In vivo wide-area cellular imaging by side-view endomicroscopy

Abstract

In vivo imaging of small animals offers several possibilities for studying normal and disease biology, but visualizing organs with single-cell resolution is challenging. We describe rotational side-view confocal endomicroscopy, which enables cellular imaging of gastrointestinal and respiratory tracts and may be extended to imaging organ parenchyma such as cerebral cortex in mice. We monitored cell infiltration, vascular changes and tumor progression during inflammation and tumorigenesis in colon over several months.

The gastrointestinal tract and respiratory airways are major sites of immunological challenge. The interplay among microorganisms, the epithelial barrier, immunity, and genetics is critical to control organismal homeostasis. Impairment of one or some of these factors can lead to homeostatic imbalance, causing disease such as inflammatory bowel diseases¹, diet problems, infectious lung diseases², and cancer. To investigate the complex mucosal immune system and diseases related to it, small animal models, particularly mice, have been widely used. Animal studies have primarily relied on histological examinations of excised tissues ex vivo. Although well established, this approach provides only static information at a specific time point and therefore is inadequate for investigating dynamic longitudinal events involved, for example, in host-microbial interactions, immune reactions and tumor development. Real-time intravital fluorescence microscopy could be a powerful technique for visualizing such processes in natural environments^3,4. Until now, however, in vivo cellular imaging of the mucosa in small animals has been difficult due to the lack of a non-invasive endoscopic method with high resolution and easy maneuverability.

Recently, significant effort has been made to realize high-resolution minimally invasive endoscopy in mice. Laser-scanning confocal endomicroscopy, based on a resonantly vibrating fiber or a fiber bundle, has shown a potential for cellular examination of the colon^5,6. The front-view configuration of the instruments requires direct contact of the probe perpendicular to the intestinal wall. Although such a contact probe would be viable in a human patient⁵, it has proven very difficult to maneuver in small animals, such as mice, due to their small lumen diameters. Non-contact endoscopes have been developed to provide a fish-eye view similar to conventional clinical colonoscopy^7,8. But this approach requires a large depth of field for a given limited aperture size and, therefore, microscopic resolution could not be achieved. Microendoscopy using graded-index (GRIN) lenses has been demonstrated for imaging brain neural circuitry⁹ and muscle kinetics¹⁰ in mice. However, the field of view (FOV) of such a high-resolution probe is typically only 5% of its cross-sectional area, seriously limiting the size of tissue interrogated at a given insertion site.

Here, we describe a new approach based on a side-view microprobe that overcomes the limitations of current endoscopes and enables wide-area cellular-level fluorescence imaging of tissue in live mice. Contact between the view window and luminal wall makes it easy to navigate along the tract by rotation and translation of the probe. This allowed us to obtain a comprehensive map of fluorescently labeled cells and microvasculature in the mucosa in vivo at multiple time points. We demonstrate the new possibilities enabled by this technology in mouse models of colitis and colorectal tumor.

To fabricate a high-resolution side-view endoscopic probe, we modified a 1-mm-diameter triplet GRIN lens microendoscope⁹ and attached an aluminum-coated right-angle prism at the distal end (Fig. 1a). The optical unit was then sealed in a stainless protection sleeve and transparent epoxy window. The assembled rigid endoscope has an outer diameter of 1.25 mm and a length of 50 mm. We integrated the endoscope into a custom-built video-rate scanning-laser confocal microscope¹¹. The laser beam is raster scanned over a fixed x-y plane at the proximal end, so we used a simple image rotation to convert the x-y to ϕ-ζ frame (Fig. 1b). The endoscope had a FOV of 250 μm and transverse and axial resolutions of about 1 and 10 μm, respectively, in the air. The endoscope could be rotated endlessly to change the imaging plane along the circumference (ϕ). To change the view plane along the lumen (ζ), we moved the animal axially by using a motorized translation stage. The focal depth (z) within the tissue was controlled externally without having to move the endoscope or mouse, simply by translating the coupling 40X objective lens to change the distance to the endoscope (Fig. 1c and Supplementary Fig. 1).

