Macrophage Mitochondrial Oxidative Stress Promotes Atherosclerosis and NF-κB-Mediated Inflammation in Macrophages

Abstract

Rationale

Mitochondrial oxidative stress (mitoOS) has been shown to correlate with the progression of human atherosclerosis. However, definitive cell-type specific causation studies in vivo are lacking, and the molecular mechanisms of potential pro-atherogenic effects remain to be determined.

Objective

To assess the importance of macrophage mitoOS in atherogenesis and explore the underlying molecular mechanisms.

Methods & Results

We first validated Western-type diet-fed Ldlr^-/- mice as a model of human mitoOS-atherosclerosis association by showing that a marker of mitoOS in lesional macrophages, non-nuclear oxidative DNA damage, correlates with aortic root lesion development. To investigate the importance of macrophage-mitoOS, we used a genetic engineering strategy in which the OS suppressor catalase was ectopically expressed in mitochondria (mCAT) in macrophages. MitoOS in lesional macrophages was successfully suppressed in these mice, and this led to a significant reduction in aortic root lesional area. The mCAT lesions had less monocyte-derived cells, less Ly6c^hi monocyte infiltration into lesions, and lower levels of the monocyte chemotactic protein-1 (MCP-1). The decrease in lesional MCP-1 was associated with suppression of other markers of inflammation and with decreased phosphorylation of RelA (NF-κB p65), indicating decreased activation of the pro-inflammatory NF-κB pathway. Using models of mitoOS in cultured macrophages, we showed that mCAT suppressed MCP-1 expression by decreasing activation of the Iκ-kinase-RelA NF-κB pathway.

Conclusions

MitoOS in lesional macrophages amplifies atherosclerotic lesion development by promoting NF-κB-mediated entry of monocytes and other inflammatory processes. In view of the mitoOS-atherosclerosis link in human atheromata, these findings reveal a potentially new therapeutic target to prevent the progression of atherosclerosis.

Introduction

Oxidative phosphorylation in the mitochondria produces limited, physiologic levels of superoxide, most of which is converted to hydrogen peroxide by superoxide dismutase (SOD).¹ While this process is adaptive under normal conditions, excessive mitochondrial oxidative stress (mitoOS) has been correlated with a number of diseases, including atherosclerotic vascular disease in humans.^{2, 3} However, definitive evidence of causation and cell-specific pro-atherogenic mechanisms of mitoOS require further investigation.^{4, 5} For example, while several important studies demonstrated that genetic targeting of Mn-SOD or uncoupling protein-2 increases mitoOS and worsens atherosclerosis,^{6, 7} the role of endogenous mitoOS is not addressed by this experimental strategy. Another elegant study showed that endothelial-targeted overexpression of thioredoxin 2, an anti-oxidant enzyme that has been identified in mitochondria, increased total anti-oxidant activity, lowered ROS, promoted NO formation, and improved endothelial function.³ When crossed onto the Apoe^-/- background, thoracic aortic rings showed improved relaxation, and atherosclerotic lesion size was decreased. Whether the atherosclerosis endpoint was mechanistically related to lesional endothelial mitoOS, the aortic ring data, or other possible mechanisms remains to be determined in this model.

The high level of interest in this topic, the human relevance, and the potential therapeutic implications prompted us to explore causation and mechanism with a focus on the key inflammatory cell type in atherosclerosis, the macrophage. For this purpose, we used a recently described model, the mitochondrial catalase (mCAT) transgenic mouse, that decreases mitoOS in vivo.⁸ Normally, glutathione perioxidase is the endogenous mitochondrial enzyme that catalyzes the reduction of H₂O₂ and prevents its conversion into the most detrimental ROS hydroxyl nitrites. Catalase can carry out this role in peroxisomes, where it is exclusively located. The mCAT transgenic mouse expresses human catalase with a mitochondrial matrix-targeting motif, which quenches mitoOS and protects against mitoOS-induced damage. To focus on myeloid-derived cells in atherosclerosis, we used two strategies: transplantation of mCAT transgenic bone marrow cells into atheroprone Ldlr^-/- mice and crossing Ldlr^-/- mice with an mCAT^fl/-LysMCre model that expresses mCAT only in lysozyme M-expressing cells, notably differentiated macrophages. Both models demonstrated evidence of decreased mitoOS in lesional macrophages, decreased atherosclerosis, and suppression of inflammatory monocyte infiltration. In vitro and in vivo mechanistic studies suggest that macrophage mitoOS promotes monocyte chemotactic protein-1 (MCP-1) production through enhancing the Iκ-kinase (IKK)-RelA NF-κB pathway.

Methods

Animals and diets

C57BL/6J (000664) and Ldlr^-/- (002381) mice on the C57BL/6J background were purchased from Jackson Laboratory. mCAT transgenic and floxed mice were generated as described previously ^{8, 9} and were backcrossed >10 times onto the C57BL/6J background. For the atherosclerosis study, mCAT transgenic and age/gender-matched littermates were used as donors. Ldlr^-/- male mice, at 14 weeks of age and 6 weeks after the bone marrow transplant, were placed on a Western-type diet (TD88137; Harlan Teklad) for the indicated periods of time.

