Dissection, MicroCT Scanning and Morphometric Analyses of the Baculum

Summary

Many biological structures lack easily definable landmarks, making it difficult to apply modern morphometric methods. Here we illustrate methods to study the mouse baculum (a bone in the penis), including dissection and microCT scanning, followed by computational methods to define semi-landmarks that are used to quantify size and shape variation.

Abstract

Modern morphometrics provides powerful methods to quantify size and shape variation. A basic requirement is a list of coordinates that define landmarks; however such coordinates must represent homologous structures across specimens. While many biological objects consist of easily identified landmarks to satisfy the assumption of homology, many lack such structures. One potential solution is to mathematically place semi-landmarks on an object that represent the same morphological region across specimens. Here, we illustrate a recently developed pipeline to mathematically define semi-landmarks from the mouse baculum (penis bone). Our methods should be applicable to a wide range of objects.

Introduction

The field of morphometrics includes a diversity of methods to quantify the size and shape of the biological form, a fundamental step in scientific inquiry^1,2,3,4,5,6. Traditionally, the statistical analysis of size and shape begins by identifying landmarks on a biological structure, and then measuring linear distances, angles and ratios, which could be analyzed in a multivariate framework. Landmark-based Geometric Morphometrics is an approach that retains the spatial position of landmarks, preserving geometric information from data collection through analysis and visualization⁵. Generalized Procrustes Analysis (GPA) can be applied to remove variation in location, scale, and rotation of landmarks to produce an alignment between specimens that minimizes their squared differences - what remains is shape dissimilarity⁷.

An important concept of any morphometric analysis is homology, or the idea that one can reliably identify landmarks representing biologically meaningful and discrete features that correspond between specimens or structures. For example, human skulls have homologous processes, foramina, sutures, and ducts that can enable morphometric analyses. Unfortunately, the identification of corresponding landmarks is difficult across many biological structures, especially those with smooth surfaces or curves^8,9,10.

We approach this problem below using computational geometry. The general workflow is to generate a three dimensional scan of the object that can be represented as a cloud of points, and then rotate and transform that point cloud so that all specimens are oriented on a common coordinate system. Then we mathematically define semi-landmarks from specific regions of the object. Discrete semi-landmarks placed on such regions are biologically arbitrary¹¹. Conducting GPA and subsequent statistical analyses can produce undesirable artifacts^8,12 because arbitrarily placed landmarks may not be biologically homologous. Therefore, we allow these semi-landmarks to mathematically "slide". This procedure minimizes the potential difference between structures. As argued elsewhere the sliding algorithm used here is appropriate to quantify similar anatomical regions lacking easily identified corresponding landmarks^{3,6,8,10,11,12}. These methods have their limitations¹³, but should be adaptable to objects of different size and shape.

Here, we illustrate how this method was applied in a recent study of the mouse baculum¹⁴, a bone in the penis that has been gained and lost multiple independent times during mammalian evolution¹⁵. We discuss the dissection and preparation of a specific bone, the baculum (Protocol 1), the generation of microCT images (Protocol 2), and the conversion of these images to a format that enables all downstream computational geometry (Protocols 3 and 4). After these steps, each specimen is represented by ~100K x-y-z coordinates. We then walk through a series of transformations that effectively align all specimens into a common orientation (Protocol 5), then define semi-landmarks from aligned specimens (Protocol 6). Protocols 1-4 should be similar regardless of the object being analyzed. Protocol 5 and Protocol 6 are specifically designed for a baculum, but it is our hope that by detailing these steps, investigators can imagine modifications that would be relevant for their object of interest. For example, modifications of these methods were applied to study whale pelvic bones and rib bones¹⁶.

Protocol

All procedures and personnel were approved by the University of Southern California's Institute for Animal Care and Use Committee (IACUC), protocol #11394.

1. Baculum Dissection and Preparation

1. Euthanize a sexually mature male mouse via carbon dioxide over-exposure, according to protocols set forth by the relevant Institutional Animal Care and Use Committee (IACUC).
2. Lay the animal in the supine position, and protract the penis by applying pressure with thumbs lateral to the preputial opening.
3. Once the penis is protracted, extend the tissue through the prepuce as far as possible.
4. With scissors, cut the penile body proximal to the glans penis where the baculum resides.
5. Transfer the dissected penis to a 1.7 mL tube and add 200 µL tap water. Make certain that the penis is fully submerged in the liquid.
6. Incubate the tissue in water at ~50 °C for 3-5 days.
7. After proper incubation, remove the surrounding tissue from the baculum, using forceps under a dissection microscope. Gently squirt 70% ethanol to push off remaining tissue and clean the bone.
8. Place the dissected baculum into a new microcentrifuge tube with the cap open. Leave cap open O/N to dry out bone.

