An anti-carcinoembryonic antigen (CEA) antibody fragment, the anti-CEA diabody, was fused to the bioluminescence enzyme Renilla luciferase (RLuc) to generate a novel optical imaging probe. Native RLuc or one of two stabilized variants (RLucC124A, RLuc8) was used as the bioluminescent moiety. A bioluminescence ELISA showed that diabody-luciferase could simultaneously bind to CEA and emit light. In vivo optical imaging of tumor-bearing mice demonstrated specific targeting of diabody-RLuc8 to CEA-positive xenografts, with a tumor:background ratio of 6.0 ± 0.8 at 6 h after intravenous injection, compared with antigen-negative tumors at 1.0 ± 0.1 (P = 0.05). Targeting and distribution was also evaluated by microPET imaging using 124I-diabody-RLuc8 and confirmed that the optical signal was due to antibody-mediated localization of luciferase. Renilla luciferase, fused to biospecific sequences such as engineered antibodies, can be administered systemically to provide a novel, sensitive method for optical imaging based on expression of cell surface receptors in living organisms.
Over the past decade, the number of recombinant proteins used for cancer imaging and therapy has grown considerably. Genetically engineered antibodies represent an important class of biomolecules that can be used to target specific cell surface proteins on diseased tissue and have spawned an ever-growing list of FDA-approved antibody therapeutics, particularly in cancer therapy (Adams and Weiner, 2005). More recently, engineering antibodies into smaller, defined fragments has yielded proteins with properties optimized for radiolabeling and in vivo imaging. Single-chain antibody Fv fragments (scFv, comprising the light and heavy chain variable regions joined by a peptide linker), and in particular dimers of scFvs called ‘diabodies’ (Holliger et al., 1993), retain antigen binding in a much smaller format. We have previously described an anti-CEA (carcinoembryonic antigen) diabody based on the murine T84.66 anti-CEA antibody (Wu et al., 1999). When radiolabeled with 123I, this anti-CEA diabody was highly effective in targeting antigen-expressing tumors in vivo, reaching a high uptake level in CEA-positive LS174T xenografts in mice at 2 h post-injection, and rapid blood clearance resulted in a high tumor:blood uptake ratio at 24 h (Wu et al., 1999). Switching the label to the positron emitter 124I (t1/2 = 4.2 days) enabled high contrast imaging of LS174T xenografts in vivo using small animal positron emission tomography (PET) (Sundaresan et al., 2003).
Radioactivity-based imaging modalities such as PET offer high sensitivity, tomographic imaging, are quantitative and have the potential for direct clinical translation. Optical imaging, on the other hand, offers a low-cost and simple alternative to PET for real-time imaging in small animal models and has the potential for higher sensitivity than PET approaches (Massoud and Gambhir, 2003). Disadvantages of optical imaging include limited depth penetration of visible light, due to scatter and absorption in tissue, and the lack of tomographic imaging at the present time. In addition, optical detection approaches may face barriers to clinical development such as the potential immunogenicity of fluorescent/bioluminescent proteins.
In vivo bioluminescence imaging has largely employed the luciferase from the firefly (FLuc), Photinus pyralis, which in the presence of O2, magnesium and ATP, catalyzes the oxidation of d-Luciferin to yield light. Another luciferase, also used in optical imaging, is that of the sea pansy Renilla reniformis, a soft coral that displays bioluminescence upon mechanical stimulation. The 36 kDa Renilla luciferase (RLuc) protein catalyzes the oxidative decarboxylation of its substrate, coelenterazine, to release light with a maximum intensity at 480 nm (Matthews et al., 1977). The light emission exhibits flash kinetics in vitro, peaking within the first second after the addition of coelenterazine (Bhaumik and Gambhir, 2002). RLuc has been used as a marker of gene expression in bacteria, yeast, plant and mammalian cells (Lorenz et al., 1991, 1996; Mayerhofer et al., 1995; Liu et al., 1997). Furthermore, Renilla luciferase has been used for in vivo imaging applications as a reporter gene, as well as to tag cells (Massoud and Gambhir, 2003; Gross and Piwnica-Worms, 2005; Shah et al., 2005). Another coelenterazine-dependent luciferase, from the marine copepod Gaussia princeps (Verhaegen and Christopoulos, 2002), has recently been validated as a reporter gene for in vivo imaging applications and also holds promise (Tannous et al., 2005).
