X. laevis adults were purchased from a local breeder (Watanabe Zoshoku, Hyogo, Japan). Fertilized eggs were obtained by means of insemination and fed until they developed into larvae or young adults. X. tropicalis was provided by Amphibian Research Center (Hiroshima University) through the National Bio-Resource Project of MEXT, Japan. H. japonica, G. rugosa, and F. kawamurai adults were caught in a rice field near Yamaguchi University.

Blood samples were collected from the hearts of anesthetized frogs after injecting the animals with 0.2% MS222 (ethyl 3-aminobenzoate methanesulfonate; catalog no. E1052, Sigma-Aldrich) in 100% De Bore’s solution [110 mM NaCl, 1.3 mM KCl, 5.7 mM Tris, pH 7.5 (NaOH)] with 5 mg/ml ethylenediamine-N,N,N′,N′-tetraacetic acid (EDTA). In case of larva, the body near the heart was cut using tweezers and the blood samples were collected by means of pipetting with 5 μl of 0.8 × phosphate-buffered saline [1 × PBS; 8.04 mM Na₂HPO₄, 1.763 mM KH₂PO₄, 136.8 mM NaCl, 268 μM KCl, pH 7.4 (NaOH)]. The collected solution was transferred to 100% De Bore’s solution with 5 mg/ml EDTA. The solution was centrifuged at 300 × g, 4°C for 15 min. The pellet was resuspended in 0.8 × PBS and centrifuged at 1,000 × g, 4°C for 15 min. This washing step was repeated twice. The final pellet was resuspended in a designated volume (almost the same as the collected blood volume) of 0.8 × PBS.

For tissue samples, pieces of liver or kidney were dissected from anesthetized X. laevis adult frogs and homogenized in 0.8 × PBS using a micro-tube pestle. These suspensions were used for further experiments. For embryonic samples, fertilized eggs were obtained by in vitro insemination of freshly laid X. laevis eggs with crushed X. laevis testes. Thirty minutes after insemination, the embryos were dejellyed with 2% cysteine [pH 7.8 (KOH)] and washed with 0.2 × MMR [1 × MMR: 0.25 mM 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES)-KOH, 0.1 mM EDTA, 5 mM NaCl, 2 mM KCl, 1 mM MgCl₂, 2 mM CaCl₂, pH 7.8] 3–5 times. The embryos were incubated in 0.2 × MMR at 18°C for ∼8.5 days so that the developmental stages of the swimming tadpole embryos could be confirmed. Then, the embryos were homogenized in 0.8 × PBS using a micro-tube pestle. These suspensions were used for further experiments.

For observation of erythrocytes, the prepared erythrocyte solution was mixed with the same volume of a solution containing 4% formaldehyde and 15% glycerol with 5 μg/ml Hoechst 33342 (catalog no. H1399, Invitrogen), and 10 μg/ml 3,3′-dihexyloxacarbocyanine iodide [DiOC₆(3), catalog no. 318426, Sigma-Aldrich]. Approximately 6 μl of the mixture was placed on a glass slide and covered with a coverslip. The observation slide was subjected to fluorescence imaging using an Eclipse Ti-E (Nikon) wide-field microscope equipped with a ×40 objective lens and an sCMOS camera Zyla4.2 (Andor) motorized by Micro-Manager. To measure the three-dimensional volumes of erythrocytes and nuclei, the slides were subjected to z-sectional imaging using a FV1200 confocal microscope (Olympus) equipped with a ×60 objective lens. For immunostaining of erythrocytes and tissue nuclei, 50 μl of the prepared erythrocyte and tissue or embryo samples were gently mixed with 1 ml of 30% glycerol and 5% formaldehyde. After incubation for 15 min at 23°C, the fixed fractions were sedimented on poly L-lysine-coated coverslips by means of centrifugation at 300 × g and 16°C for 15 min. The coverslips were subsequently fixed in ice-cold methanol for at least 15 min, at −20°C. After washing with 1 × PBS, the coverslips were incubated in blocking solution (PBS containing 2% bovine serum albumin and 0.1% Triton™ X-100) for at least 1 h. The coverslips were subsequently treated with anti-nucleoporin antibodies (clone: mab414; 500 × dilution, BioLegend) or anti-pan lamin antibodies (catalog no. 65147; 50 × dilution, PROGEN Biotechnik) for 1 h at 23°C. Anti-pan lamin antibodies are multi-epitope cocktails that recognize multiple lamin isotypes (LA, LI, LII, LIII, B1, B2, and C) expressed in X. laevis. Antibodies for nucleoporins can recognize conserved domain FXFG repeats in nucleoporins. Because the NPCs consist of multiple nucleoporins, including FXFG repeats, the antibodies were expected to interact with all NPCs. After washing with PBS containing 0.1% Tween-20 three times, the coverslips were subsequently treated with Alexa Fluor™ 633 anti-mouse IgG antibody (catalog no. A11003; 1,000 × dilution; Thermo Fisher Scientific) for 1 h at 23°C. Finally, the coverslips were stained with 10 μg/ml Hoechst 33342 and 10 μg/ml DiOC₆(3) for 10 min at 23°C. The stained coverslips were subjected to z-sectional imaging using the above-mentioned wide-field microscope equipped with a ×40 objective lens, and deconvolution was performed using the Microvolution ImageJ plugin (Microvolution, CA, United States).

