Glioma is the most common malignant form of primary brain tumor with the highest mortality rate (1). Grade IV glioblastoma multiforme (GBM) is considered the most aggressive, invasive, and lethal form of glioma, and it accounts for ∼50% of all glioma cases (2–4). GBM is considered almost invariably incurable, and patient survival after diagnosis is ∼15–18 months despite aggressive treatment paradigms such as combined surgical resection (5), radiotherapy, and temozolomide (TMZ) chemotherapy (6, 7). These highly aggressive malignancies are exceedingly invasive and proliferative (8, 9), have robust angiogenesis, and are resistant to apoptosis (10, 11). Combined, these factors contribute to therapeutic resistance and nearly 100% recurrence rates. To better study the natural progression of this devastating disease and for enhanced evaluation of new treatment regimens, ways of noninvasively and longitudinally characterizing GBM tumors in both preclinical GBM models and GBM patients are highly sought.
Preclinical glioma models have been invaluable for understanding patient GBM progression, for evaluating promising therapeutic strategies, and for developing new noninvasive imaging techniques to aid better tumor characterization (12–14). In particular, the orthotopic C6 rat model has provided considerable insight into the progression of GBM and the biological mechanisms regulating it (8, 13, 15–17). The C6 cell line was first developed by the Sweet Laboratory in the 1960s by means of repeated intravenous administration of the carcinogenic alkylating agent N-methyl-N-nitrosourea to outbred adult Wistar rats (18, 19). Since then, C6 cells have been isolated and grown in culture, and these can be used to generate orthotopic syngeneic tumors in rats. Orthotopic C6 tumors share several specific tumor markers found in human GBM such as increased Ras pathway activity, as well as increased expression of platelet-derived growth factor-beta, insulin-like growth factor 1, epidermal growth factor receptor, human epidermal growth factor receptor 3, and glial fibrillary acidic protein (13). C6 tumors also share histopathological features found in human GBM, particularly a diffuse infiltrative pattern (8). Because of these favorable characteristics, C6 tumor-bearing rats have been used to evaluate therapeutic efficacy of various treatment regimens such as chemotherapy (15), radiotherapy (16), and cytotoxic gene therapy (17), among others.
To facilitate improved evaluation of treatments and to better understand tumor progression in the C6 model, numerous groups have used clinically relevant imaging tools such as positron emission tomography (20), computed tomography (21), and magnetic resonance imaging (MRI) (22). MRI has become the standard of care for the detection, staging, and assessment of treatment response in patients with glioma (23). The benefits of MRI include the ability to collect high-resolution images (hundreds of microns) with excellent soft tissue contrast without the use of ionizing radiation, allowing for changes in tumor morphology to be sensitively monitored over time. Various MRI contrast mechanisms are available, which can accentuate different tumor features, allowing better delineation of tumor boundaries and intratumoral features such as edema, necrosis, or hemorrhage. Multiparametric MRI (mpMRI) describes the collection of a suite of MRI images with different contrast mechanisms, providing both anatomical and functional information about tumors that can include total tumor volume, contrast-enhanced tumor volume, apparent diffusion coefficients (ADC), and measures of perfusion (24). mpMRI has previously been applied to the orthotopic C6 rat model (22, 25–27). For instance, Liao et al. used mpMRI to gather information about tumor volume and location, blood–brain barrier integrity, and edema in the C6 model (28).
In addition to MRI, preclinical molecular imaging modalities can provide complementary information on tumor biology. Bioluminescence imaging (BLI) requires cells to be engineered to express a luciferase reporter gene before implantation into animals. Thereafter, following administration of the appropriate luciferin substrate, one can collect the light produced by these engineered cells by means of a cooled charged-coupled device camera (29). BLI provides the ability to sensitively monitor the relative viability of cancer cells over time because of the requirement of cells to actively transcribe the reporter, as well as adenosine triphosphate (ATP) as a cofactor for light production. The main purposes of this study were to use BLI to track the viability of C6 tumors growing orthotopically in Wistar rats, as well as the combined use of BLI and mpMRI to explore a relationship between tumor viability and tumor volume in this model. We found that the combined use of this suite of imaging tools provided new insights into individual rat tumor development in this model and will be useful for more precise evaluation of the effectiveness of new treatment strategies aimed at halting or preventing tumor progression.
