1.

Introduction

Due to the absence of background signal and facile operation, small animal bioluminescence imaging of well-known luciferase gene reporter systems has significantly advanced studies.¹ However, both bioluminescence and fluorescence gene reporter imaging (referred to here as BGRI and FGRI, respectively) suffer from strong light attenuation since most gene reporters emit light in the visible wavelength region where tissue absorbance is high.² In addition, strong tissue autofluorescence results from excitation at visible light wavelengths, further compromising the performance of FGRI. Upon using excitation light at far-red or near-infrared (NIR) wavelengths, tissue absorption and autofluorescence are significantly reduced, enabling fluorescence gene reporters that can be excited at these wavelengths ideal for FGRI. However, the limited availability of far-red and NIR fluorescence gene reporters has restricted the development of FGRI. Recently, Roger Tsien and coworkers³ developed a far-red fluorescence gene reporter, IFP1.4, which could be excited by a red laser diode emitting at 690 nm for collection of 710 to 720 nm fluorescent photons. The protein was created through structure-based engineering of a bacteriophytochrome that fluoresces when it associates with ubiquitous biliverdin, the catabolic by-product of hemoglobin metabolism. More recently, Filonov et al.⁴ developed an analog of IFP1.4, iRFP, with greater fluorescent yield that further enhances the opportunity for gene reporter fluorescence tomography (GRFT).

Fluorescence tomography makes use of the surface measurement of emitted light for mathematical reconstruction of the source of light emission.⁵ Compared to bioluminescence tomography, GRFT could potentially result in more facile and robust three-dimensional (3-D) image reconstructions due to a higher photon count rate and the ability to conduct time-dependent measurements, as well as the possible combinations of multiple incident excitation patterns with multiple projection measurements of emitted light. However, acquisition of multiple projections requires multiple excitation sources/detectors,⁶ a conical mirror assembly,⁷ or a rotating gantry-based imaging system, the latter of which could also benefit with the integration of other imaging modalities, such as nuclear [positron emission tomography (PET), single-photon emission computed tomography (SPECT)] and x-ray computed tomography (CT).

Herein, we present the first tomographic small animal reconstructions of far-red fluorescence gene reporters in an orthotopic prostate cancer tumor model stably expressing IFP1.4 or iRFP. The methods employed a trimodal ($\text{fluorescence}/\mu \mathrm{PET}/\mu \mathrm{CT}$) gantry-based imaging system and a linear, regularization-free reconstruction algorithm employing the third-order simplified harmonics spherical approximation (${\mathrm{SP}}_3$) to the radiative transfer equation (RTE) and a priori anatomical information obtained from $\mu$CT. With IFP1.4 and iRFP stable expression in orthotopically implanted human tumor cells, the tomographic results demonstrate that far-red GRFT can provide good 3-D reconstruction for deeply located primary tumors as validated in vivo using an exogenous PET agent as well ex vivo fluorescence imaging.

2.

Materials and Methods

Human prostate cancer (PC3) cells were transfected by the IFP1.4 or iRFP reporter genes for our experiments. For IFP1.4 gene reporter, a mammalian viral vector was used to transfect cells and high expressing cells were selected and sorted on the basis of 710 nm fluorescence. Transfection with the plasmid was not efficient as it was for the iRFP gene reporter. For iRFP gene reporter, piRFP plasmid (Addgene, Cambridge, MA) was used to deliver iRFP reporter gene into the cells by using the electroporation method. Transfected cells were grown under G418 selection in Dulbecco’s modified eagle medium-F12/10% fetal bovine serum growing medium and were sorted depending on 710 nm fluorescence intensity. Male nu/nu mice (6 to 8 weeks old; Charles River Laboratories, Wilmington, MA) were prepared for the experiments. The transfected PC3 cells (${10}^6$ cells) were orthotopically implanted in the dorsal prostate of each mouse. Planar fluorescence imaging was used to longitudinally monitor tumor growth every week and animals were fed an alfalfa-free food to reduce autofluorescence for imaging^8,9 and tomography measurements. Animal studies were approved by the Animal Welfare Committee at the University of Texas Health Science Center at Houston and the Association for Assessment and Accreditation of Laboratory Animal Care.

