A leukocyte-mimetic MRI contrast agent homes rapidly to activated endothelium and tracks with atherosclerotic lesion macrophage content

Abstract

Objective

Endothelial cell activation is an important mediator of monocyte recruitment to sites of vascular inflammation. We hypothesised that high-affinity dual-ligand microparticles of iron oxide, targeted to P-selectin and vascular cell adhesion molecule-1 (VCAM-1) (PV-MPIO), would identify activated endothelial cells during atherosclerosis progression.

Methods and Results

In vivo MRI in apolipoprotein E-deficient mice showed rapid binding of PV-MPIO to the aortic root, which was maximal 30 minutes post-MPIO injection and maintained at 60 minutes. Minimal binding was observed for control IgG-MPIO. Intensely low MR signal areas, corresponding to PV-MPIO binding, were detected early (14 weeks), during foam cell formation. Contrast effects increased at 20 weeks during fibrofatty lesion development (P < 0.05), but reduced by 30 weeks (P < 0.01). Across all lesion severities, MRI contrast effects correlated with lesion macrophage area quantified by immunohistochemistry (R = 0.53; P < 0.01). Near-infrared fluorescently labeled PV-MPIO were shown, by flow cytometry, to bind only activated endothelial cells, and not to macrophages. Using en face immunofluorescence, we further demonstrate selective PV-MPIO accumulation at atherosclerosis-susceptible sites, with minimal binding to atherosclerosis-spared regions.

Conclusions

This high affinity leukocyte mimetic MRI agent reveals endothelial activation. PV-MPIO demonstrate exceptionally rapid in vivo steady state accumulation, providing conspicuous MR contrast effects that can be objectively quantified. In atherosclerosis progression, PV-MPIO tracked closely with the burden and distribution of plaque macrophages, not merely plaque size. On a biocompatible platform, this approach has potential for quantitative MRI of inflammatory disease activity.

Correspondence: Professor Robin Choudhury Department of Cardiovascular Medicine, Level 6 West Wing, John Radcliffe Hospital, Oxford. OX3 9DU United Kingdom Telephone: +44-1865-234663 Fax: +44-1865-234667 robin.choudhury@cardiov.ox.ac.uk

Introduction

Inflammation, notably macrophage infiltration, is an important determinant in the pathogenesis of atherosclerosis.^{1, 2} Macrophages are involved in all stages of atherosclerotic lesion development and may trigger clinical events such as myocardial infarction or stroke by promoting fibrous cap degradation and plaque disruption.^1,3,4-6 Conversely, interventions that regress atherosclerosis and stabilize plaques have been associated with reduced inflammation and a diminution in plaque macrophage content.^{7, 8} There is increasing evidence that the lesion macrophage population is not static, but is involved in ongoing influx and excursion.⁹ Despite the critical role played by macrophages, non-invasive MRI techniques for their accurate quantification are still imperfect.¹⁰

Monocyte recruitment to the vascular wall is promoted by upregulation of endothelial adhesion molecules such as vascular cell adhesion molecule-1 (VCAM-1; CD106) and P-selectin (CD62P) at atherosclerosis-prone sites.^11-13 Initial monocyte-endothelial interactions are mediated by P-selectin, which stimulates monocyte rolling along the activated endothelium,¹³ whereas firm adhesion of monocytes depends on the engagement of integrin α4β1 (also termed very late antigen-4, VLA-4) with endothelial VCAM-1, preceding their transmigration to the nascent lesion.^{14, 15} Appreciation of these mechanisms and their relevance to atherogenesis lays a foundation for the design of molecular imaging probes that can determine, non-invasively, in vivo, the propensity of regional arterial endothelial activation to attract monocytes.^16-18

