Imprinted Grb10, encoding growth factor receptor bound protein 10, regulates fetal growth independently of the insulin-like growth factor type 1 receptor (Igf1r) and insulin receptor (Insr) genes

Abstract

Background

Optimal size at birth dictates perinatal survival and long-term risk of developing common disorders such as obesity, type 2 diabetes and cardiovascular disease. The imprinted Grb10 gene encodes a signalling adaptor protein capable of inhibiting receptor tyrosine kinases, including the insulin receptor (Insr) and insulin-like growth factor type 1 receptor (Igf1r). Grb10 restricts fetal growth such that Grb10 knockout (KO) mice are at birth some 25-35% larger than wild type. Using a mouse genetic approach, we test the widely held assumption that Grb10 influences growth through interaction with Igf1r, which has a highly conserved growth promoting role.

Results

Should Grb10 interact with Igf1r to regulate growth Grb10:Igf1r double mutant mice should be indistinguishable from Igf1r KO single mutants, which are around half normal size at birth. Instead, Grb10:Igf1r double mutants were intermediate in size between Grb10 KO and Igf1r KO single mutants, indicating additive effects of the two signalling proteins having opposite actions in separate pathways. Some organs examined followed a similar pattern, though Grb10 KO neonates exhibited sparing of the brain and kidneys, whereas the influence of Igf1r extended to all organs. An interaction between Grb10 and Insr was similarly investigated. While there was no general evidence for a major interaction for fetal growth regulation, the liver was an exception. The liver in Grb10 KO mutants was disproportionately overgrown with evidence of excess lipid storage in hepatocytes, whereas Grb10:Insr double mutants were indistinguishable from Insr single mutants or wild types.

Conclusions

Grb10 acts largely independently of Igf1r or Insr to control fetal growth and has a more variable influence on individual organs. Only the disproportionate overgrowth and excess lipid storage seen in the Grb10 KO neonatal liver can be explained through an interaction between Grb10 and the Insr. Our findings are important for understanding how positive and negative influences on fetal growth dictate size and tissue proportions at birth.

Background

Mammalian fetal growth is a highly regulated process influenced positively and negatively by genetic and environmental factors, including maternal nutrient supply. Attaining an appropriate size is strongly correlated with infant survival [1] and minimises the risk in later life of common disorders including obesity, diabetes and cardiovascular disease (see [2, 3]). The insulin/insulin-like growth factor (Ins/IGF) signalling pathway is conserved, most likely throughout animal species, to regulate growth and energy homeostasis, as well as being a major determinant of longevity [4, 5]. Involvement of the target of rapamycin complex (TOR or mTOR in mammals) is similarly broadly conserved, linking nutrient sensing, growth factor signalling and protein translation control with the same processes [5]. The invertebrate pathway involves a single Ins/Igf receptor that mediates all of these functions. In mammals the regulation of energy metabolism is a separate function of insulin acting through the insulin receptor (Insr), while the structurally related Igf1r is the primary mediator of fetal growth (Fig. 1A). This was established through a series of elegant mouse genetic experiments that also linked fetal growth regulation with genomic imprinting [6]. These experiments proved that Igf1 and Igf2 stimulate fetal growth through the Igf1r, while a second, structurally unrelated receptor, Igf2r, inhibits growth by acting as a sink for Igf2. Further, they revealed that both Igf2 and Igf2r are regulated by genomic imprinting, a form of epigenetic gene regulation that restricts expression to only one of the two parental alleles. The mouse genome contains around 150 imprinted genes, with just over half expressed predominantly from the paternally inherited allele and the rest expressed from the maternally inherited allele [7, 8]. Imprinted genes are diverse in their functions and the products they encode, but notable among them are genes encoding signaling proteins that regulate growth of the fetus, placenta, or both. These genes tend to fit with the most widely accepted hypothesis for the evolution of genomic imprinting in mammals, which posits a conflict between parental alleles in offspring that can influence nutrient acquisition from the mother [9, 10]. Noting that a female may have multiple mates, it is in the father’s interest to maximise fitness of his offspring in an opportunistic manner, whereas the mother favours a more even distribution of resources to offspring throughout her reproductive span. These pressures have resulted in the expression in developing offspring of growth-promoting genes from paternally inherited alleles, such as Igf2 and Dlk1, and growth restricting genes from maternally inherited alleles, such as Cdkn1c, Grb10, Igf2r and Phlda2 [11, 12].

