Spinal cord samples were collected opportunistically from odontocetes that stranded in good body condition (not emaciated) with a Smithsonian or National Stranding Code ≤2 (a fresh dead carcass) (Geraci & Lounsbury, 2005). Specimens included 10 odontocetes from five different species (two individuals of each species), and two pigs (Table 1—listed by shallow divers, deep divers, and terrestrial species) (Piscitelli et al., 2010). For the odontocetes, necropsies were performed, and the vertebral column was removed, wrapped tightly in plastic bags to avoid desiccation, and stored at −20°C or −80°C until further dissection. The frozen vertebral column was then separated into individual vertebrae using a saw or large knife (Rowlands et al., 2021). Each sample was allowed to partially thaw, and using a scalpel, the spinal cord and meninges were carefully excised from the vertebral canal (Figure 1) and the meninges removed. The spinal cord sample was then cut transversely to collect an approximately 1 cm cross-section, which was placed into a 15 ml cryovial and frozen in a −10°C freezer until further analysis. Samples were variably collected from the cervical, thoracic, and lumbar sections of the spinal cord (Table 1; Figure 2). The exception to this sampling protocol were the pigs. Samples were obtained from a commercial abattoir. Due to health regulations, pig spinal cords were removed from vertebral columns by the commercial processor, and therefore, it was not possible to determine the exact location of samples taken; however, for data analysis purposes, it was assumed pig spinal cord was thoracic. When received, sections of spinal cord were taken equally spaced along the length of the cord.

The frozen, unfixed spinal cord cross-sectional samples from all animals were placed in a Leica Cryocut 1800 (Lecia Biosystems Inc, Buffalo Grove, IL), covered with Optimal Cutting Temperature Compound (Sakura Finetek, Torrence, CA), and allowed to freeze to −27°C. Each spinal cord sample was sectioned to obtain five noncontiguous 20 µm sections, which were placed on gelatin-coated superfrost plus microscope slides (Fisher Scientific, Waltham, MA). All steps following were performed at room temperature (∼20°C). Sections were incubated in BLOXALL Blocking Solution (Vector Laboratories, Burlingame, CA) for 10 min then rinsed in PBS for 5 min. After rinsing, sections were incubated in Normal Horse Serum, 2.5% (Vector Laboratories, Burlingame, CA) for 20 min. The sections were then incubated in CD34 antibody (R&D Systems, Minneapolis, MN) diluted to 1:200 in PBS for 4 h. The sections were then rinsed in PBS for 5 min and incubated in ImmPRESS universal Polymer Reagent (Vector Laboratories, Burlingame, CA) for 30 min. This step was followed by two 5-min rinses in PBS. Afterwards, sections were incubated in ImmPACT DAB EqV (Vector Laboratories, Burlingame, CA) for 5 min. Lastly, sections were rinsed in PBS twice for 5 min and placed under coverslips with trisglycerol.

Stained sections were viewed with an Olympus BX60 brightfield microscope (Olympus America, Center Valley, PA) and digital images were taken of each section using a Diagnostic Instruments SPOT RT digital camera (Diagnostics Instruments, Sterling Heights, MI). All images were analyzed using ImageJ software (National Institutes of Health, Bethesda, MD). A stereological analysis was utilized by overlaying a grid of 5,000 μm² onto the images. Microvascular density (% microvascularity or area of tissue occupied by microvasculature) was calculated by counting microvessels that hit the overlaying grid in five sections of tissue for each sample to get a total count of at least 100 microvessels (following the methods of (McClelland et al., 2012; Gabler et al., 2018). If the total did not reach 100 with five sections, then more sections were analyzed. To quantify the degree of microvascular branching, five microvessels were randomly selected for each of the five noncontiguous samples. The number of terminal branches per microvessel counted for these 25 microvessels were averaged to determine the number of branches per sample. Microvessel diameter was measured with ImageJ software (Shahlaee et al., 2016). Five microvessels from one image were randomly selected and the diameter was determined by tracing the width of the microvessel from one edge to the other (Shahlaee et al., 2016).

