Gigantic animal cells suggest organellar scaling mechanisms across a 50-fold range in cell volume

Abstract

The largest cells are orders of magnitude bigger than the smallest cells. Organelle content scales to maintain cell function, with different organelles increasing in volume, length, or number as cells increase in size. Scaling may also reflect functional demands placed on organelles by increased cell size. Amphibians exhibit exceptional diversity in cell size. Using transmission electron microscopy, we analyzed 3 species whose enterocyte cell volumes range from 228 to 10,593 μm³. We show that nuclear volume increases by an increase in radius while mitochondrial volume increases by an increase in total network length; the endoplasmic reticulum and Golgi apparatus, with their complex shapes, are intermediate. Notably, all 4 organelle types increase in total volume proportional to cell volume, despite variation in functional (i.e., metabolic, transport) demands. This pattern suggests that organellar building blocks are incorporated into more or larger organelles following the same rules across species that vary ~50-fold in cell sizes, consistent with a “limited precursor” model for organellar scaling that, in turn, assumes equivalent cytoplasmic concentrations of organellar building block proteins. Taken together, our results lead us to hypothesize that salamanders have evolved increased biosynthetic capacity to maintain functional protein concentrations despite huge cell volumes.

Introduction

Cell size shows spectacular diversity across the tree of life. In both unicellular organisms and multicellular organisms—where cell size shapes the basic building blocks of tissues and organs—cells vary up to 1,000,000-fold in volume (Malerba & Marshall, 2021). As cell size changes, assuming shape remains constant, the surface area to volume ratio (SA:V) scales in predictable ways dictated by the cell’s shape. This scaling of SA:V impacts many aspects of a cell’s biology including movement across the cell membrane, metabolic flux, the reach of microtubules, and nutrient exchange (Marshall, 2020; Marshall et al., 2012).

As cell size increases, organelles must scale to maintain overall cell functionality (Reber & Goehring, 2015). Within the natural range of cell size, scaling of organelles among smaller cells has been shown to follow different patterns: increasing in volume, linear network length, or copy number (Chan & Marshall, 2010). Specific examples include organelles like the nucleus and nucleolus increasing in volume (Jorgensen et al., 2007; Noel et al., 1971), while mitotic spindles and mitochondria increase in network length (Hara & Kimura, 2009; Rafelski et al., 2012) and multiple-unit organelles like peroxisomes increase in copy number (Titorenko et al., 2000). Other organelles of more complex shapes like the endoplasmic reticulum (ER) and Golgi apparatus have been the focus of less research, and their scaling patterns are less clear (Chan & Marshall, 2010; Marshall, 2015). Patterns of organellar scaling at the larger end of the natural range of cell size remain understudied.

Marshall (2020) laid out five mechanistic models as possible explanations for organellar scaling patterns with cell size. In order of simplest to most complex, these are (a) the limiting precursor model—large cells have higher biosynthetic capacity and, thus, make more organellar building blocks, which are incorporated into more/larger organelles; (b) the relative growth model—cells and organelles follow their own independent growth trajectories; (c) the demand-driven model—organelles grow or multiply to meet the demands of their cell’s size; (d) the size measurement model—the size of the cell is measured, and this measurement is used to set organelle size/number; and (e) the programmed scaling model—cells adjust biosynthesis of organellar building blocks to achieve a match between organelle size/number and target cell size. In the absence of mechanistic data, the simplest model that can explain observed patterns in the data is to be favored (Marshall, 2020).

As with an increase in cell volume, an increase in organellar volume also alters organellar surface area in predictable ways that differ based on shape; an organelle shaped like a sphere shows a SA:V ratio that scales as r²:r³, where r is the radius of the sphere. In contrast, an organelle shaped like a tube maintains a SA:V ratio of 1:1 as it grows in length (Chan & Marshall, 2010). These two basic structures account for most organelles and reflect their functions; the nucleus is a large space that houses DNA and its associated molecules, and it tends to be more spherical. The mitochondria weave throughout the cell providing energy, and they tend to be more tubular (Chan & Marshall, 2010; Marshall, 2015). The ER and Golgi apparatus, however, are made up of cisternae, which are flattened membrane vacuoles that share attributes of both structures (Day et al., 2013; Schwarz & Blower, 2016).

