Near-infrared optical imaging is a newer imaging technique that, coupled with sensitive enzymatically specific fluorescent beacons, shows much promise for earlier detection of many cancers and their in situ characterization. On the basis of animal studies demonstrating visualization of micrometastasis-sized tumors and the ability to evaluate therapeutic enzyme inhibition real-time, such imaging may be incorporated in the clinical imaging paradigm in the future, both to improve cancer screening as well as for monitoring therapy in individual patients. This review details some of the related biology, optical probe design, and required hardware, with in vivo cathepsin and matrix metalloprotinease imaging used as examples.
Imaging plays a key role in helping oncologists meet several goals: detection of solid tumors; detection of recurrence; and evaluation of the success of a treatment regimen. Imaging traditionally has represented a quantum leap in how disease can be detected, quantitated, and characterized compared with the physical exam. Given this intertwining between oncology and imaging, it is imperative that those involved with cancer care have an understanding of what imaging can achieve today, and the impressive technological advances, currently evaluated in mouse models, that are on the horizon for significantly altering imaging-use paradigms in clinical practice in the future. Here, we focus on a new detection technology that is composed of imaging hardware for visualization of near-infrared light coupled with i.v.-injected chemical probes that allow in vivo detection of specific protease activity.
For perspective, it is important to remember that much of the first 100 years of radiology since William Roentgen discovered X-rays has focused on anatomy, with pathology visualized by differences in the physical parameters of neoplastic tissue compared with normal adjacent parenchyma. During this time, there has been an incredible improvement in detailing anatomy secondary to this focus; today, routine clinical computed tomography and MRI2 systems can easily obtain images of specific organs with submillimeter resolution in three dimensions. However, the measured anatomical changes that are used to detect pathophysiology and treatment response reflect very late manifestations of the stepwise molecular alterations within cancer cells.
This understanding has led to a molecular imaging approach (1) that complements the molecular targeting seen in current development of newer anticancer therapies. The field attempts to noninvasively map cellular and subcellular events, especially those targets responsible for pathogenesis that are key points for therapeutic intervention. Thus, relatively crude parameters of tumor growth and development (tumor burden, anatomical location, and so forth) may be supplemented by more specific parameters (e.g., levels and types of secreted proteases, growth factor receptors, cell surface markers, transduction signals, cell cycle proteins, and so forth). Determination of this additional information will allow assessment of therapeutic efficacy at a molecular level, long before phenotypic changes typically occur; a major impact in drug development and testing is expected. The goal of individualized medicine, with therapy chosen based on a person’s gene expression levels, will be partially realized through molecular imaging techniques; as a major example of this approach, optical imaging of protease activity is detailed here.
Overall, there are three general classes of imaging probes that are used to aid in visualization. The most common agents are nonspecific, usually smaller molecules, which have vascular distribution with leakage into the extracellular space. These increase contrast of pathology by differential rates of tissue/tumor perfusion or vascular leakage. The problems associated with this approach include the quite limited tumor:background ratio (making visualization more difficult) and the lack of any specific molecular information. Targeted approaches have been developed to increase the localization of image contrast-enhancing molecules in tumors and to reduce their uptake in normal tissues (2). One potential caveat of using targeted conjugates such as monoclonal antibodies is the fact that target-to-background ratios can be limited by receptor density and/or availability, limited clearance kinetics from the interstitial space, and/or nonspecific cellular uptake or adhesion of certain fluorescent probes. In particular, it may be difficult to differentiate specifically bound from unbound ligands, and this is the reason why imaging is usually performed after nonspecifically distributed excess probe has cleared.
The third general class of imaging contrast agents are smart probes. These agents change their physical properties after specific molecular interaction and are sometimes referred to as molecular beacons (3, 4). In vivo optical (near-infrared) smart probes that have been developed by our group for the detection of protease activity are based on a quenching-dequenching paradigm, as shown in Fig. 1. The probes are optically silent in their native (quenched) state and become highly fluorescent after enzyme-mediated release of fluorochromes, resulting in signal amplification of up to several 100-fold, depending upon the specific design. Of note, nonspecific and targeted agents have no amplification, and newly developed experimental MRI probes have <3-fold amplification (5, 6). Quenching of fluorescence is secondary to fluorescence resonance energy transfer, which occurs because the fluorochromes on the intact probe are spatially near one another. Enzyme specificity is imparted through the use of enzyme cleavage-specific peptide sequences, which can be varied depending upon the desired protease to be visualized. Moreover, other enzymatic pathways are amenable to this activation scheme. This approach has several major advantages over simple targeting: (a) a single enzyme molecule can cleave multiple fluorochromes, resulting in one form of signal amplification; (b) reduction of background signal of several orders of magnitude is possible because the quenched probe is optically silent when injected and remains so until it is activated by its target; and (c) very specific enzyme activities can potentially be interrogated. All of these lead to better visualization of tumors based on their enzyme overexpression profile.
