1.

Introduction

Many areas of biomedical research have been held back by limited possibilities to extract simultaneous global and regional information of a studied biological process or event. Commonly, high-resolution studies of specific cell niches or regions of interest (ROIs) within a greater body of tissue does not allow for the resultant data to be put into a larger context. Due to technical limitations, information on spatial origin of the studied region, neighboring cell types, and the global state of the organ studied is often lost. Conversely, technologies allowing whole organ imaging do not generally provide the resolution required for detailed cell-level analyses. A protocol allowing for high-resolution analysis to be directed by a prior global assessment of the disease state of an entire organ would thus constitute a significant asset for many areas of research. A good example comes from different murine models for development of type-1 diabetes (T1D). Our present understanding of the natural history of this disease¹ rests to a large extent on analysis of rodent models, in particular the nonobese diabetic (NOD) mouse.^{2, 3} This model is characterized by the autoreactive T-cell mediated destruction of pancreatic $\beta$ -cells, leading to diabetes.^{4, 5, 6} While it is well established that insulitis progresses over an extended time period, detailed understanding of its kinetics, as well as the cellular and molecular mechanisms underlying observed regional variations, is largely lacking due to limitations in existing technology. In particular, high-magnification imaging alone [e.g., confocal laser scanning microscopy (CLSM) on isolated islets of Langerhans⁷], while providing excellent details about individual islets, does not provide an overview of disease progression at the level of the whole pancreas. Further, it does not generate information regarding the spatial origin of the analyzed islet, and importantly, does not permit studies of neighboring/interacting cell types and molecules.

We have previously adapted the concept of optical projection tomography (OPT)^{8, 9} to allow for imaging of specifically labeled structures within intact mouse organs, including the pancreas.¹⁰ To develop a means by which both the details of the autoimmune attack at the islet level, and also information of the global state of the pancreatic gland, could be studied in the same specimen, we postulated that four criteria would have to be met. The methodology should allow for; 1. global (whole organ) assessments of molecularly labeled pancreatic constituents (e.g., insulin producing islets); 2. selection of ROIs based on the expression of specific markers and the spatial position, shape, and size of individual objects; 3. high resolution analyses of user-defined ROIs identified based on the global assessment; and 4. high resolution 2-D and 3-D analyses of two or more molecular markers within the ROI. To develop a protocol that would meet these criteria, we turned to the NOD mouse model for T1D.

2.

Experimental Methods

3.

Results

Obtaining parameters for region-of-interest selection

The OPT scan data [Fig. 1a ] were reconstructed providing tomographic sections throughout the volume of each pancreas [Figs. 1a through 1g, 1i, 1j and other data not shown]. 3-D volume reconstructions were subsequently generated based on the signal from the insulin-specific antibodies and from tissue autofluorescence [Fig. 1b]. These are fully interactive and allow for free tilting, zoom, and virtual clipping of the specimen (Video 1 ). Based on the tomographic data, specific islet $\beta$ -volumes were quantified. This generated a full histogram of islet $\beta$ -cell volumes and information of corresponding centroid coordinates [Fig. 1h and other data not shown]. In the example depicted in Fig. 1b, the pancreas ( $16\text{week}$ duodenal-NOD) contained 93 insulin labeled islets having islet $\beta$ -cell volumes ranging from $2.9\times {10}^3\mu {\mathrm{m}}^3$ (which equals the maximum spatial resolution for the current sample at utilized zoom factor) to $2.56\times {10}^7\mu {\mathrm{m}}^3$ [Fig. 1h]. The total islet $\beta$ -cell volume for the current specimen was $3.46\times {10}^8\mu {\mathrm{m}}^3$ . As compared to stage matched NOD H2b congenic pancreata, this volume corresponds to approximately a 50% reduction in insulin cell mass. The control specimen depicted in Figs. 1i and 1j contained 1063 islets with a total $\beta$ -cell volume of $6.73\times {10}^8\mu {\mathrm{m}}^3$ .

Fig. 1

ROIs can be selected at will within the entire volume of the intact pancreas based on the spatial position, shape, and volume of specifically labeled objects. (a) and (c) show single images from the OPT scan datasets before (a) and after (c) microdissection of the selected ROI (insets). (b) and (d) show volume reconstructions of the pancreas based on the background autofluorescence (dark gray) and the signal from insulin-specific antibodies (white-islets) before (b) and after (d) isolation of the ROI (insets). The selected islets have been assigned red and yellow pseudocolors. (e), (f), and (g) show orthogonal section planes depicting the selected islets (red and yellow) from (e) coronal, (f) saggittal, and (g) transaxial views. (h) Histogram of individual islet $\beta$ -cell volumes. Volume data for the “red” and “yellow” islets depicted in, (b), (e), (f), and (g) have been assigned corresponding color codes. (i) and (j) show single images from (i) the OPT scan dataset and (j) corresponding volume reconstruction of a control NOD H2b pancreata. Scale bar in (a) is $2000\mu \mathrm{m}$ in (a) through (d), (i), and (j), inset $800\mu \mathrm{m}$ and $970\mu \mathrm{m}$ in (e), (f), and (g).

