Chemicals and reagents
Chemicals
Cholesterol-PEG2000-maleimide (Nanocs); Na2HPO4, NaH2PO4, glycerol (CarlRoth GmbH); thrombin, ADP (Sigma Aldrich); E80S (Lipoid GmbH); perfluoro-15-crown-5 ether (PFCE), perfluoro-1,3,5-trimethylcyclohexane (PFCH), perfluorooctyl bromide (PFOB), diblock, i.e. F6H10, perfluoro-n-hexyl decane (Abcr chemicals). Buffer: Phosphate glycerol buffer (3 mM NaH2PO4, 7 mM Na2HPO4, glycerol 2.5% m/V, pH 7.4), phosphate buffer (PBS, CarlRoth). Targeting peptides: Peptides which were derived from α2-antiplasmin (α2AP = Ac–Gly–Asn–Gln–Glu–Gln–Val–Ser–Pro–Leu–Thr–Leu–Leu–Lys(Cys)–Trp–Lys13,15) and a peptide which was modified on the basis of EP2104-R57 (here called fbn = Tyr–D-Glu–Cys–Hyp–Tyr[3-Cl]–Gly–Leu–Cys–Tyr–Ile–Gln–Gly–Gly–Gly–Cys) were utilized for targeting. As control for α2AP, we used a peptide where the glutamic acid at position Q3 which is cross-linked to the fibrin network was replaced by alanine. The control peptide for fbn (FAM–Ser–Asp–Gly–Gly–Als–Asn–Leu–Gly–Gly–Ile–Glu–Gly–Gly–Gly–Cys) was generated by replacing and randomizing all of the amino acids which are responsible for fibrin binding58. All peptides were commercially manufactured (Genaxxon Bioscience) and fbn and its control were equipped with an additional C-terminal GGG-Cys tag and a carboxyfluorescein at the N-terminus to enable fluorescence imaging. The FXIIIa specific peptide α2AP and its control contained an additional Cys- on the ε-amino group of K13 and the carboxyfluorescein was conjugated to K15. Delivered lyophilized peptides were suspended in phosphate glycerol buffer and stored at −20 °C.
Preparation and characterization of perfluorocarbon nanoemulsions (PFCs)
Preparation
PFCs which contained perfluoro-15-crown-5 ether (PFCE), perfluorooctyl bromide (PFOB) or perfluoro-1,3,6-trimethylcyclohexane (PFCH) were prepared by microfluidization with E80S lipids as emulsifier using a laboratory processor (LV1, Microfluidics Corp.) operating at 1000 bar. PFCH and PFOB emulsions were further stabilized by adding (equimolar to the lipid content) a semifluorinated fluorocarbon/hydrocarbon diblock59. In brief, 2.4% (w/w) E80S was dissolved in 10 mM phosphate glycerol buffer. Subsequently, either PFCE (20% w/w), PFCH (40% w/w) or PFOB (40% w/w) was added to the lipid suspension. For PFCH and PFOB 35 mM diblock was additionally added. The suspensions were preemulsified by high shear mixing (Ultra Turrax TP 18/10) and processed with the LV1 at 1000 bar and for 5 cycles. PFCs were autoclaved (121 °C, 1 bar, 30 min), aliquoted in glass vials, and stored at 4 °C under light protection.
Characterization
PFCs were characterized by dynamic light scattering (DLS) using a Nanotrac instrument (Microtrac) to determine the hydrodynamic diameter, the polydispersity index (PDI) and the ζ potential. For this, 20 µl of PFCs were diluted in 1 ml Milli-Q water and analyzed for 90 s. These measurements were repeated for 10 times for calculation of mean values. To gain information on several preparations, the mean values and standard deviations of individual particle size measurements were averaged. For determination of the 19F amount, 10 µl of PFCE, PFOB, PFCH were diluted with 20 µl H2O and filled into a 200 µl reaction tube. Additionally, a 30 µl mixture of PFCs (10 µl each) was prepared. Subsequently, the reaction tubes were fixed to a 2 ml water tube and subjected to 1H/19F MRI (see below).
