Longitudinal in-vivo OCM imaging of glioblastoma development in the mouse brain

: We present in-vivo imaging of the mouse brain using custom made Gaussian beam optical coherence microscopy (OCM) with 800nm wavelength. We applied new instrumentation to longitudinal imaging of the glioblastoma (GBM) tumor microvasculature in the mouse brain. We have introduced new morphometric biomarkers that enable quantitative analysis of the development of GBM. We conﬁrmed quantitatively an intensive angiogenesis in the tumor area between 3 and 14 days after GBM cells injection conﬁrmed by considerably increased of morphometric parameters. Moreover, the OCM setup revealed heterogeneity and abnormality of newly formed vessels.


Introduction
Almost all of the primary and essential biological functions of developed living organisms (e.g., breathing, heartbeat, sleep, etc.) are controlled by the brain -center of the nervous system. Several millions of people around the world are affected each year due to medical conditions like stroke, tumor, neurological disorders, which can be interrelated affecting the brain and the same on the activity of the entire organism. There is only a limited state of art imaging techniques that can penetrate through skull like computed tomography (CT), positron emission tomography (PET), magnetic resonance imaging (MRI).
In vivo, optical imaging is still an extremely difficult challenge due to the highly scattering properties of the skull and the relatively high optical density of the brain's cortex itself. While the first problem is still unresolved, the maintenance of relatively deep brain penetration above 3 mm and high spatial resolution has been demonstrated by using optical coherence tomography (OCT) [1]. OCT is a non-invasive technique that uses near-infrared light to optimize competing resolution and penetration depth parameters. OCT also provides the advantage of 4-D information on brain dynamics without using exogenous contrasting agents [2]. These advantages make OCT techniques unique and useful for monitoring tumor progression in the brain, as typically, the tumor scattering properties increase during longitudinal progression resulting in a reduced depth of imaging for methods using visible light such as confocal microscopy or fluorescence imaging [3]. Optical coherence microscopy (OCM) circumvents some of the limitations of optical coherence tomography (OCT) with achieving the micron-scale transverse resolution, especially vital in biological tissue imaging [4,5]. In turn, optical coherence angiography (OCTA) is an OCT modality that enables to access function of the sample like visualization of the microvasculature malignant area over time. This study focuses on the quantitative and qualitative analysis of the longitudinal progression of angiogenesis accompanying the growth of GBM tumors. The advantage of our method is to keep the brain functionally undisturbed by using a transcranial window and by keeping the dura intact, giving an insight into the tumor under real conditions [45]. To the best of our knowledge, the presented results are the first in the world to show the monitoring of changes in cerebral microcirculation during the growth of a GBM tumor in the brain up to 2 weeks using OCTA. We quantified the tumor with a set of vascular parameters obtained by fractal analysis. Thanks to this work, we have taken the first step towards biomarkers, which enable quantitative analysis of GBM development. During these studies, intensive angiogenesis was found in the tumor area between 3 and 14 days after GBM cell injection, confirmed by considerably increased morphometric parameters. Moreover, the OCM system showed heterogeneity and abnormalities of newly formed vessels.