In plant species and some fungal species, cell size can be heavily impacted by large fluid-filled vacuoles, which can comprise 30%–90% of the cell volume; additionally, storage plastids also contribute to intracellular volume (Klionsky et al., 1990; O’Connor et al., 2010). The size of these vacuoles can be highly variable and can cause overall cell size to shift dramatically (Chan & Marshall, 2014; Tan et al., 2019); although the cells are enclosed by a cell wall, the walls themselves have the ability to expand and contract rapidly to accommodate changes in volume (Cosgrove, 1993; Marshall et al., 2012). In contrast, animal cells do not contain these storage organelles and plastids (O’Connor et al., 2010; Wise, 2007), meaning their cell volume is almost exclusively derived from cytosol and membrane-bound organelles used for intra- and intercellular functions other than storage. In addition, animal cells closely regulate their osmotic gradient and maintain tight control on their cell size (Alberts et al., 2017; Cooper & Adams, 2022). Thus, animal cells do not undergo rapid size change outside of growth and size reduction associated with the cell cycle, remaining at a more stable equilibrium size than plant cells.

Two animal groups are composed of somatic cells at the farthest end of the cell size spectrum—salamanders and lungfishes. Salamanders are one of the three extant clades of amphibians, which include 8,695 species (Amphibiaweb, 2023) with genome sizes ranging from 0.95 to 120 Gb; because genome size and cell size are correlated, the clade also exhibits a huge range of cell sizes (Gregory, 2023). Amphibians have been used for decades as a model taxon to examine how cell size affects an organism at the tissue and organismal levels (Decena-Segarra et al., 2020; Hanken, 1983; Itgen et al., 2022; Miller et al., 2020; Roth et al., 1994; Womack et al., 2019). We continue to leverage this natural diversity of cell size to quantify how organelles scale with cell volume using transmission electron microscopy (TEM) and stereology (Howard & Reed, 2004; Russ & Dehoff, 2012; Winey et al., 2014). Our focal taxa are as follows: the widely used western clawed frog, Silurana tropicalis (genome size = 1.2 Gb); the northern gray-cheeked salamander, Plethodon montanus (genome size = 35 Gb); and the western waterdog, Necturus beyeri (genome size ~100 Gb based on congeners that range from 80.5 to 120.6 Gb); cell size is ~50-fold larger in N. beyeri than in S. tropicalis. We focus our analyses on the nucleus, mitochondria, ER, and Golgi apparatus. The nucleus is the primary site for DNA and RNA synthesis as well as the cell’s central hub for detecting deformations in cell shape (Venturini et al., 2020). Mitochondria are a network of membrane-bound organelles whose functions include ATP synthesis, apoptosis, and signaling (Brand et al., 2013). The ER is a network of membranous sacs and tubules that functions in lipid synthesis and protein synthesis, modification, and sorting (Schwarz & Blower, 2016). The Golgi apparatus is a collection of stacked membrane-bound organelles that functions in protein sorting and trafficking and carbohydrate and lipid synthesis (Day et al., 2013). Larger cell size increases intracellular trafficking distance as well as increasing per cell demand for ATP, transcript and protein production, and lipid and carbohydrate synthesis (Guo & Fang, 2014); thus, functional demands on all of these organelles are likely impacted by evolutionary increases in cell size. Taken together, our results reveal patterns of organellar scaling that maintain cell functionality at the extremes of animal cell size, which we interpret in light of proposed mechanistic models.

Materials and methods

Tissue sampling, fixation, staining, and imaging

Intestinal tissue was chosen for analysis as it is made up of only four cell types, and 80% of the total cell population is enterocytes, resulting in a relatively homogenous population of cells (De Santa Barbara et al., 2003). Our focal taxa span much of the range of amphibian genome and cell sizes: the western clawed frog, S. tropicalis (genome size = 1.2 Gb); the northern gray-cheeked salamander, P. montanus (35 Gb); and the western waterdog, N. beyeri (~100 Gb). Silurana tropicalis were obtained from a lab-reared colony following standard husbandry conditions and N. beyeri were obtained commercially. Plethodon montanus were field collected between May and August of 2018 in Avery County, North Carolina under the wildlife collection license # 18-SC01250 issued by the North Carolina Wildlife Resources Commission. One individual was sampled per species, and all specimens were euthanized in MS222. Work was carried out in accordance with Colorado State University (P. montanus, N. beyeri) and University of Wyoming (S. tropicalis) IACUC protocols (17-7189A and 20200714DL00443-01, respectively).

