In Vivo Testing of Renilla Luciferase Substrate Analogs in an Orthotopic Murine Model of Human Glioblastoma

Abstract

In vivo bioluminescent imaging using cells expressing Renilla luciferase is becoming increasingly common. Hindrances to the more widespread use of Renilla luciferase are the high autoluminescence of its natural substrate, coelenterazine, in plasma; the relatively high absorbance by tissue of the light emitted by the enzyme–substrate reaction; rapid clearance of the substrate; and significant cost. These factors, save for the cost, which has its own limiting effect on use, can combine to reduce the sensitivity of in vivo assays utilizing this reporter system, and methods of increasing light output or decreasing autoluminescence could be of great benefit. A number of analogs of coelenterazine are being investigated that may accomplish one or both of these goals. In this study, we report on the testing of two new substrate analogs, EnduRen™ and ViviRen™, manufactured by Promega Corporation, in an orthotopic murine model of human glioblastoma expressing Renilla luciferase. We have tested these analogs in this cell line, both in vitro and in vivo, and find that the substrate ViviRen results in significantly greater light output than the natural substrate or the other analog EnduRen. This new substrate could be valuable for studies where greater sensitivity is important.

Introduction

In vivo bioluminescence imaging is a widely used method for measuring light output of cells induced to express bioluminescent reporter enzymes upon substrate addition. Serial bioluminescence imaging of animal models of disease can provide information not easily obtained by other methods and is an efficient, high-throughput method with perhaps the highest sensitivity of detection of any animal imaging modality [1]. The increasing application of bioluminescent imaging to in vivo studies has resulted in a growing interest in using two or more reporter proteins in the same animals. The two most commonly used bioluminescent reporters are firefly and Renilla luciferase [1–10]. The substrate for firefly luciferase, luciferin, is a small molecule that readily passes through cell membranes and barriers, including the blood–brain barrier (BBB), by simple diffusion, whereas the substrate for Renilla luciferase, coelenterazine, may not [3,7].

The use of Renilla luciferase can be hindered by the fact that its substrate, coelenterazine, is unstable in serum and plasma and exhibits a high degree of autoluminescence [11]. This negatively affects the signal-to-noise ratio and can make it difficult to ascertain where true signal is versus that of background, particularly in cases where high sensitivity of detection is required. Renilla luciferase catalyzes the monooxygenation of coelenterazine to coelenteramide and results in the creation of a photon of light. Molecular oxygen is required for this reaction [3]. There is an increasing number of substrate analogs available for Renilla luciferase, one of which has been reported to have increased signal emission in vivo [11]. We report here on the use of two new substrate analogs that have been designed specifically to increase bioluminescent signal output upon reaction with Renilla luciferase while reducing the amount of autoluminescence.

These two coelenterazine analogs, ViviRen™ and EnduRen™, developed by Promega Corporation (Madison, WI), were engineered to reduce the autoluminescence found with coelenterazine, as well as to have higher inherent light output in the presence of the enzyme Renilla luciferase, and longer kinetics. They contain small blocking groups that protect the oxygenation site within coelenterazine, reducing the rate of degradation and the spontaneous emission of light that causes high background luminescence. The blocking groups are cleaved by esterases and lipases within cells, thus allowing formation of coelenteramide and light emission with the enzyme Renilla luciferase. In this study, we report on the usefulness of these two analogs in our orthotopic model of human glioblastoma. We have tested these compounds in a nude mouse model of human glioblastoma and found that EnduRen results in longer time to peak signal but lower light output than the native substrate, whereas ViviRen consistently exhibits approximately 0.5–1.0 log units higher signal at the same concentration as coelenterazine. Use of the analog ViviRen has allowed us to increase our sensitivity of detection using Renilla-expressing cells, and we are able to detect tumors in the brain of at least 0.6 mm in diameter. The substrate analog ViviRen should be useful in studies where increased sensitivity is important.

