Magnetic Resonance Imaging of Atherosclerotic Plaque at Clinically Relevant Field Strengths (1T) by Targeting the Integrin α4β1

Inflammation drives the degradation of atherosclerotic plaque, yet there are no non-invasive techniques available for imaging overall inflammation in atherosclerotic plaques, especially in the coronary arteries. To address this, we have developed a clinically relevant system to image overall inflammatory cell burden in plaque. Here, we describe a targeted contrast agent (THI0567-targeted liposomal-Gd) that is suitable for magnetic resonance (MR) imaging and binds with high affinity and selectivity to the integrin α4β1(very late antigen-4, VLA-4), a key integrin involved in recruiting inflammatory cells to atherosclerotic plaques. This liposomal contrast agent has a high T1 relaxivity (~2 × 105 mM−1s−1 on a particle basis) resulting in the ability to image liposomes at a clinically relevant MR field strength. We were able to visualize atherosclerotic plaques in various regions of the aorta in atherosclerosis-prone ApoE−/− mice on a 1 Tesla small animal MRI scanner. These enhanced signals corresponded to the accumulation of monocyte/macrophages in the subendothelial layer of atherosclerotic plaques in vivo, whereas non-targeted liposomal nanoparticles did not demonstrate comparable signal enhancement. An inflammatory cell-targeted method that has the specificity and sensitivity to measure the inflammatory burden of a plaque could be used to noninvasively identify patients at risk of an acute ischemic event.


Supplementary Materials and Methods
Supplementary Table S1. Selectivity data of antagonists tested against a panel of integrin targets. Supplementary Table S2. Constituents of THI0567-targeted and non-targeted liposomal-Gd constructs.
Compound 10: To a solution of 9 (2.97 g, 5.80 mmol) in absolute ethanol (39 mL) at room temperature under argon, glacial acetic acid (0.5 mL), palladium metal on carbon (Degussa type E101 NE/W, 50% H2O, 10% Pd dry weight basis, 0.98 g, 0.46 mmol Pd). The atmosphere was replaced with hydrogen (toggling between vacuum and hydrogen from a balloon several 8 times), and the reaction was stirred overnight. The mixture was filtered through Celite ® , washing with ethanol, and the filtrate was concentrated under reduced pressure. The residue was recrystallized from diethyl ether and hexanes to give 10 (1.373 g) as a white crystalline solid. No further attempts were made to isolate additional material from the mother liquor. 1  Compound 11: A solution of 6 (1.003 g, 2.46 mmol) and 10 (890 mg, 2.24 mmol) in DMF (12.3 mL) and N,N-diisopropylethylamine (DIPEA) (0.59 mL, 3.36 mmol) under argon was heated to 55°C for 8 hours. An aliquot indicated unreacted 10, so additional 6 (100 mg, 0.25 mmol) was added. Then, the mixture was heated to 55°C overnight, cooled to room temperature, and diluted with 1:1 hexanes:ethyl acetate and HCl (2N). The organic layer was washed with water (3 times) and brine, dried over MgSO4 and filtered. The filtrate was concentrated under reduced pressure, and the resulting residue was purified automated chromatography on silica gel (Biotage ® , SNAP100 KP-Sil, eluting with 25-50% ethyl acetate in hexanes). A few fractions containing the desired product also contained an impurity. The fractions were concentrated and repurified (Biotage ® , SNAP10 Ultra, eluting with 30-50% ethyl acetate in hexanes). Fractions from both separations containing only the desired product were combined and concentrated to give 11 (1.25 g) as a pale yellow foam. This material contained approximately 4% ethyl acetate by weight but was used as is Compound 12: To a solution of 11 (26 mg, 0.035 mmol) in dichloromethane (0.2 mL) at room temperature, trifluoroacetic acid (0.2 mL) was added. The mixture was stirred at room temperature for 4 hours and then was concentrated. The residue was dissolved in dichloromethane and concentrated. The residue was then taken up in a 1:1 mixture of acetonitrile and water (2 mL) and allowed to stand overnight. The resulting mixture was diluted with water (2 mL) and then was frozen in a dry ice/acetone bath and lyophilized to give 12 (19.6 mg) as a The synthetic scheme to generate Compounds 13 through 19 is shown in Figure S5.