Fig. 1

Signalling interactions within the insulin/insulin-like growth factor pathway inferred from biochemical and mouse genetic studies. A The Igf1 and Igf2 ligands bind and activate Igf1r to promote fetal growth, whereas insulin (Ins) activates the Insr predominantly to regulate energy homeostasis (solid arrows). In the placenta, Igf2 also binds Insr, though with lower affinity than it does the structurally related Igf1r, to promote fetal growth (shown by the dashed arrows). Igf2 is also bound by Igf2r and thereby targeted for lysosomal degradation, such that Igf2r has an inhibitory action on fetal growth through sequestration of Igf2. Products of imprinted genes, paternally expressed Igf2 and maternally expressed Igf2r, are shaded (grey). B Fetal growth outcomes expressed as mass at birth in mice of genotypes relevant to this study. Knockouts of either the Igf1r or Insr (Rec KO) previously shown to be growth restricted to 60% [41] and 90% [43] the size of wild type animals, respectively, while Grb10 KO pups are enlarged at 135% [24,25,26]. If Grb10 should act predominantly on either receptor to inhibit growth then double knockout (DKO) mice, generated in the present study, should be indistinguishable from the respective receptor single KO pups

The importance of IGF signalling and imprinting for human fetal development is exemplified by characteristic overgrowth in Beckwith-Wiedemann syndrome (BWS), associated with excess IGF2 expression, and growth restriction in Silver-Russell syndrome (SRS) associated with loss of IGF2 expression [11,12,13].

Growth factor receptor-bound protein 10 (Grb10) is a signaling adaptor protein, capable of interacting with numerous different receptor tyrosine kinases (RTKs), typically inhibiting receptor activity and downstream signalling (reviewed in [14,15,16]), in at least some cases through a mechanism involving phosphorylation of Grb10 by the mTORC1 complex [17,18,19,20,21]. Grb10 is unusual among imprinted genes in being expressed predominantly from the paternal allele in the developing and adult central nervous system (CNS), and from the maternal allele in tissues outside of the CNS [22,23,24]. Mice with a germline knockout of the maternal Grb10 allele (Grb10^m/+) are at birth around 30% larger by weight than wild type littermates [24,25,26], establishing a role for maternal Grb10 as a potent inhibitor of fetal growth. While the mass of organs such as lungs and heart increased roughly in line with the whole body, brain size did not increase significantly and was small relative to the body in Grb10 KO pups. This correlates with the lack of expression from the Grb10 maternal allele in the CNS, though interestingly no obvious effect on brain size at birth was seen in Grb10^+/p pups and instead paternal Grb10 expression in CNS has been associated with specific behavioural changes [24, 27,28,29]. In contrast to brain, Grb10^m/+ liver mass was at birth over twice that of wild type littermates [24,25,26]. This disproportionate enlargement was associated with excessive accumulation of lipid by hepatocytes whereas generally the excess growth involved changes in cell cycle and increased cell number during fetal development [26, 30]. Notably, skeletal muscle mass was increased at birth due to an increase in myofiber number, without changes in myofiber size or in the ratio of fast- and slow-twitch fibres [31], and this increase in muscle or lean mass persists into adulthood [26, 31,32,33].

Mice overexpressing Grb10, due to deletion of imprinting control regions that normally suppress expression of the paternally inherited allele, are born small (around 60% the mass of wild type littermates) and remain small into adulthood, modelling the situation in around 10-20% of growth restricted SRS patients who inherit two maternal copies of the chromosome 7 region containing GRB10 [12, 13]. This illustrates a conserved role for GRB10 in fetal growth control that is emphasized by genome-wide association studies in which GRB10 has been linked with birth weight or body size in several mammalian populations, including human [34], pig [35], sheep [36]and Arctic ringed seal [37].

Mouse studies have shown that Grb10 regulates the Insr in vivo to influence glucose regulation through actions on peripheral tissues [19, 32, 38] and the endocrine pancreas [39], and are consistent with human population studies linking GRB10 with energy homeostasis and endocrine pancreas function (e.g. [40]). Grb10 has also been shown to inhibit Igf1r activity in adult tissues [32, 39] and it is widely assumed that Grb10 influences fetal growth by acting on the Igf1r (Fig. 1B). We previously tested this assumption by performing crosses between Grb10 KO and Igf2 KO mouse mutants [25]. Resulting Grb10^m/+:Igf2^+/p double knockout (DKO) pups were intermediate in size at birth, compared to Grb10^m/+ (large) and Igf2^+/p (small) pups, indicating additive effects of two growth regulators acting largely independently of each other. Since both Igf1 and Igf2 influence fetal growth equally through the Igf1r [41,42,43] (Fig. 1A), these experiments formed only an indirect assessment of the potential for Grb10 to act via Igf1r. Given the unexpected nature of this result and the potential for some form of compensation occurring at the level of the receptor, here we tested directly for epistatic genetic interactions between Grb10 and either Igf1r or Insr. We present two key findings. First, our data support the conclusion that Grb10 acts largely independently of Igf1r or Insr signaling to regulate fetal growth. Second, excessive lipid accumulation in the neonatal Grb10^m/+ liver was found to be Insr-dependent, meaning that Grb10 modulation of Insr-regulated metabolism begins during fetal development. These findings are important for the understanding of fetal growth regulation and its impact on tissue proportions and life-long metabolic health.