C6 rat glioma cells (CCL-107, American Type Culture Collection, Manassas, VA) were grown in Ham's F-12K (Kaighn's) medium (ThermoFisher Scientific, Waltham, MA) containing 2.5% fetal bovine serum, 15% horse serum, and 1% penicillin–streptomycin. All cells were grown at 37°C in a 5% CO2 incubator. Cells were regularly confirmed to be mycoplasma-negative using the MycoAlert Mycoplasma Detection Kit (Lonza, Basel, Switzerland).
Self-inactivating lentiviral particles coexpressing the red-shifted Luciola italica luciferase (Luc) and green fluorescent protein (GFP) under the control of the Ubiquitin C promoter were used (RediFect Red-FLuc-GFP; PerkinElmer, Waltham, MA; Figure 1A). For efficient coexpression, Luc and GFP are separated by a T2A self-cleaving peptide in this vector. C6 glioma cells were transduced for 24 hours at a viral multiplicity of infection of 50 using polybrene (8 μg/mL). Following transduction, fluorescence-activated cell sorting (FACS) was performed twice on modified C6 cells to select for the highest GFP-expressing cells using a BD FACSAria™ III sorter (BD Biosciences, Franklin Lakes, NJ).
Cell Doubling Rate
Naïve C6, modified luciferase-expressing C6 cells (C6Luc) without D-Luciferin and C6Luc cells incubated with 150 μg/mL of D-luciferin (C6Luc150) were seeded into individual wells of a 6-well plate (2 × 103 cells/well; n = 6 per cell condition). Every 24 hours for up to 144 hours, the total cell number from 1 well per cell condition was quantified using a hemocytometer. The doubling rates of the cell conditions were determined using the model y = a × 2t/b; here, y is the cell count, a is the number of cells in the initial seed, t is the amount of time elapsed in hours, and b is the doubling time of the cells in hours. A MATLAB script using the nonlinear least squares trust-region algorithm from the Curve Fitting Toolbox (MATLAB 2016b, The MathWorks Inc., Natick, MA) was used to determine parameters a and b.
BLI Signal vs Cell Number
C6Luc cells (1 × 106, 5 × 105, 2.5 × 105, 1.25 × 105, 6.25 × 104, and 3.13 × 104) were seeded into individual wells and incubated for 24 hours (n = 3). D-luciferin (150 μg/mL) was added to each well, and plates were imaged 5 minutes later with an IVIS Lumina XRMS In Vivo Imaging System (PerkinElmer). Images were analyzed using Living Image Software (IVIS Imaging Systems, PerkinElmer) to obtain the average radiance (photons/sec/cm2/steradian) per well.
Orthotopic C6 Glioma Rat Model
All animal procedures were performed in accordance with relevant guidelines and regulations stipulated by an animal use protocol approved by the University Council on Animal Care, Animal Use Subcommittee at Western University (Animal Use Protocol: 2010-040). Male Wistar rats (n = 12; age, 4–6 weeks; weight, 200–225 g; Charles River Laboratories, Wilmington, MA) were used. Rats were anesthetized with 5% isoflurane, maintained at 2% isoflurane (1 L/min oxygen), and placed in a stereotactic frame for C6Luc cell implantation (Stoelting Co., IL). Fur on the superior side of the head was removed to expose the skin, and the bregma was landmarked after incision. A burr hole was drilled 1-mm anterior and 3-mm lateral to the bregma on the right. A microsyringe (Hamilton 1700 series) containing 106 C6Luc cells suspended in 10 μL of Hank's buffered salt solution at 37°C was advanced to a depth of 4 mm in the brain. The cells were injected at a rate of 3 μL/min. Bone wax was used to fill the burr hole to prevent cell suspension reflux. The incision was sutured and 1 mL of Meloxicam (5 mg/mL; Metacam, Boehringer Ingelheim Vetmedica, Ingelheim, Germany) was administered via a subcutaneous injection.