Intestinal tissue from each individual was dissected and immersion fixed in 2.5% glutaraldehyde/2% formaldehyde. The tissues then underwent secondary fixation and staining in 1% OsO₄ in a 0.1 M cacodylate buffered solution followed by embedding in PELCO Eponate 12 epoxy (Cushing et al., 2014). Thin sections (60–80 nm) of resin-embedded samples were cut using a Leica UCT ultramicrotome, collected onto Formvar-coated TEM slot grids, and poststained with 2% aqueous uranyl acetate followed by Reynold’s lead citrate.

Sample preparation, fixation, and mounting were done at Colorado State University. The samples were then sectioned, stained, and imaged at the University of Colorado, Boulder Electron Microscopy Services Core Facility. Sections were imaged using a Tecnai T12 Spirit transmission electron microscope, operating at 100 kV, with an AMT CCD digital camera. Silurana tropicalis and P. montanus tissues were imaged at 9,300× direct magnification. Necturus beyeri tissue was imaged at 6,800× direct magnification because the larger cell sizes could not be captured at the higher magnification. All images were evaluated for quality, ensuring intact tissues undamaged by the fixation process.

Stereological approach

Stereology is an unbiased sampling method that uses two-dimensional image sampling protocols to estimate the surface area and volume of three-dimensional shapes (Howard & Reed, 2004; Russ & Dehoff, 2012). Grids are superimposed randomly onto preselected, nonoverlapping TEM images that include structures of interest. Depending on the type of probes used in the grid (i.e., point probes or cycloid probes) and how the user measures the probes’ interactions with the images, the volume, surface area, length, and number of three-dimensional structures can be estimated from the 2D TEM images. We emphasize that stereology does not produce detailed measurements of individual cells; rather, it produces mean estimates from populations of cells. For the individual representing each species, 60 TEM images were randomly selected (each of which encompassed >1 individual cell), and each image was overlayed with a grid of probes using the IMOD image processing program (Kremer et al., 1996). All measurements for the nucleus, mitochondria, ER, and Golgi apparatus were taken using the 3dmod stereology plugin in IMOD (Noske, 2010).

Organelle volume fraction estimation

In stereology, the term “volume fraction” $(V^V)$ refers to the amount of a total volume (i.e., total cell volume) that consists of a component part (i.e., the volume occupied by a specific type of organelle). Volume fractions were measured for the nucleus, mitochondria, ER, and Golgi apparatus. To this end, each of the 60 images for each individual was fitted with a 7 × 7 grid using crosshair probes (points) for a total of 2,940 probes per species. Next, each crosshair was visually identified and manually assigned to one of the following categories: nucleus, mitochondria, ER, or Golgi apparatus. If a crosshair fell on a part of the image that included damaged cellular material or extracellular material, it was categorized as “not in bounds” and was removed from the dataset. The center of the crosshair was used to define the object category, and only one class was permitted per point. A denser grid was used for the Golgi apparatus due to the rarity of the organelle; the same 60 images per individual were overlaid with 18 × 18 grids for a total of 19,440 crosshair probes. The volume fraction $(V^V)$ of each organelle was calculated using the following equation:

where P(ref) = points hitting the reference volume (i.e., the whole cell) and P(Y) = points hitting the subregion (i.e., organelle of interest) (Howard & Reed, 2004; Russ & Dehoff, 2012).

Organelle and cell surface area density estimation

In stereology, the term “surface area density” $(SV)$ refers to the amount of surface area of a focal object per reference volume; here, we calculate the amount of surface area of each specific organelle per unit of cell volume. We also calculate the amount of surface area of the cell itself (i.e., the plasma membrane) per unit of cell volume (i.e., the cell’s SA:V ratio). For measuring surface area densities, the same 60 images per individual were used as for volume fraction estimates. The nucleus, mitochondria, ER, and plasma membrane were measured simultaneously using alternating cycloid probes in a 2 × 2 grid per image for a total of 240 probes. The Golgi apparatus was again measured using a denser grid due to the organelle’s scarcity, using a 6 × 6 grid of alternating cycloids on each image for a total of 2,160 cycloids per species. The cycloids were visually identified and manually assigned as being “in bounds” and “not in bounds” as in the volume fraction estimation. From there, cycloids were marked with intercepts each time a cycloid interacted with the boundary of an organelle or the cell itself; if a cycloid went entirely through an organelle, then an intercept was marked both entering and leaving the structure. Each intercept was then categorized as contacting one of the four organelles or the plasma membrane. Cycloids that were fully inside an organelle and did not contact the outside of the organelle were not marked with intercepts. Surface area density $(SV)$ was calculated using the following equation:

where $pl$ = points per length of the cycloid, $Ii$ = intercepts with the surface of the organelle of interest (both entering and leaving structure), and $Pi$ = number of points falling in the reference volume (i.e., the whole cell) (Howard & Reed, 2004; Russ & Dehoff, 2012).