The present work seeks to extend bioluminescence imaging beyond reporter gene applications, which require engineering of cells and animals to introduce foreign luciferase genes. We have enabled the imaging of endogenous proteins and processes by using the luciferase protein itself as an imaging probe. In particular, we have developed a novel probe for optical imaging of cancer by fusing a cancer-targeting engineered antibody, the anti-CEA diabody, to RLuc. We have chosen to use Renilla luciferase in the present work owing to its smaller size compared with FLuc and the fact that the oxidation reaction is not dependent on ATP, which is critical considering CEA is a cell surface target and the extracellular environment contains very low levels of ATP. Three versions of Renilla luciferase were incorporated in this fusion: native RLuc, RLucC124A or RLuc8. RLucC124A is a previously described stabilized mutant (Liu and Escher, 1999), and RLuc8 is a novel 8-mutation variant with 200-fold increased enzymatic stability (Loening et al., 2006). A bifunctional fusion protein demonstrating CEA binding and bioluminescence activity was produced and characterized. Following intravenous administration of this probe, CEA-expressing tumor xenografts in mice could be detected by optical imaging. The probe was additionally studied for tumor targeting and biodistribution by radiolabeling the protein and performing in vivo microPET imaging.
Materials and methods
Construction and expression of anti-CEA diabody-Renilla luciferase (Db-RLuc) fusion proteins
Genes encoding the T84.66 anti-CEA diabody (Wu et al., 1999) were fused to the RLuc gene from the phRL-CMV vector (Promega Corporation, Madison, WI) by splice overlap PCR, incorporating linkers of two different lengths (5 or 18 amino acid residues) (Figure 1A). The two Db-RLuc variants (Db-5-RLuc and Db-18-RLuc) were inserted into the bacterial expression plasmid pKKtac (Takkinen et al., 1991) using HindIII and EcoRI sites. The C124A mutation was introduced using a Quikchange XL mutagenesis kit (Stratagene, La Jolla, CA) to generate Db-5-RLucC124A and Db-18-RLucC124A. An additional, further stabilized variant of RLuc, (RLuc8) containing a total of 8 amino acid substitutions (Loening et al., 2006) was fused to the diabody using the 18 amino acid linker to generate Db-18-RLuc8. All bacterial expression constructs included a pelB leader sequence for periplasmic secretion and a hexahistidine tag to facilitate purification.
For bacterial expression, Escherichia coli BL21 cells were grown in Luria–Bertani broth (LB) to an OD600 of ∼0.7, induced with a final concentration of 0.25% β-lactose, and grown 2.5 h at 29°C. Periplasmic extracts were prepared using Peripreps Periplasting Kits (Epicentre, Madison, WI) and fusion proteins purified by Ni-NTA chromatography (Qiagen, Valencia, CA). Eluted proteins were stored in aliquots with 1.0% human serum albumin (Bayer Corporation, Elkhart, IN) as carrier at 4°C. The Db-18-RLuc8 construct was also produced by secretion from NS0 murine myeloma cells using the pEE12 expression vector as previously described (Yazaki et al., 2001). Culture supernatants were centrifuged, filtered, chromatographed on a Protein L column (Pierce, Rockford, IL) and eluted using a pH gradient. Final concentrations of the purified proteins were determined by absorbance at 280 nm. For in vivo studies, buffer was exchanged to phosphate-buffered saline (PBS) using Microcon YM-30 Centrifugal Filter Devices (Millipore, Bedford, MA).
Binding, bioluminescence and stability of Db-RLuc fusion proteins
Bioluminescence measurements were carried out as previously described (Bhaumik and Gambhir, 2002), using a Flash & Glow LB955 luminometer equipped with an autoinjector (Berthold Technologies, Oak Ridge, TN). Results are reported in relative light units (RLU) measured with a 10 s integration time.
Purified proteins were electrophoresed under non-reducing and reducing (0.1 mM DTT) SDS–PAGE conditions on pre-cast 4–20% polyacrylamide Ready-Gels (Bio-Rad Laboratories, Hercules, CA). For visualization of protein bands, gels were stained with Microwave Blue (Protiga Inc., Frederick, MD). For western blots, gels were electroblotted onto nitrocellulose membranes (Bio-Rad Laboratories) and incubated with a 1 : 2000 dilution of mouse anti-Renilla luciferase (Chemicon International, Temecula, CA) and a 1 : 5000 dilution of the secondary AP-conjugated goat anti-mouse IgG (Fc specific) antibodies (Sigma, St Louis, MO) for detection of Renilla luciferase. Membranes were developed using BCIP/NBT alkaline phosphatase substrate (Promega, Madison, WI).