The probes typically consist of three biocompatible building blocks: (a) a delivery vehicle; (b) near-infrared fluorochromes; and (c) enzyme-specific peptide substrates coupling the two. In, general, high-affinity ligands have to be able to reach their intended target at sufficient concentrations and for sufficient lengths of time to be detectable in vivo. Low molecular weight probes are typically subject to fast excretion in vivo, given renal clearance of small molecules and reticuloendothelial system clearance of nonimmunologically shielded compounds. To improve tumoral delivery of the NIRF probes, the delivery vehicles used for all of the probes described in this review are higher molecular weight novel, long-circulating synthetic-protected graft copolymers that have recently been tested in clinical trials (7). The copolymers accumulate in tumors by extravasation through permeable neovasculature, with uptake of the polymer comparable in magnitude to that of tumor-specific internalizing monoclonal antibodies (8).
Many tumors have been shown to have elevated levels of proteolytic enzymes, presumably in adaptation to rapid cell cycling and for secretion to sustain invasion, metastasis formation, and angiogenesis (9–12). Because they are present at high levels in tumors and are elevated at an early stage, proteolytic enzymes represent an attractive target for antitumor imaging and therapeutic strategies (13–16). An additional benefit of targeting proteases for imaging is the relative ease compared with other compartments of probe delivery to the lysosomal and extracellular compartments where protease activity and concentration are highest. Over the last 2 years, our group has prepared a number of protease sensors that are summarized in Table 1. Specific examples, related to tumor aggressiveness and early changes in tumors, including cathepsin B, cathepsin D, and MMP-2 imaging, are discussed in more detail.
Imaging Cathepsin B
Cathepsin B is a lysosomal cysteine protease involved in cellular protein turnover and degradation (17, 18). It is overexpressed in many tumors as well as overexpressed by host cells associated with tumors; its near ubiquity makes it a very attractive target for tumor detection. It has been implicated in tumor progression: both metastases formation (19, 20) and in vitro growth (19–21) decrease in the presence of cathepsin B inhibitors. Several studies (22–25) have demonstrated that high levels of cathepsin B expression correlate with aggressive tumor behavior. High levels of cathepsin B expression also correlate inversely with patient survival (26–29).
Given cathepsin B’s close relationship with early cancers and metastasis with resultant host response, the first optical probes constructed for targeting protease activity were specific for cathepsin B (30, 31). Fig. 2 shows a 2-mm xenograft of a LX-1 human lung carcinoma implanted in a nude mouse. In this typical acquisition, 2 nmol of quenched probe was injected i.v. 24 h before near-infrared imaging was performed. The time for imaging after probe administration may be as short as 2 h, depending upon the location of the enzyme being imaged (32); optimal timing after injection depends as well upon the disease imaged (33, 34). Image acquisition times on newer systems is in the subsecond time frame, allowing real-time visualization of fluorescent anatomical detail. The bright fluorescence, as this example shows, helps locate this micrometastasis-sized tumor, based upon its cathepsin B activity. Other interesting screening and therapy relevant applications of cathepsin B activity are discussed below, with respect to specific diseases.
A probe specific for cathepsin D has also been synthesized and tested in mouse models (35, 36). The main function of cathepsin D is in protein catabolism; however, a 2–50-fold increase in enzyme levels has been reported in breast cancers (37, 38), and overexpression has been associated with higher metastatic potential (39). Imaging of cathepsin D was analogous to cathepsin B imaging in Fig. 2. However, these studies additionally demonstrated that it was possible to selectively image tumors based on a difference of a single gene expression.
Imaging MMP-2 and Its Inhibition
MMPs (40) similar to the cathepsins detailed above, are proteases overexpressed in many cancers. The level of MMP expression has been shown to be related to tumor stage (41) and metastasis (42). Among the subtypes, MMP-2 (gelatinase) has been identified as one of the key MMPs. Numerous clinical studies show a clear correlation between MMP-2 expression and poor outcome of disease (43–46). A number of MMP inhibitors, some of which advanced to Phase III clinical trials, have been developed (47–49). Given its importance in tumor pathology, a MMP-2-selective probe was constructed. Imaging of MMP-2 activity was demonstrated, starting 2 h after probe administration (32, 50). Most importantly, a series of experiments showed in parallel cohorts of mice that fluorescent signal intensity after MMP-2 probe administration significantly decreased in tumors from animals that were pretreated with a MMP-2 inhibitor compared with controls, as visualized in Fig. 3. In this figure, both mice had two identical tumors, known to overexpress MMP-2, implanted in the anterior chest wall. Thus, essentially real-time protease inhibition may be imaged noninvasively, instead of waiting several months to evaluate anatomical response, as is traditionally done in individual patients, as well as in in vivo animal trials of inhibitor therapy. Such imaging has implications for faster, more accurate titration of inhibitor dosing in human drug trials, as well as potentially routine care in the future for individual patients.