Materials and Methods

Cell Culture and Brain Tumor Injection

The human brain tumor cell line U87MG (glioblastoma) expressing Renilla luciferase (Rluc) was kindly provided by Michael Jensen, MD, of the City of Hope Cancer Center, Duarte, CA. Cells were transfected to express Rluc and zeocin (zeo) resistance genes (InvivoGen, San Diego, CA) as previously reported [12]. Cell lines were grown in Dulbecco's modified Eagle's medium (DMEM) with 10% fetal bovine serum and 0.2 mg/mL Zeocin in 5% CO₂ at 37°C.

Seven mice were injected with human glioblastoma for this study. Female (6–8 weeks) nu/nu mice were purchased from Harlan Sprague–Dawley (Indianapolis, IN). Details of the orthotopic xenotransplant model in nu/nu have been described previously [13]. In summary, 1.5 × 10⁵ or 6 × 10⁴ tumor cells were suspended in 2 or 1 μL, respectively, of DMEM. Originally, we injected four mice with 1.5 × 10⁵ tumor cells, but for later experiments where we wished to assess the substrates on small tumors (< 1.0 mm in diameter), we injected three new mice with 6 × 10⁴ tumor cells. Tumor cells were injected using a stereotaxic apparatus over 5 min into the forebrain at 2.0 mm lateral and 0.5 mm anterior to the bregma, at a depth of 3.3 mm. Mice were kept under general anesthesia (100 mg/kg ketamine and 10 mg/kg xylazine) during these procedures. All animal studies were done according to NIH guidelines and approved by the local animal care committee.

Substrates and Preparation

All three substrates were kindly provided by Promega (Figure 1). We used equimolar amounts (0.295 mM working concentration) of each substrate for the imaging studies performed here. This is equivalent to 1 mg coelenterazine/kg body weight, the reported amount used for in vivo imaging of Rluc expression [11]. All substrates were diluted in dimethyl sulfoxide or ethanol and aliquots were made using phosphate-buffered saline (PBS) and 0.1% bovine serum albumin, and these aliquots were used for both in vitro and in vivo studies. All dissolved compounds were stored in either −80°C or 20°C freezers prior to thawing for experiments.

Bioluminescent Imaging

Bioluminescent imaging was performed using a prototype imaging system (IVIS) manufactured by Xenogen Corporation (Alameda, CA). The system is linked to a PC and runs Living Image software (version 2.50, Xenogen) and Igor (Wavemetrics, Lake Oswego, OR) using Microsoft Windows 2000 (Microsoft Corp., Redmond, WA). This prototype system uses the same Roper Scientific (Roper Industries, Inc., Duluth, GA) cooled, back-thinned CCD camera as do most of the IVIS systems (Xenogen has recently switched to another manufacturer of cameras).

In Vitro Studies

In order to determine whether the substrates produced signal in our U87MG cells, we first performed in vitro tests. For these studies, U87MG Rluc cells were plated overnight (50,000 cells/well) in black 24-well plates (Greiner Bio-One, Longwood, FL). Black plates were used to avoid luminescent signal “bleed” from nearby wells of the plate containing luminescent cells that can occur with bright cell lines and transparent plates.

Figure 1.

Molecular structures of native coelenterazine and the substrate analogs EnduRen and ViviRen. Coelenterazine analogs EnduRen and ViviRen have small groups added, as outlined by boxes. These groups are cleaved by lipases and esterases within cells.

Two plates were used for duplication. For each plate, three wells contained U87MG Rluc cells (50,000 cells each). Wells containing nontransfected U87MG cells and others containing medium only served as controls. To the three U87MG Rluc-containing wells on each plate, one substrate per well was added (20 μL per well). The individual substrates were added in the same fashion to control wells to test for background autoluminescence. One additional cell plate was originally used, with wells containing the same groups of cells as above; however, due to an error in calculation, 50 μL per well of substrate was added. This well plate is not included in this analysis for this reason, but the results were the same as with the other two plates, save for an exception discussed in the Results section.

Bioluminescent imaging was performed on each plate immediately following substrate delivery to all of the wells containing cells or medium. Each plate was imaged by placing it in the IVIS imaging system and acquiring images for 10 sec each at 5-min intervals for 30 min. Images were analyzed by region-of-interest (ROI) analysis using the Living Image and Igor software packages. Quantification of light output was obtained and expressed in photons per second per square centimeter per steradian (photons/sec/cm²/sr).