Linear mixed models using repeated measures followed with Type III Analysis of Variance with Satterthwaite’s method (to retrieve p values) were run using R version 1.2.5019 (RStudio Inc.—packages lme4 and lmertest) to determine if there were differences in the dependent variables (microvascular density, diameter, and branching) in the three sampling areas of the spinal cords (cervical, thoracic, and lumbar) between the species, and between the different dive behaviors (shallow, deep, terrestrial). The unit of statistical analysis was the average of the dependent variable (density, branching, or diameter) at each location within each individual. The model used to achieve results was as follows: “Parameter (density, branching, or diameter) ∼ Dive Behavior * Location * Dive Behavior:Location + (1/Individual)”. Density, diameter, and branching data were examined for normality using Q-Q plots of residuals versus theoretical quantiles and tested for equality of variance using Levene’s test. If there were significant differences in microvascular density, diameter, or branching between the different sampling areas (along the spinal cord) or dive behavior, on each of the variables, post hoc tests were carried out to determine where differences existed. Linear mixed models with random effects (species) of repeated data from multiple animals and multiple tissues were conducted to determine whether there were differences between spinal cord microvasculature and previously published microvasculature data from blubber (McClelland et al., 2012) and acoustic fat bodies (extramandibular fats covering the mandible, and the intramandibular fats in the mandibular fossa) (Gabler et al., 2018). The model used to achieve results was as follows: “Parameter (density, branching, or diameter) ∼ Location + (1/Species). All tests were evaluated for significance at α = 0.05.

Neither dive behavior (shallow, deep, terrestrial) or spinal cord location (cervical, thoracic, lumbar) had a significant effect on microvascular density (p = 0.09 and 0.14, respectively). There was no interaction between dive behavior and spinal cord location (p = 0.52). Finally, while D. delphis presented the lowest microvascular density and K. breviceps had the highest in all sampling locations, the microvascular density varied little between species, ranging from 9.3% to 11% (Figure 3 and Figure 4).

Between species, microvascular branching was also conserved, ranging from 1.6 to 2.1 branches/microvessel (Figure 3 and Figure 5). Dive behavior and spinal cord location had no significant effect on microvascular branching (p = 0.77 and 0.83 respectively) and there was no interaction between dive behavior and spinal cord location (p = 0.45).

The microvessel diameter for all species was between 11 and 15 µm (Figure 6). Neither dive behavior or spinal cord location had a significant effect on microvascular diameter (p = 0.93 and 0.25 respectively) and there was no interaction between dive behavior and location (p = 0.60).

All blubber data described here are from McClelland et al. (2012) and all mandibular acoustic fat data are from Gabler et al. (2018). In the acoustic fats, microvascular densities ranged from 0.60 to 1.6% in extramandibular acoustic fats and 0.50 to 1.7% in intramandibular acoustic fats. Within the blubber, microvascular density ranged from 1.60 to 3.30% in superficial (nearest the epidermis) blubber and 1.80 to 9.30% in deep (nearest the subdermal connective tissue sheath and muscle) blubber. The spinal cord microvascular density was higher than both blubber and acoustic jaw fats microvascular densities (p= < 0.001) (Figure 7). Microvasculature branching ranged from 1.6 to 2.1 branches/vessel in the spinal cord, and 2.1 to 2.9 branches/vessel in the blubber and acoustic fats, except for the middle and deep blubber layers of T. truncatus having an average of 5.65 branches/vessel (Figure 8); however, there is still a significant difference (p = 0.02) between spinal cord, blubber, and acoustic fats. The microvessel diameter of the spinal cord ranged from 11 to 16 μm, which was larger (p = < 0.001) than the microvessel diameter of the blubber and fat bodies (range from 7.90 to 10.9 µm; Figure 9).

The generally uniform microvascular characteristics between all species, marine and terrestrial, suggests that diving species may not have specialized microvasculature in their spinal cord. Microvessel branching is related to the process of angiogenesis, which is the growth of new blood vessels from pre-existing ones (Baker et al., 2012). Angiogenesis is an intricate process that includes the stimulation of endothelial cells by many different pro-angiogenic factors, including vascular endothelial growth factor (VEGF) and basic fibroblast growth factor (Baker et al., 2012). Expanding blood vessel networks are required to branch frequently to guarantee all tissues receive an adequate blood supply (Ruhrberg et al., 2002). However, these branching networks must obey strict patterning cues to match the architectural designs of the resident tissue (Ruhrberg et al., 2002). The lungs and kidneys best display this pattern demand, in which the juxtaposition of the bloodstream with alveoli or nephric tubules, respectively, is essential for organ function (Abrahamson et al., 1998; Gebb & Shannon, 2000). Thus, it is possible that when blood vessels branch in the spinal cord, they must find pathways around the structures within. Therefore, the similarity in spinal cord microvasculature between species could reflect that the vessels must form around the similar anatomical features of the cord.

