Natural size variation among embryos leads to the corresponding scaling in gene expression

Abstract

Xenopus laevis frogs from laboratory stocks normally lay eggs with extensive size variability. We find that these initial size differences subsequently affect the size of the embryos prior to the onset of growth, and the size of tadpoles during the growth period. Even though these tadpoles differ in size, their tissues, organs, and structures always seem to be properly proportioned, i.e. they display static allometry. Initial axial patterning events in Xenopus occur in a spherical embryo, allowing easy documentation of their size-dependent features. We examined the size distribution of early Xenopus laevis embryos and measured diameters that differed by about 38% with a median of about 1.43 mm. This range of embryo sizes corresponds to about a 1.9-fold difference in surface area and a 2.6-fold difference in volume. We examined the relationship between embryo size and gene expression and observed a significant correlation between diameter and RNA content during gastrula stages. In addition, we investigated the expression levels of genes that pattern the mesoderm, induce the nervous system and mediate the progression of ectodermal cells to neural precursors in large and small embryos. We found that most of these factors were expressed at levels that scaled with the different embryo sizes and total embryo RNA content. In agreement with the changes in transcript levels, the expression domains in larger embryos increased proportionally with the increase in surface area, maintaining their relative expression domain size in relation to the total size of the embryo. Thus, our study identified a mechanism for adapting gene expression domains to embryo size by adjusting the transcript levels of the genes regulating mesoderm induction and patterning. In the neural plate, besides the scaling of the expression domains, we observed similar cell sizes and cell densities in small and large embryos suggesting that additional cell divisions took place in large embryos to compensate for the increased size. Our results show in detail the size variability among Xenopus laevis embryos and the transcriptional adaptation to scale gene expression with size. The observations further support the involvement of BMP/ADMP signaling in the scaling process.

1. Introduction

An essential feature of all animals is that their component parts, i.e. organs or tissues, have a characteristic size relationship relative to each other at any given stage, known as static allometry (Huxley, 1932, Snell, 1892, Thompson, 1992). These characteristic size relationships arise in part because structures grow at different rates from one another, and from the animal as a whole (ontogenic allometry). To produce their final relative sizes, structures may grow at the same rate (isometry), or grow slower or faster (hypo-, or hyper-allometry, respectively). Alternatively, during early embryo genesis structures may be specified at their correct relative size initially. Accordingly, mechanisms that specify the initial size and relative growth rate of embryonic regions and structures play an essential role in ensuring the normal development and ultimate viability of the individual organism. During embryogenesis, individuals of the same species can exhibit size differences that require the adaptation of the onto genic allometry to size, i.e. scaling.

Xenopus laevis embryos develop from spherical eggs without growth and limited shape change until the neural plate forms in the surface ectoderm towards the end of gastrulation. Thus, the early patterning events that regulate organogenesis occur in an embryo in which the volume and surface area can be readily measured and monitored. Xenopus laevis oocytes and early embryos are commonly described as having an average diameter of 1.2 mm, with a range of 1–1.3 mm (Mitchison et al., 2015, Tassan et al., 2017; Xenbase.org), although, to the best of our knowledge, no extensive analysis has been reported documenting the size of these embryos either in nature or in laboratory specimens. In any case, this range of diameters corresponds to a dramatic, about twofold difference in volume. However, despite these large size variations, individual organs are scaled and properly proportioned.

Early during embryogenesis, regions and structures that regulate subsequent developmental events and establish morphogenetic gradients, must have the correct size relationship when they are specified. The vertebrate organizer (Spemann's organizer in amphibians) is a prime example of such an embryonic structure that would need to be initially scaled based on embryo size as it regulates multiple scalable processes including the establishment and patterning of the dorsal-ventral and anterior-posterior axes as well as mesendoderm patterning and neural induction (Harland and Gerhart, 1997, Leibovich et al., 2018, Niehrs, 2010, Tam and Loebel, 2007). It has been shown that very early experimental bisection of frog embryos into left and right halves results in two normally proportioned embryos, albeit smaller in size (De Robertis, 2006, McClendon, 1910, Spemann, 1938). Also, following dorsal-ventral bisection of Xenopus embryos, the dorsal, organizer-containing half gives rise to a proportioned but smaller embryo, whereas the ventral half develops aberrantly (Inomata et al., 2013, Reversade and De Robertis, 2005). These results show that early vertebrate embryos can scale their developmental processes and that this scaling requires the organizer (De Robertis, 2006). A number of studies have addressed the scaling of the organizer and its downstream effects with the aid of mathematical and theoretical models supported by experimental manipulations (Ben-Zvi et al., 2008, Inomata et al., 2013, Leibovich et al., 2018, Meinhardt, 2015, Zinski et al., 2017). In recent years multiple models have been proposed to explain the possible mechanisms that determine the size of embryonic tissues and scale the nervous system during its induction (Almuedo-Castillo et al., 2018, Ben-Zvi et al., 2008, Inomata et al., 2013, Ishimatsu et al., 2018, Leibovich et al., 2018). Many of these models focus on different components of the BMP signaling pathway as important regulators of scaling.

In Xenopus embryos, soon after fertilization cytoplasmic movements, i.e. cortical rotation, deliver components of the Wnt/β-catenin pathway to the dorsal midline inducing it to develop as an embryonic organizing center, Spemann's organizer (De Robertis, 2006, Fainsod and Levy, 2004, Harland and Gerhart, 1997, Niehrs, 2004). In parallel, diffusible BMPs in ventral and lateral regions increases to restrict the dorsal organizer and prevent the formation of ectopic ones (Fainsod et al., 1994, Marom et al., 2005). Importantly, the dorsal organizing center produces diffusible anti-BMP factors, including chordin, noggin, follistatin, and cerberus, that antagonize the anti-organizer activity of the BMPs that reach the dorsal side of the embryo and contribute to the establishment of the dorsal-ventral BMP signaling gradient (Bier and De Robertis, 2015, De Robertis et al., 2017, Plouhinec et al., 2013). Thus, these opposing diffusible factors seem to play a role in the feedback regulation of the relative size of the embryonic dorsal and ventral regions.

These same anti-BMP factors play a role in the induction of the nervous system as the descendants of the organizer migrate beneath the overlying dorsal ectoderm (Andoniadou and Martinez-Barbera, 2013, Pera et al., 2014). In parallel, the dorsal ectoderm expresses a cascade of genes that are associated with its transition to become the precursor of the neural plate. Some of these genes are activated during neural induction, whereas others seem to precede those inductive events (Lee et al., 2014, Moody et al., 2013). In particular, Foxd4l1.1 activates genes that maintain the ectoderm in a proliferative state (gmnn and zic2) and represses genes that promote differentiation (zic1 and zic3)(Sherman et al., 2017). This cascade of ectodermal genes may be involved in regulating the relative size of the neural plate.