Experimental setup, scanning protocol, and data analysis
In this study, we used spectral OCM system based on a Michelson interferometer configuration ( Fig. 1). We used a ∼6fs short pulse laser light along with a 75:25 fiber beam splitter to split the output into the reference and the object arm, respectively. The reflected light from the sample and the reference mirror was coupled using a fiber coupler and directed into a spectrometer. To provide a high axial resolution of 2.2 µm, a femtosecond laser (Fusion Femtolasers, Austria) centered at 795 nm with 130 nm full bandwidth was used as a light source. The interferometric signal was recorded by a custom-made spectrometer with 2048 pixels of high-speed line scan camera (Basler Sprint, Germany). Interchangeable imaging optics can be used with the OCM system; however, for this study, we used a 4x microscope objective (Olympus Plan Fluorite, NA = 0.13, equivalent full aperture = 11.7 mm, effective focal length = 45 mm, maximum transverse resolution at 800nm = 2.5 µm). To compromise between lateral resolution and depth of focus the beam entering the objective was expanded to 6 mm (at 1/e 2 of the Gaussian profile) by achromatic lenses L 3 =30 mm and L 4 =100 mm [ Fig. 1(A)]. The theoretically estimated confocal parameter is ∼100 µm for obtained transverse resolution of ∆x = 4.5 µm (at FWHM of Gaussian profile). The sample was illuminated with 10 mW average optical power. A-scans were collected at 50 kHz repetition rate. Three dimensional volumes covering area of 2.8 × 2.8 x 2 mm 3 were acquired with sampling density of 800 × 800 × 2048 pixels for X, Y and Z axis respectively. For better visualization of the Region of Interest we cropped the volumetric data to 1.7 mm x 1.7 mm x 0.4 mm containing 400 × 400 × 400 pixels. The 2mm measuring range of the OCT was selected to facilitate the positioning of the mouse brain and to be able to react during the measuring session to accidental movements of the animal. At each position along Y-axis, the B-scans were repeated six times with a repetition time of 8.2 ms. Such six repeated frames were used to calculate angiographic cross-sections or averaged to improve the quality of OCT structure images. Full 3D volume was collected within 19.2 s, but total measurement time -including data transfer and saving -was 94s. Dedicated stand-alone software, developed in LabView, allowed the control of a scanning protocol, synchronization of camera and scanners, and data processing. Data analysis enabled the volumetric reconstruction of mouse brain structure as well as angiographical mapping.
The pre-processing included standard steps for OCT systems: spectrum linearization in k-space, numerical dispersion correction, and fast fourier transform. Complex valued data was either averaged (N = 6) to provide high-quality structure information, or processed towards the visualization of brain vasculature. OCT angiographic processing was based on the absolute complex difference (ACD) method in Fig. 1(B) that for mouse brain imaging provides the best angiographic contrast, especially for the smallest capillaries [36]. All 6 B-scans acquired at a given Y location were first corrected for potential bulk motion. Then ACD was calculated as an amplitude of complex difference between consecutive pair of B-scans and further averaged over 5 of such pairs. The procedure was repeated for all Y axis locations and presented either as angiographic B-scans or en-face angiographic projections by taking maximum intensity along Z-axis. We extracted various morphological parameters from OCM en-face angiographic projections to estimate the progress of angiogenesis during tumor growth. The region of interest (ROI), of the same diameter for all time points, was identified for both Glioma and control datasets as a circular region around the injection area. The choice of the ROI was dictated by the shape of the changes introduced by the application of culture medium in the control group. We assumed that the potential protocol for monitoring tumor progression would typically be assisted by the control group to quantify the immune response to the cell injection objectively. Further, we applied a multi-scale Frangi-Hessian filter, which enhances the tube-like structures to extract only vascular information and to segment the vasculature for quantitative analysis [46]. The vessels (sigma) of the filter, which signifies the vessel diameter was adjusted to match the original vasculature size. The AngioTool (National Cancer Institute) and post-processing were made in Python to parameterize and quantitatively access the morphometric and vessel parameters like vessel area, total vessel length, number of junctions, and number of endpoints. The parameters are defined as follows: (i) "Vessel Area" relates to the total area of the segmented vessels.
(ii) "Total Number of Junctions" represents all vessel's bifurcations.
(iii) "Total Vessels Length" describes the total Euclidian distance of all the vessels calculated between pixels in the image.
(iv) "Total Number of End Points" is the total number of only open-ended segments present [47].
These angiometeric parameters estimate the quantitative and qualitative information about the growth of the tumor.

Animals
All procedures performed according to the rules established by the First Local Ethical Committee on Animal Research in Warsaw. These rules are based on national laws in full accordance with the European Union directive on animal experimentation.

Glioblastoma cell preparation
GL-261 cells were obtained from the Division of Cancer Treatment and Diagnosis (DCTD) Tumor Repository, National Cancer Institute (NCI), Frederick, Maryland. After thawing, cells were cultured in DMEM medium, supplemented with 10% Fetal Bovine Serum (FBS), 1% L-glutamine, and 1% penicillin/streptomycin on T75 flask. The cell line was maintained at a constant temperature of 37°C with a CO 2 level of 5%. When the culture reached 80-100% confluence, the cells were trypsinized with 0,05% trypsin and neutralized with culture medium. Before implantations, floating cells were centrifuged, suspended in culture media to appropriate concentration, and maintained on ice for immediate injection.