Organelle surface area to volume ratio estimation

Organelle SA:V ratios for each species were calculated by dividing the estimates for surface area density $(SV)$ by the volume fraction estimates $(V^V)$ for each image. Because organellar structures are so variable in shape (e.g., the nucleus and mitochondria exhibiting spherical vs. tubular network shapes) (Malerba & Marshall, 2021; McCarron et al., 2013), their SA:V ratios are expected to scale in different ways with increases in organellar size (i.e., SA:V for a sphere scales as r²:r³, whereas SA:V for a tube remains 1:1 as it grows in length) (Chan & Marshall, 2010; Marshall, 2020). Thus, by comparing SA:V estimates of organelles across species, we can infer whether the organelle maintains the same shape (i.e., scales as predicted) or changes shape (i.e., displays a different SA:V scaling). For the ER and Golgi, based on their intermediate shapes, the scaling predictions are less obvious. We reiterate that cell SA:V ratio was measured directly as plasma membrane surface area density (i.e., plasma membrane surface area per unit of cell volume, above).

Organelle and cell absolute volume estimation

Nuclear volume was estimated using the Nucleator probe (Gundersen et al., 1988) in the Visiopharm VIS stereology software (version 2017.7). The Nucleator randomly assigns two perpendicular rays that radiate outward from a fixed point in the nucleus, defined by the nucleolus, and uses these rays and their intersection points with the nuclear membrane to estimate mean nuclear volume. One hundred nuclei were analyzed from each species. Mean volume of the nucleus $VN¯$ was estimated using the following equation:

where $ln$ = the length of the rays from the center of the nuclear cross-section to the nuclear periphery (Howard & Reed, 2004; Russ & Dehoff, 2012). Then, using the volume fraction $(V^V)$ estimates for the nucleus (above), cell volume and all other organelle volumes were extrapolated. Similarly, using the surface area density $(SV)$ estimates (above), surface area of the cell and all organelles were extrapolated.

Among-species differences in organelle volume fraction, surface area density, and surface area to volume ratio

We tested for associations between cell size (i.e., species) and (a) organelle volume fraction, (b) organelle surface area density, and (c) organelle surface area to volume ratio using ANOVA, followed by a pairwise Tukey post hoc method of comparison; each image/sampling grid is one data point. We carried out all analyses in RStudio (Rstudio Team, 2021) using R packages emmeans (Lenth, 2021), car (Fox & Weisberg, 2018), and lme4 (Bates et al., 2015).

Results

Organelle volume fraction

The nucleus made up the largest volume fraction $(V^V)$ of the cell in all three species (nuclear $V^V$ = 0.27 for both S. tropicalis and S. beyeri, nuclear $V^V$ = 0.36 for P. montanus) (Figure 1, Table 1). There was no statistically significant difference in nuclear volume fraction across the three species based on ANOVA (Table 1). The mitochondria made up the second largest volume fraction in all three species (0.09–0.12) and showed no statistically significant difference across the three species (Figure 1, Table 1). The ER and Golgi apparatus made up the third and fourth largest volume fractions in all three species, respectively (0.04–0.07 for the ER, 0.002–0.003 for the Golgi), and neither showed any statistically significant differences across species (Figure 1, Table 1).