Atherosclerotic lesion analysis

For morphometric lesion analysis, sections were stained with Harris' hematoxylin and eosin. The total lesion area and necrotic area were quantified as previously described.¹⁰ For immunostaining, specimens were immersed in OCT and 6-μm sections were prepared and placed on glass slides. The sections were fixed and permeabilized with ice-cold acetone for 10 min. Paraffin-embedded specimens were sectioned, de-paraffinized with xylene, and rehydrated in decreasing concentrations of ethanol. Sections were then incubated overnight at 4°C with anti-Mac-3 (BD Clone M3/84, 1:200), anti-8-OHdG (EMD Millipore, AB5830, 1:200), anti-NF-κB P65 p-S536 (Cell Signaling, 3033, 1:40), FITC-labeled smooth muscle cell a-actin antibody (FITC-labeled, clone 1A4, Sigma-Aldrich, 1:1000), or anti-CD11c (PE labeled, clone HL3, BD Biosciences, 1:200) antibody. The sections were then incubated with biotinylated anti-rat, anti-goat IgG, or anti-rabbit secondary antibody (Vector) and then streptavidin-conjugated Alexa 488- or Alexa 594-labeled antibody (Life Technology). Sections were counter-stained with DAPI to identify nuclei before mounting.

Measurement of 8-OHdG in lesional and cultured macrophages

Evidence of mitoOS in lesional macrophages was obtained by assaying oxidative damage of non-nuclear (mitochondrial) DNA. Specifically, cryo-sections were stained sequentially with anti-8-hydroxydeoxyguanosine (8-OHdG) and anti-Mac-3 primary antibodies, biotinylated secondary antibodies (Vector ABC Kit), Alexa 488- and 594-labeled streptavidin, and 4′,6-diamidino-2-phenylindole (DAPI), which was used to measure the total number of cells and to identify nuclei. Sections were then imaged by confocal microscopy (Nikon A1 confocal microscope with 40× and 100× oil objectives and z-section = 0.10 μm). Data were quantified as both the percentage of total Mac3+ cells and the total number of Mac3+ cells showing 8-OHdG staining that did not overlap with DAPI, i.e., as an indicator of exposure of mitochondrial DNA to oxidative stress. To assay mitochondrial 8-OHdG in cultured macrophages, sections were fixed and permeabilized with pre-chilled acetone on ice for 10 min, stained with anti-8-OHdG and anti-ATP synthase 5α (Abcam ab110273 1:200) primary antibodies at 4°C overnight, followed by anti-goat-Alexa488 and anti-mouse-Alexa647 secondary antibodies. The sections were then counter-stained with DAPI and visualized by confocal microscopy as above.

Measurement of mitochondrial and cytosolic ROS in cultured macrophages

Peritoneal macrophages from adult female C57BL/6J mice and mCAT transgenic mice were harvested 3 days after i.p. injection of concanavalin A or four days after i.p. injection of methyl-BSA (mBSA) in mice previously immunized with mBSA.¹¹ All macrophages were grown in full medium containing Dulbecco's Modified Eagle Medium (DMEM; 25 mM glucose, phenol-red free), 10% fetal bovine serum (FBS), 20% L-cell conditioned medium, and 1% penicillin/streptomycin/glutamine solution (GIBCO) on non-tissue culture coated plates. The medium was replaced every 24 h until the cells reached 90% confluence. On the day of the experiment, the cells were pre-incubated with 5 μM of the mitochondrial superoxide indicator MitoSOX at 37°C for 30 min. The cells were then rinsed in warm culture medium, and treatments as described in the figure legends were started 6 h later. At the end of incubation period, cells were dissociated from the petri dish and subjected to flow cytometric analysis (BD Canto II) using the Phycoerythrin (PE) channel. Data were quantified as fold change of medium fluorescent intensity (MFI) compared with baseline. For live cell imaging, cells were stained with Mitotracker Green (100 nM) for 15 min, followed by three washes with warm medium. The sections were then imaged immediately at room temperature using confocal microscopy. For measuring cytosolic ROS, cells were incubated with 2.5 μM CellROX Deep Red (Life Technology) at 37°C for 30 min and then subjected to FACS analysis.

Monocyte infiltration experiment

To track newly recruited monocytes in atherosclerotic lesions, the Ly6c^hi subset of monocytes was labeled with fluorescent beads as described previously.¹² Briefly, 96 h before the end of study, the mice were injected i.v. with 250 μl clodronate-containing liposomes (http://clodronateliposomes.org/ashwindigital.asp?docid=26) to deplete monocytes. After 48 h, the mice were injected with 250 μl of a 1:4 dilution of 1 μm Fluoresbrite Plain YG microspheres (Polysciences). After another 48 h, the mice were euthanized, and peripheral blood samples were analyzed by FACS to quantify the efficiency of bead labeling of Ly6c^hi monocytes. The heart and aortic tissues were processed as described above. The newly recruited bead-labeled monocytes in atherosclerotic lesions were visualized by fluorescence microscopy and quantified using Image J.

Results

Oxidative DNA damage surrounding mitochondria in lesional macrophages correlates with atherosclerosis lesion progression in western diet-fed Ldlr^-/- mice

Oxidative damage to nuclear and mitochondrial DNA (mtDNA) can be assessed by immunostaining for nuclear and non-nuclear 8-OHdG, respectively.^{13, 14} Thus, a non-nuclear 8-OHdG immunostaining is observed when mitochondria are exposed to excessive oxidative stress, referred to here as “mitoOS.” To illustrate this assay, cultured macrophages were subjected to various treatments and then immunostained for 8-OHdG (green) and the mitochondrial marker ATP5α (red) (Online Figure 1A). For some of the treatments, the cells were assayed by flow cytometry for mitoOS using MitoSOX and for general cellular reactive oxygen species (ROS) using CellROX. Compared with vehicle control, H₂O₂ treatment, which causes general oxidative stress in cells, yielded a positive 8-OHdG signal, some of which overlapped with the mitochondrial marker (yellow staining in cytoplasm) and some of which was juxtaposed with DAPI (blue)-stained nuclei. Short-term treatment with phorbol myristate acetate (PMA) activates NADPH oxidase, not mitoOS, and we saw no 8-OHdG-mitochondrial co-localization despite robust activation of the CellROX signal. As will be described in later sections, oxidized LDL (oxLDL) is an athero-relevant inducer of oxidative stress in macrophages, and we found that it also activates mitoOS, i.e., there is ample evidence of punctate yellow staining in the cytoplasm, indicative of 8-OHdG-mitochondrial co-localization. Lipopolysaccharide (LPS) also activates mitoOS (below), and here again 8-OHdG-mitochondrial co-localization was seen. These data provide validation for the use of non-nuclear 8-OHdG, which reflects mitochondrial DNA oxidative damage, as a marker of mitoOS.

