In Vivo Tomographic Imaging of Near-Infrared Fluorescent Probes

Abstract

Fluorescence imaging is increasingly used to probe protein function and gene expression in live animals. This technology could enhance the study of pathogenesis, drug development, and therapeutic intervention. In this article, we focus on three-dimensional fluorescence observations using fluorescence-mediated molecular tomography (FMT), a novel imaging technique that can resolve molecular function in deep tissues by reconstructing fluorescent probe distributions in vivo. We have compared FMT findings with conventional fluorescence reflectance imaging (FRI) to study protease function in nude mice with subsurface implanted tumors. This validation of FMT with FRI demonstrated the spatial congruence of fluorochrome activation as determined by the two techniques.

Introduction

Fluorescence imaging has recently sparked significant interest because of its ability to assess protein function and gene expression in vivo [1]. Several elegant technologies have evolved to measure the fluorescence emitted from extrinsically administered fluorescent probes or fluorescent proteins expressed by engineered cells. Principally, there are three approaches to imaging fluorescent signals in vivo. The first utilizes microscopic observation of surface or subsurface (0–500 μm) fluorescence using confocal and/or multiphoton microscopy. These techniques have allowed unprecedented insights in in vivo biology at the cellular or subcellular level [2–4]. A similar technology, evanescent wave microscopy, has also been used for fluorescent investigations of cell membranes [5].

The second approach, fluorescence reflectance imaging (FRI), observes the macroscopic surface distribution of fluorescence in live animals. The technique generally uses planar wave illumination to excite fluorochromes at or near the tissue surface, and obtains fluorescence images using photographic methods at the emission wavelength of the targeted fluorochrome. We have previously shown that this technique can be applied to macroscopic imaging of cathepsins [6], [7], matrix metalloproteinases [8], and, potentially, other enzymes [1]. Other investigators have used similar technologies for in vivo receptor mapping [9], [10]. One of the advantages of FRI is that it allows for high animal throughput and simple operation. Conversely, the resolution of the technique is depth dependent and typical penetration depths are limited to 5–8 mm depending on experimental specifics and wavelength. Depth limits are generally not due to tissue attenuation, since light can penetrate tissues for several centimeters, especially in the low tissue absorption near-infrared spectral window, but because of the inability of reflectance imaging to resolve depth. This is because only a single image is obtained that contains information from multiple layers. Moreover, tissue scatters photons significantly in near-infrared and visible wavelengths. Thus photographic methods have difficulty in gauging size and concentration of fluorochromes in deep tissues, even in the absence of surface fluorescence.

The third potential imaging approach, fluorescence-mediated molecular tomography (FMT), aims at volumetrically imaging and three-dimensionally resolving molecular function by reconstructing the distribution of fluorescent molecular probes. The technique uses principles of diffraction tomography [11], [12] and collects light that propagated through tissue (transillumination) at multiple projections. Several investigators have performed tomography of diffuse media in fluorescence mode using simulations or phantoms [13–19]. It has been further hypothesized that tomography of fluorescent molecular probes could prove a powerful modality for imaging tissue function and gene expression [20], [21]. Diffuse optical tomography of tricarbocyanine dye tumoral accumulation in human breast cancers has been applied in vivo [22]. This study imaged the absorption perturbation due to the dye administration and the findings were confirmed with concurrent magnetic resonance imaging (MRI). More recently, tomographic measurements at both the excitation and emission wavelengths demonstrated improved capacity for quantitative in vivo fluorescence imaging of contrast agents and fluorescent molecular probes [15]. Subsequently, the ability to resolve molecular function using FMT of activatable near-infrared probes was confirmed in animal models [23] using a prototype FMT mouse scanner [24].

Herein, we corroborate the FMT capacity to resolve molecular function in animal models with subsurface tumors using correlative MRI, FRI, and Western blotting assays. This is the first report on in vivo validation of FMT with FRI and demonstrates congruence of signals from targeted malignant lesions with established cathepsin-B up-regulation. This validation was necessary since the anatomical appearance of tumors may not coincide with their fluorescence emission pattern due to the diffusive nature of light propagation in tissues and differences in structure and enzymatic biodistribution. The findings of this study suggest that FMT could be a superior method to FRI for in vivo molecular investigations since it attains deeper penetration [25], higher resolution [24], and quantification capacity [15].