Surface area density of organelles and the cell

The mitochondria had the highest surface area density $(SV)$ among the four organelles in all three species and showed no statistically significant difference across species based on ANOVA (⁠$SV$ = 1.19, 1.01, and 1.15 μm²/μm³ for S. tropicalis, P. montanus, and N. beyeri respectively; p = .585) (Figure 2, Table 1). The ER had the second highest surface area density $(SV)$ among the four organelles in all three species, and ER $SV$ showed a statistically significant decrease as cell size increased based on ANOVA (⁠$SV$ = 0.71, 0.57, and 0.34 μm²/μm³ for S. tropicalis, P. montanus, and N. beyeri, respectively; p = .007) (Figure 2, Table 1; Tukey post hoc tests show a significant difference between S. tropicalis and N. beyeri, p = .005). The nucleus had the third highest surface area density $(SV)$ among the four organelles in all three species, and nuclear $SV$ showed a statistically significant decrease as cell size increased based on ANOVA (⁠$SV$ = 0.46, 0.25, and 0.1 μm²/μm³ for S. tropicalis, P. montanus, and N. beyeri respectively; p << .001) (Figure 2, Table 1; Tukey post hoc tests show significant differences for all pairwise comparisons, p < .015). The Golgi apparatus had the lowest surface area density $(SV)$ among the four organelles in all three species and showed no statistically significant difference across species based on ANOVA (⁠$SV$ = 0.07, 0.07, and 0.03 μm²/μm³ for S. tropicalis, P. montanus, and N. beyeri, respectively; p = .32) (Figure 2, Table 1). The surface area density of the entire cell (i.e., the cell SA:V ratio) showed a statistically significant decrease as cell size increased based on ANOVA (⁠$SV$ = 0.93, 0.61, and 0.27 μm²/μm³ for S. tropicalis, P. montanus, and N. beyeri respectively; p << .001) (Figure 2, Table 1; Tukey post hoc tests show a significant difference for all pairwise comparisons, p ≤ .002).

Surface area to volume ratio of the organelles

The mitochondria had the highest surface area to volume ratio (SA:V) among the organelles in all three species and showed no statistically significant difference across species based on ANOVA (SA:V = 12.31, 11.54, and 12.57 μm²/μm³ for S. tropicalis, P. montanus, and N. beyeri respectively; p = .899) (Figure 3, Table 1). The ER had the second highest SA:V among the organelles in all three species and showed no statistically significant difference across species based on ANOVA (SA:V = 10.05, 7.50, and 4.83 μm²/μm³ for S. tropicalis, P. montanus, and N. beyeri respectively; p = .568) (Figure 3, Table 1). The nucleus had the lowest SA:V among the organelles in all three species, and nuclear SA:V showed a statistically significant decrease as cell size increased based on ANOVA (SA:V = 1.52, 0.83, and 0.26 μm²/μm³ for S. tropicalis, P. montanus, and N. beyeri respectively; p << .001) (Figure 3, Table 1; Tukey post hoc tests show significant differences between S. tropicalis and P. montanus as well as between S. tropicalis and N. beyeri, p < .035). Because of its scarcity, the Golgi was absent from many grids, so per grid SA:V ratios were not calculated.

Estimates of absolute cell and organellar volume and surface area

All estimates of absolute volume and surface area for cells and organelles are listed in Table 2. Cell volume in N. beyeri is 46 times larger than S. tropicalis. The majority of cell volume estimates from diverse taxa in the literature are for erythrocytes; comparison of our enterocyte volumes with erythrocyte volumes for the same or congeneric species reveals that enterocytes are larger (S. tropicalis 228 μm³ vs. 122 μm³; P. montanus 3,191 μm³ vs. 1,476–3,204 μm³ for congeners P. cinereus (20 Gb) and P. dunni (47.5 Gb) [Gregory, 2023]), but that our estimates are in keeping with other empirical estimates. Our nuclear surface area and volume estimates suggest that the nucleus in each case is an ellipsoid rather than a perfect sphere. Assuming a sphere, calculating the radius of the nucleus for each species from the empirical measurements of both surface area (⁠$SA=4πr2$⁠) and volume (⁠$V=43πr3$⁠) yields larger radius estimates for each species based on surface area, suggesting that a larger perfect sphere would be required to yield the empirical estimates of surface area than volume.

Discussion

Scaling of the nucleus

Our results on scaling of the nucleus across this ~50-fold increase in cell size are consistent with earlier findings in two ways: (a) The nucleus shows the decreased SA:V ratio predicted by a sphere or ellipsoid increasing in size. (b) The volume fraction $(V^V)$ is the same across cell sizes, i.e., the nucleocytoplasmic ratio is maintained, consistent with its apparent functional significance (Balachandra et al., 2022). Mechanistically, the size of the nucleus is set by several interacting forces: DNA content, the state (i.e., compaction) of the chromatin, the total cellular amount of cytoplasmic factors involved in nuclear growth, and overall nucleocytoplasmic transport (Cantwell & Nurse, 2019b, 2019a; Chen et al., 2024; Hara, 2020; Heijo et al., 2022). Across species, the varying contributions of these forces as well as the differences in the relevant cytoplasmic factors can be understood in light of a general model for maintenance of the nucleocytoplasmic ratio in which the balance of mechanical forces generated by osmotic pressure within the nucleus and within the cytoplasm sets nuclear volume (Deviri & Safran, 2022).

