Ultrasound Molecular Imaging of Atherosclerosis With Nanobodies: Translatable Microbubble Targeting Murine and Human VCAM (Vascular Cell Adhesion Molecule) 1.

OBJECTIVE
Contrast-enhanced ultrasound molecular imaging (CEUMI) of endothelial expression of VCAM (vascular cell adhesion molecule)-1 could improve risk stratification for atherosclerosis. The microbubble contrast agents developed for preclinical studies are not suitable for clinical translation. Our aim was to characterize and validate a microbubble contrast agent using a clinically translatable single-variable domain immunoglobulin (nanobody) ligand. Approach and Results: Microbubble with a nanobody targeting VCAM-1 (MBcAbVcam1-5) and microbubble with a control nanobody (MBVHH2E7) were prepared and characterized in vitro. Attachment efficiency to VCAM-1 under continuous and pulsatile flow was investigated using activated murine endothelial cells. In vivo CEUMI of the aorta was performed in atherosclerotic double knockout and wild-type mice after injection of MBcAbVcam1-5 and MBVHH2E7. Ex vivo CEUMI of human endarterectomy specimens was performed in a closed-loop circulation model. The surface density of the nanobody ligand was 3.5×105 per microbubble. Compared with MBVHH2E7, MBcAbVcam1-5 showed increased attachment under continuous flow with increasing shear stress of 1-8 dynes/cm2 while under pulsatile flow attachment occurred at higher shear stress. CEUMI in double knockout mice showed signal enhancement for MBcAbVcam1-5 in early (P=0.0003 versus MBVHH2E7) and late atherosclerosis (P=0.007 versus MBVHH2E7); in wild-type mice, there were no differences between MBcAbVcam1-5 and MBVHH2E7. CEUMI in human endarterectomy specimens showed a 100% increase in signal for MBcAbVcam1-5versus MBVHH2E7 (20.6±27.7 versus 9.6±14.7, P=0.0156).


CONCLUSIONS
CEUMI of the expression of VCAM-1 is feasible in murine models of atherosclerosis and on human tissue using a clinically translatable microbubble bearing a VCAM-1 targeted nanobody.

C omplications of atherosclerotic disease (myocardial infarction, stroke) are responsible for a large proportion of morbidity and mortality. Progress has been made in the treatment of acute complications of atherosclerosis, which has led to a reduction in mortality, particularly in patients suffering a myocardial infarction. 1 At the same time, assessment of an individual's risk for atherosclerotic disease complications continues to suffer from major drawbacks with risk models placing a large proportion of adults in western countries in an intermediate risk category. 2,3 In this group, methods such as noninvasive imaging tools that detect early atherosclerotic changes are a clinical need. The detection of an atherogenic phenotype at an early stage may also be essential for selecting patients for emerging therapies that are aimed at interrupting pathophysiologic mediators of plaque formation or that may lead to plaque regression. Endothelial inflammatory activation with upregulation of VCAM (vascular cell adhesion molecule)-1 promotes recruitment of monocytes into the arterial wall, 4 precedes plaque development, and plays a role in plaque initiation and progression. 5 Therefore, imaging techniques that detect the expression of VCAM-1 could be used to assess an individual's risk for future atherosclerosis-related events. In recent years, the feasibility of contrast-enhanced ultrasound molecular imaging (CEUMI) for detecting the expression of VCAM-1 using full-size antibodies has been shown in animal models. [6][7][8][9] However, no clinical studies using this technique have been performed thus far. For clinical translation, biotin-streptavidin linking of ligands to the microbubble used in the aforementioned studies will need to be replaced as it could potentially lead to binding of endogenous biotin, 10 and, more importantly, regarding full-size antibody ligands, there is concern about the use in humans both in terms of safety and costs. 11 Nanobodies or single domain antibodies are antibody fragments consisting of a single monomeric variable antibody domain derived from heavy-chain-only antibodies that are by nature present in camelids. Nanobodies are the smallest possible (10 to 15 kDa) antibody-derived polypeptide structure that binds to a specific antigen. 12 Nanobodies possess advantages over conventional antibodies for clinical applications. They lack an Fc region and therefore do not induce complement-triggered cytotoxicity nor bind to Fc receptors on immune and other type of cells. In addition, there is homology between camelid and human antibody heavy chains, which makes humanization of nanobodies unproblematic. 13 In preclinical studies, nanobodies targeted to VCAM-1 have already been used for imaging of atherosclerotic lesions using single photon emission computed tomography or positron emission tomography imaging, [14][15][16][17] and a GMP-produced variant is on track for testing in a phase I nuclear imaging study in patients. VCAM-1 targeting with nanobodies attached to microbubbles using biotin-streptavidin bridging has been recently accomplished in a mouse tumor model. 18 In the current study, we, therefore, studied the use of a microbubble with maleimide-thiol conjugation of an anti-VCAM-1 nanobody to detect VCAM-1 expression in a mouse model of atherosclerosis and ex vivo in human endarterectomy specimens using noninvasive ultrasound imaging.

