Noninvasive near-infrared fluorescent protein-based imaging of tumor progression and metastases in deep organs and intraosseous tissues

Abstract. Whole-body imaging of experimental tumor growth is more feasible within the near-infrared (NIR) optical window because of the highest transparency of mammalian tissues within this wavelength spectrum, mainly due to improved tissue penetration and lower autofluorescence. We took advantage from the recently cloned infrared fluorescent protein (iRFP) together with a human immunodeficiency virus (HIV)-based lentiviral vector to produce virally transduced tumor cells that permanently express this protein. We then noninvasively explored metastatic spread as well as primary tumor growth in deep organs and behind bone barriers. Intrabone tumor growth was investigated through intracranial and intratibial injections of glioblastoma and osteosarcoma cells, respectively, and metastasis was assessed by tail vein injection of melanoma cells. We found that the emitted fluorescence is captured as sharp images regardless of the organ or tissue considered. Furthermore, by overlaying fluorescence spots with the white light, it was possible to afford whole-body images yet never observed before. This approach allowed us to continuously monitor the growth and dissemination of tumor cells with a small number of animals, minimal animal handling, and without the need for any additive. This iRFP-based system provides high-resolution readouts of tumorigenesis that should greatly facilitate preclinical trials with anticancer therapeutic molecules.


Introduction
Animal models allow us to recapitulate many of human pathologies including cancer and, although they do not perfectly mimic all aspects of cancer development, they greatly contributed to our understanding of many tumorigenesis mechanisms. 1 Initial studies consisted of implantation of human cancer cells or tumor biopsies into nude mice and evaluation of tumor progression by palpation and/or measurement of tumor sizes by devices such as a caliper. 2 However, these manipulations could not be applied to study deep tumors or metastatic deposits, unless the animal was dissected to inspect for tumor lesions in several tissues and organs, therefore rendering these approaches laborious and animal consuming. 2 Hence, there was a need to find out alternative methodologies that would ease early tumor detection and evaluation of cancer-linked events such as neoangiogenesis, invasion and metastatic dissemination. 3 Accordingly, interest has increased within the past decade in sensitive noninvasive techniques such as fluorescence-based imaging with proteins of the enhanced green fluorescent protein (EGFP) family, 4 the Discosoma sp. DsRed or their shifted wavelength spectra variants 5 as well as luciferase-based bioluminescence. 3 Yet, for bioluminescence to happen, it is obligatory to inject the substrates luciferin or coelentarazin prior to imaging in order to elicit luciferase reaction. 6 Fluorescence imaging of deep tumors using EGFP, DsRed or their mutants is hampered by the limited light penetration in the body, scattering, and tissue absorption. 7 Another problem encountered with EGFP/DsRedbased imaging is autofluorescence which mostly occurs in the green part of the spectrum. 8 Recently, a series of very bright red-shifted fluorescent proteins, such as mKate, 9 E2-Crimson and mNeptune, 7 with an average excitation/emission in the range of 600∕650 nm, have been developed with significantly increased tissue penetration and reduced autofluorescence.
With regard to these limitations, probes with the wavelengths in the near-infrared (NIR) to infrared (IR) ranges (i.e., 650 to 900 nm), would be superior to others for deep-tissue imaging. Recently, two NIR fluorescent proteins, IFP1.4 (Ref. 10) and iRFP (Ref. 11), have emerged with fluorescence characteristics laying within an NIR transparency of mammalian tissues. 12 The first described in 2009 was the bacteriophytochrome IFP1.4 followed, 2 years later by iRFP, cloned from the Rhodopseudomonas palustris and engineered through a series of mutations. 11 Although fluorescence spectra of the two molecules are quite similar, IFP1.4 must incorporate its chromophore biliverdin (which needs to be exogenously supplied) to further emit fluorescence, whereas the physiological level of biliverdin 13 is sufficient to fully activate iRFP. This property, added to the high brightness, prompted us to use iRFP as a tool to track primary tumors and metastasis in deep tissues and behind natural barriers such as bones.
We, herein, used in vivo NIR optical imaging together with a human immunodeficiency virus (HIV)-based lentiviral vector to deliver the iRFP gene into different tumor cell models including glioblastoma, osteosarcoma, and melanoma cells. We then used these cells to explore intracranial and intratibial primary tumor growth as well as metastatic dissemination of melanoma cells.

Vector Construction
The HIV-derived recombinant lentivirus vector expressing iRFP was made by replacing the deoxyribose nucleic acid (DNA) of EGFP in the lentiviral vector pRRLSIN.cPPT.PGK-GFP.WPRE (referred to as pRRL, kindly provided by Dr. Trono, University of Geneva, Switzerland) by the iRFP DNA from the pShuttle-CMV-iRFP plasmid (a gift of Dr. Verkhusha, Addgene plasmid # 31856). To this end, iRFP DNA was excised with BglII and SalI, gel purified and ligated into BamHI/SalI-digested pRRL resulting in pRRL/iRFP construct [ Fig. 1(a)].

