Circulation Research Manuscript #CIRCRESAHA/2011/247577 - R3 Regular Article

JOURNAL OF THE AMERICAN HEART ASSOCIATION

Manuscript Number: CIRCRESAHA/2011/247577 - R3 Article Type: Regular Article

Title: Activating Transcription Factor 1 directs Mhem atheroprotective macrophages via coordinated iron handling and foam cell protection Authors: Joseph J Boyle, Michael Johns, Theresa Kampfer, Aivi T Nguyen, Laurence Game, Dominik J Schaer, Justin C Mason, and Dorian O Haskard

This paper includes the following additional materials for review purposes: · Online Data Supplement(s) · Abstract Supplement(s) · Potentially Overlapping Work(s) To view additional materials: 1. Go to http://submit-circres.ahajournals.org and log in 2. Enter the Reviewer Area and click on Active Reviews 3. Click on the [Download Supplemental Files] link and select supplements to download

Circulation Research Online Submission: http://submit-circres.ahajournals.org Circulation Research Homepage: http://circres.ahajournals.org Circulation Research Editorial Office 3355 Keswick Rd, Main Bldg 103 Baltimore, MD 21211 [email protected] Telephone: 410-327-5005 Fax: 410-327-9322 Circulation Research Manuscript #CIRCRESAHA/2011/247577 - R3 Regular Article

Activating Transcription Factor 1 directs Mhem atheroprotective macrophages via

coordinated iron handling and foam cell protection

Joseph J Boyle

Michael Johns

Theresa Kampfer

Aivi T Nguyen

Laurence Game

Dominik J Schaer ¶

Justin C Mason ¶

Dorian O Haskard

¶ Equal contribution

Running title:

ATF-1 drives atheroprotective macrophages

Circulation Research Manuscript #CIRCRESAHA/2011/247577 - R3 Regular Article

Abstract

Rationale. Intraplaque hemorrhage (IPH) drives atherosclerosis via the dual metabolic stresses of cholesterol-enriched erythrocyte membranes and pro-oxidant heme/iron. When clearing tissue hemorrhage, macrophages are typically seen storing either iron or lipid. We have recently defined hemorrhage-associated macrophages (HA-mac) as a plaque macrophage population that responds adaptively to IPH. Objective. This study aimed to define the key transcription factor(s) involved in HO-1 induction by heme. Methods and

Results. To address this question, we used microarray analysis, and transfection with siRNA and plasmids. To maintain physiological relevance, we focused on human blood-derived monocytes. We found that heme stimulates monocytes via induction of activating transcription factor 1 (ATF-1). ATF-1 co-induces Heme Oxygenase-1 (HO-1) and Liver X

Receptor Beta (LXR-β). Heme-induced HO-1 and LXR-β were suppressed by knockdown of

ATF-1, and HO-1 and LXR-β were induced by ATF-1 transfection. ATF-1 required phosphorylation for full functional activity. Expression of LXR-β in turn led to induction of other genes central to cholesterol efflux, such as LXR-α and ABCA1. This heme-directed state was distinct from known macrophage states (M1, M2, Mox) and, following the same format, we have designated them Mhem. Conclusions. These results show that ATF-1 mediates HO-1 induction by heme, and drives macrophage adaptation to intraplaque hemorrhage. Our definition of an ATF-1 mediated pathway for linked protection from foam cell formation and oxidant stress may have therapeutic potential.

(226 words)

Key words: macrophages, atherosclerosis, heme oxygenase-1, lipids, activating transcription factor-1.

Circulation Research Manuscript #CIRCRESAHA/2011/247577 - R3 Regular Article

Introduction

Atherosclerosis is an inflammatory disease of large artery walls, driven mainly by lipid peroxidant stress from oxidized low density lipoprotein (OxLDL) 1. Intraplaque hemorrhage

(IPH) is particularly important in promoting both atherosclerotic lesion progression and destabilization 2. Thus, erythrocytes provide a potent combination of cholesterol-enriched membrane lipids and heme-iron, which together pose a serious metabolic challenge in a pathology largely driven by oxidized cholesterols 2. Along with intracranial hemorrhage, IPH is one of the most important examples of tissue damage due to extravasated blood 2. Heme- iron is an effective peroxidant catalyst, via hydrogen peroxide coordination and Fenton chemistry 3; cholesterol is modified by peroxidation to potently cytotoxic and inflammatory oxysterols (5’, 6’-epoxycholesterol, 7’-ketocholesterol and 21’-keto-cholesterol) 4; and macrophages are abundant in atherosclerosis and generate hydrogen peroxide when activated.

The combination of cholesterol loading, heme/iron loading and macrophage activation would therefore promote lipid peroxidation 5. How monocytes entering plaques differentiate adaptively to clear hemorrhage-related iron and lipid may therefore be a key transcriptional decision in atherosclerosis.

Heme Oxygenase-1 (HO-1) is a vital enzyme for iron homeostasis and protection from oxidant stress. It catalyzes pro-oxidant heme and generates biliverdin, free iron and carbon monoxide as reaction products 6. HO-1 activity also stimulates upregulation of ferritin genes, leading to the safe chelation of iron 6. Biliverdin is processed further to bilirubin, a direct antioxidant 6. Like nitric oxide, low levels of carbon monoxide activate soluble guanylate cyclase to signal vascular protective responses 6. For all these reasons, induction of HO-1 by heme provides a major pathophysiological homeostatic response in hemorrhage.

Circulation Research Manuscript #CIRCRESAHA/2011/247577 - R3 Regular Article

We have recently identified a unique state of macrophages in areas of hemorrhage within human atherosclerotic plaques that we originally designated HA-mac, but have now renamed

Mhem by analogy to M1, M2 and Mox states 7. These cells have increased intracellular iron, but are likely to be atheroprotective, as they show increased HO-1 expression, reduced oxidative stress, suppressed inflammatory activation, and increased release of the anti- inflammatory cytokine interleukin-10 7. Furthermore, they do not show lipid accumulation of the type that defines foam cells. We have shown that the same state can be induced in human blood-derived monocytes in vitro by culturing in the presence of hemoglobin-haptoglobin complexes over 4-8 days 7. The in vitro generation of Mhem was found to be dependent upon an autocrine loop involving release of the anti-inflammatory cytokine IL-10 7. We have recently found that heme is the active stimulant for Mhem development, inducing HO-1 mRNA expression by a cycloheximide-sensitive mechanism suggesting involvement of a trans-activating protein8.

Morphometric analysis of macrophages in atherosclerotic lesions with IPH showed that

Mhem cells are distinct from classical lipid-laden macrophages, and occupy separate microdomains 7. This observation has a parallel in the standard pathological teaching that resolving hematomas contain distinct specialized macrophages containing either lipid (foam cells) or iron (siderophages) 2. The mechanisms regulating this functional specialization are not known. In this paper we have used transfection of human blood-derived monocytes with siRNA and plasmids, together with microarray analysis to identify activating transcription factor 1 (ATF-1) as a key transcription factor mediating heme-induced HO-1 expression.

ATF-1 is a close, but relatively little studied, homolog of cyclic-3’-5’-adenosine monophosphate (cyclic-AMP, cAMP) binding protein (CREB1), and both mediate transcriptional activation by cyclic-AMP by binding to the cyclic-AMP response element

(CRE). Importantly, we have found that ATF-1 also induces a programme of gene expression

Circulation Research Manuscript #CIRCRESAHA/2011/247577 - R3 Regular Article

responsible for lipid export, providing a mechanism explaining the reciprocal nature of iron and lipid handling by the cell. Together, these mechanisms may underlie homeostatic adaptation to IPH in macrophages.

Methods

A full version of the Methods is at http://www.circres.ahajournals.org

Cell isolation and culture.

Peripheral blood-derived monocytes were purified and cultured from normal human volunteers as before, with informed consent and ethical clearance 7.

Microarray and qPCR

Standard molecular biology methods were used. Macrophages were lysed, RNA purified and quality controled, and then parallel aliquots analysed by both Agilent 4x44k and Affymetrix

1.0ST microarrays following respective manufacturer’s instructions. Results were validated by reverse-transcription-quantitative polymerase chain reaction (qPCR). Microarrays were analysed with Agilent GeneSpring GX10 and genelists exported to specialized bioinformatics programs. For selected genes, qPCR was done over a full time course in a further 5 donors.

Western analysis

Westerns blots were by amendment of our previous methods 7. Cells were lysed and run by

SDS-PAGE and blotted to PVDF using a commercial kit (Novex, Invitrogen). Blots were probed with multiple antibodies as indicated and visualized with ECL+ and Hyperfilm (GE

Healthcare).

Quantitative chromatin immunoprecipitation (qChIP)

ChIP was performed following a minor amendment of published protocols 9. Macrophages were fixed in 1% formaldehyde at 4°C for 5mins, stopped with excess glycine, lysed in

Circulation Research Manuscript #CIRCRESAHA/2011/247577 - R3 Regular Article

nuclear lysis buffer AM-1, sheared by 28G needle three times (giving 1 kb fragments) and

ChIP carried out with the magnetic non-enzymatic kit (Active Motif) using Protein G beads

(Invitrogen). Immunoprecipitates were DNA–purified and qPCR was as above.

Quantitation of ATF-1 / DNA binding

The 96-well plate assay for ATF-1 binding to promoter sequences was reverse engineered from TransAm kit (Active Motif) for CREB1 activation (CREB1 and ATF-1 bind to a common motif). Oligos corresponding to human HO-1 and LXR-β enhancers or their ΔATF-

1 mutants were bound to ELISA plates, macrophage lysates added and ATF-1 / DNA binding detected with anti-p-ATF-1 (Abcam).

Plasmids and siRNA

Plasmids (kind gifts, Y. Tsuji, North Carolina State University, USA.; A. Agarwal, University of Alabama at Birmingham, USA, or purchased from Clontech) were amplified and purified by standard molecular biology methods. Methods for transfection of volunteer blood-derived macrophages with plasmid DNA or siRNA were optimized and validated. Oligos for siRNA were from Dharmacon (Abgene) as indicated.