We next assessed nuclear and mitochondrial oxidative DNA damage in atherosclerotic lesional macrophages in aortic root lesions from 8-wk Western diet (WD)-fed Ldlr^-/- mice. Sections were immunostained with the macrophage marker anti-Mac3, the nuclear marker DAPI, and anti-8-OHdG, or the respective isotype-matched IgGs as negative controls, as illustrated by the images in Online Figure 1B. For quantification, lesional sections from multiple mice were viewed and quantified by confocal fluorescence microscopy to look for punctate 8-OHdG staining that was either cytoplasmic or nuclear, i.e., similar to the pattern of non-nuclear or nuclear 8-OHdG staining in culture macrophages, respectively. We found that lesional macrophages displayed clear evidence of non-nuclear 8-OHdG (Figure 1A). Increasing WD feeding for 12 and 16 wks, which is known to increase aortic root lesion area¹⁵ (data not shown), led to a progressive increase in both the percent and total number of macrophages showing this pattern (Figure 1B). By comparison, the percent of macrophages with nuclear 8-OHdG staining showed similar levels at 8 and 12 wks of WD feeding and then an increase above that level at 16 wks, while the total number of nuclear 8-OHdG macrophages continuously increased as lesions progressed (Online Figure IC). These data validate the use of the WD-fed Ldlr^-/- model to further study a known feature of human atherosclerotic lesions,⁶ namely, a progressive increase in lesional mitoOS.

Figure 1. Oxidative damage to mitochondrial DNA in lesional macrophages correlates with atherosclerosis progression in Ldlr^-/- mice.

(A) Aortic root lesions of 8-wk WD-fed Ldlr^-/- mice were subjected to immunofluorescence staining using anti-8-oxyhydrodioxy guanosine (8-OHdG), a marker of DNA oxidative damage (green). Macrophages were stained using anti-Mac3 (red), and nuclei were stained with DAPI (blue). The upper row of images shows a representative lesional section at low magnification, with the intima outlined with the dotted line. Bar, 10 μm. The two boxed areas in the fourth low-magnification image are shown at higher magnification in the lower two rows of images. In the merged image, when the green 8-OHdG signal is nuclear, it retains its green fluorescence and is juxtaposed with the blue nuclei (arrows), whereas when it is non-nuclear, the green fluorescence “merges” with the red cytoplasmic fluorescence (Mac3) and appears as yellow dots (arrowhead). Bars, 10 μm for the first row of images and 1 μm for the bottom two rows of images. (B) Aortic root lesions from 8-, 12-, and 16-wk WD-fed Ldlr^-/- mice were quantified for the percentage of non-nuclear 8-OHdG⁺ Mac3⁺ cells among all lesional Mac3⁺ macrophages and total number of non-nuclear 8-OHdG⁺ Mac3⁺ cells per section; the number of mice examined for each of the three WD durations were 4, 5, and 5, respectively (*P<0.05 vs. 8-wk group; ^#P<0.05 vs. 12-wk group; n = 4, 5, and 5 mice for 8-wk, 12-wk, and 16-wl lesions, respectively).

Suppression of MitoOS in myeloid cells protects against atherosclerosis

To test the functional importance of mitoOS in lesional myeloid-derived cells, we transplanted bone marrow from mCAT transgenic or littermate control mice into Ldlr^-/- recipients. Six weeks after transplantation, the mice were placed on a high-fat Western-type diet (WD) for 8 weeks. Bone marrow-derived macrophages (BMDMs) from the mice showed expression of human catalase mRNA only in the mCAT group, and immunoblot assay of total catalase showed a higher level in the mCAT vs. control macrophages (Figure 2A). Only macrophages from the mCAT mice showed co-localization of catalase with the mitochondrial marker ATP synthase 5α (ATP5α) (Figure 2B). Using mRNA captured from aortic root lesions by laser-capture microdissection (LCM), we found that human catalase mRNA was expressed only in the mCAT group (Figure 2C), whereas lesional murine catalase mRNA did not differ significantly between the two groups (data not shown).

Figure 2. Suppression of myeloid cell mitoOS protects against early atherogenesis in mCAT transgenic → Ldlr^-/- chimeric mice.

mCAT transgenic (mCAT) → Ldlr^-/- and littermate control → Ldlr-/- chimeric mice were fed the WD for 8 wks, and bone marrow-derived macrophages (BMDMs) and aortic root lesions were analyzed as below. (A) Relative expression of human catalase (Hucat) mRNA and protein in BMDMs (n = 4 control vs. 5 mCAT mice; *P<0.05). (B) Representative confocal microscopic images of catalase and ATP5α (mitochondria marker) in BMDMs. The 4^th column of images is a higher magnification of the boxed sections in the 3^rd column of images. Bars, 5 μm for the first 3 column of images and 2 μm for the last column. (C) Relative human catalase mRNA in aortic root lesions by laser capture microscopy (LCM). (D) Quantification of nuclear and non-nuclear 8-OHdG-positive Mac3+ macrophages as a percentage of total Mac3+ cells (two graphs on the left) and per section (two graphs on the right) in aortic root lesions (n = 10 control vs. 9 mCAT mice; *P<0.05; N.S., non-significant). (E) Representative H&E-stained aortic root lesions, with the intima marked by dotted lines, and total lesion area quantification (n = 19 mice/group; *P<0.05). Bar, 40 μm. (F) Graphs of nuclear and non-nuclear 8-OHdG vs. lesion area.