Each animal was imaged 4, 8, 11, 15, and 18 days post cell implantation. Before each imaging session, all animals were anesthetized with isofluorane and a tail-vein catheter inserted and secured to enable injection of the MRI contrast agent, gadolinium-diethylenetriaminepentacetate (Gd-DTPA). During longitudinal imaging, 1 rat was euthanized before the end point because of a rapidly increasing tumor burden and deteriorating health; data from this animal were omitted.
Anesthetized rats were given an intraperitoneal injection of D-Luciferin (150 mg/kg; PerkinElmer) and imaged using the same imaging system mentioned above. To avoid imaging through the burr hole, animals were placed on their left side (tumor side closest to the camera) and images were collected for up to 30 minutes post injection with the following parameters: exposure time = 1 minute, binning factor = 8, f-stop = 1, and field-of-view = 12.5 cm.
Animals were imaged on a 3-Tesla (T) MRI scanner (Discovery MR750 3.0 T, General Electric Healthcare, IL). The heads of the anesthetized rats were placed inside an 8-rung, 33-mm-inside-diameter bird cage radiofrequency coil (Morris Instruments Inc., Ontario, Canada), which was located within a custom-built insertable gradient coil (30). Each MRI session consisted of the following: T2-weighted imaging, diffusion tensor imaging (DTI); and pre- and postcontrast-enhanced (CE) T1-weighted imaging after intravenous tail-vein administration of Gd-DTPA (0.5 mmol/kg; Magnevist, Bayer, Leverkusen, Germany). 3D T2-weighted images were collected using a 3D fast spin-echo (CUBE, General Electric Healthcare) sequence with the following parameters: repetition time (TR) = 4000 milliseconds, echo time (TE) = 62 milliseconds, echo train length = 160, field of view (FOV) = 60 × 30 × 30 mm, acquisition matrix = 128 × 128 × 100, section thickness = 0.6 mm, bandwidth = 62.5 kHz, and number of averages (NEX) = 9. DTI images were obtained using a 2D spin-echo single-section echo-planar imaging sequence with the following parameters: TR = 6000 milliseconds, TE = 71.1 milliseconds, FOV = 60 × 60 mm, acquisition matrix = 120 × 120, coronal sections = 130, section thickness = 3 mm, bandwidth = 166.7 kHz, NEX = 1, b-value = 1000 s/mm2. Pre-CE (n = 5) and post-CE T1-weighted images were obtained using a 3D ultrafast gradient-recalled echo sequence for brain volume imaging (Fast GRE BRAVO, General Electric Healthcare) with the following parameters: TR = 6.9 milliseconds, TE = 2.9 milliseconds, preparation time = 450 milliseconds, FOV = 60 × 30 × 30 mm, acquisition matrix = 120 × 120 × 60, section thickness = 0.5 mm, bandwidth = 62.50 kHz, NEX = 9, flip angle = 25°. Post-CE T1-weighted images were obtained ∼4 minutes after Gd-DTPA injection.
For BLI, the average radiance was measured by drawing a region of interest (ROI) over the head using the software Living Image (PerkinElmer). For MRI, ADC values were calculated using Functool, a vendor-provided software analysis package for brain DTI (General Electric Healthcare). Tumor volume measurements were obtained from T2-weighted and post-CE T1-weighted images using ITK-SNAP 3.6.0, a freeware image analysis tool (www.itksnap.org). Tumor boundaries, chosen on the basis of post-CE T1-weighted images, were manually contoured in ITK-SNAP and tumor volumes were calculated. If noncontrast-enhanced (NCE) regions were present within tumors, separate ROIs were drawn to measure CE and NCE tumor volumes. Microscopic images of histological sections were coregistered to post-CE T1-weighted magnetic resonance (MR) images with a nonrigid, interactive, thin-plate spline extension by Gibson in the software 3D Slicer (Surgical Planning Laboratory, Harvard Medical School, Boston, MA) (31).