Functionalization of PFCs
PFCs were equipped with peptide ligands using the sterol-based post-insertion technique (SPIT)15,60. For this, peptides were coupled via free cysteines to cholesterol-PEG2000-maleimide anchors exploiting the formation of a stable thioether after reaction of the cysteine SH-group with the maleimide of the anchor. For conjugation, 280 µg of the peptide (suspended in phosphate glycerol buffer) was added to 300 µl of cholesterol-PEG2000-maleimide (1 µg/µl, suspended in phosphate glycerol buffer) and the volume was adjusted to 650 µl with phosphate glycerol buffer. The solution was incubated overnight at 22 °C under constant stirring. On the next day, 650 µl of the cholesterol-PEG2000-peptide conjugate was mixed with 650 µl of preformed PFCs and incubated at 22 °C for 3 h which led to the spontaneous insertion of the cholesterol conjugate into the lipid layer of the PFCs15,60. For targeting of activated platelets, PFCs were equipped with a unique human single-chain antibody (scFv) as previously described28. Functionalized PFCs were stored at 4 °C under light protection.
Ethics
Animal experiments were performed in accordance with the European Union guidelines described in the directive 2010/63/EU and were approved by North Rhine Westphalian State Agency for Nature, Environment and Consumer Protection (LANUV = Landesamt für Natur, Umwelt und Verbraucherschutz Nordrhein-Westfalen), Germany. In vitro studies with blood samples from healthy humans were conducted to the declaration of Helsinki and approved by Ethics Committee of the Heinrich Heine University, Düsseldorf, Germany. All healthy volunteers gave written informed consent to the collection and use of blood samples in this study.
Animals
Mice expressing a hypomorphic mutant form of ApoE (ApoE-Rh/h) and lacking the scavenger receptor class B type I (SR-BI−/−) were bred and housed with a 12L/12D cycle (light switch 6:00/18:00) at an ambient temperature of approximately 22 ± 2 °C and a humidity of 55% at the central animal facility of the Heinrich Heine University Düsseldorf, Germany. Mice were fed a standard chow diet and received tap water ad libidum. For induction of the disease model16,17, mice were exposed to a high fat/high cholesterol diet (7.5% cocoa butter, 15.8% fat, 1.25% cholesterol, 0.5% sodium cholate; Altromin). For visualization of thromboinflammatory processes by 19F MRI, mice received an intravenous bolus injection of PFCs (3 mM/kg BW) 48 h prior to MRI to ensure appropriate PFC deposition at the sites of interest9.
A total of 33 mice (males and females, 25–30 g body weight; 8–10 weeks of age) were used in this study. Twenty-three of them were subjected to the main protocol illustrated in Supplementary Fig. 1. All mice received baseline 1H MRI, PFC injections on day 3 after onset of the diet and 1H/19F MRI on day 5. Thereafter, 6 mice were sacrificed for 3D high resolution 1H/19F MRI and histology. For follow-up, 17 mice were subjected to PFC injection on day 8 and 1H/19F MRI on day 10, whereupon again 6 mice were sacrificed for post mortem analyses. Afterwards, one mouse died spontaneously, so that 10 mice received the final PFC injection on day 13 and 1H/19F MRI on day 15. Thereafter, the remaining mice were sacrificed for 3D 1H/19F MRI and pathohistological analyses. Additional 4 mice were used for the explorative study of the fourth color with scFvPFCs (Supplementary Fig. 15) and further 6 mice served as control and were not subjected to the diet but received PFC injections as indicated above (Supplementary Fig. 7).