Compound 14: To a solution of 11 (1.25 g, 96% by weight, 1.61 mmol) and 13 (918 mg, 3.22 mmol) in DMF (8 mL) at room temperature under argon, potassium carbonate (668 mg, 4.83 mmol) was added. The resulting mixture was heated to 80°C overnight, cooled to room temperature, diluted with ethyl acetate, and washed with water (3 times) and brine. The organic layer was dried over magnesium sulfate, filtered, and concentrated under reduced pressure. Compound 15: To a solution of 14 (1.16 g, 1.29 mmol) in THF (12.9 mL) at room temperature under argon, triphenylphosphine (508 mg, 1.94 mmol) was added. The mixture was stirred for 1.5 hours, water (0.26 mL) was added, and stirring was continued overnight. The mixture was concentrated, and the residue was purified by automated chromatography on silica gel (Biotage ® , SNAP25 KP-Sil, eluting with 75-100% ethyl acetate in hexanes, then 0-10% methanol with 2% added triethylamine in ethyl acetate, then 10-20% methanol with 2% added triethylamine in dichloromethane) to give 15 (1.101 g) as a brownish yellow viscous oil. NMR Compound 19: To a solution of 18 (680 mg, approximately 0.13 mmol) in dichloromethane (9 mL), trifluoroacetic acid (9 mL) was added. The mixture was stirred at room temperature for 4 hours and then was concentrated. The residue was dissolved in dichloromethane and concentrated (5 times). The residue was then taken up in a 1:1 mixture of acetonitrile and water (30 mL) and allowed to stand overnight. The resulting mixture was diluted with water (60 mL), and then the resulting mixture was frozen in a dry ice/acetone bath and lyophilized. The resulting powder was purified by size-exclusion chromatography (Sephadex LH-20) in 2 portions, eluting with methanol. Fractions were spotted as described above, and fractions containing material that was both UV and PMA active were combined and concentrated. The residue was taken up in water (50 mL) and acetonitrile (15 mL), and the resulting mixture was frozen in a dry ice/acetone bath and lyophilized to give 19 (572 mg) as an off-white solid. MALDI (Positive mode, sinapic acid): central mass of distribution: 4867.8.
The synthetic scheme to generate Compounds 20 through 25 is shown in Figure S6.
Compound 20: To a suspension of 3-ethoxycinnamic acid (2.028 g, 10.6 mmol) in toluene (13.3 mL) at room temperature under argon, tert-butyl 2,2,2-trichloroacetimidate (2.37 mL, 13.3 mmol) was added. The mixture was heated to 50°C overnight, at which time TLC analysis revealed partial conversion. Additional tert-butyl 2,2,2-trichloroacetimidate (1.2 mL) was added, and heating was continued for 24 hours. The reaction was still not complete, so more tert-butyl 2,2,2-trichloroacetimidate (1.2 mL) was added, and heating was continued for an additional 24 hours. The resulting mixture was filtered, washing with toluene, and the filtrate was concentrated under reduced pressure. The residue was purified by chromatography on silica gel, eluting with 14 10% ethyl acetate in hexanes to give 20 ( Compound 22: Following the general procedure for catalytic hydrogenolysis of the benzyl groups, a mixture of 21 (1.50 g, 3.26 mmol), palladium metal on carbon (Degussa type E101 NE/W, 50% H2O, 10% Pd dry weight basis, 0.55 g, 0.26 mmol Pd), and glacial acetic acid (0.2 mL) in absolute ethanol (22 mL) was stirred under a hydrogen atmosphere overnight. After undergoing filtering and concentrating, the residue was taken up in a 1:1 mixture of ethyl acetate and hexanes and was washed with aqueous sodium hydroxide, water, and brine. The organic phase was dried over magnesium sulfate, filtered, and concentrated to give 22 (798 mg) as a light-yellow oil. NMR (300 MHz, CD3SOCD3): δ 7.17 (t, J = 7.8 Hz, 1H), 6.86-6.94 (m, 2H), 6.74 (ddd, J = 7.8, 2.7, 0.9 Hz, 1H), 4.09 (t, J = 7.1 Hz, 1H), 4.00 (q, J = 6.9 Hz, 2H), 2.37-2.53 (m, 2H), 1.93 (br. s, 2H), 1,27-1.37 (m, 12H). The synthetic scheme to generate Compounds 24 through 30 is shown in Figure   S7.Compound 26: Following the general procedure for alkylation of the pyridone hydroxyl, a mixture of 24 (710 mg, 1.18 mmol), tert-butyl (6-bromohexyl)carbamate (992 mg, 3.54 mmol) and potassium carbonate (326 mg, 2.36 mmol) in dimethylformamide (4 mL mmol) was added. The reaction was stirred for 2 hours and then was diluted with ethyl acetate and washed with aqueous sodium hydroxide, water and brine. The organic layer was dried over magnesium sulfate, filtered, and concentrated