A Dorsal Skinfold Window Chamber Tumor Mouse Model for Combined Intravital Microscopy and Magnetic Resonance Imaging in Translational Cancer Research

Abstract

Preclinical intravital imaging such as microscopy and optical coherence tomography have proven to be valuable tools in cancer research for visualizing the tumor microenvironment and its response to therapy. These imaging modalities have micron-scale resolution but have limited use in the clinic due to their shallow penetration depth into tissue. More clinically applicable imaging modalities such as CT, MRI, and PET have much greater penetration depth but have comparatively lower spatial resolution (mm scale).

To translate preclinical intravital imaging findings into the clinic, new methods must be developed to bridge this micro-to-macro resolution gap. Here we describe a dorsal skinfold window chamber tumor mouse model designed to enable preclinical intravital and clinically applicable (CT and MR) imaging in the same animal, and the image analysis platform that links these two disparate visualization methods. Importantly, the described window chamber approach enables the different imaging modalities to be co-registered in 3D using fiducial markers on the window chamber for direct spatial concordance. This model can be used for validation of existing clinical imaging methods, as well as for the development of new ones through direct correlation with "ground truth" high-resolution intravital findings.

Finally, the tumor response to various treatments-chemotherapy, radiotherapy, photodynamic therapy-can be monitored longitudinally with this methodology using preclinical and clinically applicable imaging modalities. The dorsal skinfold window chamber tumor mouse model and imaging platforms described here can thus be used in a variety of cancer research studies, for example, in translating preclinical intravital microscopy findings to more clinically applicable imaging modalities such as CT or MRI.

Introduction

Tumor microvasculature is an important component of the tumor microenvironment that can be a target for therapy and a determinant of treatment response. In the preclinical setting, the microvasculature is typically studied using intravital microscopy in orthotopic or heterotopic window chamber animal models^1,2. This has several advantages over histological studies since the imaging is done in live tissues and the tumor can be monitored longitudinally over several weeks or even months^2,3. These studies can leverage the high-resolution imaging capabilities of intravital microscopy to study the delivery of therapeutics to the tumor^4,5, the causes of treatment resistance⁶, and the response of the micro vessels to therapies such as antiangiogenic treatment^7,8 and radiotherapy^2,9.

Intravital microscopy clearly plays an important role in preclinical cancer research; however, how can tumor microenvironmental features be measured in the clinic? Microvascular information would be useful in the clinic for measuring blood supply and tumor cell hypoxia, which is important for determining treatment resistance in radiotherapy¹⁰, as well as the ability of the microvasculature to deliver chemotherapeutic agents to the surrounding tumor cells¹¹. For example, in radiotherapy, spatial information on the structure and function of the tumor microvasculature may help personalize a patient's treatment plan by adjusting the fractionation schedule or by preferentially boosting the dose to avascular and likely hypoxic regions¹².

Intravital microscopy can measure these important microvascular features since it has a very high resolution (μm scale); however, its depth penetration into tissue is limited to several hundred microns or a few millimeters, at most making clinical implementation challenging. Indeed, there are some novel applications of intravital microscopy in the clinic¹³; however, these are still limited to examinations of near-surface level tissue such as the skin¹⁴ or mucosal/endothelial linings of various body cavities via flexible catheters/endoscopes^15,16.

More commonly, the microvasculature is studied using imaging modalities such as CT¹⁷ or MRI¹⁸. These clinical imaging modalities can image to any depth within the body, but they have a much lower spatial resolution (mm scale). Thus, there is a need to bridge this resolution gap between preclinical intravital microscopy and clinical imaging modalities to bring high-resolution and detailed microvascular information into the clinic¹⁹. Several functional imaging methods have been developed to improve the microvascular imaging capabilities of clinical imaging modalities such as dynamic contrast-enhanced (DCE) MRI and CT²⁰, and Intravoxel incoherent motion (IVIM) MRI²¹. However, these are model-based methods that provide indirect measurements of the microvasculature and thus, must be validated with appropriate "ground truth" measurements of the microvasculature^19,22.

We have developed a dorsal skinfold window chamber (DSFC) tumor mouse model to bridge this gap between preclinical intravital microscopy and clinically applicable imaging modalities such as CT and MRI. The DSFC provides direct access to the tumor for high-resolution, intravital microscopy imaging through a glass window but also clinically applicable imaging such as MRI as it is made of MR-compatible materials (plastic and glass). Furthermore, an included MATLAB code performs multimodality 3D co-registration for direct spatial correlations between preclinical intravital microscopy and clinically applicable imaging modalities. Here we will describe the design and surgery to install the DSFC as well as the procedure to co-register intravital microscopy and clinically applicable imaging modalities.

Protocol

All animal procedures were performed in accordance with the Guide to the Care and Use of Experimental Animals which is set forth by the Canadian Council on Animal Care. Experiments were performed according to a protocol approved by the University Health Network Institutional Animal Care and Use Committee in Toronto, Canada.

1. Tumor inoculation landmarking

NOTE: "Landmarking" refers to the process of marking the skin of the mouse to indicate where the tumor cells should be injected to optimize DSFC placement. This landmarking procedure should be done on the same day or 1 day before the inoculation. The immunocompromised NOD.Cg-Rag1tm1Mom Il2rgtm1Wjl/SzJ (NRG) female mouse was used for this work.

