Mammary gland development in the rodent has typically been evaluated using descriptive assessments or by measuring basic physical attributes. Branching density is an indicator of mammary development that is difficult to quantify objectively. This protocol describes a reliable method for the quantitative assessment of mammary gland branching characteristics.
An increasing number of studies are utilizing the rodent mammary gland as an endpoint for assessing the developmental toxicity of a chemical exposure. The effects these exposures have on mammary gland development are typically evaluated using either basic dimensional measurements or by scoring morphological characteristics. However, the broad range of methods for interpreting developmental changes could lead to inconsistent translations across laboratories. A common method of assessment is needed so that proper interpretations can be formed from data being compared across studies. The present study describes the application of the Sholl analysis method to quantify mammary gland branching characteristics. The Sholl method was originally developed for use in quantifying neuronal dendritic patterns. By using ImageJ, an open-source image analysis software package, and a plugin developed for this analysis, the mammary gland branching density and the complexity of a mammary gland from a peripubertal female rat were determined. The methods described here will enable the use of the Sholl analysis as an effective tool for quantifying an important characteristic of mammary gland development.
Mammary gland branching is a characteristic that is commonly assessed as an indicator of gland development, but it is difficult to objectively quantify. In 1953, Sholl1 described a method for measuring neuronal dendritic arborization in the visual and motor cortices of the cat, and a plugin for this technique was developed by Ferriera et al2. Because both neurons and mammary glands exhibit a similar tree-like structure, the plugin was employed to quantify mammary epithelial branching densities in 2D images of the peripubertal rat mammary gland. The peripubertal stage was chosen for analysis because weaning is a life stage that is often assessed in academic laboratories and test guideline studies. The Sholl analysis is a plugin distributed with FIJI, which is the open-source image processing package ImageJ, with additional plugins included. The plugin creates a series of concentric rings encircling a predefined center (typically the soma of a neuron or the origin of the primary duct of a mammary gland) and extending out to the distal-most portion of the object (the enclosing radius). It then counts the number of intersections (N) that occur on each of the rings. The plugin also returns a Sholl regression coefficient (k), which is a measurement of the rate of decay of epithelial branching.
Using ImageJ, a skeletonized image of a mammary gland whole-mount is created and the mammary epithelial area (MEA) is measured. The image is analyzed using the Sholl analysis plugin, and values for N and k, among other values not utilized here, are returned. Mammary epithelial branching density is determined by calculating N/MEA. The extent to which branching continues in the outer regions of the glandular epithelium is the branching complexity and is an indicator of uniform distal epithelial growth. As k is a measure of the distal decrease in epithelial branching, it is an effective measure of the branching complexity and a reliable indicator of mammary development.
This protocol describes a computer-assisted method for creating skeletonized images of mammary gland whole-mounts and quantitatively evaluating mammary branching characteristics in peripubertal male and female rats. This method is relatively rapid and does not require the use of specialized microscopy equipment. Development and validation of this method are described in Stanko et al. (2015)3. This report also describes preparation of rat mammary gland whole=mounts. Similar mammary whole-mount procedures have been described in de Assis et al. (2010)4 and Plante et al. (2011)5.
All animal use and procedures for this study were approved by the NIEHS Laboratory Animal Care and Use Committee and conducted in an Association for Assessment and Accreditation of Laboratory Animal Care-accredited facility.
1. Excise Mammary Glands
2. Prepare Mammary Whole-Mounts
3. Prepare Whole-mount Images for Analysis
4. Conduct the Sholl Analysis
5. Measuring the MEA
6. Reporting Data
The values for the measured enclosed radius, MEA, N, k, and calculated branching density for the mammary gland analyzed in this protocol are reported in Table 1. The Sholl analysis generates linear and semi-log plots of the number of intersections at each radius (Figure 9) and, if selected, a heat map of the intersections (Figure 10). Less-developed glands exhibit fewer intersections within the same MEA and therefore have a lower branching density. A well-developed gland will continue to branch uniformly throughout the entire glandular epithelium, particularly in the distal regions. The extent to which branching continues in these regions can be described as branching complexity and decreases in complexity are conveyed as a rate of decay (or Sholl regression coefficient, k). The rate of decay reflects the change in distal epithelial branching and is measured as the slope of the line of the number of intersections plotted against the enclosing radius (i.e., the longitudinal growth of the epithelium). Thus, the Sholl regression coefficient is calculated by taking the slope of the line of the plot of log(N/S) versus the radial distance (r), where log(N/S) = –k r + m, with N being the number of intersections for each ring of radius r and area S (πr2), and m being the intercept. Because the slope –k describes the decay of the intersections, a value of –k = 0 would indicate zero decay and uniform branching from the center of analysis to the edge of the epithelium. In sparsely developed glands, branching decay is increased; there are fewer intersections in the distal region of the epithelial tree; and the slope, k, is increased. Therefore, values of k approaching 0 are indicative of greater distal branching (i.e., branching complexity) and a well-developed mammary gland.