Because of salamanders’ enormous cell volumes, the conserved nucleocytoplasmic ratio yields a disproportionately low nuclear surface area relative to both nuclear and cell volumes. In addition, nuclear pore complexes, which mediate the import and export of macromolecules into and out of the nucleus, are sparse across the salamander nuclear envelope relative to their density in other vertebrates (Maul & Deaven, 1977). Nonetheless, the conserved nucleocytoplasmic ratio in salamander cells—as well as overall cell functionality—indicates the existence of effective nucleocytoplasmic transport, demonstrating that movement through the nuclear pore complexes is not a rate-limiting factor for salamander cell physiology, despite proportionally low nuclear surface area and low nuclear pore complex density.

Although NC ratio is broadly conserved across the tree of life, larger cells do typically have proportionally smaller nuclei (Hara, 2020; Malerba & Marshall, 2021; Sessions, 2008). Our data do not follow this pattern. The most likely explanation for this mismatch is that we analyzed only three species and given the noise that exists in the cell volume:nucleocytoplasmic ratio relationship (Hara, 2020; Malerba & Marshall, 2021), we could have selected three species that do not fit the pattern. More generally, differences in nucleocytoplasmic ratio have been hypothesized to reflect, in part, differences in the stiffness of both chromatin and the nuclear lamina (Hara, 2020), which alter the balance between chromatin’s outward force and the incorporation of new lamins to drive nuclear growth (Chen et al., 2024). The effects of the state of chromatin—which behaves as a viscoelastic spring—on nuclear size are not completely straightforward; somewhat unintuitively, some degree of heterochromatin formation (i.e., chromatin condensation) is associated with increasing nuclear size, whereas more extreme condensation is associated with decreasing nuclear size (Chen et al., 2024). The relationship between chromatin state and nuclear size may be elucidated through studies of amphibians. Genome sizes across the clade vary over two orders of magnitude in large part because of differences in the amounts of transposable elements, parasitic and self-replicating sequences whose activity is silenced through various alterations to chromatin (e.g., CpG methylation of the DNA sequence itself as well as histone modifications) (Nowoshilow et al., 2018; Sotero-Caio et al., 2017; Sun & Mueller, 2014; Wang et al., 2023). Thus, amphibians likely vary in the proportion of condensed relative to open chromatin, yielding variable effects of chromatin state on nuclear size.

Scaling of the mitochondria

Across the tree of life, across cell sizes that span four orders of magnitude, mitochondrial volume increases proportionally with cell size, comprising roughly 10% of overall cell volume (Okie et al., 2016). Our results are largely consistent with these broad patterns; mitochondrial volume fractions in our representative amphibians range from 0.09 to 0.12 (Figure 1, Table 1). Mitochondrial surface area has also been shown to scale roughly proportionally with cell surface area across a dataset of eukaryotes that includes unicellular organisms, green algae and land plants, and mammals (Lynch & Marinov, 2017); our results show somewhat higher mitochondrial surface area per cell surface area than predicted from this broad relationship, but given the sparse sampling and noise inherent in this growing but limited empirical dataset, we do not attach any functional interpretation to this pattern.

In heterotrophic organisms, the increase in overall mitochondrial volume accompanying increased cell volume is primarily accomplished through an increase in mitochondrial number rather than size (Okie et al., 2016). In addition, experimental manipulations in yeast demonstrate that mitochondrial shape stays the same across increasing cell and overall mitochondrial volumes (Seel et al., 2022). Our results showing unchanged mitochondrial SA:V ratio across our model amphibians also suggest a mitochondrial network that is growing in total length, with no other changes to mitochondrial shape or size, consistent with these previous results.

Despite the presence of general scaling rules suggested by these broadscale phylogenetic patterns, mitochondrial number is also known to vary at a finer scale among cells, individuals, and species in association with metabolic demand (Schwerzmann et al., 1989; Scott et al., 2018). Salamanders have lower metabolic rates than frogs, and in fact show the lowest metabolic rates among tetrapod vertebrates (Chong & Mueller, 2013; Gatten et al., 1992; Pough, 1980). Accordingly, they have long been studied with the goal of connecting metabolic rate to cell physiology (Goniakowska, 1973; Szarski, 1970). The association between low metabolic rate and huge cells in salamanders prompted the “frugal metabolic strategy” hypothesis that natural selection favors large cell size because it relaxes energetic requirements (Szarski, 1983)—in contrast to selection at the other physiological extreme favoring small cell size to facilitate high metabolic output in bats, birds, and pterosaurs (Hughes & Hughes, 1995; Kapusta et al., 2017; Organ & Shedlock, 2009; Wright et al., 2014). Since then, mechanistic hypotheses connecting large cell size to low-metabolic rate have been proposed based on both energy supply (e.g., constraints on intracellular resource transport) and demand (e.g., decreased relative cost of Na⁺–K⁺ gradient maintenance across the plasma membrane) (Glazier, 2022).