1. Anesthetize the mouse using 5% isoflurane for induction and 2% isoflurane for maintenance (oxygen flow rate set to 0.5 L/min). Maintain the body temperature by placing the mouse on a heating pad wrapped in an autoclaved surgical mat.
2. Prepare the mouse for tumor cell inoculation by shaving followed by the application of medical hair removal cream. Thoroughly remove the hair removal cream after 30-60 s using a wet sterilized paper towel. Apply veterinary eye lubricant to prevent dryness.
3. Disinfect the skin with an alcohol swab.
4. Gently straighten the body of the mouse and lift the skin along the spine of the mouse. Using a surgical marker, draw a single point on one side of the mouse in the center of the tented skin. The point should be located at approximately the center of the thoracic spine of the mouse.
   NOTE: This will be the desired location of the tumor.

2. Tumor inoculation

NOTE: In this study, we are using a human pancreatic cancer cell line (BxPC3). Other cell lines can also be used; however, specific cell culture steps may vary across different cell lines. Refer to the instructions included with cells for modifications to the below procedure.

1. Grow the cells starting 2 weeks prior to inoculation using complete growth media (Roswell Park Memorial Institute 1640 medium with 10% fetal bovine serum and 1% Penicillin/streptomycin) in a 75 mL flask at 37 °C and 5% CO₂. Ensure that the cells are maintained in the exponential growth phase.
   NOTE: Passage number should be restricted to 10-15x and the 75 mL flask should contain approximately 6 million cells (~70% confluence) at the time of inoculation.
2. Aspirate the media and wash the cells with 5 mL of phosphate-buffered saline (PBS) without calcium or magnesium.
3. Add 5 mL of cell disassociation agent and incubate at 37 °C and 5% CO₂ for 6-7 min. Confirm the cells have detached using a microscope and add 5 mL of complete growth media.
   NOTE: Gently tapping the side of the flask can help with detaching the cells.
4. Transfer the suspension to a 15 mL centrifuge tube and centrifuge at 500 × g for 5 min to pellet the cells. Aspirate the media and resuspend the cells in 5 mL of complete growth media.
5. Determine the cell concentration and total cell count using a hemocytometer.
6. Centrifuge at 500 × g for 5 min to pellet cells and aspirate the media.
7. Based on the total cell count obtained from step 2.5, add the appropriate amount of complete growth media to reach a concentration of    where n is the total number of mice to be inoculated. Resuspend the cells.
   NOTE: The target number of cells to be inoculated per mouse is 250,000.
8. Transfer 1.5 mL containing 250,000 cells × n into a 1.5 mL microcentrifuge tube. Centrifuge the 1.5 ml microcentrifuge tube at 500 × g for 5 min and aspirate the media.
   NOTE: Any excess cell suspension can be kept in a hot water bath warmed to 37 °C to be used for additional inoculations if needed.
9. Resuspend the cells in 10 μL × n of PBS. Place the cells on ice to cool the suspension.
   NOTE: Once the cells have been chilled on ice, they must be injected into the mouse within 20 min.
10. Add 10 μL × n of solubilized basement membrane using chilled pipette tips. Load chilled 29-gauge insulin needles with 20 μL each of the cell suspensions. Keep the syringes on ice.
11. Anesthetize the mouse using 5% isoflurane for induction and 2% isoflurane for maintenance (oxygen flow rate set to 0.5 L/min).
12. Place the mouse on a heating pad wrapped in a sterilized surgical mat.
13. Disinfect the skin at the injection site with an alcohol swab.
14. Insert the needle ~1 cm before the landmarked point and move the needle under the skin until it is at the landmark. Position the bevel side of the needle upwards and inject the 20 μL of cell suspension.
15. Wait 45 s to allow the solubilized basement membrane to solidify before removing the syringe.
16. Remove the mouse from anesthesia, wait until the mouse becomes ambulatory, and return it to its cage with the other mice.
17. Monitor the tumor daily by palpation and allow the tumor to grow for 4 weeks or until the tumor is 4-8 mm in diameter. Euthanize the mouse if any of the following conditions are met: tumor size exceeds 1.5 cm, ulcerated tumor, or any signs of systemic illness (lethargic, weight loss exceeding 20% normal body weight, impaired ambulation, inability to regulate body temperature, anorexia, hunched posture, visible signs of pain (facial expressions, etc.) or dehydration).

3. Window chamber surgery

NOTE: The DSFC consists of four 3D-printed parts as shown in Figure 1. Schematics of each part are included in Supplementary File 1. All parts are printed with a biocompatible clear plastic resin. The main window chamber assembly consists of three parts (Figure 1A-C) with an additional fiducial marker ring (Figure 1D) that can be affixed during MRI or CT imaging.

Figure 1: Dorsal skinfold window chamber schematic. The main window chamber contains three parts. First, (A) the front frame is sutured underneath the skin of the mouse and contains a glass coverslip affixed using UV-cured glue. (B) The back frame is sutured to the front frame on the outside of the skin. (C) The support clip affixes to the bottom of the back frame and keeps the DSFC upright on the mouse body. (D) The fiducial marker ring contains seven 'wells' where fiducial markers can be inserted. The fiducial marker ring can be affixed to the front frame of the DSFC using the three support posts. (E) The full DSFC assembly with a fiducial marker ring is shown. Scale bars = 1 cm (A-D, at bottom left; E). Abbreviation: DSFC = dorsal skinfold window chamber. Please click here to view a larger version of this figure.