We next analyzed non-nuclear 8-OHdG in lesional macrophages and found suppression of this marker in the mCAT group (Figure 2D). In contrast, nuclear 8-OHdG was similar between the two groups. These data support both the usefulness of the non-nuclear 8-OHdG marker in lesions and the overall strategy of the experimental design. Most importantly, aortic root lesion area was, on average, ∼2.5-fold lower in the mCAT group (Figure 2E). The decrease in atherosclerosis in the mCAT group was maintained after 16 wks of WD feeding (Online Figure II). The two groups of mice had similar weights, fasting plasma glucose levels, and plasma lipids and lipoprotein concentrations after 8 wks of WD feeding (Online Figure III). As is usually the case with mouse models of atherosclerosis, the lesion area data showed a wide range of variability, and we took advantage of this spread to test whether there was a correlation between non-nuclear 8-OHdG and lesion area in the combined group of mice (Figure 2F). This analysis revealed a strong positive correlation between these two parameters, whereas there was no correlation between nuclear 8-OHdG and lesion area. Finally, we tested the effect of macrophage mCAT using a non-BMT model. For this purpose, Ldlr^-/- mice were crossed with a cre-lox model that expresses mCAT in cells expressing lysozyme M, which, in the setting of atherosclerosis, are mostly macrophages.¹⁶ Thus, 8-wk-WD-fed mCAT^fl/-LysMCre^+/-Ldlr^-/- mice were compared with control mCAT ^fl/-Ldlr^-/- mice. The two groups of mice did not differ with respect to body weight, plasma lipids, or fasting glucose (data not shown). The atherosclerotic lesion data were very similar to the those with the BMT model: the lesions from the LysMCre mice contained macrophages having lower levels of non-nuclear but not nuclear 8-OHdG, and the lesions were smaller in a manner that correlated strongly with lesional macrophage non-nuclear 8-OHdG (Figure 3). In summary, expression of mitochondria-targeted catalase in myeloid cells lowers a marker of mitoOS in lesional macrophages and, in direct proportion to this parameter, decreases atherosclerotic lesion size.

Figure 3. Suppression of mitoOS in lysozyme M-expressing cells protects against early atherogenesis in Ldlr^-/- mice.

mCAT^fl/-Ldlr^-/- (control) and mCAT^fl/-LysMCre^+/-Ldlr^-/- mice were fed the WD for 8 wks, and aortic root lesions were analyzed as below. (A) Quantification of nuclear and non-nuclear 8-OHdG -positive Mac3⁺ macrophages as a percentage of total Mac3⁺ cells (upper two graphs) and per section (lower two graphs) in the aortic root lesions of the two groups of mice (n = 14 Cre- vs. 10 Cre+ mice; *P<0.05; N.S., non-significant). (B) Representative images of nuclear and non-nuclear 8-OHdG staining in lesional macrophages in the two groups of mice using the same staining procedure as in Figure 1. The first row of images are low-magnification, with the intima outlined by the dotted line; the second row of images is a higher magnification of the boxed areas. As in Figure 1A, nuclear 8-OHdG is depicted by arrows and non-nuclear 8-OHdG by the arrowhead. Bars, 20 μm for the first row of images and 1 μm for the second row. (C) Representative images, with outlined intima, and quantification of total lesion area in the two groups of mice (n = 14 Cre- vs. 10 Cre+ mice; *P<0.05). (D) Graph of nuclear and non-nuclear 8-OHdG vs. lesion area.

Suppression of MitoOS in myeloid cells decreases monocyte infiltration, inflammation, and RelA NF-κB activation in atherosclerotic lesions

To explore the mechanisms of how suppression of myeloid cell-derived mitoOS decreases atherosclerosis, we analyzed the cells and extracellular matrix of aortic root lesions from 8-wk WD-fed mCAT → Ldlr^-/- and wild-type littermate → Ldlr^-/- chimeric mice. At this stage of atherosclerosis, most of the variability in aortic root lesion area among individual mice can be explained by the number of lesional cells (Figure 4A), whereas the extracellular matrix area was very small and not noticeably different between the two groups of lesions (data not shown). In particular, the mCAT-positive lesions had smaller numbers of total cells, Mac3+ cells (macrophages), and CD11c+ cells (cells having properties of dendritic cells¹⁷) (Figure 4B). In contrast, the numbers of lesional smooth muscle cells and CD3+ T cells were similar between the two groups of mice (Online Figure IVA).

Figure 4. Lesional monocyte infiltration and inflammation are decreased in mCAT transgenic → Ldlr^-/- chimeric mice.

mCAT transgenic (mCAT) → Ldlr^-/- and littermate control → Ldlr-/- chimeric mice were fed the WD for 8 wks, and aortic root lesions were analyzed as below. Circulating Ly6^hi monocytes in the mice were labeled with green fluorescent beads in vivo prior to sacrifice (see text and Methods). (A) Graph of lesional cell number vs. lesion area. (B) Quantification of the number of monocyte-derived cells (Mac3⁺ and/or CD11c⁺) in aortic root lesions (n = 10 control vs. 9 mCAT mice/group; *P<0.05). (C) Representative images, with intima outlined and bead-labeled cells depicted by arrows, and quantification of the number of bead-labeled cells in aortic root lesions (n = 11 vs. 10 mice/group). Bar, 40 μm. (D) Relative level of lesional Mcp1, Tnfa, and Inos mRNA in the two groups of mice (n = 5 mice/group; *P<0.05). (E) Representative immunofluorescence images and quantification of nuclear NF-κB RelA p-S536 in lesional myeloid vs. non-myeloid cells (smooth muscle cells and endothelial cells) (n = 14 vs. 10 mice/group; *P<0.05; N.S., non-significant). Bar, 20 μm.