At the end point (18 days post cell injection), animals were anesthetized with isoflurane and intravenously injected with pimonidazole (60 mg/kg; HypoxyProbe™-1; Hypoxyprobe Inc., MA). After 30 minutes, animals were euthanized with an overdose of isoflurane followed by a 1-mL tail-vein injection of 1-M potassium chloride to ensure cardiac arrest. Each rat was pressure-perfused through the left ventricle with 4% paraformaldehyde in a phosphate-buffered solution. Following perfusion, the brain was excised and stored at 4°C in 4% paraformaldehyde for an additional 24 hours. Brains were placed in solutions of progressively increasing glucose concentration (10%, 20%, and 30% w/v) for 1, 1 and 24 hours, respectively, embedded in optimal cutting temperature compound and frozen with a mixture of dry ice and methanol. Further 10-μm sections were obtained on a cryostat (Leica CM1860, Wetzlar, Germany) and stained for the following: hematoxylin and eosin (H&E), 4′,6-diamidino-2-phenylindole (DAPI), and hypoxia using the HypoxyProbe™-1 kit (according to manufacturer's instructions, Hyproxyprobe, Burlington, MA). Bright-field and epifluorescence microscopy was performed on an upright Zeiss Axio Imager Z1 (Carl Zeiss CG, Oberkochen, Germany).
Statistical analyses were performed using GraphPad Prism v7 (GraphPad Software, Inc., CA). A 1-way ANOVA and Tukey post hoc test were performed to compare doubling time across conditions. For most imaging measurements over time, a 1-way repeated-measures ANOVA and Tukey post hoc test was performed. A 2-way repeated-measures ANOVA and Sidak post hoc test were performed to compare the ADC values between the CE and NCE tumor regions. For correlational analysis, Pearson product–moment analysis was performed.
C6Luc cells coexpressing GFP and luciferase were obtained after lentiviral transduction and 2 rounds of FACS for GFP-positive cells (Figure 1). The first FACS enabled the selection and isolation of the highest GFP-expressing cells (4.52% of all cells). In a subsequent second sort, after the first population had been expanded, cells with the top 10% fluorescence intensity were selected and expanded (Figure 1, B and D). Next, the relationship between BLI signal intensity and cell number (Figure 2, A and B), as well as the growth rates of naïve C6, C6Luc, and C6Luc150 cells (Figure 2C), was evaluated. A strong positive correlation was found between cell number and BLI signal (Figure 2, A and B; R2 = 0.985, P = .001). No significant differences were observed between the doubling rates of C6 (11.9 ± 1.7 hours), C6Luc (12.4 ± 1.7 hours), and C6Luc150 cells (10.3 ± 1.0 h; P = .801).
In total, 12 Wistar rats were implanted orthotopically with 106 C6Luc cells and imaged with BLI and mpMRI on days 4, 8, 11, 15, and 18 following cell implantations. BLI images of a glioma-bearing rat are shown in Figure 3. Owing to significant variability across animals, no significant changes in BLI signal were detectable across time (Figure 3B; P > .05). However, within 7 of the 11 rats, the average radiance peaked near day 11 (group average, 5.2 ± 5.6 × 105 p/s/cm2/sr) and on average dropped 2.48-fold by day 15 (group average, 2.1 ± 2.3 × 105 p/s/cm2/sr) and 2.74-fold by day 18 (Figure 3C; pattern 1; group average, 1.9 ± 2.4 × 105 p/s/cm2/sr). In contrast, BLI signal increased gradually and plateaued around days 15 and 18 for 2 rats (Figure 3C; pattern 2); and slowly decreased after day 4 for the other 2 rats (Figure 3C; pattern 3).