Figure 1. In vivo side-view endomicroscopy.

(a) Schematic of a laser-scanning side-viewing endoscope. The raster-scanned beam in x and y is relayed by grade-index lenses in the probe and directed by a 90° prism to a side-view window. ζ and ϕ represent the axial and circumferential coordinates, respectively, in the imaging plane. θ denotes the rotation angle of the probe, or the angle between x and ϕ axes. (b) Coordinate transform between the proximal (x-y) and distal (ζ-ϕ) imaging planes. The trace of raster-scanned beam is depicted in blue solid lines. (c) Imaging set up. The laser beam is projected to an animal stage. Dotted lines depict the outline of the beam diverging after going through the imaging plane. (d) 3-dimensional rendered fluorescence image of the vasculature in the descending colon of a normal C57B6/L mouse. (e) A fly-through rendered image of d. Fluorescence images of the microvasculature in the esophagus (f), vasculature of villi in the small intestine (g) and MHC-II-GFP expression in dendritic cells in the trachea (h). Blood vessels are visualized by intravenously injected FITC-dextran conjugates. In f-h, the horizontal axis represents the circumferential angle ϕ; scale bars, 200 μm.

To perform wide-area imaging, we used rotational and axial scanning; the probe scans along the luminal length repeatedly at different angles (θ) or alternatively along a helical path. During scanning, we continuously acquired a movie at 30 frames per second (576 × 576 pixels per frame). Each frame is rotated according to the probe angle, and image registration was applied to display the entire scanned area (Supplementary Fig. 2). We obtained a wide-area vasculature image in the descending colon of an 8 week-old mouse in vivo after intravenous injection of FITC-dextran conjugates. Figure 1d depicts a 3-dimensional rendered image of the microvasculature, constructed from a total of 60,000 frames acquired in 2,000 s over a 12 mm-long section. The dataset can be presented in various ways such as a flythrough view (Fig. 1e and Supplementary Video 1) and a 2-dimensional presentation. The high spatiotemporal resolution allows for quantitative analysis of vascular parameters such as flow velocity and vessel diameters (Supplementary Video 2 and Supplementary Fig. 3). The small-diameter probe is also well suited for esophageal imaging via transoral endoscopy (Fig. 1f and Supplementary Video 3). We also imaged the extensive network of the vessels and mucosal dendritic cells in the small intestine through a gastric “feeding” tube (Fig. 1g and Supplementary Fig. 4) and in the airway by bronchoscopy (Fig. 1h). The high resolution of the side-view probe allowed us to visualize the interaction of mucosal dendritic cells with fluorescently labeled antigens (Supplementary Fig. 5 and Supplementary Video 4).

To apply the probe to quantitative cellular imaging, we performed wide-area colonoscopy in FoxP3:GFP⁺ mice over the course of acute colitis induced by adding dextran sodium sulfate (DSS) to the drinking water for 5 days. The FoxP3⁺ regulatory T cells (T_reg) were visualized over a 2 mm (ϕ) × 5 mm (ζ) × 50 μm volume of the colonic mucosa repeatedly at Days −1 (normal colon), 3 (acute phase of colitis), 7 and 14 (recovery phase). In the normal state, T_reg cells were found to be distributed sparsely (Figs. 2a, b). We observed a significant increase in the T_reg cells during the course of DSS colitis (Fig. 2c and Supplementary Fig. 6), which supports a recent finding that a breach in the epithelial barrier enhances the expansion of T_reg cells in the colon¹².

Figure 2. Visualization of FoxP3⁺ T_reg cells in a DSS-induced colitis model.

(a) Large-area map of FoxP3:GFP⁺ T_reg cells taken before DSS treatment (day −1). (b) 2.5x image of the area (red square) marked in a. Inset is 10x image showing T_reg cells (green) and blood vasculature (red). (c) A typical image of T_reg cells at day 7 in the recovery phase of colitis. Blood vessels are visualized by intravenously injected Evans Blue. Scale bars, 200 μm; 20 μm in the inset.