The decrease in myeloid-derived cells in the mCAT-positive lesions could, in theory, be due to increased apoptosis, followed by rapid efferocytosis,¹⁸ or to decreased proliferation. However, TUNEL-positive staining as a marker of apoptosis was barely detectable in these early lesions (Online Figure IVB). The number of Ki67-positive lesional macrophages as a marker of macrophage proliferation was similar between the two groups of mice, and cultured macrophages from mCAT-positive and control mice had similar proliferation rates (Online Figure IVC). Another mechanism could be decreased retention or increased egress of lesional macrophages, but the mRNA level for a key molecule that mediates retention, netrin-1,¹⁹ was not decreased in the mCAT lesions, and the mRNA level for the egress mediator CCR7²⁰ was similar between the two groups (Online Figure VB). Interestingly, there was a marked increase in netirn-1, which might represent a compensatory response that is subservient to the dominant mechanism of lesional myeloid cell decrease described below.

We next turned our attention to the hypothesis that suppression of mitoOS by mCAT decreased blood monocyte infiltration into lesions, with an emphasis on the Ly6c^hi subpopulation of monocytes, which contributes to lesion progression.²¹ Monocyte infiltration into lesions involves both endothelial cell monocyte-adhesion molecules and chemokine-mediated monocyte migration (chemokinesis). In consideration of the former mechanism, we assayed the expression levels of intercellular adhesion molecule 1 (ICAM-1) and vascular cell adhesion molecule 1 (VCAM-1) on lesional endothelium, but the levels were not decreased in the lesions of mCAT mice (Online Figure IVD). To test whether chemokinesis was lower in mCAT mice, mCAT transgenic → Ldlr^-/- and wild-type littermate → Ldlr^-/- chimeric mice were fed the WD for 6 wk and then injected with fluorescent beads. In particular, we used a protocol, pioneered by Randolph and colleagues in which the injected beads preferentially label Ly6c^hi monocytes.¹² Lesions were than analyzed for labeled cells 48 h later (Figure 4C). Total lesion area significantly correlated with the number of beads (Figure 4C, left graph), consistent with the important role of monocyte chemokinesis in lesion progression.²¹ Most importantly, mCAT-positive lesions had a significantly lower number of bead-labeled cells, suggesting decreased chemokinesis (Figure 4C, right graph). In theory, this finding could be explained by a decrease in peripheral monocyte count,²² but the number of circulating total leukocytes and subsets (both Ly6c^hi and Ly6c^low) were similar between the two groups of mice (Online Figure IVE).

MCP-1 is a major monocyte chemokine in atherosclerosis,^{23, 24} and we recently showed that immunoneutralization of MCP-1 in WD-fed Ldlr^-/- mice decreased the entry of monocytes into lesions¹⁷. In this context, we interrogated lesions for Mcp1 mRNA and found a marked decrease in the lesions of mCAT mice (Figure 4D, graph #1). Note that plasma MCP-1 and lesional mRNAs for other chemokines and their receptors, including Ccl5, Cx3cl1, Cxcl1, Ccr5, Cx3cr1, were not different between the two groups of mice (Online Figure V). MCP-1 is induced in response to activation of inflammatory pathways in macrophages, and so we reasoned that mCAT-mediated suppression of Mcp1 might be part of a larger program of inflammation suppression. Consistent with this idea, the mRNAs of two other inflammatory markers, Tnfa and Inos, were also markedly reduced in the lesions of the mCAT group (Figure 4D, graphs #2-3).

A key inflammatory pathway involves the transcription factor RelA (p65) of the NF-κB pathway. To assess whether this pathway was affected by mCAT, we assayed a marker of pathway activation, namely, nuclear localization of Ser536-phosphorylated RelA.^{25, 26} Analysis of macrophage-rich areas of lesions for co-localization of DAPI and p-RelA showed a clear decrease in nuclear RelA in the lesions of mCAT mice (Figure 4E). In contrast, SMC-rich areas and the endothelium showed no difference in nuclear p-RelA between the two groups (Figure 4E). These combined data are consistent with the idea that mCAT expression in lesional myeloid cells suppresses inflammation in general and NF-κB and MCP-1 expression in particular, which is likely a key mechanism of decreased atherosclerosis in myeloid mCAT-expressing mice.

Quenching MitoOS in cultured macrophages suppresses inflammatory cytokine and chemokine expression and decreases activation of the IKK-RelA pathway