Qualitative differences in MR tumor morphology were noted across animals. Nine rats displayed tissue damage caused by the needle tract injection till the end point. Evidence of a hemorrhage was seen in 1 rat on day 11 on the basis of T2-weighted MRI, but recovery was noticed by day 18 (not shown). Tumor volumes were measured on both T2-weighted and post-CE T1-weighted MR images (Figure 4A). Total tumor volume based on T2-weighted MRI significantly increased at each time point from day 4 to day 15 (Figure 4B; P < .05), whereas no significant change was detected from day 15 to day 18. Post-CE T1-weighted tumor volumes exhibited a statistically significant increase from 23 ± 10 mm3 on day 4 to 145 ± 87 mm3 on day 11 (Figure 4C; P < .05), and no significant changes were detected beyond day 11. Post-CE T1- and T2-weighted tumor volumes were strongly correlated (Figure 5A; R2 = 0.884, P = .001). When all 5 imaging time points were considered, T2-weighted tumor volume data did not correlate to BLI signal data (Figure 5B; R2 = 0.027, P = .226) and post-CE T1-weighted tumor volume data were poorly correlated to BLI signal data (Figure 5C; R2 = 0.074, P = .045). In contrast, T1- and T2-weighted tumor volume data from only days 4 to 11 were moderately correlated to BLI signal data (Figure 5D; R2 = 0.305, P = .001; Figure 5E; R2 = 0.237, P = .004), highlighting a relationship in tumor volume and BLI signal during early, but not late, tumor progression (ie, beyond day 11).
We noted in post-CE T1-weighted MR images that 7 of the 11 rodents displayed an NCE region on days 15 and 18. A representative rodent is shown in Figure 6A. Therefore, we separately measured the CE and NCE tumor volumes over time. CE tumor volume significantly increased from 23 ± 10 mm3 on day 4 to 143 ± 85 mm3 on day 11 (Figure 6B; P < .05), and no significant changes were observed beyond day 11. NCE tumor volume was not significantly different between days 15 and 18 (Figure 6C). CE tumor volume data were poorly correlated to BLI signal data (Figure 7A; R2 = 0.088, P = .028).
Next, we evaluated changes in tumor cellularity using longitudinal DTI data (Figure 7). An ADC map and matched post-CE T1-weighted MR images, with CE and NCE regions contoured, are shown in Figure 7A. On day 4, a large variability was noted in ADC values, presumably owing to tissue damage from the needle tract during the stereotactic surgery. Whole-tumor ADC values significantly increased from days 8 and 11 to day 18 (Figure 7B; P < .05). Of the 7 rodents that displayed NCE regions within tumors at days 15 and 18, the ADC values within these NCE regions were significantly higher than those in the surrounding CE regions (Figure 7C; P = .001).
MR-matched whole-brain H&E images showed large necrotic areas within tumors (Figure 8). DAPI and GFP images also showed the presence of an acellular core and many GFP-positive cells at the periphery of tumors, respectively. Pimonidazole staining showed fewer hypoxic regions, predominately adjacent to regions lacking nuclei. Coregistered ADC maps showed regions of increased ADC values that corresponded to NCE regions on post-Gd T1-weighted MR images and to necrotic regions lacking GFP-positive cells on histology.
Several groups have previously described the use of MRI-derived measures of tumor volume in the C6 model (32), and BLI has been used in some C6 tumors (33, 34). Our study describes the first combined use of mpMRI and BLI to quantitatively characterize orthotopic C6 tumor growth in rats. These 2 imaging modalities allowed longitudinal multiparametric assessment of tumor growth, including: total tumor volume; CE and NCE tumor volumes; ADC measures for whole tumor and CE and NCE regions; and relative cellular viability using BLI. Our results showed that MR measurements of tumor volume did not correlate with BLI signal beyond day 11 (Figure 5). In fact, many rats displayed a decrease in BLI signal while tumor volume continued to increase or remained stable. Previous studies have reported a linear relationship between tumor volume and BLI signal for glioma models, whereas other studies have shown a lack of a relationship (35–39). In particular, Jost et al. reported a small subset of animals (n = 2) showing decreased BLI signal despite increasing tumor volume and contributed it to hemorrhage and necrosis (39). Likewise, Rehemtulla et al. supported these findings and concluded that BLI signal was not always concordant with MRI tumor volume measurements, particularly for treated tumors or large untreated tumors that displayed sizeable necrotic cores (40). Supported by our post-Gd T1-weighted images, ADC data, and histology, we found that beyond day 11, the day when BLI signal decreased, numerous animals formed an ill-perfused, necrotic core. Khalil et al. reported that hypoxic regions within the tumor may also have a negative influence on BLI signal, as oxygen is an important cofactor in the luciferin–luciferase reaction (41). We also found small regions of hypoxia, which were predominately adjacent to necrotic regions (Figure 8; #1 and #2). We also note that BLI only measures the viability of luciferase-expressing cancer cells, whereas tumors are composed of different types of stromal cells that can all contribute to tumor volume. There is a possibility that stromal cells growing faster than tumor cells can contribute to a larger fraction of tumor volume over time, making tumors appear larger in MR images while the BLI signal goes down. Loss of reporter expression in some of the transplanted cancer cells may also explain our results in part; however, this is less likely, as the mammalian Ubiquitin promoter used to drive reporter expression is unlikely to be silenced compared with viral promoters (eg, cytomegalovirus promoter). Realistically, it is likely that all these factors (necrosis, hypoxia, and limited substrate delivery) are contributors to the decreased BLI signal we see in our study at later time points in the majority of animals. However, combining post-Gd T1-weighted MRI, ADC mapping and BLI provides significant added value to studies that solely rely on MRI measures of tumor burden, as tumors can clearly increase in size but decrease in cellularity and overall cellular viability. Overall, our results highlight the variability in tumor growth across C6 tumor-bearing animals, as well as the value of a multimodality imaging approach to noninvasively characterize tumor development in individual animals.
