Compared with other medical imaging modalities, ultrasound imaging has several advantages. It does not involve radiation, has high spatial resolution and is generally considered cheap, safe and widely available. The introduction of gas-filled microbubbles as contrast agents for ultrasound imaging has further improved the performance and increased the versatility of ultrasound imaging.

In in vivo molecular imaging, targeted probes are used for non-invasive imaging of molecules overexpressed in disease. This principle can be used also in ultrasound imaging. Conjugating microbubbles with specific ligands allows non-invasive imaging of target expression with high sensitivity. However, microbubbles are confined to the intravascular compartment. Their use as molecular contrast agents is therefore restricted to diseases which are directly or indirectly associated with altered phenotype of cells present in the intravascular compartment. Nevertheless, several interesting indications for targeted microbubbles exist. For example in atherosclerosis, subendothelial lesions cause inflammatory changes in the endothelium. In solid tumors, the cancer cells stimulate the endothelium and induce neoangiogenesis, leading to upregulation of several molecules on the endothelial surface.

Over the last decade, molecular contrast-enhanced ultrasound imaging has been established as a useful technique for imaging intravascular angiogenic markers in cancer. Proof-of-concept studies in preclinical models have repeatedly demonstrated that retention of targeted microbubbles, and thus the signal intensity of the ultrasound images, reflects the expression of target biomarkers. Here, we summarize the findings from preclinical studies of molecular ultrasound contrast agents with the intention to describe the potential clinical utility of these agents in ultrasound imaging of tumor vasculature.

Contrast agents for ultrasound consist of micrometer-sized gas bubbles stabilized by a thin encapsulating shell made of lipid, albumin or biocompatible polymers which are intravenously administered. Such microbubbles can undergo strong oscillations when exposed to an ultrasound beam. This leads to strong back-scattering of waves enabling detection of the microbubbles and thereby assessment of micro-circulation and perfusion. In the early eighties, Feinstein [1,2] was one of the first to develop air microbubbles encapsulated in protective shells small enough for transpulmonary passage. About 10 years later, the first generation of commercial contrast agents (such as Echovist^®, Albunex^®, Levovist^®) was available. A few microbubble agents with mean diameter of 1–4 microns (Definity^®, Optison^®) are currently the only FDA approved contrast agents for ultrasound. The flow pattern of conventional non-targeted microbubbles is typically similar to erythrocytes within the circulation system [3]. Targeting can be achieved through conjugation of disease-specific ligands for the target molecule to the microbubble shell. Molecular imaging with ultrasound relies on detection of targeted microbubbles. Because of the large size of the microbubble, it is only possible to target molecules occurring in plasma or on the surface of the endothelial cells. Seconds after an intravenous bolus administration, contrast agent inflow provides real-time information of blood flow patterns within a region of interest. A few minutes later, imaging of microbubbles attached to molecular targets may be performed and 5–15 minutes after administration most of the free circulating microbubbles are typically cleared from the bloodstream (Figure 1). Differentiating targeted bubbles retained in diseased tissue from free circulating bubbles can be done either by waiting until all circulating bubbles are cleared from the bloodstream or by using a subtraction technique comparing the back-scattered signal before and after a high amplitude destruction pulse has been applied. Free flowing bubbles and bubbles bound to molecular targets at the endothelial cells will potentially have different resonance frequencies due to different oscillation conditions. It is then, for example with the use of dual frequency band detection techniques, possible to differentiate the received echoes from unbound and retained bubbles [4].

Cancer can be defined as diseases where cells divide in an uncontrolled manner and are able to invade other tissues. Solid tumors depend on formation of new blood vessels in order to grow beyond a size of 1–2 mm [19,20]. As changes in vascular architecture and function in the affected tissue is an integrated part of the pathology, assessment of blood vessels and their function is a useful approach in imaging of cancer [21,22]. The progression to invasive and eventually metastatic cancer depends on angiogenesis, and imaging of this process is an important approach for early detection and characterization of cancer.

Development of angiogenic capability is typically acquired at a relatively late time point during tumorigenesis [19]. To explain this, the “angiogenic switch” model has been developed [23]. The relative balance between inhibitors and activators of angiogenesis is in general in favor of vascular quiescence. At some point in tumor development, loss of inhibitors or increased amount of activators (for example due to cumulative mutations in oncogenic signaling pathways) turns the balance in favor of angiogenesis, ultimately tipping it in favor of new blood vessel growth. This understanding has led to the development of antiangiogenic drugs, which impair the vascularisation of tumors. Although this in theory should deprive tumors of oxygen and nutrients, the clinical efficacy of antiangiogenic drugs has turned out to be smaller than expected [24].