MATERIAL AND METHODS
The data that support the findings of this study are available from the corresponding author upon reasonable request.

Study Design
Microbubbles carrying VCAM-1 targeted nanobodies were characterized using in vitro fluorescence and flow-chamber studies. All animal experiments were performed in accordance with Swiss Federal Legislation and approved by the local Animal Care and Use Committee of the University Hospital of Basel and the ethics committee of the Veterinary Office of the Canton of Basel. Thirty-two male mice (B6;129S-Ldlr tm1Her Apob tm2Sgy /J) referred to as double knockout (DKO) mice, underwent molecular imaging either at the age of 10 weeks (n=15) or 40 weeks (n=17). These mice are deficient in low-density lipoprotein receptor and apo B 48 . Deficiency in apo B 48 is accomplished by introduction of a point mutation into the apo B 48 -editing codon and thus only apo B 100 can be synthesized. DKO mice develop atherosclerosis in a time-dependent fashion while on a normal laboratory diet (cat. no. 3436; Provimi Kliba AG, Switzerland). 19 Male mice were chosen based on the extensive experience in our laborartory regarding the time course of plaque development in males. Reproducible plaque development with mild intimal thickening at age 10 weeks and widespread lesions with luminal encroachment at 40 weeks in the ascending aorta have been shown in this animal model using histological methods. 7 Thirty-two C57BL/6J wild-type (WT) male mice on normal laboratory diet were used as controls (n=15 at the age of 10 weeks, n=17 at 40 weeks). For ex vivo validation using human tissue, carotid (n=5) and femoral (n=2) endarterectomy specimens were obtained from patients undergoing clinically indicated endarterectomy at the University Basel Hospital Switzerland. The study conformed to the Declaration of Helsinki and was approved by the Ethical Committee of Northwestern and Central Switzerland. All subjects gave written informed consent for participation.

Nanobody Density on the Microbubble Surface
For assessing the amount of nanobody (cAbVcam1-5) required to saturate maleimide binding sites, microbubbles were incubated with increasing amounts of fluorescently labeled cAbVcam1-5 (1, 10, 20, 60, 100, 140 µg per 1×10 8 microbubbles). After washing, the amount of fluorescence retained on the microbubble surface was measured with Accuri C6 flow cytometer (BD Biosciences). To quantify the site density of nanobodies on the microbubble surface after conjugation to maleimide, cAbVcam1-5 was fluorescently labeled using N-Hydroxysuccinimide-fluorescein (Thermo Fisher Scientific). This procedure resulted in a ratio of 2 fluorescein molecules per nanobody. MB cAbVcam1-5 obtained using fluorescently labeled cAbVcam1-5 were destroyed by application of pressure. The concentration of fluorescein-cAbVcam1-5 was then measured using a Synergy H1 microplate reader (Biotek) and compared with a reference standard of known concentrations of fluorescently labeled cAbVcam1-5. Microbubble concentration and size measurements obtained before destruction were used to calculate ligand surface density and number of ligands conjugated per microbubble.