The spinal cord is composed of two general regions: the gray matter and the white matter (Evans & de Lahunta, 2013). The gray matter helps to form the core of the spinal cord and is composed of cell bodies, neurons, glial cells, and a limited amount of myelinated axons (Evans & de Lahunta, 2013). On the other hand, the white matter is situated superficially in the spinal cord and is packed with myelinated axons (Evans & de Lahunta, 2013). One of the key differences between the white and gray matter is that the gray matter of mammals, such as the well-studied dog, has a relatively rich blood supply (Evans & de Lahunta, 2013). Perhaps the differences in the organization of the gray and white matter present developing blood vessels with regionally specific patterning rules. Furthermore, the arrangement could be based on the metabolic requirement of cell bodies and axons (Turnbull, 1971). The white matter microvasculature follows the pattern of the axons—following the direction of the nerve fibers, while the gray matter microvasculature relies on the location of the cell bodies. Images were taken during image acquisition to capture microvessels in both white and grey matter by segregating the spinal cord into nine sections (Figure 10). While all images from areas with the white matter and gray matter were consistent in terms of density and branching, gray matter seems to have microvessels with a greater diameter than vessels of white matter (Figure 11; although we were unable to test whether this difference was significant due to not knowing the exact area of white and gray matter). Greater microvessel diameter in gray matter could correlate to the higher metabolic requirement of the gray matter and blood flow (Marcus et al., 1977). In mammals such as dogs and sheep, the vascularity of the gray matter differs from the white matter since the tissue has a larger metabolic requirement that requires a higher blood flow in this region (Marcus et al., 1977). To accommodate this increase in blood flow, the diameter of vessels increases, allowing lower resistance and constant flow (Marcus et al., 1977).

The average diameter of a red blood cell in odontocetes ranges from 7.3 to 10 µm (Lenfant, 1969) and 5.1 to 8.5 µm in pigs (Scarborough, 1931). Therefore, the microvessel diameters within the spinal cord of both odontocetes and pigs are larger than their average red blood cell diameter, which could be an adaptation related to oxygen and nutrient absorption within the spinal cord. The capillaries are where the exchange of oxygen and nutrients between blood and tissue occurs (Yuan & Rigor, 2010). During dives, marine mammals undergo a well-developed dive response characterized by apnea, bradycardia, and peripheral vasoconstriction (Scholander, 1963; Kooyman, 1985; Hooker et al., 2012; Ponganis, 2015). This results in a redistribution of blood flow so that oxygen sensitive tissues, such as the brain and spinal cord, are adequately perfused with blood, while other tissues experience ischemia and hypoxemia (Zapol et al., 1979; Meir et al., 2009). For example, studies on the Weddell seal (Leptonychotes weddelli) show that organs such as the brain and spinal cord maintain or even receive increased blood flow during dives, while organs like the kidneys, spleen, and liver experience decreased blood flow (Zapol et al., 1979). Therefore, the brain and spinal cord receive adequate blood flow during a dive, and the increase in diameter results in lower resistance to further increase the blood flow in these tissues. This allows the tissues like the spinal cord to receive oxygen during times of hypoxia. While terrestrial mammals do not naturally experience hypoxia, it has been shown that hypoxia results in increased blood flow to the brain and spinal cord of terrestrial mammals (Griffiths, 1973; Marcus et al., 1977); mammalian spinal cords may simply be well protected in case of hypoxic events.

The consistency of all microvascular characteristics between species and along the length of the spinal cord parallels a recent study on the lipid composition of spinal cord tissue in similar diving and non-diving species used in this study (Glandon et al., 2021). These authors demonstrated consistency of lipid content and lipid class composition of neural tissues in 12 marine and two terrestrial mammalian species. The five classes of lipids in neural tissue (in terrestrial mammals including dogs and humans and the odontocetes used in Glandon’s study) include sphingomyelin, sulfatides, cerebrosides, phospholipids, and cholesterol. Additionally, the spinal nerve tissue includes triacylglycerol and wax ester/sterol ester (Glandon et al., 2021). Lipid class composition was not significantly different between shallow and deep divers, despite clade-specific differences in the classes of lipids in other tissues, such as blubber (Pond, 1998; Koopman, 2007). Lastly, there was no significant difference in lipid composition across sampling location along the length of the spinal cord (Glandon et al., 2021). Thus, at least from the lipid and microvascular perspectives, the spinal cord appears to be a very conserved tissue across mammals.

