BoneKEy-Osteovision | Article
Application of digital imaging to problems in bone biology
David W Rowe
Zana Kalajzic
Yu-Hsiung Wang
Kathleen Buhl
DOI:10.1138/2002023
Our research group is developing promoter transgene constructs that activate in cells at different stages of differentiation within the osteoprogenitor pathway (). With the advent of GFP, it is now possible to observe cellular differentiation in real-time in primary cell culture and correlate this lineage information with the GFP expression pattern of the same transgene construct in intact bone. The GFP markers allow one to appreciate the microheterogeneity of cell differentiation within the osteoprogenitor lineage that is present both in primary cell culture and intact bone. One of the problems we have encountered is convincing other investigators that a particular promoter fragment has a characteristic expression pattern when we are limited by the number images we can afford to publish in standard print journals. How can our experience of looking at many samples both in cell culture and histological sections be transferred to a critical reviewer based on the few images that we select as being representative? In this brief report, we wish to propose a solution to the limitation placed by standard print journals for examining images either of intact bone or cell culture. The goal is to give the reviewer the same experience of examining the entire tissue specimen as the primary investigator had when the original sample was viewed through the microscope.
We have developed transgenic mice harboring various Col1A1 promoter fragments driving GFP that activate in preosteoblastic cells (pOBCol3.6GFP) or at the stage of full osteoblast/osteocyte differentiation (pOBCol2.3GFP) (). Figure 1 is a composite image showing the characteristic expression patterns of the two transgenes in primary cell culture and intact bone. Differences in intensity of GFP fluoresence and the shape of the positive cell can be appreciated in the pOBCol3.6GFP transgene. Specifically preosteoblastic cells have a weak signal intensity and an elongated fibroblast like morphology. They are found surrounding nodules in primary culture and within the periosteum of intact bone. The expression of this transgenes in osteoblastic cells is much stronger than in preosteoblastic cells. These cells are found within bone nodules in primary culture and on the bone surface of intact bone. In contrast, pOBCol2.3GFP shows one level of strong intensity and is located in cells within mineralized nodules of primary culture and on the surface of and within the bone matrix in intact bone. However, the accuracy of these statements based on the images as presented in Figure 1 is often unconvincing and does not convey the temporal expression in primary culture that is so characteristic of the two transgenes.
We have recently acquired a new microscopic system that may overcome the difficulty of providing images that are as convincing to a reviewer as they are to the investigator. We are employing a Zeiss Axiovert 200 (Gottingen, Germany), which is their new inverted microscope with outstanding fluorescent optics, that can be utilized equally well for imaging tissue culture or histological sections. All of the operations of the microscope are motor controlled and in turn these motors are controlled by Improvision's Openlab software (Coventry, England). Using their ingenious system for managing image files and a well-designed macros language, routines can be built to obtain images in a systematic and reproducible manner. Whether one is examining histological sections of bone or a defined area of a tissue culture plate, a series of adjacent images are taken at high-power so as to cover the entire field of interest. The images are exported from the Improvision environment into a powerful public domain software program developed for Macintosh (GraphicConverter, Lemke Software, Peine, Germany) that automatically aligns these images (tiled) to produce a single image which is examined in Photoshop (Adobe, San Jose, CA). Figure 2 shows the sequence of steps we have used to image a large bone section. Points are selected that mark the beginning and end points of a series of images that encompass the desired bone section. The images are captured, saved to disk, imported into GraphicConverter and tiled into the single Photoshop image. The image is rotated, cropped, reduced in size and compressed so that the initial 50-75 MB file can be preserved as a 1-4 MB file.
Figure 3 is an H&E stained bone section that was imaged with a 10X bright field ocular and represents 48 original images. Once you have downloaded this image into a PC or Macintosh platform, open it in Photoshop where the file may expand back to 10-20 MB. A fast processor and plenty of RAM is required to make this an enjoyable experience. It will appear as 5-10% of the total image size and display a low-power image of the total bone section. The image can be magnified up to 100% and remain unpixilated allowing the observer to look at any region of bone in the same way we observed the bone. Note that the full size of this image is 30 by 8 inches. The original image is 123 by 33 inches and produces ever greater cellular detail.