In addition to viewing the potential of near-infrared protease imaging from a biological perspective, it is useful to understand possible clinical imaging-use paradigms in the future based on cancers of different types and locations. The role of protease imaging in the future will be several fold, including screening, given the near ubiquitous overexpression in tumors and associated host cells, in situ characterization of lesions, measurement of early response, and evaluation of therapeutic efficacy measured by protease (and other enzymatic) inhibition in individual patients. Some of these future roles are exemplified in specific cancers below. Overall, evaluation and detection of a number of cancers may potentially benefit from in situ optical protease imaging, including prostate, bladder, colon, esophageal, oropharyngeal, breast, bronchial, brain, ocular, musculoskeletal, and cervical, as a partial list.
Colorectal cancer is the second most common cause of cancer death in the United States (51), with ∼140,000 new cases annually and >55,000 deaths (52). It is generally agreed that most colonic cancers develop from adenomatous polyps (53). Identification and removal of colonic adenomatous polyps during endoscopy has been shown to reduce the incidence of colorectal cancer (54). Adenomatous polyps are particularly worrisome if they are large (>1 cm) or multiple, have extensive villous components (55, 56), and/or are highly dysplastic (54). Apart from the size criterion, however, it is currently clinically difficult to ascertain the extent of dysplastic features (57) during colonoscopy in vivo. Moreover, colonoscopy has been shown to miss many lesions, with an overall miss rate of 24% (58), especially if small (<1 cm; Ref. 59).
Protease imaging of the colon will have two roles that are related. The first is screening, with the goal of finding smaller and earlier lesions; the second is in situ characterization of the identified lesions. In a recent study (60), the role of cathepsin B imaging in intestinal adenomas was evaluated. Using the APC min mouse model in which adenomas form secondary to a truncation of the APC protein, intestines of mice were excised and imaged. The mice received no injection, injection of the cathepsin B-activatable probe, or injection of ICG, a fluorescent perfusion agent approved for human use. The cathepsin B probe helped detect lesions as small as 50 μm in diameter, based on an increase in fluorescence of 100–300% compared with background tissue. Fig. 4 shows the comparison of white light images (as would be seen by a gastroenterologist performing a colonoscopy) and a fluorescent view of the same. Many of the smaller lesions were visualized only by cathepsin B imaging and were not seen by standard white light imaging nor imaging using the nonspecific perfusion agent ICG. This may potentially markedly decrease the miss rate of polyps (58) when used for screening colonoscopy. Additional early findings3 suggest that differentiation of the aggressiveness of colonic lesions may also be visualized and characterized by this method.
Beast cancer is the most diagnosed cancer in women in the United States, with ∼190,000 new cases annually and >40,000 deaths (52). Early detection has been shown to save lives (61). Optical protease imaging will have several roles in breast cancer evaluation and screening. In a technically elegant study using the nonspecific perfusion fluorochrome ICG (62), human breast optical fluorescence tomography was performed, as shown in Fig. 5. Although an 8-mm ductal carcinoma was found in this case, the sensitivity and specificity of perfusion imaging may be similar to that of perfusion imaging using other techniques such as MRI; the real strength of optical tomography will be in its combination with molecularly specific probes such as the protease specific probes detailed in this review. The imparted information is very difficult to obtain otherwise, in some cases, even with biopsy. The technical aspects of optical imaging deep in tissue is briefly discussed below.
Another example of an application to breast tissue was reported recently (25). Using the cathepsin B-sensitive probe, clear differences were seen by NIRF imaging between two models of aggressiveness, as visualized in Fig. 6. A well-differentiated human breast cancer (BT20) and highly invasive metastatic human breast cancer (DU4475) were imaged 24 h after i.v. administration of the probe. Both types of breast cancers activated the probe so that both tumors became readily detectable. However, in equisized tumors, there was a statistically significant 1.5-fold higher fluorescence signal in the highly invasive breast cancer compared with the well-differentiated lesion, which correlated with a 1.4-fold higher cathepsin-B protein content in the aggressive tumors as seen by Western blotting. Thus, important information about likelihood of metastases (19, 20, 63) and the rate of patient survival (26–29), which correlate with cathepsin B expression, may be determined noninvasively in individual patients.
