In vivo longitudinal depth-wise visualization of tumorigenesis by needle-shaped side-view confocal endomicroscopy

In vivo, longitudinal observation of tumorigenesis in a live mouse model over an extended time period has been actively pursued to obtain a better understanding of the cellular and molecular mechanism in a highly complex tumor microenvironment. However, common intravital imaging approaches based on a conventional laser scanning confocal or a twophoton microscope have been mostly limited to the observation of superficial parts of the solid tumor tissue. In this work, we implemented a small diameter needle-shaped side-view confocal endomicroscope that can be directly inserted into a solid tumor in a minimallyinvasive manner in vivo. By inserting the side-view endomicroscope into the breast tumor from the surface, we achieved in vivo depth-wise cellular-level visualization of microvasculature and fluorescently labeled tumor cells located deeply inside the tumor. In addition, we successfully performed longitudinal depth-wise visualization of a growing breast tumor over three weeks in a live mouse model, which revealed dynamic changes in microvasculature such as a decreasing amount of intratumoral blood vessels over time. © 2019 Optical Society of America under the terms of the OSA Open Access Publishing Agreement


Introduction
Tumor microenvironments are constituted by not only tumor cells but also numerous stromal cells and immune cells with blood vessels and various soluble factors such as hypoxiainducible factors. In the development of a tumor, called tumorigenesis, not only tumor cell but also diverse types of cells including immune, vascular and stromal cells interact with each other in a highly complex and dynamic manner at the cellular and molecular level [1][2][3][4][5][6][7][8]. It is imperative to improve our understanding of the tumor microenvironment to develop a more effective target therapy and optimize treatment strategies. Direct in vivo observation of cellular-level phenomena inside tumor tissue is highly desirable as it can provide new insight into the complex and dynamic processes in tumorigenesis [9][10][11][12][13][14][15]. In addition, to observe the single tumor-bearing mouse over an extended period of time, several imaging window techniques have been developed for various organs and tissues such as subcutaneous tissue [16,17], brain [18], mammary fat pad [19], intestine [20], and liver [21]. However, highresolution intravital imaging has been mostly limited to the superficial portion of the tumor tissues. This is mainly due to a rapidly deteriorating signal-to-background ratio in the image acquired at a deeper portion of tissues, which is mainly caused by increased multiple light scattering. To overcome this limitation, a miniature optical probe based on graded-index (GRIN) rod lenses was suggested to enable cellular-level deep tissue imaging [22][23][24][25][26][27][28][29]. However, a longitudinal in vivo cellular-level visualization of deep tumor tissue during tumorigenesis to investigate the dynamically changing microenvironment and complex pathophysiology has remained as a challenging task.
In this study, by utilizing GRIN lenses based side-view endomicroscope integrated into a small-diameter needle, we successfully achieved minimally-invasive cellular-level depth-wise visualization of microvasculature and fluorescently labeled tumor cells deeply placed inside a solid tumor. The side-view endomicroscope was fabricated by attaching 350 μm diameter triplet GRIN lenses consisting of a coupling lens, a relay lens, and an imaging lens with an aluminum coated 90° micro-prism. It was then packaged into a commercial hypodermal 22 gauge needle tip for direct and stable insertion to the deep region of solid tissue in a live mouse. To demonstrate the depth-wise imaging performance of the side-view endomicroscope into the deep tissue, it was inserted into the dorsal skin layer of a H2B-GFP transgenic mouse in which fluorescent nuclei were successfully visualized from the epidermis to the deeper dermis and microvasculature simultaneously. Finally, we successfully performed repetitive and longitudinal depth-wise visualization of a growing tumor in the orthotropic mouse model of breast cancer during three weeks, which revealed dynamic changes in microvasculature such as a decrease of intratumoral blood vessels over the timecourse of tumorigenesis.