1. Administer 1 mg/kg sustained-release buprenorphine subcutaneously 1-3 h before surgery. Ensure that the mouse is of sufficient body size and weight to tolerate the DSFC without thoracic constriction.
2. Sterilize the main window chamber parts (Figure 1A-C) in a high-level liquid disinfectant for 12 min followed by soaking in isopropyl alcohol for at least 5 min.
3. Anesthetize the mouse using 5% isoflurane for induction and 2% isoflurane for maintenance (oxygen flow rate set to 0.5 L/min).
4. Transfer the mouse to a sterile surgical mat with a heating pad underneath.
5. Shave the mouse and apply medical hair removal cream if hair is present. After 30-60 s, remove the hair removal cream using sterile wet paper towels.
6. Apply veterinary eye lubricant to prevent dryness. Reapply every 30 min or as needed throughout the surgery.
7. Disinfect the skin by applying 7.5% povidone-iodine scrub and rinsing away with 70% isopropyl alcohol. Allow the skin to air dry.
8. Apply 10% povidone-iodine solution to the skin and allow it to air dry (Figure 2A).
9. Replace the surgical mat with a new sterilized mat.
10. Ensure that the mouse has reached the surgical plane of anesthesia using a toe pinch.
11. Lift the skin of the mouse along the spine and look for the tumor growing on one of the two lateral skin folds. Place the back frame of the DSFC (Figure 2B) on the same side of the skin that the tumor is growing on, making sure to center the tumor in the frame.
12. Add three sutures to each of the top three holes of the DSFC and the surgical guide to fix the position of the back frame (Figure 2B).
13. Insert three needles through the back frame of the DSFC in the holes containing the spacers (Figure 2B).
    NOTE: There are three 'spacers' on the back frame of the DSFC to maintain a gap between the front and back frames to ensure that blood flow is not restricted to the tissue inside the DSFC. The spacers are visible in Figure 1B.
14. Using a surgical marker, mark the six points where the DSFC back frame spacers will be located on both sides of the skin (points 1-6 in Figure 2C,D).
    NOTE: The inserted needles should be used as a guide for the precise location of these points.
15. Draw a 1 cm diameter circle to indicate the skin that will be removed on the side opposite to the back frame (Figure 2D).
16. Remove the needles and the back frame from the skin by cutting the sutures.
17. Cut the 1 cm diameter hole in the skin that was marked in step 3.15 using surgical micro scissors (Figure 2E).
18. Place the temporary suture to secure all parts of the window chamber to the skin. Insert a suture through point 4, through the top hole of the front frame of the DSFC, and then through point 1 on the opposite side of the mouse (Figure 2F and steps 1-2 in Supplementary Figure S1).
19. Thread the suture through the top support post of the back frame and then back through the support post on the back frame that is closest to the head of the mouse (Figure 2F).
20. Insert the suture through point 2. Then, thread the suture through the corresponding hole in the front frame (Figure 2G and step 3 in Supplementary Figure S1).
21. From the inside of the skin, thread the suture through point 6 as indicated in Figure 2H and steps 3-4 in Supplementary Figure S1.
22. Thread the suture through point 5 as indicated in Figure 2I and step 4 in Supplementary Figure S1.
23. Bring the suture through the front frame, point 3, and then through the back frame as indicated in Figure 2J and step 5 in Supplementary Figure S1.
24. Bring the suture back through point 1, the front frame of the DSFC, and then out through point 4 as indicated in Figure 2K and step 6 in Supplementary Figure S1.
25. Tighten the entire assembly with this suture and slip the front frame underneath the skin through the hole that was created in step 3.17 (Figure 2L,M).
26. Tie the two ends of this suture together and cut any excess string.
27. Perform permanent sutures through the holes around the perimeter of the front and back frames of the DSFC. Suture the two frames together in the pattern shown in Figure 2N and Supplementary Figure S2.
28. Cut and remove the temporary suture.
29. Affix the support clip to the back frame of the DSFC by sliding it into the protrusion from the back frame (Figure 2O).
    NOTE: An image of the mouse 2 weeks post surgery is shown in Figure 2P,Q. The support clip is used to keep the DSFC upright on the mouse to reduce skin tension and discomfort.
30. Administer 5 mg/kg body weight Meloxicam, a non-steroidal anti-inflammatory drug, subcutaneously for pain and inflammation reduction.
31. Remove the mouse from anesthesia, wait until the mouse becomes ambulatory, and return it to its cage.
32. Check on the mouse 2-3 h after surgery and then daily for at least 1 week. Euthanize the mouse if any of the conditions listed in step 2.17 are met.

Figure 2: DSFC surgery procedure. (A) The mouse is prepared for surgery by removing the hair and disinfecting the skin. The subcutaneous tumor is indicated by the arrow. (B) The back frame is placed in the appropriate position and secured by three syringes as well as temporary sutures affixed to the black surgical guide. (C,D) The spacer locations (points 1-6) and hole are marked on both sides of the skin. (E) The skin is removed. (F-K) A temporary suture is threaded through the two layers of skin, front, and back frames of the DSFC to secure all the parts together. (L,M) The temporary suture is tightened, and the front frame is inserted underneath the skin. (N) Eight permanent sutures are placed to secure the DSFC. (O) Finally, the temporary suture is removed, and the support clip is attached. (P,Q) The same mouse is shown 2 weeks after surgery from both sides. Abbreviation: DSFC = dorsal skinfold window chamber. Please click here to view a larger version of this figure.