Despite solid theoretical predictions for lowered metabolic rates in large cells (Kozłowski et al., 2003), empirical data have both supported and rejected a causal relationship between cell size and metabolic rate in amphibians. For example, studies of Xenopus frog embryos that differ in cell size and ploidy show lower mass-specific metabolic rates because of decreased relative Na⁺–K⁺ ATPase activity costs in larger cells (Cadart et al., 2023). In contrast, studies of salamanders fail to show a consistent relationship between mass-specific metabolic rate and cell size’s most commonly used proxy, genome size (Gardner et al., 2020; Gregory, 2003; Johnson et al., 2021; Licht & Lowcock, 1991; Uyeda et al., 2017). Our results reveal no effect of lower metabolic rate—whatever its underlying mechanistic cause—on mitochondrial volume fraction, surface area density, or shape (SA:V) in amphibians, suggesting that differences in metabolic demand across species do not drive cellular mitochondrial scaling in the clade. Our results are more consistent with mitochondrial volume increasing proportionally with cell volume through an increase in network length, independent of any specific features of cell physiology.

Mitochondrial abundance and functional output can be decoupled (Miettinen & Björklund, 2016), and salamander mitochondrial oxidative phosphorylation genes show evidence of relaxed purifying selection relative to frogs, consistent with lower functional demand on ATP synthesis (Chong & Mueller, 2013). Thus, uniform mitochondrial volume fraction, surface area density, and SA:V should not be interpreted as uniform capacity for ATP synthesis across amphibians. More generally, mitochondrial fission and fusion events maintain connectivity among the mitochondria to maximize overall function (Miettinen & Björklund, 2016), possibly through the transmission of the proton motive force itself from one region of the mitochondrial network to another faster than the corresponding metabolites and oxygen could reach distant mitochondria by diffusion (Glancy et al., 2015; Miettinen & Björklund, 2017). Our results are consistent with mitochondrial volume scaling proportionally with cell volume, which yields a mitochondrial network that comprises ~10% of the cell’s volume. This volume fraction, in turn, produces a spatial distribution of organelles that can undergo fission and fusion events to maintain an effective functional network throughout the cell—even at the extremes of animal cell size and metabolic rate.

Scaling of the endoplasmic reticulum and Golgi apparatus

The largest cell (N. beyeri) has significantly less ER surface area density than the smallest cell (S. tropicalis). The ER is a complex organelle consisting of different interconnected structures that perform different functions: the nuclear envelope, peripheral cisternae around the nucleus, and a tubular network that extends throughout the cytoplasm (English & Voeltz, 2013; Gubas & Dikic, 2022). Our analyses were not able to distinguish among these different components of the overall ER, limiting our ability to functionally interpret differences in ER surface area density. However, decreased ER surface area density in the largest cells suggests that ER functions that take place in the membrane (e.g., membrane lipid synthesis) (English & Voeltz, 2013) are operating at a lesser capacity per unit of cytoplasm in larger cells relative to smaller cells. This pattern is consistent with the existence of relatively less plasma and nuclear membrane in larger cells (Table 1). In contrast, ER volume scales proportionally with cell volume (i.e., there is no difference in the ER volume fraction $V^V$ across species), suggesting that ER functions that take place in the lumen (e.g., protein folding, processing, and assembly) are operating at the same capacity per unit of cytoplasm in larger and smaller cells. Taken together, these surface area and volume patterns suggest that the overall ER network shape is different in larger cells, trending toward less tubular morphology. Overall, our results are consistent with ER volume scaling proportionally with cell volume, which yields an ER network that comprises ~6% of the cell’s volume, albeit with decreased SA density in the largest cells. This volume fraction, in turn, produces a spatial distribution of tubules and cisternae that can maintain the physical connections between the ER and the mitochondria, plasma membrane, nucleus, and Golgi apparatus that are necessary for the functionality of all of these organelles (Heald & Cohen-Fix, 2014).