The trimodal ($\text{fluorescence}/\mu \mathrm{PET}/\mu \mathrm{CT}$) imaging system was developed using a commercial, Siemens Inveon scanner as previously described,¹⁰ but modified to enable far-red GRFT. Briefly, the SPECT component of the commercial imaging system was replaced with a miniaturized Gen III intensified charge-coupled device (ICCD) and an illumination system within the CT gantry. For this work, the illumination system consisted of a 690-nm laser diode (Intense Inc., North Brunswick, NJ), laser mount, diode driver, temperature controller (Thorlabs, TCLDM9, LDC205, TED200, Newton, NJ), and two scannable mirrors. A 690-nm bandpass filter (Semrock Inc., Rochester, NY) was used to insure the monochromatic light and the collected light were passed through the 720-nm filters (Semrock Inc.) before being incident on the photocathode stage of the image intensifier.^8,9 Although the fluorescence imaging system is capable of time-dependent measurements through gain modulation of the laser diode and photocathode, herein we conducted time-independent, continuous wave measurements since we expected limitations in the number of fluorescent photons counted. With the gantry-based system design, acquisition of multiple projection images were made by transilluminating an incident point of excitation light and collecting fluorescent photons on the opposite side of the animal as the gantry was rotated 0, 45, 180, and 315 deg around the stationary animal. The fluorescent photon distributions were mapped onto the surfaces defined by CT as previously described.¹⁰

In order to perform GRFT, a linear, regularization-free reconstruction strategy was developed by neglecting the absorption coefficient of the fluorescence gene reporter at the excitation wavelength. In other words, the attenuation of excitation light from the gene reporter was assumed to be small compared to that from endogenous chromophores. With this assumption, the high-order ${\mathrm{SP}}_3$ approximation achieves more accurate reconstruction quality when compared to the classic diffusion approximation (DA)¹¹ because a more precise solution to the forward problem of photon propagation is obtained from the ${\mathrm{SP}}_3$. ^12–14 Briefly, the linear, regularization-free reconstruction method needs to minimize the following least-squares problem:

For tumor imaging, the antibody-based agents were prepared by attaching a 1,4,7,10-tetraazacyclododecane-1,4,7,10-tetraacetic acid (DOTA) chelating agent to antibodies that target epithelial cell adhesion molecule using prior methods.¹⁶ A previously described flow cytometry method¹⁷ was used to assess the biological activity of the antibody conjugates and showed that conjugation of a DOTA chelating agent did not affect binding compared to unlabeled antibody. ${}^{64}\mathrm{Cu}$-labeling was performed with high radiochemical yields ($>75\%$), and radiochemical purity was routinely $>95\%$ as determined by radio-thin-layer chromatography. The imaging agent was injected into the tail vein of tumor-bearing mice and trimodal imaging was performed 24 h after administration.

3.

Results and Discussion

3.1.

IFP1.4 Gene Reporter Fluorescence Tomography with In Vivo and Ex Vivo Validation

IFP1.4 mouse experiments were performed six weeks after orthotopic implantation of IFP1.4 transfected prostate tumor cells. Figure 1(a) to 1(d) illustrates the mapped IFP1.4 fluorescent photon distribution onto the mouse surface. Since the tumor locations varied from site of excitation and surface collection of fluorescence, distinct differences in fluorescent photon distribution occured in the different views. The range of the ICCD exposure time was 600 to 1200 ms to maintain a similar photon count rate and five frames of data were averaged for each projection image to improve the signal-to-noise ratio. Figure 1(e) to 1(g) shows the in vivo trimodal tomographical results and ex vivo white and fluorescence images of the prostate tumor. As is typical with antibody imaging, clearance occurs through the liver; hence the PET signal can be seen within the abdomen, but GRFT imaging showed consistent location of tumors with PET signal in the prostate region. Ex vivo experiments show that the tumor has strong fluorescent signals. We found that PC3-IFP1.4 cells exhibited fast tumor growth compared to wild type (WT) or prior orthotopic studies employing PC3-DsRed.¹⁶ The potential reason may be the effect of the virus transfection method and DNA combination. Although in vivo PET imaging can provide validation information, it is not always possible to discriminate bladder uptake from prostate uptake in the PET/CT images. Nonetheless, one can see that the fluorescence tomographic results in Fig. 1(e) are consistent with PET imaging of the tissue region containing the bladder and prostate. In addition, ex vivo imaging confirms the IFP1.4 fluorescence arises from the prostate tumor [Fig. 1(f) and 1(g)].