The tumor vasculature has a different morphology and biology than mature, quiescent vessels [25]. It is tortuous and leaky, and the blood flow is often compromised. The phenotype of the endothelial cells is changed in response to the angiogenic activator molecules, and they overexpress a number of proteins compared to normal endothelium. These proteins are potential targets for drugs or contrast agents. Findings from in vivo studies of ultrasound contrast agents targeting angiogenesis are summarized in Table 1.

The principle of personalized medicine in cancer has led to an increased need for methods that can detect response to treatment, including antivascular and antiangiogenic drugs. Identification of responders to targeted anticancer drugs will be increasingly important as more drugs reach clinical practice. Ultrasound imaging is a suitable modality for this purpose, since it is a non-invasive, portable and non-radiative modality. In addition, it can be used to assess vascular function at several levels, including perfusion, blood volume and the expression of endothelial vascular markers [60,61]. Several studies have demonstrated that targeted microbubbles can be used for non-invasive assessment of angiogenic markers. Korpanty et al. [30] demonstrated that accumulation of both endoglin- and VEGF-targeting microbubbles correlated with target expression and microvessel density (MVD) in pancreatic xenograft tumors. Treatment with anti-VEGF antibodies detectably reduced the binding of microbubbles to endothelial cells. Similar findings have been reported using the VEGFR-targeting BR55 microbubble in colon cancer xenografts, where both VEGFR expression, MVD and imaging signal was significantly reduced after anti-VEGF antibody treatment [26]. Using an α_vβ₃-targeted contrast agent, molecular ultrasound imaging has been shown to discriminate between responding and non-responding xenograft tumors [62]. Following bevacizumab treatment, decrease in α_vβ₃-expressing vasculature was paralleled by a relative reduction in tumor blood volume in the responding tumor model only. In these experiments, adhesion of targeted microbubbles was a more consistent marker for response to treatment than the relative blood volume. Combining several vascular parameters obtained in the same ultrasound examination may further improve the sensitivity and specificity, and possibly also allow prediction of response based on pre-treatment data [63].

Detection of microbubbles in ultrasound imaging is complex, depending on both biological (target expression) and physical (imaging protocol, microbubble characteristics) parameters. For evaluation of the abovementioned intravascular markers as targets for ultrasound contrast agents, direct comparisons in the same disease models using the same experimental conditions can nevertheless be valuable. As an example, microbubbles targeting VEGFR2 and α_vβ₃ have been studied in subcutaneous squamous cell carcinoma xenografts, where VEGFR2-targeted microbubbles were found to give 1.6-fold higher signal intensity [32]. Similar findings have been reported in an ovarian cancer xenograft model, in a study of dual VEGFR- and α_vβ₃-targeting microbubbles. Here, microbubbles targeting VEGFR alone were found to give higher signal intensity than microbubbles targeting α_vβ₃ alone. Dual-targeted microbubbles were superior to both the two single-targeted microbubbles [37]. However, a longitudinal study of antibody-based microbubbles targeting α_vβ₃, VEGFR2 or endoglin in xenograft models of breast, pancreatic and ovarian cancer did not show any difference in the performance of α_vβ₃ and VEGFR2-targeted microbubbles in any model at any time point [34]. In addition to addressing the relative imageability of three potentially useful angiogenic markers, this study demonstrated the importance of experimental model characterization. The signal intensity after administration of endoglin-targeting microbubbles decreased with tumor size, which may reflect both reduced endoglin expression and a relative decrease in tumor vascularity, which is a typical feature of experimental subcutaneous tumors. The study also demonstrated that microbubble retention reflect differences in target expression between different cancers. Endoglin-targeting microbubbles were therefore associated with higher signal intensity than α_vβ₃-targeting microbubbles in ovarian cancer xenografts, whereas this pattern was reversed in pancreatic cancer xenografts. This illustrates the importance of animal model characterisation and biomarker validation during development of targeted contrast agents. If the receptor expression in an animal model is not adequately described during the selection and optimization phases of development, extrapolation of the findings may lead to false conclusions. Finally, selection of the correct ligand has been shown to have great impact on the imaging performance and specificity of targeted microbubbles. This was illustrated by Willmann et al. in a study comparing three different ligands for the α_vβ₃ integrin [14]. In this study, knottin ligands outperformed both an RGD peptide and an anti-α_vβ₃-antibody. In summary, the findings from preclinical studies comparing microbubbles must be interpreted with caution. In most of the studies, direct comparison of microbubbles has not been the primary objective, and the studies have been designed to assess other aspects of molecular ultrasound imaging. Due to the variability in angiogenic biomarker expression between cancers, no superior target for molecular ultrasound imaging can be identified from these studies. The body of data demonstrates that microbubbles targeting different angiogenic markers reflect the target expression of the biological systems, a feature which can be of great clinical value.