Needle-shaped side-view endomicroscope based on GRIN lens
A side-view endomicroscope was fabricated based on miniature GRIN lenses, as shown in Fig. 1(a). It consists of a high NA coupling lens (NA: 0.5), a 1.5 pitch relay lens (NA: 0.2), and an imaging lens (NA: 0.43) with a diameter of 350 μm attached to a 90° micro-prism (base length: 300 μm) with a reflective aluminum coating on the hypotenuse surface to reflect light for side-viewing. The working distance of the coupling lens is 100 μm when the focal point made by imaging lens is on the surface of the micro-prism. The fabricated side-view endomicroscope with a total length of 13.7 mm was packaged into a hypodermal 22 gauge needle tip (PrecisionGlide syringe needle, BD) by using a UV curing optical adhesive (NOA61, Norland Products). The NA of the imaging lens was calculated by ray-tracing software, which was polished to achieve the working distance of 100 μm for coupling lens when the focal point made by the imaging lens is on the surface of attached micro-prism. The fabricated optics with total length of 13.7 mm and diameter of 350 μm was packaged into a hypodermal 22 gauge needle tip (PrecisionGlide syringe needle, BD) with the nominal inner diameter of 0.41 mm using a UV curing optical adhesive (NOA61, Norland Products). To avoid excessive damage to the tissue and to protect the micro-prism, the empty space between the needle tip and the prism was filled with a UV-curing adhesive. The finally packaged needle-shaped side-view endomicroscope had a diameter of 0.7 mm and a length of 15.4 mm. As shown in Fig. 1(b), for in vivo deep tissue imaging with an anesthetized mouse, the sideview endomicroscope was fixed to a custom-design mount consisting of XY translation mount (CXY1, Thorlabs) and Z-axis translation stages (9061-COM-M, Newport) for precise adjustment of the position under the objective lens. After integration into the imaging system, the needle-shaped side-view endomicroscope with the magnification of 0.83 provided a circular-shaped field of view (FOV) with diameter of 180 μm and transverse and axial resolutions of about 1 and 10 µm, respectively.

Custom-built laser-scanning confocal endomicroscopy
A custom-built video-rate laser-scanning confocal microscope system previously implemented [27,[30][31][32] was used as an imaging system for the needle-type side-view endomicroscope. Figure 1(b) shows a schematic illustration of the custom-built confocal microscope system and photos of the side-view endomicroscope positioned under the objective lens. As excitation sources for multi-color fluorescence imaging, four continuous wave laser modules with output wavelengths of 405, 488, 561, and 640 nm (Coherent OBIS, Cobolt MLD, Cobolt Jive and Cobolt MLD) were used. The excitation lights were combined by dichroic beam splitters (DBS1, FF593-Di03; DBS2, Di01-R405; DBS3, FF520-Di02, Semrock) and then delivered to a multi-edge dichroic beam splitter (DBS4, Di01-R405/488/561/635, Semrock). 2D laser beam scanning in a raster pattern was achieved by a fast rotating 36 facets polygonal mirror scanner (MC-5, Lincoln Laser) providing 17.28 kHz lateral scanning in the X-axis and a galvanometer mirror scanner (6230H, Cambridge Technology) providing 30 Hz vertical scanning in the Y-axis. The scanning laser beams were then delivered to the back aperture of a commercial objective lens (LUMFLN 60XW, Olympus). The multi-color fluorescence signals from the imaging sample were epi-detected and separated from the excitation laser beams by DBS4. The multi-color fluorescence signals were then split into the single-color signal by dichroic beam splitters (DBS5, FF484-Di01; DBS6, FF560-Di01; DBS7, FF649-Di01, Semrock) and detected by photomultiplier tubes (PMT; R9110, Hamamatsu) through band pass filters (BPF1, FF01-442/46; BPF2, FF02-525/50; BPF3, FF01-600/37; BPF4, FF01-685/40, Semrock). Electronic outputs from the PMTs were digitized by a 4 channel frame grabber (Solios, Matrox) with a sampling rate of 10 MHz. The video-rate images with a frame rate of 30 Hz and a size of 512 × 512 pixels were displayed and recorded in real time by custom-written software using Matrox Imaging Library (MIL9, Matrox).

Animal model
Mice were housed and bred in an institutional animal facility in Korea Advanced Institute of Science and Technology (KAIST). BALB/c mice were purchased from OrientBio (Suwon, Korea) or bred in-house. Histone H2B-eGFP transgenic mice (Stock no. 006069, Jackson Laboratory [33]), which endogenously express eGFP in nuclei, were used for in vivo cellularlevel depth-wise visualization from the epidermis to the deeper dermal layer of dorsal skin. For in vivo longitudinal and repetitive endomicroscopic depth-wise visualization of tumorigenesis in the single mouse, the breast cancer cells labeled with fluorescence protein, 4T1-Luc2-GFP cells (2.5 × 10 6 ), were orthotopically injected into the mammary fat pad of wild-type BALB/c mice. All animal experiments were performed in accordance with the standard guidelines for the care and use of laboratory animals and were approved by the Animal Care and Use Committee of KAIST (protocol no. KA2016-51). All imaging and surgical procedures were performed under anesthesia, and all efforts were made to minimize suffering.