Magnetic resonance imaging
General
Data were recorded at a Bruker AVANCEIII 9.4T wide bore NMR spectrometer driven by ParaVision 5.1 (Bruker) and operating at frequencies of 400.21 MHz for 1H and 376.54 MHz for 19F measurements9,15. Images were acquired using the Bruker microimaging unit Micro 2.5 with actively shielded gradient sets (1.5 T/m) and a 25 mm birdcage resonator (Bruker) tunable to both 1H and 19F. Mice were anaesthetized with 1.5% isoflurane and kept at 37 °C. To prevent dehydration of eye surfaces and mucous membranes, the anesthesia gas mixture was moisturized by passing the mixture through a gas-washing bottle filled with water. The front-paws and the left hind-paw were attached to ECG electrodes (Klear-Trace; CAS Medical Systems) and respiration was monitored by means of a pneumatic pillow positioned at the animal’s back. Vital functions were acquired by a M1025 system (SA Instruments) and used to synchronize data acquisition with cardiac and respiratory motion if necessary. After acquisition of initial 1H pilot scans, the resonator was immediately tuned to 19F to screen the entire thorax above the liver up to the carotid bifurcation for 19F hotspots. Thereafter, the resonator was tuned back to 1H to acquire the corresponding anatomical information for detected 19F signals and functional analysis. For superimposing the images of both nuclei, the ‘hot iron’ color look-up table provided by ParaVision was applied to 19F images. To fade out the background noise from 19F images a constant threshold was applied to 19F data and signals originating from the liver were masked for sake of clarity (cf. Supplementary Fig. 7 for a representative example of a mouse 10 days after onset of the diet illustrating a comparable 19F signal intensity in inflamed myocardium as compared to the liver as major site of PFC deposition). For subsequent intraperitoneal Gd contrast agent application (bolus of 0.2 mmol Gd-DTPA per kg body weight), the animal handling system was shortly removed from the magnet and afterwards re-inserted with exactly the same positioning. The entire scanning protocol took around 80 min and was well tolerated by all animals, which recovered within 2 min from anesthesia.
For functional and morphometric analysis, high resolution images of mouse hearts were acquired using an ECG- and respiratory-gated segmented fast gradient echo cine sequence with steady state precession (FISP). A flip angle (FA) of 15°, echo time (TE) of 1.2 ms, 128 segments and a repetition time (TR) of about 6-8 ms (depending on the heart rate) were used to acquire 16 frames per heart cycle with a field of view (FOV), 30 × 30 mm2; matrix size after zero filling (MS), 256 × 256; slice thickness (ST), 1 mm; number of averages (NA), 2; acquisition time (TAcq) per slice for one cine loop, ∼1 min. In cases of inadequate ECG recording, images were acquired with a retrospectively gated fast low angle shot sequence (IntragateFLASH, Bruker) using a FA 10°, TE 1.26 ms, TR 5.82 ms, FOV 30 × 30 mm2, MS 256 × 256, TAcq 1.5 min. For retrospective gating, a navigator slice was placed near the base of the heart, where the signal is primarily modulated by the periodically changing atrial and aortic blood volume61. Routinely, 8–10 contiguous short axis slices were required for complete coverage of the LV. Longitudinal slices orientated perpendicular to the atrio-ventricular level served to ensure appropriate scanning of the entire heart from base to apex. For evaluation of functional parameters, ventricular demarcations in end-diastole and -systole were manually drawn with the ParaVision Region-of-Interest (ROI) tool.
MR angiography (MRA) of aortic arch and carotid arteries was carried out using a flow-compensated 2D time-of-flight (TOF) FLASH sequence (FA 80°, TE 2.2 ms, TR 12.3 ms, NA 4, TAcq 4.4 min, FOV 30 × 30 mm2, MS 384 × 384)62,63. In z-direction the FOV was set to cover the entire vessel system from the aortic root up to carotid bifurcation (40 overlapping slices with a slice thickness (ST) of 0.4 mm and an interslice distance of 0.25 mm resulting in a package extent of 10.15 mm). For 3D visualization, data were imported into Amira 4.0 (Mercury Computer Systems) and resampled to isotropic voxel size using a Lanzcos filter. Afterwards, 1H MRA data were visualized in greyscale by texture-based volume rendering (VRT). For overlay, anatomical corresponding 19F MRI data (see below) were rendered by application of the ‘glow’ colormap.