Importantly, as seen in other cancer models (39, 40), changes in BLI measures of cellular viability may or may not differ from MRI measures of total tumor burden measures, meaning BLI provides complementary information to MRI for tracking tumor progression. Independently, BLI has been previously applied to the C6 model. Xi et al. used BLI to track treatment response to vincristine administered to Fischer 344 rats carrying luciferase-expressing C6 tumors in the brainstem (33, 34). Hwang et al. also applied BLI to monitor tumor growth in mice bearing luciferase-expressing C6 cells in the hind limb (42). However, our study highlights the large variability of disease progression across animals in this model. This is important information and it highlights that studies assessing treatments can benefit from the use of combined BLI and mpMRI and analysis of response for individual animal. Similar to Response Evaluation Criteria In Solid Tumors (RECIST) criteria in humans, the evaluation of treatment response in preclinical models based solely on morphological changes in tumor volume is problematic, as volume changes can be slow to occur and this method provides little to no information on the viability of cancer cells.
A limitation of this study is the lack of histological samples obtained at each time point from different animal cohorts. Evaluation of the histological tissue could help confirm the development of hypoxia or necrosis or lack of perfusion at each time point and explain the discrepancies obtained using BLI and mpMRI over time. Furthermore, the use of additional molecular imaging tools to identify and measure tumor growth characteristics would provide a deeper understanding of tumor development in this rat model. Some examples are 18F-fluorodeoxygluocose, which enables metabolic imaging of the tumor through glucose activity; 18F-fluoromisonodazole, which probes regional hypoxia; and 18F-galacto-arginine-glycine-aspartic acid, which is a probe for angiogenesis (43). Another imaging modality that would be useful in improving the ability to characterize tumor progression and treatment response is hyperpolarized (HP) 13C magnetic resonance spectroscopic imaging (MRSI). HP 13C-pyruvate MRSI is a metabolic imaging modality that measures the conversation of pyruvate to lactate, which provides information on cellular glycolysis (44). The imaging of metabolism with 18F-fluorodeoxygluocose positron emission tomography or HP 13C-pyruvate MRSI provides a method to monitor the biological activity of the tumor, which can be a useful surrogate measure of treatment response (45, 46).
Among the currently available intracranial glioma tumor models, there is no single model that can fully recapitulate all the distinctive characteristics of human GBM (13). The C6 glioma model has value and relevancy within this landscape because this model displays several characteristics of human GBM, such as the aggressive and diffuse infiltrative growth pattern (47), which is important for assessing therapies. In particular, Gieryng et al. concluded that the rat C6 glioma model is useful for evaluating immunotherapies that require an immunocompetent host (48). The use of our imaging tools in combination with the C6 model may be useful for evaluating current and emerging immunotherapies.
In conclusion, this was the first longitudinal study that applied mpMRI and BLI for characterizing orthotopic C6 tumor growth in rats. Future studies should apply this multimodality imaging approach to obtain as many independent, accurate, and complementary measures of tumor response to novel therapeutics.