In Vivo Studies

Mice bearing tumors (N = 7) were imaged for Rluc activity using the cryogenically cooled, high-efficiency CCD camera of the IVIS device. Imaging was performed on subsets of injected mice (n = 3) at different days post tumor cell injection to assess the substrate utility on tumors of different sizes. For each portion of the studies (intravenous [iv] tests, intraperitoneal [ip] tests, small tumors, etc.), animals were imaged with all three substrates within 2 days to minimize effects of tumor size change. Animals were given iv or ip injections of 200 μL of substrate or substrate analogs (0.295 mM) and then placed in the light-tight chamber of the imaging system for image acquisitions beginning at 1 min postinjection. Successive images were acquired with 2-min acquisition times until after peak signal had been achieved. Mice were anesthetized with 2% isoflurane immediately prior to injection and for the duration of the scans. A grayscale photograph of the mice was first collected in the imaging box under dim light-emitting diode (LED) illumination, followed by acquisition and overlay of the pseudocolor luminescent image.

ROIs were drawn around the heads of all animals for the in vivo experiments or around the relevant wells of the 24-well plates for the in vitro studies. Quantification of light output was obtained using the Living Image and Igor software, and expressed in photons per second per square centimeter per steradian. Region size was the same in all animals.

Magnetic Resonance Imaging

Magnetic resonance imaging (MRI) was performed using a Bruker 7T Pharmascan (Bruker BioSpin, Ettlingen, Germany). Mice were sedated with and maintained under 2% isoflurane for the duration of the scans. Scanning was done ∼15–20 min post iv injection of 100 μL gadopentate dimeglumine to allow for peak concentration in the brain tumor, as we have determined through past experience [14]. Scans were performed according to the following protocol: TR = 3076.5 msec, TE = 36.2 msec, and NEX = 4. The field of view was 2.6 cm, and the final voxel sizes were 0.100 × 0.200 × 0.670 mm slice thickness, with contiguous slices.

MRI of the mouse brains with tumors were visualized using the Bruker Pharmascan software, and largest tumor axial dimensions were measured using this software for comparison to bioluminescent signal intensity.

Euthanization and Histology

For euthanization and histologic preparation of mouse brains, animals were anesthetized with ketamine/xylazine (100 and 10 mg/kg, respectively), the thoracic cavity and chest exposed surgically, the right atrium cut open, and 10 mL of saline flushed immediately into the left ventricle followed by 10 mL of 10% buffered formalin. Brains were then removed, fixed in buffered formalin, and embedded in paraffin for hematoxylin and eosin staining.

Statistical Analysis

One-way analysis of variance (ANOVA) and Student's t tests were used to assess the statistical significance of the bioluminescent results for the in vitro and in vivo tests. Peak signal intensities for each substrate and each route of administration were used for statistical analysis. The Pearson product–moment correlation was calculated to test for reproducibility and to assess light output versus tumor size. Linear regression was also performed on the bioluminescent signal output versus tumor size data. All data are presented as means ± standard deviation. p Values of < 0.05 were considered significant.

Results

In Vitro Studies

Bioluminescent signal measured from plates with U87MG Rluc-expressing cells was highly reproducible. Signal from wells with the substrate ViviRen added was highest, with an average peak approximately 1 and 2 log units above that of coelenterazine and EnduRen, respectively (p < .01) (Figure 2). Signal from cells with EnduRen was lowest among the wells with cells expressing Rluc, exhibiting a longer time to peak than either coelenterazine or ViviRen (30 min vs. 15 and 4, respectively). Signal from coelenterazine was approximately stable from 8 to 30 min, whereas signal from ViviRen was decreasing over this period.

Peak signal from coelenterazine wells was only 8% of the ViviRen peak, whereas peak signal from EnduRen-containing wells was 2% of ViviRen signal. Peak signal from the control wells with nontransduced U87MG cells and medium alone containing substrates was always less than 1% of peak ViviRen signal and below that of EnduRen and coelenterazine signal peaks.