Materials and Methods

Experimental Protocol and Animals

Two million HT1080 fibrosarcoma cells were injected subcutaneously in the mammary fat pad or the upper posterior thorax of nude mice. HT1080 cells were obtained from the American Type Culture Collection (ATCC, Manassas, VA) and cultured in MEM medium with 2 mM l-glutamine and Earle's BSS adjusted to contain 1.5 g/L sodium bicarbonate, 0.1 mM nonessential amino acids, 1.0 mM sodium pyruvate, and 10% heat-inactivated fetal bovine serum. After implantation, tumors were allowed to grow for 1–2 weeks. Lesion growth was followed with MRI. One day prior to imaging experiments, animals were anaesthetized by an intraperitoneal injection of ketamine (90 mg/kg) and xylazine (10 mg/kg), and the imaging probe (100 nmol/kg) was injected intravenously.

Animals were anaesthetized and imaged with FRI. Imaging acquisition lasted 1 sec for the intrinsic image (image obtained at the excitation wavelength) and 60 sec for the fluorescence image. Subsequently, animals were imaged by FMT. The animals were suspended into the imaging bore, tail first. The region of interest was placed within the field-of-view (FOV) of the imager. Once the subject was secured, the optical bore was filled with a scattering fluid consisting of an intralipid solution and India Ink in order to match the optical properties of the animal (μ_a = 0.15 cm⁻¹, μ_s′ = 8 cm⁻¹). The fluid covered the whole animal body except the top part of the head to allow free breathing. The purpose of the fluid was to facilitate optimum photon coupling between the sources, animal body, and detectors by minimizing index of refraction and photon wave propagation vector mismatches along the cylinder. In this manner, the measurement obtained is always resolved through a highly diffuse, tissue-like medium. Tomographic imaging was performed at the intrinsic and fluorescence wavelength. Acquisition times per source were 0.1 and 5 sec, respectively. The tomographic measurement lasted approximately 4 min, including the time required for fiber switching and CCD chip read operation. At the end of the imaging experiment, background noise signals were acquired for postprocessing subtraction.

Protease Activated Near-Infrared Fluorescent Probes

We employed an enzyme-sensing activatable (cathepsin-B) near-infrared fluorescent imaging probe reported previously [6]. The probe was synthesized by binding multiple Cy 5.5 fluorochrome residues to a long circulating graft copolymer consisting of a poly-lysine (PL) backbone sterically shielded through methoxypoly-ethylene glycol side chains (MPEG). Unmodified lysines were present in each probe serving as cleavage sites for K–K recognizing proteases including cathepsin-B. The probe is designed to fluoresce only when interacting with the specific molecular target. This is achieved by loading the graft copolymer with multiple fluorochromes in close proximity to each other resulting in energy resonance transfer. Therefore, the fluorochromes are quenched in their native state and do not fluoresce in the presence of excitation light. The fluorochromes are released upon interaction of the probe with the specific molecular target due to enzymatic cleavage of the backbone, resulting in a fluorescence emission.

Fluorescence-Mediated Molecular Tomography

A schematic of the tomographic system employed is given in Figure 1a. The light source (i) was a 670-nm laser diode (B&W TEK, Newark, DE) operating in continuous wave mode with maximum power 200 mW. The light from the laser diode was directed to a 50/50 fused silica coupler (ii) (OZ Optics, Ontario, Canada) that split the signal to the reference branch (iii) and to the optical switch (iv). The optical switch (Dicon FiberOptics, CA) can multiplex one input to any of 32 outputs. In the current implementation, light was switched sequentially to 24 outputs and directed onto the optical bore (vi) using twenty-four 200/230 μm multimode optical fibers (v) (Fiber Instrument Sales, NY). The optical fibers were inserted through the Delrin in two parallel rings of 12 fibers each, and were flushed with the inner Delrin wall. Sequential selection of source fibers illuminated the tissue of investigation at different angles using a full 360° projection scheme. For every source fiber that was used, light was collected in parallel around the boundary using 3 mm diameter optical fiber bundles (Dollan–Jenner Industries, MA) (vii) arranged in three parallel rings of 12 detector fibers around the optical bore, with the collection tip also flushed with the inner optical bore surface. The detector fiber rings were interlaced with the source rings at 0.5-cm intervals. The minimum distance between source–detector pairs was 8.2 mm, therefore within diffusion approximation limits [26]. Figure 1b depicts the optical bore with source fibers and detector fiber bundles attached.

Figure 1.