GL-261 cell injection and cranial window implantation
C57BL/J6 mice (age 6-10 weeks old, weight 20-25 g) were used in the experiments. Two mice were used for GL-261 implantation (M1, M2), and two served as a control (M3, M4). To prepare the mice for OCM imaging, a cranial window was implanted over the parietal-temporal lobe and centered on the barrel cortex of the right hemisphere. The cranial window was located at 2.5 mm posterior to bregma. The mice have been placed in a stereotaxic frame and deeply anesthetized with isoflurane (4% for induction; 1.5-2% for surgery). Dexamethasone (0.2 mg/kg) and Carprofen (5 mg/kg) were administered subcutaneously to prevent an inflammatory response and brain edema. A warming pad with an anal probe containing an automated temperature control feedback system was used for monitoring body temperature (maintained at 37°C level during surgery). After removing the skin over the top of the skull, a drop of 1% lidocaine: epinephrine (1:10 5 ) solution was applied onto the periosteum to avoid excessive bleeding or pain. A circular portion of the skull (diameter 4 mm) was removed to expose the dura mater. The dura mater was incised, and 5 × 10 3 GL-261 cells in 0,3 µl culture medium were injected with Hamilton syringe (Fig. 2, d0) to depth of 300-800 microns into the somatosensory cortex. Control mice were injected with the cell culture medium only to the same depth. Then the brain surface was covered with 0.9% NaCl and a cover glass (diameter 5 mm) was attached with cyanoacrylate-based glue and sealed with dental acrylic. After the surgical procedure, the mice were transferred to cages and monitored until total recovery. Figure 2 demonstrates the glioma cell injection procedure and tumor appearance and growth across 14 days.

In-vivo OCT imaging
OCM measurements were acquired at 1, 3, 11, 14 days from G261 implantation and the injection of the cell culture medium. Each animal is placed in a stereotaxic frame with a warming pad and was deeply anesthetized with isoflurane (4% for induction; 1.5-2% for the procedure). The

Histology
Histological hematoxylin-eosin (HE) staining was performed 14 days after GL-261 cell implantations. The mice were anesthetized with pentobarbital and perfused with warm (37°C) phosphate-buffered saline (PBS) for 5 min, followed by 4% paraformaldehyde in PBS for 15 min (at room temperature). The brains were postfixed overnight in the same fixative solution and impregnated in 30% sucrose for 3-5 days. Next, mouse brains were cut on cryostat into serial coronal sections with 40 µm thickness. The sections were mounted on SuperFrost (Fisher Scientific, Hampton, New Hampshire /USA) slides, air-dried on a slide warmer at 58°C for at least 30 min. Following this, the sections were hydrated in tap water, then stained in Mayer's hematoxylin (Diapath, Martinengo, Italy) for 6 min, and washed in running tap water for 5 min. Next, the slides were stained in Shandon Eosin Y (Thermo Fischer Scientific, Waltham, Massachusetts, USA) for 2 min, washed in tap water for 2 min, dehydrated in baths with increasing alcohol concentrations (1 min in each bath of 50%, 80%, 90% and 100% ethanol), and cleared in xylene. Finally, the sections were mounted in Shandon Consul-Mount media (Thermo Fischer Scientific), a non-aqueous, non-fluorescent plastic mounting media.

Results
To completely confirm the presence of a malignant brain tumor after cell implantation, we evaluated the structure and size of glioma tumor based on histology (HE staining) and compared it to structural OCM reconstructions. After 14 days of glioma growth, histology revealed intense violet staining of the tumor with hematoxylin, clearly darker than the surrounding tissue (Fig. 3). The size of the tumor was approximately 1.8 mm x 2 mm x 3 mm (length, width, and depth, respectively). The tumor occupied the cortex's entire thickness with part of the lateral ventricle [ Fig. 3(b), 3(c)]. Structural OCM B-scans also revealed a characteristic structure of the glioma, visible as a hyporeflective signal area surrounded by the bright scattered signal of neighbor healthy tissue [Figs. 3(e), 3(f)]. Larger vessels surrounding the tumor can be easily identified in the structural OCM images and are seen on the cross-sections as a "butterfly pattern" with a dark shadow signal below [ Fig. 3(g)]. A different pattern of scattering signal was obtained for the control mouse injected with the cell culture medium. Figure 3(d) shows the needle trace surrounded by HE stained healthy tissue, and Fig. 3(h) shows the corresponding structural B-scan.
To follow up on the tumor progression, the OCM measurements were taken over two weeks period at the following time points  Fig. 4(k)] compared to day 11 [ Fig. 4(i)]. Panels m and n in Fig. 4 also shows the OCTA data for two mice that had a tumor. In both cases we observed quite similar neovascularization process which led to the formation of irregular, tortuous and tangled microvasculature morphologically in the area of tumor proliferation. We assessed the diversity of the tumor structure and vasculature at various cortical depths after 14 days of GBM development. In the first superficial layer (0-50 um below the surface), we observed the highest density of newly formed thin meningeal vessels with loops characteristic for tumor [ Fig. 5(a)]. The Field of view (FoV) for the angiomaps in Figs. 5(a)-5(d) is 2.8 × 2.8 × 0.4 mm in XYZ axes. In the deeper layer (50-100 um), there were evidently fewer vessels visible within the tumor, due to increased degeneration in this region. Deepest layers that were analyzed (100-200 um and 200-300 um) reveal maps of rare radially arranged, tortuous vessels. Vascularization abnormality was confirmed on HE stained sections of the glioma [ Fig. 6(b)]. The healthy tissue closely surrounding the glioma area contains much narrower and more numerous vessels [ Fig. 6(a)].
To assess the angiogenesis quantitatively in the tumor region, we performed an analysis of morphological parameters characterizing the vessel network (vessel area, total mumber of  junctions, total vessels Length, total number of endpoints) presented on en-face angiographic projections (Fig. 7). Selected areas [ROI, Region of Interest, Fig. 7(a)] for initial days (1-3) are avascular and analyzed parameters at low levels, similar for glioma and control groups. The diameter of the avascular region in ROI was around ∼0.7 mm. On day 11, we observed a dramatic increase in all of the parameters indicating potent angiogenesis. Further increase of parameters was observed for day 14. All characterization parameters for the control mice remained at low levels over the whole investigation period [ Fig. 7(c)].