Figure 1: Ventral View. Image of the ventral portion of an adult female Sprague Dawley rat, illustrating how to secure the rat on the dissecting surface and the location of the 12 mammary glands, with the nipples circled. * The nipples of glands 6 and 7 are not visible. Please click here to view a larger version of this figure.
Figure 2: Female Rat Mammary Gland. Illustration of exposed mammary glands 4 (MG4) and 5 (MG5), with the skin pinned to the dissecting surface above MG4 and just below MG5. The glands should be removed from the skin beginning with MG5 and continuing up and dorsally until MG5 and 4 are completely removed. The nipple is in the distal area of gland 4, and care should be exercised to collect this area. The lymph node is indicated for reference. Please click here to view a larger version of this figure.
Figure 3: Mammary Gland Whole-mount. A whole-mount image of a mammary gland collected from a postnatal day 25 female Sprague Dawley rat. Scale bar =1 mm. Please click here to view a larger version of this figure.
Figure 4: Removal of Noise. The blue color channel of a mammary whole-mount image, with the background subtracted. (A) demonstrates examples with noise. The arrows indicate noise created by blood vessels, and the more heavily shaded region surrounding the ductal ends is an example of noise created by subtracting the background. (B) illustrates the image after the noise has been removed. Please click here to view a larger version of this figure.
Figure 5: Image Reconstruction. Reconstruction of the erased portions of the thresholded image. (A) The red arrows indicate regions where portions of the image were lost due to thresholding. Image reconstruction should be performed at these regions. (B) Mammary image after reconstructing the deleted regions. Image reconstruction should be conducted carefully and on a minimal basis so as to maintain the integrity of the original image Please click here to view a larger version of this figure.
Figure 6: Overlay of a Skeletonized Image. Overlay image showing a skeletonized image overlaid onto the original whole-mount image. This image demonstrates that the skeletonized gland reflects the branching of the actual gland with a high degree of accuracy. Please click here to view a larger version of this figure.
Figure 7: Enclosing Radius. Skeletonized image of a mammary whole-mount showing where the enclosing radius is measured (yellow). The line should begin at the base of the epithelial tree (center of analysis) and extend to the most distal point of the epithelium. Please click here to view a larger version of this figure.
Figure 8: Mammary Epithelial Area. Skeletonized image showing a polygon traced around the epithelial tree to determine the MEA. Please click here to view a larger version of this figure.
Figure 9: Sholl Plot Output. Sholl output of linear (A) and semi-log (B) plots of the number of intersections at each radial increment. The red dot in panel (A) is the abscissa of the centroid (geometric center). In panel (B), the blue line is the linear regression over the full range of data, while the red line is the linear regression over the 10th-90th percentile. Please click here to view a larger version of this figure.
Figure 10: Intersections Mask. When the "Create Intersections Mask" option is selected (step 4.3.7), the analysis will output a heat map of the number of intersections across the enclosing radius of the image. This heat map reflects the density of branching intersections throughout the epithelium (red = hot = high density; blue = cold = low density). The entire epithelium would be the same color in a heat map of an image where k = 0. Please click here to view a larger version of this figure.
Enclosing Radius (mm) | MEA (mm2) | N | k | Branching Density (N/mm2) |
7.4 | 71.7 | 2381 | 0.73 | 33.2 |
Table 1: Sholl Analysis Parameters. The values are the reported data for the Sholl analysis. The Enclosing Radius (step 4.2) and MEA (step 5.1) are measured values, N and k are Sholl analysis results and are returned in the Sholl analysis results window (step 6.1.), and Branching Density is calculated using the formula N/(MEA-LNA) (step 6.1.2).
