Combined in Vivo Bioluminescence and Fluorescence Imaging for Cancer Gene Therapy

Abstract

Nasopharyngeal carcinoma is intimately associated with the Epstein-Barr virus (EBV), which we have exploited therapeutically by constructing an EBV-specific synthetic enhancer sequence, within an adenoviral vector, denoted as adv.oriP. The achievement of tumor targeting provides therapeutic potential when delivered systemically, which could impact on distant metastases. We demonstrate here the feasibility and potential utility of combined, minimally invasive in vivo bioluminescence and fluorescence imaging to monitor adenoviral infection of subcutaneous C666-1 nasopharyngeal xenograft tumors stably expressing the DsRed2 gene. Fluorescence imaging was used to monitor the location and size of the C666-1.DsRed2 tumors, whereas bioluminescence imaging demonstrated the distribution and specificity of a transcriptionally targeted adenoviral vector, adv.oriP.fluc, expressing the firefly luciferase gene. Fluorescence, bioluminescence, and photographic images were aligned using grids to examine co-localization of adenovirus and tumors. Bioluminescence and fluorescence co-localized in 92% (11/12) of tumors at 24 hr and 100% (12/12) at 96 hr after adv.oriP.fluc (10⁹ ifu) was administered intravenously. Nonspecific luciferase signal was detected in the liver area. The combined imaging was therefore successful in monitoring the uptake of systemically administered adenovirus in implanted tumors. This may ultimately lead to an effective noninvasive method to monitor the response of metastases to adenoviral gene therapy.

Introduction

Nasopharyngeal carcinoma (NPC) is an epithelial neoplasm of the head and neck region with unique environmental and geographical epidemiologic features [1]. A key element is its strong association with the Epstein-Barr virus (EBV); nearly all cases of NPC harbor the virus in a latent form within each cancer cell [1]. We have previously exploited the presence of EBV for therapeutic advantage using a transcriptional targeting strategy [2]. This was achieved by the construction of a synthetic regulatory sequence consisting of the EBV FR enhancer elements in conjunction with a basal CMV promoter, denoted as oriP [2]. This successfully achieved EBV-specific NPC targeted toxicity, both in vitro and in vivo, within the context of an adenoviral vector system [2–4].

Given that a significant proportion of NPC patients will die from systemic disease, we need to develop viral therapies which will treat distant metastases [5]. With the emergence of molecular imaging modalities, it is now possible to ascertain, rapidly and noninvasively, the kinetics and biodistribution of adenoviral vectors, when carrying the appropriate reporter genes.

The aim of this study was to demonstrate the combination of two molecular imaging modalities, bioluminescence and fluorescence, to visualize the biodistribution and tumor specificity of an oriP-driven, firefly luciferase-expressing adenovirus (adv.oriP.fluc), injected systemically into tumor-bearing SCID mice. This involves the establishment of NPC tumors that stably express the DsRed2 fluorescent protein, followed by the systemic injection of adv.oriP.fluc. Tumor location is spatially determined by fluorescence imaging (FI); bioluminescence imaging (BLI) can evaluate the biodistribution of adv.oriP.fluc in the same animals. By examining co-localization between viral expression and tumor presence at various subcutaneous sites throughout the mouse using tandem BLI and FI, insight into the utility of this technique in monitoring metastases can be obtained.

Materials and Methods

All animal studies were carried out with the approval of the Animal Care Committee, University Health Network, Toronto.

Cell Lines

DsRed2-expressing C666-1 cells were generated using Lipofectamine 2000 (60 μL; Invitrogen, Carlsbad, CA) transfection with pDsRed2 (24 μg; BD Biosciences, Franklin Lakes, NJ) and subsequent selection in G418 (300 μg/mL; Gibco, Grand Island, NY). Fluorescent colonies were selected using a digital fluorescence stereomicroscope (MZFLIII; Leica Microsystems, Bannockburn, IL) and the single strongest DsRed2-expressing colony was cultured, expanded, and used in all subsequent experiments. Stable C666-1-DsRed2 cells were maintained in RPMI1640 media with 10% fetal bovine serum, in a 37°C, 5% CO₂ incubator.

