Specimen information
The imaged coelacanth specimen is an adult male West Indian Ocean coelacanth, Latimeria chalumnae Smith, 1939, CCC 23, fished at 250 m of depth and 700 m off the coast between Iconi and Moroni, Grande Comore (Ngazidja) at 1:00 h on June 23, 1960. Total length was 130 cm. The specimen was reportedly only preserved in 70% ethanol without the use of formaldehyde [43]. Since this was unconventional for the era, this was tested using a formaldehyde assay test kit (see chemical analysis section below). The left side of the abdomen has been opened with a longitudinal cut, and in the head, one gill arch has been bent; however, all inner organs remain intact (see Additional files 7 and 8). The specimen was received as a gift on June 16, 1962, at the Zoological Museum, University of Copenhagen, Denmark, from the Muséum National d'Histoire Naturelle, Paris, France. It has been on display at Danmarks Akvarium and later on the Natural History Museum of Denmark, but is currently stored in the collection (catalog number ZMUC P1112). For all imaging procedures, the coelacanth specimen was placed on its left side (same position as it was presumably fixed in); however, to better display chemical content values in a typical snout-to-tail/left-to-right reading direction, datasets were flipped along the horizontal axis to achieve this. This has no effect on any calculations. Additional species were from the collection of the Natural History Museum of Denmark: Neoceratodus forsteri (Krefft, 1870) (ZMUC Journal# 4), Lepidosiren paradoxa Fitzinger, 1837 (ZMUC P1124), Protopterus annectens (Owen, 1839) (ZMUC P1122), Dissostichus eleginoides Smitt, 1898 (ZMUC P63215), Hoplostethus atlanticus Collett, 1889 (ZMUC P40335), Spectrunculus grandis (Günther, 1877) (ZMUC P77701), Myctophum humboldti (Risso, 1810) (ZMUC P2397221), Thalassobathia pelagica Cohen, 1963 (ZMUC P77853), Hexanchus griseus (Bonnaterre, 1788) (uncat.), Typhlonus nasus Günther, 1878 (ZMUCP77447), Solivomer arenidens Miller, 1947 (ZMUC P2341238), Neopagetopsis ionah Nybelin, 1947 (ZMUC P7641), Aeoliscus strigatus (Günther, 1861) (ZMUC P39373), Beryx decadactylus Cuvier, 1829 (ZMUC P4013), and a beached Mola mola (Linnaeus, 1758) (uncat.). Nine specimens of Sparus aurata Linnaeus, 1758 (uncat.) were purchased fresh and ungutted from a local fish market for the chemical analysis and validation of methods for bone mineral and lipid content measurements. Putative time calibrated phylogeny of extant non-tetrapod sarcopterygians was acquired from TimeTree (http://timetree.org/ accessed on June 22, 2021).
Imaging
For lipid/water mapping, MRI was performed on an Siemens Magnetom Skyra system equipped with a row of coils in the scanner table and two Siemens Body 18 and one Flex Large 4 surface coils using a Dixon flash 3D sequence with the following parameters: field strength: 3 T, repetition time = 5 ms, echo time = 1.23 ms, flip angle = 10°, field-of-view = 423.3 × 500.0 × 318.1 mm3, spatial resolution 1.42 mm isotropic, number of averages = 3, acquisition time = 1.4 h pr. scan. Four scans were performed to image the entire specimen, and these were subsequently concatenated using ImageJ 1.50e. An ethanol dilution series of 0–100% v/v ethanol in distilled water in steps of 10% v/v was imaged using the same Dixon flash 3D sequence for subsequent lipid signal correction of the effect of tissue ethanol. The same MRI system quipped with a Flex Small 4 surface coil was used to acquire high-resolution images of the vestigial lung and stomach object using a T2-weighted spin echo sequence with the following parameters: field strength: 3 T, repetition time = 1000 ms, echo time = 139 ms, field-of-view = 134 × 134 × 67.2 mm3, spatial resolution 0.35 mm isotropic, number of averages = 4, acquisition time = 4 h pr. scan.