2. MicroCT Scanning

1. Press a microCT scan cylindrical holder into a brick of florist foam to create a cylinder of florist foam.
2. Extract the cylinder of florist foam and cut slices ~2-5 cm thick.
3. Push dried bacula into the florist foam, around the periphery of an individual slice to minimize interference during scanning. The precise orientation of bones should be remarked allowing for proper identification of individual specimens in Protocol 4.
4. Gently place the slice with the embedded bones into the microCT holder.
5. Acquire microCT scans. In the case of mouse bacula¹⁴, we used a uCT50 scanner (Scanco Medical AG, Bruttisellen, Switzerland) at the USC Molecular Imaging Center under the following settings: 90 kVp, 155 µA, 0.5 mm Al filter, 750 projections per 180 (360 coverage), exposure time of 500 ms, and voxel size of 15.5 mm.

3. MicroCT Processing: Converting a .DCM Stack to a Single .xyz File

NOTE: Each microCT scan produces a stack of .DCM, or "dicom", files that represent image slices taken through the object. All downstream computational geometry requires flat .xyz files, which is simply a text file that contains four columns – the x, y, and z coordinates of each pixel, and the intensity of the pixel, ranging from -5,000 (black) to +5,000 (white). A pixel threshold above 3,000 generally works well as a threshold for defining bones.

1. Install Python (www.python.org) and the PYTHON modules COMMANDS, DICOM, PYLAB, SYS, and NUMPY.
2. Open "01_process_dicom.py"{Figshare} with any text editor. Under the Variables section, change path, pixel thresholds, and directory names as necessary.
3. Run "python 01_process_dicom.py". Progress will be printed to screen. Within each directory named in step 3.2, a new file is made named; for example, directory_name.PT3000.xyz, where PT3000 indicates the pixel threshold indicated in step 3.2.

4. MicroCT processing: Segmenting-out Individual Specimen .xyz Files

1. Install R (https://www.r-project.org/) with the library RGL.
2. Open the file '02_segment_dicoms.r'{Figshare} with any text editor. Under the Variables section, change the path name to point to the .xyz file created in Protocol 3 above.
3. From within R, run the command "source('02_segment_dicoms.r')" (without the double quotes).
4. After the three dimensional image of the .xyz file created in Protocol 3 appears, enter the number of specimens in the overall .xyz file. Then label and select the points from each specimen using the scroll and zoom functions.
   NOTE: In the background, separate .xyz files will be made for each specimen. These appear in a directory named, for example, XYZ_FILES_PT3000, where PT3000 indicates the pixel threshold used.

5. "Aligning" Specimen .xyz Files to Common Coordinates.

1. Open the Python script "03_transform.py"{Figshare}, which requires the additional module mattdean_modules.py{Figshare}, as well as two self-standing applications: "rotate_translate_cylindrical" (https://github.com/timydaley/dean_cylindrical_tranform) and “qconvex” (www.qhull.org/html/qconvex.htm) that are used by this script.
2. Under the Variables section, identify the full path names to mattdean_modules.py, rotate_path and qconvex_dir. In addition, identify the full path to the directory containing the individual .xyz files created in step 4.
3. Run 03_transform.py, which creates a new file per specimen with the suffix .TRANSFORMED.xyz.

6. "Slicing" Aligned Specimen .xyz Files to Identify Semi-landmarks.

1. Open and run the Python script "04_identify_landmarks.py"{Figshare}. In the Variables section, identify full path names to the directory containing the .TRANSFORMED.xyz files. This script identifies 802 semi-landmarks that can be used to quantify size and shape of the structure.

Results

The x-y-z coordinates of the semi-landmarks produced in Protocol 6 can be directly imported into any landmark-based geometric morphometrics analysis¹⁷. The computational pipeline above has been applied to study mouse bacula¹⁴, as well as whale pelvic and rib bones¹⁶. More details on the computational definition of semi-landmarks are presented here, in an attempt to help researchers visualize steps that might be modifi...

Discussion

The critical steps in the above protocol are 1) dissecting the bacula, 2) gathering the microCT images, 3) converting the microCT output to a flat file of x-y-z coordinates, 4) segmenting out each specimen's point cloud, 5) transforming each specimen to a standardized coordinate system, and 6) defining semi-landmarks. These steps are easily modified to accommodate different objects.

These methods can likely be applied to any object that is essentially "rod-shaped", or at least not ...

Materials

List of materials used in this article

Name	Company	Catalog Number	Comments
Dissecting scissors	VWR	470106-338	Most sizes should work
Dissecting Forceps, Fine Tip, Curved	VWR	82027-406
1.7 mL microcentrifuge tube	VWR	87003-294
Absolute Ethanol	Fisher Scientific	CAS 64-17-5	To be diluted to 70% for dissections
Floral Foam	Wholesale Floral	6002-48-07
uCT50 scanner	Scanco Medical AG, Bruttisellen, Switzerland