Flow-Chamber Experiments
Parallel plate flow-chamber adhesion assays using activated bEnd.3 (ATCC CRL2299) murine endothelial cells expressing VCAM-1 were performed. Fibronectin coated (45µg/mL) 35×10 mm culture dishes (Corning) were seeded with bEnd.3 murine endothelial cells. Cells were grown to confluence in DMEM (ATCC20-2002) containing 10% FBS and 1% penicillin/streptomycin. The bEnd.3 endothelial cells were activated with tumor necrosis factor-alpha (50 ng/mL) 240 min before flow-chamber experiments. The culture plates were mounted in an inverted position in a parallel plate flow chamber (Glycotech, Inc) with a 0.254 mm gasket thickness and a 2.5 mm channel width. MB cAbVcam1-5 or MB VHH2E7 (3×10 6 /mL) were drawn through the flow chamber using a syringe pump (Genie plus, Kent Scientific). The speed of the syringe pump was set to obtain wall shear stresses of 1, 2, 4, and 8 dynes/cm 2 for adhesion studies. Binding of fluorescently labeled microbubbles was quantified on 20 randomly selected visual fields using a fluorescent microscope (Olympus BX51WI) and a ×40 magnification.
In separate experiments on separate culture plates, binding of MB cAbVcam1-5 versus MB mAbVcam1 microbubbles was assessed at 2 dynes/cm 2 . Because aortic flow is pulsatile, adhesion at the highest shear rate of 8 dynes/cm 2 was also assessed after transient (5 s) reductions of shear to <0.5 dynes/cm 2 . After maximal attachment of MB cAbVcam1-5 at lowest shear stress of 0 to 0.5 dynes/cm 2 , the detachment characteristics of the microbubbles were determined by sequentially increasing the shear stress by 5 dynes/cm 2 up to a maximum of 43 dynes/cm 2 .

Animal Instrumentation
For echocardiography and CEUMI, mice were anesthetized with inhaled isoflurane (1% to 1.5% in room air for maintenance), and body temperature was maintained at 37°C with a heating pad. The chest was depilated. For microbubble injections, the right internal jugular vein was cannulated (PE 50 tubing), and the animals were transferred onto a temperaturecontrolled imaging stage (Vevo Imaging Station). The heart rate was monitored.

Echocardiography
High-frequency (40 MHz, MS 550D transducer) ultrasound imaging (Vevo 2100, VisualSonics, Inc) was performed to assess the left ventricular ejection fraction, as well as aortic arch diameter and flow velocity. M-Mode images of the left ventricle in the parasternal short-axis plane at the midpapillary muscle level were used to measure left ventricular internal diameters at end-diastole and at end-systole. Left ventricular ejection fraction was calculated using the cube formula (left ventricular ejection fraction (%)=100×[(left ventricular internal diameters at end-diastole 3 -left ventricular internal diameters at end-systole 3 ) / left ventricular internal diameters at end-diastole 3 ]). 22 Parasternal long-axis images of the aortic arch were used to measure internal diameter and the centerline peak systolic velocity in the aortic arch just distal to the brachiocephalic artery was measured by pulsed-wave doppler.

CEUMI in Murine Atherosclerosis
CEUMI (Sequoia Acuson C512; Siemens Medical Systems) was done with a high-frequency linear-array probe (15L8) secured in place by a railed gantry system. The ascending aorta of the mice, including the sinus of valsalva, and the takeoff of the brachiocephalic artery was imaged in a long-axis plane from a right parasternal window. Using a combination of power modulation and pulse inversion (Contrast Pulse Sequence), the contrast microbubbles were imaged at a transmission frequency of 7 MHz and a dynamic range of 50 dB. The gain settings were adjusted to levels just below noise speckle and maintained constant. MB cAbVcam1-5 and MB VHH2E7 (1×10 6 microbubbles per injection) were injected intravenously in random order. Ultrasound imaging was paused from the beginning of microbubble injection for 8 minutes. Imaging was then resumed at a mechanical index of 0.87. The first acquired image frame was used to derive the total amount of microbubbles present within the aorta. The microbubbles in the ultrasound beam were then destroyed using several (>10) image frames at a mechanical index of 0.87. Several image frames (n=3) at a long-pulsing interval (10 s) were then captured to measure signal from freely circulating microbubbles. The video intensities were log-linear converted, and frames representing freely circulating microbubbles were averaged and subtracted digitally from the first image to derive signal from attached microbubbles alone. Contrast intensity was measured from a region of interest encompassing the sinus of Valsalva, the ascending aorta, and extending into the origin of the brachiocephalic artery. The selection of the region of interest was guided by fundamental frequency anatomic images of the ascending aorta and the aortic arch acquired at 14 MHz at the end of each individual imaging sequence.