Quantitative CT was performed on the coelacanth specimen using a Siemens Somatom Definition Dual Energy system with the following parameters: X-ray tube voltage = 120 kVp, X-ray tube current = 428 mA, integration time = 1000 ms, field-of-view = 102.0 × 102.0 × 207.6 mm3, spatial resolution = 0.2 mm isotropic, convolution kernel = B45s, acquisition time = 90 s pr. scan. A total of 53 tile scans of small portions of the coelacanth specimen were performed and these tiles were subsequently combined using ImageJ 1.50e creating a dataset covering the entire specimen at a high resolution. A Mindways QCT Pro bone mineral calibration phantom placed underneath the specimen was used to calibrate X-ray attenuation values to bone mineral density (mg/mm3 equivalent aqueous K2HPO4). An ethanol dilution series of 0–100% v/v ethanol in distilled water in steps of 5% v/v was imaged using the same imaging protocol together with a sample of the preservation liquid for the coelacanth specimen. Additional CT scans were performed on species with relevance for the study on either the same CT system with similar parameters but varying resolution depending on sample size (Protopterus annectens, 0.39 × 0.39 × 0.6 mm3; Dissostichus eleginoides, 0.94 × 0.94 × 0.6 mm3; Neopagetopsis ionah, 0.5 mm isotropic; Mola mola, 0.97 × 0.97 × 0.6 mm3; Sparus aurata, 0.6 mm isotropic) or another CT system (Toshiba Aquillon Prime SP) with similar parameters but varying resolution depending on sample size (Neoceratodus forsteri, 0.6 × 0.6 × 0.5 mm3; Lepidosiren paradoxa, 0.2 mm isotropic; Hoplostethus atlanticus, 0.4 mm isotropic; Spectrunculus grandis, 0.3 mm isotropic; Thalassobathia pelagica, 0.3 mm isotropic; Hexanchus griseus, 0.6 × 0.6 × 0.5 mm3; Typhlonus nasus, 0.3 mm isotropic; Beryx decadactylus, 0.5 mm isotropic) or for the smaller species (Myctophum humboldti, Solivomer arenidens, Aeoliscus strigatus) a high-resolution clinical extremity CT system (Scanco Medical XtremeCT; Scanco, Brüttisellen, Switzerland) with the following parameters: X-ray tube voltage = 59.4 kVp, X-ray tube current = 119 µA, integration time = 132 ms, field-of-view = 70 × 70 × 150 mm3, spatial resolution = 0.082 mm isotropic, acquisition time = 1.5 h pr. scan.
Photogrammetry of the right side of the coelacanth specimen was performed using 48 photos at 6000 × 4000 px2 resolution captured at different angles using a Canon DSLR camera. Photos were assembled to a three-dimensional model using Autodesk Recap Photo.
Nuclear magnetic resonance spectroscopy
1H nuclear magnetic resonance spectroscopy was performed using both single-voxel spectroscopy at selected organs (fatty organ, post ocular, muscle, notochord) as well as chemical shift imaging across the entire coelacanth body at slices spaced 51 mm apart (16 anteriormost slices) or 102 mm apart (2 posteriormost slices). The same imaging system as for MRI, a Siemens Magnetom Skyra system equipped with body and surface coils, was used for both single-voxel spectroscopy and chemical shift imaging. Single-voxel spectroscopy was performed using a PRESS sequence with the following parameters: field strength: 3 T, repetition time = 2000 ms, echo time = 33 ms, bandwidth = 1200 Hz, flip angle = 90°, voxel size = 10 × 10 × 10 mm3, number of averages = 128, acquisition time = 256 s. Acquisition was performed both with and without water suppression. Chemical shift imaging was performed using a chemical shift imaging spin echo sequence with the following parameters: field strength: 3 T, repetition time = 1500 ms, echo time = 35 ms, bandwidth = 1200 Hz, flip angle = 90°, field-of-view = 180 × 180 × 10 mm3, spatial resolution 11.25 × 11.25 × 10 mm3, number of averages = 4, acquisition time = 1536 s. Acquisition was performed both with and without water suppression. At each spectroscopy step, an anatomical reference image was acquired using the Dixon sequence described above.