Bifunctionality was assessed by capturing fusion protein on microtiter plates coated with a recombinant CEA fragment, N-A3 (100 μl of 2 μg/ml in PBS) (You et al., 1998). Bioluminescence was detected by auto-injection of coelenterazine and images were acquired using a FLUOstar Optima plate reader (BMG Labtechnologies, Offenburg, Germany). Competition experiments were performed in triplicate with the addition of 0–100 nM intact cT84.66 antibody.
To assess binding of fusion proteins to live cells, LS174T human colon carcinoma cells (ATCC, Manassas, VA; #CCL-188) were incubated with purified RLuc8 or Db-18-RLuc8 (5 μg) for 45 min on ice. Fusion protein was detected using murine anti-Renilla luciferase monoclonal antibody (Chemicon, Temecula, CA) followed by goat anti-mouse antibody conjugated to phycoerythrin (Jackson Immunoresearch, West Grove, PA) on a MoFlo flow cytometer (Cytomation, Fort Collins, CO). Murine anti-CEA T84.1 antibody was used as a positive control.
Stability was assessed by incubating aliquots of fusion protein with equal volumes of mouse serum (Calbiochem, EMD Biosciences, La Jolla, CA) at 37°C. At each time point, samples were withdrawn for bioluminescence assay and western blot. Half-lives of the proteins were determined using a least-squares fit to the log RLU with respect to time.
Optical imaging of tumor-bearing mice
All animal handling was performed in accordance with University of California Los Angeles (UCLA) Chancellor's Animal Research Committee guidelines. Tumors were induced in 7–8-week-old female athymic mice (Charles River Laboratories, Wilmington, MA) using LS174T human colon carcinoma cells or C6 rat glioma cells (1–2 × 106 cells) as previously described (Wu et al., 2000). Tumor masses were allowed to develop for 7–10 days and reached ∼140–200 mg. At time zero, mice were injected via tail vein with 100 μl concentrated mammalian expressed Db-18-RLuc8 (17 μg of total protein, 9.1 × 108 RLU). For the control experiments, 2.5 × 109 RLU of bacterially expressed RLuc8 was injected. After 2, 4, 6, 8 and 24 h, mice were anesthetized, injected via tail vein with 20 μg coelenterazine in 0.1 ml in PBS (0.7 mg/kg), and immediately imaged using a cooled CCD camera as described previously (Bhaumik and Gambhir, 2002). A region of interest (ROI) was selected over the tumor and the intensity was recorded as maximum photons/s/cm2/steradian. Three separate non-tumor areas were selected (one in the head region and two in the lower back region) and the intensity of these areas were averaged and used as the background value.
Serial and dynamic microPET imaging with 124I-labeled Db-18-RLuc8 and RLuc8
Purified RLuc8 (200 μg), and two batches of Db-18-RLuc8 (25 and 50 μg) were radiolabeled with 124I (Advanced Nuclide Technologies, Indianapolis, IN) as described (Kenanova et al., 2005). Labeling efficiencies were 34.7, 59.9 and 82.5%, respectively. Unincorporated iodine was removed by centrifugation through Sephadex multi-spin G-10 separation columns (Island Scientific, Bainbridge Island, WA). Prior to imaging tumor-bearing mice, thyroid and stomach uptake of radioiodine was blocked as described previously (Kenanova et al., 2005). For static scans, mice were injected via the tail vein with 140–150 μCi of 124I-Db-18-RLuc8 in saline/1% HSA. After 4 or 21 h, mice were anesthetized using 2% isoflurane and scanned for 10 min using a FOCUS 220 microPET (Concorde Microsystems, Inc, Knoxville, TN). Dynamic scanning was conducted on anesthetized non-tumor-bearing mice by initiating image acquisition immediately upon tail-vein injection of the radiolabeled protein (90 μCi of 124I-Db-18-RLuc8 or 140 μCi of control 124I-RLuc8) and acquiring emission data for 2 h. Following microPET imaging, a subset of the mice was immediately scanned using the MicroCAT™ II tomograph (ImTek Inc, Knoxville, TN) as previously described (Chow et al., 2005). MicroPET images were reconstructed using an FBP algorithm (Kinahan and Rogers, 1989; Defrise et al., 1997), and microPET and CT images were coregistered and displayed using AMIDE (Loening and Gambhir, 2003). After final imaging, animals were euthanized and the tumors, spleen, liver, kidneys, lungs and blood were recovered, weighed and counted for radioactivity (Wizard 3, Perkin-Elmer Life Sciences). Percent injected dose per gram (%ID/g) was calculated after decay correction. Tumor to background ratios were determined for the positive and negative tumors by selecting three ellipsoid ROIs over each tumor and a non-tumor background area.