Despite the similarity in microvascular density within the spinal cord, there is significant variation of vascularity between different tissues within odontocetes. The microvascular characteristics have been studied within the acoustic fats and blubber of odontocetes (McClelland et al., 2012; Gabler et al., 2018). A possible explanation for these differences between tissues is that the spinal cord must have higher vascularity to allow for adequate oxygenation of the central nervous system at all times, especially during times of stress on the body (Griffiths, 1973; Marcus et al., 1977; Williams & Davis, 2020).

Preserving blood flow to critical body tissues is supported in diving mammals through the dive response which causes redistribution of blood flow during dives towards oxygen-sensitive tissues (i.e., the central nervous system) and ischemia to hypoxia tolerant tissues (i.e., blubber) (Zapol et al., 1979). The acoustic fats, however, likely sustain perfusion during dives to maintain thermal gradients and acoustic function, though the amount of flow and its consistency is unknown (Houser et al., 2004). Additionally, terrestrial mammals experiencing hyperthermia maintain blood flow to oxygen-sensitive tissues (i.e., the central nervous system) while blood flow to regions of the skeletal muscle and adipose tissue is reduced significantly (Hales et al., 1979). Cerebral autoregulation is not altered while blood flow in other tissues decrease significantly (Ogoh et al., 2013). Thus, increased vascularity of the central nervous system in both marine and terrestrial mammals can aid in autoregulating blood flow to the central nervous system when experiencing environmental stressors. While there are differences in the microvascular density of the various tissues of marine mammals, the branching and diameter are generally consistent between tissues.

Nitrogen gas is approximately five times more soluble in fat than other tissue of the human body (Ikels, 1964). Therefore, since there is more area for gases to diffuse, at most blood/tissue interfaces nitrogen gas will diffuse into the fat, if the partial pressure gradient supports it. This interface occurs at the level of microvasculature (Yuan & Rigor, 2010). Why then have marine mammals not developed their microvasculature as an extra defense against gas embolism?

In diving mammals, the primary determinant of how high the partial pressure of nitrogen rises in blood is determined by the depth of alveolar collapse (Zimmer & Tyack, 2007). The dissolved nitrogen will be distributed throughout perfused tissues based on partial pressure gradients. Bubble nuclei might always be present in diving mammals, but the risk of DCS is dependent on the volume and size of bubbles, resulting in gas supersaturation (Zimmer & Tyack, 2007; Bostrom et al., 2008; Fahlman et al., 2009). The depth of alveolar collapse is dependent on the dead space in the lungs in relation to the total volume of the lung during the dive, thus, different lung volumes will alter the depth of alveolar collapse and produce different equilibrium nitrogen tensions and finally, have an overall influence on the size of any bubbles formed (Zimmer & Tyack, 2007).

It is also proposed that some deep divers, such as the beaked whales, may tolerate sporadic nonlethal bubbles and that normal dive profiles for diving animals show low risk of acute and severe DCS (Zimmer & Tyack, 2007). It has been hypothesized that DCS would be most likely to occur from repetitive shallow dives—dives that are too shallow for alveolar collapse (Zimmer & Tyack, 2007; Fahlman et al., 2009). These cases with limited surface durations and repetitive short dives—correlated to avoidance behavior of acoustic events—are associated with the central circulation and the brain (Frantzis, 1998; Houser et al., 2001; Fernández et al., 2005; Zimmer & Tyack, 2007). This can be seen by looking at the diffusion halftimes of these tissues (Zimmer & Tyack, 2007). The central circulation tissues and brain may exceed their long-term equilibrium tensions during descent which may increase the opportunity for bubble growth at the surface (Zimmer & Tyack, 2007). Despite millions of years of potential selective pressure on the microvasculature and lipids of the cetacean spinal cord, we did not find any evidence of such adaptations for diving in this tissue. We propose that there are physiological limits and trade-offs to the spinal cord, because of the need to deliver blood flow consistently to this oxygen-sensitive tissue, that can override even strong evolutionary pressures associated with diving.

With the lack of microvascular adaptations within the spinal cord, diving vertebrates must avoid gas embolism by minimizing nitrogen uptake and its movement. This can be done by 1) diving with a low diving lung volume, 2) alveolar collapse, and 3) cardiovascular management of systemic blood flow (Fahlman et al., 2021).