Fig. 1

IFP1.4 gene reporter fluorescence tomography six weeks after orthotopic cell injection with in vivo and ex vivo validation. (a) to (d) show the mapped fluorescent photon distribution on the mouse surface at four different views. (e) shows the trimodal ($\text{fluorescence}/\mu \mathrm{PET}/\mu \mathrm{CT}$) tomographic image results. Blue represents skeletal information from CT images; yellow represents PET imaging information; and red represents the reconstructed results of fluorescence tomography. The artifacts on the mouse surface are removed for better demonstration. (f) and (g) show the ex vivo white and fluorescent images of the prostate with tumor in ex vivo experiments.

3.2.

iRFP Gene Reporter Fluorescence Tomography with In Vivo and Ex Vivo Validation

Because of the uncharacteristically rapid growth rate of PC3-IFP1.4 cells, we subsequently transfected cells with iRFP. Figure 2 shows the fluorescence imaging of the transfected cells. Fluorescence microscopy [Fig. 2(a) and 2(b)] confirms iRFP fluorescent protein expression. In order to analyze the stability of the cells, PC3-iRFP cells were carried in culture for several passages. Levels of iRFP expression were examined by flow cytometry using parental PC3 cells as a negative control. Figure 2(c) shows similar levels of iRFP expression in the transfectoma after 15 versus $>50$ passages in culture, compared to the parental PC3 cell line.

Fig. 2

PC3-iRFP cell imaging and stability analysis. (a) and (b) are the white and fluorescent images of the transfected PC3 cells. (c) is the stability analysis of the transfected PC3 cells by using flow cytometry. “Alexa Fluor 700-A” means that the fluorescent signals are collected by using Alexa Fluor 700-A filters. The table shows the mean channel intensity of each sample.

Five weeks after being orthotopically implanted with PC3-iRFP cells, mice were imaged as described above. The trimodal reconstruction results are shown in Fig. 3(a) with GRFT conducted with four projections. Figure 3(b) and 3(c) shows ex vivo white and fluorescent images of the prostate with tumor ($\sim 3.0\mathrm{mm}$ diameter) and strong fluorescent signals. When compared to an IFP1.4-based mouse model, although there is a one-week growth difference, the tumor size in an iRFP-based mouse model is smaller but similar to that previously reported by our group using PC3-DsRed. Although it is difficult to distinguish the bladder and prostate tumor regions from the PET and CT images, the reconstructed results in fluorescence tomography have similar size with the tumor and are close to the bladder and prostate tumor region in PET images [Fig. 3(a)].

Fig. 3

iRFP gene reporter fluorescence tomography five weeks after orthotopic cell injection with in vivo and ex vivo validation. (a) shows the trimodal ($\text{fluorescence}/\mu \mathrm{PET}/\mu \mathrm{CT}$) tomographical results. Blue represents skeletal information from CT images; yellow represents the PET imaging information; and red represents the reconstructed results of fluorescence tomography. The artifacts on the mouse surface are removed for better demonstration. (b) and (c) show the white and fluorescent images of the prostate with tumor in ex vivo experiments.

3.3.

Fluorescence Tomography for iRFP Gene Reporter at Different Tumor Stages

Imaging was performed at 4 ($N=2$) and 10 to 12 weeks ($N=2$) after orthotopic PC3-iRFP cell implantation. Figure 4(a) and 4(b) shows the trimodal reconstructed results for mice imaged 4 weeks and 10 weeks after implantation, respectively. When the tumor is at its early stage, the reconstructed results from fluorescence tomography agree well with PET imaging and are similar to the results shown in Fig. 3(a). However, later stage tumors were found after euthanasia to be massive (diameter $\sim 30\mathrm{mm}$) [Fig. 4(c)]. Although it is difficult to distinguish the bladder and tumor regions in Fig. 4(b), the in vivo PET images appeared to be restricted to the tumor boundary and excluded from the interior of the tumor due to the well-known restricted transport of macromolecules in solid tumors.¹⁸ In addition, the reconstructed iRFP source does not correspond to the size of the tumor for reasons which remain to be investigated. Nonetheless, iRFP signal originates from the center of the tumor region demarking the differences in localization expected between exogenously administrated imaging agents and a gene reporter.