The cross-sectional area of the Hoechst 33342-stained nuclei was measured using elliptical ROI sections in the ImageJ software (National Institutes of Health). The elliptical ROI was set on the rim of the Hoechst 33342-positive chromatin region. To measure the erythrocytic areas, the cross-sectional area of erythrocytes in phase contrast images was measured using the elliptical sections tool of the ImageJ software. To measure the shape parameters in erythrocytes, the lengths along a long axis, defined as length, and along a short axis (perpendicular to the long axis), defined as width, were measured using a straight-line ROI in the ImageJ software. The aspect ratio and contour length of the erythrocytes were calculated using formulas as (length)/(width) and 2 × π × [(length/2) ² + (width/2) ² ] ^0.5 , respectively. The shape index was calculated using the formula as (contour length)/[π × (length/2) × (width/2)]. We measured at least 150 erythrocytes from individual frogs, and calculated the mean value for each frog. The thickness of the disc-shaped erythrocytes was measured directly, with the erythrocytes standing vertically inside the sample slide. Ordinary least-squares regression was performed to generate a regression line and calculate the correlation coefficient (r) and coefficient of determination (R ² ) using the regression analysis in the Excel software (Microsoft). To measure the volumes of erythrocytes and nuclei, the z-sectional images made using DiOC₆(3) and Hoechst 33342 were subjected to three-dimensional reconstruction, and their volumes and surface areas were determined using the IMARIS software (Oxford Instruments). For quantification of immunostaining signal intensities, the images were acquired under the same exposure conditions as the samples that they were compared with. ROIs were selected by tracing the nucleus rim using the segmented line tool of ImageJ. The mean intensity per pixel on the line ROIs was measured, followed by subtraction of the background signal from it. To normalize the signal intensity between erythrocyte and tissue or embryo samples, the calculated intensities in each erythrocyte nucleus and liver nucleus were divided by the mean intensity of erythrocyte nuclei from individual frogs. Significant differences in erythrocyte and nuclear size parameters, and signal intensities among samples were determined by non-parametric Wilcoxon tests using the R software.

For comparison of size scaling parameters among wider species classes, we collected the data of erythrocytic and nuclear sizes from the Cell Size Database (http://www.genomesize.com/cellsize/) and published papers. In cases where the lengths of the long (length) and short axes (width) were previously described in the literature, the cross-sectional area was calculated using the formula, π × (length/2) × (width/2). When only images were shown in the papers, we measured the nuclear and erythrocytic cross-sectional areas using the above-mentioned methods from the published images and then calibrated the absolute area to the scale bar. The genomic content was estimated based on the C-value in the Animal Genome Size Database (http://www.genomesize.com/). All values used in this study are listed in the Supplementary Tables.