19F MRI
For the initial experiments with PFCE only, 19F RARE images were essentially recorded as previously described9,64,65,66 using the following parameters: RARE factor 32, TR 2500 ms; TE 4.37 ms, MS 128 × 128, ST 2 mm, averages 256, TAcq 21.3 min. For multi-color experiments, the mCSSI pulse sequence schematically shown in Fig. 2a was developed. It can be considered as a multi-slice 3D RARE variant in which slice selection is replaced by frequency selective excitation. This allows for simultaneous imaging of molecules with complex spectra (Fig. 2b) and/or different substances. Narrow bandwidth excitation is followed by phase encoding of the third dimension and the echo train acquisition with selective refocusing pulses. Gradient spoiling is required to destroy remaining transverse magnetization at the end of each cycle. This procedure is repeated for the number of excitation frequencies within one repetition time which is similar to TR of the standard RARE sequence. Reconstruction is straightforwardly performed like for a standard multislice RARE dataset, in which slice selection is replaced by different narrow band excitation loops, and provides either a complete image set for each frequency or—to omit a manual calculation step afterwards—the user can also choose to directly create sum images from individual frequencies of a PFC with multiple resonances (such as PFOB) to improve SNR. For this, dedicated reconstruction software was developed in-house based on the LabVIEW package (National Instruments). Multi-color 19F datasets were recorded with this sequence using the following parameters: TR 2500 ms, RARE factor 32, TE 4.49 ms; bandwidths 1370 and 1610 Hz for excitation and refocussing Gaussian pulses, respectively; MS 64 × 64, ST 2 mm, NA 512; TAcq 21.3 min. The frequencies for excitation/refocussing were adjusted to the resonances of PFOB, PFCE, PFCH, and PFTBCH as listed in Supplementary Table 1. All chemical shifts relative to the CF3 resonance of PFOB were derived from non-selective spectroscopic measurements. The standard deviation of the average frequencies never exceeded 40 Hz and were well below the bandwidth of the Lorentzian-shaped lines. For 19F mCSSI, it is therefore adequate to define only one absolute excitation frequency and to assume constant chemical shifts for the other resonances.
For post mortem high resolutions measurements and subsequent histology, hearts were excised and fixed in a 1.5 ml Eppendorf cap filled with paraformaldehyde. 19F data were acquired with a 3D RARE sequence (RARE factor 32, FOV 22 × 16 × 16 mm3, MS 384 × 256 × 256, TE 21.4 ms, TR 2500 ms, NA 72, TAcq 25.36 h), while anatomical reference data were recorded with a 3D FISP sequence from the same FOV using the following parameters: TE 3.4 ms, TR, 6.8 ms, MS 512 × 384 × 256, NA 24, TAcq 51 min. After MRI measurements, serial 5 µm sections of the entire heart were stained with hematoxylin and eosin (H&E).
MRI data were imported into 3D visualization software Amira and resampled to isotropic voxel size using a Lanzcos filter. 1H MRI data were visualized in greyscale by VRT adapting the lower and upper intensity thresholds for rendering in a way that the surrounding fluid got invisible. For overlay, the ‘glow’ colormap was used to render the morphologic corresponding 19F data. After coregistration, data were analyzed in multi-planar view for exact anatomical localization of the 19F label in the hearts. For confirmation of the 19F label to be localized at the luminal site of the aorta (Supplemental Fig. 3), the vessel wall was manually segmented and its surface calculated with unconstrained smoothing. For overlay, semi-transparent surface views of the wall were generated and anatomic corresponding 19F data were rendered as described above.