Functional assays

Oxidative stress assays were carried out as before 7, 8 using aminophenyl fluorescein a selective dye assay for hypochlorite, hydroxyl radicals, and peroxynitrite (Invitrogen).

Cholesterol export was measured by a minor modification of established methods 10.

Macrophages were incubated with human recombinant ApoA1, an acceptor for cholesterol export, and cholesterol assayed with a specific high sensitivity fluorescent kit (Invitrogen) as described 11. Luciferase analysis was a Dual-Luciferase kit (Promega) and a 96-well luminometer (BioTek Synergy HT). For the foam cell assays, macrophages were cultured on

Circulation Research Manuscript #CIRCRESAHA/2011/247577 - R3 Regular Article

tissue culture glass slides, challenged with Ox-LDL, stained with 10μM Nile Red (Sigma-

Aldrich), and imaged by confocal microscopy (Zeiss LSM510) based on previous methods 7.

Circulation Research Manuscript #CIRCRESAHA/2011/247577 - R3 Regular Article

Results

Heme induces HO-1 via activating transcription factor 1 (ATF-1)

Macrophages in atherosclerotic plaques tend to store either iron (siderophages) or lipid (foam cells) (Figure 1A, B). To investigate the mechanisms underlying this distinction further, we stimulated freshly isolated human blood-derived monocytes with heme and measured gene expression after one and four hours using microarray technology. Genes induced at one hour were generally only modestly upregulated and consisted mainly of signalling molecules and transcription factors (Figure 1C, and Supplemental Figures I-V). In contrast, those upregulated at four hours included a cluster of highly expressed effector genes such as

HMOX1, the gene for HO-1 (Figure 1D). When genes induced at one hour were explored in more detail using Gene Ontology, we found the most upregulated transcription factor to be

ATF-1 (Figure 1E). The most upregulated effector genes are shown in Figure 1F and

Supplemental Table I. In contrast, macrophage pro-inflammatory activation genes were consistently repressed (Supplemental Figure VI).

The microarray data was corroborated with 2 microarray systems and with qPCR

(Supplemental Figure VII). We then investigated the relationship between ATF-1 and HO-1 by qPCR in heme-stimulated monocytes from five additional donors. ATF-1 expression was transient and succeeded by a wave in HO-1 expression (Figure 2A) with a concomitant high time-lagged correlation coefficient (R=0.71, p<0.05) 12. Therefore, we tested the hypothesis that ATF-1 expression causally drives HO-1.

ATF-1 is known to be activated by phosphorylation at Ser63. By western analysis, ATF-1 and p-Ser63-ATF-1 protein both peaked at about four hours after heme-stimulation, corresponding with emergence of HO-1 mRNA (Supplemental Figure VIII A, B). Of note, while total ATF-1 protein was principally cytoplasmic, p-ATF-1 was located in the nucleus,

Circulation Research Manuscript #CIRCRESAHA/2011/247577 - R3 Regular Article

consistent with functional activity (Supplemental Figure VIII C,D). We then used siRNA perturbation analysis to test the dependency of HO-1 induction by heme on ATF-1 induction.

Pooled ATF-1 siRNA, and two independent ATF-1 siRNA oligonucleotides each suppressed both ATF-1 and p-ATF-1 (Figures 2B-E and Supplemental Figure IX), and prevented HO-1 mRNA and protein induction in response to heme (Figure 2B-E). This had a concomitant stimulatory effect on release of highly reactive oxygen species (hROS) by monocytes and led to reduced cell survival (Supplemental Figure X). Conversely, transfection (validated in

Figure 2F) of an ATF-1-expressing plasmid into monocytes led to an increase in p-ATF-1 and HO-1 mRNA and protein (Figure 2G, H). Finally, transfection of ATF-1, but not unrelated plasmids or a Ser63-Ala non-phosphorylatable mutant (designated DN-ATF-1), also suppressed hROS (Figure 2I). In contrast, transfection of a close homolog of ATF-1, cAMP response element binding protein (CREB1), led to an increase in hROS (Figure 2I).

Heme induces a set of genes distinct from M1, M2 or Mox

Analysis of the microarray data using FDR-adjusted ANOVA showed that 2400 genes reached statistical significance. Therefore further conservative cut-offs on fold-regulation

(over 1.4-fold up or down) and intensity (above the level of known muscle-markers) were applied to exclude weakly regulated or poorly expressed genes. These more stringent criteria indicated that heme significantly modulated approximately 800 genes (Heatmap in

Supplemental Figure XI A, B). To make sense of this large number of genes, we first used a

Venn diagram to compare the Mhem transcriptome with published human M1 and M2 transcriptomes (GEO database, NCBI). This revealed that Mhem had approximately 700 unique genes and shared only 1 gene with M2, which was LHFPL2, a gene of essentially unknown function (Supplemental Figure XI C and Supplemental Table II). The 44 genes in the Mhem / M1 intersection were counter-regulated in Mhem and M1 (Supplemental Table

II). Thus Mhem cells appear to be distinct from canonical M1 or M2 macrophages. An

Circulation Research Manuscript #CIRCRESAHA/2011/247577 - R3 Regular Article

equivalent Venn diagram for Mox 13 and Mhem is shown in Supplemental Figure XI-D.

Approximately 800 genes were uniquely regulated in Mhem, 157 genes were uniquely regulated in Mox, and 7 were coregulated (namely HMOX1, GCLM, OSGIN1, IHPLP2,

DCK, NINJ1, TRAF5). The fraction (approximately 1%) of shared genes was no different to chance (χ2, p>0.95). Thus, even allowing for experimental variation between laboratories and platforms, although Mhem and Mox share occasional key genes, such as HO-1, they are distinctly regulated.

We probed the relationship of the heme-response state to M1, M2 and Mox states experimentally (Supplemental Figure XII A-I). Heme suppressed HLA-DR as we have previously reported (Supplemental Figure XII A) 7. Either IL-4 treatment (M2) or interferon-

γ / LPS treatment (M1) increased HLA-DR (Supplemental Figure XII B). Itself, this indicated the heme-treated state is distinct to M1 or M2. Heme suppressed HLA-DR in either M1 or

M2 macrophages (Supplemental Figure XII C). Heme increased surface CD163 as previously reported (Supplemental Figure XII D) 7, 8. We found that in either M1 or M2 macrophages, heme did not increase CD163 expression (Supplemental Figure XII E). As a control, we assessed the plasticity of M1 versus M2 commitment. We cultured human blood-derived macrophages with M1 and M2 stimuli for 3 days, swapped the stimuli and carried out flow cytometry after another 3 days for a canonical M1-M2 marker macrophage mannose receptor

(MMR) (Supplemental Figure XII F). This indicated that MMR expression was conditioned by the initial stimulus, i.e. M1 and M2 are committed. Culture of human blood-derived macrophages with OxLDL prevented heme from either suppressing HLA-DR or inducing

CD163, also indicating commitment (Supplemental Figure XII G, H).

Interestingly, culture of macrophages with OxLDL, or with M1 conditions caused macrophages to induce p-ATF-1 more strongly in response to heme (Supplemental Figure

XII - I). In contrast, culture with POVPC (to induce Mox), or with M2 conditions suppressed

Circulation Research Manuscript #CIRCRESAHA/2011/247577 - R3 Regular Article

p-ATF-1 induction by heme (Supplemental Figure XII I, J). This indicates that Mhem are actually antagonised by Mox- or M2- differentiation. Thus, while the responses are complex and a full analysis would require a dedicated paper, Mhem are clearly distinct to M1, M2 or

Mox states, and M1, M2 or Mox differentiation prevents the full Mhem state.

When added to macrophages, heme immediately induces hROS, before a later (biologically- mediated) suppression 8 . We then examined whether heme-induced ATF-1 was mediated by heme-induced oxidative stress. We assessed a prototypical direct antioxidant (N-acetyl- cysteine, (NAC) free radical scavenger) and prototypical indirect antioxidant (curcumin, anti- inflammatory and inducer of endogenous antioxidants). We first validated that NAC suppresses hROS in a cell-free system (Supplemental Figure XIII A) and then in macrophages (Supplemental Figure XIII B). Both NAC and curcumin induced p-ATF-1 and suppressed additional heme-induced p-ATF-1 (Supplemental Figure XIII C, D). Since NAC suppresses hROS but induces p-ATF-1, it seems highly improbable that heme induces ATF-1 via a simple oxidant stress mechanism alone.

Heme stimulates an ATF-1 dependent gene regulatory network that coordinates iron handling with lipid handling

We then used standard cluster analysis methods to define secondary response genes induced by heme at 4h. When this cluster was analyzed by Gene Ontology (GO) classification to assign biological meaning, we unexpectedly found that it was significantly enriched in genes involved in lipid handling (Figure 3A, listed in Supplemental Table III and Supplemental

Figure XIV). These included the transcription factor liver X receptor beta (LXR-β), a ‘master regulator’ of lipid metabolism (Figure 3B).

We extracted the regulatory sequences of each gene between human and mouse from the

ENSEMBL database, identified conserved motifs, and analysed these for transcription factor

Circulation Research Manuscript #CIRCRESAHA/2011/247577 - R3 Regular Article

binding sites. Other than LXR-β itself (the ATF-1 responsive initiating gene), all of the key heme-inducible lipid handling genes had predicted LXR response elements (Figure 3C). We thus postulated that ATF-1 co-induces LXR-β and HO-1 and thereby directs a gene network that integrates lipid and iron handling (Figure 3C) and went on to test this linkage.

In time-course studies, ATF-1 preceded the largest part of the LXR-β rise (Figure 4A).

Involvement of ATF-1 in LXR-β induction by heme was confirmed using ATF-1 siRNA

(Figures 4B, C), and, again conversely, ATF-1-transfection induced LXR-β (Figure 4D, E).

To exclude off-target effects, the siRNA data were corroborated with two siRNA oligos for

ATF-1, two control oligos and pooled oligos (Dharmacon SMARTpool) as for ATF-1 / HO-

1, shown as previously in Supplemental Figure IX).