4. Optical imaging

1. Allow the mouse to heal and the inflammation to decrease for at least 5 days after surgery before imaging.
2. Anesthetize the mouse using 5% isoflurane for induction and 2% isoflurane for maintenance (oxygen flow rate set to 0.5 L/min).
3. Apply veterinary eye lubricant to prevent dryness. Reapply every 30 min or as needed.
4. Secure the mouse in an imaging stage with a gas anesthesia attachment as shown in Supplementary Figure S3.
5. Obtain a wide field of view (>1.5 cm width) brightfield microscopy image. Ensure that the fiducial marker divots on the front frame of the DSFC are visible.
   NOTE: There are seven divots around the perimeter of the glass on the front frame that align with the seven fiducial marker wells on the fiducial marker ring attachment. These divots are visible in Figure 1A.
6. On the same day, obtain a microvascular image using the intravital microscopy modality of choice. Use the same imaging stage from step 4.4 (Supplementary Figure S3) for this.
   NOTE: We use speckle variance optical coherence tomography (svOCT) for high-resolution 3D microvascular images.

5. Magnetic resonance imaging

1. Place the mouse cage under a heat lamp to warm the mice for approximately 15 min before anesthetizing the mouse.
   NOTE: Warming promotes vasodilation, which will aid in tail vein catheter placement.
2. Anesthetize the mouse using 5% isoflurane for induction and 2% isoflurane for maintenance (oxygen flow rate set to 0.5 L/min).
3. Record the weight of the mouse using an electronic scale for correct drug dosing.
   NOTE: For accurate body weight measurements, make sure to subtract the weight of the DSFC assembly (0.83 g).
4. Place the mouse on the MRI bed and apply veterinary eye lubricant to prevent dryness.
5. Maintain the body temperature of the mouse using a water heater and pump system.
6. Position a respiratory monitoring pillow underneath the diaphragm of the mouse and maintain at 30 ± 5 breaths/min.
7. If using a contrast agent, insert a 27 G butterfly needle into the tail vein with microtubing attached (30 μL of dead volume prefilled with 1% herapin-saline solution). Secure the needle and microtubing to the MRI bed with surgical tape.
8. Secure the window chamber in a 3D printed 'immobilization device' as shown in Figure 3.
9. Using an 18 G needle, inject veterinary eye lubricant into the seven tubes on the MRI fiducial marker.
10. Affix the fiducial marker to the DSFC by aligning the three square connectors to the three posts protruding from the front frame (Figure 3).
11. Connect the tail vein catheter to the drug delivery line and syringe in the automatic pump. Insert the bed into a 7-tesla preclinical MRI scanner.
12. Acquire coronal and axial T2-weighted (T2w) image sets to visualize the plane of the window chamber (echo time TE = 25 ms; repetition time TR = 2,500 ms; 40 x 40 mm field-of-view with a 64 x 64 matrix for 0.5 x 0.5 mm in-plane resolution; 0.5 mm slice thickness; 25 s). Prescribe a sagittal T2-weighted set (TE = 25 ms; TR = 2,500 ms; 32 x 32 mm field-of-view with 128 x 128 matrix for 0.25 x 0.25 mm in-plane resolution; at least 11 imaging slices; 0.5 mm slice thickness; 87 s), which is then rotated into the plane of the DSFC and fiducial markers, based on the coronal and axial views, as shown in Figure 4. Iteratively re-orient and position the sagittal set until the imaging slices are fully aligned so that slice 5 fully contains tissue signal in the DSFC and slice 6 contains no tissue signal in the DSFC.
13. Perform microvascular imaging using the microvascular MRI method of choice.
    NOTE: For the microvascular MR acquisitions, the fiducial marker slices do not need to be imaged since the images are acquired in the same imaging frame-of-reference as the T2w registration images.
    1. Acquire all acquisitions detailed in steps 5.12 and 5.13 over a 32 x 32 mm field-of-view with a 64 x 64 matrix for 0.5 x 0.5 mm in-plane resolution. For all MRI acquisitions, use consistent RF pulses to improve through-plane geometric consistency (sinc excitation; sinc3 refocusing; 2,484 Hz bandwidth).
14. For DCE imaging:
    1. If measurements of gadolinium concentration are desired, obtain T1 maps using 2D-RARE images acquired at variable repetition times (TE = 7 ms; RARE factor = 2; TR = 350, 500, 750, 1,000, 1,500, 2,500, and 4,000 ms; 8 min 28 s).
    2. Perform time-series imaging using 2D RARE images (TE = 8.1 ms; RARE factor = 2; TR = 200 ms; flip angle = 90°; temporal resolution = 12.8 s; 188 repetitions; total monitoring time = 40 min 6 s).
    3. Inject 0.75 mmol/kg body weight Gadobutrol over 10 s via the tail vein after completion of five image repetitions using an automated MR-compatible syringe pump.
15. For IVIM MRI:
    1. Perform diffusion-weighted imaging with the following B-values: 0, 20, 40, 60, 80, 100, 150, 200, 400, 600, 800, 1,000 s/mm² with three averages of the B = 0 s/mm² and isotropic sampling (TE = 16 ms; TR = 800 ms; gradient duration = 2.2 ms; gradient separation = 9 ms; 61 min).
       NOTE: Diffusion-weighted imaging is acquired using a 2D Fourier transform readout, rather than with distortion-prone echo-planar imaging, to ensure geometric consistency of the tumor and surrounding tissue signals within the DSFC across the image sets and with the intravital microscopy images.

Figure 3: DSFC MR imaging setup. (A) Side and (B) top views of the mouse positioned on the MRI bed with DSFC secured and immobilized. The mouse has a tail vein catheter for contrast agent injection and the fiducial maker ring is affixed to the front frame of the DSFC. Abbreviations: DSFC = dorsal skinfold window chamber; MR = magnetic resonance imaging. Please click here to view a larger version of this figure.