To image vascular endothelial cell activation with magnetic resonance imaging (MRI), we previously developed prototypic 4.5 μm diameter microparticles of iron oxide (MPIO) conjugated to antibodies against both VCAM-1 and P-selectin.¹⁹ Although this approach proved promising in feasibility experiments, demonstrated by ex vivo MRI of explanted mouse aortas (with advanced atherosclerosis), retention of MPIO was insufficient at acceptable iron doses for reliable in vivo molecular MRI. To overcome this limitation, we have developed a second-generation of smaller (1.0 μm) MPIO that have higher surface area to volume ratio for polyvalent ligand conjugation and which, we hypothesized, would be less buoyant under conditions of flow and high shear stress. These micron size-range particles should be distinguished from the targeted²⁰ or untargeted^{21, 22} nano-scale particles that have more commonly been used for atherosclerosis imaging. Compared to nano-scale particles, MPIO offer distinct advantages: (a) the payload of iron and, therefore, sensitivity is high;^{23, 24} (b) the clearance of MPIO from circulation is very rapid so “background” blood phase contrast is minimal;²⁵ (c) the obligate intravascular MPIO are much more tractable for endothelial molecular imaging than nanoparticles, which are susceptible to passive accumulation^{26, 27} and (d) they are readily functionalized allowing conjugation of one or more high valency targeting ligands.^28-30

Accordingly, we have developed a leukocyte mimetic contrast agent, based on size and surface ligands, which targets both VCAM-1 and P-selectin. We test the extent to which dual-ligand leukocyte mimetic MPIO home to activated endothelium and reflect inflammatory cell content in vivo across a range of atherosclerotic lesion complexities in apolipoprotein E^−/− mice. We further determine cellular binding patterns of dual-ligand MPIO in regions of the aorta that are susceptible to atherosclerotic lesion development. In so doing, we have sought to use the leukocyte mimetic properties of MPIO to map vascular inflammation in atherosclerosis.

Methods

A detailed Supplemental Methods section is available online at http://atvb.ahajournals.org.

Preparation of MPIO

Rat anti-mouse monoclonal VCAM-1 (clone M/K2) (Cambridge Bioscience Ltd, UK) and P-selectin antibodies (CD62P clone RB40.34) (BD Biosciences, UK) were covalently conjugated to the surface of tosyl activated MPIO (1 μm diameter) (Invitrogen, UK) in a 50:50 ratio to produce dual-ligand MPIO (PV-MPIO) as previously described.^{26, 31} Isotype control rat IgG-1 antibody (clone Lo-DNP-1) (Serotec, UK) conjugated MPIO (IgG-MPIO) were also prepared. For fluorescence imaging using en face immunofluorescence and flow cytometry, near infrared fluorescently labeled dual-ligand MPIO were developed. P-selectin and VCAM-1 monoclonal antibodies (50:50 mix) were combined with a near infra-red Alexa Fluor 750 dye for 60 minutes at RT according to the Small Animal In Vivo Imaging (SAIVI™) Alexa Fluor 750 antibody labeling kit (Invitrogen), which is azide-free and thus suitable for in vivo applications. Approximately two Alexa Fluor 750 molecules were coupled to each antibody molecule, according to the manufacturers protocol. The SAIVI™ 750 labeled P-selectin and VCAM-1 antibody mix was purified by gel column filtration and then conjugated to the surface of tosyl MPIO, as above. The SAIVI™ 750 dual-ligand MPIO were stored in the dark at 4°C. For in vitro flow chamber experiments, MPIO conjugated to mouse anti-human P-selectin and VCAM-1 antibodies or their counter-ligands, P-selectin glycoprotein ligand-1 (PSGL-1) or very late antigen 4 (VLA-4; α₄β₁) were developed (Supplemental Methods).

Mouse atherosclerosis

Female apolipoprotein knockout (apoE^−/−) mice on a C57BL/6 background were weaned at 3 weeks of age and fed a standard rodent chow diet ad libitum. Mice underwent MR imaging after 8, 14, 20 and 30 weeks on chow diet to determine the ability of dual-ligand MPIO to track endothelial adhesion molecule expression from early foam cell formation to more advanced atherosclerotic lesions. Animal procedures were performed in accordance with UK Home Office Animals (Scientific Procedures) Act 1986.

