Evaluation of TgH(CX3CR1-EGFP) mice implanted with mCherry-GL261 cells as an in vivo model for morphometrical analysis of glioma-microglia interaction

This work was conducted at the University of Saarland in strict accordance to the European and German guidelines for the welfare of experimental animals under the license 65/2013, approved by the Saarland state’s “Landesamt fuer Gesundheit und Verbraucherschutz” in Saarbrücken/Germany.

This study used the mCherry-GL261 tumor cells [19], a glioma cell line with expression of the fluorescent protein mCherry, kindly provided by Helmut Kettenmann (Max-Delbrück-Center for Molecular Medicine, Berlin, Germany). Cells were maintained on 75 cm³ culture flasks with DMEM/F12 [supplemented with 10 % (volume/volume) heat-inactivated fetal calf serum, and 1 % of penicillin/streptomycin solution, all obtained from Invitrogen (Karlsruhe, Germany)]. When the confluence reached about 90 %, the cells were harvested using 0.25 % Trypsin solution (Invitrogen, Karlsruhe, Germany). A pellet was formed by centrifugation at 1200 g for 3 min (Hettich Universal 30 F, Tuttlingen, Germany). Cells were diluted in 1.0 ml PBS and counted in Moxi-Z (Orflo, Hailey, USA). Aliquots containing 5x10⁴ cells in PBS were made and centrifuged at 1200 g for 3 min (Eppendorf MiniSpin, Hamburg, Germany). The pellet was re-suspended in 3 μL of PBS by gently mixing, and the 5 μL resulting solution was immediately injected into the mouse brain by using a microliter syringe (Hamilton) with a 25 gauge needle.

Adult TgH(CX3CR1-EGFP) heterozygous mice [12] backcrossed to C57BL/6 N background for more than 10 generations were maintained at a temperature and light controlled animal facility, and received food and water ad libitum.

All animals were anesthetized with Ketamine/Xylazine (Bayer, Germany) (140 mg/10 mg/kg body weight) by intraperitoneal injection. They had the top of head shaved, and the surgery site cleaned with iodine antiseptic. Bepanthen® cream (Bayer, Germany) was used to cover and protect the eyes. After proper mouse fixation in the stereotaxic instrument, a single skin cut of about 0.5 cm was performed by using scissors, followed by gently divulsion of the sub-cutaneous tissue. A 2.0 mm hand drill (Fine Science Tools, Heidelberg, Germany) was used to thin the skull at the injection site, approximately at 2.0 mm posterior and 1.5 mm lateral of the bregma. Five microliters of the mCherry-GL261 cell suspension (a total of 50.000 cells) were aspired with a micro-syringe and slowly injected, first 2.0 mm below skull surface into the cortex, and then the needle was pushed back 0.5 mm to inject the rest of the volume. About 5 min were spent to complete the injection and two additional min for totally removing of the needle. The sub-cutaneous tissue was gently washed with NaCl, 0.9 % and the skin closed with simple interrupted sutures. Iodine antiseptic was applied on the suture after the surgery to prevent infections. Immediately afterwards, the mice were kept on a heating plate until woken up. To release the pain, buprenorphine (0.09 g/30 g body weight) was administrated after the surgery and at the following day. Once the tumor cells were injected, the mice were housed individually in the same conditions as described before. For in vivo imaging, the procedure to inject the glioma cells was the same. However, after injection a cranial window was formed in the same area, to acquire the images as described below. Mice, which received injections of mCherry-GL261 cells, were daily monitored for weight and clinical status. If an animal’s weight dropped 15% below the baseline or became symptomatic, it was euthanized.

The procedures were carried out under anesthesia initiated by Ketamine/Xylazine (Bayer, Germany) (140 mg/10 mg/kg body weight). Afterwards mice were placed on a heating pad, heads were stabilized in a stereotactic frame using ear bars, artificial ventilation was continued with a gas mixture of O₂ (50%), and N₂O (50%) at 120 strokes/min (100–160 μl/stroke depending on the body weight). A longitudinal incision of the skin was performed between the occiput and the forehead. The subcutaneous membrane was completely removed. A 0.5 mm hand drill (Fine Science Tools, Heidelberg, Germany) was used to thin the skull and to open a 5.0 mm diameter window, approximately at 2.0 mm posterior and 1.5 mm lateral of the bregma. The dura mater was continuously rinsed with artificial cerebro-spinal fluid: 125 mM NaCl, 25 mM NaHCO₃, 2.5 mM KCl, 1.25 KH₂PO₄, 1 mM MgCl₂, 2 mM CaCl₂*H₂O and 10 mM glucose. To cover the window and make the surface flat, a 5.0 mm cover glass (Thermo-Scientific, Germany) was fixed by Self-Adhesive Resin Cement ESPE RelyX U200 (3 M, Seefeld, Germany). In addition, the head was rigidly fixed with a custom-made clamp. The rectal body temperature was measured and kept between 36 and 38 °C by a heating plate. The depth of anesthesia was tested by provoking the corneal reflex and reactions to noxious stimuli.