From birth until puberty, mammary gland growth is allometric. After puberty, the mammary gland develops through extensive ductal branching and elongation, which continue until the mammary epithelium occupies the entire fat pad. Branching characteristics are an important aspect of mammary gland development, and the ability to objectively quantify these characteristics can be highly useful for assessing normal mammary development and for identifying abnormal development following early life exposures to mammary toxicants.
Scoring morphological characteristics, quantifying basic dimensional measurements, and counting mammary structures are typical methods for evaluating mammary gland development. However, these methods are not especially sensitive due to the considerable variation in the size and shape of rodent mammary glands and developmental interpretation may be difficult for an inexperienced evaluator. Furthermore, the potential for bias exists in studies that are not blinded properly. The Sholl analysis method provides an efficient protocol for accurately quantifying mammary epithelial branching density and branching complexity, discrete morphological characteristics of mammary gland development, which can easily be compared across studies and laboratories.
There are critical steps within several sections of this protocol. The first and most important relates to the condition of the mammary gland whole-mount. The accuracy of this method relies upon a mammary gland that is collected wholly intact, mounted with no defects, properly fixed and stained, and demonstrates no oxidation of the gland or significant discoloration of the mounting medium. If the gland is torn or folded, an accurate measure of the branching density cannot be obtained. If the ductal ends are not present, the value for k will not be representative of the entire gland. Thresholding will be difficult in glands that have not been fully fixed due to a lack of staining contrast in the ductal epithelium. And finally, if oxidation or discoloration is present, these blemishes could prevent the analysis from measuring intersections in the affected area.
When suitable whole-mounts have been prepared, the next critical step is capturing the images at the same magnification. It is common practice to capture digital images at the highest resolution possible. However, for the Sholl analysis, it is more important that all images be captured at the same magnification. As described in Stanko et al. (2015)3, a caveat was discovered where images of smaller glands captured at high magnification exhibited greater branching densities than images of larger glands captured at a lower magnification, even though they visually appeared to be less developed. We hypothesized that the higher magnification resulted in greater detail, which carried over into the skeletonized image and resulted in a higher N, which over-represented the branching density of the smaller glands. This issue is alleviated by capturing all images at the same magnification.
While the basis of an accurate analysis lies within the whole-mount, the greatest potential for user-influenced changes in intersection data lie within the steps for noise removal. All images contain noise, to some extent, due to staining intensity, non-relevant physiological entities (e.g., blood vessels), and artifacts of thresholding. Each image must be addressed independently due to variations in the amount of noise between images. Care must be taken not to remove too little or too much noise, as this can skew the number of intersections and, consequently, the branching density. However, the extent to which noise affects the interpretation has not been examined. The user should decide how meticulous to be with noise removal and should also exercise consistency to maintain the integrity of the images. It is highly recommended that the user be blinded to treatment when conducting noise removal, as this will minimize the potential for bias. Noise removal is described in detail the ImageJ User's Guide7. In this procedure, noise is removed primarily from the background-subtracted image. Additionally, the thresholding process itself may remove segments of the gland. Portions of the gland where only a few pixels have been removed will be reconstructed automatically when the skeletonized image is dilated. However, expansive gaps may require manual reconstruction. The user should decide whether and to what extent to reconstruct these segments, again maintaining the integrity of the original images.
Although this is not critical, it is important to maintain software updates, as ImageJ software is updated frequently. The methods described here are based on version 1.48v. FIJI and the Sholl analysis plugin are also updated regularly, and the protocol described here is based on v3.4.1. Changes made in later versions of both ImageJ and the Sholl Analysis plugin can affect these methods. ImageJ automatically checks for updates, but updates for FIJI should be conducted regularly, and changes between the current versions and those utilized here should be addressed as needed. All parameters are defined in subheadings on the Sholl Analysis webpage6. Parameter settings within this procedure are based on images captured from mammary gland whole-mounts created in our laboratory and are not absolute. Whole-mount preparation varies from lab to lab, and these parameters may be adjusted accordingly to optimize images and output.