Results are expressed as the mean plus or minus the standard error of the mean. One-tailed Student's t-tests were used for all hypotheses testing.
Expression and stability of Db-RLuc fusion proteins
Bacterial periplasmic secretion was initially employed for expression of fusions of anti-CEA diabody to Renilla luciferase, assembled using peptide linkers of 5 or 18 amino acid residues (Figure 1A). In addition to the native Renilla luciferase sequence, a cysteine to alanine mutation at position 124 [position 152 in (Liu and Escher, 1999)] was also utilized based on the previously reported increase in stability conferred by this mutation (Liu and Escher, 1999). Total recoveries of bioluminescence enzyme activity from a standard bacterial culture volume were roughly 5-fold higher for the Db-5-RLucC124A and Db-18-RLucC124A variants compared with their parental counterparts Db-5-RLuc and Db-18-RLuc.
Stability of the bioluminescence activity over time was evaluated by incubation of samples with equal volumes of mouse serum at 37°C. The activity half-lives increased ∼12-fold after incorporation of the C124A mutation, with the half-life of Db-5-RLuc increasing from 0.7 to 9.1 h, and that of Db-18-RLuc increasing from 1.4 to 15.6 h (Figure 2). Western blot analysis of Db-5-RLucC124A and Db-18-RLucC124A at the corresponding time points showed that this reduction was not due to proteolysis, as no change in the pattern of protein bands was observed over time (data not shown).
Subsequently, a homology-based mutagenesis approach was used to identify a set of additional mutations in Renilla luciferase that confer substantially improved enzymatic stability. Combination of eight amino acid substitutions (A55T, C124A, S130A, K136R, A143M, M185V, M253L, S287L) resulted in a variant, designated RLuc8, that exhibited 4-fold brighter light emission compared with the native enzyme and an activity half-life in mouse serum of >200 h (Loening et al., 2006). RLuc8 was incorporated into an anti-CEA diabody-luciferase fusion protein (Figure 1A), and recoveries of bacterially expressed Db-18-RLuc8 luciferase activity were an additional 2-fold higher than yields of the corresponding Db-18-RLucC124A protein. Mammalian expression of Db-18-RLuc8 proved especially suited for recovery of intact fusion protein. Protein L chromatography resulted in higher protein recoveries and separation of Db-18-RLuc8 from RLuc8 cleavage products. Typical recoveries were 1 mg purified Db-18-RLuc8 from 100 ml culture supernatant following Protein L chromatography. When examined by SDS–PAGE under reducing conditions, the Db-18-RLuc8 existed mainly as a protein band of ∼62 kDa corresponding to the size of the intact monomer and a band of 25 kDa corresponding to the size of the diabody. Western blots probed with anti-RLuc antibodies detected the 62 kDa band along with lower molecular weight bands corresponding to cleavage products (Figure 1B).
Functional characterization of Db-RLuc fusion proteins
A bioluminescence ELISA assay was developed to determine whether the Db-RLuc proteins retained the activities of both fusion partners: CEA-binding and light production. Db-5-RLucC124A and Db-18-RLucC124A and Db-18-RLuc8 proteins captured in wells coated with recombinant CEA N-A3 protein retained the ability to oxidatively decarboxylate coelenterazine and emit light. Moreover, the fusion proteins were displaced and the bioluminescent signal reduced by competition with increasing concentrations (0.1–100 nM) of T84.66 anti-CEA antibody (Figure 3), confirming that the Db-RLuc fusion proteins retain the epitope specificity of the parental antibody. Flow cytometric analysis using the primary anti-Renilla luciferase monoclonal antibody confirmed binding of Db-18-RLuc8 to live LS174T colon carcinoma cells that express CEA on the cell membrane. When gated as shown, 83.8% of the LS174T cells stained positive for the Db-18-RLuc8 fusion protein (similar to the positive control, the T84.1 murine IgG1 anti-CEA antibody), as compared with 23.5% positive with RLuc8 alone (background staining was 19.7%) (Figure 4).