Phantom experiments
To prove reliable quantification of the 19F mCSSI signals, phantom measurements were performed using the same PFCs as in in vivo experiments. PFCs were filled into 200 µl microfuge tubes and fixed in the resonator with in-house constructed specimen holder. MRI data were acquired with same parameters given above and afterwards SNR was determined for each individual excitation frequency (see below).
19F MR image analysis
19F MRI data were analyzed using an in-house developed software module based on the LabVIEW package (National Instruments) which allowed (i) an overlay on anatomical corresponding 1H MRI data, (ii) an easy separation and image reconstruction for each frequency, and (iii) also a summation of the individual free induction decays to increase SNR for PFCs with multiple signals. For quantification of the 19F signal, ROIs were drawn around the area of interest, whereas background ROIs of similar geometry were placed outside the samples. The signal-to-noise ratio was calculated from the mean of the background-corrected 19F signal divided by the standard deviation of the noise. For phantom measurements, single frequency as well as sum images were quantified from identical ROIs covering the cross sections of the individual PFC-filled vials. For a more detailed description of the 19F MRI approach, acquisition parameters, and quantification procedures, please refer to references9,64.
Cellular uptake of neat, PEGylated and PEGylated functionalized PFCs
Preparation of Atto647PFCs
Atto647 labeled PFCs were prepared as described above. In brief 100 µg of Atto647-DPPE (DPPE = 1,2-dipalmitoyl-sn-glycero-3-phosphoethanolamine; ATTO-TEC GmbH) was resuspended in phosphate-glycerol buffer containing solubilized E80S lipids and stirred for 30 min at room temperature. Subsequently, PFCs were added and a pre-emulsion was generated by high-shear mixing. The crude emulsion was subjected to five cycles of microfluidization (LV1, Microfluidics) at 1000 bar.
PEGylation and functionalization of PFCs
PEGylated PFCs were generated by incubation of 100 µl of Atto647PFCs with 100 µl of cholesterol-PEG2000-CH3 (45 µl PEG + 55 µl buffer). To generate ligand-functionalized PFCs, Fbn or α2AP were conjugated to cholesterol-PEG2000-maleimide and then inserted into the lipid surface of the PFCs by SPIT. To this end, 12.2 µl Fbn or 13.4 µl α2AP (2.5 µg/µl each) were incubated with 39.2 µl cholesterol-PEG2000-maleimide. The volume was adjusted to 100 µl with phosphate glycerol buffer and incubated overnight at 20 °C under constant motion. On the next day, 100 µl of the Fbn/α2AP conjugate and additional 12 µl of cholesterol-PEG2000-CH3 (10 µg/µl) were added to 100 µl of Atto647PFCs and incubated for 3 h at 20 °C.
Cellular uptake of PFCs by RAW macrophages and murine monocytes
RAW macrophages were detached from the culture flask by strong shaking of the flask. Approximately 5 × 105 cells were taken up in 500 µl of culture medium and incubated with 25 µl of PFCs at 37 °C for 20 min. Subsequently, cells were transferred into a FACS tube filled with 1 ml of ice cold FACS buffer (PBS, 2% FBS, 1% EDTA). Cells were washed twice with FACS buffer, finally resuspended in 500 µl FACS buffer with DAPI (1 µg/ml; to label dead cells) and analyzed on a FACS Canto II (BD Biosciences). For analysis of PFC uptake by murine monocytes, blood was obtained by venous puncture of heparinized mice. Subsequently, the withdrawn blood (~100 µl) was subjected to hypotonic erythrocyte lysis (3 ml) for 5 min at room temperature, pelleted by centrifugation and washed with ice cold FACS buffer. Incubation with PFCs was again carried out for 20 min at 37 °C. Afterward, cells were washed and labeled with antibodies against CD11b (CD11b-PE, clone M1/17; BD Biosciences) and Ly6G (Ly6G-FITC, clone 1A8, BioLegend) for 30 min at 4 °C, washed twice with FACS buffer, stained with DAPI and analyzed by flow cytometry. Monocytes were identified as CD11b+/SSClow and Ly6G negative. Mean fluorescence signals of RAW macrophages and murine monocytes were recorded in the APC channel and fluorescence signals of cells which were not exposed to PFCs were subtracted from the data of cells incubated with neat, PEGylated, and PEGylated functionalized PFCs. Please refer to Supplementary Fig. 16 for the gating scheme used to analyse RAW macrophages and murine monocytes.