To explore causation and mechanistic links among mitoOS, NF-κB, and MCP-1 expression, we turned to cultured primary macrophages harvested from control or macrophage-mCAT-expressing mice (mCAT^fl/- or mCAT ^fl/-LysMCre^+/- mice, respectively). MitoOS was quantified by FACS analysis of macrophages incubated the mitochondria-targeted fluorescence sensor of superoxide, MitoSOX.²⁷ The first goal was to find activators of mitoOS that were relevant to atherosclerosis. One possibility would be a circulating factor induced by hypercholesterolemia, but when macrophages incubated with serum from WD-fed vs. chow diet-fed Ldlr^-/- mice, the level of MitoSOX fluorescence was similar (Online Figure VI). We therefore tested molecules or other factors that are either known to accumulate in atherosclerotic lesions or mimic processes known to occur in these lesions (Figure 5A). Six of these factors increased mitoOS in macrophages: Lp(a), which is a highly atherogenic lipoprotein known to carry oxidized phospholipids (oxPLs)²⁸; oxPAPC (1-palmitoyl-2-arachidonoyl-sn-phosphatidylcholine), which is a non-lipoprotein-bound oxPL²⁹; the combination of an inducer of endoplasmic reticulum stress, thapsigargin, and another type of oxPL, KodiA-PC (1-[palmitoyl]-2-[5-keto-6-octene-dioyl]phosphatidylcholine)³⁰; 7-ketocholesterol (7KC), which is an oxysterol that accumulates in atheromata³¹; lipopolysaccharide (LPS), which is a model of toll-like receptor activation in atherosclerosis³² and oxidized LDL (oxLDL)³³, which turned out to be the most potent activator of mitoOS in this screen. In contrast, a minimally oxidized form of LDL (mmLDL)³⁴REF YURY MILLER/WITZTUM and two inflammatory cytokines, TNF-α and interleukin-1β (IL-1β), did not induce mitoOS in macrophages.

Figure 5. Cultured macrophages from mCAT ^fl/-LysMCre^+/- mice have less LPS-induced mitoOS, Mcp1, p-RelA, and p-IKK.

(A) Cultured macrophages from wild-type mice were pre-incubated with MitoSOX for 30 min and then treated with the indicated stimuli for either 4 h (mmLDL) or 12 h (other stimuli). The doses were: Lp(a) 25 μg/ml, oxPAPC 25 μg/ml, KOdiA-PC 50 μg/ml, thapsigargin 0.5 μM, mmLDL 50 μg/ml, oxLDL 50 μg/ml, 7KC 35 μg/ml, IL-1β 250 ng/ml, TNF-α 40 ng/ml, LPS 100 ng/ml. The cells were then analyzed by flow cytometry, and the data were quantified as MitoSOX mean fluorescence intensity (n = 3 sets of macrophages/group; *P<0.05). (B) Representative images of macrophages stained with MitoSOX and then treated with LPS (100 ng/ml) and oxLDL (50 mg/ml). After the indicated incubation times, the cells were stained with Mitotracker Green and DAPI and viewed by fluorescence microscopy. The 3^rd and 5^th columns of images are higher magnifications of the boxed areas in the 2^nd and 4^th column of images, respectively. Bars, 10 μm for the first 4 columns of images and 2 μm for the fifth column. The flow cytometry data are shown in the graphs (n = 3/group; *P<0.05). (C) Similar to (B) except that mitOS was measured, showing that mCAT ^fl/-LysMCre^+/- macrophages are protected from LPS- and oxLDL-induced mitoOS (n = 3 sets of macrophages in each group). (D) Time course of Mcp1 and Tnfa mRNA levels, relative to Gapdh, after LPS and oxLDL treatment (n = 3/group; *P<0.05). (E) Immunoblots showing decreased phosphorylation of IKKβ (p-S177) and NF-κB RelA (p-S536) in mCAT^fl/-LysMCre^+/- macrophages after 6 h of LPS or oxLDL treatment; also shown are total catalase and, as a loading control, β-actin. Densitometric quantification of the phospho:total ratio of RelA and IKK from the immunoblots is shown in the bar graphs (n number/group indicated below the graphs; *P<0.05).

For the mechanistic studies that follow, we focused on two of the inducers, oxLDL and LPS. We verified the increase in mitoOS by these stimulators using both non-nuclear 8-OHdG and MitoSOX and showed these signals co-localized with the mitochondria markers ATP5α and Mitotracker Green (Online Figure IA and Figure 5B). Most importantly, MitoSOX fluorescence induced by LPS and oxLDL were lower in macrophages from mCAT-expressing mice (Figure 5C). We then determined whether this model could mimic a major mechanistic finding in our in vivo studies, namely, that mCAT suppresses pro-inflammatory cytokines and chemokines and RelA activation. The data show that Mcp1 and Tnfa mRNA were induced by LPS and oxLDL, and both mRNAs were decreased in mCAT-expressing macrophages (Figure 5D). The decrease of Mcp1 and Tnfa by mCAT was greatest at later time points, suggesting that mitoOS maybe most important for mediating sustained expression of these molecules. As an important negative control for this mCAT effect, we tested mmLDL, which induces Mcp1 but does activate mitoOS (Figure 5A and Online Figure VIII). Consistent with the specificity of the model, mCAT does not suppress Mcp1 in mmLDL-treated macrophages.

We investigated a possible link between mitoOS and inflammation by examining NF-κB RelA (p65) activation. As expected, LPS caused an increase in two markers of activation of this pathway, p-Ser¹⁷⁷-Iκ kinase (IKK) β and p-Ser⁵³⁶-RelA. Most importantly, both of these markers were decreased in mCAT-expression macrophages (Figure 5E). In contrast, the phosphorylation levels of two other LPS-TLR signaling molecules, mitogen-activate protein kinase (MAPK) p38 and c-Jun N-terminal kinase (JNK), were not suppressed by mCAT (Online Figure VII). OxLDL-induced p-Ser⁵³⁶-RelA was also decreased by mCAT (Figure 5E), while mmLDL was unable to activate RelA (data not shown). Thus, the macrophages incubated with LPS or oxLDL capture the essential mechanistic features that were found in the macrophages of control vs. mCAT atherosclerotic lesions.