Video 1

Interactive OPT-generated 3-D renderings facilitate identification and isolation of individual islets within the full volume of the pancreas. The movie shows a screen recording of the specimen seen in Fig. 1a and illustrates how volume renderings of pancreas specimen can be interactively manipulated to facilitate identification and isolation of specific ROIs. Here, the specimen is first rotated (aligned with the field of view in the stereomicroscope) and thereafter the background fluorescence channel (green) is made semitransparent to lay bare the insulin labeled islets (red). Finally, the magnification is increased (QuickTime, $4.8\mathrm{MB}$ ). 1

10.1117/1.3000430.1

Isolation of regions of interest

To address the spatial character of infiltrating cell types, individual islets were selected based on their position within the whole pancreas (e.g., peripheral location or near the main duct), their shape (e.g., spherical or elongated), or their size. In the current example, two neighboring islets with different shapes and volumes were selected [smaller elongated (assigned yellow pseudocolor in Fig. 1b) and dumbbell-shaped larger (assigned red pseudocolor in Fig. 1b)]. 3-D quantification of the two islets provided information of their individual $\beta$ -cell volume. In the current example, the “red” islet $\beta$ -cell volume was $2.56\times {10}^7\mu {\mathrm{m}}^3$ and the spherical “yellow” islet $1.04\times {10}^7\mu {\mathrm{m}}^3$ . Volumetric data could be cross-referenced back to the 3-D model and served as a converse selection criterion [red and yellow in Figs. 1b and 1e through 1h]. Although these data provide both global and regional information, the current resolution of the OPT technology does not permit cell-resolution analyses of individual islets when addressing specimens on a current scale. Selected islets were therefore isolated as microbiopsies, in the current example containing both the “red” and “yellow” islets. This was made possible by using the interactive volume reconstructions as reference guides to accurately guide the biopsy capillary into the fixed pancreas specimen [Video 1 and Fig. 2a ]. After the biopsy was removed, the pancreas was subjected to another round of OPT scanning to confirm that the selected ROI had been properly isolated [Figs. 1c and 1d].

Fig. 2

High-resolution 3-D assessment of two individual islets and interacting infiltrate. (a) Schematic illustrating the use of interactive OPT datasets for isolation of individual islets within the full volume of the pancreas. The interactive datasets allow for free rotation, zoom, and virtual clipping of the pancreas outline and the labeled islets. This facilitates the identification and isolation of individual islets in a view matching the field of view of the agarose-embedded specimen in the stereo microscope. (b) through (k) show CLSM data of the highlighted islets in Fig. 1. For each islet, (b) and (g) volumetric and (c) through (k) 3-D isosurface data are shown. In all cases, green represents the infiltrate (CD3), and the islet $\beta$ -cell volumes (insulin) are shown as yellow or red to correspond with Fig. 1. In (f) and (k), the isosurfaces have been rendered as wire frames to facilitate a complete interpretation of the data. (For clarity, the wire frame renderings have been based on a larger voxel size to reduce the number of triangles). Note that a fiber in Fig. 1g has been digitally removed. Scale bar in (k) is $150\mu \mathrm{m}$ in (b), (c), and (f), $75\mu \mathrm{m}$ in (d) and (e), $100\mu \mathrm{m}$ in (g), (h), and (k), and $50\mu \mathrm{m}$ in (i) and (j).

Confocal analyses of regions of interest

To address the amount and spatial distribution of infiltrating cell types in relationship to the selected islets, the biopsies were stained with antibodies against a marker for infiltrating T-lymphocytes (CD3), and CLSM scanned. This allowed for the generation of high-resolution orthogonal sections planes (data not shown) throughout the ROI (down to $\sim 0.5\text{-}\mathrm{mm}$ $z$ depth when using the $20\times$ objective), as well as for the generation of volume-, isosurface-, and wire-frame renderings of the selected islet $\beta$ -cell volumes and surrounding infiltrate [Figs. 2b through 2k], altogether facilitating a detailed 3-D assessment of the infiltration process (islets depicted correspond to the “red” and “yellow” islets in Fig. 1). In the current example, the two islets displayed insulitis varying from partly [Figs. 2b through 2f] to fully T-cell enclosed islets [Figs. 2g through 2k]. Comparing OPT with CLSM data of 24 islets indicated a 1-D variation of $\pm 3\text{-}5\%$ , which translates into a volumetric variation of $\pm 10\text{-}15\%$ per islet between the two techniques.

4.

Discussion

In this study we provide a protocol that allows for site-directed analysis based on a global overview of the composition of the pancreas (i.e., volume and spatial distribution of molecularly labeled objects within the full volume of the gland). By applying this approach to the NOD model of type-1 diabetes, we demonstrate that all required criteria for such analyses have been fulfilled, and we provide detailed 3-D images of the autoimmune attack on individual islets of Langerhans. Although not falling within the scope of this brief work, it should be possible to expand these analyses to cover the infiltration dynamics of other subpopulations of the autoimmune attack during this process. Further, we anticipate that the developed methodology could easily be adapted to studies of combined global and region-specific character also in other organ systems, e.g., when high-resolution analyses of rare cell niches or events in a larger body of tissue are required.