The Golgi apparatus receives processed proteins from the ER via membrane-bound vesicles and completes the final stages of the secretory pathway: additional protein processing, sorting, and export in vesicles to the proteins’ final destinations in the cell (e.g., the plasma membrane for secretion or the lysosomes for degradation) (Sengupta & Linstedt, 2011). Thus, Golgi size is proximately held at a dynamic equilibrium by the balanced influx and outflux of cargo in membrane-bound vesicles, and ER and Golgi size are functionally interconnected (Reber & Goehring, 2015; Sengupta & Linstedt, 2011). Ultimately, though, Golgi size does scale with cell volume during the growth phase of the mammalian cell cycle (Sin & Harrison, 2016), and we show here that Golgi volume also scales proportionally with cell volume across evolutionary increases in cell size, maintaining the same volume fraction $V^V$ despite vast increases in intracellular transport distance.

Mechanistic models as possible explanations for organellar scaling patterns with cell size

The simplest mechanistic model—the “limited precursor” model (Marshall, 2020)—provides an explanation for the patterns observed for the nucleus, mitochondria, and Golgi. In all three cases, the organelles scale proportionally with volume (i.e., maintain a constant volume fraction across species) and show no evidence of differences in shape (i.e., have an SA:V ratio that scales as predicted by an increase in radius for a spherical/ellipsoid nucleus and an increase in length for a tubular mitochondrial network). This pattern can be explained by the cytoplasm of larger cells containing more organellar building blocks, which are simply incorporated into larger organelles following the same rules across cell sizes. This uniformity exists despite differences in metabolic rate and transport distances that likely place different functional demands on the mitochondrial network and Golgi, respectively, rendering the “demand-driven” model a poor fit for these organelles. The uniformity in volume fraction also renders the “relative growth” model a poor fit, as it is not expected to yield uniform volume fractions across a large range of cell volumes. With our data, we are unable to eliminate the more complex “size measurement” or “programmed scaling” models; in the absence of additional data, however, we favor the simplest model as a working hypothesis (Marshall, 2020). For the ER, the pattern is somewhat consistent with both the “limited precursor” model (the volume fraction is the same across cell sizes) and the “demand-driven” model (the largest cells have decreased ER surface area density, consistent with lower functional demands for membrane lipid synthesis).

Organellar scaling mechanism yields insights into the biosynthetic capacity of gigantic animal cells

We note that, although the limited precursor model is introduced as the simplest of the mechanistic models by Marshall (2020), the model assumes that the limited precursors themselves (i.e., the building blocks of the organelles, such as lamin proteins for the nucleus; Chen et al., 2024) are present in the same concentrations in large cells as in small cells. This assumption is not trivial, and it is not consistent with experimentally manipulated increases in cell size, which lead to proteome dilution and perturbation so severe the cells appear senescent (Cheng et al., 2021; Lanz et al., 2022; Neurohr et al., 2019; Xie et al., 2022). Thus, even invoking the simplest limited precursor model implies that salamanders have evolved the increased biosynthetic capacity to maintain concentration homeostasis of organellar building blocks, despite a huge cell volume. Invoking the more complex models implies the same increases in biosynthetic capacity. Further work is required to reveal the dynamics of RNA and protein synthesis and stability that underlie this hypothesized alteration to biosynthetic output accompanying the evolution of gigantic cells.

Data availability

Data and code have been archived: https://doi.org/10.5061/dryad.cz8w9gj91.

Author contributions

A.N.A contributions: study conception, experimental design, data collection, data analysis, initial draft of manuscript, revision of manuscript. B.J.S. contributions: experimental design, data analysis, revision of manuscript. T.J.R. contributions: data collection, data analysis, revision of manuscript. R.L.M. contributions: study conception, experimental design, initial draft of manuscript, revision of manuscript, funding.

Funding

Funding for this project was provided by National Science Foundation grant 1911585 to R.L.M. and Colorado State University.

Conflict of interest: The authors declare no conflict of interest.

Acknowledgments

We thank members of A. Adams’ dissertation committee K. Hoke, D. Sloan, and J. Hansen for helpful discussion and feedback throughout the project. We thank D. Levy for S. tropicalis tissues. We thank M. Itgen for P. montanus tissues. Electron microscopy was done at the University of Colorado, Boulder EM Services Core Facility in the Molecular, Cellular and Developmental Department, with the technical assistance of facility staff.

References