to give a 4:1 mixture of 27:26 (492 mg) as a yellow oil. In general, this material was used without purification, but a small amount from a previous reaction was purified for analysis by reverse phase HPLC (Symmetry Shield RP18, 7 µm, 30x250mm, 30-80% acetonitrile in water with 0.1% trifluoroacetic acid). Fractions containing the desired material were combined, diluted with water and ethyl acetate, made basic with aqueous sodium hydroxide, and shaken in a separatory funnel. Then, the phases were separated. The organic layer was washed with water (3 times) and brine, dried over magnesium mg, 0.041 mmol) in dimethylformamide (4.1 mL), and the mixture was stirred for 2 days.

Cell adhesion assays.
Cell adhesion assays were performed as previously described 77 . Wells were coated either directly with substrate (fibronectin, vitronectin, collagen) or with the appropriate anti-IgG antibody to subsequently capture IgG fusion proteins (VCAM-1-Ig, MAdCAM-1-Ig, ICAM-1-Ig). The concentration of substrate or Ig fusion protein added to the wells was equivalent to the EC50 as previously determined by dose-dependent binding curves.
Mouse integrin adhesion assays were performed identically, except that mouse VCAM-1 was Prime module was used in the construction of the α4β1 model, which was based on the α4β7 and α5β1 crystal structures 78,79 (entries 3v4v and 3vi4, respectively) as obtained from the Protein Data Bank. The α4 integrin sequence (P13612) and the β1 integrin sequence (P05556) were obtained from UniProt (http://www.uniprot.org) and downloaded in the FASTA format. The sequence for α4 was truncated to residues 1-587 and read into Prime. Chain A of α4β7 crystal structure was selected as the template for this part of the model. Similarly, the β1 integrin sequence was truncated to residues 4-445, and Chain D in the α5β1 crystal structure was used as the template for that portion of the model. No changes to the initial alignment were necessary since the sequences were identical to their respective templates. During the build-structure phase, an energy-based build was used for α4 with the 0DU ligand and for β1 with both the calcium and magnesium ions. Heteromultimer modeling was entered, and the α4 and β1 runs were selected.
After construction of the raw model, two loops underwent structural refinement: Loop1 (chain A, residues 33-42) and Loop2 (chain B, residues 553-560). These loops were distal from the ligand binding site and not expected to affect the binding region. At this point, we noticed that some bonds in the ligand were incorrect. The bonds were corrected, and the protein portion of the model was prepared. This involved basic protein preparation; assignment of heteroatom states; and H-bond assignment, including PROPKA to assign sidechain ionization. This was followed by Impref "H-only" minimization and then Impref "Minimize All" to root mean square deviation=0.5. Docking THI0565 into integrin α4β1. The initial geometry optimization of THI565 was performed in PRODRG ( Supplementary Fig. S9). Quantum mechanical calculations were completed in FIREFLY with a Pople's 6-N31G* (6 Gaussians) split valence basis set applying Hartree-Fock theory. The derived partial atomic charges were computed from a least-squares fit of the electrostatic potential (ESP) from the Lowdin atomic population and aggregated in charge groups that matched the GROMACS force field. THI565 was docked into the input molecular model of α4β1 by using Autodock Vina 1.1.2 with an unbiased search box ( Supplementary Fig.   S10) with center (6,40,10)A and size (100,100,100)A and then refined by using a focused box encompassing the putative binding sites, identified from the first round at the interface between the α4 and β1 domain with a center of (2,35,72)A and size of (40,32,40)A. Each trial was assigned an independent random seed. An exhaustiveness of 100 was used with a mode number 21 of 50. The top 20 poses were selected for each trial. Molecular dynamics simulations were run using GROMACS 5.1.4 with a GROMOS96 43a1 force field. The total system charge was neutralized using the appropriate number of sodium or chloride counterions. An energy minimization was first performed to remove interatomic clashes by using 500 steps of the steepest descent algorithm or until a threshold was met. Next, NVT and NPT energy minimization was done to prepare the complex for equilibration and production MD simulation.