Results

Genetic interaction tests show that Grb10 inhibits fetal growth independently of Igf1r

To directly assess the possibility that Grb10 interacts with the Igf1r to influence growth we performed genetic crosses between both Grb10Δ2-4 and Grb10ins7 (collectively referred to as Grb10 KO strains) and Igf1r KO mice. Grb10Δ2-4 offspring were analysed at PN1 and e17.5 whereas Grb10ins7 offspring were analysed at PN1 only. To increase statistical power, both sexes were pooled together and considered in a single analysis, with mean weights ± standard error of the mean stated in the text and shown graphically for offspring genotype groups. PN1 data were consistent between offspring of the two Grb10 KO strains (as summarised in Table 1) and consequently all subsequent experiments were carried out with only the Grb10Δ2-4 strain.

Table 1 Summary of PN1 body and organ weight data for progeny of crosses between Grb10 KO strains and Igf1r KO mice. Mean weights are shown for each genotype together with changes relative to wild type (%WT) for each mutant genotype. A) Grb10ins7 KO data. B) Grb10Δ2-4 KO data

Grb10ins7 KO x Igf1r KO offspring PN1 body mass

Progeny of crosses between Grb10ins7^+/p:Igf1r^+/- females and Grb10ins7^+/+:Igf1r^+/- males were collected at PN1 for body and organ weight analysis (Fig. 2). Progeny with six genotypes were reduced to four groups by pooling Grb10ins7^+/+:Igf1r^+/- with Grb10ins7^+/+:Igf1r^+/+ (wild type group) and Grb10ins7^m/+:Igf1r^+/- and Grb10in7^m/+:Igf1r^+/+ (Grb10ins7 KO group), for comparison with the Igf1r KO and Grb10ins7:Igf1r DKO groups (Table 2A). This was done following initial analysis of the data which confirmed that Igf1r^+/- animals had a normal fetal growth phenotype (Additional file 1: Fig.S1), as previously shown [41]. Pooling allowed us to strengthen statistical analyses, while simplifying data analysis and presentation, without materially affecting the outcome. If Grb10 regulates growth through an interaction with the Igf1r, Grb10:Igf1r DKO animals would be expected to be phenotypically indistinguishable from Igf1r KO animals (Fig. 1B). Body mass data (Fig. 2A; Table 1A) immediately indicated that we should reject this hypothesis. Grb10ins7 KO pups (mean weight 1.7670±0.0360g) were approximately 26% larger (p<0.0001) and Igf1r KOs (0.6395±0.0267g) 54% smaller (p<0.01) than wild type controls (1.401±0.0297g), respectively, whereas Grb10ins7:Igf1r DKO mutants were intermediate in size (1.1650±0.0554g). Thus, Grb10ins7:Igf1r DKO pups displayed an additive effect of both parental genotypes, being significantly different from Grb10ins7 KO (p<0.0001) single mutants, but not from both Igf1r KOand wild type neonates (Fig. 2A). This was supported by a two-way ANOVA test which showed both Grb10 (p<0.0001) and Igf1r (p<0.0001) are significant factors affecting body weight, in opposite directions, but detected no interaction between the two genotypes (p=0.1017).

Fig. 2

Weights at PN1 from progeny of crosses between Grb10ins7 KO and Igf1r KO mice. Data were pooled into four groups for analysis as described in the Methods, wild type, Igf1r KO, Grb10 KO and Grb10:Igf1r double knockouts (DKO). Body weights are shown for the four offspring genotype groups (A). Actual weights of brain (B), liver (C), lungs (D), heart (E) and kidneys (F) are shown alongside relative weights of the same organs, expressed as a percentage of body mass (G-K). Values represent means and SEM, tested by one-way ANOVA using Kruskal-Wallis and Dunn’s post hoc statistical tests. Summaries of Two-way ANOVA outcomes beneath each graph show the percentage of total variation (%var) and a p value for each source, namely the two single KO genotypes and any interaction (Inter.) between the two (values significant at p<0.05 in bold). Sample sizes were, for body, wild type (WT) n=38, Igf1r KO n=7, Grb10 KO n=26, Grb10:Igf1r DKO n=12; brain, WT n=38, Igf1r KO n=3, Grb10 KO n=25, Grb10:Igf1r DKO n=8; liver, WT n=38, Igf1r KO n=2, Grb10 KO n=25, Grb10:Igf1r DKO n=7; lungs, WT n=38, Igf1r KO n=7, Grb10 KO n=12, Grb10:Igf1r DKO n=7; heart, WT n=37, Igf1r KO n=2, Grb10 KO n=8, Grb10:Igf1r DKO n=7; kidneys, WT n=38, Igf1r KO n=2, Grb10 KO n=25, Grb10:Igf1r DKO n=7. Asterisks indicate p-values, *p <0.05, **p <0.01, ***p <0.001, ****p<0.0001