As mentioned previously, one additional plate had 50 μL per well added rather than 20 μL per well. The relative results were identical to those above, except that in the two control wells with the substrate (ViviRen) added (with nontransduced U87MG and medium alone) the peak signal was 1.5–2 times greater than signal from the wells containing the substrate EnduRen and Rluc-expressing U87MG cells, demonstrating that ViviRen has relatively high autoluminescence in vitro to go along with its high true signal.

Figure 2.

In vitro testing of native substrate and analogs on human glioblastoma cell line. Plated cells had the three different substrates added as listed. Wells with U87MG Rluc cells and ViviRen added demonstrated the highest signal after substrate addition. Control wells with medium alone and nontransduced U87MG cells had significantly lower signal than wells with transduced cells following administration of substrates, save for when the substrate ViviRen was added. In this case, the autoluminescence from ViviRen was brighter than the signal from EnduRen in Rluc-expressing cells. True ViviRen signal in Rluc-expressing cells was, however, much greater than this autoluminescent signal in control wells. All points represent means ± SD.

In Vivo Studies

Intraperitoneal injections. Intraperitoneal injections overall resulted in lower bioluminescent signal output than iv injections (Figure 3). ViviRen produced the highest average peak signal following ip administration of substrate (n = 3) (p = .04). Compared to peak signal from mice following ip injection of ViviRen, peak signals from EnduRen and coelenterazine were 55% and 91%, respectively. Intraperitoneal injections exhibited a longer time to peak than iv injections, with the mean time to peak for ip administrations of coelenterazine and ViviRen at ∼20 min, and approximately 30–35 min for EnduRen.

To test for reproducibility of ip injections, one mouse was injected ip with ViviRen on successive days. Correlation of bioluminescent signal output was high at R = .99 (p < .001).

Intravenous injections. Following iv injections, bioluminescent signal exhibited an extremely rapid rise to peak in all animals for the substrates coelenterazine and ViviRen. Peak occurred almost immediately postinjection, with rapid decay of signal to levels comparable to ip within minutes (Figure 3).

As with ip injections, the highest peak signal in all mice was achieved following injection of the substrate ViviRen (p = .05). Mice with tumors greater than 1 mm in largest diameter (n = 3) injected with coelenterazine demonstrated signal similar to these same mice following injection with EnduRen, whereas mice with smaller tumors (< 1 mm) (n = 3) had significantly higher peak signal with coelenterazine than with EnduRen (p = .04). In fact, EnduRen signal was not significantly different from background in these mice with small tumors (see below).

We tested for reproducibility of iv injections by injecting the same mouse with ViviRen iv on successive days as above for the ip injections. Reproducibility was also high, with a correlation of R = .98 (p < .001).

To test if we could visualize small tumors (< 1 mm) with the substrates, we performed bioluminescent imaging on animals with small brain tumors as determined by MRI. Mice had tumors ranging in size from 0.6 to 0.9 mm (maximum diameter). Three mice were imaged following iv injections of the substrates. Intraperitoneal injections were not performed, as preliminary testing by ip injection of ViviRen demonstrated no signal from these small tumors. Intravenous injection with EnduRen did not enable detection of these small tumors, whereas with coelenterazine signal was approximately double that of background and that of EnduRen (p = .04). Following injection of ViviRen, signal increased dramatically and significantly, and tumors were easily visualized (Figure 4).

Figure 3.

Comparison of ip and iv injections of coelenterazine and its analogs in mice bearing brain tumors. Signal from iv injections showed greater peak signal compared to ip injections. Bioluminescent images are of representative mouse following administration of each substrate by iv and ip method. Note that peak signals post iv injections were greater than peak signals from ip injections in all cases. Intravenous ViviRen-induced signal was much greater than signal with any other substrate regardless of injection method (iv or ip). Coelenterazine and EnduRen signals were not significantly different from each other for either iv or ip injections in mice with these larger (>1.0 mm) tumors. Images of representative mouse show peak signals obtained with each substrate by each injection method and demonstrate that signal from iv ViviRen was greatest. All points represent means ± SD.