We isolated erythrocytes from individual X. laevis adult frogs and measured the cross-sectional areas of the nuclei and whole erythrocytes (Figure 1A). Erythrocytic size is known to vary in different environments such as nutrient conditions (Maekawa et al., 2012; Janiga et al., 2017) and during metamorphosis and larval development (Hota et al., 2013). To collect samples of different sizes, we isolated erythrocytes from different developmental stages, larvae at metamorphosis stages, young adults after metamorphosis (less than 1-year of age), fully grown adults (more than 1-year of age). In the larval samples, the measured mean values and the coefficient of variation (CV) of erythrocytic areas [mean (±SD): 211 ± 20 μm²; CV (±SD): 14.5 ± 1.5%] were significantly higher than those in fully grown adults (179 ± 12 μm²; 9.8 ± 2.2%) and young adults (170 ± 13 μm²; 10.6 ± 3.1%) (Supplementary Table S1; Supplementary Figure S1A). Furthermore, when the lengths of erythrocytes along the long (length) and short axes (width) were measured, the aspect ratio [(width)/(length)] of the erythrocytes was not significantly different when these developmental groups were compared (Supplementary Table S1; Supplementary Figure S1B). Although variation in the aspect ratio was detected among individual frogs, there were fewer differences in erythrocytic shape during development. Considering that the conversion of larval erythrocytes to adult erythrocytes occurs after metamorphosis (Tamori and Wakahara, 2000), the observed changes in the mean erythrocytic area may be caused by the conversion of erythrocytes during development. Furthermore, when the volumes of erythrocytes isolated from a single adult frog were estimated using three-dimensional reconstruction, the measured cross-sectional areas were almost proportional to the estimated volumes (Supplementary Figure S1D). We also detected fewer differences in erythrocyte heights through direct measurement among samples from different adult frogs (Supplementary Figure S1E). Therefore, our measurements for two-dimensional sectional areas of the erythrocytes revealed almost similar results for the three-dimensional volume. It should be noted that even when processing the automatic reconstruction of erythrocyte volume, variations in the measured volumes were still observed (Supplementary Figure S1D). Therefore, the measured variation in erythrocyte size was mainly caused by the intrinsic variation of the erythrocytes rather than measurement noise. The mean value of the measured nuclear area was also significantly larger in erythrocytes from larvae [mean (±SD): 23.4 ± 2.5 μm²] than those from fully grown adults (22.1 ± 2.5 μm²) and young adults (21.4 ± 1.5 μm²) (Supplementary Figure S1C). In addition to the erythrocytic shape, the estimated three-dimensional volumes of the nucleus were plotted against the measured nuclear cross-sectional areas (Supplementary Figure S1F). Although a coefficient of determination R ² , which is the ratio of the explained variation from the regression line, is smaller in nuclear parameters than that in erythrocytic parameters, we detected a significant proportionality between nuclear area and volume. This result suggested that the nuclear cross-sectional areas revealed similar tendencies for the nuclear volume. When comparing the nuclear areas with the erythrocytic areas in the datasets from individual frogs were compared, very weak (correlation coefficient r < 0.4) but statistically significant (p < 0.05) correlations were observed in some samples of all three developmental stages [mean correlation coefficient r (±SD) = 0.22 ± 0.16 in fully grown adults; 0.14 ± 0.13 in young adults; 0.02 ± 0.10 in larvae; Figure 1A]. Although the significant correlations were detected in some samples, the slope values of the regression lines were very small in all samples [mean slope (±SD) = 0.20 ± 0.29 in fully grown adults; 0.03 ± 0.03 in young adults; −0.00 ± 0.01 in larvae]. Together with the very small coefficient of determination R ² in each frog (0.06 ± 0.05 in fully grown adults; 0.04 ± 0.02 in young adults; 0.02 ± 0.04 in larvae; Figure 1A), the measured nuclear areas did not change significantly when the erythrocytic areas increased. Furthermore, an analysis of the ratio of nuclear area-to-erythrocytic (cell) area (N/C area ratio) showed a negative correlation with the erythrocytic area in all three categories (r = −0.39 ± 0.18 in fully grown adults; −0.51 ± 0.07 in young adults; −0.76 ± 0.21 in larvae; Figure 1B). This negative correlation could be mathematically explained by the observed less correlation between the nuclear area and the erythrocytic area. Next, to evaluate the known effects of cell surface area and the ratio of surface area to cell volume to control the nuclear size (Brownlee and Heald, 2019), we plotted the measured nuclear areas with the calculated contour lengths of the erythrocyte sections or the shape index, which is a ratio of the contour length to the cell area (Supplementary Figure S2A). However, these two parameters showed a very weak correlation with nuclear area in most samples from each developmental stage (Supplementary Figure S2A). Furthermore, the nuclear area showed no significant correlation with the estimated volume and surface area of erythrocytes from three-dimensional reconstruction (Supplementary Figure S2B). Collectively, the observed less correlation of the nuclear area with any cell size parameter in X. laevis erythrocytes is distinct from the previously observed apparent correlations of nuclear size with cell size in specific cell types of eukaryotic species.