Table 2 Genetic crosses used in the study, showing parent and offspring genotypes with their expected Mendelian ratios. A) crosses between either Grb10 KO strain, (Grb10Δ2-4 and Grb10ins7) and the Igf1r KO strain. For statistical analysis Igf1r^+/- heterozygous offspring were grouped with their respective Igf1r^+/+ wild type counterparts, as indicated. B) Crosses between the Grb10Δ2-4 KO and Insr KO strains. For statistical analysis Insr^+/- heterozygous offspring were grouped with their respective Insr^+/+ wild type counterparts, Grb10^+/p with respective Grb10^+/+ wild types and Grb10^m/p with respective Grb10^m/+ as Grb10 maternal allele knockouts, as indicated. DKO = double knockout

Grb10ins7 KO x Igf1r KO offspring PN1 organ mass

To assess body proportions selected individual organs (brain, liver, lungs, heart, kidneys) were dissected at PN1 and their weights were analysed directly (Fig. 2B-F) and as a percentage of total body weight (Fig. 2G-K). The pattern of organ weight difference across the genotypes was again consistent with the DKO pups having an additive phenotype, comprising the sum of the two single KO phenotypes (summarised in Table 1A). First, the brain from Grb10ins7 KO (mean mass 0.0860±0.0017g) pups was spared from the general overgrowth phenotype indicated by body mass and was only 4% larger than wild type brain (0.0824±0.0019g) (Fig. 2B). Meanwhile, brains from Igf1r KO (0.0491±0.0021g) and Grb10ins7:Igf1r DKO (0.0535±0.0012g) pups were strikingly similar, being smaller than wild type brain by 40% (p<0.05 ), and 35% (p<0.001), respectively. Thus, while Igf1r KO brains were roughly proportionate with body size, both Grb10ins7 KO (p<0.0001) and Grb10ins7:Igf1r DKO (p<0.0001) brains were disproportionately small within larger bodies (Fig. 2G). In other words, the Grb10ins7:Igf1r DKO phenotype was dominated by brain size being severely reduced, as in Igf1r KO pups, which can therefore be attributed to loss of Igf1r expression. In keeping with this Two-way ANOVA indicated that brain weight was influenced mainly by Igf1r (p<0.0001).

In direct contrast, the livers of Grb10ins7 KO (0.1231±0.0051g) and Grb10ins7:Igf1r DKO (0.1115±0.0083g) pups were each at least double, by 124% (p<0.0001) and 103% (p<0.01), respectively, the size of wild type (0.0550±0.0016g), while the Igf1r KO (0.0469±0.0036g) liver was some 15% smaller (Fig. 2C). Consequently, while the liver was disproportionately enlarged within the heavier Grb10ins7 KO body (p<0.0001), liver disproportion was exaggerated in Grb10ins7:Igf1r DKO (p<0.0001) pups, due to DKOs having a body size similar to wild type (Fig. 2H). Due to their greatly reduced body mass relative to wild types, although Igf1r KO livers were smaller in actual mass than in wild type controls, Igf1r KO pups also had disproportionately large livers. Although neither actual nor relative liver weight was significantly different between Igf1r KO and wild type (likely due to the small Igf1r KO small sample size), the Grb10ins7:Igf1r DKO liver weight phenotype was clearly dominated by the massive size increase also seen in Grb10 KO single mutants and therefore associated with loss of the maternal Grb10ins7 allele. Two-way ANOVA analysis reflected this with only Grb10 significantly (p<0.0001) contributing to liver weight.

The remaining organs followed a pattern of size difference like that seen in the body mass data, in that Grb10ins7:Igf1r DKO mass was intermediate between that of the two single KO values. Compared to wild type (0.0387±0.0014g) lungs from a single Igf1r KO sample (0.0079g) were 80% lighter (not statistically significant due to very small samples size) and Grb10ins7 KO (0.0480±0.0019g) 24% heavier (p<0.01), whereas Grb10ins7:Igf1r DKO (0.0287±0.0013g) lungs were 26% smaller (p<0.05) and intermediate in size (Fig. 2D). Relative to total body mass, Igf1r KO lungs appeared disproportionately small while Grb10ins7 KO and Grb10ins7:Igf1r DKO lungs were roughly proportionate with their respective body sizes (F igure 2I). According to two-way ANOVA, both Grb10 (p<0.0024) and Igf1r (p<0.0001) contributed significantly to lung weight. Similarly, in comparison with wild type (0.0092±0.0005g), hearts from Igf1r KO (0.0078±0.0004g) pups were some 15% smaller and Grb10ins7 KO hearts (0.0127±0.0006g) 39% larger (p<0.0001), with Grb10ins7:Igf1r DKO hearts (0.0094±0.0006g), intermediate in size, being only 2% larger than wild type (Fig. 2E). While these weight differences were not all statistically significant, in relative terms, the heart from Grb10ins7 KO (p<0.05) and Igf1 KO (p<0.05) single mutants were disproportionately large, whereas the Grb10ins7:Igf1r DKO heart was not (Fig. 2J). Two-way ANOVA indicated Igf1r (p<0.0395) as the major contributor to heart weight.