We also tested whether bioluminescent signal increased over time as tumors grew. One additional mouse was imaged at Week 9 post tumor injection and then again at Week 10. MRI showed tumor size had increased from 2.0 to 2.4 mm in this time. Bioluminescent signal following iv administration of substrate increased as well (Figure 5). This was true for all three substrates.

Figure 4.

Comparison of bioluminescent signal in mice with small brain tumors. Graph shows mean (± SD) signal achieved in brains of three mice with tumors less than 1.0 mm in maximum diameter following iv administration of substrates. Peak signal post ViviRen injection was significantly (p< .05) highest in these small (<1.0 mm) tumors, as it was in larger tumors.

Tumor size versus bioluminescent signal resulting from administration of each substrate in all seven animals was also plotted. Tumor size as determined by MRI approached a significant correlation with bioluminescent signal from ViviRen (R = .68, p = .06) and EnduRen (R = .74, p = .06) injection. Correlation of signal to tumor size was not significant with coelenterazine (R = .59, p = .12).

Upon histologic examination, all tumors were found to be largely composed of healthy tumor cells, have well-delineated borders, and have limited areas of necrosis.

Discussion

In this study we found that bioluminescent signal emitted from tumor-bearing mice was greatest in all cases with the substrate ViviRen. This was true for both ip and iv injections. Signal from EnduRen was consistently equal to or lower than signal following coelenterazine administration. EnduRen signal took longer to peak than either other substrate-induced signal, both with ip and iv injections. The reasons for this latter point are unclear but may be due to less efficient cleavage of one of the additional subgroups added.

The shorter time to peak following iv administration is not surprising because administration in this manner means that a bolus of agent is directly introduced into the blood stream, and the diffusion processes that dominate the early phases of ip injection are avoided. With iv injections we noted high signal in the lungs of the mice in the first images acquired after substrate administration (and in ip injections we noted high signal in the abdominal area), presumably due to substrate autoluminescence. This autoluminescence was not insignificant, and although the two substrate analogs were created in part to have lower autoluminescence than the natural substrate in vivo, in our studies it appeared that each retained relatively high background levels in regions where substrate accumulated. In our brain tumor model this was not a problem, but in models necessitating high sensitivity in thoracic or abdominal regions this could cause difficulty in detecting true signal. It is important to note that the signal we detected in the brains of our mice was indeed true signal and not autoluminescence, as control mice injected with the substrates did not show significant brain signal at all.

Figure 5.

Change in bioluminescent signal with tumor size change. Plots of bioluminescent signal following iv substrate administration in one mouse 9 weeks after brain tumor injection and repeated 1 week later (“wk 10”) to test if signal had increased along with size increase as measured by MRI. Bioluminescent and MRI images show brain tumor and signal at weeks 9 (a) and 10 (b). Although we observed signal increase with all substrates in this mouse from week 9 to week 10, only iv ViviRen and EnduRen signals approached a significant correlation with tumor size overall in our study.

It is thought that the luminescence due to autooxidation of coelenterazine is in part increased by the presence of albumin. In our studies, we limited the amount of albumin, used as a carrier to keep the substrate in solution, to 0.1% or lower. In this solution, all three substrates demonstrated similar amounts of autoluminescence. In future studies, the autoluminescence of substrate in various other solutions, such as PBS alone, should be tested.

We found good, although not quite statistically significant, correlation of bioluminescent light output with the substrates ViviRen and EnduRen to tumor size as determined by MRI. The significance of the ViviRen signal correlation to tumor size as measured on MR images became significant (p < .05) when an additional mouse with brain tumor was imaged following iv administration of ViviRen. This mouse was a pilot study mouse and was not included in our data. Taken as a whole, however, our data demonstrate that bioluminescent signal from Rluc-expressing brain tumor cells most likely reflects relative tumor sizes in vivo. It is quite possible that with larger numbers of mice, even the correlation of tumor size measured with MRI to coelenterazine-induced signal would be significant, although we could not support this with our relatively small sample size. The correlation of light emission to tumor size is similar to the results reported by others and found in prior work of ours comparing MRI to bioluminescent signal in animals containing brain tumors that expressed firefly luciferase [15]. To our knowledge, this is the first report demonstrating this for Renilla-luciferase-expressing brain tumors.