Heme and ATF-1 transcriptionally activate HO-1 and LXR genes

To further test causality in this transcriptional network, we next assessed luciferase constructs. We first used an easily transfectable HEK cell line (not shown), which yielded more light-signal but obtained qualitatively similar results with human blood-derived monocytes, which were more physiologically relevant (Figure 5). These we transfected with a series of luciferase constructs covering sequentially larger fragments of the 5kb enhancer- promoter of HO-1 (Figure 5A-C). Heme-stimulation (Figure 5B) or ATF-1 cotransfection

(Figure 5C) induced robust luciferase activity only with the full-length construct (-4.9kb), but not with the shorter constructs (-2.2kb and -0.3kb), indicating a requirement for the enhancer sequence between -2.2kb and -4.9kb. ATF-1 is known to bind DNA via the cyclic-AMP response element (CRE): TGACGTCA, or its half-sites (TGACG / GTCA)14, which is conserved in both HO-1 and LXR-β in both mouse and human (Supplemental Figure XV).

As shown in Figure 5D, heme failed to activate the -4.9kb reporter construct when containing

Circulation Research Manuscript #CIRCRESAHA/2011/247577 - R3 Regular Article

a mutated ATF-1 response site (-4.9ΔATF-1 mutant). A similar pattern was seen with ATF-

1-cotransfection (not shown).

ATF-1 binds to the enhancers for the genes for HO-1 and LXR-β

We next tested whether ATF-1 binds to its predicted response elements in the HO-1 and

LXR-β enhancers (Figure 5E, F). Chromatin immunoprecipitation (ChIP) is a method of detecting complexes between specific proteins and DNA sequences in live cells. Using an anti-p-ATF-1 antibody in a quantitative ChIP assay, we found that binding of p-ATF-1 to the

ATF-1 site of the HO-1 enhancer was increased in human blood-derived monocytes approximately 10-fold four hours after heme stimulation, but not after only one hour (Figure

5E). Moreover, binding of p-ATF-1 to a proximal (-0.3kb) control site in the HO-1 promoter was undetectable at either time-point (not shown). Measurement by qPCR of the LXR-β enhancer promoter in the same ChIP samples also showed significant enrichment after heme treatment, although the peak of heme-stimulated p-ATF-1 / LXR-β binding was earlier, consistent with the earlier induction of LXR-β mRNA (Figure 5E). Similar data were obtained with a solid phase assay quantifying p-ATF-1 / DNA binding capacity. Immobilized oligonucleotides from HO-1 or LXR-β enhancers, but not the corresponding mutants with mutated ATF-1 response elements, captured more p-ATF-1 complexes from heme-stimulated blood-derived monocytes than controls (Figure 5F).

The heme / ATF-1 pathway drives lipid export

So far our bioinformatic data had suggested a coregulatory mechanism where ATF-1 contributes to induction of LXR-β as well as HO-1, with each in turn driving many other genes regulating iron handling and lipid transport (Figure 3). Consistent with the latter, time- course qPCR showed that ApoE expression followed LXR-β induction (Figure 6A). We also

Circulation Research Manuscript #CIRCRESAHA/2011/247577 - R3 Regular Article

examined the predicted LXR-β–responsive genes LXR-α and ABCA1, and found strong late induction of both LXR-target genes LXR-α and ABCA1 at 48 hours (Figure 6B). Heme- induced ApoE was prevented by ATF-1 siRNA (Supplemental Figure XVI). LXR-β-specific siRNA was found to suppress levels of mRNA for both LXR-α and ABCA1 at 48 hours after heme stimulation, with western blotting confirming the results for the latter at protein level

(Supplemental Figure XVII). Furthermore, we found that heme and ATF-1-cotransfection both induced LXR-α−luciferase plasmids at 48h (Figure 6C).

As ABCA1 is a key cholesterol exporter to HDL, we next transfected ATF-1 or a control green fluorescent protein (GFP) construct into blood-derived monocytes to ask whether this pathway promoted cholesterol export, challenging the cells with OxLDL. We found that heme transfection suppressed cell-associated cholesterol (Figure 6D) and increased cholesterol efflux (Figure 6E), and both were prevented by ATF-1-siRNA (Figure 6D, E).

Furthermore, ATF-1-transfected cells were resistant to foam cell formation when compared to controls (Figure 6F, G). These effects are similar to those induced by the LXR-agonist R- hydroxycholesterol and were inhibited by the LXR-antagonist S-hydroxycholesterol (not shown).

Mhem macrophages in human intraplaque hemorrhage are protected from foam cell change and co-express p-ATF1, HO-1 and ABCA-1

Lastly, we went back to the human culprit plaques in which Mhem macrophages had first been described 7. Because Mhem macrophages and foam cells are found in distinct zones 7, we could easily re-examine the same cells in adjacent serial sections. We first compared the sizes of macrophages identified as Mhem with those identified as conventional inflammatory plaque macrophages, and found that Mhem cells were smaller, consistent with resistance to

Circulation Research Manuscript #CIRCRESAHA/2011/247577 - R3 Regular Article

foam cell change (Figure 7A). We found p-ATF-1 in Mhem macrophages, but not in control foam cell macrophages with expanded cytoplasm due to accumulated lipid (Figure 7B).

Furthermore, p-ATF-1 was nuclear in the Mhem macrophages, consistent with transcriptional activation (Figure 7B). Relative to the foam cells, the Mhem macrophages expressed increased HO-1, ABCA1, ApoE and LXR-β (Figure 7C and Supplemental Figures XX-XXI).

Using 4-channel confocal, we then also formally showed there was colocalization of p-ATF-1 with HO-1 and ABCA1 (Figure 7D). Thus our mechanistic insights are likely to apply to macrophages within vulnerable atherosclerotic plaques.

Circulation Research Manuscript #CIRCRESAHA/2011/247577 - R3 Regular Article

Discussion

Atherosclerosis, the main pathology causing cardiovascular morbidity and mortality, largely comprises macrophage-mediated inflammation driven by oxidized lipids. The state of the infiltrating macrophages is a key pathophysiological determinant. IPH is pathogenic in advanced atherosclerosis via a combination of loading with iron, a lipid peroxidant, and erythrocyte membrane cholesterol. We have described a macrophage phenotypic adaptation to IPH that appears to partially counter its pathogenic effects 7.

Coordination of lipid and iron homeostasis

We show here that the hemorrhage-induced macrophage state is primarily directed by ATF-1 mediated co-induction of HO-1 and LXR-β, Together, these coordinate the safe handling of iron in parallel with lipid export from the cell (Figure 8). The segregation of lipid storage in foamy macrophages versus iron storage in hemosiderin-laden macrophages makes functional sense as a strategy to help prevent iron-mediated lipid peroxidation. Heme / ATF-1 mediated co-induction of HO-1 and LXR-β provides a molecular mechanism for this phenomenon. We defined this pathway by genetic modification of blood-derived monocytes from normal human volunteers, to allow both pathophysiological relevance and specific molecular control.

Pathophysiological relevance of the model

The in vitro model of challenge with 10μM heme is physiologically relevant, being approximately 1/40th of that measured in subarachnoid hematoma in vivo 15. Furthermore, recent transcriptome analyses of plaque macrophages and pooled cell line microarray data indicate that gene expression in freshly isolated blood-derived monocytes is relevant to plaques, in sharp contrast to gene expression in transformed cell lines 16, 17. Indeed, the macrophage cell lines we have examined (eg U937, THP-1, RAW264) have constitutively expressed very high levels of p-ATF-1, consistent with its role as a proto-oncogene 18 (not

Circulation Research Manuscript #CIRCRESAHA/2011/247577 - R3 Regular Article

shown). Thus although human blood monocytes are difficult to transfect, they are likely to be to be far more informative about human pathophysiology. Of note, siRNA is in essence a perturbation experiment, and thus useful even in the absence of complete knockout 19.

Mhem macrophages mitigate accelerant effects of intraplaque hemorrhage

The macrophage state we are now defining as Mhem was first described associated with human fatal coronary intraplaque haemorrhages (hemorrhage-associated macrophages, HA- mac)7. These had less oxidative stress than foamy macrophages – in the face of increased iron storage; higher HO-1 expression; less myeloperoxidase expression and higher IL-10 expression. We modelled these in vitro, and found that hemoglobin:haptoglobin complexes evoked cells with similar characteristics, switched by an autocrine IL-10 / CD163 feedback loop7. These cleared Hb more quickly than control macrophages and we proposed that they were adaptively differentiated to mitigate the otherwise accelerant effects of intraplaque hemorrhage on plaque development. We now show that the initiating transcriptional events involve induction and phosphorylation and ATF-1 and subsequent co-induction of an HO-1 cascade and an LXR / ABCA1 / ApoE / foam cell protection cascade. The HA-mac / Mhem cells in intraplaque hemorrhage indeed co-express p-ATF-1, HO-1, and the canonical LXR targets ApoE and ABCA1, and have less foam cell change, a feature we did not actively comment on before in the absence of mechanistic evidence.

Cyclic-AMP response element mediated regulation of vascular disease

ATF-1 is named for activating transcription in response to cyclic-AMP 20. Its nearest homolog, cAMP response element binding protein (CREB) has been much more extensively studied, and is thought to be the major cAMP activated transcription factor, regulating key metabolic genes, e.g. gluconeogenesis 21-23 and lipoprotein lipase 24. Interestingly, Ruffell et.al. described dependency of wound-resolving macrophages on CREB, with a targeted

Circulation Research Manuscript #CIRCRESAHA/2011/247577 - R3 Regular Article

inactivating mutation of its response element in a single target gene preventing in vivo healing 25, setting a precedent for a role of the CREB family in macrophage differentiation.