This review has focused on the biological targets of protease activated optical probes and the implications for cancer imaging. Excellent reviews on tomographic reconstruction provide detail on the physics and technical aspects of the imaging devices (64, 65). Here, we provide a very brief review of the technology that allows imaging of a number of different cancers. This can be divided into surface/subsurface imaging and deeper imaging. Fig. 7 shows why deeper imaging is possible using near-infrared probes. Light in the visible range is partially absorbed by naturally abundant fluorochromes, including hemoglobin. Photons in the infrared region of the electromagnetic spectrum are partially absorbed by water. The near-infrared region of the electromagnetic spectrum provides a window of opportunity with greater tissue penetration. The fluorochromes reported on in this review typically fluoresce in the 700–800-nm range, wavelengths that allow for tissue penetration on the order of 10–15 cm.
The technically easiest systems are reflectance systems in which light at one wavelength illuminates a surface and fluorescent photons at a second wavelength are recorded (31). Such a system for mouse imaging is shown in Fig. 8. Advantages of such devices include ease of use, rapid data acquisition for screening of many animals, and straightforward data analysis. Even early systems could detect subpicomole amounts of fluorochromes, making optical imaging very attractive in terms of sensitivity. Reflectance systems are not intrinsically quantitative for tissue fluorochrome concentration but can provide somewhat quantitative data when imaging of fluorescence is near a surface and are even more accurate when a second imaging channel is also used. Importantly, such reflectance systems for mice are analogous to fluorescent endoscopy for humans (66). Many epithelial cancers are near surfaces: in addition to the obvious case of examining skin lesions in this fashion, colonoscopy, bronchoscopy, upper gastrointestinal endoscopy, and laparoscopy all provide surface illumination of epithelial tissue at risk for neoplasia. Moreover, near-infrared goggles with appropriate filters are also analogous to the reflectance system and may be used intraoperatively, for example, during neurosurgery, for tumor margin evaluation in real time.
Although it is difficult with many imaging modalities, including MRI, computed tomography, positron emission tomography, and ultrasound, to interrogate more than one parameter simultaneously, optical imaging lends itself to such multiplexing of even molecular information secondary to the ease of separating wavelengths of light. Thus, as was demonstrated recently (67), it is possible to use near-infrared fluorescent smart probes to image multiple gene expressions simultaneously and independently. A second channel allows greater quantitation of concentration based on signal intensity acquired during surface imaging. More importantly, miniarrays of several injected probes for in vivo target assessment may be implemented. For example, a breast lesion may be better characterized with a combination of enzyme activities such as MMP-2 combined with tyrosine kinase evaluation rather than either one alone. It is likely that each group of cancers will have a several targets that would help define the imaging miniarray most suited for characterization of that disease. These multichannel techniques are applicable for both subsurface imaging and imaging of deeper tissue.
Fluorescence-mediated molecular tomography (62, 68, 69) is a technique that intrinsically provides quantitative fluorochrome concentration in deep tissue. Although the coefficient of absorption of near-infrared light is on the order of 1 cm−1 through tissue, scattering occurs much more often, approximately every millimeter. Thus, reconstruction must take into account this high scatter, which truly limits deep imaging without advanced reconstruction techniques. As in the reflectance mode, fluorescence-mediated molecular tomography has a detection limit in the range of picomoles, comparing favorably to other modalities such as positron emission tomography (68). As a simplistic view of the reconstruction, multiple excitation fibers are placed around an object (human breast or whole mouse, for example), and multiple rows of detectors simultaneously record fluorescent signal. This is repeated sequentially for all excitation fibers, and the entire data set is processed using algorithms based upon diffusion-model equations to form images with relatively high resolution (1–3 mm in mice and on the order of 5 mm in human breast). Given the molecular specificity of the probes, this spatial resolution is more than adequate to localize and characterize lesions.
Imaging in the future will not merely report on the success or failure of therapy several months after it has been initiated but will play a crucial role in detecting lesions based upon their molecular signatures, will characterize lesions in situ to aid in treatment decisions, and will help define successful therapeutic drug levels on an individual basis. The near-infrared protease imaging technology reviewed here shows examples of molecularly specific probes and some example clinical-use paradigms for their implementation into practice. Treatment and screening approaches of number of cancers may benefit in the near future from these tools.
|Specificity .||Enzyme substrate .|
|Cathepsin B||K↓K, R↓R|
|Herpes simplex virus protease||LVLA*SSSFGY|
|Interleukin 1 β converting enzyme||GWEHD*G|
|Specificity .||Enzyme substrate .|
|Cathepsin B||K↓K, R↓R|
|Herpes simplex virus protease||LVLA*SSSFGY|
|Interleukin 1 β converting enzyme||GWEHD*G|
The abbreviations used are: MRI, magnetic resonance imaging; MMP, matrix metalloproteinase; APC, adenomatous polyposis coli; ICG, indocyanine green; NIRF, near-infrared fluorescence.
U. M. and R. W., unpublished results.