Photographic overview images were captured using an epifluorescence microscope Axio Imager Z2 (Zeiss, Oberkochen, Germany) equipped with a 5x objective. Confocal images were taken using a laser-scanning microscope LSM-710 (Zeiss, Oberkochen, Germany) with appropriate excitation and emission filters. Z-stack images were taken at 0.8–2.0 μm intervals and processed with the ZEN software (Zeiss, Oberkochen, Germany). All data were collected from three randomly selected pictures, from three different slices of at least three mice per group.

For skeletonization, first, images from brain regions were acquired with a confocal microscope. Maximum intensity projections were treated to remove background noise. The images were converted to binary, presented to skeletonization, and analyzed by the ImageJ plugin AnalyzeSkeleton [20]. The parameter end points express the complexity of microglial cell structure. It corresponds to branch ends, determined by voxels with less than two neighbors [21‐23]. To measure the soma size, sequences of 14 stacks were examined with the ZEN Software (Zeiss, Oberkochen, Germany). Each microglial soma was marked with Spline contour tool, which revealed the area in μm². Finally, the total length of each microglia was quantified, with the maximum intensity projections, by using the ImageJ plugin ROI manager.

High resolution in vivo imaging was performed using a custom-made 2P-LSM equipped with fs-pulsed titanium-sapphire laser (Chameleon Ultra II; Coherent, USA). For 2P-recordings, a Zeiss W Plan Apochromat 20× (NA 1.0) water immersion objective was used. Laser excitation was set at 895 ± 5 nm for EGFP and mCherry detection. Emitted light was split by a 520 nm longpass dichroic mirror (Semrock, Rochester, USA) and collected by photo-multiplier tubes (Hamamatsu, Japan) through two bandpass filters: a 494 ± 20.5 nm (FF01-494/41-25) and a 542 ± 25 nm (FF01-542/50-25 (Semrock). In parallel, uniformly spaced (1.5–2.4 μm) planes of 100*100 to 600*600 μm² regions were recorded, digitized and processed to obtain z-stacks of images (256 × 256 to 1024 × 1024 pixels in size). Voxel sizes ranged from 0.2 × 0.2 × 1.5 to 1.17 × 1.17 × 2.4 μm for the xyz-axes. Recordings of, at most, 100 μm stack depth were obtained.

First, our GL261/CX3CR1 model allowed imaging of microglia and glioma cells, easily discriminated by specific fluorescences. Intracortically implanted mCherry-GL261 cells formed clusters of red-fluorescent glioma cells, as shown in Fig. 1. Green-fluorescent microglia interacted with these tumors, as examined at 3, 6 and 18 days post injection (dpi). Microglial cells presented a time- and space-dependent pattern of infiltration into developing tumor areas. In sections examined three days after inoculation, we identified microglial cells next to mCherry-GL261 cells (Fig. 1a, d). The density of microglia close to developing glioma was higher in comparison with other brain regions. Indeed, all cells avoided to cross the tumor borders. However, six days after tumor injection, many activated microglial cells had infiltrated the tumor mass (Fig. 1b, e). Some activated microglia also remained growing around the tumor, in close contact with glioma cells. At 18 dpi, we found an increased number of microglial cells into the enlarged and diffuse mass of glioma cells (Fig. 1c, f).

The mCherry-GL261 cells used in our model mimicked the typical aggressive behavior also reported in patients with glioblastoma multiforme. Tumor cells rapidly grew and infiltrated surrounding tissues. As shown in Fig. 1c, at day 18 the tumor mass compressed the left hippocampus, deforming regions CA1, CA2, CA3, and the dentate gyrus. Implanted mCherry-GL261 cells also developed secondary tumors apart from the main tumor mass, formed by cell migration or metastasis. They were present in the border of left lateral ventricle in the early 3 dpi (Fig. 1a, red arrow). Secondary tumors also appeared at 6 dpi in the margins of the right lateral ventricle (Fig. 1b, red arrows). Finally, a more prominent mass apart from the injection site developed next to hypothalamus at 18 dpi (Fig. 1c, red outlined arrow).

The present GL261/CX3CR1 glioma model also allowed imaging of microglial cells present in tumor areas for analysis of morphology and cell density. Confocal microscopy successfully captured cell fluorescences, organized in stacks of images from fixed brain slices mounted on glass slides. We first focused on cells present in non-inoculated brain regions, like region 1 (Fig. 2a, c, and f). These surveilling microglia showed a ramified morphology, with a small cell body and fine processes. In contrast, cells near the tumor regions assumed an amoeboid shape, a sign of microglial activation. Cell bodies were enlarged and the cellular processes displayed reduced lengths and ramifications. Amoeboid cells were present either on the borders as shown in region 2 (Fig. 2a, d, and g) or infiltrated into the tumor core, region 3 (Fig. 2a, e, and h) at 14 dpi.