Here, we examined the scaling of morphogens in normal Xenopus laevis embryos by extensive analysis across large numbers of embryos that naturally exhibited size variation. Our results reveal a wide variability in size among Xenopus laevis embryos, larger than was previously assumed, and this variability is independent of maternal size. To address the question of adaptation to size changes, we identified a correlation between total RNA content and size, and specific mesodermal and neuroectodermal markers showed scaled changes in transcript abundance. These observations suggest a regulation of the transcriptional output with size to compensate for volume changes, and the necessary adaptation of signaling intensity and morphogen gradients to function over the longer distances of larger embryos. The embryos ultimately establish size-adapted morphogen gradients and give rise to normally proportioned organs. Finally, we found that the neural plates of larger embryos are composed of cells of the same size and density as those of smaller embryos, indicating that in larger neural plates some compensatory cell divisions took place to account for the increased size. Collectively, these findings illustrate how embryonic regions and structures scale to embryo size during early patterning events.

2. Materials and methods

2.1. Embryo culture

Xenopus laevis frogs were purchased from Xenopus I or NASCO (Dexter, MI or Fort Atkinson, WI). Experiments were performed after approval and under the supervision of the Institutional Animal Care and Use Committee (IACUC) of the Hebrew University (Ethics approval no. MD-17-15282-3), or the IACUC of GWU (approval no. A233). Embryos were obtained by in vitro fertilization, incubated in 0.1% MBSH and staged according to Nieuwkoop and Faber (1967).

2.2. Total RNA purification from embryos and cDNA preparation

For each sample, 5–10 staged embryos were collected and stored at −80 °C. RNA purification was performed using the Bio-Rad Aurum Total RNA Mini Kit (according to the manufacturer’s instructions). RNA samples were used for cDNA synthesis using the Bio-Rad iScriptTM Reverse Transcription Supermix for RT-qPCR kit (according to the manufacturer’s instructions).

2.3. Expression analysis

Whole-mount in situ hybridization (WISH) analysis of gene expression was performed as described previously (Epstein et al., 1997). Probes for in situ hybridization were prepared from the H7 clone for gsc (Cho et al., 1991), Δ59 clone for chrd.1 (Sasai et al., 1994), pCR-Script-ADMP clone for admp (Moos et al., 1995), XP5 (pGEM-Xwnt8) for wnt8a (Christian and Moon, 1993), pCS2-ZICR1 for zic1 (Mizuseki et al., 1998), pCS2-foxD5aORF for foxd4l1.1 (Sullivan et al., 2001), and pCS2-SOX2 for sox2 (Mizuseki et al., 1998).

Quantitative real-time RT-PCR (qPCR) was performed using the Bio-Rad CFX384 thermal cycler and the iTaq Universal SYBR Green Supermix (Bio-Rad). All samples were processed in triplicate and analyzed as described previously (Livak and Schmittgen, 2001). All experiments were repeated with at least three different embryo batches. qPCR primers used are listed in (Table 1).

Table 1.

qPCR Primers pairs.

Genea	Forward primer	Reverse primer
admp	GCCTTCCGAGCAAGCTTACTT	CCTTGTGGCAACTGTATCTTATTTTTA
acvrl1	GCTCTGGGGAAACTTGTGTT	CAACGCTCCTTTATGCTGTT
acvr1	TGTTATGGGCAGCAGTGTT	GATGTTCAAGTTACAGAGGTCACT
bmpr1a	TGGCTCAGGGCTACCATTATT	CACCTTCTCTCCTCTCCATTTTC
bmpr1b	ACAGCAGGAAGGAAGACACA	ACAGTGGTGGTGGCAGTAAC
bmp2	ACACGGACAGCAGAAAACCA	AACAGCAGCAGGAGCAGAGA
bmp4	GCAGCCCAGTAAGGATGT	CTTCTGTGCCTGGTAGATTC
bmp7	TCTCCTTTGGACATACTTCTTGTG	CGCAACCTCCTCTGGATAAA
cer1	CTGGTGCCAAGATGTTCTGGAA	CGGCAAGCAATGGGAACAAGTA
chrd.1	ACTGCCAGGACTGGATGGT	GGCAGGATTTAGAGTTGCTTC
fst	CAGCGACAACACGACTTACC	TCCTCCTCTGTATCTTCAACAATG
gapdh	GCTCCTCTCGCAAAGGTCAT	GGGCCATCCACTGTCTTCTG
gsc	TTCACCGATGAACAACTGGA	TTCCACTTTTGGGCATTTTC
myod1	CCCTGTTTCAATACCTCAGACAT	CGTGCTCATCCTCGTTATGG
nog	CAATGCCAGCGGAAATCA	ATGTGTAAAGGACAGGACAGAAGGT
sia1	CTGTCCTACAAGAGACTCTG	TGTTGACTGCAGACTGTTGA
szl	AACAAGGTCTGCTCCTTCCA	CTGTGGGTCTGGTCCGTATC
ventx1.2	AGGCAGGAGTTCACAGGAAA	AATGCCTGTTCCAGTTTGCTT
ventx2.2	AGAGAGCAGCCAAGCAAAGT	GAAGAAGGGGACACATCACTGT
ventx3.2	CCCAGCCAGCACCACAA	AGCATCTTCATCACACACAGGTTT
wnt8a	CTACACCGCAGAGTATTCCA	ATCTCAGGACAGACCAATCG
lhx1	CCCTGGCAGCAACTATGACT	GGGCACAGAGGAAGGTACAA
zic1	CGCCCAGCACAGTCTATT	TGTCCGTTCACCACATTG
zic2	GTCGCACGGACATGCTACTA	GTGGTGATGTCGGGCAAATG
sox2	GCATGTCCTACTCCCAACAAG	GGGAAGAAGAGGTGACTACAGG
sox3	TTGGAATCTGTGTGGCTGTT	GGCTCTTGATGTCGGTGTC
sox11	CGACCCCAACTCAGGAAAC	TAATCGGGGAAGTCGAAGTG
gmnn	TTGAAAGGCTCACTGGAAATG	CTTCTGCCATGTCTGCTTCA
foxd4l1.1	TCAGCAGCAAGTTCCCTTAC	CCTGGTTCCCGTGGTATT
Meis3	GTGGCTCTTTCAGCACCTCT	CTGGACAATGCGTCTTCTTG
fst	CAGCGACAACACGACTTA	TCCTCCTCTGTATCTTCAACAATG
nodal3.1	GAGCACCGTTCCACCTAA	CCCATCCGATCTTCTGAA
not	CCGTGTGGCTCATCAAAC	GAATGGCAATACTCTCTCTTTACAG
slc35b1	CGCAATTTCCAAACAGGCTCC	CAAGAAGTCCCAGAGCTCGC

^aGene names according to Xenbase.org.

2.4. Embryo measurements

Developing embryos were measured from the animal side after dejellying and ensuring a firm consistency. Otherwise, fixed embryos were measured mainly from the vegetal side following whole-mount in situ hybridization. Diameter length average for each embryo was determined after two perpendicular measurements to ensure crossing the embryo center. No significant embryo size differences were observed between the clutches measured in the two laboratories.