Next, we applied side-view colonoscopy to investigate spontaneous colorectal tumorigenesis. We used an inducible tumor model¹³ in which the floxed adenomatosis polyposis coli (Apc) gene is inactivated by administration of adenoviral Cre into the colon. This mouse model mimics the somatic mutation observed in most patients with colorectal cancer and enables us to control the location and timing of the onset of tumorigenesis. Necropsy taken at 16–18 weeks after introduction of adeno-Cre revealed that all of the floxed Apc mice treated with adeno-Cre (n=13) developed a few large adenomatous polyps in the descending colon. Sham operated control mice that received saline only (n=4) did not show polyps. We performed in vivo colonoscopy in these mice every 2 weeks from week 9 till week 19, and imaged the blood vessels in a segment of descending colon about 5 to 25 mm from the anus. In the Apc-conditional knock-out mice, but not in the control mice, regions with abnormal vasculature were observed. Figure 3a depicts a vascular image taken at 11 weeks, revealing a small lesion of about 500 μm in diameter, characterized by vessel dilation and increased spacing between vessels. In the same mouse imaged at 13 weeks, this anomalous area had grown to a size of about 1.5 mm; we observed a strong fluorescent signal from fluorescent tracer (FITC-dextran) that had presumably leaked from the vessels and transiently accumulated in middle of the lesion (Fig. 3b). We confirmed the formation of tumors by examining sections of dissected colon (Supplementary Fig. 7). In another mouse, at week 17, we observed a larger lesion about 4 mm in diameter showing severe vessel dilation, tortuous vasculature, and elevated leakage of fluorescent tracer (Fig. 3c and Supplementary Video 5). These features are typical for angiogenic vessels associated with tumor development, which was confirmed by histology (Supplementary Fig. 7). We further engineered the mouse model so that the Apc inactivation is accompanied by constitutive expression of GFP. Side-view endomicroscopy enabled us to observe the formation of GFP-expressing Apc-knockout cells and to monitor their progression over time. Multiple groups of GFP⁺ cells were found typically within the first two weeks after Cre-administration (Fig. 3d). From z-stack images (50 μm), we could determine the volume of small nodules over time. Some of these lesions continued to grow, while others apparently shrank or vanished (Fig. 3d and Supplementary Fig. 8). These results demonstrate the possibility of using our side-view endoscope to monitor the fate of these cells in vivo from the moment of genetic mutation to the formation of large adenomas.

Figure 3. Longitudinal imaging of colorectal tumorigenesis.

(a-c) Fluorescence image of colorectal vasculature in a floxed Apc mouse at 11 weeks (a) and 13 weeks (b) after adeno-cre administration, and in a large lesion at week 17 (c) in another Cre-treated mouse. (d) Apc-knockout GFP⁺ cells (green) and blood vessels (red) at the same site in the descending colon, observed at days 10, 12, 14, and 28. Each image is a projection view of 50-μm z-stack. The serial images reveal a GFP⁺ lesion that grows in size over time (*). Other GFP⁺ nodules shrink (▲) or vanish (circle) at day 28. Blood vessels are visualized by intravenously injected FITC-dextran in a-c and by TAMRA-dextran in d. Scale bars, 200 μm.

In addition to genetic predisposition, the stromal microenvironment plays critical roles in the initiation and progression of tumors⁴. Side-view endomicroscopy can visualize a variety of events, such as vascular changes, matrix modulation, and circulating cell infiltration, during development and treatment, and thus is expected to be a powerful research tool in oncology. Also, it may also be used to directly monitor the transport of orally- or systemically-administered therapeutic agents, including nanoparticles, across the epithelial and endothelial barriers in the intestine and other organs, providing insights into new therapeutics¹⁴. Finally, “needle” endoscopy⁹ is an effective way to access internal tissues of solid organs. The side-view microprobe offers the possibility of visualizing the entire tissue surface along the insertion hole, which may prove useful in investigating brain diseases and neural signaling (Supplementary Fig. 9)¹⁵.

