All studies were performed on an approved protocol in accordance with the Boston College Institutional Animal Care and Use Committee (IACUC; Protocol #2020-005). Zebrafish were reared at 28°C on a 14:10 light:dark cycle and fed a diet of marine rotifers and adult pellet food flakes three times a day. Zebrafish were of the genetic background Tg(tg:nVenus-2a-nfnB) ^wp.rt8 (McMenamin et al., 2014) and originated from several matings of the same parental breeding stock. Individuals were treated with 1% DMSO at 4 dpf, which does not activate the transgenic nitroreductase system (McMenamin et al., 2014). To ensure that these individuals were representative and that the transgenic background or the DMSO-treatment did not cause gross skeletal mis-patterning, we scanned representative stages from the Tübingen wild-type line for comparison. The two strains were overall comparable morphologically and in terms of relative density (see Supplementary Figure S1).

Capturing individuals at a range of developmental stages requires precise and repeatable methods for measuring development. Days post fertilization is an unreliable measure of developmental progress in zebrafish, particularly during later stages of development (Parichy et al., 2009; McMenamin et al., 2016). We used SL as a proxy for development, and samples were measured both before scanning and in the scans themselves (see Supplementary Figure S2). Although staging according to the postembryonic normal table is likely a more accurate measurement of developmental progress than length (Parichy et al., 2009; McMenamin et al., 2016), we chose to use fixed SL because it is a continuous and quantitative proxy for development which may be easily obtained from scans. On average, most of the 12 mm SL fish were 2 months old; 16 mm SL fish were between 4 and 5 months age and 24 mm SL fish were 6–9 months old when fixed. The sex of the individuals was also recorded when it was possible to discern sex, starting approximately around 17.5 mm SL (Supplementary Figure S3).

Fish were euthanized by MS-222, and fixed in 4% paraformaldehyde for 24 h. SL was measured in fixed samples according to Parichy et al. (2009) before scanning, and was also measured digitally in the scans themselves (see Supplementary Figure S1). Fish samples shrink slightly during the fixation process; note that the reported fixed SL values may be converted to corresponding “fresh” SL by adding 0.29 mm (Parichy et al., 2009). A total of 62 specimens were scanned, ranging from 12 to 25 mm SL with a minimum of one scan for every half millimeter. Fixed specimens were placed in low-density foam molds and inserted into either a 1.5 ml centrifuge tube (for specimens 12–14 mm SL) or a 15 ml conical tube (for specimens >14 mm SL). Scans were performed on a SkyScan 1275 high resolution microCT system (Bruker, Kontich, Belgium) at a scanning resolution of 10.5 μm with an x-ray source voltage of 45 kV and current of 200 mA. Projection images were generated over 360° with a 0.1° rotation step and 6 averaging frames. Thresholding, ring artifact reduction, and beam hardening corrections were consistent across all scans during reconstruction using NRecon (Bruker, Kontich, Belgium).

Cross section images were generated in the open source software 3D slicer (Kikinis et al., 2014). Vertebral diameter measurements were taken using Amira 6.5 (Thermo Fisher Scientific FEI, Hillsboro, Oregon, United States) using the orthoslice module to view the transverse cross section and the canal width was quantified with the 2D measurement tool, measuring the diameter of the vertebral canals of all rib bearing vertebrae in zebrafish at four representative sizes (12, 16, 20, and 24 mm SL). Multi-level modeling was performed with pairwise post-hoc analysis to determine significant differences in vertebral diameters at each vertebra among the four standard length sizes. These measurements could alternatively be made using 3D Slicer.

Relative density heatmaps were generated with the volume rendering module and physics load transfer function in Amira with a threshold range of 20–120. Mean gray value was also used to show relative density between scans. Mean gray value was calculated from imported cross section slices using the measurement tool in ImageJ (Version 1.8.0_172, National Institutes of Health, Bethesda, Maryland, United States). Volume measurements were taken with the Material Statistics module in Amira.

Individual bones were segmented using the Segmentation Editor in Amira 6.5 (Thermo Fisher Scientific FEI, Hillsboro, Oregon, United States). Briefly, the entire scan volume was loaded into the program and a pixel threshold was determined to differentiate bone from soft tissue. The lasso tool was then used to select the corresponding pixels of a specific skeletal element and added to the appropriate material label. Segmented bones include the basibrachials, branchial arches, basioccipital, dentary, dermatocranium, ectopterygoid, exocciptal, entopterygoid, hyoid, hyomandibula, infraorbital, interopercle, kinethmoid, lateral ethmoid, maxilla, metapterygoid, opercle, orbitosphenoid, pharyngeal jaws, premaxilla, preopercle, parasphenoid, quadrate, supracleithrum, supraoccipital, subopercle, and supraorbital. Pearson correlation coefficients were calculated to show the correlation between density or volume and SL.