Mesodermal genes

Embryos to be measured were photographed during their incubation to the desired stage or after WISH by placing them in either MBSH or PBS solution. Each embryo in its vitelline membrane was photographed individually using a Nikon Digital Sight DS-Fi2 camera attached to a Nikon SMZ800 binocular microscope. A millimeter scale was placed under the Petri dish with the embryos during this process. The size of the embryo (diameter) and expression domain size (degrees) were determined using the ImageJ software (NIH) after setting the correct scale (pixel to mm) using the mm ruler in the picture. This setup allowed the measurement of embryo diameters and marginal zone expression domain angles.

Neural Genes

During cleavage stages, embryos in their vitelline membranes were screened for size differences by quickly measuring them with an eye-piece micrometer on a dissecting microscope. Embryos that measured 1.4 mm or less were classified as small, and those 1.6 mm or more were classified as large. These separate groups were fixed and processed for in situ hybridization as described above. Following in situ hybridization, computer-assisted morphometry was performed on each embryo using either an Olympus SZH12 stereomicroscope with CellSense software or a Nikon SMZ800 with NIS Elements software. After diameter measurement, the microscope was then re-focused to the edge of the expression domain. The border of the labeled probe was outlined to determine the area of the expression domain. Each embryo was photographed along with its measurements.

2.5. Neural plate cell size and density

During cleavage stages, embryos were measured and separated into large and small groups as above. At early neurula stages (stage 14), embryos were fixed in 4% formaldehyde in PBS and incubated overnight at 4 °C in 0.4% phalloidin/AlexaFluor 488 plus 0.05% DAPI and 0.1% Triton-X100 in PBS. For cell density estimations, phalloidin-stained and DAPI labeled cells were counted in an area of 5X10⁴-10⁵pixels in the anterior region of the neural plate. For cell size measurements, the long axis of about 50 cells per neural plate was measured from 11 large and 11 small embryos.

2.6. Total RNA per embryo and “embryo equivalent RNA value”

To calculate the RNA content in small and big embryos we ensured that each sample collected originated from 5 embryos, and the clutches were generated from parallel fertilizations performed on the same day, such that the sperm utilized was from the same male but the big and small embryos originated from different females. Care was taken to compare embryos at the same developmental stage. The RNA concentration in each sample was determined using a Tecan Infinite F200 Pro plate reader/nanodrop. This measurement allowed us to calculate the RNA content/embryo. Relative expression levels were corrected for the amount of RNA per embryo, “embryo equivalents”, to account for the differences in RNA content between small and large embryos and the technical limitation of the qPCR procedure in which the initial 1 μg of RNA represents a different proportion of the small and large embryos.

2.7. Statistical analysis

All statistical analysis was performed using the Prism (GraphPad) software package. Results are displayed as mean ± standard error of the mean (SEM) or in the case of violin plots and boxplots as median with the interquartile range and where applicable, whiskers mark the 10th to the 90th percentile range. Data were analyzed using Student's two-sample t-test and one-way ANOVA (Bonferroni's multiple comparisons). Results were considered statistically significant when p < 0.05.

3. Results

3.1. Embryo size distribution in Xenopus laevis laboratory populations

To study and characterize the natural size variation between Xenopus laevis embryos from laboratory-bred individuals, we collected embryo batches (clutches) from numerous fertilizations (N = 33). Each clutch was comprised of eggs laid by a single female. Embryos from each clutch were photographed and measured before mid-gastrula stages, and the diameters of the embryos were determined. These measurements (n = 2239 embryos) illustrated an extensive size variation between embryos obtained from randomly chosen Xenopus laevis females (Fig. 1A). The diameters ranged from 1.23 mm to 1.70 mm with a median of 1.44 mm and exhibit a normal distribution (Fig. 1A, inset). This range represents a 38.4% difference in diameter that results in an almost 2-fold difference in surface area and a 2.6-fold difference in volume between extremes. To determine whether embryo size depends on the mother, we analyzed the size distribution within a clutch of eggs from a single female (Fig. 1B). The results showed that each female usually laid eggs within a relatively small range of sizes (SD = 0.03 mm–0.065 mm). The size distribution defined a median for the clutch which was dependent on the female. This dependence may reflect genetic variability, maternal size, nutrition, age or other female-specific properties. For 29 independent clutches, we measured the snout-vent length of the female and the diameter of her embryos which allowed us to study the possible correlation between maternal size and embryo size (Fig. 1C); our results show no correlation. In a few instances, the large and small embryos were monitored and photographed from fertilization to initial tadpole stages (st. 41). Comparative analysis revealed that the initial size of the embryo influenced the size of the larva; small eggs gave rise to small embryos and large eggs gave rise to large embryos (Fig. 1D).

Fig. 1. Natural size variability in Xenopus laevis.

(A–C) Embryos obtained from multiple clutches (fertilizations) were photographed and measured before mid gastrula stages. (A) The distribution in diameters of 2239 embryos (n) from 33 fertilizations (N) is shown in a violin plot. The median value is shown as a dashed line and the interquartile range is demarcated by the dotted lines. The inset shows the normal distribution of the embryo diameters. (B) The size distribution of the embryos in each clutch from a single female was plotted relative to the embryo clutches from other females (N = 33). For each clutch, a boxplot of the interquartile range and the median diameter for the clutch is shown. Whiskers mark the range from the 10th to the 90th percentile for each clutch. Clutches were ranked along the Y axis based on their median value. Clutch xvi (filled boxplot) had a median diameter close to the total median diameter of all clutches together and the significance of the size variation was calculated to it. ****, p < 0.0001; **, p < 0.01; *, p < 0.05; ns, not significant. (C) Correlation between female snout-vent length and clutch embryo mean diameter (SEM shown). Simple linear regression suggests a lack of correlation, p-value provided is the significance of a non-zero slope. (D) Embryos were monitored from fertilization to advanced larval stages (st. 41). Embryos from various clutches differing in size were photographed at several developmental stages to determine whether with the progression of embryogenesis the initial size differences disappear.

3.2. Molecular evidence of natural size scaling

These natural size differences raise a number of questions about the corresponding molecular, gene regulatory, and biochemical adaptations elicited, and how these changes affect and regulate ontogenic allometry in large and small embryos. Multiple studies by us and others have focused on the scaling mechanism by proposing theoretical models, performing experimental manipulations or developing computer models (Almuedo-Castillo et al., 2018; Ben-Zvi et al., 2008; Inomata et al., 2013; Ishimatsu et al., 2018; Jevtić and Levy, 2015; Leibovich et al., 2018; Mitchison et al., 2015; Uygur et al., 2016). Surprisingly, natural scaling without experimental intervention has not been quantitatively documented. For this reason, we collected evidence supporting the natural adaptation, i.e., scaling, to size differences. First, we studied the relationship between embryo size (volume) and whole embryo RNA content. RNA was extracted from different clutches selecting a group of similarly sized embryos from the same clutch, and the total amount of RNA for each embryo was estimated. In parallel, the mean volume of the clutch was determined after measuring the embryonic diameter. This determination was performed for multiple clutches during early/mid-gastrula (st. 10.5; N = 21 clutches), late gastrula (st. 12; N = 14), and early neurula (st. 14–15; N = 15). At stage 10.5, the RNA content/embryo shows a strong and statistically significant correlation to the volume of the embryo (Fig. 2A). At stage 12, a similar significant correlation between embryo volume and RNA content was observed (Fig. 2B). Interestingly, soon thereafter, at the onset of neurula stages (st. 14–15), the correlation between RNA content and volume is no longer statistically significant (Fig. 2C). The strong correlation between embryo size and RNA content during gastrula stages suggests an increase in overall gene expression with increased volume.