In vivo MRI

Mice underwent in vivo MR imaging of the aortic root using a 9.4 Tesla horizontal magnet interfaced to a VNMRS DirectDrive MR system (Varian Inc. USA). Animals were anesthetized with 4% isofluorane in O₂ gas mixture and maintained with 1.5-1.8% isofluorane in 100% O₂ (3 L/minute) delivered through a nosecone and positioned in a cradle in a 33 mm birdcage coil (Rapid Biomedical, Rimpar, Germany). Cardiac and respiratory signals were continuously monitored using an in-house developed ECG- and respiratory gating device.³² The aortic root was identified in a coronal section on a localizing sequence. T2* weighted images were acquired in end-diastole using a double-gated, flow-compensated, segmented 3D gradient-echo sequence (TE / TR = 2.4 / 4.0 ms, repetition time per segment 120 ms to 150 ms depending on heart rate, 2 k-space lines per cardiac cycle; matrix size 256 × 256 × 48; number of signal averages 1, slice thickness 8 mm, field of view 30 × 30 ×12 mm; flip angle 15°; reconstructed resolution 59 × 59 × 125 μm³). After pre-contrast imaging, the mice were administered either dual-ligand PV-MPIO (n = 6 - 7 per group) or control IgG-MPIO (n = 2-3 per group) (3.3 mg Fe / kg body weight) via tail vein injection. Post-contrast images were acquired at ~30 minutes and ~60 minutes after MPIO injection. Mice were terminally anesthetized with isofluorane and perfusion fixed with 4% paraformaldehyde (PFA). Histopathological examination of aortic roots was performed (Supplemental Methods).

MR image analysis

Aortic root low signal intensity areas were measured by drawing an area of interest around the outer wall of the aortic root using ImagePro Plus version 6.1. The threshold of low signal, corresponding to MPIO binding, was defined to be 3 standard deviations below the mean signal intensity of the pre-contrast root on signal intensity histograms. This threshold was applied to sequential MR slices through the aortic root, before and after MPIO contrast, to determine the mean area of MPIO low signal contrast. Pseudo-colored maps of bound MPIO were generated using the same greyscale thresholds, to visualise the distribution and magnitude of MPIO signal intensities in the aortic root wall, using a pseudocolor range from 0-255 colours. The contrast to noise ratios (CNR) of identified MPIO-positive lesion areas were compared to the CNR of equivalent areas on the pre-contrast images as CNR = blood pool signal – lesion signal/ standard deviation of the noise as previously described.³³

Flow cytometry

To assess the cellular distribution of dual-ligand MPIO binding, flow cytometry experiments were performed on cell suspensions from disaggregated whole aortas (Supplemental Methods). Whole blood from apoE^−/− mice was also assessed to establish whether there was binding of dual-ligand MPIO to circulating platelets. Apo E^−/− mice (n = 3), 20 wks on chow diet, were injected i.v. with near-infrared SAIVI™ 750 PV-MPIO (3.3 mg/kg). After 60 minutes circulation, mice were terminally anesthetized and perfused with ice-cold PBS. Aortas from apo E^−/− mice (n = 3) without MPIO injection served as controls.

En Face Immunofluorescence

ApoE^−/− mice, 20 weeks on chow diet, were injected with near-infrared SAIVI™ 750 labelled PV-MPIO (3.3 mg / kg). After 60 minutes, mice were terminally anesthetised and aortas perfused in situ with PBS and then 4% PFA via the left ventricle. Aortas were further fixed in 4% PFA for 30 minutes. Mouse aortas were examined using en face immunofluorescence (Supplemental Methods) to assess the binding patterns of dual-ligand MPIO binding in regions of the proximal and descending aorta with high and low probability (HP and LP) for atherosclerotic lesion development.^{34, 35}

Statistical analysis

Results are expressed as mean ± S.D. GraphPad Prism 5 (GraphPad Software, Inc., San Diego, USA) was used for statistical analysis. For imaging experiments with multiple groups, differences were evaluated using two-way analysis of variance (ANOVA). Unpaired 2-tailed Student’s t-tests were applied to determine differences between two groups and paired 2-tailed Student’s t-tests to compare differences in MRI data within a group before and after MPIO administration. Spearman rank non-parametric correlations were performed to correlate MRI and histological quantification of aortic root dual-ligand MPIO binding and the relationship between aortic root lesion macrophage content and dual-ligand MPIO binding. The significance level in all tests was P < 0.05.