These macrophage models were then used to address a critical causation question, namely, whether restoration of the RelA pathway could blunt the ability of mCAT to suppress Mcp1 and Tnfa. We began by testing our restoration strategies in control macrophages. The first strategy used RelA transfection, which we found increased both p-RelA and also Mcp1 and Tnfa in response to LPS (Online Figure IXA). The second strategy used IKK transfection, which increased the level of p-IKK and p-RelA and the expression of Mcp1 (Online Figure IXB). We then applied these strategies to control vs. mCAT-expressing macrophages. In macrophages transfected with control vector, mCAT lowered both LPS-induced p-RelA and Mcp1 and Tnfa as above (Figure 6A, left half of blot and first pair of bars in each group in the graphs). In macrophages transfected with RelA, however, there was an increase in LPS-induced p-RelA, Mcp1, and Tnfa, and, most importantly, mCAT did not suppress the cytokine mRNA levels under these conditions (Figure 6A, right half of blot and second pair of bars in each group). Similar results were obtained when restoration of the RelA pathway was accomplished using IKK transfection (Figure 6B) and when the experiment was conducted using the oxLDL-macrophage model (Figure 6C).

Figure 6. Restoration of RelA abrogates the difference in LPS or oxLDL-induced Mcp1 expression in control vs. mCAT-expressing macrophages.

(A) Macrophages were transfected with control (pcDNA3) or RelA-encoding plasmid and then assayed for phospho- and total RelA by immunoblot, with GAPDH as the loading control, and Mcp1 mRNA by RT-qPCR, relative to Gapdh, before and after 6 h of LPS treatment (n = 3 sets of macrophages in each group). (B) The indicated groups of macrophages were transfected with control or IKK-encoding plasmid and then assayed for phospho- and total IKK and RelA and for Mcp1 MRNA before and after 6 h of LPS treatment (n = 3/group; *P<0.05). (C) The indicated groups of macrophages were transfected with control or RelA-encoding plasmid and then assayed for phospho- and total RelA and for Mcp1 MRNA before and after 6 h of oxLDL treatment (n = 3/group; *P<0.05; N.S., non-significant).

Interestingly, Kanters et al.³⁵ found that macrophage IKKβ deletion actually increased lesion area in WD-fed Ldlr^-/- mice. Although the mechanism remains to be determined, macrophages from these mice, when stimulated with LPS in vitro, secreted lower levels of the anti-inflammatory/anti-atherogenic cytokine, interleukin 10 (IL-10). In contrast, we found that the suppression of the IKK pathway by mCAT in macrophages was associated with a slight but significant increase in IL-10 (Online Figure XA), which might contribute to the beneficial effect of mCAT. In another study consistent with our overall findings on the role of RelA in atherosclerosis, Goossens et al.³⁶ demonstrated that myeloid deficiency of IκBα, a negative regulator of RelA, promoted atherogenesis by enhancing leukocyte recruitment to the plaques. Although this study did not probe mechanism in vivo, IκBα deficiency in LPS-treated cultured macrophages was associated with an increase in the chemokine CCL5 but not MCP-1. This finding contrasts with the effects of suppressing RelA via mCAT, which we showed decreases Mcp1 but not Ccl5 both in vivo and in the LPS-macrophage model (Online Figures V and XB).

To test ROS-source specificity in activating RelA-MCP-1 signaling, we transfected macrophages with cytosolic catalase (cCAT), which we showed suppresses LPS-induced cytosolic ROS (cellROX) but not mitoOS (MitoSOX) (Online Figure XIA). cCAT did not suppress RelA and actually enhanced the induction of Mcp1 and Tnfa mRNAs (Online Figure XIB-C), which is consistent with previous studies suggesting non-mitochondrial ROS can inhibit pro-inflammatory cytokine induction^{37, 38}. Thus, mitoOS has distinct roles in inflammatory signaling compared with other sources of cellular oxidative stress.

These combined in vivo and in vitro data support the hypothesis that an important pro-atherogenic mechanism of macrophage mitoOS is the enhancement of IKK/RelA signaling, leading to increased inflammation, including MCP-1-induced monocyte recruitment.

Discussion

In view of the association between markers of mitoOS and the progression of human atherosclerosis² and the importance of inflammatory macrophages in atherosclerosis, the goal of the current study was to provide causation data in vivo for the role of endogenous macrophage mitoOS in atherosclerosis and to explore mechanism. Our data indicate that macrophage mitoOS is atherogenic and that a major mechanism involves activation of an NF-κB—MCP1 pathway. Whether mitoOS in other lesional cell types also contributes to atherogenesis remains to be determined.^39-41 In an elegant study, Sessa and colleagues³ showed that overexpression of thioredoxin-2 in endothelium lowered oxidative stress and enhanced production of NO in endothelial cells, improved endothelial function in aortic rings, and lessened atherosclerosis in Apoe^-/- mice. Although this study did not examine mitoOS per se, the fact that thioredoxin-2 is localized to mitochondria raises the possibility that pro-atherogenic effects of mitoOS in endothelial cells might complement the effects revealed here for macrophages. Moreover, Bennett and colleagues⁴² have shown recently that mitochondrial DNA damage independently of mitoOS can lead to pro-atherogenic changes in smooth muscle cells and monocytes. In the context of that report and our finding that nuclear 8-OHdG becomes very high in the most advanced lesions (Online Figure IC), it is highly likely that non-mitochondrial sources of oxidative stress are important in atherosclerosis, perhaps particularly so in advanced lesions, which is supported by previous studies examining how targeting NADPH oxidase subunits affects atherosclerosis.^43-45

Our mechanistic studies support a model in which lesional macrophage mitoOS promotes inflammation in general and MCP-1 induction in particular, which would then amplify the inflammatory milieu of lesions by promoting additional monocyte entry. Although recent studies using Ccr2^-/- mice showed that MCP-1 can promote the release of Li6c^hi monocytes from the bone marrow⁴⁶, MCP-1 can also contribute to monocyte entry into local sites of inflammation, including atherosclerosis^47-49. Indeed, we found here that mCAT expression in myeloid cells did not affect the level of plasma MCP-1 or the number of circulating monocytes, and we showed recently that injection of anti-MCP-1 neutralizing antibody into a similar model decreased monocyte entry into atherosclerotic lesions¹⁷. Moreover, the level of other potentially atherogenic chemokines and their receptors, including CCR5/CCL5 (RANTES), CX3CR1/CX3CL1 (fraktalkine), and CXCL1/CXCR2^50-52, were similar between control and mCAT lesions. Thus, suppression of the NF-κB—MCP-1—chemokinesis pathway is likely an important mechanism behind the decrease in lesion cellularity in macrophage-mCAT mice.