Our model also allowed to quantify microglial cell numbers in regions 1, 2, and 3. Both tumor regions (border and core) presented a higher number of microglial cells in comparison with the contralateral brain hemisphere (Fig. 2b, p < 0.05). In addition, cell density was significantly higher in the core region of tumors compared to the borders. These results reinforced the notion that glioma cells recruit microglia and induce its activation.

We examined microglial cells at 14 dpi to quantify parameters of microglial activation. First, we confirmed that EGFP-positive cells of the GL261/CX3CR1 glioma model express Iba1 (Fig. 3a–c). Iba1 is a marker of microglia/macrophages and can also indicate the activated status of microglia. Iba1 expression was higher in tumor regions in comparison with the contralateral hemisphere. The mean intensity values of anti-Iba1 were 611.4 ± 49.0 and 453.0 ± 6.0 in tumor border (Fig. 3e) and tumor core (Fig. 3f), respectively, in comparison with 24.8 ± 2.8 found in non-inoculated sites (Fig. 3d, g; p < 0.05).

To quantify the changes in microglial morphology, we used maximum intensity projections of confocal images (EGFP channel). We first determined the number of endpoints per cell, that represents the branch ends. Microglial cells in the border or the tumor core showed less end points (35.7 ± 3.1 and 26.0 ± 0.5, respectively) in comparison with those in the contralateral control hemisphere (108.0 ± 16.8) (Fig. 3h; p < 0.05). Microglial cells in the tumor core region also showed an enlarged soma size (1310 ± 146.0 μm²) in comparison with those in tumor borders and the contralateral hemisphere (661 ± 37.4 μm² and 366 ± 0.0 μm², respectively) (Fig. 3i; p < 0.05). Finally, the cell perimeter lengths in tumor regions (74.3 ± 7.0 μm in tumor center and 86.7 ± 13.4 μm in the border) were smaller than those found in the contralateral control site (268.0 ± 38.8 μm) (Fig. 3j; p < 0.05).

In summary, our morphometric results showed that microglial cells apart from glioma tissues have a more ramified morphology. They present a higher number of thin processes extending away from their small soma. When growing next to tumor regions, cells reduced the number of processes and enlarged their soma, assuming a typical amoeboid shape. These data suggest that glioma tumors influence microglial cells to become activated and change their morphology.

The time-course of microglial growth in tumor regions was examined by confocal laser scanning microscopy at 3, 6, 9, 12, 14, and 18 dpi (Fig. 4a–r). At 3 dpi, we found activated microglia (in green) surrounding glioma mCherry-GL261 cells without infiltrating the tumor mass (Fig. 4a, b, c). From 6–18 dpi, green fluorescent microglial cells infiltrated the tumor mass (in red), as shown in merge images of Fig. 4f, i, l, o, r. Indeed, cells in close contact with glioma presented a typical activated morphology, with amoeboid shape and pseudopodia (Fig. 4, right column).

Diameter and mean intensity value of glioma cells were determined by using the ImageJ plugin ROI. For each time point, 10 cells (mCherry channel) per image, of three animals were randomly analyzed. mCherry-GL261 tumor cells displayed significant differences in cell morphology during their growth. At 3 dpi, they showed similar sizes and shapes, presenting a uniform and round shaped morphology (Fig. 4b); diameters of cell bodies ranged from 10–20 μm. Tumors increased in size in later stages, and glioma cells assumed clear morphological changes, as occurred at 14 dpi (Fig. 4n). The tumor core contained a population of cells heterogeneous in size and morphology; at this stage, diameters of glioma cell bodies varied from 14–52 μm. Cells also differed in fluorescence intensity as the tumor grows, ranging from 167 at 3 dpi to 626 at 14 dpi. In addition, we found an another cell type in tumor region at 12 dpi. They were polykaryocytes, i.e. multinucleated cells, with a bizarre nuclei structure and enlarged cytoplasm up to 60 μm of diameter (Fig. 4k, l).

The GL261/CX3CR1 model proposed in our study also provided brain tumor images of living animals. A 2P-LSM microscopy captured the fluorescence emitted by microglia and glioma cells during tumorigenesis, allowing an intravital- and noninvasive imaging. First, we could track the early stages of microglial activation. Already 30 min after injection, we noted microglia (Fig. 5a) in close contact with mCherry-GL261 cells (Fig. 5b and c, white triangles showing glioma cells). At 36 h, microglial cells presented different morphologies, as shown in Fig. 5d, squares. Some cells were typically amoeboid (Fig. 5d, orange outlined square); others had a ramified morphology, with small processes (Fig. 5d, white outlined square).

Regarding mCherry-GL261 cells, they infiltrated into the brain parenchyma and, in addition, were present at the margins of blood capillaries (Fig. 5e and f, white arrows). This anatomic location favors the metastatic behavior of glioma cells cited above (see Fig. 1c). At 36 h after injection, microglia remained growing in close contact with tumor cells (Fig. 5g, h and i).