Results

Antibody conjugated MPIO show high binding affinity to stimulated endothelial cells in vitro

To test surface orientation of the cognate ligands VLA-4 and PSGL-1 conjugated to MPIO, the particles were incubated with human recombinant VCAM-1-Fc or human recombinant P-selectin-Fc chimeras. Figure 1A shows green fluorescent MPIO confirming binding of VLA-4 MPIO to recombinant VCAM-1 and PSGL-1 MPIO to recombinant P-selectin. There was no binding of PSGL-1-MPIO to recombinant VCAM-1-Fc but VLA-4 MPIO showed some cross reactivity to recombinant P-selectin-Fc. In cellular binding experiments, quantification of MPIO adherent to stimulated HUVEC after 5 min of flow showed that antibody conjugated MPIO were retained significantly more than either VLA-4 MPIO (18-fold; P < 0.0001) or PSGL-1 MPIO (185-fold; P < 0.01) (Figure 1C). Antibody conjugated MPIO were therefore used for in vivo experiments in mice.

Figure 1. Antibody-conjugated MPIO bind with greater affinity in vitro than cognate ligand MPIO.

(A) VLA-4 MPIO and PSGL-1 MPIO were incubated with either Fc-VCAM or Fc-P-selectin. Green fluorescence indicates binding of Fc-VCAM to VLA-4 MPIO and Fc-P-selectin to PSGL-1 MPIO, with no MPIO binding to secondary antibody only or to non-his-tagged protein. (B) Immunofluorescent staining of VCAM-1 and P-selectin expression using cognate ligands and antibodies. VLA-4 binding was detected using an anti-human his tagged FITC while anti-mouse Alexa Fluor 594 was used to detect primary anti-human VCAM-1. PSGL-1 binding was detected using an anti-human FITC antibody and anti-P-selectin detected with anti-mouse Alexa Fluor 488. (C) Representative images of MPIO binding to stimulated HUVEC after 5 min of flow (flow rate 1 dyne cm⁻²). Antibody conjugated MPIO binding was significantly greater than VLA-4 MPIO (P < 0.0001) or PSGL-1 MPIO (P < 0.01).

In vivo imaging of adhesion molecules in atherosclerotic mice using dual-ligand MPIO

To determine whether dual-ligand MPIO can detect endothelial activation in vivo, MRI was performed at 8, 14, 20 and 30 weeks during atherosclerotic lesion development. Representative MR images of aortic root atherosclerotic lesions pre- and post-injection of either dual-ligand MPIO or negative control IgG-MPIO are shown in Supplemental Figure 1. MR images showed highly conspicuous areas of low signal after dual-ligand MPIO injection, with minimal MR signal effects observed in pre-contrast images and in aortic root images of mice injected with control IgG-MPIO. Sequential MR images through the root showed strong low signal areas 30 minutes post-MPIO injection, which persisted at similar levels at 60 minutes (Figure 2A). Pseudo-color maps illustrated the localisation and magnitude of the T₂* signal intensity enhancement in the aortic root wall (Figure 2B). Contrast effects were confined to regions of the root affected by atherosclerosis and not to lesion-free areas, as confirmed by histology (Figure 2C). The contrast-to-noise ratios (CNR) of MPIO-positive lesion areas and adjacent blood pool increased by 87% (P<0.05), 100% (P<0.01) and 96% (P<0.01) at 14, 20 and 30 weeks respectively compared to equivalent pre-contrast aortic lesion areas. No difference in lesion CNR was observed between time-points.

Figure 2. In vivo MR images of aortic root post dual-ligand MPIO injection.

(A) Consecutive MR images show low signal areas in the aortic root in apoE^−/− mouse (14 weeks on chow diet) at 30 minutes post PV-MPIO injection, which are maintained at 60 minutes. (B) Pseudo-colored maps (0-255 range) show the localisation and magnitude of the T₂* signal intensity in the aortic root, with red indicating the greatest and blue the least signal intensities. (C) The distribution of MR signal intensity is confined to atherosclerotic foam cell lesions and not to lesion-free endothelium, as confirmed by histology (arrows) using combined Masson’s trichrome and Elastin stain. (D) Contrast-to-noise ratio (CNR) of MPIO-positive lesion areas was significantly increased (P<0.05 at 14 weeks; P<0.01 at 20 and 30 weeks) after injection of PV-MPIO compared to equivalent lesion areas on pre-contrast images (mean ± SD), with no significant difference in post-MPIO contrast CNR between time-points. Scale bars = 1 mm.