Because the processed digital image is captured, it can be maneuvered to bring out important differences between two different specimens. Figure 4 shows the same control bone now placed beside a bone derived from a ganciclovir treated Col2.3Δtk transgenic mouse (). Upon drug treatment, the active osteoblasts are destroyed while the osteocytes survive. An interesting side consequence of the lost osteoblastic activity is a lost of bone marrow activity.
Figure 5 is an unstained fluorescent image of pOBCol2.3GFP expression in bone from a mature three month-old transgenic mouse. The overall pattern of GFP expression can be appreciated at low-power while the cellular detail at any region of the bone can be appreciated by magnifying the image. The examination shows that the expression pattern of this transgene is limited to cells at the bone surface and within the bone matrix. More activity is seen along the endosteal surface in the metaphyseal region of bone, although limited areas of activity can be appreciated in the diaphyseal region.
Another example is the distribution pattern of osteoclasts in an entire bone section. Figure 6 is a TRAP stained section of normal bone that is contrasted with the bones derived from Col2.3Δtk mice after withdrawal from ganciclovir treatment during which time a wave of new bone formation develops that eventually is remodeled back to a normal bone architecture.
Figure 7 is a 10x phase image of primary calvarial cell culture derived from either pOBCol2.3GFP or pOBCol3.6GFP transgenic mice. A fluorescent image has been merged with the grayscaled phase image to show the relationship of the fluorescent pattern with the cellular detail. In this case 9 images are tiled together to produce a more representative distribution in which pOBCol3.6GFP expression is seen within and between bone nodules while pOBCol2.3GFP is only present within bone nodules.
As we have become more experienced with the microscope and the macros that control it, we are beginning to appreciate the impacts this type of analysis can have for studying bone biology. Because the entire section can be repetitively imaged, it should be possible to examine the histological section for GFP expression and then process the same section for enzymatic staining, immunostaining or in situ hybridization followed by standard histological staining. At each stage an image is taken and subsequently digitally overlaid to determine relationships between cells and structures of bone. Similarly, identical areas of tissue culture cells can be repetitively imaged over time to obtain the temporal expression of the transgenes. Here the output can be either a tiled series of images or a QuickTime movie that shows the progression of nodule development and GFP expression.
The power of acquiring these tiled digital images is realized when image analysis techniques are used to quantitate a visual impression. Thus in a histological section, the number of GFP positive cells and their relationship to the bone matrix, osteoclasts, BrdU labeled cells or an in situ marker can be ascertained and related to specific regions within the entire bone section. Multiple sections of bone can be recorded from which a 3-D reconstruction can be assembled that relates GFP expression to specific regions of bone. In tissue culture the tempo and magnitude of cell differentiation will also be amenable to quantitation. It is likely that subtle mutations of genes affecting bone formation will be expressed in primary culture as an alteration of tempo of differentiation as much as a change in magnitude when full differentiation is attained.
The generation of these data intense files cannot be accommodated in standard print journals. Just as websites have been developed to house primary data from microarray experiments, a similar need can be anticipated as digital analysis of histological and primary culture data are developed. This provides a strong rationale for a web based repository such as the BoneKEy initiative or for print journals to provide such a service. Otherwise this information can only be provided through links to the individual investigator which are unlikely to be as well maintained or as uniform in presentation as can be achieved with a common site.
Tips for saving images:
right click on your mouse over the "save image" link, select "save target as…" or "save as…" menu choice in your browser. save the image in your hard disk and open the image with Photoshop note that for a 2.5 MB image, download time is appropriately 10 minutes when you are using a 56K modem connection
This work is licensed under a
Creative Commons Attribution-Noncommercial-No Derivative Works 3.0 United States License.