Similarly, the cytoprotective role of ATF-1 we describe here has precedents in pro-survival and mitogenic effects in other lineages 18, 26-29. In contrast to ATF-1 dependent cytoprotection in human blood monocytes, ATF-1 suppresses induction of ferritin heavy chain and the HO-1 gene via antioxidant response elements in HEK cells 30. Again in HEK cells, CREB-1 binding to the cyclic-AMP response element at ∼-4kb mediates the induction of HO-1 by the oxylipid nitrolinoleic acid 31. Here, we found an alternative use of the same response element (GTCA) in mediating heme-induced HO-1. As there has been little attention paid to determining how heme induces HO-1 6, the identification of the mechanism of this important physiological negative feedback loop is highly significant.

LXRs in vascular disease

Atherosclerosis is thought to largely hinge on phagocytosis of OxLDL by plaque macrophages via scavenger receptors 32, causing pro-inflammatory activation, endoplasmic reticulum stress and oxidant stress, promoting death and necrotic core formation 33. However, macrophages also protect against inflammatory activation and lipid overload by lipid export to high density lipoprotein (HDL) in reverse cholesterol transport 34, via adenosine- triphosphate-binding-cassette-transporter (ABC) proteins -A1 and -G1 34. ABCA1 and

ABCG1 are upregulated by LXRs, former orphan nuclear receptors ligated by oxysterols and by cholesterol at ‘overload’ concentrations 2, 5, 35, 36. Thus LXRs sense cholesterol overload and cholesterol oxidation and negatively feedback to evoke cholesterol export and suppress inflammatory activation 37.

Hemorrhage / heme-challenge feeds forward into this key defence via a novel heme → ATF-

1 → LXR-β → LXR-α cascade. Notably, in a landmark paper, peroxisome proliferation

Circulation Research Manuscript #CIRCRESAHA/2011/247577 - R3 Regular Article

activation receptor γ (PPAR-γ) induced lipid export via transcriptional induction of LXR-α

35. The Heme → ATF-1 pathway in our study induced greater fold-change in LXR-α mRNA

(11-fold) than was previously observed via PPAR-γ (5-fold) 35. Moreover, this appears to be the first documentation of LXR-β isoform induction by any stimulus. The magnitude of effect of heme → ATF-1 on cholesterol efflux is similar to that in the PPAR-γ → LXR → ABCA1 axis 35 and comparable to differences between normal versus ABCG-deficient mice 38, 39. The heme → ATF-1 → LXR-β cascade was devoid of induction of fatty acid synthesis genes, a major side-effect of some pharmacological LXR agonists that is via LXR-α activation 41-44, suggesting that heme → ATF-1 → LXR-β mediated activation may be more beneficial.

Macrophage polarization

The Mhem anti-inflammatory state appears to be a polar opposite of the recently described

M1-like iron-overload state that impairs wound healing 45. We suspect that this is what ATF-

1 protects against, and therefore ATF-1-deficient mice may not cope with hemorrhage 46.

However, this is an area of active investigation beyond the scope of the current paper.

Although plaque M2 macrophages may overexpress PPAR-γ, which also may induce LXR-α, our data indicate that Mhem is very distinct from an M2 state 47. Furthermore, although

Leitinger’s group recently described a HO-1-expressing putatively protective mouse macrophage state Mox driven by oxidized phospholipids via Nrf2 13, transcriptome comparisons indicate that Mhem are not Mox. However, we have shown elsewhere that Nrf2 plays a role in heme-induced HO-1 8. It is possible therefore that ATF-1 and Nrf2 may be cooperative in HO-1 induction, as seen for NFκB / AP-1 cooperativity in macrophage responses to minimally modified LDL 48. The ATF-1 site immediately adjacent to a putative

Bach / Nrf2 site (Supplemental Figure XVIII), so cooperative binding would be plausible.

Moreover, ATF-1 and Nrf2 are structurally related, basic zipper transcription factors, raising the more speculative hypothesis of ATF-1 / Nrf2 heterodimerization. Extending the idea of

Circulation Research Manuscript #CIRCRESAHA/2011/247577 - R3 Regular Article

transcription factor cooperativity more globally, we found by bioinformatic prediction from

ChIP-Seq databases, overrepresentation of up to 10 transcription factor binding sites in the

Mhem geneset (Supplemental Figure XIX). This sets ATF-1 in a context of a more complex response comprising a network of several transcriptional factors that may modulate responses.

As far as can be assessed without a very large (and largely repetitious) microarray experiment, Mhem appear to be distinct to recognized polarization patterns (M1, M2, Mox).

Both M1 and M2 were HLA-DR-high activation states distinct to the HLA-DR-low Mhem state and either M1 or M2 commitment prevented acquisition of a full Mhem state. However, both M1 and M2 cells committed relative to one another could be dynamically converted to

HLA-DR-low by heme-treatment, suggesting that the activation-deactivation axis may show some plasticity. Interestingly, M1- or foam cell differentiation both enhanced heme-induced p-ATF-1 responses, while M2- or Mox-differentiation suppressed both basal and heme- stimulated p-ATF-1. Furthermore, two relevant antioxidants, N-acetyl-cysteine (a prototypical direct antioxidant, i.e. free radical scavenger) and curcumin (a prototypical indirect antioxidant and nutriceutical): both induced p-ATF-1 and prevented heme from inducing it any further. Although these data suggest that ATF-1 may also play a role in the mechanism of action of these important agents, testing this was beyond the scope of the current manuscript. Venn diagrams identified a large number of apparently Mhem specific genes, the most fold regulated and highly expressed was HTRA3, which could be assessed as a useful Mhem marker in future.

What kinases mediate ATF-1 regulation by heme ?

Although a study of heme-regulated ATF-1 kinases was beyond the scope of the present paper, heme-induced ATF-1 could be activated by several serine kinases: protein kinase A

Circulation Research Manuscript #CIRCRESAHA/2011/247577 - R3 Regular Article

(PKA) 14; several stress kinases including adenosine monophosphate activated kinase

(AMPK) in skeletal muscle during exercise 49, p38 mitogen activated protein kinase in leptin signalling 50, salt-inducible kinase / mitogen stress activated kinase (SIK / MSK) downstream of ultraviolet light 51, 52, and by protein kinase D 53. Tyrosine phosphorylation of ATF-1 by homeodomain interacting protein kinase (HIPK) has also been described in response to genotoxic stress 54. Understanding the role of upstream kinases in the regulation of Mhem is a major priority for further work because of its potential for experimental and eventual therapeutic modulation by small molecules.

Therapeutic potential

LXR-agonists are being actively developed to initiate reverse cholesterol transport 41; and

HO-1 is a target of interest for a variety of existing therapies 6. Considering that there is significant therapeutic interest in each of the mediators alone, our finding that ATF-1 coregulates HO-1 and LXR suggests that therapeutic manipulation of p-ATF-1 may provide a strong benefit / risk ratio in atherosclerosis. In this light, our findings that p-ATF-1 is induced by curcumin (a dietary flavonoid)55 and NAC (established treatment for paracetamol overdose and interstitial lung disease)56 is interesting.

Circulation Research Manuscript #CIRCRESAHA/2011/247577 - R3 Regular Article

Conclusions

In conclusion, Mhem is an atheroprotective macrophage state, distinct from M1, M2 and

Mox, that develops in specific adaptation to IPH. Induction of ATF-1 co-induces HO-1 and

LXR-β, interlinking lipid and iron handling for the first time. By acting as a novel transcriptional integrator of iron and lipid homeostasis, ATF-1 is a key driver of human plaque monocytes to acquire the atheroprotective Mhem macrophage state.

Circulation Research Manuscript #CIRCRESAHA/2011/247577 - R3 Regular Article

Acknowledgements

None

Sources of funding

We thank the British Heart Foundation for financial support (Gerry Turner Intermediate

Fellowship FS07/010 to JJB and Professorial Core Support to DOH) and thank the

Cambridge Biomedical Research Centre and Imperial College NIHR Biomedical Research

Centre for support.

Disclosures

None

Circulation Research Manuscript #CIRCRESAHA/2011/247577 - R3 Regular Article

Non standard abbreviations and non-standard acronyms used in the text

AMPK, adenosine monophosphate activated protein kinase

APF, aminophenylfluorescein

ATF-1, activating transcription factor 1

ChIP, chromatin immunoprecipitation

CRE, cyclic-3’-5’-adenosine-monophosphate (cAMP) response element

CREB, cyclic-3’-5’-adenosine-monophosphate (cAMP) response element binding protein

FC, foam cell

HO-1, heme oxygenase 1 (HMOX1) hROS, highly reactive oxygen species (hypochlorite, peroxynitrite, hydroxyl radicals)

IPH, intraplaque hemorrhage

LSM, laser scanning microscope

LXR-α, Liver X Receptor alpha

LXR-β, Liver X Receptor beta

M1, classic macrophages

M2 alternative macrophages

Mhem, hemorrhage-specialist macrophages

NAC, N-acetyl-cysteine

Oligos, oligonucleotides

OxLDL, oxidised low density lipoproteins

PEI, polyethyleneimine

PEI-Mann mannosylated polyethyleneimine

PKA, protein kinase A q-PCR quantitative polymerase chain reaction qChIP, quantitative chromatin immunoprecipitation

TMB, tetramethylbenzidine

Circulation Research Manuscript #CIRCRESAHA/2011/247577 - R3 Regular Article

Figure Legends

Figure 1. Microarray analysis defines Mhem human macrophages (A,B) histopathological images of macrophages in an area of hemorrhage within an atherosclerotic plaque. Macrophages are either foam cells laden with clear intracellular lipid (A), or siderophages laden with brown hemosiderin (B); Insets, cultured human blood-derived macrophage incubated with erythrocytes and then dual-stained for both lipid (red) and iron

(blue). The cells were either large and lipid-laden (red) or small and iron-laden (blue) replicating the in vivo findings. All scalebars = 100μm. (C, D) show the transcriptional response in peripheral blood-derived monocytes one and four hours after stimulation with heme (10μM), relating mean gene expression levels (X axis) to fold change (Y axis). The blue triangle at one hour shows the early moderate induction of low expression genes identified in Gene Ontology as mainly transcriptional regulators and signalling molecules.

The red box at four hours shows later higher level induction of higher expression genes identified as mainly effector molecules; (E) top 20 transcription factors by fold induction.