In vivo optical imaging with Db-18-RLuc8
Db-18-RLuc8 obtained by mammalian expression and Protein L purification was evaluated for its tumor-targeting properties in athymic mice bearing LS174T (CEA-expressing human colorectal carcinoma) and C6 (CEA-negative rat glioma) subcutaneous xenografts. Proteins were injected into xenograft-bearing mice (n = 7) via the tail vein, and 2, 4, 6, 8 and 24 h later optical images were acquired using the cooled CCD camera immediately after tail vein injection of coelenterazine (0.7 mg/kg) (Figure 5A). Signal was present in the positive tumor as early as 2 h after injection of fusion protein and a strong signal was evident at 4 h. At all time points, the calculated CEA-positive tumor to background ratio was significantly higher than the CEA-negative tumor to background ratio, with an increase in the CEA-positive tumor to background ratio from 2 to 6 h and then a decline over the next 18 h (Figure 6A). The CEA-positive tumor to background ratio was highest at 6 h (6.0 ± 0.8) after injection and significantly higher than the CEA-negative tumor to background ratio (1.0 ± 0.1) (P = 0.001) (n = 7). By 24 h, targeting was still seen, although the overall intensity was diminished, with a LS174T tumor to background ratio of 2.8 ± 0.7 and a C6 tumor to background ratio of 1.0 ± 0.1 (P = 0.03) (n = 7) (Figure 6A). Background was determined over non-tumor areas as described in Materials and methods.
A separate set of tumor-bearing mice was injected with RLuc8 in order to assess uptake in tumors and normal tissues by the luciferase alone. ROI analysis shows that the positive tumor to background ratio (1.2 ± 0.1) was not significantly different from the negative tumor to background ratio (1.2 ± 0.3) (P = 0.44) (n = 4) (Figure 7A). As a further control, four mice bearing CEA-positive and negative tumors were imaged before and after injection of coelenterazine; no specific signal was detected in either tumor or elsewhere in the body, with all values below 4 × 103 photons/s/cm2/steradian.
In vivo microPET imaging with 124I-labeled Db-18-RLuc8
Db-18-RLuc8 was radiolabeled with the positron emitter 124I. MicroPET imaging studies were conducted on four athymic mice carrying LS174T tumors averaging 130 mg (range 125–140 mg) and C6 xenografts averaging 70 mg (range 45–100 mg) to confirm localization and distribution of the fusion protein. Specific targeting to the LS174T xenograft was seen at 4 and 21 h post-injection, with no significant targeting to the CEA-negative tumor at either time point (Figure 5B). ROIs were drawn over the CEA-positive and CEA-negative tumors, and the tumor to background ratios were calculated for each. At 4 h, the LS174T tumor to background ratio of 3.8 ± 0.8 was significantly higher than the C6 tumor to background ratio of 1.2 ± 0.3 (P = 0.02) (n = 4). At 21 h, antigen-specific targeting was maintained, with a positive tumor to background ratio of 18.0 ± 5.3 and a negative tumor to background ratio of 1.6 ± 0.3 (P = 0.03) (n = 4) (Figure 6B). Following the 21 h time point, the mice were sacrificed and the dissected tumors and organs were weighed and counted. Uptake was significantly higher in the CEA-positive tumor than in the CEA-negative tumor (P = 0.01) with very little uptake in other tissues (Table I).
|LS174T||12.9 ± 1.80|
|C6||1.7 ± 0.40|
|Spleen||0.4 ± 0.04|
|Liver||0.4 ± 0.06|
|Kidneys||0.6 ± 0.07|
|Lung||0.8 ± 0.11|
|LS174T||12.9 ± 1.80|
|C6||1.7 ± 0.40|
|Spleen||0.4 ± 0.04|
|Liver||0.4 ± 0.06|
|Kidneys||0.6 ± 0.07|
|Lung||0.8 ± 0.11|
aAfter microPET imaging of 124I-Db-18-RLuc8, tissue biodistribution was determined as described in Materials and methods.
bValues are given as percent injected dose per gram (%ID/g) ± SEM.