Next, to analyze the scaling of the nuclear size to the erythrocytic area within X. laevis species, we calculated the mean values in each frog and plotted the mean nuclear areas against mean erythrocytic areas in a log-log plot (Figure 2A). In this plot, the dataset showed a statistically significant (p < 0.001) correlation between the nuclear and erythrocytic areas. From a correlation analysis estimated as (nuclear area) = A × (erythrocytic area) ^B , we detected a “hypoallometric” relationship, where the scaling exponent B was less than 1 (B = 0.43, r = 0.47, p < 0.001, the regression line shown in black in Figure 2A), suggesting that the relative nuclear area against erythrocytic area decreases when erythrocytic area increases. In line with this, when we calculated mean N/C area ratios in each frog and plotted them against mean erythrocytic areas, the N/C area ratio correlated negatively with the erythrocytic area in the dataset including all developmental data (B = −0.57, r = −0.57, p < 0.001, the regression line shown in black in Figure 2B). It should be noted that a statistically significant correlation between nuclear area and erythrocytic area was also obtained when the dataset was plotted on a normal graph with linear regression (r = 0.48, p < 0.001; Supplementary Figure S2C). The analysis of the correlation between the nuclear areas and cell areas at each developmental stage showed that the scaling exponent and strength of the correlation obtained from the dataset of larvae during morphogenesis (B = 0.62, r = 0.59, p < 0.05) were higher than those in fully grown adults (B = 0.28, r = 0.17, p = 0.39) and young adults (B = 0.40, r = 0.43, p = 0.08). Since the measured range of erythrocytic areas in the larval samples was wider than that in adults, the correlation of the nuclear areas with the erythrocytic areas might be more apparent in the larval samples. Although the size scaling feature of the nucleus changes during development, the intra-species correlation of the nuclear size with the erythrocytic size could be detected within the dataset from a single X. laevis, including intrinsic size variations during development.