Fig. 2. RNA content as a function of size during early development.

Groups of embryos from multiple clutches were collected during (A) early gastrula, (B) late gastrula, and (C) early neurula. The average embryo diameter of the clutch was determined and RNA was extracted. The volume and the average amount of RNA/embryo were calculated. At each developmental stage, the RNA content/embryo was plotted as a function of volume. N, number of clutches/RNA samples analyzed; n, number of embryos used for RNA extraction and in parenthesis the number of embryos measured to determine the average volume; r², coefficient of determination; p-value provided is the significance of a non-zero slope.

3.3. Changes in marginal zone gene expression as a function of embryo size

To further study the effects of size differences on gene expression, we analyzed multiple early expressed genes by quantitative RT-PCR (qPCR) focusing on organizer and marginal zone (ventro-lateral) genes mainly belonging to the BMP network suspected as important players in the scaling process (Ben-Zvi et al., 2008, Inomata et al., 2013, Leibovich et al., 2018, Reversade and De Robertis, 2005). Pairs of clutches were selected by visual screening such that they exhibited clear size differences, and were processed in parallel from fertilization to gene expression analysis to ensure technical uniformity. Embryos (at least 10) from each clutch were photographed before the onset of gastrulation to determine the mean diameter of the clutch. During early/mid-gastrula (st. 10.5), late gastrula (st. 12), and early neurula (st. 14–15), RNA was extracted for gene expression analysis. Relative expression of dorsal (organizer), ventral and lateral genes between large and small embryos was calculated normalizing for the amount of RNA/embryo, i.e., embryo equivalents (Figs. 2 and 3). Analysis of nine large/small clutch pairs that spanned 0.93 mm³–2.09 mm³ in volume and 2.2–5.7 μg of RNA per embryo revealed two categories of genes based on expression levels. At early/mid-gastrula, 16 out of the 22 of genes studied (acvr1, cer1, gsc, lhx1, admp, sia1, chrd.1, nodal3.1, bmp4, wnt8a, ventx2.2, ventx1.2, bmpr1b, bmpr1a, acvrl1, and sizzled) exhibited a significant increase in transcript abundance (1.65–3.25 fold) in large embryos that correlated with the increase in volume (Fig. 3A). Averaging all significant expression changes gave a 2.33 fold change which is closely similar to the fold volume change in this series of experiments (2.3 fold). The remaining genes (not, noggin, bmp2, bmp7, myod1, and ventx3.2) did not significantly alter their expression level in adaptation to size (Fig. 3A).

Fig. 3. Changes in marginal zone gene expression in response to size changes as a function of developmental stage.

Pairs of embryo clutches differing in size were collected (N = 18) and processed in parallel during (A) early/mid gastrula (st. 10.5), (B) late gastrula (st. 12), and early (C) neurula (st. 14–15). Gene expression levels were determined by qPCR and the relative expression level between large and small embryos after correcting for the amount of RNA/embryo. In all three graphs the order of the genes and their categorization as DMZ, LMZ, and VMZ (dorsal, lateral and ventral marginal zone respectively) were retained from early gastrula although with development the expression domains change. *, p < 0.05; **, p < 0.01; ***, p < 0.001; ****, p < 0.0001; ns, not significant.

By late gastrula, all genes studied exhibited a significant transcript increase in large embryos relative to small embryos. Many genes exhibited expression scaling in the 2–4.6 fold range (Fig. 3B). Interestingly, a few genes (cer1, gsc, admp, sia1, nodal3.1, and acvrl1) exhibited increased expression in the 6.3–12.7 fold range. By early neurula, most genes exhibited increased expression in the 1.3 to the 3.6 fold range (Fig. 3C). Only admp and its putative receptor acvrl1 (Leibovich et al., 2018) continued to exhibit strongly enhanced expression (7.1 and 6.4 fold respectively) in large embryos. These results show that many components of Spemann's organizer and the BMP signaling network normally expressed along the marginal zone adapt their expression levels to compensate for the difference in volume. Interestingly, admp, chrd.1 and sizzled, as well as the four type I BMP receptors studied, acvr1, acvrl1, bmpr1a, and bmpr1b, exhibited the greatest expression changes and many of them have been linked to the scaling process (Ben-Zvi et al., 2014, Inomata et al., 2013, Leibovich et al., 2018, 2017, Reversade and De Robertis, 2005).

3.4. Expanded expression domains of blastopore lip genes in large embryos

The observed increase in gene expression levels with size raises several possibilities for the process of scaling. For genes expressed in spatially restricted patterns, the increase in gene expression levels could translate into either higher expression in a domain of similar size irrespective of embryo size change, or an expansion of the expression domain. The spherical geometry of the Xenopus embryo provides an exquisite, yet simple system in which to distinguish these possibilities by studying the normal pattern of gene expression around the blastopore lip. Initially, we analyzed the expression pattern of wnt8a during mid-gastrula (st. 11). At this stage, wnt8a expression covers about 70% of the blastopore circumference, avoiding the dorsal-most region (Fig. 4A)(Christian et al., 1991, Marom et al., 1999). A large group of mixed-sized embryos (n = 116), from multiple clutches with diameters ranging from 1.36 mm to 1.71 mm, were processed for in situ hybridization with a wnt8a specific probe. At stage 11, embryos were fixed, photographed and the angle describing the domain of wnt8a was measured (Fig. 4A). In addition, the diameter of the embryo and the diameter of the blastopore were measured. Based on the calculated circumference, the mixed embryo group was divided into small, (the lower 45th percentile), and large (the upper 45th percentile); the middle 10% were omitted from the analysis (Fig. 4B). The mean wnt8a expression angle in the small group was 257.7°, whereas the large embryos exhibited a 259.9° expression angle (Fig. 4C). Statistical analysis (t-test) showed no significant difference in the wnt8a expression angle between small and large embryos. From the angle of wnt8a expression and the diameter of the embryo, the length (arc) of the expression domain along the outer edge of the embryo also was calculated. These results show a highly significant (p < 0.0001) correlation between the arc of the wnt8a expression domain and the size of the embryo (Fig. 4D). Calculation of the percent circumference covered by the wnt8a expression domain in small and large embryos revealed no significant difference (Fig. 4E). These results show that due to the spherical geometry of the mid-gastrula Xenopus embryo, maintenance of a constant wnt8a angle of expression results in a scaled expression domain and can explain, at least partially, the increase in transcript levels (see Fig. 3A and B). These results show natural scaling of the wnt8a expression domain with embryo size without any experimental manipulation.

Fig. 4. Blastopore gene expression scales proportionally with embryo size.