Semi-quantitative analysis of the area of PV-MPIO contrast effects in the aortic roots showed minimal detectable MR contrast effects at 8 weeks, with a 3.9-fold increase in the area of low MR contrast effects at 14 weeks (P < 0.05), and a further 1.8-fold increase by 20 weeks (P < 0.05). However, at 30 weeks, there was a 1.8-fold reduction in MR contrast effects compared to 20 weeks (P < 0.01) (Figure 3A).

Figure 3. MRI and histological quantification of MPIO binding to aortic roots.

(A) Mean area (± SD) of low MR signal areas in aortic roots at baseline and 30 and 60 minutes after injection of PV-MPIO. Dynamic MRI identified rapid PV-MPIO binding to aortic roots of apoE^−/− mice by 30 minutes post-injection, which persisted at 60 minutes (*P < 0.05, **P < 0.01; n = 6-7 apoE^−/− mice per group injected with PV-MPIO; n = 2-3 apoE^−/− mice per group injected with IgG-MPIO; n=2 C57Bl/6 wild type mice injected with PV-MPIO. (B) Histological quantification found a significant increase in the number of bound PV-MPIO between 8 and 20 weeks, and a significant reduction between 20 and 30 weeks. No MPIO binding was observed to plaque-free areas of the endothelium in aortic roots of apoE^−/− mice or control wild type mice. (C) Low MR signal areas significantly correlated with the number of bound PV-MPIO by histology. (D) Immunohistochemistry confirmed aortic root VCAM-1 expression.

To test whether MR contrast effects reflected MPIO retention, the number of bound PV-MPIO were quantified by histology (Figure 3B). Significant increases in the number of PV-MPIO binding were observed at 14 and 20 weeks, with a reduction at 30 weeks, reflecting the MR contrast effects. The number of bound PV-MPIO quantified by histology correlated with MR signal intensity areas corresponding to aortic root PV-MPIO binding (P < 0.001) (Figure 3C). Irrespective of age, dual-ligand MPIO localized only to endothelium overlying atherosclerotic lesions, with no MPIO seen within lesions or bound to lesion-free endothelium. Immunohistochemistry for VCAM-1 confirmed endothelial activation on atherosclerotic lesions (Figure 3D).

Dual-ligand MPIO binding tracks quantitatively with lesion macrophage content

To investigate whether dual-ligand MPIO accumulation tracked with macrophage content within the lesion over time, aortic root sections were immunostained for Gal-3 (Figure 4A). Lesion macrophage area increased between 8 and 14 weeks, with a further increase by 20 weeks. However, between 20 and 30 weeks, there was a significant reduction in total macrophage area, which reflected the pattern of dual-ligand MPIO binding to the overlying endothelium (Figure 3B). Total lesion size also increased up to 20 weeks, but between 20 and 30 weeks, plaque size remained constant (Fig 4B). Lesion macrophage area correlated with dual-ligand MPIO binding to the aortic root assessed by in vivo MRI (R = 0.53; P < 0.01) (Figure 4C).

Figure 4. Dual-ligand MPIO binding quantitatively reflects lesion macrophage content.

(A) Representative images of aortic roots with lesion macrophage immunostained using Gal-3 antibody. Scale bar = 0.5 mm. Quantification of total lesion macrophage area (mm²) showed significant increases in lesion macrophage content from 8 to 20 wks and a significant reduction between 20 and 30 weeks. Magnification x 5. (B) Bound dual-ligand MPIO (yellow spheres) localized to endothelium overlying atherosclerotic lesions, with no binding observed to macrophages within the lesion. Magnification x 100. (C) Lesion size also increased from 8 to 20 weeks, but there was no change between 20 to 30 weeks. (D) Lesion macrophage content correlated significantly with dual-ligand MPIO MRI signal area.