Figure 4: MRI slice locations with respect to fiducial markers and window chamber. (A) A diagram of the DSFC with fiducial marker ring attachment with the 11 overlaid MRI slices. Several T2-weighted images must be acquired to ensure that the slices are correctly aligned with the DSFC and tissue. (B,C) Correct positioning of the 11 slices with respect to the tissue in the DSFC from different orientations. (D) Slice 5 is the most superficial slice where intermodality correlation analysis will be performed. (E) Slice 6 contains no tissue signal indicating that it is properly aligned with the DSFC. (F) Finally, the 7 fiducial markers are clearly visible in slice 9. Scale bars = 5 mm. An 'X' on the axis indicates that the axis is going into the page and a circle indicates that the axis is coming out of the page. Abbreviations: DSFC = dorsal skinfold window chamber; MRI = magnetic resonance imaging. Please click here to view a larger version of this figure.

6. MRI to intravital microscopy co-registration

1. In MATLAB, open the Multimodal_Image_Register.m file contained in Supplementary File 2.
2. To the workspace, load in the microvascular image (Figure 5A), brightfield microscopy image (Figure 5B), and microvascular MRI data (IVIM and/or DCE MRI parameter maps).
3. Click the Run button.
4. Using the pop-up file finder, navigate to the file containing the T2w MRI slices.
5. Select up to four T2w MRI slices that clearly display the fiducial markers (Figure 5C).
6. A user interface showing the microscopy image (Figure 5B) and the depth-averaged T2w MRI slices (Figure 5C) will be displayed. Place a point on each of the seven fiducial markers in the MRI image and match them to their corresponding point on the microscopy image (fiducial marker divots around the perimeter of the glass on the front frame of the DSFC) as indicated by the green points in Figure 5B,C.
7. Close the window.
8. A figure containing the overlaid MRI and microscopy images will appear to help with assessing the quality of the registration between the two datasets. If the registration is adequate, continue the code by typing y, then pressing enter in the command window. Otherwise, retry this step by typing n in the command window and pressing enter.
   NOTE: Successful co-registration of the MRI dataset with the microscopy image means that the seven bright fiducial markers from the MRI dataset are centered and fully contained within their corresponding divots on the front frame of the DSFC.
9. A user interface showing the microscopy image (Figure 5B) and the svOCT microvascular imaging dataset (Figure 5A) will be displayed. Select at least three microvascular landmark points on the svOCT microvascular image and the corresponding points in the microscopy image as indicated by the blue points in Figure 5A,B.
10. Close the window.
11. A figure containing the overlaid svOCT and microscopy images will appear to help with assessing the quality of the co-registration between the two datasets. If the registration is adequate, continue the code by typing y, then pressing enter in the command window. Otherwise, retry this step by typing n in the command window and pressing enter.
    NOTE: Successful co-registration of the svOCT dataset with the microscopy image means that the vessels from both datasets are perfectly overlaid with each other.
12. Close the window.
13. The co-registered microcopy image will then appear. Contour the tumor in this image.
14. Close the window.
15. Several figures will then be displayed: the svOCT microvascular dataset (Figure 5D), the co-registered microscopy image (Figure 5E), and the co-registered MRI parameter maps (Figure 5F). Save the maps for later analysis.
    NOTE: The displayed MRI maps are restricted to the user-drawn tumor contour (Figure 5F).

Figure 5: Multimodal point-based co-registration. (A) Color depth-encoded microvascular svOCT dataset; scale bar = 1 mm. (B) Brightfield microscopy image of the window chamber; scale bar = 2 mm. (C) Average of T2w MRI slices 8-11 showing the seven fiducial markers contained in the fiducial marker ring; scale bar = 5 mm. (C) First, the 'moving' T2w MRI dataset is co-registered to the 'fixed' brightfield microscopy image using the user-inputted green markers on both image sets. Next, the 'moving' brightfield microscopy image and co-registered MRI image are co-registered to the 'fixed svOCT dataset' using the blue markers in A and B. The final co-registered dataset includes the (D) svOCT, (E) brightfield microscopy image, and (F) functional MRI parameter map. The black voxels in F are outside of the tumor and are therefore not considered in the analysis. For D-F, scale bar = 1 mm. Abbreviations: svOCT = speckle variance optical coherence tomography; MRI = magnetic resonance imaging. Please click here to view a larger version of this figure.

Representative Results

Speckle variance optical coherence tomography (svOCT) was performed to obtain large field-of-view (FOV) 3D microvascular images (6 x 6 mm² lateral x 1 mm depth). To obtain these images, a previously described swept source OCT system based on a quadrature interferometer was used²³. OCT images were acquired by stitching together two laterally adjacent 3 x 6 mm² FOV scans. Each B-scan consisted of 400 A-scans and was performed 24x per location (25 ms apart) to enable accurate speckle variance processing as described in our previous work²⁴.

To co-register the MRI dataset to the microvascular svOCT dataset several steps were performed. The first step was done in step 5.12 and is illustrated in Figure 4. This step co-registers the two datasets along the depth direction (z-axis in Figure 4). To co-register the datasets along the lateral dimensions (xy-plane in Figure 4), the fiducial marker ring along with a series of two point-based affine registrations are used. First, the MRI dataset is co-registered to the brightfield microscopy image by matching the fiducial marker locations in the ring, with the corresponding divots around the perimeter of the glass window on the front frame (green points in Figure 5B,C). Next, vascular landmarks that are identifiable in both the svOCT microvascular dataset and the brightfield microscopy image are used to co-register the brightfield image (and corresponding MRI dataset) to the svOCT dataset (blue points in Figure 5A,B). Therefore, the brightfield microscopy image acts as an intermediate-resolution image that contains features from both the MRI (divot locations for fiducial markers) and svOCT microvascular image (microvascular landmarks) to enable multimodality co-registration across vastly different imaging resolutions.