Groups of embryos of a wide range of sizes from multiple clutches were fixed during mid gastrula (st. 11) for wnt8a analysis (A–E) or early gastrula (st. 10.25) for chrd.1 (F–J), gsc (K–O) and admp (P–T) analysis by in situ hybridization with gene-specific probes. For each gene, a group of mix-sized embryos was processed for analysis. Based on calculated circumference, the lower 45th percentile comprised the small embryos, and the upper 45th percentile were the large embryos. The middle 10% were omitted from the analysis. For each embryo, the angle describing the expression domain, the embryo diameter and the diameter of the blastopore were measured. Plots comparing the calculated circumference (B, G, L, Q), the angle describing the expression domain (C, H, M, R), the calculated arc of the expression domain (D, I, N, S), and the percent circumference covered by the expression domain (arc)(E, J, O, T) between small and large embryos are shown. **, p < 0.01; ***, p < 0.001; ****, p < 0.0001; ns, not significant.

A similar analysis was performed for the expression domains of three organizer genes: gsc (Cho et al., 1991), chrd.1 (Sasai et al., 1994) and admp (Moos et al., 1995). Also in these cases, we determined the angle of expression and diameter for each embryo. In the case of chrd.1 (Fig. 4F-J; n = 286), the angle of the expression domain did not change significantly with size (Fig. 4H), whereas the length (arc) of the expression domain increased with size and exhibited a statistically significant correlation with embryo size (p = 0.0003; Fig. 4I). For gsc (Fig. 4K-O; n = 117), the measurements showed no significant change in the angle of the expression domain in relation with the size of the embryo (Fig. 4M), whereas the arc of the gsc expression domain exhibited a significant (p = 0.0005) correlation with embryo size (Fig. 4N). Also for admp (Fig. 4P-T; n = 51), the angle of expression did not change significantly with embryo size (Fig. 4R), whereas there was a significant correlation (p = 0.0090) between the expression domain (arc) and embryo size (Fig. 4S). These results further support the conclusion that the expression domains of these three genes increase with embryo size while retaining an expression domain size in the same proportion to embryo size (Fig. 4J, O, T).

3.5. Molecular response of the prospective neuroectoderm to size changes

Scaling of the BMP signaling pathway will affect the patterning of the mesoderm and the relative size of the tissues derived from it. During gastrula, BMP signaling also plays a role in the induction and initial patterning of the neuroectoderm (Sasai et al., 1995, De Robertis and Kuroda, 2004, Fainsod et al., 1997, Sasai et al., 1995). Therefore, understanding the scaling of the nervous system between large and small embryos and the components involved in the early specification and differentiation of this tissue is important. We previously described components of the early gene regulatory network involved in induction, specification and early differentiation events of the neural plate (Klein and Moody, 2015, Neilson et al., 2012, Sherman et al., 2017). To better understand the response of the neural genes to embryo size, we first determined their temporal patterns of expression during embryogenesis (Fig. 5). The expression of nog, gmnn, sox3, sox2, fst, meis3, sox11, foxd4l1.1, zic1, zic2, and zic3 was determined by qPCR of RNA isolated from embryos from stage 5 (16-cell) to stage 19 (mid-neurula). This temporal analysis identified three main expression patterns during gastrula and early neurula stages. The first group (nog, gmnn, sox3, and zic2) are already expressed at the onset of gastrulation (st. 10) and their expression declines towards early neurula stages (Fig. 5A). This temporal pattern is consistent with their roles in neural induction and maintenance of a pluripotent state (Buitrago-Delgado et al., 2015). In contrast, three genes (foxd4l1.1, zic1, and zic3) exhibit a sharp increase in transcript levels at the onset of gastrulation and remain highly expressed until early neurula stages (Fig. 5B). This pattern is consistent with their reported roles in establishing neural plate stem cells prior to and during initial neural differentiation. Finally, the third group (sox2, fst, meis3, and sox11) is weakly expressed at the onset of gastrulation and their expression increases towards late gastrula (st. 12) or early neurula (st.14–15)(Fig. 5C), consistent with their described roles in the onset of neural differentiation and neural plate patterning (Lee et al., 2014, Moody et al., 2013, Sasai, 1998, Yan et al., 2009).

Fig. 5. Temporal expression pattern of the neural genes.

The temporal pattern of expression of the neural genes was determined by quantitative real-time RT-PCR (qPCR) with primers specific for each gene. The relative expression level was calculated normalizing to the stage 19 sample. (A) Temporal expression pattern for nog, sox3, zic2, and gmnn (A); for foxd4l1.1, zic1, and zic3 (B); and for fst, meis3, sox11, and sox2 (C).

To study the involvement of the neural factors in the scaling of the neural plate, we used qPCR to compare their expression between small and large embryos. We focused on three categories of genes: genes that maintain proliferation (foxd4l1.1, gmnn, and zic2), transition genes (sox2, sox3, and sox11), and genes that promote differentiation (zic1 and zic3). We collected pairs of embryo clutches that differed in size, determined the mean diameter of the clutch prior to the onset of gastrulation (st. 9–10, and collected RNA samples at early/mid-gastrula (st. 10.5), late gastrula (st. 12) and early neurula (st. 14–15). As above, relative expression was normalized to the amount of RNA/embryo. During early/mid-gastrula stages, all genes studied exhibited a statistically significant increase in transcript levels in large embryos compared to small ones (Fig. 6A). By late gastrula, about half of the factors studied (sox11, zic1, gmnn, and zic3) exhibited a significant expression increase in large relative to small embryos (Fig. 6B). At the same developmental stage, expression of fst, zic2, sox2, sox3, and foxd4l1.1 showed a slight but not statistically significant increase in large embryos compared to small ones (Fig. 6B). By neural plate stages, seven of the nine neural factors studied exhibited a significant increase in expression in large embryos compared to small ones (Fig. 6C). These results suggest that with the transition into neural plate stages, the gene network involved in early neural differentiation has adapted to the size of the neural plate of the embryo.

Fig. 6. Expression scaling among neural genes.

RNA samples were collected from pairs of embryo clutches differing in size. The embryos were collected during (A) early/mid gastrula (st. 10.5), (B) late gastrula (st. 12), and (C) early neurula (st. 14–15) stages. qPCR was performed for members of the early neural network. Results were normalized to the amount of RNA/embryo, i.e. embryo equivalents. *, p < 0.05; **, p < 0.01; ***, p < 0.001; ****, p < 0.0001; ns, not significant.

3.6. Scaled size of neural gene expression domains

To gain a better understanding of how the expression of neural genes differs between large and small embryos and the importance of the increased expression, we measured the size of their expression domains. We focused on the expression domain of foxd4l1.1, which maintains proliferation, sox2, a transition gene, and zic1, which promotes differentiation (Fig. 7A, H, O). The analysis was performed at early/mid- and late gastrula (st. 10–11, st. 12, respectively), which represent the transition of proliferative ectodermal progenitors to differentiating neurons. For each embryo, we calculated its surface area (Fig. 7B, E, I, L, P, S), determined the area of the expression domain (Fig. 7C, F, J, M, Q, T) and calculated the percent of the embryonic surface area occupied by the gene expression domain (Fig. 7D, G, K, N, R, U).

Fig. 7. Scaling of neural expression domains with embryo size.