Contrast-Enhanced Ultrasound in Human Endarterectomy Specimens
Endarterectomy specimens were obtained directly from the operating theater, stored in NaCl 0.9% at 4°C, and the imaging procedures were performed within 2 hours. The tissue was put inside a polyethylene tubing (inner diameter 10 mm) with the endothelial layer facing inside and immobilized using rexolite adhesive (San Diego Plastics) applied on the abluminal surface resulting in a lumen diameter from 5.5 to 8 mm depending on the thickness of the endarterectomy specimen. The tubing was immersed into a water bath (37°C) and connected to a closed-circuit filled with pooled human serum (6 donors, blood group AB) obtained from the local blood bank. Inlet and outlet ports in the circuit were used for microbubble administration and removal. A peristaltic pump (Masterflex, easy load II) was used to generate flow within the circuit (total circuit volume 25 mL, shear stress ≈3-9 dynes/ cm 2 ). A 15L8 probe immersed into the water bath was used to acquire B-mode images (14MHz) of the endarterectomy specimens ( Figure I in the online-only Data Supplement for experimental setup). Boluses of 2.5×10 7 MB cAbVcam1-5 and MB VHH2E7 were injected into the circuit in random order. Seven minutes after microbubble injection, freely circulating microbubbles were reduced by replacing the fluid volume in the circuit by injection of 25 mL of serum. Microbubble signal from attached and remaining circulating microbubbles was recorded using CPS imaging at 7 MHz with a mechanical index of 0.87. Attached microbubbles were destroyed with >10 frames (mechanical index of 0.87).
Signal from remaining circulating microbubbles was recorded at a 10-s pulsing interval. Signal from attached microbubbles was derived by subtracting 3 averaged postdestruction frames from the first image frame before microbubble destruction. Targeted signal was measured from a region of interest encompassing the tissue-fluid interface derived from 14MHz fundamental frequency images obtained at the end of each sequence.

Immunohistology
For excision of the thoracic aorta, a limited thoracotomy was performed under deep sedation. The blood volume was removed with direct cardiac puncture, and the aorta was perfused with 10% formalin. Connective tissue and fat were removed from the thoracic aorta and the ascending portion, and arch, including the takeoff of the arch vessels, was excised. Immersion fixation of the tissue in 4% paraformaldehyde was performed overnight.
Tissues were subsequently embedded in paraffin and cut into 5-µm sections. Immunohistochemistry was performed to qualitatively assess the endothelial expression of VCAM-1 by using a rabbit monoclonal antibody to mouse VCAM-1 (ab134047; Abcam) or a matched isotype IgG control (ab172730; Abcam, Figure III in the online-only Data Supplement) at a 1.4 µg/mL concentration, and a goat anti-rabbit IgG-horseradish peroxidase as a secondary antibody (P0448, Dako). Before insertion of endarterectomy specimen into the closed-circuit, part of each tissue was set aside and snap-frozen in optimal cutting temperature medium by immersion in liquid nitrogen. Ten micrometer sections were stained with Masson trichrome and adjacent sections incubated with a rabbit polyclonal to human VCAM-1 (ab106777; Abcam) or a matched isotype IgG control (011-000-003, Jackson ImmunoResearch, Figure IV in the online-only Data Supplement) at a 5 µg/mL concentration and a goat anti-rabbit IgG-Alexa Fluor 546 as a secondary antibody (A-11010, Thermo Fisher Scientific) plus DAPI (4',6-diamidine-2'-phenylindole dihydrochloride; D1306, Thermo Fisher Scientific). Microscopy imaging was performed with a Nikon TI microscope.