When control 124I-labeled RLuc8 was injected into a non-tumor-bearing mouse, the protein cleared rapidly through the kidneys. Dynamic microPET scanning showed that the kidney signal peaked at 20 min with residual signal at 60 min (Figure 7B). 124I-labeled Db-18-RLuc8 injected into a non-tumor-bearing mouse showed a peak in activity in the liver at 20 min with a decreased signal at 60 min (Figure 7C). These observations are consistent with our expectation that in vivo the smaller RLuc8 (36 kDa) should experience rapid first-pass renal clearance, while the Db-18-RLuc8 should display primary hepatic clearance based on its larger size (120 kDa dimeric molecular weight).
We describe a bifunctional, engineered antibody-luciferase fusion protein (Db-18-RLuc8), which, following systemic administration in tumor-bearing mice, localizes preferentially to antigen-positive tumors and generates a bioluminescent signal that can be detected in the living animal. Renilla luciferase has been extensively employed as a reporter gene, which requires introduction of reporter gene constructs into cell lines or transgenic animals and relies on synthesis of the luciferase by the modified cells. In the current application, the luciferase protein itself is used, and its fusion to cancer-specific antibodies enables in vivo localization to tumors following intravenous injection. Fusion of target-specific protein or peptide moieties to luciferase represents a new class of optical imaging probes that can be used to detect expression of endogenous tissue-specific or disease-specific markers.
Successful development of an antibody-luciferase fusion required attention to its three components: the antibody, the luciferase and the linkage between the two. Carcinoembryonic antigen was selected as an appropriate antibody target as it is a highly expressed and well established cell surface marker in colon cancer. The anti-CEA diabody (non-covalent dimer of scFvs) used in this work was chosen as the minimal, bivalent engineered antibody fragment that is still capable of efficient localization to CEA-expressing tumors in vivo (Wu et al., 1999). Furthermore, examination of the X-ray crystallographic structure of the anti-CEA diabody suggested it would be well suited as the N-terminal partner in a fusion protein (Carmichael et al., 2003). The C-termini of the diabody subunits are ∼70 Å apart and on an alternate face of the protein from the antigen combining site (Carmichael et al., 2003). Thus, fusion of an additional protein domain to the C-termini of the diabody should not interfere with binding.
As the bioluminescence partner in this fusion protein, Renilla luciferase was selected over Firefly luciferase (FLuc), which has an emission spectrum more suitable for in vivo imaging, for several reasons. The smaller size of RLuc (36 kDa) was preferable to that of FLuc (61 kDa), in order to minimize the overall size of the fusion protein, particularly since the fusion protein is expected to dimerize through the diabody domains. Additionally, as noted earlier, a key feature of RLuc is its ability to catalyze the oxidation of coelenterazine in the absence of ATP, which is not expected to be available in extracellular settings.
Initially, bacterial periplasmic secretion was employed for expression of the Db-RLuc fusion proteins. Bacterial expression allowed for purification of small amounts of the Db-RLuc fusion proteins, although only about half of the fusion protein was recovered intact. Owing to the low yield of intact fusion protein, mammalian expression was explored as an alternative. Recoveries of intact Db-RLuc were greater and, owing to the lack of proteases, the protein purified via mammalian expression remained stable for several months when stored at 4°C. For these reasons, we chose to express the Db-18-RLuc8 using only the mammalian expression system. This produced enough intact fusion protein for use in our in vivo studies.
A major challenge was the lability of the enzyme activity at 37°C in serum. We incorporated a previously described cysteine to alanine mutation that improves the half-life of native RLuc as measured in vitro in cell culture medium 6-fold (Liu and Escher, 1999). Further work determined that seven additional amino acid substitutions within the Renilla luciferase yielded a protein with even greater stability and improved light output (Loening et al., 2006). Inclusion of the improved RLuc8 in the Db-RLuc fusion protein similarly improved the bioluminescence activity and provided the stability required for in vivo experimentation.