In vitro generation and labeling of thrombi
Platelet rich plasma (PRP) was obtained by venous puncture from healthy volunteers and immediately centrifuged at 200 g for 10 min. The white PRP on the top of the sample was carefully withdrawn, separated in 1 ml aliquots and stored at −20 °C. For generation of acute thrombi, 1 ml of the PRP was melted on ice and mixed with 100 µl of thrombin (5 U/ml) and 10 µl of ADP (1 mM). Then 100 µl samples were transferred to wells of a round-bottom 96-well plate and incubated for 5 min. Next, α2APPFOB and/or fbnPFCH were added and incubated for additional 90 min. Thrombi were removed and washed three times with 5 ml of PBS and analyzed by IVIS (see below) or 19F MRI. To obtain subacute thrombi, PRP was activated with thrombin and ADP (as described above) and 100 µl samples were incubated in a round-bottom 96-well plate for 90 min. Formed thrombi were removed from the well, washed with PBS and incubated with fbn or ctr peptide or fbnPFCH/α2APPFOB (or the respective controls). To determine the binding of PFCs to chronic thrombi, thrombi were stored in PBS for 24 h, three days and five days at 4 °C and then incubated with the targeting probes, intensively washed with PBS and analyzed by IVIS.
Imaging of in vitro generated thrombi
Thrombi labeled with fbn/α2AP-peptides or fbn/α2APPFCs were spotted on a glass plate and the fluorescence signal of the carboxyfluorescein was determined by an IVIS Lumina II imaging system (Perkin Elmer). For quantification, ROIs were drawn around the thrombus-associated fluorescence signals and the background-corrected mean values were determined. For 19F MRI, thrombi were placed in 200 µl reaction tubes filled with PBS which were fixed at a 2 mL water tube.
Induction of deep venous thrombi in vivo
Mice were kept under anesthesia with 1.5% isoflurane and buprenorphin (0.3 mg/kg) was injected s.c. 30 min prior to surgery for analgesia. A median laparotomy was performed, and the inferior vena cava was exposed at the anatomical level of both kidneys. Subsequently, a filter paper (1 × 2 mm2) soaked with 10% FeCl3 was placed on the top of the vessel and incubated for 4 min15. To prevent any leakage of the FeCl3 to the surrounding tissue, two stretches of parafilm were placed on both sides of the vessel. After removal of the filter paper the vessel was washed with 0.9% NaCl to remove residual FeCl3. PFCs (3 mmol/kg body weight) were injected into the tail vein approximately 5 min prior to or 24 h post thrombus induction. Combined 1H/19F MRI scans were performed 24 h after PFC injection.
Data reporting
No statistical methods were used to predetermine sample size. The experiments were not randomized and the investigators were not blinded during experiments and outcome assessment. Due to the highly variable phenotype of the used animal model, each animal was analyzed on its own and included into the study.
Statistics and reproducibility
Unless otherwise indicated, all values are given as mean ± standard deviation (SD). Statistical and regression analysis was performed using OriginPro 2016 (Originlab Corporation, Wellesley Hills, USA). Data were tested for Gaussian distribution using D’Agostino and Pearson omnibus normality test. Statistical significance was assessed by Student’s two-sided t test compared to the respective control group. Histology was carried out in 22 mice and the representative findings displayed in the supplement were confirmed in at least 6 independent experiments.
Reporting summary
Further information on research design is available in the Nature Research Reporting Summary linked to this article.