Images of the successful co-registration between all three modalities (microvascular imaging, widefield microscopy, and MRI) for a window chamber preparation mouse with no tumor (healthy bare skin) and tumor are shown in Figure 6A-C and Figure 6D-F, respectively. Once all three modalities are co-registered, correlation analysis can be performed. To take advantage of the co-registered datasets, spatial correlations should be performed either on a voxel-by-voxel basis or by using appropriate averaging techniques over selected volumes of interest. Figure 7 shows some representative results of the correlation between svOCT-derived microvascular metrics and functional MRI-derived macrovascular metrics. These plots include the data from two tumor-bearing mice and one 'healthy' bare skin mouse. These plots were generated using a 1 mm² lateral x 500 μm depth sliding window volume of interest (VOI) technique with a 500 μm lateral step size with each point representing one position of the VOI. For the healthy mouse, vascular analysis was restricted to a 3 x 3 mm² region of interest in the center of the FOV to match the 3D spatial extent of the examined tumors. A more detailed study involving svOCT to MRI correlations using this analysis method can be found in our recent work²².

Figure 6: Co-registered healthy and tumor imaging datasets. The top row shows the (A) color depth-encoded svOCT, (B) co-registered brightfield microscopy image with analysis region shown by the blue dotted line, and (C) co-registered DCE-MRI parameter map for a healthy bare skin mouse. The bottom row shows the (D) color depth-encoded svOCT, (E) co-registered brightfield microscopy image with tumor contour shown by the blue dotted line, and (F) co-registered DCE-MRI parameter map for a tumor-bearing mouse. The black voxels in C and F are outside the analysis region and tumor, respectively, and are, therefore, not considered. Scale bars = 1 mm. Abbreviations: svOCT = speckle variance optical coherence tomography; DCE-MRI = dynamic contrast-enhanced magnetic resonance imaging. Please click here to view a larger version of this figure.

Figure 7A shows MRI's k_trans parameter versus the svOCT-measured vascular volume fraction (VVF). VVF is the proportion of tissue that is occupied by vessels and k_trans is the rate constant for gadolinium moving from the intravascular space to the extravascular space. k_trans is obtained by fitting of the Toft's model to the gadolinium time concentration curves on a voxel-by-voxel basis^25,26. VVF measurements from svOCT show that tumors have a smaller VVF than healthy tissue, potentially indicating the presence of poorly vascularized regions inside the tumor. As k_trans is dependent on blood flow, vascular permeability, and capillary surface area^25,26, a positive correlation between k_trans and VVF is expected since a larger VVF implies a larger vascular surface area and blood flow. However, more studies such as Dextran extravasation experiments are needed to further elucidate the impact of tumor vessel permeability in this relationship²⁷.

Figure 7B shows the correlation between MRI's time to peak contrast enhancement (TTP) versus svOCT's mean distance to the nearest vessel (). TTP is a semi-quantitative metric that measures the time required for the gadolinium time concentration curve to reach its maximum value and this is calculated on a voxel-by-voxel basis²⁶. The is a svOCT microvascular metric that measures the average distance to the nearest vessel within the analyzed volume²². In general, there is a smaller in healthy tissue compared to tumor tissue, indicating that the healthy microvasculature is more densely packed leading to more uniformly oxygenated tissue, and the tumor tissue is less well vascularized²⁸. The higher values in tumor tissue may indicate the presence of avascular tissue regions. Figure 7B shows a positive correlation between TTP and . This demonstrates that healthy tissue is well vascularized with gadolinium washing into and out of the tissue efficiently and quickly. However, tumor microvasculature is more malformed with larger intervascular distances, which increases the time for gadolinium to reach the tissue thus resulting in an increase in TTP.

Figure 7: MRI to svOCT macro to micro-vascular spatial correlations. These plots contain the findings from two tumor-bearing mice and one healthy bare-skin mouse using a x 1 x 1 mm² lateral × 500 μm depth, sliding window VOI with a 500 μm step size, with each point representing the average values of the given metric for one position of the sliding window VOI. Abbreviations: svOCT = speckle variance optical coherence tomography; DCE-MRI = dynamic contrast-enhanced magnetic resonance imaging; VOI = volume of interest. Please click here to view a larger version of this figure.

Supplementary File 1: 3D STL files of window chamber components. A zip file containing the 4 components of the DSFC chamber (front frame, back frame, support clip, and fiducial marker ring), which was used to 3D print the parts. Abbreviation: DSFC = dorsal skinfold window chamber. Please click here to download this File.

Supplementary File 2: MATLAB script for multimodal image co-registration. This MATLAB script (Multimodal_Image_Register.m) takes in the brightfield microscopy, microvascular image data, and structural and functional (DCE and/or IVIM parameter maps) MR images and outputs all datasets co-registered to the microvascular dataset. The MATLAB script is written to support the following file formats: MRI - .dcm, svOCT - .mat, and microscopy - .jpg, or .TIF. The MATLAB code is written to accommodate the file sizes/resolutions that are described in the protocol section: MRI - 32 x 32 mm field-of-view with 64 x 64 matrix for 0.5 x 0.5 mm in-plane resolution; 10 contiguous imaging slices, microscopy - 2D image with >1.5 cm² field of view (any resolution acceptable), and svOCT - 6 x 6 mm field of view 7.5 µm resolution. The MATLAB code should be modified if different file formats, image resolution, or imaging fields of view are used. Abbreviations: DCE = dynamic contrast-enhanced; IVIM = intravoxel incoherent motion; MRI = magnetic resonance imaging; svOCT = speckle variance optical coherence tomography. Please click here to download this File.