Statistical Analysis
Data are expressed either as medians and 25th to 75th percentile or as mean±SD, as appropriate.

Characterization of Targeted Microbubbles
The mean size of microbubbles used for characterization was 2.88±1.61 µm, the mean concentration 8.5×10 7 microbubbles/mL. Flow-cytometry analysis demonstrated the optimal conjugation saturation of the microbubble surface with nanobodies ( Figure 1A) to be reached at an incubation with 140 µg nanobody per 1×10 8 microbubbles. At this concentration, the fluorescent spectroscopy calculations revealed 3.5×10 5 cAbVcam1-5 molecules coupled to the microbubble surface which corresponded to ≈1 cAb-Vcam1-5 molecule per 74 nm 2 (8.6×8.6 nm) or 1.3×10 4 cAbVcam1-5 molecules per µm 2 of the microbubble surface ( Figure 1B) and a calculated coupling yield of 0.7%.  Figure 2B). After adhesion at low shear stress, MB cAbVcam1-5 showed minimal detachment up to a shear rate of 25 dynes/cm 2 , with increased detachment at 30 dynes/cm 2 , and significant detachment compared with baseline reached after 35 dynes/cm 2 (P<0.05 versus baseline; Figure 2C). When compared with microbubbles carrying a monoclonal antibody to VCAM-1 (MB mAbVcam1 ), retention of MB cAbVcam1-5 tended to be less; however, this difference was statistically nonsignificant ( Figure II in the online-only Data Supplement).

In Vivo CEUMI of VCAM-1 Expression in the Mouse Aorta
Body weight and heart rate were not different between DKO and WT mice at age 10 and 40 weeks. Left ventricular ejection fraction was in the normal range in both DKO and WT mice at 10 and 40 weeks but was significantly lower by 7 to 8 percentage points in DKO mice. At 40 weeks of age, internal aortic diameters and peak aortic flow velocity were increased in DKO mice (Table). Microbubble preparations used for in vivo CEUMI had mean sizes of 2.63±1.34 µm for MB cAbVcam1-5 and 2.44±1.24 µm for MB VHH2E7 (P=0.2 for comparison of the 2 mean sizes); microbubble concentrations were 2.7×10 8 and 2.5×10 8 microbubble/mL, respectively. CEUMI showed selective signal enhancement for MB cAb-Vcam1-5 in DKO mice only, both in very early stages of atherosclerosis at age 10 weeks and in established atherosclerosis at age 40 weeks. Signal enhancement was approximately 3-fold increased over MB VHH2E7 control signal at 10 weeks, whereas at 40 weeks this ratio was ≈2-fold. In WT mice, there were no differences in signal between MB cAbVcam1-5 and MB VHH2E7 both at 10 and 40 weeks, but the signal tended to increase for both microbubbles at 40 weeks (Figures 3 and 4).
Histology of 40 weeks old DKO mice ( Figure 5A) showed large atherosclerotic plaques and abundant expression of VCAM-1 both on the endothelial surface and in plaque macrophages compared with normal histology and low endothelial staining for VCAM-1 in WT mice ( Figure 5B). At 10 weeks, DKO mice ( Figure 5C) showed a moderate increase in endothelial expression of VCAM-1 in comparison to WT mice ( Figure 5D).

Ex Vivo CEUMI of VCAM-1 Expression in Endarterectomy Specimens
Imaging was performed in a total of 7 specimens: 5 from carotid endarterectomy and 2 from femoral endarterectomy. Overall, there was a statistically significant 100% signal increase, P=0.0156 for MB cAbVcam1-5 as compared to MB VHH2E7 from microbubble attachment to the endothelial surface of the human endarterectomy specimens  ( Figure 6A). In all specimens, retained microbubbles could be appreciated visually as a linear contrast enhancement on the endothelium ( Figure 6B). There were large differences in signal obtained between specimens; however, as shown in Table I in the online-only Data Supplement, in every single tissue, signal for MB cAbVcam1-5 was larger compared to MB VHH2E7 . Immunohistology of the endarterectomy specimens showed atherosclerotic plaques and expression of VCAM-1 both on the endothelial surface as well as in plaque macrophages ( Figure 6C).