METHODS

Side-view endomicroscope

We fabricated the side-view endoscopes in house using 1-mm-diameter graded-index lenses (NSG America) and aluminum-coated right-angle microprisms (base length = 0.7 mm). The lens unit has a triplet structure with a magnification of 0.84, comprising a proximal coupling lens (ILW, pitch=0.25), a relay lens (SRL, pitch=1), and a distal imaging lens (ILW, pitch=0.16). We attached the microprism to the end surface of the imaging lens. The assembled optical unit was inserted into a stainless sleeve and glued. UV epoxy (Norland 81) was used to create a protective and transparent window at the distal tip (as shown in Fig. 1). The finished endoscope was mounted on a custom-built XYZ-translational and rotational stage and coupled into a custom-built confocal imaging system¹¹.

Imaging system

Our imaging system was built on a video-rate scanning laser confocal microscope platform previously described. The system has two continuous-wave lasers with emission at both 491 nm and 532 nm (Dual-Caylpso, Cobolt) and 635 nm (Radius, Coherent), respectively. For video-rate raster scanning, we used a custom-developed scanner comprised of an aluminum-coated polygonal mirror (MC-5, Lincoln Laser) and a galvanometer (6220H, Cambridge Technology). The scanner was configured to provide a field-of-view (FOV) of 250 × 250 μm² at the focal plane of a 40X objective lens (LUCPlanFl, NA=0.6, Olympus). The objective lens was mounted on a linear translational stage for fine control of the distance from the GRIN probe. Three photomultiplier tubes (PMT, R9110, Hamamatsu) were used as fluorescence detectors, which were placed after multilayer color filters (Semrock) and confocal pinholes. The PMT outputs were digitized by an 8-bit 3-ch frame grabber (Solios, Matrox) at 10 MS/s each channel. Images were displayed on a computer monitor at a frame rate of 30 Hz (512 × 512 pixels per frame, 3 windows) and stored in a hard disk in real time by using custom-written software based on Matrox Imaging Library (MIL9, Matrox).

Wide-area image acquisition and processing

We used image registration to produce a wide-area image from the movie acquired by the rotational and pullback operation of a side-view endoscope. The scan speed of the endoscope was controlled to be typically 100 – 200 μm/sec. With a frame rate of 30 fps and a field-of-view of 250 μm, this ensured that there is significant overlap in position between adjacent frames. Individual images were rotated according to the angular position of the rotational stage of the endoscope (Fig. 1 and Supplementary Fig. 2). Then, we applied an image registration algorithm¹⁶ written in Matlab (MathWorks) to merge the individual frames to produce a wide-area image. This process enhances image contrast by averaging the noise over multiple frames and also effectively removes artifacts due to tissue motion such as breathing (for more information, see Supplementary Fig. 2). 3D presentation of vasculature images (Fig. 1d and e) was made by using SketchUp Pro 7 (Google).

DSS mouse model of acute colitis

We used FoxP3:GFP⁺ knock-in mice (5 weeks old, female), kindly provided by T. B. Strom (Harvard Institutes of Medicine, Boston, MA). 3.5% wt/vol of dextran sulfate sodium (DSS; 36,000–50,000 MW, MP Biomedicals) was added to drinking water for 5 days. The DSS treatment was terminated from Day 6 to induce the recovery process from the inflammation.