In vivo imaging procedure
For in vivo imaging, the mouse was anesthetized with a mixture of zoletil (30 mg/kg) and xylazine (10 mg/kg) by intraperitoneal injection. Body temperature of the anesthetized mouse was maintained at 36°C by a homeothermic temperature monitoring and control system (RightTemp, Kent Scientific). For the depth-wise visualization of the skin, the skin was shaved by a hair clipper and hair removal cream, and then the side-view endomicroscope was slowly inserted into the dorsal flank. Before the insertion, the target site was cleaned by alcohol swab. For the depth-wise visualization of solid breast tumor tissue, the side-view endomicroscope was gently inserted from the top surface to the bottom of the tumor until it penetrated the whole tumor tissue. For longitudinal observation of tumorigenesis, the growing tumor of the signal mouse was repeatedly imaged from day 9 to day 21 after injection of the cancer cells with 3 day intervals. At each imaging session, the side-view endomicroscope was inserted at the center of the tumor tissue while avoiding the previous insertion site. To simultaneously visualize the microvasculature with the GFP signal in the skin and the tumor, anti-CD31 antibody (Stock no. 553708, BD Biosciences) conjugated with a far-red color fluorophore, AlexaFluor 647 (Stock no. A20006, Invitrogen), was intravenously injected into the mouse at 2 hours before the imaging.

Preparation of cryosectioned tumor tissue slices
Whole tumor tissues were dissected from the mouse perfused with 4% paraformaldehyde (PFA). Overnight dehydration was carried out twice repeatedly by immersing the tumor tissue while changing the concentration of PFA to 15% and 30%. The tissue was then placed in a cryomold with optimal cutting temperature compound (OCT compound, Tissue-Tek TM ) for freezing. After freezing the tumor tissue filled with OCT, it was sliced using a cryotome machine (FSE Cryostats, Thermo Scientific) to a thickness of 20 μm.

In vivo cellular-level depth-wise visualization of dorsal skin tissue by needleshaped side-view endomicroscopy
A H2B-eGFP transgenic mouse endogenously expressing eGFP in nuclei was used to demonstrate the depth-wise deep tissue imaging capability of the newly fabricated needleshaped side-view endomicroscope. As shown in Fig. 1(c), the side-view endomicroscope could be inserted into the dorsal flank skin of an anesthetized H2B-eGFP mouse from the epidermis to the deeper dermis in a minimally invasive manner. In addition to the nuclei expressing eGFP, the microvasculature in the dermis was fluorescently labeled in vivo by intravenously injecting anti-CD31 antibody conjugated with a far-red fluorophore at 2 hours before the insertion of the side-view endomicroscope. The dotted circle in Fig. 1(c) shows a representative single FOV image obtained in vivo at the deeper dermis in which nuclei of individual stromal cells and capillaries were clearly visible. While inserting the side-view endomicroscope, we could continuously acquire two-color fluorescence images of nuclei and blood vessels at different depths in the skin. To mitigate the chromatic aberration induced focal plane difference between two-color channels, we adjusted the distance between objective lens and surface of coupling lens to sequentially visualize different focal planes by translating only the position of the objective lens. The acquired single FOV images of each channel were projected together then successfully stitched to generate a large-area mosaic image. It shows the depth-wise cellular-level visualization of the skin from the lower epidermis to the deeper dermis, as shown in Fig. 1(d), with a length of 740 μm along the depth-axis at the insertion site.

In vivo repetitive and longitudinal cellular-level depth-wise visualization of orthotropic breast tumor by needle-shaped side-view endomicroscopy
For the in vivo longitudinal depth-wise visualization of tumorigenesis, an orthotopic breast tumor model established by injecting 4T1-Luc2-GFP cancer cells into the mammary fat pad of wild-type BALB/c mice were used. The needle-shaped side-view endomicroscope was repetitively inserted into the growing tumor in a single mouse model from day 9 to day 21 with 3 day intervals after injection of the cancer cells, as depicted in Fig. 2(a). As shown in Fig. 2(b), which presents representative single FOV endomicroscopic images obtained at day 9 and 21, 4T1-Luc2-GFP cancer cells and highly dilated tumor blood vessels in the deep tumor tissue were clearly distinguishable with cellular-level resolution.