Genes identified in (C) were selected out using GO = ‘regulator of transcription, DNA dependent’ and plotted by relative rank of fold induction. The most strongly induced regulator was activating transcription factor 1 (ATF-1). ATF-1 was also the only such regulator in the top 10 that was annotated as a validated transcription factor

(http://www.ncbi.nlm.nih.gov/gene/466 ); (F) identity and fold induction of genes induced over 2-fold at four hours. Full names and annotation are given in Supplemental Tables I-III.

Circulation Research Manuscript #CIRCRESAHA/2011/247577 - R3 Regular Article

Figure 2. Heme induces HO-1 via ATF-1 (A) Validation by qPCR of the time-course of

ATF-1 and HO-1 expression after heme challenge, using monocytes from five separate donors (all values are mean + SE, n=5 donors). The time-lagged correlation coefficient between ATF-1 and HO-1 was 0.8, consistent with ATF-1 driving HO-1. (B-E) ATF-1 knockdown. Peripheral blood-derived monocytes were transfected with ATF-1 or control siRNA and then stimulated with heme (10μM for four hours) or vehicle (Veh). (B) ATF-1 knock-down was confirmed by qPCR. Data are representative of four independent experiments, each with five donors (n = 20 total); (C) shows HO-1 mRNA expression in the same samples, with reduction of HO-1 induction due to ATF-1 knock-down. *p<0.05, paired

Wilcoxon; (D) Western analyses showing that ATF-1 knockdown also suppressed heme- induced total ATF-1, phospho-ATF-1 and HO-1 protein levels. The blots are representative of n=12 independent experiments with different donors; (E) Quantification by western blots densitometry for p-ATF-1, ATF-1, HO-1, as indicated. *P<0.05, Student’s paired t-test for indicated comparison; (F) Validation of transfection: after 2 days culture, blood-derived monocytes were transfected. GFP = EGFP expression plasmid 1μg; PEI-mann = mannosylated polyethyleneimine transfection reagent 1μl; Complexes of pEGFP and PEI- mann were formed (Methods) before addition to culture. GFP fluorescence is found only in the cells given complexes; (G) ATF-1 transfection of peripheral blood-derived monocytes induced HO-1 mRNA. Indicated plasmids were added as PEI-Mann complexes at 1μg / well.

Data are mean ± SE of five donors; * p<0.05, Student’s paired t-test; (H) Western blotting showing dose-dependent induction of ATF-1 and HO-1 caused by ATF-1 transfection; (I)

ATF-1 transfection of peripheral blood-derived monocytes suppressed hROS release, as measured by APF (see methods). Note the lack of effect of transfection with enhanced GFP control plasmid (GFP); and the kinase-dead mutant of ATF-1 (DN-ATF-1) or the kinase-dead

Circulation Research Manuscript #CIRCRESAHA/2011/247577 - R3 Regular Article

mutant of CREB1 (DN-CREB) and the effect of CREB1 (CREB) in inducing hROS. Data are mean ± SE. *p<0.05 compared to GFP control.

Figure 3. Bioinformatic analysis of heme regulated macrophage genes (A) Gene ontology

(GO) analysis of transcripts upregulated by heme. The bars show genes upregulated one and four hours after heme stimulation. *significant enrichment with false discovery rate (FDR)

<0.05 (a p-value adjusted for multiple simultaneous comparisons used in microarray analysis). As might be expected, the primary response gene set was enriched for transcription factors, signalling and nucleotide metabolism. The secondary response gene set was enriched in genes coding for lipid handling, including lipid export and beta-oxidation pathways. The largest single category was genes that are still not annotated; (B) genes extracted from microarray data using GO term “regulation of DNA dependent RNA transcription” were essentially transcription factors, and ordered by fold induction by heme at 4h as indicated.

ATF-1 is the most induced transcription factor. (C) ATF-1-mediated gene regulatory network initiated by ATF-1 mediated co-induction of HO-1 and LXR-β. Red nodes = gene induced by heme. Green nodes = gene repressed by heme. Straight thick blue blue arrows = induces via cyclic-AMP response element (CRE). Curved thick blue arrow = self-inducing via CRE.

Curved thick orange arrow = self inducing via LXRE. Straight fine orange arrows = induces via LXRE, with either biochemical validation or bioinformatic prediction. Grey arrows = known dependences (PubMed) but exact mechanism uncertain. Green box = novel data.

Known interactions are in the orange (lipid) and red (heme) boxes. This was inferred in silico by bioinformatic analysis of microarray data, qPCR data, siRNA data (Supplemental

Methods). Of note, an ATF-1 site was identified upstream of ATF-1 itself, suggesting putative autoinduction.

Figure 4 ATF-1 induces LXR-β (A) Time-course of LXR-β mRNA expression in human blood-derived macrophages relation to ATF-1, following heme stimulation. ATF-1 and LXR-

Circulation Research Manuscript #CIRCRESAHA/2011/247577 - R3 Regular Article

β were induced in parallel waves with the ATF-1 wave leading the main LXR-β wave. A late secondary LXR-β wave is also seen, peaking at 48h. Time-lagged correlation coefficient =

0.85, p=0.014; (B, C) ATF-1 knockdown in peripheral blood-derived monocytes, as in Figure

2A, led to inhibition of LXR-β (C) mRNA and (D) protein induction by heme. *p<0.05, paired Wilcoxon; RNA Data are mean ± SE of five donors; Protein data are representative of n=12 independent experiments from different donors. (E) ATF-1 transfection (1 μg) induced

LXR-β mRNA. Monocytes were transfected with 1μg.mL-1 plasmid DNA, as indicated, and examined by qPCR for LXR-β, * p<0.05, Student’s paired t-test.

Figure 5. ATF-1 binds to and activates LXR-β and HO-1 enhancers. (A) Diagram of reporter constructs formed by fusing different lengths of the 5’-enhancer/promoter sequence from the HO-1 gene (hmox1) to drive the coding sequence from firefly luciferase (Luc); (B)

Human peripheral blood-derived monocytes were transfected with the human HO-1 luciferase reporter constructs indicated in (A), and then stimulated with heme (10μM) for four hours.

Data are mean ± SE of four donors. Light levels were normalized for cotransfected pGL3-

Renilla (control), and to respective untreated controls. Lengths are given in kb relative to the transcription start site (TSS). Effective heme-induction of luciferase is seen with the longest

(4.9kb) construct, indicating that the functional heme response element is between -2.2 to

4.9kb distal to the transcription start site; (C) cotransfection of blood-derived monocytes with

HO-1-luciferase reporter plasmids in (A) and (B) with an expression plasmid for ATF-1 produces the same pattern as challenge with heme, consistent with ATF-1 mediating HO-1- luc induction by heme, n=3 donors; (D) In the -4.9kb HO-1 luciferase plasmid, the indicated inactivating mutations of the ATF-1 response site at ∼-4kb (ΔATF-1) prevented HO-1-luc induction by 10μM heme at 4h in n=3 donors; (E) Chromatin immunoprecipitation (ChIP) analysis of heme-stimulated blood-derived monocytes, using anti-p-ATF-1 as in the diagram.

The figures show the specific fold enrichment at one and four hours of HO-1 and LXR-β

Circulation Research Manuscript #CIRCRESAHA/2011/247577 - R3 Regular Article

relative to IgG control and relative to control sequence (3’ end of ORF); (F) ATF-1 binding to oligonucleotides matching HO-1 or LXR-β regulatory sequences. Y-axis, specific immunodetection of p-ATF-1 bound to solid-phase oligonucleotides, as in diagram

(absorbance units, TMB peroxidase substrate). Cons ATF-1 = consensus ATF-1 response element (= consensus CRE) from the rat somatostatin promoter. X-axis, control or mutated oligos (sequences in Supplemental). * significant Student’s paired t-test, p<0.05. Data are mean ± SE of five donors.

Figure 6. Heme directs foam-cell protection via ATF-1 (A,B) Blood-derived monocytes were stimulated with heme, after which mRNA levels for LXR-β were related to those for

(A) ApoE or (B) LXR-α, ABCA1 and FASN (fatty acid synthase). ApoE succeeds and parallels LXR-β giving rise to a high time-lagged correlation coefficient (0.8). Likewise,

LXR-α and canonical target ABCA1 have a very tight late temporal correlation, with induction of 10-20 fold at 48h. LXR-β shows an early peak, as detected by microarray, then falls and has second peak at 48h coincident with the LXR-α and ABCA1 spikes. FASN was not induced at any point; (C) Heme-challenge and ATF-1 cotransfection produced equivalent induction of an LXR-α luciferase reporter construct at 48 hours; (D,E) peripheral blood- derived monocytes were transfected with ATF-1 siRNA and then challenged with heme followed by OxLDL (10μg.mL-1). We then measured cholesterol accumulation (D) and cholesterol efflux (E) *p<0.05, paired t-test; (F) ATF-1 transfection induced protection against OxLDL mediated foam cell formation. Human peripheral blood-derived monocytes were cultured for 48 hours, transfected with enhanced GFP or ATF-1. Cells were then incubated with OxLDL (hypochlorite-modified LDL 30μg.mL-1) for 96h, after which they were fixed, stained for lipid with Nile-Red and imaged by confocal microscopy. The images are representative of nine experiments. (G) counting of foam cells (Nile-Red bright macrophages over 20μm diameter) in nine monocyte donors. Y-axis, % of cells defined as

Circulation Research Manuscript #CIRCRESAHA/2011/247577 - R3 Regular Article

foam cells. X-axis, Ox-LDL = culture in 30μg.mL-1 hypochlorite-modified LDL. UTx = untransfected; GFP = transfected with pEGFP; ATF-1 = transfected with ATF-1; He = heme

10μM. All incubated for 6 days. * P<0.05 paired t-test relative to respective control (ATF-1 transfection -GFP) or (heme – vehicle). Data are mean ± SE of five donors.

Figure 7. Activated ATF-1 defines plaque Mhem macrophages (A) Size difference between Mhem and conventional inflammatory plaque macrophages (foam cells, FC).