Analysis of the spatial expression domains of foxd4l1.1 (A–G), sox2 (H–N), and zic1 (O–U) in embryos of different sizes. Representative embryos illustrating the measurement of the foxd4l1.1 (A), sox2 (H), and zic1 (O) expression domains. Analysis of the expression domain in embryos at early/mid gastrula (st. 10–11; B–D, I–K, P–R) and late gastrula (st. 12; E–G, L–N, S–U) were hybridized in situ with probes specific for foxd4l1.1 (B–G), sox2 (I–N), and zic1 (P–U). For each embryo, the area of gene expression and the diameter of the embryo were measured. From these parameters, the surface area of the embryo (B, E, I, L, P, S), the area of the expression domain (C, F, J, M, Q, T), and the percent of the surface area covered by the expression domain (D, G, K, N, R, U) were calculated. *, p < 0.05; **, p < 0.01; ***, p < 0.001; ****, p < 0.0001; ns, not significant.

During early/mid-gastrula, the expression domain of foxd4l1.1 was significantly larger in the large embryos (Fig. 7C). The percent surface area occupied by the foxd4l1.1 expression domain was not significantly different between large and small embryos (about 7.1%; Fig. 7D), indicating that the foxd4l1.1 domain expanded proportionally to match the embryo surface area. Measurements of late gastrula stages also show that the foxd4l1.1 expression domain increased in large compared to small embryos (Fig. 7F). Thus, the foxd4l1.1 expression domain covers about the same relative surface area irrespective of the size of the embryo, indicating a proportional expansion (9.4% for both large and small; Fig. 7G).

Analysis of sox2 expression revealed an enlarged expression domain in large embryos (Fig. 7J, M) at all stages examined. At early/mid-gastrula, the sox2 expression domain covered about 8.5% of the surface area of both large and small embryos (Fig. 7K) whereas, at stage 12 it comprised about 15.8% of the embryonic surface area irrespective of embryo size (Fig. 7N). These observations suggest that both the foxd4l1.1 and sox2 expression domains scale from a relatively early stage. In contrast, the zic1 expression domain revealed a relatively late scaling response. During early/mid-gastrula, the expression domain of zic1 was slightly larger in small embryos (Fig. 7Q), covering 11% in small embryos, but only 8% of the surface of large embryos (Fig. 7R). By late gastrula (st. 12), the zic1 expression domain was proportionate, comprising about 13.5% of the surface area of both large and small embryos (Fig. 7T, U).

3.7. Larger embryos have more neural plate cells

Based on the observation that large embryos have larger neural plates (Fig. 7), we determined whether it is larger because it contains more cells or because each cell is larger. We compared cell density and cell size in the neural plates of small and large embryos that were stained with phalloidin to visualize cell borders, and with DAPI to visualize cell nuclei (Fig. 8A and B). Based on phalloidin staining, small embryos had an average of 215 cells/10⁵ μm²; large embryos had an average of 204 cells/10⁵ μm². Based on DAPI staining, small embryos had an average of 301 cells/10⁵ μm², and the large embryos had an average of 339.5 cells/10⁵ μm². These small differences are not statistically significant (p > 0.05; t-test), indicating that the neural plate cell density does not differ between large and small embryos (Fig. 8C).

Fig. 8. Cell density and size in small and large embryos.

(A) Phalloidin stained neural plate. (B) DAPI stained neural plate. (C) Cell density in a comparable area of the anterior neural plate was calculated in phalloidin and DAPI stained large and small embryos at early neurula stages (st. 14). (D) The long axis of cells within the neural plate of large and small embryos was measured in phalloidin stained embryos, ns, not significant (p > 0.05, two-sided t-test).

Comparable cell densities suggest that overall there might be more cells in the neural plate of large embryos. With each cell division from fertilization to mid-gastrula stages, cells become smaller as embryo size remains constant. By neurula stages (st. 14) when these measurements were performed, embryo size is increasing and cell size could change and adapt. For these reasons, the size of the long axis of individual cells in the same region of the anterior neural plate was measured in phalloidin-stained embryos (Fig. 8A, D). The average neural plate cell length was not significantly different between small (25.27 ± 6.4 μm) and large (25.35 ± 6.1 μm) embryos (p > 0.05)(Fig. 8D). The similar size and density of neural plate cells of large and small embryos suggests that neural plate cells of large embryos have gone through additional cell divisions to compensate for the larger size of the embryo.

4. Discussion

Size variability among individuals of a single species is a very common occurrence in the animal and plant kingdoms. Many factors are known to affect adult size, among them, gender, diet, and genetic background. In numerous species, size differences are already evident at the earliest embryonic stages and even when the eggs are laid. Importantly, embryos of different sizes from the same species will give rise to normally proportioned body plans. In the present study, we performed an extensive quantitative characterization of the early size differences between Xenopus laevis embryos and characterized several cellular and molecular events involved in the size differences and probably the adaptation to different sizes, i.e. scaling.

4.1. Early embryonic size differences have a maternal origin

Analysis of several thousands of embryos from 33 independent clutches provides the first detailed quantitative description of the size differences in early Xenopus laevis embryos bred in the laboratory. Most measurements were performed in the developmental window spanning from fertilization to early gastrula when it is known that Xenopus laevis embryos exhibit no size changes (Newport and Kirschner, 1982, Nieuwkoop and Faber, 1967, Satoh, 1977). The size measurements revealed a difference of almost 0.5 mm in diameter between the smallest and the largest embryos, with a median diameter close to 1.43 mm. This range of diameters translates into a 38.4% difference in circumference, a 1.9-fold difference in surface area, and a 2.65-fold difference in volume. This wide range of sizes raised a number of questions regarding the outcome and the source of the size differences, the existence of compensation, referred to as scaling, and the mechanism(s) by which this compensation is accomplished. We monitored the development of several clutches and observed that small embryos gave rise to smaller tadpoles compared to those that developed from large embryos. Importantly, the embryos across all sizes progressed normally through embryogenesis and gave rise to similarly proportioned tadpoles; organs and structures of the different-sized embryos were properly proportioned to one another and to the tadpole as a whole, i.e. static allometry.

Throughout our observations, we ensured that each clutch originating from a single female was kept separate so we could trace the size differences to the maternal origin. A number of factors could affect the size of the eggs laid by a female such as age, size, nutritional status, and genetic polymorphisms. Since the husbandry of all individuals in the colony was kept constant, we believe this rules out nutritional status as the source of the embryo size differences. We directly studied the possible link between maternal size and embryo size but we found no significant correlation. The two frog colonies used for this study are composed of mixed populations from several commercial sources purchased over close to a decade. Thus, we cannot rule out age as a confounding parameter in the generation of size differences.

We analyzed the size distribution within each clutch; each female lays eggs within a restricted range of sizes, supporting a genetic contribution in determining egg size. Commercial Xenopus laevis sources normally provide outbred individuals exhibiting extensive genetic polymorphism (Savova et al., 2017); although inbred strains are available (Session et al., 2016), they were not included in this study. We recently showed that maternal origin also affects the length of Xenopus larvae during tailbud stages (Shukrun et al., 2019). Although the length would be affected by the original egg size, it is also affected by its genetic background and its ability to grow. For these reasons, we conclude that the naturally-occurring embryonic size differences observed in laboratory-bred individuals originate with the female laying the eggs, influenced at least in part by her genetic composition.