To examine the consistency of the detected nuclear size scaling features in other frog species, we measured erythrocytes of other clawed frog species, X. tropicalis, and regional frog species H. japonica, G. rugosa, and F. kawamurai. The measured erythrocytic areas in G. rugosa were substantially larger than those in X. laevis, whereas those in X. tropicalis and F. kawamurai were smaller (Figure 2C). In each dataset obtained from the four frogs, the measured nuclear areas correlated very weakly (r < 0.4) with the erythrocytic areas in some samples of four frogs [X. tropicalis: r (±SD) = 0.23 ± 0.10; H. japonica: 0.16 ± 0.18; G. rugosa: 0.14 ± 0.10; F. kawamurai: 0.01 ± 0.10], similar to that in X. laevis erythrocytes (Figure 2C; Supplementary Table S1). Although the significant correlations (p < 0.05) were detected in some samples, the calculated slope values of the regression lines were very small in all frog samples [mean slope (±SD) = 0.05 ± 0.02 in X. tropicalis; 0.09 ± 0.19 in H. japonica; −0.01 ± 0.29 in F. kawamurai; 0.02 ± 0.02 in G. rugosa]. This result suggested that the measured nuclear area did not change significantly when the erythrocytic area increased in all frog species. In line with this, the calculated N/C area ratios showed a negative correlation with the erythrocytic area in every species, excluding samples of X. tropicalis, which showed a narrower erythrocytic size variation (Figure 2D; Supplementary Table S1). Furthermore, when we compared the measured nuclear areas with the calculated contour length or shape index, the nuclear areas still showed very weak correlations with the parameters for each erythrocytic shape (Supplementary Figure S3A). When we analyzed the intra-species correlation in a log-log plot using the mean values obtained from individual frogs, we observed a correlation between the nuclear areas and erythrocytic area in G. rugosa (r = 0.78, p < 0.005; Figure 2E). However, no statistically significant correlation was obtained in other species (X. tropicalis: r = 0.01, p = 0.99; F. kawamurai: r = 0.27, p = 0.31; H. japonica: r = 0.24, p = 0.51; Figure 2E). The calculated scaling exponent B of the nuclear areas to the erythrocytic areas revealed hypoallometry, with B < 1 in every species, although the values were distinct in each species. Furthermore, the N/C area ratio showed a negative correlation with the erythrocytic area in most species (Figure 2F), supporting the observed hypoallometric relationship between the nuclear area and the erythrocytic area. The dataset for G. rugosa showed no correlation between N/C area ratio and erythrocyte area (r = -0.08, p = 0.82), which is thought to be caused by the apparent correlation between the nuclear area and erythrocyte area. Overall, the data suggest that although the quantitative features of nuclear size scaling are different in each species, the lower correlation of nuclear size with erythrocytic size within individuals and the hypoallometric scaling within species might be conserved across frog species.

To evaluate the inter-species size scaling relationship, we calculated the mean values for each species and plotted the mean nuclear areas against the mean erythrocytic areas in a log-log plot (Figure 3A). In contrast to the low scaling exponents in the intra-species datasets, the inter-species scaling exponent B was more than one among the five different species (B = 1.49, r = 0.98, p < 0.005), indicating a hyperallometric relationship in which the relative nuclear area versus the erythrocytic area increases when the erythrocytic area increases. In line with this, the calculated N/C area ratio revealed a positive correlation with erythrocytic area (Figure 3B; B = 0.49, r = 0.87, p = 0.06). Next, to consider the effects of genomic content, which is intrinsically different in each species, we compared the mean nuclear areas of each species with the genomic content within the nucleus (Figure 3C). In the plots, the nuclear area correlated well with the genomic content (r = 0.98, p < 0.005). In accordance with the well-known relationship between erythrocytic size and genomic content, our measured erythrocytic area also correlated with the genomic content (Figure 3D; r = 0.94, p < 0.05). Although both the nuclear and erythrocytic areas qualitatively correlated with the genomic content, the calculated scaling exponent of the erythrocytic area to the genomic content (B = 0.43) was substantially smaller than that of the nuclear area to the genomic content (B = 0.68). This difference in the scaling exponents indicates that the relative changes in the nuclear and erythrocytic areas are distinct when genomic content increases. This scaling feature was also supported by the positive correlation of the calculated N/C area ratio with the genomic content among the five frog species (Figure 3E; B = 0.25, r = 0.97, p < 0.05). It should be noted that statistically significant correlations between the nuclear area and erythrocytic area were also seen when the dataset was plotted on a normal graph using a linear regression (Supplementary Figure S3B). Overall, based on our inter-species comparisons with the genomic content, we found differences in the nuclear size scaling between intra- and inter-species and in the scaling exponents of the nuclear and erythrocytic sizes with the genomic content among the five limited frog species.