For prediction of 1H nuclear magnetic resonance spectra for ethanol and oleyl oleate, molfiles of each molecule were downloaded from PubChem (https://pubchem.ncbi.nlm.nih.gov), imported into nmrdb (https://www.nmrdb.org) and their spectrum were simulated using parameters as similar as for the magnetic resonance spectroscopy acquisition: spectrometer frequency = 100 MHz, line width = 1 and 5 Hz, range = 0–6 ppm. Labile protons, e.g., OH, are not predicted in nmrdb, thus no prediction of the hydroxy group in ethanol was expected.
Chemical and physical analyses of tissue
It was imperative to keep the coelacanth specimen intact; thus, in addition to the non-invasive imaging procedures undertaken on the specimen in this study, only a small ventral muscle sample was taken for chemical analysis.
Lipid extraction was performed to measure lipid fraction of the ventral muscle sample of the coelacanth specimen using a chloroform-free lipid extraction kit validated to perform equally well as the Folch method of lipid extraction (ab211044, Abcam). The extracted amount of lipid of the ventral muscle sample of the coelacanth was insufficient to perform temperature-dependent density and solubility testing; thus, pure oleyl oleate (Sigma-Aldrich O3380), the dominant wax ester found in coelacanth tissue [34], and Hoplostethus atlanticus, extracted swim bladder lipid which is dominated by wax esters quite similar to those of the coelacanth [41], were taken as proxies for physical behavior of lipids found in the coelacanth. Average density of oleyl oleate and the extracted Hoplostethus atlanticus swim bladder lipid at different temperatures was measured by adjusting lipid temperature on a thermoshaker and then pipetting out and weighing subsamples of known volume. Measurements were pressure adjusted by assuming similar physical behavior as wax esters found in the marine copepod Calanus plumchrus Marukawa, 1921 that has been measured in a range of deep sea relevant pressures [82]. Temperature, salinity, and depth information in the habitat of the coelacanth was taken from reference values [42]. Comoran seawater density at varying temperature, salinity and pressure was calculated according to standard procedures [83].
Tissue ethanol and formaldehyde concentration was measured in the coelacanth muscle sample with colorimetry kits (Formaldehyde: K-FRHYD 01/21 and Ethanol: K-ETOH 08/18, both produced by Megazyme) using catalyst-mediated formation of the reduced form of nicotinamide adenine dinucleotide that can be measured by the increase in absorbance at 340 nm. To compare tissue formaldehyde concentration in the coelacanth with concentrations found in fish tissue after different preservation methods, the nine Sparus aurata specimens were preserved using three different methods: physical preservation by freezing (n = 3), chemical preservation using phosphate buffered formalin (4% v/v formaldehyde) fixation followed by storage in 70% v/v ethanol (n = 3), or another method of chemical preservation by direct immersion in 70% v/v ethanol followed by storage in 70% v/v ethanol together with formalin fixed specimen (n = 3). Tissue formaldehyde concentration was measured in muscle samples from these physically and chemically preserved Sparus aurata specimens using the same colorimetry kits as for the coelacanth sample.
The effect of preservation method on bone mineral content was measured by non-destructive quantification using quantitative CT on the nine Sparus aurata specimens before and after preservation and thereafter validated to actual amount of bone mineral using ashing (12 h at 90 °C then 72 h at 580 °C) of homogenized samples. The effect of preservation method on lipid content was measured by performing lipid extraction on the nine Sparus aurata specimens after homogenization.
Solubility of pure oleyl oleate, extracted swim bladder lipid of Hoplostethus atlanticus and extracted full body lipid of Sparus aurata (in contrast to the wax esters of coelacanth and Hoplostethus atlanticus, lipids of Sparus aurata are dominated by triglycerides and phospholipids [84], thus representing other types of lipids also found in the coelacanth although at a lower concentration than wax esters [34]), in 70% v/v EtOH at varying temperatures from 20 to 40 °C in steps of 5 °C was measured by incubating lipids with 70% v/v EtOH in separate containers on a thermoshaker using 3 h of shaking (500 RPM) and 1 h of rest pr. temperature level. Samples of the lipid mixtures were drawn at each temperature and dried (6 h at 78 °C and 12 h at 99.9 °C). Dissolved lipid content of coelacanth, Hoplostethus atlanticus and Sparus aurata 70% v/v ethanol storage solutions was measured by drying (6 h at 78 °C and 12 h at 99.9 °C) a portion of each storage solution and subsequently performing lipid extraction to remove debris and dissolved proteins.