4.2. RNA content compensation with size

We measured the RNA content/embryo and observed a significant correlation with the embryonic volume during early/mid-gastrula and late gastrula stages. Thus, RNA content changes very precisely with embryo size, presumably to compensate for dramatic differences in volume. Interestingly, with the end of gastrulation and the transition into neurulation, this correlation degrades. We believe that an important contribution to the early correlation is that from fertilization to late gastrulation Xenopus embryos do not grow (Newport and Kirschner, 1982, Nieuwkoop and Faber, 1967, Satoh, 1977). However, towards the end of gastrulation, growth commences as tissues and organs begin to form, grow and elongate, a process called ontogenic allometry.

The source of the compensatory RNA during early/mid-gastrula, which includes mRNAs, non-coding RNAs, and ribosomal RNAs, is of particular interest. Two mechanisms, not necessarily exclusive, seem feasible. First, as oocytes mature and grow, the maternally contributed cytoplasm provides a constant RNA concentration resulting in more RNA in larger eggs. Second, when zygotic transcription begins at the mid-blastula transition (MBT) (Newport and Kirschner, 1982), a transcriptional scaling mechanism is in place to adapt to embryo size. It is well established that 12 synchronous cell divisions will take place until the Xenopus laevis embryo reaches the MBT (Newport and Kirschner, 1982). At MBT, embryos will have the same number of cells, which in larger embryos will be larger in size. Since total RNA extractions are commonly accepted to include 95% or more ribosomal RNA (rRNA), it would be reasonable to assume that a major component of our total RNA samples is rRNA. However, the ribosomal content during Xenopus laevis oogenesis has been extensively studied and is thought to be constant across oocytes (Brown, 1967). This suggests that maternal rRNA might not contribute to size compensation at the level of RNA content. On the other hand, it would be difficult to explain a 2–3 fold change in RNA content based simply on a maternal contribution of protein-coding and non-coding RNAs. Since we show a transcriptional change that correlates with size for about 30 genes, we propose that there is a zygotic transcriptional mechanism that probably works in parallel with a yet unidentified maternal scaling mechanism.

4.3. Organizer and BMP signaling adaptation with size

Several mechanisms have been proposed to explain normal development after reducing embryo size by bisection into dorsal and ventral, or left and right halves. These embryos are normally proportioned albeit smaller in size, providing an experimental demonstration of scaling (Moriyama and De Robertis, 2018, Reversade and De Robertis, 2005). Many of the mechanisms proposed to account for scaling in manipulated embryos implicate components of the BMP signaling pathway or interactions with other TGFβ signaling components (Almuedo-Castillo et al., 2018, Ben-Zvi et al., 2008, Inomata et al., 2013, Inui et al., 2012, Ishimatsu et al., 2018, Leibovich et al., 2018). Most of these studies relied on extensive experimental manipulation of the embryo and were complemented by computational modeling to further support the conclusions. We studied many of the proposed candidate genes in embryos of different sizes without experimental manipulation, analyzing many components of the embryonic BMP signaling pathway and additional genes expressed along the blastopore lip. Most marginal zone genes studied exhibited a significant change in their transcript abundance in correlation with size differences during early/mid-gastrula. The expression levels increased by about 1.6–3.2 fold in large embryos compared to small ones. Among the genes exhibiting significantly increased expression in large embryos were szl, admp, and chrd.1, which were previously proposed in experimental models as important components of the scaling mechanism (Ben-Zvi et al., 2008, Inomata et al., 2013, Inui et al., 2012, Leibovich et al., 2018). The type I BMP4 and ADMP receptors, bmpr1a, bmpr1b, acvr1, and acvrl1, exhibited some of the most significant size-dependent expression level changes in agreement with their involvement in the scaling mechanism (Graff et al., 1994, Heldin and Moustakas, 2016, Leibovich et al., 2018, 2017, Schille et al., 2016).

By late gastrula, all genes studied along the blastopore lip showed increased expression in large embryos (2.23–12.69 fold). At this stage, a subset of BMP signaling components exhibit some of the highest expression changes. This condition becomes more refined by early neurula when all genes studied exhibit a statistically significant size compensation in their expression. Importantly, by this stage two genes exhibit the strongest expression change, admp and acvrl1 (7.14 and 6.41 fold, respectively) both of which we have proposed to play an important role as a ligand/receptor pair in the scaling of Spemann's organizer (Leibovich et al., 2018). This sequence of expression changes from early gastrula to early neurula describes early transcriptional events in the scaling process to a stabilization of the adaptation to size. Accordingly, we hypothesize that scaled expression levels of BMP signaling components, together with organizer-dependent events that affect the size of this region, work to shape these morphogenetic gradients (Fig. 9)(Inomata et al., 2013, Inui et al., 2012, Leibovich et al., 2018, Reversade and De Robertis, 2005). Consequently, the organizer domain will scale with the size of the embryo, and the scaled BMP gradient will modulate mesodermal patterning, which, in turn, will induce a scaled neural plate (Fig. 9).

Fig. 9. Consequences of scaling of the BMP morphogen gradient with size.

Schematic representation of the size-scaled BMP gradient and the effects on the organizer domain as determined by the arc length along the circumference and the induced neural domain.

Our study indicates that one mechanism to adjust the transcription levels of the genes to embryo size is to scale gene expression domains in correlation with the size of the embryo. Marginal zone (blastopore) expressed genes in combination with the spherical shape of the Xenopus embryo provide an ideal combination to study this further requirement of scaling. We studied a number of well-known genes during early or mid-gastrula stages. As all the genes studied (gsc, chrd.1, ADMP, and wnt8a) are expressed along the marginal zone, their expression covers part of the embryonic circumference (arc) and their domain can be described as an angle. For each gene studied, a specific angle defined its expression domain irrespective of the embryonic size (circumference). This observation showed that the gene expression domain always occupies the same proportion of the embryonic circumference and exhibits very accurate scaling with embryo size. For these domains to scale, the concentration gradients of secreted factors that set up the dorsal-ventral axis must adjust to the volume of the embryo. Our qPCR data support that this is accomplished by providing additional transcripts encoding either secreted factors that regulate dorsal-ventral patterning (e.g., chrd.1, ADMP, wnt8a) or transcription factors that regulate the expression of secreted factors (e.g., gsc). Production of additional diffusible proteins at their dorsal-ventral sources would adjust the concentration gradients, and hence expression domains, to embryo size (Fig. 9).