3D models of the microCT reconstructed scans were generated in Amira 6.5 using the Segmentation Editor and Generate Surface module (Thermo Fisher Scientific FEI, Hillsboro, Oregon, United States). Meshes were simplified using MeshLab (Cignoni et al., 2008; Callieri, 2013). These models were converted to .u3d files and imported as interactive 3D PDF using Acrobat Pro DC (Version 2021.005.20058, Adobe Inc., San Jose, CA, United States).

Whole, raw microCT scans for individuals from every half mm SL are available for download (see Table 1; MorphoSource project URL https://www.morphosource.org/projects/000415918?locale=en). When multiple scans were available for each size category (see Supplementary Figure S3 and Supplementary Table S1), we selected the highest quality scan for upload to MorphoSource. Additionally, individuals of four representative sizes (12, 16, 20, and 24 mm SL) were used to generate 3D PDFs (see Table 1; Supplementary Figures S5–S8). These interactive PDFs can be viewed with any standard PDF viewer, including Adobe Acrobat Reader (Adobe Inc., San Jose, CA, United States). These 3D PDFs allow users to turn, rotate, and zoom in to the embedded 3D models.

The small size of the zebrafish can pose a barrier to measuring small anatomical elements in 3D. However, microCT technology allows visualization and analysis of elements of interest. MicroCT scans generate cross sections that can be accessed using a variety of programs such as DataViewer (Bruker, Kontich, Belgium), Amira (Thermo Fisher Scientific FEI, Hillsboro, Oregon, United States) or ImageJ (National Institutes of Health, Bethesda, Maryland, United States). Any of these programs will allow a user to scroll through the stacks of cross-section images from the scans in any anatomical plane (e.g., see Figures 1A–D). These cross-sections capture details at a resolution of 10.5 μm, which allows anatomical measurements even in relatively small bones. To test these types of measurements, we focused on the morphological changes of vertebrae along the anterio-posterior axis. We examined sagittal cross sections (as in Figure 1D) from scans of adult zebrafish at four representative sizes (12, 26, 20 and 24 mm SL), measuring the diameter of the vertebral canal of vertebrae 2 through 10 (the rib-bearing vertebrae; Figure 1E). These widths increase markedly in more posterior vertebrae (Figure 1E).

MicroCT datasets can be used to determine relative density. Our samples were all scanned under consistent parameters, so density can be directly compared between scans. We hypothesized that overall skeletal density would continue increasing throughout stages of adult growth. Indeed, density (as measured in mean grey values) increased markedly with increased size; regions of increasing density were particularly notable in the dermatocranium, ribs, and hypural complex (Figure 2A, Supplementary Videos S1, S2). Quantifying overall density of the skeleton as a function of body size (SL), we found that relative density increases roughly linearly throughout juvenile and adult development (Figure 3A). We next asked how density was distributed along the anterio-posterior axis of the skeleton, and whether such patterns change with growth. We found that density was highest in anterior regions of the body, corresponding to the craniofacial skeleton (Figure 2B). The high density of the head corresponds to the many plate bones in this region. The head also contains three pairs of otoliths; these dense, highly mineralized bony elements are used for hearing and vestibular function (Vasconcelos-Filho et al., 2019), and contribute to the overall density of the head. We note that while maximum density increases in increasingly large individuals, the distribution of density across the skeleton remains largely consistent (Figure 2B). Overall, we also see the same patterns of density distribution when comparing density to distance from the anterior portion of the zebrafish, and when normalizing density to the proportion of body length (Supplementary Figure S4).

MicroCT scans can be used to calculate the volume of tissues within a specified density range. Volumetric renderings of the skeleton highlighted the new appearance of bones in the skull and fins as fish continue to grow (Figure 2C; also see Figures 4A,B). We further quantified the changes in overall skeletal volume, finding a roughly exponential increase in bone volume (a 3D measurement) with linear fish growth (a 2D measurement; Figure 3B).

3D models can be digitally segmented into individual elements. We segmented an adult skull into the 74 component bones. This segmented model captures the association of each element in 3D space and captures the anatomy of the adult craniofacial skeleton (Figure 4C). After segmentation, bones can be examined individually. To visualize how an individual bone changes shape as development progresses, we “virtually dissected” the lower jaws and caudal vertebrae from fish at a range of sizes (Figures 5A,B). We note that as adult zebrafish continue to grow, the anguloarticular prominence of the lower jaw becomes considerably more pronounced and the posterior end of the jaw widens (Figure 5A), while the caudal vertebrae do not undergo significant shape change during juvenile and adult development (Figure 5B).