C666-1 Xenograft Tumors

Similar to previous studies [2,3], 10⁷ C666-1-DsRed2 cells in 100 μL PBS were injected subcutaneously through a 28-gauge needle in 6- to 8-week-old BALB/c SCID mice (n = 3; Animal Research Colony, Ontario Cancer Institute, Toronto, Canada) in four different locations, to simulate multiple metastatic lesions. One mouse had dorsally engrafted tumors whereas the other two had them placed ventrally. Experiments were conducted 3 days after injection, yielding solid tumors ~5 mm in diameter.

Adenoviral Vector and Infection

Detailed methods to construct the adv.oriP.fluc have been described previously [2]. Adv.oriP.fluc (10⁹ ifu) in 100 μL PBS was injected via a 25 5/8-gauge needle into the tail-vein of each mouse.

In Vivo Bioluminescence Imaging

One day prior to imaging, the mouse fur was sheared and chemically removed using Nair (Church and Dwight, USA). At 24 and 96 hr after adv.oriP.fluc injection, the animals were anaesthetized and then injected with 100 mg/kg d-luciferin (Promega, Fitchburg, WI), intraperitoneally. To protect these immunocompromised animals from the external environment, they were restrained within clear Lucite cylinders, and then placed within the BLI system (IVIS 1.0: Xenogen, Alameda, CA). Bioluminescence and photographic images were taken 15 min after luciferin injection using a 20-cm field of view, 4 pixel binning, 600–900 sec bioluminescence exposure time (to maximize signal), and 0.2 sec photographic exposure time. Elliptical regions of interest (ROIs) were drawn and counts integrated around the tumors and adjacent non-involved regions (as background) using Living Image 2.0 software (Xenogen).

In Vivo Fluorescence Imaging

Following BLI, the animals were transported to the stereomicroscope. Green light (Leica Filter HQ560/55, λ = 540–580 nm, 0.1 kW mercury lamp) was used as the excitation source, and red fluorescence was collected (Leica Filter HQ645/75, λ = 600–680 nm) using a 1 × objective lens and 10.4 sec exposure time to maximize saturation-free signal. Following FI acquisition, a photographic image was obtained (250-350 msec exposure). Two fields of view were required to image the entire animal. Elliptical ROIs of approximately the same dimensions as those used for BLI were drawn and integrated around the tumors using Image-J software (NIH, http://rsb.info.nih.gov/ij/).

Image Processing

Photographic and bioluminescence TIFF files, saved with Living Image 2.0, were imported into Image Pro Express (Media Cybernetics, Silver Springs, MD). The noise was reduced with a median filter which replaces each pixel with the median value of the neighboring pixels.

Image Alignment

Photographic images of a 2.5 × 2.5 cm grid were obtained using the Xenogen and Leica systems to measure differences in magnification and spherical distortion due to the different camera-mouse distances. A reverse spherical projection transformation, followed by size scaling, was applied to the IVIS-image grid, using Photoshop 7.0 (Adobe, San Jose, CA), for dimensional and curvature alignment with the grid obtained with the Leica system. All BLIs were subsequently transformed using identical settings. Adjustments to correct small shifts in the mouse position during transfer between the imaging systems were applied by aligning the corresponding photographic images. The bioluminescence, fluorescence, and photographic images were superimposed, using Image-Pro Express and Photoshop 7.0.

Results

Figure 1 illustrates the alignment procedure for the three images on a mouse bearing dorsal tumors 96 hr after adv.oriP.fluc injection; the BLI (Figure 1A) was filtered for noise, aligned with the FI (Figure 1B) using the superimposed grids, then overlaid onto the photographic image (Figure 1C). Figure 2 shows the combined images for the three mice, with the bioluminescence and fluorescence shown in different false colors. Each of the 12 DsRed2-expressing xenografts, at various locations on each mouse, was successfully visualized with the fluorescent stereomicroscope prior to adenoviral infection. There were no false-positive fluorescent loci. Although autofluorescence was present from the remaining fur on the head and extremities, its intensity was 20-50% of tumor fluorescence.