Condition Apc knockout model of colorectal tumor

We used floxed Apc mice originally generated by the Kucherlapati group. The administration of adenoviral Cre into the colon epithelia can inactivate the Apc gene, which mimics somatic Apc mutation in human patients with sporadic colorectal cancer. Apc-GFP mice were homozygous for a floxed exon 14 of the Apc allele and heterozygous for a latent GFP reporter allele^17,18. For adeno-Cre delivery, we anesthetized the mouse at the age of 8 weeks by inhalation anesthesia (4% isoflurane, oxygen flow 1.5L/min) and made a 3–4 cm midline incision along about one-half the length of the lower abdomen. A 5 × 0.8 mm 10 g pressure vessel clip was placed proximally around the colon about 2 cm from the anus. A 3 cm long polyethylene tube (0.28 mm inner diameter) attached to a 1 cc syringe was introduced into the colon through anus until the extent of the proximal clip. A second clip was placed 1 cm distal to the first one. 100 μl of adenovirus (Ad5CMVCre, Gene Transfer Vector Core, Univ. of Iowa, Iowa, IA) was infused and incubated for 30 minutes. After removing both clips, the wound was closed with 4.0 silk sutures in two layers. Sham operated control mice were prepared by the same surgical procedure without adeno-Cre transduction.

In vivo mouse imaging

Prior to in vivo imaging, mice were anesthetized by intraperitoneal injection of ketamine/xylazine anesthesia (90mg/9mg per kg body weight). Eye ointment was applied to protect the cornea from dehydration during anaesthetization. The mice were placed on a heated plate of a motorized XYZ translational stage. For vasculature imaging, we injected fluorescent tracer intravenously. Images were typically acquired within 5 to 40 minutes after the injection. Fluorescein Isothiocyanate (FITC) dextran conjugates (500 μg/100 μl, 2,000,000 MW, FD2000S, Sigma Aldrich) was used for Figs. 1d-g, 3a-c; Tetramethylrhodamine (TAMRA) dextran conjugates (500 μg/100 μl, 2,000,000 MW, Invitrogen) was used for Fig. 3d; and Evans Blue dye (1 μg/100 μl, E2129, Sigma Aldrich) was used for Fig. 2b. GFP-expressing cells and the FITC vessel tracer were visualized by excitation at 491 nm and detection through a bandpass filter at 502-537 nm (Semrock). TAMRA was imaged by excitation at 532 nm and detection through a bandpass filter at 562-596 nm (Semrock). Evans Blue was excited at 635 nm and detected at 672-712 nm. All animal experiments were performed in compliance with institutional guidelines and approved by the subcommittee on research animal care at the Massachusetts General Hospital.

In vivo esophageal imaging

The mouse was laid down on the heated plate. A small block of gauze was placed under the neck. The plate was tilted to make a straight access for the endoscope through the mouth into the esophageal tract. We opened the mouth using forceps, gently pulled and pushed down the tongue using a cotton tip, and inserted a side-view endoscope through the mouth. Care was taken not to obstruct the airway. Breathing of the mouse was monitored during imaging.

In vivo percutaneous imaging of the small intestine

Small-intestine imaging was done via laparotomy. The mouse was laid down on the heated plate. After making a midline incision in the skin, we placed the jejunum over the center of the surgical opening by using cotton tips. Saline was applied to the exposed jejunum regularly during imaging to prevent the tissue from drying out. With an 18 gauge needle, we made a small hole on the jejunum. Through the hole, we injected 0.5 ml of distilled water, which serves as a lubricant, and inserted a side-view endoscope.

In vivo colonoscopy

Prior to gastrointestinal imaging, mice were starved for 24 hours before imaging to avoid strong auto-fluorescence signal of mouse excrement and digested food. We dilated the colon by injecting 0.5 ml of distilled water via enema using a rubber-tipped needle. The injected water also serves as a lubricant during endomicroscopy, avoiding tissue damage. We applied a small drop of 2% methylcellulose (Methocel, Dow) onto the tip of a side-view endoscope, which also serves as a lubricant and helps the endoscope pass through the anus.

In vivo tracheal imaging

Prior to imaging, we performed tracheotomy to open a direct airway for breathing through a small incision on the trachea. We then inserted a side-view probe into the trachea through the mouth.

In vivo brain imaging

We used a H2B:GFP⁺ transgenic mouse (12 weeks old, female, Jackson Laboratory) in which all the cell nuclei express green fluorescence. A craniotomy was performed under anesthesia prior to imaging. After applying a drop of saline, we inserted a GRIN side-view probe deep into the brain. We then performed wide-area imaging by a helical pullback scanning of the probe,