4.4. Scaling of the neural plate with size

One of the earliest morphologically distinguishable tissues forming in the embryo is the neural plate, whose size also depends on concentration gradients of secreted factors from the embryonic mesoderm. Ventral mesoderm factors promote non-neural (epidermal) ectoderm formation, whereas dorsal mesoderm (organizer) factors promote neural ectoderm formation (Stern, 2006, Gaulden and Reiter, 2008, Kiecker and Niehrs, 2001, Moody et al., 2013, Sasai, 1998, Stern, 2006). Therefore, we analyzed genes involved in the induction and differentiation of the neural plate in parallel with the marginal zone mesodermal genes. During early/mid-gastrula, all genes studied exhibited expression levels that were significantly different between large and small embryos. Three factors may contribute to these differences. First, some members of the neural induction and differentiation network exist in the oocyte as maternal transcripts (e.g., gmnn, sox3, sox11, zic2, and foxd4l1.1)(Kroll et al., 1998, Miyata et al., 1996, Nakata et al., 1998, Penzel et al., 1997, Sullivan et al., 2001) and thus could be contributed at higher amounts by the female during large egg oogenesis. Second, some of these factors are known to perform additional non-neural regulatory functions (e.g., gmnn, fst, and sox2)(Fainsod et al., 1997, Kerns et al., 2012, Masui et al., 2007). Third, higher levels of organizer factors likely directly up-regulate the expression of the earliest expressed neural genes (Klein and Moody, 2015). Thus, the neural gene levels observed at neural induction stages is an expected and necessary aspect of scaling; neural plate size must match mesoderm size. In contrast, during late gastrula (stage 12), only a few genes (sox11, zic1, gmnn, and zic3) showed a significant up-regulation in large compared to small embryos, and at early neural plate (stage 14–15), foxd4l1.1, sox3, and sox2 additionally became significantly upregulated. Some contributing factors to these temporal changes in transcript level scaling include: the degradation of most maternal transcripts between stages 10 and 12 (Session et al., 2016), the onset of neural plate expansion by proliferation of committed neural plate stem cells, and the onset of neural plate elongation and neural differentiation. Previous studies have demonstrated the involvement of gmnn, sox3, foxd4l1.1, sox11, and zic2 in biasing embryonic ectoderm to a neural fate and maintaining it in a proliferative state, and sox2, sox3, sox11, zic1 and zic3 in the progression from stem cell to neural differentiation (Gaur et al., 2016, Yan et al., 2009). Our results are the first demonstration that many components of the neural induction and differentiation regulatory network scale their expression with size, and that the scaling response of individual genes is temporally regulated during ontogenic allometry. Based on a previously proposed neural ectodermal regulatory hierarchy (Klein and Moody, 2015, Yan et al., 2009) and the time course of expression, we suggest that at gastrulation, neural plate scaling is achieved by maternal contributions at oogenesis as well as induction by already-scaled levels of zygotic mesodermal secreted factors and transcription factors (Fig. 9). During late gastrula, there appears to be a transcriptional up-regulation of neural genes required for proliferation and neural stemness (gmnn, sox3, foxd4l1.1, sox11, sox2, and zic2), and by early neurula, proliferation, stemness, and differentiation are required for the neural plate to elongate and primary neurons to become post-mitotic. Thus, scaling of the marginal zone genes precedes and directly influences the scaling of the nascent neural ectoderm, and then growth and differentiation of this early organ continue the process.

As is the case for marginal zone genes, the expression domains of two of the analyzed neural factors, foxd4l1.1 and sox2, are expanded in large embryos. Their expression domains are proportionate to embryo size at both gastrula and early neurula stages, demonstrating the high accuracy of the scaling mechanism. An interesting exception is zic1. This gene, which is induced during gastrulation by organizer factors (Kuo et al., 1998), has a significantly smaller expression domain in large embryos during mid-gastrula and then becomes proportionate by late gastrula. Although by qPCR there are quantitatively more transcripts in large embryos at early/mid-gastrula, this does not translate into a larger expression domain until late gastrula. Since this gene has been implicated in controlling the onset of neural differentiation and is repressed by foxd4l1.1, we suggest that part of the scaling mechanism to create a larger neural plate is to delay the expansion of the zic1 expression domain until late gastrula when it will upregulate bHLH neural differentiation factors.

Ultimately, tissues and organs in the embryo and subsequently, the adult, have to adapt in size to achieve static allometry. Neural plate development allowed us to examine how the initial organ anlage scales at the cellular level. In Xenopus, cell divisions are synchronous without cell growth until MBT just before gastrulation (Newport and Kirschner, 1982). Thus, cell sizes in the prospective neural ectoderm will be larger in larger embryos. If the number of cell divisions in the neural ectoderm was maintained between gastrula and neural stages in large and small embryos, then neural plate cells would be about twice the size in large embryos (based on our estimate that embryonic surface area changes about two-fold between small and large embryos). Our measurements in the neural plate, using two independent staining methods, revealed no differences in cell size or cell density between small and large embryos, indicating that larger embryos “generated” more cells by additional cell divisions to build a larger neural plate. These observations indicate that after the neural ectoderm is induced, additional cell divisions in large embryos contribute to increase of the size of the neural plate.

4.5. Proportionate or disproportionate scaling responses?

Some important questions or parameters to understand in the context of scaling are how size differences affect gene expression and activity. We performed numerous measurements and gene expression determinations to better understand the natural process of size scaling without experimental manipulation. For each embryonic measurement or gene expression analysis, we collected samples and replicates to obtain statistically significant quantitative results that we finally summarized as a fold change between the smallest and largest embryo sampled to determine that specific parameter. For many of our measurements, the fold change in gene expression level, expression domain or RNA content matched the change in size, i.e., showed a proportionate change. For example, in experiments in which the expression domains were measured following in situ hybridization, we could conclusively show a proportionate enlargement of the domain such that the percentage of the embryonic surface area expressing these genes remained constant through all embryo sizes. Analysis of the expression levels of the same genes by qPCR also showed proportionate scaling. In these cases, in which the embryo volumes changed 1.8–2.3 fold, most marginal zone gene expression levels at early/mid-gastrula changed from 1.94 to 2.47 fold, and most neural gene expression levels at late gastrula also changed proportionately. In some cases we observed disproportionate scaling, which could be caused by a some imprecision in staging and sample collection. Identifying stages can be somewhat subjective because gastrula stages are separated by small time windows, sometimes less than an hour (Nieuwkoop and Faber, 1967), and many of the genes expressed during gastrula stages exhibit very dynamic expression patterns, including transcript abundance, over short periods of time. Sample sizes, as in the case of RNA extractions, and geometric measurements of imperfect spheres, also could introduce variability that could make changes appear disproportionate. Despite these possibilities, we observed that components of the BMP signaling pathway along the ventrolateral marginal zone, and ADMP signaling in the organizer, seem to increase more than the increase in embryo size, total embryo RNA content or expression domain, in accord with previous studies that have identified players in the scaling process (Ben-Zvi et al., 2008, Inomata et al., 2013, Leibovich et al., 2018, Reversade and De Robertis, 2005). Multiple studies across several animal models have identified signaling gradients that play important roles in the scaling to size (Almuedo-Castillo et al., 2018, Gregor et al., 2005, Huang and Umulis, 2019, Lauschke et al., 2013, Uygur et al., 2016). Many of these studies have identified BMP and other members of the TGFβ family as signals involved in scaling during embryogenesis. Our observations show that a disproportionate change in the expression levels of BMP and ADMP signaling pathway components drive the proportionate changes involved in natural scaling.