In the case of kidneys, those from Grb10ins7 KO (0.0166±0.0006g) were only slightly enlarged, by 5%, compared with wild type (0.0159±0.0005g), while both Igf1r KO (0.0111±0.00134g) and DKO (0.01287±0.0009g), were smaller by 31% and 20%, respectively (Fig. 2F). The only significant difference in kidney weights was between Grb10ins7 KO and Grb10ins7:Igf1r DKO (p<0.05). Relative to wild type body mass, this meant that Grb10ins7 KO pups alone had disproportionately small kidneys (p<0.001) (Fig. 2K). Two-way ANOVA indicated Igf1r (p<0.001) as the major contributor to kidney weight. For each individual organ two-way ANOVA tests indicated there was no interaction between the genotypes, just as for the whole body (Fig. 2A-F).

Grb10Δ2-4 KO x Igf1r KO offspring PN1 body mass

To corroborate data from the Grb10ins7 strain, similar PN1 data were collected using the Grb10Δ2-4 strain. Progeny of crosses between Grb10Δ2-4^+/p:Igf1r^+/- females and Grb10Δ2-4^+/+:Igf1r^+/- males were again collected at PN1 and whole body weights recorded along with weights of selected organs (Fig. 3). As before, data for the six offspring genotypes were pooled to generate four groups for analysis, combining Grb10Δ2-4^+/+:Igf1r^+/- with Grb10Δ2-4^+/+:Igf1r^+/+ (wild type group) and Grb10Δ2-4^m/+:Igf1r^+/- with Grb10Δ2-4^m/+:Igf1r^+/+ (Grb10Δ2-4 KO group) progeny (Table 1B), which was again supported by our initial data analysis (Additional file 1: Fig. S2). As for the previous cross, while Grb10Δ2-4 KO pups (mean weight 1.887 ±0.0239g) were around 33% larger (p<0.0001) and Igf1r KOs (0.6205±0.0192g) 56% smaller (p<0.001), respectively, than wild type controls (1.422±0.0189g), Grb10Δ2-4:Igf1r DKO mutants (1.278±0.0381g) were intermediate in size, just 10% smaller than wild type (Fig. 3A; Table 1B). Grb10Δ2-4:Igf1r DKO pups were smaller than Grb10Δ2-4 KO pups (p<0.0001) but not significantly smaller than wild type neonates (Fig. 3A, B), while Grb10Δ2-4 KO pups were significantly larger than both wild type (p<0.0001) and Igf1r KO (p<0.0001) pups. The two-way ANOVA test showed both Grb10 (p<0.0001) and Igf1r (p<0.0001) contributed significantly to body weight and indicated a possible interaction between the genotypes, but at a relatively high significance level (p=0.0135). Despite this, it was clear that Grb10Δ2-4:Igf1r DKO pups were not small, to the extent consistently shown for Igf1r KO pups, and instead their intermediate size must result from an additive effect of the two mutant parental genotypes.

Fig. 3

Weights at PN1 from progeny of crosses between Grb10Δ2-4 KO and Igf1r KO mice. Data were pooled into four groups for analysis as described in the Methods, wild type, Igf1r KO, Grb10 KO and Grb10:Igf1r double knockouts (DKO). Body weights are shown for the four offspring genotype groups (A). Gross physical appearance of typical WT (Grb10^+/+:Igf1r^+/+) and Grb10:Igf1r DKO (Grb10^-/-:Igf1r^-/-) pups, noting in the DKO the small head relative to body size and enlarged liver (l) obscuring the milk filled stomach (s), clearly visible through the skin of the wild type (B). Actual weights of brain (C), liver (D), lungs (E), heart (F) and kidneys (G) are shown alongside relative weights of the same organs, expressed as a percentage of body mass (H-L). Values represent means and SEM, tested by one-way ANOVA using Kruskal-Wallis and Dunn’s post hoc statistical tests. Summaries of Two-way ANOVA outcomes beneath each graph show the percentage of total variation (%var) and a p value (values significant at p<0.05 in bold) for each source, namely the two single KO genotypes and any interaction (Inter.) between the two. Sample sizes were, for body wild type (WT) n=104, Igf1r KO n=13, Grb10 KO n=92, Grb10:Igf1r DKO n=28; brain, WT n=102, Igf1r KO n=6, Grb10 KO n=90, Grb10:Igf1r DKO n=24; liver, WT n=104, Igf1r KO n=5, Grb10 KO n=90, Grb10:Igf1r DKO n=23; lungs, WT n=104, Igf1r KO n=4, Grb10 KO n=90, Grb10:Igf1r DKO n=23; heart, WT n=103, Igf1r KO n=5, Grb10 KO n=88, Grb10:Igf1r DKO n=23; kidneys, WT n=100, Igf1r KO n=5, Grb10 KO n=90, Grb10:Igf1r DKO n=23. Asterisks indicate p-values, *p <0.05, **p <0.01, ***p <0.001, ****p<0.0001