Fig. 4

iRFP gene reporter fluorescence tomography overlaid on CT and PET at different tumor stages. (a) and (b) are reconstructed results 4 and 10 weeks after cell implantation, respectively. Blue represents the skeletal information from CT images; yellow represents PET imaging information; and red represents the reconstructed results of fluorescence tomography. The artifacts on the mouse surface are removed for better demonstration. (c) In situ white light image for euthanized mouse depicted in (b) (the liver and intestine were removed).

3.4.

Gene Reporter Fluorescence Tomography with Control Mouse

In order to validate the stability of the developed GRFT and account for the effect of the excitation light leakage in the reconstructed results,^11,19–21 a nontumor-bearing mouse was used as a control subject. Because the intensifier gain and exposure time of the imaging system affect the imaging counts, we set the intensifier gain and exposure time for the nontumor-bearing mouse to the averages of two tumor-bearing mice. The detailed settings and acquired maximal counts can be found in Table 1 and the acquired raw images from the ventral side of the mice are shown in Fig. 5(a) to 5(c). Although there is some excitation light leakage, the maximal counts in two tumor-bearing mice are higher than that in the nontumor-bearing mouse. The reconstructed results acquired with four-projections of surface fluorescent photon distributions are shown in Fig. 5(d). One can find that all the reconstructed values in fluorescence tomography appear artifactually on the mouse surface, which shows the robustness of the developed fluorescence tomography.

Table 1

Summary of fluorescent images in Fig. 5(a), 5(b), and 5(c).

Animal model	Intensifier gain	CCD integration times	Maximum photon counts
Tumor bearing
Fig. 5(a)	7.4 V	1000 ms	44441.0
Fig. 5(b)	6.0 V	400 ms	52830.0
Nontumor-bearing control
Fig. 5(c)	6.7 V	700 ms	7148.0

Fig. 5

Fluorescence imaging comparison between tumor- [(a) and (b)] and nontumor- (c) bearing mice and fluorescence tomography with nontumor-bearing mouse (d). The images were acquired from the ventral side of the mice. Blue represents skeletal information from CT images; yellow represents PET imaging information; and red regions are the reconstructed results of fluorescence tomography.

4.

Discussion and Conclusion

Herein we demonstrated for the first time fluorescence tomography using far-red fluorescence gene reporters IFP1.4 and iRFP to image deep-seated orthotopic primary human prostate cancer in a mouse model. Due to the imprecise solutions with the classical DA, several high-order approximation models to the RTE have been developed to improve the reconstruction quality.⁵ Herein, we used the ${\mathrm{SP}}_3$ high-order approximation due to its improved accuracy.¹⁴ Regularization methods are also popular for current implementation of nonlinear reconstruction strategies because of the ill-posed characteristics of fluorescence tomography. Although stable solution can be achieved from the regularization, the approach suffers from the need to select regularization parameters to achieve the appropriate image and therefore is susceptible to the “inverse” imaging crime.²² While the heterogeneous optical properties can affect time-independent measurements as made herein, anatomical information from CT or magnetic resonance imaging²³ and precise photon propagation models may further improve the reconstruction quality with a priori information. Although the image reconstructions conducted herein were performed with the assumption of a homogeneous optical property distribution, it is possible to develop more complex reconstruction strategies using anatomical information.

Preliminary studies using IFP1.4 as a far-red gene reporter provided results similar to iRFP. However, the PC3-IFP1.4 cells exhibited more rapid tumor growth than PC3-iRFP or WT PC3 and resulted in centimeter-sized primary tumors, which may not be suitable for clinically relevant studies. Finally, the development of a near-infrared fluorescent gene reporter with excitation at wavelengths longer than 750 nm could further enhance fluorescence tomography by reducing the noise floor associated with autofluorescence that results from far-red excitation.