Sections were dual stained with CD163 (blue) and HLA-DR (red) or stained with CD68

(brown). Scalebars = lengths indicated. The Mhem cells (CD163-high HLA-DR-low i.e. blue) are obviously smaller and have smaller amounts of less foamy cytoplasm. The smaller size of Mhem is borne out on image analysis (right hand side) *P<0.05, Student’s paired t- test; (B) phospho-ATF-1 is selectively found in nuclei of Mhem cells by confocal microscopy. Sections of hemorrhaged human atherosclerotic culprit lesions were incubated in anti-p-ATF-1 (red) followed by ant-rabbit alkaline phosphatase and Vector Red; counterstained with macrophage lysosomal marker anti-CD68-FITC (green); (C) Expression of p-ATF-1 target proteins, and relative quantification, in Mhem cells relative to foam cells;

(D) Coexpression of p-ATF-1 and key effectors HO-1 and ABCA-1 by Mhem cells but not foam cells, using multichannel confocal. Conditions for FC and Mhem images are equivalent; channels are as indicated. Staining for LXR-β (which was only transiently 2-fold regulated in culture) is shown in Supplemental Figure XX. Corroborative quantification of HO-1 from images of brightfield immunohistochemistry with peroxidase / DAB is shown in

Supplemental Figure XXI.

Figure 8 Schema of macrophage regulation following erythrophagocytosis.

Macrophages ingest erythrocytes (RBC), which are then trafficked to the lysosome (Ly) 8.

Within the lysosome, heme is released causing ‘heme stress’; which is associated with

Circulation Research Manuscript #CIRCRESAHA/2011/247577 - R3 Regular Article

oxidative stress 8, but causes a more complex response (Figure XIII). Heme stress induces, and causes phosphorylation of, ATF-1. Heme stress activates Nrf2 in parallel 8. ATF-1 translocates to the nucleus (blue arrows) and binds to the enhancers of key transcription factors (HO-1 and LXR) and drives transcription of genes (fine arrows) leading to heme disposal into ferritin (HO-1 / Fe pathway), releasing biliverdin (green) and bilirubin (orange) to diffuse out the cell. In parallel, (second gene icon) ATF-1 induces cholesterol export via

ABCA-1 (membrane cylinder), driving local HDL synthesis and reverse cholesterol transport

(HDL is essentially cholesterol docked onto ApoA). Thus, iron and lipid handling are linked via ATF-1. This explains the histological dichotomy between lipid-laden macrophages and iron-laden macrophages, and also points to key therapeutic avenues linking antioxidant protection, anti-inflammatory protection and foam cell protection.

Circulation Research Manuscript #CIRCRESAHA/2011/247577 - R3 Regular Article

Reference List

(1) Glass CK, Witztum JL. Atherosclerosis. the road ahead. Cell 2001;104(4):503-16.

(2) Kolodgie FD, Gold HK, Burke AP, Fowler DR, Kruth HS, Weber DK, Farb A, Guerrero LJ, Hayase M, Kutys R, Narula J, Finn AV, Virmani R. Intraplaque hemorrhage and progression of coronary atheroma. N Engl J Med 2003;349(24):2316- 25.

(3) Sadrzadeh SM, Graf E, Panter SS, Hallaway PE, Eaton JW. Hemoglobin. A biologic fenton reagent. J Biol Chem 1984;259(23):14354-6.

(4) Daugherty A, Rateri DL, Lu H. As macrophages indulge, atherosclerotic lesions bulge. Circ Res 2008;102(12):1445-7.

(5) Levy AP, Levy JE, Kalet-Litman S, Miller-Lotan R, Levy NS, Asaf R, Guetta J, Yang C, Purushothaman KR, Fuster V, Moreno PR. Haptoglobin genotype is a determinant of iron, lipid peroxidation, and macrophage accumulation in the atherosclerotic plaque. Arterioscler Thromb Vasc Biol 2007;27(1):134-40.

(6) Otterbein LE, Soares MP, Yamashita K, Bach FH. Heme oxygenase-1: unleashing the protective properties of heme. Trends Immunol 2003;24(8):449-55.

(7) Boyle JJ, Harrington HA, Piper E, Elderfield K, Stark J, Landis RC, Haskard DO. Coronary intraplaque hemorrhage evokes a novel atheroprotective macrophage phenotype. Am J Pathol 2009;174(3):1097-108.

(8) Boyle JJ, Johns M, Lo J, Chiodini A, Ambrose N, Evans PC, Mason JC, Haskard DO. Heme Induces Heme Oxygenase 1 via Nrf2: Role in the Homeostatic Macrophage Response to Intraplaque Hemorrhage. Arterioscler Thromb Vasc Biol 2011.

(9) Ito K, Yamamura S, Essilfie-Quaye S, Cosio B, Ito M, Barnes PJ, Adcock IM. Histone deacetylase 2-mediated deacetylation of the glucocorticoid receptor enables NF-kappaB suppression. J Exp Med 2006;203(1):7-13.

(10) Yvan-Charvet L, Ranalletta M, Wang N, Han S, Terasaka N, Li R, Welch C, Tall AR. Combined deficiency of ABCA1 and ABCG1 promotes foam cell accumulation and accelerates atherosclerosis in mice. J Clin Invest 2007;117(12):3900-8.

(11) Becker L, Gharib SA, Irwin AD, Wijsman E, Vaisar T, Oram JF, Heinecke JW. A macrophage sterol-responsive network linked to atherogenesis. Cell Metab 2010;11(2):125-35.

(12) Ramsey SA, Klemm SL, Zak DE, Kennedy KA, Thorsson V, Li B, Gilchrist M, Gold ES, Johnson CD, Litvak V, Navarro G, Roach JC, Rosenberger CM, Rust AG, Yudkovsky N, Aderem A, Shmulevich I. Uncovering a macrophage transcriptional program by integrating evidence from motif scanning and expression dynamics. PLoS Comput Biol 2008;4(3):e1000021.

Circulation Research Manuscript #CIRCRESAHA/2011/247577 - R3 Regular Article

(13) Kadl A, Meher AK, Sharma PR, Lee MY, Doran AC, Johnstone SR, Elliott MR, Gruber F, Han J, Chen W, Kensler T, Ravichandran KS, Isakson BE, Wamhoff BR, Leitinger N. Identification of a novel macrophage phenotype that develops in response to atherogenic phospholipids via Nrf2. Circ Res 2010;107(6):737-46.

(14) Liu F, Thompson MA, Wagner S, Greenberg ME, Green MR. Activating transcription factor-1 can mediate Ca(2+)- and cAMP-inducible transcriptional activation. J Biol Chem 1993;268(9):6714-20.

(15) Letarte PB, Lieberman K, Nagatani K, Haworth RA, Odell GB, Duff TA. Hemin: levels in experimental subarachnoid hematoma and effects on dissociated vascular smooth-muscle cells. J Neurosurg 1993;79(2):252-5.

(16) Eijgelaar WJ, Horrevoets AJ, Bijnens AP, Daemen MJ, Verhaegh WF. Equivalence testing in microarray analysis: Similarities in the transcriptome of human atherosclerotic and non-atherosclerotic macrophages. Physiol Genomics 2010.

(17) Lukk M, Kapushesky M, Nikkila J, Parkinson H, Goncalves A, Huber W, Ukkonen E, Brazma A. A global map of human gene expression. Nat Biotechnol 2010;28(4):322- 4.

(18) Dunham C, Hussong J, Seiff M, Pfeifer J, Perry A. Primary intracerebral angiomatoid fibrous histiocytoma: report of a case with a t(12;22)(q13;q12) causing type 1 fusion of the EWS and ATF-1 genes. Am J Surg Pathol 2008;32(3):478-84.

(19) Tewari M, Hu PJ, Ahn JS, yivi-Guedehoussou N, Vidalain PO, Li S, Milstein S, Armstrong CM, Boxem M, Butler MD, Busiguina S, Rual JF, Ibarrola N, Chaklos ST, Bertin N, Vaglio P, Edgley ML, King KV, Albert PS, Vandenhaute J, Pandey A, Riddle DL, Ruvkun G, Vidal M. Systematic interactome mapping and genetic perturbation analysis of a C. elegans TGF-beta signaling network. Mol Cell 2004;13(4):469-82.

(20) Rehfuss RP, Walton KM, Loriaux MM, Goodman RH. The cAMP-regulated enhancer-binding protein ATF-1 activates transcription in response to cAMP- dependent protein kinase A. J Biol Chem 1991;266(28):18431-4.

(21) Hanson RW. Biological role of the isoforms of C/EBP minireview series. J Biol Chem 1998;273(44):28543.

(22) Leahy P, Crawford DR, Grossman G, Gronostajski RM, Hanson RW. CREB binding protein coordinates the function of multiple transcription factors including nuclear factor I to regulate phosphoenolpyruvate carboxykinase (GTP) gene transcription. J Biol Chem 1999;274(13):8813-22.

(23) Wang Y, Inoue H, Ravnskjaer K, Viste K, Miller N, Liu Y, Hedrick S, Vera L, Montminy M. Targeted disruption of the CREB coactivator Crtc2 increases insulin sensitivity. Proc Natl Acad Sci U S A 2010;107(7):3087-92.

(24) Kim SJ, Nian C, McIntosh CH. Glucose-dependent insulinotropic polypeptide (GIP) increases human adipocyte lipoprotein lipase (LPL) expression through cyclic AMP (cAMP) response element binding protein (CREB) and cAMP-responsive CREB coactivator 2 (TORC2) mediated trans-activation of the LPL gene. J Lipid Res 2010.

Circulation Research Manuscript #CIRCRESAHA/2011/247577 - R3 Regular Article

(25) Ruffell D, Mourkioti F, Gambardella A, Kirstetter P, Lopez RG, Rosenthal N, Nerlov C. A CREB-C/EBPbeta cascade induces M2 macrophage-specific gene expression and promotes muscle injury repair. Proc Natl Acad Sci U S A 2009;106(41):17475-80.