Supplementary Figure S1: Temporary sutures to affix window chamber to the skin of the mouse. The suture starts at (1) and goes through the front side skin of the mouse and through the front frame of the window chamber. (2) The suture then goes through the back skin of the mouse and out the back frame of the window chamber. The suture is then threaded back through the back frame, (3) into the backside skin, through the front frame, and (4) back out the front skin. (5) The suture is then brought back through both DSFC frames and skin layers. (6) Finally, the suture is brought back through the top hole of the window chamber and all layers of the skin. The DSFC + skin assembly is then brought together by tightening and knotting the sutures at (1). Abbreviation: DSFC = dorsal skinfold window chamber. Please click here to download this File.

Supplementary Figure S2: Optimal placement of permanent sutures. The permanent sutures should be threaded through the front skin/frame out of the back skin/frame, back through the entire assembly, and knotted in the front. Each number represents one individual suture. The sutures should be placed in the order of the numbers to evenly apply pressure to the entire assembly as it is secured. Please click here to download this File.

Supplementary Figure S3: 3D printed stage for intravital microscopy imaging. (A) A 3D printed stage with isoflurane gas anesthesia attachment is used for intravital microscopy imaging. (B) The window chamber is secured to the stage by three bolts to immobilize it and eliminate any breathing motion. (C) Two heating elements connected in series to a DC power source are glued to the underside of the stage to maintain the mouse's body temperature while anesthetized. Please click here to download this File.

Discussion

In this work, we have developed a workflow to perform both intravital microscopy and clinically applicable imaging (CT, MRI, and PET) in the same animal. This was done with the goal of translating preclinical microscopy findings to the clinic by direct correlation of intravital microscopy with clinical imaging modalities such as MRI. Although conventional DSFC designs are made of metal^2,3, we have adapted the DSFC to be MR-compatible by using 3D-printed window chambers made of a biocompatible plastic resin. The registration of the modalities was accomplished by the use of fiducial markers for the clinical imaging modalities and by obtaining a widefield microscopy image to match the fiducial marker locations on the DSFC with the vascular landmarks that can be seen on the widefield microcopy image as well as the high-resolution intravital microscopy microvascular images. A MATLAB code to co-register the modalities was developed and is included in Supplementary File 2.

DSFCs are not new in preclinical research^{2,3,27,29,30,31,32} with some studies developing MR-compatible DSFCs out of biocompatible plastics for intravital microscopy and clinically applicable imaging^{27,29,30,31,32}. However, these studies perform whole-tumor level correlations with MRI and intravital microscopy findings. The novelty of this work is our ability to directly co-register intravital microcopy images to clinically applicable imaging modalities using a fiducial marker attachment to the DSFC enabling direct spatial correlation analysis. Gaustad et al. have also performed spatial correlation analysis between MRI and intravital microscopy; however, specific detailed steps on the co-registration process between modalities are not included²⁹. This work provides a protocol for spatially correlating multiple modalities with each other using the fiducial marker ring attachment and includes a MATLAB pipeline.

There are several details of the DSFC design and procedure that offer improved imaging and tumor growth characteristics compared to other similar studies. First, most window chamber methods involve inoculation of the tumor after window chamber surgery. This is typically done since it is easier to center the tumor in the window chamber FOV; however, previous work has shown that this is suboptimal, leading to an 'unrealistic' mono-layer tumor growth with spontaneous regression in many instances³³. Instead, we allow the tumor to grow unrestricted in the skin for several weeks before DSFC implantation, which allows for a more realistic 3D tumor with improved growth characteristics³³.

This DSFC design is also extremely lightweight at 0.83 g with a large imaging window FOV (1.2 cm diameter glass coverslip). A unique characteristic of our DSFC design is that it does not include a snap ring to secure the glass coverslip to the front frame. Instead, the glass coverslip is permanently affixed using UV-cured glue allowing the bottom portion of the front frame to be completely flat. Furthermore, the front frame is secured underneath the skin of the mouse. This was done to minimize the chance of any air gap between the tumor-glass interface. It has been noted in previous designs that if there is a gap at the tumor-glass interface there is a high probability of fluid buildup in this region. This significantly impacts results by attenuating the intravital microscopy signal thus reducing image quality. Additionally, it makes it more difficult to target the MRI slices to the tumor since this fluid buildup will appear bright on the T2w localization scans. This makes it challenging to identify which slices are imaging the tissue and which slices are imaging the fluid buildup in the DSFC. One drawback to this design is that if the glass coverslip breaks, it cannot be easily replaced. We have also designed a front frame that is made entirely out of an optically clear plastic thus eliminating the glass breakage issue³⁴.

Throughout this procedure, several crucial steps are more technically challenging than others. One step that may cause difficulties is during the DSFC surgery when suturing the front frame to the skin (protocol steps 3.18-3.24). It is important to keep some extra loose skin along the top of the front frame of the DSFC. If this is not done, then the skin in this region will lose blood supply and become necrotic, thus ruining the preparation. Another technically challenging task is the targeting of the MRI slice package to include the most superficial part of tissue underneath the glass. If this slice is not targeted precisely to the region, then partial volume effects with the glass may occur, thus leading to erroneous MRI results. Uneven surfaces of the tissue underneath the glass of the DSFC may make this even more challenging. It is important to ensure that the skin is taut inside the DSFC to keep the tissue surface as flat and uniform as possible to reduce this source of error.