(a) Schematic diagram of the FMT scanner employed. The scanner uses 24 × 36 source-detector channels, a 200-mW/670-nm laser diode light source, and CCD-based detection (see text for details). (b) Photograph of the optical chamber used for animal investigations. The source fibers (blue) and detection fiber bundles (black) are shown installed. (c) Photograph of the fiber bundle array arranged for imaging by the CCD camera.

The detection fibers were arranged in a 6 × 6 array onto a flat plate (viii) made of Delrin (shown in Figure 1c). The fiber array was imaged with a Vers Array 512B back-illuminated CCD camera (ix) (Roper Scientific, NJ) with ~80% quantum efficiency at the 700-nm spectral range, using an F-mount Micro-Nikkor 60 mm f/2.8D AF lens (Nikon, Japan). Fluorescence detection used a three-cavity band-pass interference filter (×) with center frequency at 710 nm and a full-width half-maximum response of ±5 nm (Andover, NH). System timing and data storage were performed with a personal computer (xi).

Fluorescence reconstruction was based on an algorithm previously developed for tomographic investigations of fluorescent probes [15] using principles of diffracting sources [11]. In this particular implementation, the diffusion equation was solved for the linear perturbative case and the corresponding greens functions were obtained with a finite-difference solution of the diffusion equation [27] for the cylindrical boundary at both the excitation and emission wavelengths. Subsequently, the algorithm inverted an appropriately constructed vector of measurements obtained. The volume reconstructed was a cylinder of diameter 2.5 cm and height 2 cm. A total of 900 1.5 × 1.5 × 5 mm³ voxels were inverted using 2000 iteration steps of the algebraic reconstruction technique with positive restriction [11].

The capacity of the algorithm and system employed to resolve fluorochrome concentration in three dimensions has been shown in the past [15]. Furthermore, it has been shown [24] that the technology can detect fluorochrome quantities in the femtomole range in diffuse media and does not photobleach or saturate fluorochromes at concentrations higher than 800 nM and incident light powers that exceed 100 mW. The resolution of the imager is less 3 mm across the optical bore.

FRI System

A previously described surface-weighted fluorescence imager was used for correlative reflectance imaging [28]. Briefly, the imager used a 150-W halogen white light source and an excitation band-pass filter (610-650 nm, Omega Optical, Brattleboro, VT). An optical fiber directed light from the source into a light-tight chamber where the animals were placed. Virtually homogeneous illumination was achieved over the FOV using light diffusers. Fluorescence images were acquired using a 12-bit monochrome CCD camera (Kodak, Rochester, NY) equipped with an f/1.2 12.5–75 mm zoom lens and an emission long-pass filter at 700 nm (Omega Optical). Images were acquired over 30 sec using commercially available software (Kodak Digital Science 1D software, Rochester, NY). Exposure time was manually adjusted to optimize signal-to-noise ratio. Intrinsic light images (images at the excitation wavelength) were also acquired by removing the long-pass filter from the CCD detector.

Magnetic Resonance Imaging

MRI was performed with a 1.5-T superconducting magnet (Signa 5.0, GE Medical Systems, Milwaukee, WI) using a custom-built 1.5-in. surface coil, an FOV of 6 cm and a 512 × 512 matrix. The imaging protocol consisted of a T1-weighted axial 3-D fast spoiled gradient recalled sequence (3D-FSPGR) with a TR/TE/FA of 8/2.3/60°. Imaging time was approximately 7 min.

Western Blotting

Western blotting was carried out in order to measure cathepsin-B expression in tumors and adjacent muscle tissue. Tissues were homogenized in a 50-mM Tris buffer (pH 7.5) containing 0.25% Triton X-100 and complete protease inhibitor (Boehringer-Mannheim). Minced tissue was then repeatedly sonicated and centrifuged at 10,000 rpm for 10 min. The protein content of tumor homogenates was measured with a BCA assay (Pierce, Rockford, IL). Equal amounts of protein were applied to a 12% SDS-PAGE gel. Following separation, proteins were blotted onto a PVDF membrane (Bio-Rad, Hercules, CA). After blocking, immunoblots were incubated with a primary anti-cathepsin-B polyclonal antibody (Santa Cruz Biotechnology, Santa Cruz, CA) followed by incubation with a peroxidase labeled secondary antibody (Vector Laboratories, Burlingame, CA). Binding of the antibodies was revealed by diaminobenzidine serving as a peroxidase substrate (Vecstastain, Vector Laboratories, Burlingame, CA). Blots were scanned and evaluated digitally using commercially available software (Kodak Digital Science 1D software).