(26) Zucman J, Delattre O, Desmaze C, Epstein AL, Stenman G, Speleman F, Fletchers CD, Aurias A, Thomas G. EWS and ATF-1 gene fusion induced by t(12;22) translocation in malignant melanoma of soft parts. Nat Genet 1993;4(4):341-5.

(27) Waters BL, Panagopoulos I, Allen EF. Genetic characterization of angiomatoid fibrous histiocytoma identifies fusion of the FUS and ATF-1 genes induced by a chromosomal translocation involving bands 12q13 and 16p11. Cancer Genet Cytogenet 2000;121(2):109-16.

(28) Chan CP, Chang MC, Wang YJ, Chen LI, Tsai YL, Lee JJ, Jia HW, Jeng JH. Thrombin activates Ras-CREB/ATF-1 signaling and stimulates c-fos, c-jun, and c- myc expression in human gingival fibroblasts. J Periodontol 2008;79(7):1248-54.

(29) Ghosh SK, Gadiparthi L, Zeng ZZ, Bhanoori M, Tellez C, Bar-Eli M, Rao GN. ATF- 1 mediates protease-activated receptor-1 but not receptor tyrosine kinase-induced DNA synthesis in vascular smooth muscle cells. J Biol Chem 2002;277(24):21325-31.

(30) Iwasaki K, Hailemariam K, Tsuji Y. PIAS3 interacts with ATF1 and regulates the human ferritin H gene through an antioxidant-responsive element. J Biol Chem 2007;282(31):22335-43.

(31) Schopfer FJ, Lin Y, Baker PR, Cui T, Garcia-Barrio M, Zhang J, Chen K, Chen YE, Freeman BA. Nitrolinoleic acid: an endogenous peroxisome proliferator-activated receptor gamma ligand. Proc Natl Acad Sci U S A 2005;102(7):2340-5.

(32) Greaves DR, Gordon S. Thematic review series: the immune system and atherogenesis. Recent insights into the biology of macrophage scavenger receptors. J Lipid Res 2005;46(1):11-20.

(33) Hansson GK, Libby P. The immune response in atherosclerosis: a double-edged sword. Nat Rev Immunol 2006;6(7):508-19.

(34) Wang X, Collins HL, Ranalletta M, Fuki IV, Billheimer JT, Rothblat GH, Tall AR, Rader DJ. Macrophage ABCA1 and ABCG1, but not SR-BI, promote macrophage reverse cholesterol transport in vivo. J Clin Invest 2007;117(8):2216-24.

(35) Chawla A, Boisvert WA, Lee CH, Laffitte BA, Barak Y, Joseph SB, Liao D, Nagy L, Edwards PA, Curtiss LK, Evans RM, Tontonoz P. A PPAR gamma-LXR-ABCA1 pathway in macrophages is involved in cholesterol efflux and atherogenesis. Mol Cell 2001;7(1):161-71.

(36) Lin HL, Xu XS, Lu HX, Zhang L, Li CJ, Tang MX, Sun HW, Liu Y, Zhang Y. Pathological mechanisms and dose dependency of erythrocyte-induced vulnerability of atherosclerotic plaques. J Mol Cell Cardiol 2007;43(3):272-80.

(37) Im SS, Osborne TF. Liver x receptors in atherosclerosis and inflammation. Circ Res 2011;108(8):996-1001.

Circulation Research Manuscript #CIRCRESAHA/2011/247577 - R3 Regular Article

(38) Wang N, Ranalletta M, Matsuura F, Peng F, Tall AR. LXR-induced redistribution of ABCG1 to plasma membrane in macrophages enhances cholesterol mass efflux to HDL. Arterioscler Thromb Vasc Biol 2006;26(6):1310-6.

(39) Terasaka N, Wang N, Yvan-Charvet L, Tall AR. High-density lipoprotein protects macrophages from oxidized low-density lipoprotein-induced apoptosis by promoting efflux of 7-ketocholesterol via ABCG1. Proc Natl Acad Sci U S A 2007;104(38):15093-8.

(40) Hu T, Foxworthy P, Siesky A, Ficorilli JV, Gao H, Li S, Christe M, Ryan T, Cao G, Eacho P, Michael MD, Michael LF. Hepatic peroxisomal fatty acid beta-oxidation is regulated by liver X receptor alpha. Endocrinology 2005;146(12):5380-7.

(41) Calkin AC, Tontonoz P. Liver x receptor signaling pathways and atherosclerosis. Arterioscler Thromb Vasc Biol 2010;30(8):1513-8.

(42) Bischoff ED, Daige CL, Petrowski M, Dedman H, Pattison J, Juliano J, Li AC, Schulman IG. Non-redundant roles for LXRalpha and LXRbeta in atherosclerosis susceptibility in low density lipoprotein receptor knockout mice. J Lipid Res 2010;51(5):900-6.

(43) Wrobel J, Steffan R, Bowen SM, Magolda R, Matelan E, Unwalla R, Basso M, Clerin V, Gardell SJ, Nambi P, Quinet E, Reminick JI, Vlasuk GP, Wang S, Feingold I, Huselton C, Bonn T, Farnegardh M, Hansson T, Nilsson AG, Wilhelmsson A, Zamaratski E, Evans MJ. Indazole-based liver X receptor (LXR) modulators with maintained atherosclerotic lesion reduction activity but diminished stimulation of hepatic triglyceride synthesis. J Med Chem 2008;51(22):7161-8.

(44) Lund EG, Peterson LB, Adams AD, Lam MH, Burton CA, Chin J, Guo Q, Huang S, Latham M, Lopez JC, Menke JG, Milot DP, Mitnaul LJ, Rex-Rabe SE, Rosa RL, Tian JY, Wright SD, Sparrow CP. Different roles of liver X receptor alpha and beta in lipid metabolism: effects of an alpha-selective and a dual agonist in mice deficient in each subtype. Biochem Pharmacol 2006;71(4):453-63.

(45) Sindrilaru A, Peters T, Wieschalka S, Baican C, Baican A, Peter H, Hainzl A, Schatz S, Qi Y, Schlecht A, Weiss JM, Wlaschek M, Sunderkotter C, Scharffetter-Kochanek K. An unrestrained proinflammatory M1 macrophage population induced by iron impairs wound healing in humans and mice. J Clin Invest 2011.

(46) Matus M, Schulte J, Huyen S, Seidl M, Schutz G, Schmitz W, Muller FU. Knockout of ATF1 leads to enhanced cardiac contractility and output. FASEB J 2008;22:1155.14.

(47) Chinetti-Gbaguidi G, Baron M, Bouhlel MA, Vanhoutte J, Copin C, Sebti Y, Derudas B, Mayi T, Bories G, Tailleux A, Haulon S, Zawadzki C, Jude B, Staels B. Human Atherosclerotic Plaque Alternative Macrophages Display Low Cholesterol Handling but High Phagocytosis Because of Distinct Activities of the PPAR{gamma} and LXR{alpha} Pathways. Circ Res 2011;108(8):985-95.

(48) Wiesner P, Choi SH, Almazan F, Benner C, Huang W, Diehl CJ, Gonen A, Butler S, Witztum JL, Glass CK, Miller YI. Low doses of lipopolysaccharide and minimally

Circulation Research Manuscript #CIRCRESAHA/2011/247577 - R3 Regular Article

oxidized low-density lipoprotein cooperatively activate macrophages via nuclear factor kappa B and activator protein-1: possible mechanism for acceleration of atherosclerosis by subclinical endotoxemia. Circ Res 2010;107(1):56-65.

(49) Thomson DM, Herway ST, Fillmore N, Kim H, Brown JD, Barrow JR, Winder WW. AMP-activated protein kinase phosphorylates transcription factors of the CREB family. J Appl Physiol 2008;104(2):429-38.

(50) Cui H, Cai F, Belsham DD. Leptin signaling in neurotensin neurons involves STAT, MAP kinases ERK1/2, and p38 through c-Fos and ATF1. FASEB J 2006;20(14):2654-6.

(51) Staples CJ, Owens DM, Maier JV, Cato AC, Keyse SM. Cross-talk between the p38alpha and JNK MAPK pathways mediated by MAP kinase phosphatase-1 determines cellular sensitivity to UV radiation. J Biol Chem 2010;285(34):25928-40.

(52) Wiggin GR, Soloaga A, Foster JM, Murray-Tait V, Cohen P, Arthur JS. MSK1 and MSK2 are required for the mitogen- and stress-induced phosphorylation of CREB and ATF1 in fibroblasts. Mol Cell Biol 2002;22(8):2871-81.

(53) Johannessen M, Delghandi MP, Rykx A, Dragset M, Vandenheede JR, Van LJ, Moens U. Protein kinase D induces transcription through direct phosphorylation of the cAMP-response element-binding protein. J Biol Chem 2007;282(20):14777-87.

(54) Hailemariam K, Iwasaki K, Huang BW, Sakamoto K, Tsuji Y. Transcriptional regulation of ferritin and antioxidant genes by HIPK2 under genotoxic stress. J Cell Sci 2010;123(Pt 22):3863-71.

(55) Balogun E, Hoque M, Gong P, Killeen E, Green CJ, Foresti R, Alam J, Motterlini R. Curcumin activates the haem oxygenase-1 gene via regulation of Nrf2 and the antioxidant-responsive element. Biochem J 2003;371(Pt 3):887-95.

(56) Bast A, Weseler AR, Haenen GR, den Hartog GJ. Oxidative stress and antioxidants in interstitial lung disease. Curr Opin Pulm Med 2010;16(5):516-20.

Circulation Research Manuscript #CIRCRESAHA/2011/247577 - R3 Regular Article

Novelty and Significance

What is known ?

• Heme induces heme oxygenase 1 (HO-1), which is an enzyme that catabolises heme,

forming a key negative feedback loop.

• Induction of HO-1 by external stimuli has important antioxidant, anti-inflammatory,

and cytoprotective effects.

• The Liver X receptor alpha isoform (LXR-α) protects against lipid overload in

macrophages by activating genes for cholesterol export and also has anti-

inflammatory activities.