The studies reviewed in this paper show that microbubbles targeting intravascular disease markers consistently give several-fold higher signal intensity than isotype control microbubbles. The difference between targeted and control microbubbles is in the same order of magnitude for most intravascular targets. Microbubble retention consistently reflects the expression of the target molecule across all studies. The in vivo performance of molecularly targeted contrast agents should not be compared without taking the properties of the model systems into account. The expression of target molecules related to angiogenesis is known to differ between models. Even within the same model, target expression can vary with time due to tumor growth or disease progression. There are also several factors related to the microbubble that can affect the target-to-background achieved in vivo. For example, the size of the microbubbles may affect the signal intensity that can be achieved. Therefore, it is important to compare bubbles of similar size when evaluating different target molecules [64]. Other important factors are the wall shear rate and the ligand density of the microbubbles. It has been shown that adhesion of microbubbles increase with ligand density, and comparison of different microbubbles must therefore take the ligand density into account [65]. For comparison of imaging performance, contrast agent candidates must therefore be tested in the same animal models under the same experimental conditions, as discussed in section 6. It is, however, becoming increasingly clear that microbubbles with dual or triple targeting are superior to microbubbles carrying only one ligand. Multiple targeting facilitates multiple bindings between the bubbles and the endothelium and increases the adhesion strength to the endothelium and the retention of microbubbles in the tissue (Figure 3). It has been shown that combined targeting of VEGFR2 and α_vβ₃ is advantageous compared to microbubbles targeting only one of these targets [37]. Furthermore, triple-targeting microbubbles (P-selectin, VEGFR2 and α_vβ₃) have been demonstrated to have even higher tumor retention than dual-targeting microbubbles [38]. Interestingly, sequential administration of single-, dual- and triple-targeted microbubbles in the same individual animals has demonstrated significant synergy effects of multiple targeting. Triple-targeted microbubbles gave approximately 4-fold higher video intensity than any single-targeted microbubble, and 40% higher signal intensity than dual-targeted microbubbles [38].

The studies summarized above demonstrate that the intravascular compartment in cancer contains several targets that can be imaged using contrast-enhanced ultrasound, for diagnostic or therapy monitoring purposes. For successful development of a targeted ultrasound particle, several aspects must be taken into account. Firstly, the microbubbles must be biocompatible and have physical/acoustic properties matching the intended use. Secondly, a relevant molecular target (or several targets) must be identified. Thirdly, the microbubbles must be decorated with ligands binding to this target with sufficient specificity and affinity, and at a sufficient density, to facilitate binding to the target in vivo. Finally, the prototype microbubbles must be evaluated in clinically relevant and well-characterized preclinical models. Most importantly, the targeted microbubbles must produce clinically relevant information that has incremental diagnostic value compared to other diagnostic procedures.

Based on the research summarized in this review, no single molecular target appears to be superior for ultrasound imaging of cancer. In terms of clinical translation, VEGFR2-targeting microbubbles are in early-phase clinical trials and may be the first to reach clinical use. Dual- or triple-targeting microbubbles have repeatedly been shown to outperform single-ligand microbubbles. It is therefore possible that future research will identify combinations of ligands which optimize the imaging performance and allow tailoring of contrast agents for specific purposes. A multi-purpose microbubble decorated with ligands both for angiogenesis and inflammation is an exciting possibility, which could reduce the relative development cost, as the same microbubble product can be used in different indications.

In conclusion, molecular contrast-enhanced ultrasound is well established as a method for functional studies of diseases involving vascular pathology. This has been demonstrated in cancer, where imaging of angiogenic markers with targeted microbubbles have been evaluated both for diagnostic imaging and for monitoring response to antiangiogenic treatment. Progress in ligand and conjugation chemistry has led to the development of contrast agents with clinically desirable properties, and clinical trials have been initiated. Future optimization of ligands and microbubbles may lead to contrast agents that are valuable in management of cancer. We believe that identification of applications which utilize the inherent strengths of ultrasound imaging will be crucial for the introduction of molecular contrast-enhanced ultrasound in the clinic.