Cells and Lentiviral Transduction
The human embryonic kidney cells HEK293T, the glioblastoma U87 MG and the melanoma IGR37 cells were cultured in Dulbecco's modified Eagle medium (Cambrex, Verviers, Belgium). The osteosarcoma U2OS cells were cultured in Roswell Park Memorial Institute (RPMI) medium 1640 (Cambrex). All media were supplemented with 10% fetal calf serum and antibiotics penicillin and streptomycin (Eurobio, Les Ullis, France). Lentiviral particles were produced according to Mathieu et al. 14

Animals and Tumor Implantations
Four-to six-week old male athymic Naval Medical Research Institute nude mice (Harlan, Gannat, France) were maintained in a barrier facility on high efficiency particules arresting (HEPA)-filtered racks and fed with autoclaved alfalfa-free rodent diet to avoid digestive tract autofluorescence. 15 Animal experiments were conducted in accordance with the European Directive EEC/2010/63, under the supervision of authorized investigators. Subcutaneous (s.c) and intracranial (i.c) injections of U87 MG glioblastoma cells were performed as described earlier 16,17 using 1 × 10 5 cells (s.c) or 5 × 10 5 cells (i.c), respectively. For the experimental metastases, we injected 1.5 × 10 6 melanoma cells IGR37 through the tail veins. For intratibial xenografts, mice were anesthetized and 1 × 10 5 U2OS cells were injected into the tibia with a needle introduced at the top of the tibial metaphysics. Immunohistochemistry of melanoma tumors was performed according to Ref. 18, using S100, HBM-45, and Melan-A as markers.

In Vivo Imaging and Analyses
Unless otherwise indicated, animals were subjected to NIR imaging every 3 to 4 days with a Pearl ® Impulse Small Animal Imaging System Li-Cor ® Biosciences (LI-COR Biosciences GmbH, Bad Homburg, Germany). Excitation and emission wavelengths were fixed at 690 and 710 nm, respectively. The fluorescence signal was overlaid with the white channel light-emitting diode [(LED) light] and detected by a cooled CCD camera of the imager. During the in vivo imaging process, mice were kept on the imaging bed by isoflurane inhalation (Abbott Laboratories, Rungis, France) and warmed at 37°C. In some experiments, organs or tumor deposits were retrieved with the minimum surrounding tissue from mice sacrificed by cervical dislocation and subjected to ex vivo imaging. Excised tissues were then placed on the stage and their fluorescence captured as indicated above. Image analysis was performed according to Nakayama et al. 19 In brief, regions of interest (ROIs) were targeted on iRFP fluorescent sites and circles with the minimum size were drawn around lesions on each image. The pixel density in the selected ROI was quantitated using the software for image analysis provided with the Pearl ® Imager. The fluorescence signal on the plots was normalized to the background and presented as arbitrary units (a.u.) according to Filonov et al. 11

Statistical Analyses
Data were collected from the three independent experiments. Animal implantation experiments were performed with 3 to 4 mice per each group of glioblastoma, osteosarcoma, or melanoma tumor models and results were expressed as a standard error of the mean (SEM).

Visualization of iRFP-Expressing Tumor Xenografts in Mice
Tumor cells expressing iRFP were produced as described and analyzed for the NIR fluorescence before inoculation to animals [ Fig. 1(b)]. As shown in Fig. 1(c), fluorescence of iRFP-expressing U87 MG glioma cells, 20 not only clearly distinguishes the tumor from the host but also allows a real-time monitoring of subcutaneous tumor growth, without any additive [ Fig. 1(d)]. By subcutaneously inoculating increasing numbers of iRFPlabeled cells, we found that the minimal number of detectable cells by our NIR imaging device is 50,000 cells (data not shown). Filonov et al. advocated the use of iRFP for imaging of deep organs such as the liver. 11 We, therefore, wanted to ask whether iRFP would uncover intraosseous tumors using intracranial and intratibial inoculated iRFP-expressing tumor cells. As shown in Fig. 2 Intratibial growth of U2OS cells is shown in Fig. 2(B). Note that the presence of additional NIR fluorescent spots on the abdomen of mice between days 43 and 51, which could be due to traces of alfalfa in this animal food; 15 a problem that we subsequently solved. To further analyze the intrabone tumor progression of iRFP-expressing U2OS cells, two animals from a parallel series and corresponding to those of days 43 and 70 were sacrificed and their tibias retrieved. As shown in Fig. 2(Ba), the tumor that initially localized to the head of the bone at day 43, progressively invaded the rest of the organ a week later (at day 51), eventually leading to fracture of the bone and spreading over surrounding soft tissue at day 70 [ Fig. 2(Bb)]. Therefore, the iRFP-emitted fluorescence crosses the bone barrier, thus allowing us to reliably estimate tumor burden without sacrificing the animal [see also Fig. 2(Bc)].