The data herein have also revealed an interesting link between mitoOS and IKK-RelA activation. Although a number of studies have shown that oxidative stress can activate NF-κB signaling^53-56, the specific roles of different cellular sources of oxidative stress and how each may affect the NF-κB pathway and its various downstream targets remains to be fully explored. In this regard, NF-κB target genes can be affected by NF-κB activation kinetics, cell type, nature of the stimulus, and cofactors^{57, 58}. In a study using murine embryonic fibroblasts and human peripheral blood mononuclear cells, the mitochondria-targeted antioxidant Mitoquinone (MitoQ) was reported to suppress LPS-induced cytokine production, but deletion of the NADPH oxidase subunits gp91^phox and p22^phox in macrophages actually showed a trend toward increased cytokine production³⁷. With regard to the MitoQ result, the authors hypothesized that the mechanism involved suppression of p38 and JNK MAPK signaling, but data specifically linking mitoOS to these MAPKs were not provided. In another study, NADPH oxidase-deficient macrophages from gp91^phox-null mice and CGD patients also produced increased inflammatory cytokines in response to LPS independently of NF-κB³⁸. These data are consistent with our finding that quenching non-mitochondrial ROS by cCAT leads to enhanced LKPS-induced inflammatory cytokine induction without affecting NF-κB activation. Thus, the source and/or intracellular location of oxidative stress can have distinct effects on activation of NF-κB signaling.

We provide evidence that mitoOS is linked to NF-κB through IKK, but exactly how mitoOS activates IKK remains to be determined. Gloire et al.⁵⁹ showed that H₂O₂ treatment of Jurkat cells led to activation of IκK and RelA through a pathway involving the phosphatase SH2-containing inositol-5′-phosphatase 1 (SHIP-1). Whether mitoOS enhances the activation of IkK through SHIP-1 remains to be determined. Of interest, a recent study investigated the converse issue in macrophages, namely, activation of mitoOS by inflammatory signaling²⁶. The investigators showed that activation of certain Toll-like receptors led to recruitment of TRAF6 and ECSIT (Evolutionarily Conserved Signaling Intermediate in Toll pathways) to mitochondria, ubiquitination of ECSIT by TNF receptor-associated factor 6 (TRAF6), and induction of mitoOS, presumably by ubiquitination of ECSIT on the mitochondrial respiratory chain. However, whether the mitoOS generated by this mechanism then mediated or amplified downstream TLR-induced NF-κB signaling was not reported.

Oxidative stress in general has long been considered to be a therapeutic target for atherosclerotic vascular disease. Enthusiasm was dampened, however, by studies in humans showing that vitamin E was not protective against human coronary artery disease⁶⁰. While these data might be interpreted as proof against the role of oxidative stress in atherosclerosis, a more likely explanation is that the choice and/or timing of anti-oxidant treatment was not optimal. The association of mitoOS with human atherosclerosis progression and the causal and mechanistic insights provided by the current findings and previous studies raise the possibility that therapy targeted specifically to mitoOS may show benefit. In this regard, Bennett and colleagues⁶¹ showed that systemic MitoQ administration decreased macrophage content and cell proliferation in the atherosclerotic lesions of fat-fed Apoe^-/- mice. Although the mechanism behind these findings with regard to atherogenesis per se is difficult to ascertain in view of systemic metabolic effects of MitoQ administration, and although applicability to humans remains unexplored, continuing insight into the mechanisms and consequences of oxidative stress in atherosclerosis will lead to a more focused approach to this important area of biomedical research.

Novelty and Significance.

What Is Known?

• Excess mitochondrial oxidation (mitoOS) occurs in different types of cells in the atherosclerotic lesions in both human and animal models.

• Mouse models of atherosclerosis that have been genetically engineered to have increased mitoOS above the endogenous level have larger lesions.

What New Information Does This Article Contribute?

• This article provides causative in vivo evidence that quelling mitoOS in macrophages suppresses atherosclerosis by decreasing monocyte infiltration and lesional inflammation.

• MitoOS promotes monocyte infiltration through activation of the IKK-RelA (NF-κB) pathway, which increases expression of the chemokine Mcp1.

• Studies with cultured macrophages suggest that mitoOS can be stimulated by numerous factors in atherosclerotic lesions, including toll-like receptor activators and oxidized lipids.

MitoOS has been linked to atherosclerosis, but the causative role and pathogenic mechanisms of mitoOS in specific lesional cell types is not known. In this study, we use a transgenic mouse model in which macrophage mitoOS is quenched through mitochondria-targeted expression of catalase. The data show that macrophage mitoOS promotes atherosclerosis in fat-fed Ldlr^-/- mice by increasing monocyte infiltration. MitoOS, but not cytosolic ROS, activates NF-κB (RelA), leading to induction of the monocyte chemokine, MCP-1. These findings reveal a potentially new therapeutic target to prevent the progression of atherosclerosis.

Nonstandard Abbreviations and Acronyms

BMDM

bone marrow-derived macrophages

LCM

laser-capture microdissection

mitoOS

mitochondrial oxidative stress

mCAT

mitochondria targeted human catalase

WD

Western diet