Although previous studies showed that an 18–20 amino acid linker was preferable to 10 amino acids when Herpes Simplex Virus type 1 thymidine kinase was fused upstream from RLuc (Ray et al., 2003), we found only slight differences between fusion proteins assembled with 5 amino acid versus 18 amino acid linkers. Bioluminescence ELISAs as well as flow cytometric binding studies on CEA-positive LS174T cells demonstrated that the fusion proteins preserved the activities of both the fusion partners, as the proteins bound specifically to CEA and retained the ability to oxidize coelenterazine and emit light.
Most importantly, we demonstrated the capability of purified Db-18-RLuc8 to localize preferentially to CEA-expressing tumors in vivo by optical imaging studies in living mice. The lack of uptake in an antigen-negative tumor and absence of localization to either xenograft by the RLuc8 alone support the conclusion that the strong optical signal observed in the images was due to antibody-mediated targeting of the luciferase. Since the Km for coelenterazine and RLuc8 is 1.6 μM (Loening et al., 2006) and the global in vivo concentration of coelenterazine injected is ∼1 μM, there should be sufficient concentrations of substrate for light production.
In order to ascertain the physical localization, normal tissue distribution and clearance of the diabody-luciferase, static and dynamic microPET imaging was conducted on radioiodinated protein. The substantial uptake levels achieved in the CEA-positive tumors (12.9% ID/g at 21 h) are similar to previously reported activities achieved with anti-CEA diabody alone (Wu et al., 1999) and, again, support the conclusion that the observed optical signal was the result of antigen-specific localization of the luciferase. We also determined the CEA-positive and CEA-negative tumor:background ratios over time with both optical and microPET imaging. Using optical imaging, the CEA-positive tumor:background ratio increased over time, peaking at ∼6 h, and then decreased over the next 18 h. In the microPET imaging, the CEA-positive tumor:background ratio increased from the 4 to 21 h time point. This divergence in tumor:background ratios of the optical and microPET imaging is most likely due to the fact that over time, the luciferase portion of the fusion protein loses its enzymatic activity, leading to a decrease in the bioluminescence signal. In the microPET imaging, the tumor:background ratio reflects the physical presence of the fusion protein, rather than the enzymatic activity, resulting in a higher tumor:background ratio than that of the optical imaging.
We have demonstrated that a purified, exogenously administered luciferase protein can be targeted to cell surface markers on tissues in vivo and generate an imageable bioluminescence signal. Stability engineering of the Renilla luciferase facilitated its function in this non-native environment. The work can be readily extended to additional tissue-specific or disease-specific markers, based on the broad availability of antibodies. Additionally, other peptides or small protein-based ligands (e.g. growth factors, cytokines or protein-based ligands) can similarly be fused to the stabilized RLuc8 and are currently being studied in our laboratory. These luciferase fusion proteins could also be used to target extracellular (matrix/stromal) targets and internalizing receptors, and imaging of intracellular targets may eventually be possible. Biologically targeted luciferases should find wide applicability in monitoring small animal models of disease, with the principal advantage over similar fluorescent labeling approaches being the decreased background signal and resultant higher sensitivity of bioluminescent approaches. Potential clinical applications can be imagined; however, careful evaluation of any potential toxicity of the coelenterazine substrate, along with potential immunogenicity issues of the Renilla luciferase, would first need to be addressed. Additionally, multimodality imaging probes incorporating bioluminescence proteins along with a radiolabel could be developed in order to allow rapid transitions from high throughput evaluation using optical imaging in animals to PET imaging in humans. In summary, ligand–luciferase fusion proteins represent an important new class of imaging probes for sensitive optical imaging of cell surface receptor levels in vivo.
The authors wish to thank Dr Arion Chatziioannou, Dr David Stout, Waldemar Lladno and Judy Edwards for assistance with the microPET imaging studies. We appreciate the advice and assistance from Vania Kenanova, Karl Bauer and Felix Bergara during various stages of the work. We also thank Michael Gulrajani at the Flow Cytometry Core Facility, UCLA Jonsson Comprehensive Cancer Center. We extend our appreciation to Dr Bruce Bryan for inspiring this project. This work was supported in part by NIH Grants T32 GM 08652 (K.M.V.), P01 CA 43904 (T.O., A.M.W.), R24 CA92865 (S.S.G.), R01 CA082214 (S.S.G.), Department of Defense NDSEG Fellowship (A.M.L.) and Stanford Bio-X Graduate Fellowship (A.M.L.); A.M.W. is a member of the UCLA Jonsson Comprehensive Cancer Center (NIH P30 CA 16042).