To extensively understand whether our observed size scaling features are conserved among other amphibians and species in the other classes, we revisited the prominent open databases of nuclear size, erythrocytic size, and genomic size collected by Ryan T. Gregory across a wide range of species (Animal Genome Size Database, Cell Size Database). Furthermore, we combined our original measurement dataset of five frog species and our collected data from published papers into Gregory’s datasets. To analyze the effects of either genomic content or cell size on the determination of the nuclear size separately, we excluded the data in which one parameter was predicted from the other, unlike previous comparisons (Malerba and Marshall, 2021; Gillooly et al., 2015). Using our compiled dataset, we analyzed relationships among the three scaling parameters in each species class: Anura and Caudata amphibians, aves, fish, and reptiles (Figure 4, Supplementary Table S2). Although our compiled dataset should have variations in the size parameters, due to potential differences in data sampling, the qualitative size relationship between two parameters appeared to be conserved across the species class. In the amphibian dataset from Caudatans or Anurans, the scaling exponents of the nuclear areas to the erythrocytic areas revealed a hypoallometry with B < 1 (B = 0.71, r = 0.82, p < 0.001 in Caudatans; B = 0.83, r = 0.60, p < 0.001 in Anurans; Figure 4A). When calculating B values from the dataset of other species classes, the hypoallometric scaling of the nuclear areas to the erythrocytic areas was conserved in the fish, reptiles, or aves class, with B values maintained within the range of 0.75–0.87 (Figure 4A). Although the calculated B values revealed variations among species classes, the hypoallometric scaling of the nuclear size to the erythrocytic size was well conserved among all species classes (B = 0.91, r = 0.94, p < 0.001; black line in Figure 4A). This hypoallometry in erythrocytes is consistent with the previous calculations in inter-species datasets, including non-biased cell types (Hara, 2020; Malerba and Marshall, 2021).

Next, to analyze the contribution of the genomic content among species, the nuclear areas were plotted against the genomic contents in each species class (Figure 4B). In this plot, the datasets of fish or amphibians revealed a positive correlation of the nuclear areas to the genomic content (r = 0.50–0.83, p < 0.001), which is consistent with our original measurement from five frogs. On the contrary, in the dataset of reptiles or aves, the nuclear areas revealed a very weak correlation with the genomic content (r < 0.40, p < 0.001; Figure 4B). Furthermore, the ratios of the explained variation from the regression line is larger in fish or amphibians (R ² = 0.25–0.68) than those in reptiles or aves (R ² = 0.00–0.15). In addition to the differences in the strength of the scaling, the calculated scaling exponent in Anurans (B = 0.44) was lower than that in fish or Caudatans (B = 0.64, 0.59; Figure 4B). These calculated values of r (0.54) and B in Anurans varies with those observed in our original measurement of five frog species (r = 0.98, B = 0.68; Figure 3C), suggesting that the scaling features and underlying mechanisms might vary even within each species class and might change independently during evolution from fish into each class. We further plotted the erythrocytic areas against the genomic contents in the datasets of classes with nucleated and enucleated erythrocytes in mammalian species (Figure 4C). Positive correlations with a hypoallometric scaling relationship were detected in the dataset of nucleated erythrocytes over species classes (B = 0.70, r = 0.94, p < 0.001) and most datasets of individual classes (Figure 4C). Interestingly, the strength of the correlation and the values of the scaling exponent were different for each class. For instance, the fish dataset revealed a clear correlation of the erythrocytic area with the genomic content (r = 0.86, p < 0.001), while the correlations in other classes, including mammalians with enucleated erythrocytes, were not apparent (r = 0.25–0.47, p < 0.001). It should be noted that when the datasets were plotted on a normal graph using a linear regression, the correlations among the three parameters were conserved (Supplementary Figure S4). However, the correlation coefficients (r) and coefficient of determination (R ² ) in some samples were reduced when analyzing the normal plot that used linear regression (e.g., r = 0.54, R ² = 0.29 in scaling analysis of nuclear area with genomic content among Anura amphibians; r = 0.36, R ² = 0.13 in the normal plot). Therefore, we used scaling analyses to better understand these scaling features.