Results

Figure 2 demonstrates imaging results obtained from a mammary tumor that measured 3 mm diameter by 1.5 mm thickness on MRI imaging and was located 1 mm below the skin surface. The axial MR image passing through the tumor is shown in Figure 2e. Figure 2a and b demonstrates the FRI images at the intrinsic and fluorescence wavelengths, respectively. The fluorescence image reveals a marked fluorochrome activation congruent with the location of the lesion. Figure 2c shows a merged image that superimposes the fluorescence image above a threshold on the intrinsic light image. The fluorescence pattern appears significantly larger than the appearance and size of the tumor seen on MRI. Figure 2d depicts an axial slice passing through the tumor, its position indicated with the arrow. Such discrepancies between anatomical size and fluorescence appearance are typical in reflectance systems due to light diffusion; however, differences between anatomy and cathepsin-B biodistribution are probably also present in this particular case, based on previous experience with reflectance imaging [6], [8], [29].

Figure 2.

Imaging of an HT1080 tumor implanted in the mammary fat pad of a nude mouse. (a) Intrinsic light image obtained with reflectance imaging. (b) Fluorescence image obtained with reflectance imaging. (c) Superposition of (b) onto (a) after threshold application. (d) MRI axial slice at the level of the uppermost FMT slice acquired, with an arrow denoting the tumor location. (e) MRI sagittal slice with lines denoting the three-slice volume segmentation assumed by FMT. An arrow again denotes the tumor location. (f,g,h) Reconstructed FMT slices. The outline of the axial MRI slice is shown superimposed in (f) for registration purposes. (i) Volume rendering of the FMT dataset (red). Blue spheres denote the locations of the most superior ring of detectors.

Figure 2e indicates the positioning of the animal relative to the FMT scanner FOV. The image shows the three slices imaged by FMT superimposed on a sagittal MR slice obtained from the same animal prior to the FMT experiment. Figure 2f, g, and h depicts the three consecutive slices indicated on Figure 2e, top to bottom. The top slice (Figure 2f) is shown superimposed on the outline of Figure 2d for registration purposes. The registration of MR and FMT images is based on fiducial and anatomical markers since the MR and FMT acquisition were not obtained simultaneously. The FMT images depict high fluorescence activity congruent with the three-dimensional position of the tumor shown on the MR images. This activity also correlated with the FRI findings. The two subsequent FMT slices show minimal fluorescence activity compared with the top slice since normal tissues such as the lung or the heart do not over express cathepsin-B, as also evident from the reflectance images. Finally, Figure 2i depicts a three-dimensional rendering of the three FMT slices, interpolated along the slice dimension 15 times. The animal outline is superimposed in the middle of the FMT volume and is also rendered to enhance the visual effect. Rendering was performed using the IDL software package (Research Systems, Boulder, CO). The dots indicate the position of the top detector ring used in the acquisition, which also marks the upper end of the volume reconstructed.

Figure 3 summarizes imaging findings from another animal with a tumor implanted in the posterior upper thorax. The implantation site was deliberately deeper for the upper thorax tumors compared to the mammary pad tumors. Figure 3a and b demonstrates the FRI images at the intrinsic and fluorescence wavelengths, respectively. Figure 3c shows a merged image that superimposes the fluorescence image above a threshold on the intrinsic light image. Cathepsin-B expression is clearly evident on the fluorescent images, and the location of activation was congruent with the implantation site. The fluorescence image demonstrates a diffuse pattern, mainly due to the deep implantation site. This tumor, verified by dissection and histology, was not identifiable on the MR images (Figure 3d depicts an axial slice obtained from a slice passing through the tumor implantation site, and Figure 3e depicts a sagittal image for demonstration purposes). Western blotting confirmed cathepsin-B expression in tumors to be much higher than in muscle (Figure 4). The slices imaged by FMT are indicated on Figure 3e by horizontal slices. Three FMT consecutive slices are shown in Figure 3f, g, and h. Superimposed on Figure 3f is the outline of the axial MR slice shown in Figure 3d. The registration of MR and FMT images is again based on fiducial and anatomical markers since the MR and FMT acquisition were not performed simultaneously. The top FMT slice resolves a focus of fluorescence activation congruent with the location of implantation and the reflectance imaging findings. The middle FMT slice demonstrates very little fluorescence activity, whereas the third slice depicts minor fluorescence activation. Finally, Figure 3i depicts a three-dimensional rendering of the three FMT slices, interpolated along the slice dimension 15 times and superimposed on the intrinsic reflectance image. The top cylinder of blue dots indicates the top detector ring used in the acquisition, which also marks the upper end of the volume reconstructed.