DISCUSSION
In this study, we have characterized a microbubble contrast agent targeted to VCAM-1 with a single-variable domain immunoglobulin (nanobody) that is feasible for human use. We show that this microbubble detects the expression of VCAM-1 in murine models of both early and established atherosclerosis. In addition, we show that the same agent can be used to detect the expression of VCAM-1 on human endarterectomy specimens with ultrasound imaging (Figure 6). Nanobodies possess several advantages over full-size antibodies for human use. This includes easy large-scale production in prokaryotic and eukaryotic hosts, lack of Fc, amenability for site-directed immobilization, and a high degree of sequence homology (>80%) with human variable heavy chain region antibodies. In fact, caplacizumab, a nanobody targeting von Willebrand factor, has recently received clinical approval for treatment of adults  experiencing an episode of autoimmune thrombotic thrombocytopenic purpura. 23 Successful in vivo targeting of nanobody-contrast agents to VCAM-1 has been shown either in low-shear stress conditions or using small molecule tracers subject to less drag forces compared with microbubbles. Targeting of microbubbles to disease markers in large arteries needs to rely on high targeting efficiency with fast on-rates and low off-rates of the ligand used. We demonstrate that using maleimide coupling, 3.5×10 5 nanobodies are attached to a microbubble, which is comparable to other conjugation strategies, 18 and well above thresholds that negatively impact targeting efficiency. 24 In our flow-chamber experiments, retention of microbubbles decreased as continuous shear stress increased. However, brief reductions in shear stress resulted in sequential increases in targeted microbubble retention (fast on-rate), and firmly attached microbubbles were able to withstand shear forces up to 25 to 30 dynes/cm 2 that are within the range of peak wall shear stress values encountered in large human arteries (low off-rate), 25 indicating that the cAbVcam1-5 nanobody is feasible for VCAM-1 targeting in pulsatile high shear stress. This translated to robust signal from MB cAbVcam1-5 when noninvasively imaging the aorta in mouse models of early and late atherosclerosis, while signal was no different from control in WT animals. Interestingly, the targeted signal to control signal ratio was higher in early versus late atherosclerosis, which could potentially be explained by higher background signal from nonspecific attachment of MB VHH2E7 to leukocytes. 26 Molecular imaging of VCAM-1 in humans could cover several unmet clinical needs. This includes risk assessment in individuals that are classified as having an intermediate risk by traditional risk markers and assessment of plaque vulnerability in patients with established atherosclerotic disease. The rationale for selecting VCAM-1 for detection and risk stratification during the initial disease process is based on evidence from preclinical studies implicating VCAM-1 in very early endothelial inflammatory activation 5 and contribution to leukocyte recruitment to the vascular wall. 27 As a consequence, tracers for imaging the expression of VCAM-1 have been developed for magnetic resonance and nuclear imaging, as well as for ultrasound. 6,14,28 The novelty reported in our study is the use of an ultrasound contrast agent for VCAM-1 imaging that is fully clinically translatable which has been validated for use in high-shear conditions, and has been tested on human tissue under flow conditions. With respect to risk assessment in large populations, ultrasound imaging is well suited due to its wide availability and low cost compared to other techniques. However, ultrasound molecular imaging will have to rely on detection of the expression of a disease marker in arteries easily accessible for imaging such as the carotid or femoral arteries rather than directly assessing coronary arteries that with current ultrasound equipment demand invasive imaging and are subject to motion artifacts. Using endarterectomy specimens, we show that MB cAbVcam1-5 can be used to detect the expression of VCAM-1 present on plaques from human tissue under continuous shear conditions that are well above diastolic wall shear stress values of around 6 dynes/cm 2 and close to the peak systolic shear stress reported in one study, 29 whereas another study reported higher values for peak systolic shear stress. 