Dual-ligand MPIO bind to activated endothelial cells at atherosclerosis-susceptible sites

In order to test the cellular specificity of MPIO binding in this context, we developed dual-ligand MPIO, fluorescently labelled with a near-infrared SAIVI™ alexa fluor 750 fluorophore. Flow cytometry was performed on cellular suspensions from digested aortas and whole blood of mice 60 minutes after intravenous injection of near-infrared SAIVI™ 750 PV-MPIO. Binding was predominantly to VCAM-1 positive (CD31⁺ CD11b⁻) aortic endothelial cells, with a minority bound to non-activated VCAM-1 negative endothelial cells (P < 0.001). There was no association of MPIO with VCAM-1⁺ CD31⁻ cells and only minimal MPIO uptake observed for macrophages (CD11b⁺ F4./80⁺) (Figure 5). Analyses of whole blood found that 22% of cells were CD41⁺ CD11b⁻ platelets, of which 1% were CD41⁺ CD11b⁻ CD62P⁺. No MPIO binding to CD41⁺ CD11b⁻ or to CD41⁺ CD11b⁻ CD62P⁺ circulating platelets was detected (Supplemental Figure 2).

Figure 5. Flow cytometry shows predominant binding of near-infrared fluorescently labeled dual-ligand MPIO to activated aortic endothelial cells in apoE^−/− mice.

(A) Overlay of fluorescence signal intensity histograms in the Alexa Fluor 750 PV-MPIO channel for each cell type. (B) Mean fluorescence intensity identified activated endothelial cells (CD31⁺ VCAM⁺ CD11b⁻) as the dominant cellular signal source with minimal signal in macrophages (CD11b⁺ F4/80⁺) or other cell types.

Predisposition to atherosclerosis is asymmetric in the proximal aorta with VCAM-1 up-regulation in the atherosclerosis-prone lesser curvature and is relatively sparing in the greater curvature of the arch and descending tubular aorta. We next investigated the spatial localisation of near-infrared SAIVI™ 750 dual-ligand MPIO binding in the aorta using en face immunofluorescence microscopy. Endothelial cells in the atherosclerosis-susceptible (HP region) inner curvature of the aortic arch displayed irregular cuboidal shapes, which did not align longitudinally and showed upregulation of VCAM-1 (Figure 6A). PV-MPIO were selectively retained by the endothelial in this area compared to the atherosclerosis-resistant outer curvature (LP region) of the aortic arch and descending aorta, which displayed endothelial cells of normal shape and alignment, with minimal levels of VCAM-1 expression (Figure 6B).

Figure 6. Dual-ligand MPIO binding localises to activated endothelial cells in atherosclerosis-susceptible sites of the aortic arch.

Representative images of the spatial distribution of PV-MPIO binding to the inner curvature (high probability [HP] region) or outer curvature (low probability [LP] region) of the aortic arch and the descending aorta assessed by en face immunofluorescence. The blue colour represents endothelial cell nuclei, counterstained with DAPI. The green fluorescent staining indicates endothelial PECAM-1 expression, the red fluorescent staining indicates endothelial VCAM-1 expression and the yellow spheres represent endothelial-bound PV-MPIO (fluorescently labelled with near-infrared alexa fluorophore 750 nm). The HP area had high levels of endothelial VCAM-1 expression, with cell nuclei that are not aligned in the same direction while the LP area and the descending aorta has low levels of VCAM-1 expression and cell nuclei that are aligned in a unilateral direction. Quantification of PV-MPIO binding in 4 images per area showed co-localisation of PV-MPIO to endothelial cells expressing VCAM-1 in the HP region, with minimal PV-MPIO binding in the LP region of the aortic arch or the descending aorta.