The mammary gland whole-mount utilized in this study was from a female Sprague Dawley rat at PND 25, and the method was applied appropriately and without limitations. In rats, the mammary epithelial density increases with age to a point where it prevents thresholding the image with high enough resolution to generate an accurate skeletonized image of the gland. Therefore, we currently do not recommend using this method on glands from rats older than PND40. Although the strain of rat has been indicated here, it is irrelevant, as the authors are not currently aware of any strain-specific mammary traits that would prevent the use of this method. Furthermore, while the method described within was conducted using a female rat, it could also be applied to the mammary glands of male rats. This application has also been effectively used with whole-mounts of mice (Deirdre Tucker, personal communication) and should be suitable for mice of any age, as mammary glands in mice do not grow as dense as those in rats. However, there are two limitations with using this application in mice: 1) there may be too few branching intersections in younger mice to detect significant differences and 2) this method cannot be applied to male mice, as they do not exhibit mammary epithelium. Regardless, this automated method is faster, unbiased, and much less labor-intensive than counting branching intersections manually.
It is possible that investigators may wish to utilize the mammary gland for other experimental techniques, such as excising abnormal structures or for immunohistochemistry (IHC). Although Tucker et al. (2016) have described a method for preparing a hematoxylin-eosin-stained section from a mouse mammary gland whole-mount,8 we typically consider creating a whole-mount to be a terminal process and do not know of methods for using a whole-mounted mammary gland for additional sensitive assays, such as IHC or TUNEL assays. Where sensitive assays using mammary gland tissue are required in conjunction with whole-mounts, it is recommended to use the contralateral mammary glands.
The mammary gland continues to be the focal point in an increasing number of studies, yet differences across laboratories exist in both whole-mount preparation9,10,11,12 and developmental assessments13,14,15. The modification of the Sholl analysis described here provides a standardized method for the objective quantification of branching density, an important characteristic of mammary gland development, in the rodent mammary gland. This method can be applied to mammary whole-mounts of either male or female rodents, and though currently recommended for use in only early-postnatal to peripubertal glands from rats, it can be applied to mammary glands from mice of all ages. The application is particularly suitable for mammary glands collected from peripubertal rodents as this period is a recommended endpoint for mammary whole mount preparation in test guideline studies. Optimization of this method for use in the denser mammary glands of adult rats is currently being considered.
The authors have nothing to disclose.
The authors would like to acknowledge Dr. Michael Easterling (Social and Scientific Systems, Inc., Durham, NC) for his assistance with the validation of this method and Dr. Tiago Ferreira (McGill University, Montreal, Quebec, Canada) for his continual assistance with the Sholl application.
Dissecting board | NA | NA | A piece of styrofoam roughly 10"x12" is suitable. |
Dissecting T-Pins | Daigger | EF7419A | |
Spray bottle with ethanol | NA | NA | 70% ethanol is suitable. |
Curved dissecting scissors | Fine Science Tools | 14569-09 | |
Straight dissecing scissors | Fine Science Tools | 14568-09 | |
Curved forceps | Fine Science Tools | 11003-12 | |
Superfrost Plus 24x75x1 mm microscope slides | ThermoFisher Scientific | 4951PLUS-001 | Thermo Scientific Superfrost Plus & Colorfrost Plus slides hold tissue sections on permanently without the need for expensive coatings in IHC and Anatomical Pathology applications. This treatment reduces tissue loss during staining as well as hours of slide preparation. Slides electro-statically attract frozen tissue sections and cytology preparations and feature a chemistry similar to silane, although optimized to improve application performance. https://www.thermofisher.com/order/catalog/product/4951PLUS4. |
Fisherfinest Premium Cover Glass 24x60x1 mm | Fisher scientific | 12-548-5P | |
Bemis Parafilm M Laboratory Wrapping Film | Fisher scientific | 13-374-12 | |
Chloroform | Sigma-Aldrich | C2432 | |
Glacial acetic acid | Sigma-Aldrich | A9967 | |
Ethanol absolute, ≥99.8% (GC) | Sigma-Aldrich | 24102 | |
Xylene | Sigma-Aldrich | 214736 | |
Carmine alum | Sigma-Aldrich | C1022 | |
Aluminum potassium sulfate | Sigma-Aldrich | A6435 | |
Permount mounting media | Fisher Scientific | SP15 | |
Macroscope | Leica | Z16 APO | This is the image capturing hardware and software used in this laboratory. As there are many different options, the methods and applications may vary between laboratories. |
Digital camera | Leica | DFC295 | |
Camera software | Leica | Leica Application Suite v3.1 | |
ImageJ software | Open source | http://imagej.net/Welcome | |
Sholl analysis | Open source | http://imagej.net/Sholl_Analysis |