced in the transgenic macrophages. The mouse MPO gene was found to be unresponsive to PPAR␥ ligands, potentially explaining the previously reported absence of MPO in lesions in the LDL receptor-deficient murine model for atherosclerosis. The inability of mouse MPO to be regulated by PPAR␥ raises questions as to the validity of mouse models for this and other diseases involving MPO. Crossing the human MPO transgenics onto established murine models for atherosclerosis and Alzheimer's disease should help to humanize these models as regards MPO expression.