Figure 1.

A mouse with four dorsal xenografts 96 hr after adv.oriP.fluc injection, illustrating the use of grids to align the bioluminescence images onto fluorescence images. (A) Bioluminescence image, prenoise filtering. (B) Fluorescence image. (C) Bioluminescence, fluorescence, and photographic image composite.

Bioluminescence co-localized with fluorescence signal in 11 of the 12 tumors (92%; 4 tumors per mouse) at 24 hr postinfection and in all tumors at 96 hr. Seven xenografts were not obvious visually using the BLI detection, but in all cases the light generated exceeded the background by at least 2 standard deviations. Specific co-localization, in which the bioluminescence signal was unambiguously tumor in origin, was observed in 6/12 (50%) at 24 hr and 8/12 at 96 hr (67%). In the remaining tumors, specific co-localization was difficult to assess due to bioluminescence signals emitted from the liver and other internal organs, making some xenografts difficult to distinguish. Dorsal images were unaffected by this at 96 hr due to overlying tissue attenuation of the light from the nonspecific tissues.

No significant correlation was observed between the magnitudes of the fluorescence and bioluminescence signals at 24 and 96 hr (R² = .32 and .25, respectively). BLI values were significantly more variable (16,616 ± 10,198 bioluminescence photons/sec), when compared to fluorescence ROI values (3791 ± 898 RLU/sec), possibly due to the influence of uncontrolled variables, such as site-specific and/or tumor-specific differences in the vasculature and induction depth of the different individual tumors.

Discussion

These results presented here are the first documentation of combined BLI and FI of adenoviral cancer gene therapy. The data are based on two different time points postinjection, representing the earliest time at which a robust luciferase signal could be detected (24 hr), and the time of maximum bioluminescence signal in infected tumor tissues (96 hr) (unpublished data, January 2004).

Figure 2.

Superimposition of virally expressed bioluminescence (green) and tumor-expressed DsRed2 fluorescence (red) images in the three mice each bearing four subcutaneous tumors at (A) 24 hr and (B) 96 hr after systemic infection with 10⁹ ifu of adv.oriP.fluc. Specific co-localization (denoted with white asterisks) can be observed in 6 of 12 xenografts at 24 hr and in 8 of 12 tumors at 96 hr. The location of one tumor is not obvious in the composite at 24 hr, and is therefore demarcated with a white arrow.

At 24 hr, 1 of the 12 xenografts did not have detectable bioluminescence above background. This was most likely due to insufficient transgene expression at that time, as the same tumor was detectable at 96 hr. d-Luciferin biodistribution variability and lower fluorescence signal from that particular tumor (suggesting that depth-related signal attenuation may have been a contributing factor) are two possible explanations for this observation [6]. Specificity also improved as a function of time due to the reduction of nonspecific bioluminescence signals. At 96 hr, 4 of the 12 BLI-positive signals could not be distinguished from the nonspecific signal, suggesting that metastases in the liver area would be difficult to resolve without either tropism modification of the adenoviral vector or further modification of the oriP region to enhance tumor versus nontumor specificity.

This study has a few limitations that may reduce sensitivity and specificity. Both FI and BLI suffer from depth- and wavelength-dependent scattering and absorption of light [7]. This was the main rationale for choosing red rather than green fluorescent protein for tumor localization, because its longer excitation and emission wavelengths provide greater light penetration in tissues [8,9]. Luciferase constructs with longer wavelength emission are under development, which should improve the effective depth in BLI [10].

In conclusion, we have demonstrated, for the first time, the use of two noninvasive in vivo optical imaging modalities to assess the distribution of a systemically administered EBV-positive tumor-specific vector in solid tumors. This combined technique can be a valuable tool for the noninvasive longitudinal monitoring of tumor response to cancer gene therapy.