Grb10Δ2-4 KO x Igf1r KO offspring PN1 organ mass

As before, body proportions were assessed by dissecting and weighing selected organs at PN1. Organ weights were analysed directly (Fig. 3C-G) and as a percentage of total body weight (Fig. 3H-L). The genotype-dependent differences in organ weights were again consistent with the Grb10Δ2-4:Igf1r DKO pups having an additive phenotype in comparison with the two single KO genotypes (summarised in Table 1B). First, the brain from Grb10Δ2-4 KO (0.0931±0.001g) pups was spared from the general overgrowth phenotype indicated by body mass and was only 10% larger (p<0.0001) than wild type brain (0.0849±0.0011g) (Fig. 3C), Meanwhile, brains from Igf1r KO (0.058±0.0018g) and Grb10Δ2-4:Igf1r DKO (0.058±0.0012g) pups were strikingly similar, being smaller than wild type brain by 42%, (p<0.01) and 32% (p<0.0001), respectively. Thus, while Igf1r KO brains were proportionate to their small bodies, both Grb10Δ2-4 KO (p<0.0001) and Grb10Δ2-4:Igf1r DKO (p<0.0001) brains were small within larger bodies (Fig. 3H). In other words, the Grb10Δ2-4:Igf1r DKO phenotype was dominated by brain size being severely reduced, as in Igf1r KO pups, and is therefore associated with loss of Igf1r expression.

In contrast, the livers of DKO (0.1259±0.006g) and Grb10Δ2-4 KO (0.1279±0.0036g) pups were again each more than double (122% and 127% larger, respectively) the size of wild type (0.0568±0.0012g) liver (p<0.0001), while the Igf1r KO (0.0454±0.0016g) liver was some 20% smaller (Fig. 3D). Consequently, the liver was disproportionately enlarged within the heavier Grb10Δ2-4 KO body (p<0.0001), and in the Grb10Δ2-4:Igf1r DKO liver disproportion was exaggerated (p<0.0001), due to DKOs having a body size similar to wild type (Fig. 3I). Due to their greatly reduced body mass relative to wild types, although Igf1r KO livers were smaller in actual mass than in wild type controls, Igf1r KO pups also had disproportionately large livers (p<0.001). Similar to our findings using the Grb10ins7 KO strain, the Grb10Δ2-4:Igf1r DKO phenotype was clearly dominated by the massive size increase associated with loss of the maternal Grb10Δ2-4 allele.

The remaining organs followed a pattern of size differences like that seen in the body mass data, in that DKO mass was intermediate between that of the two single KO values. Lungs from Grb10Δ2-4:Igf1r DKO (0.0333±0.0018g) pups were only 16 % smaller than wild type (0.0398±0.0009g) but differed to those of both single mutants, with Igf1r KO (0.0085±0.0005g) approximately 79% lighter than wild type and Grb10Δ2-4 KO (0.0539±0.0012g) 35% heavier (p<0.0001) (Fig. 3E). Grb10Δ2-4 KO lung weight was significantly different to all three other genotypes (p<0.0001 in each case).. Relative to total body mass, Igf1r KO lungs were disproportionately small (p<0.01) while Grb10Δ2-4 KO lungs were roughly proportionate with their respective body sizes and Grb10Δ2-4:Igf1r DKO marginally (p<0.05), disproportionately small (Fig. 3J). Similarly, in comparison with wild type (0.009±0.0002g), hearts from Igf1r KO (0.0073±0.0005g) pups were some 18% smaller and Grb10Δ2-4 KO hearts (0.0134±0.0003g) 51% larger (p<0.0001) (Fig. 3F). Grb10Δ2-4:Igf1r DKO hearts (0.0112±0.0005g) were intermediate in size, being 26% larger (p<0.01) than wild type but not significantly different to Grb10 KO single mutants In relative terms, the hearts from Igf1r KO pups were proportionate and those of Grb10 KO (p<0.0001) and DKO (p<0.01) of Grb10 KO (p<0.0001) and DKO (p<0.01) disproportionately large compared to wild type controls. (Fig. 3K).

Compared to wild type kidneys (0.0157±0.0003g), Igf1r KO (0.0099±0.0009g) kidneys were reduced in size, by 36% (p<0.05), while Grb10Δ2-4 KO kidneys (0.0179±0.0004g) were larger by 14% (p<0.01) and Grb10Δ2-4:Igf1r DKO (0.0138±0.0006g) were intermediate, being 12% larger, but not significantly different to wild type (Fig. 3G). Notably, Grb10Δ2-4 KO kidney weights were still significantly different to those of Igf1r KO (p<0.001) and Grb10Δ2-4:Igf1r DKO (p<0.0001). Relative to body mass (Fig. 3L), this meant kidneys were proportionate in Igf1r KO pups, but disproportionately small in the larger body of Grb10Δ2-4 KO (p<0.0001) pups. As in the previous cross, two-way ANOVA tests for individual organs indicated there was no interaction between the genotypes in each case (Fig. 3C-G). In almost all cases both Grb10 and Igf1r contributed significantly to organ weight. The exception was liver where Grb10 (p<0.0001) was the major influence on weight and the influence of Igf1r did not reach significance.