Discussion

Quantitative assessment of local vascular inflammation is a key goal in the diagnosis and characterisation of vascular disease and is likely to become more important as drugs directly targeting inflammatory pathways are developed for the treatment of high-risk patients with atherosclerosis.^{16, 18, 36} The vascular endothelium is the principal regulatory interface that governs the recruitment of mononuclear cells to sites of injury and inflammation, through the up-regulation of adhesion molecules, and is therefore an important target for this purpose.^{13, 37} Here, we demonstrate specific in vivo molecular imaging of endothelial cell activation during atherosclerotic lesion development in apoE^−/− mice using leukocyte-mimetic MPIO, targeting both VCAM-1 and P-selectin. These particles rapidly achieve steady state accumulation at target and provide exceptionally conspicuous, objectively quantifiable contrast effects. The MR contrast effects generated by dual-ligand MPIO binding also tracked quantitatively with lesion macrophage content across a range of lesion complexities. However, we show through both microscopy and flow cytometry that MPIO binding is confined to activated endothelial cells, and they are not bound to, or taken up by, macrophages within plaque. Using en face techniques to evaluate anatomically distinct areas of aorta, we further demonstrate selective binding of PV-MPIO in atherosclerosis-prone segments of the aortic arch that were associated with VCAM-1 up regulation and loss of normal endothelial cell architecture.

Adhesion molecule up-regulation was identified during early foam cell lesion, at 14 weeks on chow diet, with a further increase detected at 20 weeks, in fibrofatty lesions. However, at 30 weeks, in vivo MRI quantification detected a significant, apparently paradoxical, reduction in adhesion molecule expression. However, histological examination showed that PV-MPIO accumulation faithfully reflected actual changes in macrophage content. This is consistent with regulation of monocyte influx at the level of the vascular endothelium, with overall lesion macrophage content reflecting the net effects of recruitment vs. egress, apoptosis and necrosis.⁹ In addition, the distribution of PV-MPIO was spatially related to the areas of the root showing foam cell formation, while non-lesional areas did not retain MPIO. This agrees with reports that VCAM-1 is most strongly up-regulated in vascular endothelium overlying plaque.³⁸

Molecular imaging approaches benefit from rapid accumulation and attainment of steady state at target, with clearance of unbound (blood phase) contrast. Serial imaging revealed that PV-MPIO accumulated by 30 minutes on activated vascular endothelium and was maintained at 60 minutes, thus providing a practical imaging time frame. In an alternative strategy, peptides generated by phage display have been selected to bind specifically to activated endothelium and conjugated to nanoparticles.²⁰ After binding to VCAM-1, the nanoparticles are internalized by endothelial cells and macrophages over 24-48 hours, allowing progressive concentration sufficient for noninvasive imaging applications. Our work builds on these previous approaches to imaging activated endothelium by providing exceptionally conspicuous contrast within 30 minutes (rather that 24-48 hours), without the need for intracellular accumulation and thereby maintaining specificity for the original molecular targets. Furthermore, flow cytometry of single-cell suspensions from digested aortas showed that near-infrared SAIVI™ 750 dual-ligand MPIO bound specifically to activated endothelial cells and were not taken up by macrophages or other cell types.

In common with earlier studies^{19, 33} we imaged the aortic root in its short axis, since atherosclerotic lesions develop here early in apoE^−/− mice and progress in a semi-predictable fashion.^{39, 40} Secondly, the aortic valve position in relation to the chambers of the heart provides an excellent fiducial reference to align both the imaging plane and circumferential orientation. As shown in Figure 3, an eccentric lesion identified with in vivo contrast MRI is readily confirmed in tissue section.

Endothelial activation also occurs early in regions of endothelium prone to the development of atherosclerosis. Regions of low shear stress are associated with loss of the normal endothelial cell shape, disorganised endothelial cells that do not align in the direction of blood flow and increased expression of cell adhesion molecules. Using en face immunofluorescence, we show that the atherosclerosis-prone region of the aortic arch retains dual-ligand MPIO in association with VCAM-1 up-regulation while the relatively athero-resistant portions of aorta maintained normal EC morphology; aligned in the direction of flow, low level cell adhesion molecule expression and absent MPIO binding. Previously, microbubbles targeted to either P-selectin or VCAM-1 have been applied for molecular imaging of early atherosclerosis progression, which showed preferential binding to lesion prone areas in the aortic arch, with similar increases in endothelial P-selectin and VCAM-1 expression in atherosclerotic lesions with age.⁴¹