We did not attempt to optimize sensitivity of detection for any one substrate in this study. Our purpose here was to compare the substrate analogs to the native substrate at the same concentration and using the same imaging parameters. Longer image acquisition times, increased binning (combining) of pixels on the CCD camera, and higher doses of substrate could all lead to increased signal detection for a given number of luciferase-expressing cells. Reductions in dose of substrate could also be employed to save on imaging costs, as each substrate is relatively expensive. From the results here, it is reasonable to conclude that a lower dose of ViviRen relative to coelenterazine or EnduRen could be used in our model, and thus this expense could be reduced. Optimization of imaging parameters and substrate dosing should be taken into account for any individual animal model.

Following the in vitro studies we examined the cells in wells under a light microscope and did not find significant observable differences in cell viability as assessed by attachment and morphology. This lends support to Promega's own analyses on various cell lines that show no toxicity following natural substrate or analog administration. Studies of potential toxicity of these agents should be more thoroughly performed.

In this study we tested the two substrate analogs on one cell line, the human glioblastoma line U87MG. Our cells had been selected using zeocin resistance so that we were using cells that had near-100% expression of Renilla luciferase. Other cell lines may give different results. It is possible that one of the blocking groups is easier to cleave than the other and that different cell types may be more efficient at cleaving one substrate's blocking groups more than the other. Also, other cell lines may have more or less esterases and thus cause lower or higher light output. Therefore, studies using different cell lines in vivo may have different results, and this will need to be evaluated. We are currently planning to perform such studies using multiple cell lines induced to express Renilla luciferase in mice.

This study did not attempt to specifically address the issue of diffusibility of the agents between cells of the body. Although it is known that the substrate for firefly luciferase, D-luciferin, readily passes through all cell membranes including the BBB, less is known about coelenterazine, although it is thought that it does not cross the BBB easily, if at all [3,7]. The altered size and charge distribution of the two analogs would almost certainly affect their diffusion capabilities, but this remains to be definitively tested. It is possible that the differences in bioluminescent signal observed in mice with small versus large tumors iv injected with the substrate EnduRen are due to differences in BBB permeability or degree of vascularization. We did not test for this.

There is an earlier report of different coelenterazine analogs evaluated in vivo [11]. In that study, it was found that a substrate analog, termed coelenterazine-e, produced three- to fourfold greater signal than the native substrate in vivo but also exhibited higher autoluminescence. They concluded that the native substrate would be better suited for animal studies. Although it is unclear what result this may have on their work, in that report the substrates were all administered using a 1 mg/kg body weight standard (the standard for coelenterazine, as mentioned); however, the molarity of the solutions containing the substrates is a more appropriate measure because the different molecular weights of the substrate analogs will mean that more (or fewer) molecules will be available depending on their molecular weight for a given milligram per kilogram.

In this previous report, the substrate analog coelenterazine-h was also tested in vitro in HeLa cells and in C6 cells and in vivo in livers of mice [11]. Although Rluc signal was relatively high in the cell studies, the in vivo liver results were less encouraging, with signal approximately one-fifth that of native coelenterazine; however, the above-mentioned caveat regarding molarity must be kept in mind. This result is still noteworthy because one of the intermediates formed in cells post intracellular passage of EnduRen and ViviRen is coelenterazine-h.

It is reasonable to suggest that the additional side groups added to form EnduRen and ViviRen have altered their intracellular transport capacities relative to coelenterazine-h. It is possible that the light emitted from the reaction with the substrate analogs is shifted in wavelength from the light emitted by the natural substrate, coelenterazine. We did not test for this. If the wavelength of the light were shifted to longer wavelengths than the peak with coelenterazine (∼480 nm), then the light would be less absorbed and more easily penetrate tissue, and this would confound the results [16]. Future studies should address this issue.