The organ disproportion evident in Grb10Δ2-4:Igf1r DKO PN1 pups was reflected by their appearance (Fig. 3B). Despite being similar in size to wild types, Grb10Δ2-4:Igf1r DKO pups had small, flattened heads and livers that were distended such that they largely obscured the milk-filled stomach.

Grb10Δ2-4 KO x Igf1r KO offspring e17.5 embryo and placenta

To investigate the potential for interaction between Igf1r and Grb10 to regulate growth by acting within the placenta we analysed weights of the whole embryo and placenta at e17.5 (Fig. 4). We chose a time-point late in gestation when any size differences between conceptuses of different genotypes would be relatively large. The pattern of size differences observed was very similar to that seen for pups at PN1. Grb10Δ2-4 KO embryos (1.085±0.0450g) were 35% larger than wild type (0.8031±0.0371g), whereas the single Igf1r KO (0.4029g) embryo collected was 50% smaller and Grb10Δ2-4:Igf1r DKO embryos (0.6330±0.0286g) intermediate in size, at 21% lighter than wild types (Fig. 4A). Unsurprisingly, the one Igf1r KO embryo showed no statistical differences in size compared to any of the other genotypes, however, Grb10Δ2-4 KO embryos were significantly larger than wild type (p<0.05) and Grb10Δ2-4:Igf1r DKO (p<0.0001) embryos.

Fig. 4

Weight analysis of e17.5 conceptuses from crosses between Grb10Δ2-4 KO and Igf1r KO mice. Data were pooled into four groups for analysis as described in the Methods, wild type, Igf1r KO, Grb10 KO and Grb10:Igf1r double knockouts (DKO). Weights are shown for the four offspring genotype groups for embryo (A) and placenta (B) and these have been used to calculate the embryo:placenta weight ratio as a measure of placental efficiency (C). Values represent means and SEM, tested by one-way ANOVA using Kruskal-Wallis and Dunn’s post hoc statistical tests. Summaries of Two-way ANOVA outcomes beneath each graph show the percentage of total variation (%var) and a p value (values significant at p<0.05 in bold) for each source, namely the two single KO genotypes and any interaction (Inter.) between the two. Sample sizes were, for wild type (WT) n=17, Igf1r KO n=1, Grb10 KO n=11, Grb10:Igf1r DKO n=9. Asterisks indicate p-values, *p <0.05, **p <0.01, ****p<0.0001

Placental weights followed a similar pattern (Fig. 4B), with Grb10Δ2-4 KO (0.1073±0.0060g) 22% larger than wild type (0.0882±0.0036g), the single Igf1r KO placenta (0.0729g) 17% smaller and Grb10Δ2-4:Igf1r DKO placentas (0.0916±0.0040g) in between at only 4% larger. The only statistically significant difference was between wild type and Grb10Δ2-4 KO samples (p<0.05). Next, the ratio of embryo to placental mass was calculated for each genotype as an estimate of placental efficiency (Fig. 4C). Although not statistically significant, the trend was for Grb10Δ2-4 KO placental efficiency (10.41) to be slightly higher than wild type (9.39), while both Igf1r KO (5.53) and Grb10Δ2-4:Igf1r DKO (6.98) were lower than wild type, with the only significant difference between Grb10Δ2-4 KO and Grb10Δ2-4:Igf1r DKO (p<0.01). A two-way ANOVA test found no evidence of an interaction between the genotypes for either embryo or placenta size (Fig. 4A,B).

Survival of Grb10 KO x Igf1r KO progeny at PN1 and e17.5

During collection of offspring the small, presumptive Igf1r KO pups seemed scarce and chi-square tests of observed versus expected numbers generally supported this notion (Additional file 2: Table S1). Testing of PN1 data from the Grb10Δ2-4 KO x Igf1r KO cross, which had the largest sample size (n=237), indicated that paucity of Igf1r KO pups was statistically significant (p=0.0174; Additional file 2: Table S1A), with 44% of the expected numbers surviving. The same was true for offspring collected from the same cross at e17.5 (p=0.015), though in this case the sample size was lower (n=38) and only one Igf1r KO embryo was obtained, with the expected number being closer to 5 (Additional file 2: Table S1B). In the case of the Grb10ins7 x Igf1r KO PN1 dataset (n=83), the lack of Igf1r KO pups was less evident (67% of the expected number) and the chi-square test indicated no significant deviation from expected mendelian ratios (p=0.4711; Additional file 2: Table S1C). In both crosses it was clear that Igf1r KO pups found alive on the day of birth were failing to thrive, as previously reported [42]. Strikingly, this did not appear to be true for Grb10:Igf1r DKO PN1 pups in either cross which typically had milk-filled stomachs, appeared to be doing well on PN1 and were not underrepresented (Additional file 2: Table S1).