A set of 1.0 ns production molecular dynamics simulations was conducted for the proteincompound complex ( Supplementary Fig. S11). Dihexadecanoyl-sn-Glycero-3-Phosphoethanolamine, Triethylammonium Salt (rhodamine DHPE) was purchased from ThermoFisher Scientific. All purchased reagents were used without further purification. DSPE-PEG3400-THI565 conjugate (referred to as THI0567) was synthesized as described above. DPPC, cholesterol, DSPE-MPEG2000, Gd-DTPA-BSA, and DSPE-PEG3400-THI567 were respectively constituted on the basis of the desired surface targeting ligand expression at molar proportions shown in Supplementary Table S1. We added rhodamine DHPE (1.0-2.5 mg, 0.2 mol%) to each of the lipid compositions, and particle formulation proceeded as previously described 44 . Briefly, the lipids were dissolved in ethanol (1.0 to 1.2 mL) and then hydrated at ~65°C for 40 minutes in 150 mM saline/10 mM histidine to achieve a lipid concentration of 50 mM. The mixture was then extruded in a 10-ml Lipex extruder (Northern Lipids Inc.) by using a 400-nm polycarbonate track-etch filter (5 passes) to obtain particles with a mean diameter of ~250 nm. For particles with a mean diameter of ~150 nm, the ensuing formulation was further extruded through a 200-nm polycarbonate filter (8 passes); for particles with a mean diameter of ~100 nm, the formulation was further extruded (5 times) through 100nm filters. The resulting solution was then dialyzed against 150 mM saline/10 mM histidine. The mean liposome size in the final formulation was determined by dynamic light scattering, and the gadolinium and phospholipid (equivalent phosphorus) concentrations in the formulation were quantified using inductively coupled plasma optical emission spectroscopy (ICP-OES). The number of particles/mL was computed on the basis of the mean particle size and the final lipid concentration in the formulation.
Liposome binding assays. Cells were incubated with indicated concentrations of liposome in binding buffer (50 mM HEPES, pH 7.4, 150 mM NaCl, 2.5 mM KCl, 10 mM NaHCO3, 1 mg/ml glucose) for 1h at room temperature. Background non-specific binding was determined in the presence of EDTA (20 mM). After incubation, cells were washed once in binding buffer and resuspended. Rhodamine B fluorescence was measured on a flow cytometer (LSRII, BD).
Binding data are expressed as the geometric mean fluorescence intensity (gMFI). Binding Kds were generated in Prizm Software using the saturation binding equation for "One site -Total and nonspecific binding." Total binding was fit with the equation Y=(Bmax*X/(Kd+X)) + NS*X+BKG; nonspecific binding was fit to the linear equation Y=NS*X+BKG, where X is the particle concentration of liposome, Y is Rhodamine B fluorescence, NS is nonspecific binding, and BKG is background (NS and BKG are shared). In some experiments, after in vitro binding 23 assays were performed, cells were labelled with the indicated monoclonal antibodies (anti-CD64 mAb or anti-CD3 mAb OKT3) for confocal analysis. Cells were incubated with 10 ug/ml of primary antibody in FACS buffer (PBS, 10% FCS, pH 7.4) for 1h at 4°C. After washing, secondary GAM-FITC (2 ug/ml) was incubated with cells (1h, 4°C), which were then washed, subjected to cytospin onto glass coverslips, air dried, and mounted for confocal imaging.
Fluorescent microscopy. For confocal analysis, all images were obtained by using a Leica TCS SP5 II confocal microscope. The incident laser intensity and image capture settings for each channel were kept constant for all imaging in which control and treatment groups were directly compared. Reconstruction of 3D images was performed by scanning an XY plane at multiple Z positions and utilizing the 3D-visualization software within the LAS AF software package (Leica). For standard fluorescence, images were captured on a Olympus BX51 fluorescent microscope with Cellsens Dimension imaging software.