Modeling of buoyancy and hydrostatic balance
Acquired CT and MRI data in 16-bit color depth was initially reformatted into 32 bit for precision of subsequent calculations. Thereafter, CT data outputted in Hounsfield units was converted into bone mineral density using the calibration phantom and lipid- and water-specific images of the MRI data were converted into lipid and water fraction images. Quantitative CT data was then adjusted for the effect of ethanol preservation on measurements of bone mineral content, and MRI data was adjusted for the effect of ethanol preservation on measurements of lipid content. Both datasets were cropped in each dimension to only include specimen voxels and background pixels were removed using the ImageJ Threshold plugin. Both the CT and the MRI datasets were segmented into 35 segments: main body, fatty organ, left/right pectoral fins, left/right pelvic fins, 1st and 2nd dorsal fins, anal fin, and 26 sequential caudal segments. For each segment, average proportions of bone mineral, lipid, and water (lean tissue) was measured in every slice in the transversal, sagittal, and coronal plane, respectively, using the ImageJ multi-measure function. While both CT and MRI datasets where isotropic and covered the entire coelacanth specimen, image resolution was 357.9 times higher in the CT dataset (0.2 mm vs. 1.42 mm in each spatial dimension), thus resulting in 7.1 times the number of slices spanning the specimen in each spatial dimension. To allow for the combination of bone mineral content measurement from CT with lipid and water content from MRI, each slice measurement of bone mineral content was scaled by the number of slices in the same spatial dimension of the MRI dataset and then binned by the number of slices in the CT dataset, thus resulting in slice measurements in equal numbers.
To measure absolute mass of bone mineral, lipid, and lean tissue for each segment of the specimen, slice-wise proportions of each substance was multiplied by the modeled density of the substance at the given depth (0, 190, 400, and 1000 m). Lean muscle density [58] was adjusted to pressure and temperature by the same factor as sea water. To measure absolute buoyancy for each segment, the volume displaced by the segment was multiplied by density of Comoran sea water at the given depth.
To measure center of gravity (XCOG) and center of buoyancy (XCOB) for each segment of the specimen in each anatomical axis, moments which is mass (m) times distance (x) was calculated for each slice (i) and combined in each axis (transversal, sagittal, coronal) using the equations:
$${X}_{COG}=\frac{\sum_{i=1}^{N}{(m}_{lipid,i}{+ {m}_{lean,i}+{ m}_{bone mineral,i}) \times x}_{i}}{M}$$
$${X}_{COB}=\frac{\sum_{i=1}^{N}{m}_{displaced seawater,i}{ \times x}_{i}}{M}$$
where N is the total number of slices and M is the total mass of the segment. The mass of displaced seawater was calculated as the density of Comoran seawater at the given depth multiplied by the volume of tissue when corrected for volume changes due to pressure and temperature. After relative mass and the three-dimensional location of the center of gravity and center of buoyancy of each segment was determined, the overall center of gravity and center of buoyancy of the entire specimen was calculated using the segmental method by dividing the sum of segment moments about each axis by the total relative body weight [45] (for detailed calculations see Additional file 9).
Modeling center of gravity and center of buoyancy shift at different body postures was performed as above but with the addition of adjusting the coordinates of the center of gravity and center of buoyancy to a given fin angle using trigonometric functions. Maximum fin angle relative to the coelacanth’s long axis were taken from [44, 85]: pectoral and pelvic fins: ± 120° mobility in the anterior–posterior direction, second dorsal and anal fin: 90° lateral mobility, first dorsal fin: 90° dorsoventral mobility. Caudal and epicaudal fin was assumed to be able to bend to a semicircular shape based on available photographic material [28, 44]. Torque (τ) was calculated from the force of magnitude (F) applied at a distance (r, XCOG to XCOB distance) from the axis of rotation (center of gravity) in an orientation where the arm makes the angle (θ) with the respect to the line of action of the force using the equation:
$$\tau =r\times F\times \mathit{sin}\uptheta$$