Figure 3.

Imaging of an HT1080 tumor implanted in the posterior upper thorax of a nude mouse. (a) Intrinsic light image obtained with reflectance imaging. (b) Fluorescence image obtained with reflectance imaging. (c) Superposition of (b) onto (a) after applying threshold. (d) MRI axial slice at the level of the uppermost FMT slice acquired. (e) MRI sagittal slice with lines denoting the location of acquired FMT slices. (f,g,h) Reconstructed FMT slices. (i) Volume rendering of the FMT dataset (red). Blue spheres denote the locations of the most superior ring of detectors.

Figure 4.

Western blot analysis of tumor and muscle tissue, compared to a standard. The elevated expression of cathepsin-B in tumor relative to muscle is apparent.

Discussion

The significance of fluorescence activation, and its implications to molecular medicine, have been outlined in the past [1]. Despite advances in medical imaging technologies over the last two decades, we are currently still limited in our ability to detect tumors or other diseases in their earliest stage of formation and to phenotype tumors during complex cycles of growth, invasion, and metastasis. Similar limitations also exist in neurodegenerative, cardiovascular, and immunologic diseases. Advances in these endeavors could allow the use of imaging techniques as objective therapeutic endpoints, which could in turn accelerate the drug testing process and improve the clinical management of patients with these pathologies. FMT has the potential to make significant contributions in this research. Key advances in FMT development are: (1) the engineering of “smart” targeted fluorescent molecular probes for an increasing number of targets identified by gene array profiles; (2) the miniaturization and development of highly sensitive imaging equipment that allows recording of fluorescent signals at pico- to femtomole accumulations; and (3) the development of appropriate mathematical models that allow tomographic fluorescence imaging in in vivo systems. FMT optical technology is relatively inexpensive, can be made portable, and uses nonionizing radiation and stable molecular markers. These features can allow easy laboratory and bench top use and enable monitoring of molecular events repeatedly and over time.

Fluorescence reconstructions in dense media have been validated in the past with simulations and phantom measurements [13–16,19,30]. Herein, we have demonstrated in vivo FMT findings from imaging subsurface molecular events and confirmed these findings with FRI. In vivo validation of the tomographic scanner is difficult with conventional imaging modalities due to the fundamental differences in targeted contrast. Generally, the MRI appearance of tumors could be significantly different from the fluorescence activation pattern; therefore, the cross-validation of FMT with FRI of the same animal was necessary. For the purposes of this validation, our investigations were limited to depths of a few millimeters since reflectance imaging cannot sufficiently resolve fluorescent activation deeper within tissue. Imaging of deeper tumors has been also recently confirmed [23].

The in vivo application of FMT in this work has been enabled by the development of accurate algorithms that do not require background calibration measurements from phantom experiments but utilize only data obtained from the animal measurement [15]. The performance of these algorithms is enhanced by the use of activatable fluorescent probes that minimize nonspecific background fluorescence [6]. There is strong evidence that these advancements can propagate to human applications. In vivo investigations of human tissues using near-infrared photons in the absorption/scattering mode have demonstrated the capacity of diffuse light to localize and quantify function in deep tissues. Diffuse optical tomography of the vascular distribution of extrinsically administered chromophores and of intrinsic contrast due to hemoglobin concentration deep in the human breast has shown the capacity to image deep in human tissues [22], [31]. Moreover, tomography of brain hemodynamics in humans has also been reported [32], [33]. While there is no reported in vivo imaging of fluorescence in humans, in a related study we performed recently, we have demonstrated that fluorescence signals from tumor-like lesions could be detected in the human breast and lung at diameters that exceed 15 cm and in brain or muscle at distances larger than 6 cm [25]. Therefore, besides animal applications, FMT could find clinical applications as well.

For the reported experimental studies, we utilized a previously developed probe that is activated by cathepsin-B [6]. More generally, utilizing probes specific to other biomarkers could yield an imaging modality with specificity to different molecular events. The combination of molecular beacons and optical imaging technologies has the potential to play a fundamental role in biomedicine in the next decade, and continued developments in probe design and imaging technologies will undoubtedly further expand current capabilities.