30 The results from our flow-chamber studies with incremental attachment during pulsatile flow and retention up to shear forces of 25 to 30 dynes/cm 2 together with successful targeting of VCAM-1 in the murine aorta indicate that in vivo imaging of VCAM-1 expression in humans will be feasible. However, the incremental value of molecular imaging will depend on whether inflammatory activity in a reference vessel correlates with atherosclerotic burden in vessels of interest, such as the coronary arteries. Data from large animal models show that the endothelial expression of VCAM-1 in carotid arteries correlates with plaque burden in the coronary arteries and that VCAM-1 in that respect outperforms measurement of carotid intima-media thickness or traditional risk factors, such as blood glucose and cholesterol values. 31 Cross-sectional studies in humans have shown that soluble VCAM-1 measured in the serum predicts the presence of coronary artery disease in symptomatic patients referred for coronary angiography, 32 and that values for soluble VCAM-1 are particularly elevated in patients with high levels of pro-inflammatory low-density lipoprotein triglycerides. 33 Another clinical area where noninvasive assessment of VCAM-1 expression could be valuable is in assessment of plaque vulnerability. According to current clinical guidelines, treatment of carotid artery stenosis with endarterectomy is recommended in patients with recent focal neurological symptoms and in asymptomatic patients with >70% stenosis. The benefit of surgery is, however, much reduced in asymptomatic patients where carotid artery stenosis carries an annual stroke risk of only 0.5% to 1%. 34 Further risk stratification in this population could be of value. In that respect, recent data show increased endothelial levels of VCAM-1 on histology from highrisk versus low-risk carotid endarterectomy specimens, 35 suggesting that noninvasive imaging of VCAM-1 expression may add incremental prognostic value in patients with asymptomatic carotid artery stenosis. In addition, intraplaque neovessels and neovessels inflammation are markers of plaque vulnerability and may contribute to increased signal upon in vivo imaging. 36 Several limitations of our study deserve mentioning. First, the coupling yield of the cAbVcam1-5 nanobody to the microbubbles was very low, and thus a large amount of the nanobody was lost during microbubble preparation. However, by using flow-cytometry data ( Figure 1) to select an incubation with 140 µg per 1×10 8 microbubble, we traded a lower coupling yield for a somewhat higher surface density. Figure 1 shows that using lower amounts of nanobody for incubation would not have dramatically decreased saturation, but would have resulted in a better coupling yield. Also, despite >90% desalting of the nanobody after the reduction step, remaining 2-mercaptoethylamine could have had an impact on the coupling yield. Thus, for translation purposes, microbubble preparation protocols that minimize loss of nanobody while retaining targeting efficiency will have to be developed. Second, the animal model that we used is representative for early and late stages of stable atherosclerotic disease. However, vulnerable plaques do not develop in this animal model, and therefore, studies in suitable animal models will be necessary. Third, we evaluated the targeting efficiency of MB cAbVcam1-5 versus MB VHH2E7 in human plaques but for ethical reasons did not have normal arterial tissue available. We can, therefore, not completely exclude the possibility that the high signal obtained for MB cAbVcam1-5 represents nonspecific attachment. However, this was not the case in the animal model with signal not different for MB cAbVcam1-5 versus MB VHH2E7 in WT animals. In the closed-loop model, we used pooled human serum to closely mimic in vivo conditions; however, limited availability of serum allowed for replacing the serum inside the system only once. Thus, transient attachment of microbubbles to the conduit tubing resulted in a relatively large concentration of remaining circulating microbubbles, which may have mitigated the targeted signal enhancement to some degree. Last, the technical setup of the closed-loop cardiovascular circulation model for CEUMI on human endarterectomy specimens did not allow for creation of pulsatile flow. However, the continuous shear stress of 3 to 9 dynes/ cm 2 used in the model is well above reported diastolic shear stress values in human carotid arteries, 29,30,37 and we would, therefore, not expect a negative effect on targeting efficiency in vivo.
In summary, we have characterized a targeted contrast agent using a nanobody with maleimide covalent binding to the microbubble surface for detection of VCAM-1 in large arteries. Using this contrast agent, vascular inflammation during early and late stages of stable atherosclerosis can be diagnosed with noninvasive imaging in animal models. Importantly, we show that this contrast agent allows detection of VCAM-1 on human tissue with ultrasound imaging. These findings pave the way for clinical translation of CEUMI to detect early pathophysiological changes and improve risk stratification for atherosclerotic complications.