Discussion
The specificity of in vivo OCM imaging of mice is associated with some disadvantages. First of all, to obtain a good quality image, it is necessary to immobilize the animal's head, as the breathing movements interfere with the scattering signal. There is also the possibility of unexpected death of the animal due to the rapidly growing tumor and its expansion deep into the brain, therefore it is necessary to monitor the physiological state of the mice throughout the experiment. The applied method of the GL-261 line cell implantation allowed for the precise, repeatable location of the tumor, and placement of the cells into the deep layer of the cortex was necessary to obtain solid tumors. Side effects of the method were tissue and vessel damage in the area of needle insertion, which caused the OCM images to be obscured in the first days of tumor development.
The widely recognized and used method of brain imaging through the cranial window [48,49] has enabled long-term observation of the tumor area; however, it causes astrocytosis and associated inflammation, lasting up to 3 months [45,50] Another study showed that the scattering signal was significantly correlated with the level of astrocytosis in the borderline of the infarction [47].
On the first day, we injected a suspension of glioblastoma cells into the cortex with a Hamilton syringe (needle diameter approximately 1mm), then the brain tissue was covered with glass and glued to the bone. All of these activities cause irritation of brain tissue, including the numerous micro-hematomas shown in Fig. 1 (day 1), and the leakage of a small portion of the glioblastoma cell suspension on the brain surface. It seems to us that all this disturbs the penetration of tissue by the laser beam and disturbs the scattering signal, which results in a lower quality, blurred OCM images on the day 1. Also, the immune response was individually variable. In some cases the inflammation was so high that its effects made it completely impossible to imaging the blood vessels and monitor the tumor progression.

Summary
As a result of OCM measurements, we obtained structural and angiographic images that characterize the tumor structure and vascularization. Glioblastomas are highly vascularized tumors characterized by an abundance of disorganized microvessels with chaotic organization. Highly dense regions neighbor vessel-poor areas, and vessels vary from a normally wide, irregular, and tortuous serpentine-like shape [51] what is in agreement with our observation of rare, radially arranged, tortuous vessels. In both structural B-scans and en-face projections, the tumor boundary was visible due to weaker scatter signal in the tumor area. Similarly, tumor margin was easily detected using cross-polarization OCT (CP OCT) in the rat model [52] and the mouse model using speckle-modulating OCT (SM-OCT) [53]. An attempt to identify the tumor border was made on sections of the human brain after intraoperative resection of glioma [23]. Thus, OCM imaging seems useful in determining the changes in the size of the tumor in the course of its growth. This parameter can be critical to assess the effectiveness of experimental cancer therapy.
Up to today, they are no OCM studies available related to analyzing quantitatively in-vivo neovascularization in glioblastoma development. On the other hand, the advantage of OCT in 3D imaging of sprouting of newly formed vessels in vitro has been shown [54]. Besides, OCT techniques were used in ophthalmology to diagnose neovascularization in macular degeneration [55,56].
Our OCM system is characterized by high resolution (2-3um), which allows on an accurate assessment of cerebral vascularization. It enabled us to show the tumor neovascularization progress visible on the obtained angiomaps. However, the quantitative parameters usually used to assess angiogenesis do not give high specificity due to large individual variability in vascular morphology. In our case, we have taken advantage of the elementary fact that in most of the pharma applications, it is necessary to run a control group in which the region of cell implantation remains changed for a few days after the start of the procedure. Thanks to this, we were able to show that each of them (not very sophisticated) parameters determining the state of blood vessels shows specificity. Therefore, we could confirm the neovascularization process quantitatively based on morphometric features, which demonstrated multiple increases in vessel area, total vessel length, number of junctions, and number of endpoints compared to low constant values in these parameters in control. Additionally, we confirmed the diversity of the vessel structure within the glioma in the mouse brain by histological staining.