In seeking the optimal leukocyte mimetic MPIO, we compared the interactions of the cognate ligands VLA-4 and PSGL-1 with the antibody-targeting approach. After confirming that these natural ligands were conjugated and orientated on MPIO, we found that their binding to activated endothelial cells under low shear stress conditions in vitro had neither the specificity nor affinity of antibodies. We had reasoned that the cognate ligands may show greater affinity and would be less immunogenic in any future translational application. However, even on leukocytes the mere expression of ligands is not sufficient to attain binding, which is dependent on conformational change / clustering on the cell membrane, orchestrated by the cytoskeleton.⁴² We therefore pursued an antibody-based approach, utilizing our earlier observations in a flow chamber model⁴³ and in vivo¹⁹ that a dual targeting strategy directed at both cell adhesion molecules and selectins is more effective than either alone.

The MPIO that we have used are not suitable for human use due to their non-biodegradable polyurethane coat. However, surface-functionalized biodegradable, dextran-coated MPIO have recently been developed that are actively taken up by macrophages in the liver and spleen and rapidly dispersed through the specific action of macrophage cathepsins.^{44, 45} Combined with humanized antibodies, to minimize immunogenicity, these particles are potentially suitable for human testing. The dose of MPIO administered in our study was 3.3 mg Fe / kg body weight, which is similar to that used clinically for non-targeted iron contrast agents in oncological imaging (2.6 mg / kg)⁴⁶ and significantly lower than some USPIO doses used experimentally for imaging larger animal models such as rabbits (11 to 56 mg / kg).^{21, 47}

Atherosclerotic plaque macrophage content and activity are important determinants of plaque instability and the ability to image these may also be important in predicting risk.⁴⁸ Monocytes have been shown to be actively recruited in a progressive manner that is proportional to the extent of atherosclerotic disease.⁴⁹ Treatments that reduce lesion macrophage content have also been shown to reduce VCAM-1 expression.^{50, 51} The correlation between endothelial bound dual-ligand MPIO and lesion macrophage content, and thus putatively that dual-ligand MPIO behave as a monocyte mimetic, may provide a novel imaging probe to assess the localization and extent of inflammatory atherosclerotic disease. Considerable effort has been directed towards the use of 2-[¹⁸F]fluoro-2-deoxy-d-glucose (¹⁸F-FDG-PET) for positron emission tomography (PET) assessment of atherosclerotic plaque inflammation, with success in identifying ‘acute’ lesions³⁶ and in monitoring response to therapy.^{36, 52} However, whilst ¹⁸F-FDG appears to be preferentially taken up by inflammatory macrophages, a recent in vitro study suggests that this technique may be sensitive to hypoxia in plaque macrophages rather than measuring inflammation per se.⁵³

MPIO, due to their relatively large size, do not penetrate the vessel wall and therefore remain highly specific for targeting endothelial molecular markers. This is unlike untargeted USPIO, which are subject to vascular egress and passive accumulation within atherosclerotic lesions or uptake by macrophages.^{21, 22} The rapid binding kinetics of dual-ligand MPIO, coupled with the immediate¹⁹ accessibility of endothelial cells (without the need to permeate the plaque), distinguish targeted MPIO as a powerful molecular imaging platform. The expression of endothelial cell adhesion molecules is not specific to any given stage or size of lesion. Therefore, if used in clinical practice, it would be anticipated that molecular imaging of endothelial inflammation would be combined in an integrated examination of vessel characterisation with soft tissue arterial wall imaging and MR angiography.

Conclusions

We have developed a leukocyte mimetic contrast agent, for in vivo molecular MRI, that rapidly homes to activated vascular endothelium during atherosclerotic lesion development. Although confined to the intravascular compartment and bound to endothelium, in the context of atherosclerosis in the aortic root, the accumulation of MPIO contrast agent quantitatively tracked with atherosclerotic lesion macrophage content. This approach provides new and significant benefits: (i) exceptionally rapid steady state accumulation; (ii) unusually conspicuous MR contrast effects that can be objectively quantified using semi-automated techniques and (iii) faithful reflection of endothelial activation associated with leukocyte accumulation. As degradable biocompatible particles become available this potent platform may find utility in both imaging and, potentially, drug delivery.