Comparative quantitative analysis of in vivo endomicroscopic tumor images and ex vivo microscopic cryosectioned tumor images
To validate the representativeness of the depth-wise mosaic image obtained by the insertion of the needle-shaped side-view endomicroscope with regard to reflecting the features of the whole tumor tissue, whole tumor tissue dissected from the 4T1 breast cancer mouse models was processed into cryosectioned tissue slices and visualized by a conventional laserscanning confocal microscope ex vivo. Comparative analysis of these cryosectioned whole tumor tissue slice imaging results with the depth-wise endomicroscopic tumor imaging results obtained at the same days (day 9, 12, 15, 18, and 21) after the cancer cell injection was performed. Figure 4(a) shows a representative in vivo endomicroscopic imaging result of the tumor obtained with the needle-shaped side-view endomicroscope at day 9 and a representative ex vivo confocal microscopic imaging result of cryosectioned tumor tissue slice dissected on day 9. Figure 4(b-c) present a comparative quantitative analysis of the whole tumor height and the fractional ratio of intra-tumoral vessel area inside the tumor tissue, respectively. As shown in Fig. 4(b), the average tumor heights calculated from the imaging results obtained with two different modalities, in vivo side-view endomicroscopy and ex vivo confocal microscopy, were similar to each other (n = 4). Although statistically non-significant (t-test) at all of time points from day 9 to day 21, the calculated tumor heights in the cryosectioned tumor tissue slices were slightly smaller, presumably due to the minimal tissue shrinkage induced by dehydration in the process of cryosectioning. Average tumor heights were calculated to be 1.71 ± 0.12 mm in the in vivo side-view endomicroscopic imaging group and 1.69 ± 0.04 mm in the ex vivo confocal microscopic imaging group at day 9. They continuously increased over time and were calculated to be 5.86 ± 0.54 mm with in vivo side-view endomicroscopy and 5.52 ± 0.35 mm with ex vivo confocal microscopy, respectively. Additionally, the fractional ratio of the intra-tumoral vessel areas was quantitatively calculated from the imaging results by two modalities using Image J. As was the case with tumor height, the fractional ratios of intra-tumoral vessel areas were also similar in the imaging results of the two modalities (n = 4) with statistical non-significance (t-test), as shown in Fig. 4(c). At day 9, the average fractional ratio of the intra-tumoral vessel area was calculated to be 19.04 ± 1.81% in the in vivo side-view endomicroscopic imaging group and 18.25 ± 6.32% in the ex vivo confocal microscopic imaging group. Interestingly, subsequently after only 3 days, at day 12, these values were calculated to be significantly decreased to 6.00 ± 1.01% with in vivo side-view endomicroscopy and 7.01 ± 0.58% with ex vivo confocal microscopy.

Discussion
To minimize the potential of microvascular damage, insertion sites of needle-shaped endomicroscope were carefully selected not to interfere with vessels visible through the skin over the tumor. The insertion of endomicroscope was performed with small increment of about 100 μm to minimize a sudden physical change to the tissue surrounding the endomicroscope and acquire multi-color fluorescence image at each depth. During the whole insertion and imaging process of the needle-shaped endomicroscope into the tumor, no imaging artifacts those might be caused by the bleeding such as image blurring by fluid, a sudden decrease in imaging brightness, or reduced sharpness were observed. In addition, after the removal of the needle-shaped endomicroscope, no immediate bleeding from the insertion site and no scar formation around the insertion site over time were observed. Based on these observations, collectively, we believe that the damage of microvasculature was minimal although it is still difficult to fully verify or guarantee no damage.
The 4T1-GFP breast tumor grew very rapidly in vivo in a relatively short period of time, as shown in Fig. 3 and 4(b). Rapid uncontrolled growth of tumor builds up a unique microenvironment inside the tumor tissue, which is very different from the case of normal tissue. Compared to normal tissue, the solid tumor tissue comprises a highly heterogeneous extracellular matrix with various growth factors and adhesion molecules, which greatly alters the vasculature inside the tumor [34]. As shown in Fig. 2(b) and Fig. 3, blood vessels observed deep inside the 4T1-GFP breast tumor were greatly dilated and formed a torturous network with excessive branching.
Notably, repeated longitudinal wide-area depth-wise observation of growing tumor through the entire depth by the needle-shaped side-view endomicroscopy revealed a continuous decline of the ratio of intra-tumoral vessel area from day 9 to day 21, as shown in Fig. 4(c). This is because the rapid and exponential proliferation of the 4T1-GFP cancer cells exceeds the speed of angiogenic new blood vessel formation. In addition, the uncontrolled proliferation of 4T1-GFP cancer cells in the center of the tumor compresses the neighboring blood vessels and thereby hampers blood flow. Finally, it is worth noting that the required number of mice to obtain the analysis results shown in Fig. 4(b-c) using the in vivo side-view endomicroscopy was five times less than the case with the ex vivo confocal microscopy because the single mouse could be repeatedly imaged at each of the five time points between day 9 and day 21.

Conclusion
By using a needle-shaped side-view endomicroscope integrated into a laser-scanning confocal microscope, we achieved a depth-wise visualization of microvasculature and tumor cells throughout a solid breast tumor in vivo. In a minimally-invasive manner, a repetitive and longitudinal depth-wise visualization of a growing breast tumor in the orthotropic mouse was successfully performed during three weeks, which revealed a decreasing amount of intratumoral blood vessels over the time-course of tumorigenesis. The established method could be a useful tool to image and analyze various cellular-level dynamic changes that occur at the deep center part of a large growing tumor that is normally not accessible by other conventional microscopic imaging methods.