Visualization of Metastasis
These data being acquired, we then asked whether iRFP-based imaging could be applied to the study of metastatic spread. For this purpose, we used melanoma cells IGR37, originating from a metastatic spread of the primary human melanoma cells IGR39. 21 As shown in Fig. 3(A), injection of iRFPexpressing IGR37 cells in the tail vein resulted in a tumor deposit in the liver [see m1 series and after dissection of the animal Fig. 3(Aa)], consistent with a previous work showing the liver to be a potential niche for melanoma metastasis. 22 Surprisingly, injection of the same cells to two other mice led to tumors that appeared, not in the liver, but in the space between omoplates [ Fig. 3(B), see m2 series]. Dissection of the animal revealed the presence of well-defined tumors that strongly adhered to the cervical vertebrae (data not shown) but which were positive for melanoma immunological markers such as S100, HBM-45, and Melan-A [ Fig. 3(C)]. These data show that whole-body NIR imaging with iRFP uncovers different metastatic targets for melanomas.

Discussion
Considerable progress has been made in recent years to cure cancer. Besides surgery, drug-based therapies are more and more efficient and already allow us to save lives or at least to meaningfully extend patient survival. 23 Yet, it is still highly desirable to afford new molecules with an improved bioactivity and decreased toxicity. In order to meet this need, preclinical assays involving small animals and genetic tools for whole-body imaging and tumor-emitted signal quantification would significantly improve anticancer therapeutic strategies. Several previous studies have used fluorescent proteins to track tumor lesions in mice, 24 including those related to EGFP, DsRed, or their variants. 6 However, the wavelengths of these tracers are in a range of low signal-to-noise background and low coefficient of penetration in the animal body, 7 particularly within the green part of the spectrum. 8,25 Autofluorescence, scattering, and absorbance are crucial parameters, particularly when imaging organs that lie millimeters deep under the skin layers. 7 With the NIR light, however, these parameters become less important. 25 Another problem encountered with many previous studies is the use of cells that have been transfected and selected with drugs such as neomycin. 4,24 In addition to the possible clone-to-clone variability, the main challenge with such approaches is whether the transgene expression would be kept stable throughout the study. In fact, once the drug-selected cells are deposited in the animal body, they must grow in a drugfree environment and therefore, might lose the gene of interest due to the lack of selection pressure.
To overcome these difficulties, we herein used a lentivirusmediated delivery system to permanently express the NIR fluorescent protein iRFP in cells. Another attractive characteristic of Fig. 3 Melanoma-derived metastatic lesions were formed after tail vein injection of IGR37 cells to three series of mice. One series developed a liver metastasis (A, m1 series) and the others developed a dorsal tumor (B, only the m2 series is shown). The color bar indicates the total fluorescence radiant efficiency. (C) The tumor from the mouse m2 was retrieved and subjected to hematoxylin/eosin (HE) and immunohistochemistry using S100, HBM-45 and Melan-A markers (magnification 50×).
this system is that this human immunodeficiency virus type 1 (HIV-1)-based lentiviral vector can infect virtually all kinds of cells including the nondividing ones. 26 Using this genetically encoded probe and an NIR fluorescence imaging device, we are herein reporting the usefulness of this system for the detection of deeply buried tumor lesions or those hidden behind the bone barrier. Most importantly, metastatic deposits can be easily detected and quantified [see Figs. 3(A) and 3(B)], thereby avoiding cumbersome and tedious dissection steps. Although the detection of tumors by iRFP in deep organs has been demonstrated by Filonov et al. 11 and confirmed in the present study, there is no data available using such a tool for visualization of intraosseous-implanted tumors. Intrabone emitted signals are bright enough to be detected and quantified without any contrast agent or chromophore previously administered to the animal [see Figs. 1(c) and 2(A) and 2(B)]. It is noteworthy that attempts to detect glioblastoma cell growth in the brain using IFP1.4 and biliverdin supplementation have been unsuccessful (data not shown). In conclusion, this iRFP-and lentiviral-based system present many other characteristics that combine the permanent and strong expression of iRFP with the low scattering and absorption by biological liquids and bones. Altogether, these features would allow this HIV-based iRFP-expression system to validly support the emerging therapeutic strategies in cancer treatment.