Here we have demonstrated the use of this DSFC workflow using svOCT, which has a large imaging FOV of 6 x 6 mm². This large FOV is not common in other types of intravital microscopy with many imaging methods having FOVs of 1 mm² or smaller¹³. This may cause some challenges when co-registering the widefield microscopy image to the high-resolution intravital microscopy images since microvascular landmarks may be difficult to identify. To overcome this, another widefield microscopy image may be required with a smaller FOV (~6 x 6 mm²) to act as an additional intermediate co-registration step between the imaging modalities to further refine the region of interest to the one imaged by intravital microscopy.

In this work, we have focused on imaging the tumor microvasculature, but other tumor microenvironmental components may be imaged using this model as well. For example, intravital microscopy has been shown to successfully image the tumor stroma³⁵, and diffusion-weighted MRI methods such as diffusion tensor imaging or diffusion kurtosis imaging may have the ability to provide information on the structure of the tumor stroma^36,37. This workflow may provide exciting opportunities to translate stromal tumor information into the clinic by directly co-registering high-resolution intravital microscopy images of the tumor stroma with MRI-based measurements of the stroma. However, it should be noted that because this method uses vascular landmarks for co-registration, those should also be present in intravital microscopy stromal images.

The presented workflow may prove to be an important step in bridging the gap between preclinical intravital microscopy and clinically applicable imaging modalities. This model can be used to validate clinical imaging methods with ground truth intravital microscopy images, develop new clinically applicable imaging methods through comparison with intravital microscopy, and longitudinally monitor the response of treatments using multiple imaging modalities.

Materials

Name	Company	Catalog Number	Comments
Cell Culture Materials
BxPC-3 Human Pancreatic Cancer Cells	ATCC (American Type Culture Collection)	CRL-1687
Corning Matrigel Basement Membrane Matrix, LDEV-free, 10 mL	Corning	354234
Corning Stripettor Ultra Pipet Controller	Corning	07-202-350
Dulbecco Phospphate buffered saline without Calcium, Magnesium, or phenol red, 500 mL	Gibco	14190144
Fetal Bovine Serum (Canada), 500 mL	Sigma-Aldrich	F1051-500ML
Penicillin-Streptomycin 100x (liquid,stabilized, sterile-filtered, cell culture tested)	Sigma-Aldrich	P4333-100ML
RPMI Medium 1640 (1x), liquid; with L-Glutamine, 500 mL	Gibco	11875093
TrypLE Express Enzyme, 500 mL	Gibco	12605028
Window Chamber Materials
12 mm Glass Coverslip	Harvard Apparatus	CS-12R No. 1.5
Connex 500 3D Printer	Stratasys	N/A
Biocompatible clear MED610 resin	Stratasys	RGD810
Loctite AA 3105 UV curable glue	Loctite	LCT1214249
Window chamber back frame	Trimech Inc	N/A
Window chamber fiducial marker	Trimech Inc	N/A
Window Chamber front frame	Trimech Inc	N/A
Window chamber support clip	Trimech Inc	N/A
inoculation and Surgery Materials
BD SafetyGlide Insulin Syringes with Permanently Attached Needles, 0.5 mL, 29 G x 1/2"	BD	CABD305932
Betadine Solution	Betadine	AP-B002C2R98U
Cidex OPA 14 Day Solution 3.8 L	ASP	JOH20394
Disposable Surgical Underpads 23 inch x 24 inch	Kendall	7134
Eye lubricant	Optixcare	50-218-8442
Hair removal cream	Nair	‎061700222611
Halstead Hemostatic Forceps	Almedic	7742-A12-150
Heating pad	Sunbeam	B086MCN59R
Iris Scissors	Almedic	7601-A8-690
Isoflurane	Sigma	792632
Metacam	Boehringer Ingelheim Animal Health USA Inc	NDC 0010-6015-03
NOD.Cg-Rag1tm1Mom Il2rgtm1Wjl/SzJ mouse	the Jackson laboratory	7799
Peanut Clipper & Trimmer	Wahl	8655-200
SOFSILK Nonabsorbable Surgical Suture #5-0 with 3/8" Taper point needle (17 mm) (Wax Coated,Braided Black Silk, Sterile)	Syneture	VS880
Splinter Forceps	Almedic	7725-A10-634
MR Imaging
3D printed window chamber immobilization device.			custom 3D printed, refer to figure 3 for details.
Convection heating device	3M Bair Hugger	70200791401
Drug injection system	Harvard Apparatus	PY2 70-2131	PHD 22/2200 MRI compatible Syringe Pump
Gadovist 1.0	Bayer	2241089
Respiratory monitoring system	SAII	Model 1030	MR-compatible monitoring and gating system for small animals.
Tail vein catheter (27 G 0.5" )	Terumo Medical Corp	15253
Optical Imaging
3D printed imaging stage			Custom 3D printed, refer to supplementary figure 3 for details.
12 V 7 W Flexible Polyimide Heater Plate Thin Adhesive PI Heating Film 25 mm x 50 mm	BANRIA	B09X16XCVS	Heating element used for mouse body temeprature regulation.
DC power supply	BK Precission	1761	Used to power the heating element.
Leica MZ FLIII	Leica Microsystems	15209
svOCT imaging system			In-house made imaging system. Details can be found in reference 23.
Software
MATLAB Software	MathWorks	R2020A