What new information does this article contribute ?

• Heme induces HO-1 via activating transcription factor 1 (ATF-1), in a new regulatory

mechanism for this important homeostatic loop.

• ATF-1 couples induction of HO-1 to that of an LXR-β →LXR-α pathway.

• ATF-1 thereby coordinates iron and lipid metabolism, which were previously seen as

independent, providing joint protection from oxidant stress and lipid overload.

• This ATF-1/HO-1/LXR coupling allows heme to induce a novel phenotypic state

(Mhem) in macrophages, specialised for hemorrhage clearance, atheroprotective /

anti-inflammatory activity and distinct to previously defined phenotypic states.

• Coronary plaques in patients have hemorrhage-associated macrophages (Mhem) that

exhibit coordinated expression of ATF-1, HO-1, LXR, and LXR target genes.

• Therapeutic control of ATF-1 is possible and could combine the activities of HO-1

and LXR to preventively generate Mhem atheroprotective macrophages.

Circulation Research Manuscript #CIRCRESAHA/2011/247577 - R3 Regular Article

Summary paragraph

Rationale: Heme induces heme oxygenase 1 (HO-1), which removes excess heme as part of normal homeostasis. Moreover, HO-1 induction by external stimuli is vasoprotective.

Macrophages in the vicinity of intraplaque hemorrhage show atheroprotective features even in the face of iron loading, in spite of iron being potentially pro-inflammatory. Generally, macrophages clearing hemorrhage are seen storing either lipid or iron. Summary of salient findings: We here show for the first time that heme induces HO-1 via activating transcription factor 1 (ATF-1). Moreover, ATF-1 couples induction of HO-1 to that of Liver X receptors

(LXRs), promoting foam cell protection and cholesterol export to high density lipoproteins

(HDL). ATF-1 is therefore the first known example of coordination of iron and lipid metabolism. Via ATF-1/HO-1/LXR, heme induces a novel phenotypic state in macrophages, specialised for atheroprotective / anti-inflammatory activity, hemorrhage clearance and distinct to alternately activated macrophages or Mox. Patients with fatal coronary artery disease have macrophages with coordinated expression of ATF-1 and its response genes we here define, indicating that the novel ATF-1-(HO-1/LXR) network is likely have key pathophysiological significance. Significance: Therapeutic control of ATF-1, shown here for

2 candidates, could powerfully combine the atheroprotective activities of HO-1 and LXR, switching macrophages to atheroprotection.

(180 words)

Circulation Research Manuscript #CIRCRESAHA/2011/247577 - R3 Regular Article Figure 1

A Lipid laden macrophages B Hemosiderin-laden macrophages

100μm 100μm

100μm 100μm Lipid / iron Lipid / iron

C 1h D 4h Effector Transcriptional enzymes regulators

ATF1 HMOX1 HTRA3 E CNOT6L F GCLM PBX2 BRWD2 HPCAL FUBP3 AP1S2 RASSF7 SLC35B4 ZNF593 ELOVL5 UBA1 TAF9B OSGIN1 POLR2I SERPINE1 SSSR3 RBM14 PPP1C BZW1 APOL6 SLC39A8 ZNF684 Effector Enzyme SLC48A1 ATF2 TXNDC4 Transcriptional regulator Transcriptional PML HNRNPH1 GCLM SRA1 CPVL SIRT2 ARRDC3 RARA ACOX1 COPE 123 2468101214161820 Fold induction at 1h Fold Induction at 4h Circulation Research Manuscript #CIRCRESAHA/2011/247577 - R3 Regular Article Figure 2

12 ATF-1 4 C 8 A HO-1 B 10 6 3 8 4 6 2

mRNA Fold mRNA 4 2 1 HO-1 mRNA fold mRNA HO-1

2 fold mRNA ATF-1 * 1 * 1 0 0 1 10 100 veh heme veh veh heme veh Log time (h) siRNA: siRNA: Control ATF-1 Control ATF-1

veh heme veh siRNA: E D Control Control ATF-1 ATF-1 0.3 p-ATF-1 (∼37kDa) * ATF-1

(∼37kDa) actin P-ATF-1 / / P-ATF-1 HO-1 (∼27kDa) Actin-β 0.5 (∼55kDa) * actin ATF-1 /

F G 1.6 1.6 * GFP+PEI-mann *

GFP actin PEI-Mann / HO-1 Untreated 1 veh heme veh siRNA:

HO-1 mRNA Fold mRNA HO-1 Control ATF-1 0 GFP-fluorescence GFP ATF-1

ATF-1 GFP I * H Plasmid 160 μ g 00.1 0.3 1.0 0.1 0.3 1.0 NS

HO-1 100 * β-actin

GFP ATF-1

Plasmid Normalised hROS μ g 00.1 0.3 1.0 0.1 0.3 1.0 ATF-1 0 DN- DN- β-actin UTx GFP ATF-1 CREB ATF-1 CREB Circulation Research Manuscript #CIRCRESAHA/2011/247577 - R3 Regular Article Figure 3

A Secondary Response (1h + 4h ++) GO terms Primary Response (1h++ 4h -) Unannotated Lipid metabolism Intracellular trafficking * Nucleotide metabolism Signaling Transcriptional regulator 0 20 40 60 % genes

Transcriptional regulators B Fold change (4h)

POLR2I SIRT2 LXR-β ZNF580 ATF-1

1 2 3 Fold induction

C Gene regulatory network

CRE Lipid catabolism and export Iron CRE CRE storage β

LXRE α Circulation Research Manuscript #CIRCRESAHA/2011/247577 - R3 Regular Article Figure 4

A 7 B 3 C 5 6 ATF-1 * LXR-β 4 * * 2 4 3 mRNA Fold mRNA β mRNA Fold mRNA 2 2 1 ATF-1 mRNA fold mRNA ATF-1 LXR- 1 1 0 0 0 1 10 100 veh heme veh veh heme veh Time after heme stimulation (h) control ATF-1 control ATF-1

D Heme E siRNA: control control ATF-1 ATF-1 * p-ATF-1 (∼37kDa) 2.5 LXR-β (∼55kDa) 2 Actin-β (∼55kDa)

0.5 (fold) mRNA

β 1 / β * LXR- LXR- -actin ratio

β 0 GFP ATF-1

0 Plasmid Circulation Research Manuscript #CIRCRESAHA/2011/247577 - R3 Regular Article Figure 5

A B C

Luc 0 0

HO-1 enhancer -0.3kb Luc -0.3 -0.3

HO-1 enhancer -2.2kb Luc -2.2 -2.2 * HO-1 enhancer -5kb *

Luc from TSS Distance -4.9 Distance TSS from -4.9

012345678 01234 Fold Luciferase Fold Luciferase

D 14 * HuHO-1 -3978: TGCTGC GTCA TGTT 12 Homol TGCTG GTCA TG T 10 MuHO1 -3923: TGCTGT GTCA TGGT 8 6 ΔATF-1: TGCGTA GATC TGTT 4 Fold Luciferase Fold 2 1 control -4.9 -4.9ΔATF-1 Hu HO-1 enhancer (kb) E * 10 * Bead 20 8 Anti-p- ATF-1 6 10 4 enhancer -1.2kb β 2 ChIP Erichment Fold ChIP Fold Enrichment

Native chromatin -4kb HO-1enhancer of 1

of LXR- of 1 0 14 0 1 4 Time since heme (h) Time since heme (h)

F 0.15 Heme * Control

HRP 0.10 * NS Anti-p- 0.05 NS ATF-1

0 pATF-1-DNA Binding (A650) Binding pATF-1-DNA HO-1 HO-1 LXRβ LXRβ Cons scram Biotinylated Oligos ΔATF-1 ΔATF-1 ATF-1 Circulation Research Manuscript #CIRCRESAHA/2011/247577 - R3 Regular Article Figure 6

20 A 5 ApoE 16 LXR-β C 4 LXR-β B 16 LXR-α 4 14 ABCA1 3 12 3 FASN 10 2 2 8

mRNA fold fold mRNA 6 mRNA Fold mRNA 1 4 1

2 Luciferase Fold 0 1 0 1 10 100 0 1 10 100 Veh heme GFP ATF-1 Time since heme (log h) Time since heme (log h) D E 4000 * 40 *

10 (fluorescence units) (fluorescence (% effluxed(% 24h) over Cholesterol Export (%) Export Cholesterol Total cellularTotal cholesterol 0 0 Veh heme Veh Veh heme Veh Control-siRNA ATF-1-siRNA Control-siRNA ATF-1-siRNA

G F GFP ATF-1 50 * * Control 40

OxLDL % Foam Cells 10 0 UTx GFP ATF1 He UTx GFP ATF1 He OxLDL Circulation Research Manuscript #CIRCRESAHA/2011/247577 - R3 Regular Article Figure 7 ) 2 400 * m

Mhem Mhem FC μ A Mhem 300 FC 200 FC 100 200μm 100μm 100μm 200μm Area per cell per Area cell ( 0 Mhem FC Foam cell Mhem B

CD68 P-ATF-1 PhasePhase

C 20 * 5) 15

HO-1 10 Nuc

PhasePhase (AFU x10 5 HO-1 Intensity 0 FC Mhem * 10 ) ABCA1 5 8 Nuc 6 PhasePhase 4

(AFU x 10 2 ABCA1 intensity ABCA1 intensity 0 FC Mhem 10 * ) 5 8 ApoE Nuc 6 PhasePhase 4 (AFU x10

ApoE intensity 2 0 FC Mhem

D ABCA1 PhasePhase ATF-1 ABCA1 ATF-1 PhasePhase

M-hem FC

HO-1 Merge HO-1 Merge Circulation Research Manuscript #CIRCRESAHA/2011/247577 - R3 Regular Article Figure 8

HDL Bilirubin

ABCA1 Cholesterol

Ly Biliverdin RBC HO-1

Fe Intracellular free heme “heme stress” Nrf2

Nuc ATFATF--11 pATFpATF--11