We hypothesize that such species- and/or class-specific scaling features are derived from the differences in the contribution of genomic content and erythrocytic size in the determination of the nuclear size. To distinguish the two contributions to nuclear size determination, the datasets were fitted to an equation combining two power-law parameters, [nuclear area (NA)] = A’ × [erythrocytic area (CA)] ^B’ × [genomic content (GC)] ^B’’ . Although the mutual relationship among NA, CA, and GC might be complex, in the present study, we simply assumed the equation to be a product of powered GC and powered CA. Based on the CA, GC, and NA values, the constant A’ and scaling exponents B’ and B’’ were estimated using the least-squares method in each class (Figures 4D,E). Because the estimated NA values for each group were used, the fitness of the estimated NA to the real collected NA was better than when the estimated A and B values from the equations with single-powered parameters, such as (NA) = A × (CA) ^B or (NA) = A × (GC) ^B were used (Supplementary Figure S5). The estimated scaling exponents B’ for CA in the equation with multiplied GC and CA (B’ = 0.58–0.80, Figures 4D,E) decreased from the scaling exponents B estimated using the single powered CA in each class (B = 0.71–0.87, Figure 4A). Furthermore, the estimated scaling exponents B’’ for GC in the equation (B’’ = 0–0.28, Figures 4D,E) were far smaller than B estimated using the single powered GC (B = 0.07–0.68, Figure 4B). In extreme cases in reptiles, even if both GC and CA are considered for the parameter estimation, the fitness is no longer improved (Figures 4D,E), suggesting that the contribution of the genomic contents in determination of the nuclear size might be very small in reptile erythrocytes. To clarify the difference between the two scaling exponents for CA and GC, we calculated the ratio of the two values of the scaling exponents B’’ to B’ (Figure 4E). As a result, the ratios were substantially higher in amphibians than those in aves and reptiles, suggesting that the contribution of the genomic content in determination of the erythrocyte nuclear size might have been switched during evolution from aquatic organisms to terrestrial organisms. Collectively, erythrocytes have a unique inter-species size scaling feature because of the low correlation of the genomic content, which reveals species-specificity in the degree of the contribution of the genomic content in the determination of nuclear size.

The observed variations in the contributions of the genomic content and erythrocytic size in the determination of nuclear size are expected to change the relationship between N/C area ratio with GC in each species. When the calculated N/C area ratio versus the GC was plotted, the N/C area ratio correlated very weakly with the GC in each species and for the entire dataset across species classes (r < 0.4; Figure 5A). This might result from our estimated relatively low contribution of the GC in determining the NA, as compared to CA. To ascertain whether this scaling feature of the N/C area ratio is specific for erythrocytes, we calculated the scaling exponents of the N/C area ratio to the GC using datasets including variation of the genomic contents and nuclear size from published studies with data of various cell types among species or single cell types with different ploidy levels from a single organism (Figure 5B). As a result, the most frequently calculated values of B were approximately 0, with no significant correlation (p > 0.05) in the datasets of erythrocytes and some other cell types, as observed in our erythrocyte datasets (Figure 5B). The B values in most human cell types were substantially smaller than 0 and had significant correlations (Figure 5B), indicating that the relative nuclear size versus cell size decreases when the genomic size increases. In general, in eukaryotic cells other than erythrocytes, an increase in genomic content might contribute more to the weakening increase in nuclear size. Therefore, lower correlations of the genomic content to the nuclear size and N/C area ratio could be specific features that could help characterize erythrocyte-specific size scaling and functions.

To explore which nuclear features contribute to the size scaling specificity in erythrocytes, we analyzed the nuclear structure in X. laevis erythrocytes by visualizing the nuclear membrane constituents. Immunostaining using anti-nucleoporin antibodies that recognize multiple NPC components and anti-pan lamin antibodies that recognize almost all lamin isoforms revealed nuclear pores and lamins in the peripheral region of the DNA (Figure 6A). The distribution of the nuclear pores and lamins took the form of a dot-like pattern on the rim of the erythrocyte nucleus. In contrast, these nuclear membrane constituents were located continuously on the nuclear rim without large gaps in the nuclei isolated from the liver and kidney tissues as well as tadpole embryos (Figure 6A). Furthermore, the mean intensities of both nucleoporin- and lamin-positive signals were significantly lower in erythrocytes than those in other tissues and embryos, although the intensities differed among nuclei from tissues and embryos (Figure 6B). The detected reduction in the mean signal intensity was caused by both a reduction in the quantity of nuclear pores and lamins and changes in the distribution on the nuclear rim. Overall, the observed reduction in the nuclear membrane constituents might be specific features of erythrocytes and involved in the control of erythrocyte-specific nuclear size scaling with genomic content.