Genetic interaction tests show that Grb10 inhibits fetal growth largely independently of the Insr, except in liver, where excessive enlargement in Grb10 KO neonates is due to Insr-mediated lipid accumulation

Grb10Δ2-4 KO x Insr KO offspring PN1 body mass

To address the question of whether Grb10 regulates growth in vivo through an interaction with the Insr, we next performed intercrosses between Grb10Δ2-4^+/p:Insr^+/- double heterozygous mice, giving rise to twelve offspring genotypes, which were reduced to four groups for analysis (Table 2B). In addition to combining animals with Insr^+/- and Insr^+/+ genotypes (Insr wild type groups), we also pooled Grb10Δ2-4^+/+ with Grb10Δ2-4^+/p genotypes (Grb10 wild type) and Grb10Δ2-4^m/+ with Grb10Δ2-4^m/p (Grb10 KO). This is because the Grb10 paternal allele is silent in the majority of tissues and its knockout is well established to have no effect on fetal growth [24,25,26, 49]. Similarly, only Insr^-/- animals have been shown to have a mutant phenotype affecting either growth or glucose regulation [43, 45, 50]. Initial analysis of our data prior to pooling was in line with these earlier studies (Additional file 1: Fig. S3). As asserted in the case of the Igf1r, should Grb10 regulate growth through an interaction with the Insr, Grb10Δ2-4:Insr DKO animals would be phenotypically indistinguishable from Insr KO single mutants (Fig. 1B).

Progeny were first collected at PN1 for body and organ weight analysis (Fig. 5). Just like the crosses involving the Igf1r KO, body mass data (Fig. 5A, Table 3) indicated that we should reject this hypothesis for crosses involving the Insr. Insr KO pups (1.2680±0.0483g) were not significantly different to wild type controls (1.3440±0.0297g), being only 6% smaller. In contrast, both Grb10Δ2-4 KO (1.8140±0.0447g) and Grb10Δ2-4:Insr DKO (1.6990±0.0853g) pups were substantially larger than wild type, by 35% (p<0.0001) and 26% (p<0.05), respectively, but not significantly different to each other. Thus, the overgrowth associated with loss of the maternal Grb10 allele is maintained in DKO pups despite loss of Insr expression. A two-way ANOVA test supported this, showing that body weight was mostly driven by Grb10 (p<0.0001) with little influence from Insr, and no evidence of an interaction between the genotypes (Fig. 5A).

Fig. 5

Analyses of PN1 progeny from crosses between Grb10Δ2-4 KO and Insr KO mice. Data for numerical analyses were pooled into four groups for analysis as described in the Methods, wild type (WT), Insr KO (IKO), Grb10 KO (GKO) and Grb10:Insr double knockouts (DKO). Body weights are shown for the four offspring genotype groups (A). Actual weights of brain (B), liver (C), lungs (D), heart (E) and kidneys (F) are shown alongside relative weights of the same organs, expressed as a percentage of body mass (G-K). Values represent means and SEM, tested by one-way ANOVA using Kruskal-Wallis and Dunn’s post hoc statistical tests. Summaries of Two-way ANOVA outcomes beneath each graph show the percentage of total variation (%var) and a p value (values significant at p<0.05 in bold) for each source, namely the two single KO genotypes and any interaction (Inter.) between the two. Sample sizes were, wild type (WT) n=42, Insr KO n=6, Grb10 KO n=44, Grb10:Insr DKO n=9. Histological sections of liver, stained with haematoxylin and eosin, are shown at 100x magnification for WT (L), Insr KO (M), Grb10 KO (N) and Grb10:Insr DKO (O) mice, and at 300x magnification for the same animals (L’-O’). Images are representative of at least three biological replicates per genotype and were taken at 100x magnification (scale bars show 50μm for the lower power images and 20μm for the higher power images). Asterisks indicate p-values, *p <0.05, **p <0.01, ***p <0.001, ****p<0.0001

Table 3 Summary of PN1 body and organ weight data for progeny of crosses between the Grb10Δ2-4 KO strain and Insr KO mice. Mean weights are shown for each genotype together with changes relative to wild type (%WT) for each mutant genotype

Grb10Δ2-4 KO x Insr KO offspring PN1 organ mass

As for the earlier crosses involving Igf1r KO strains, the same selection of organs was collected and weighed at PN1 to evaluate body proportions of offspring involving the Insr KO. Organ weights were analysed directly (Fig. 5B-F) and as a percentage of total body weight (Fig. 5G-K). The patterns of weight differences displayed across the genotypes was consistent with the Grb10Δ2-4:Insr DKO pups having an additive phenotype compared with the two single KOs (summarised in Table 3). The brain from Grb10Δ2-4 KO (0.0918±0.0014g) pups was once again largely spared from the general overgrowth phenotype indicated by body mass, being only 9% larger than wild type (0.0841±0.0014g), which was a significant difference (p<0.001) in this cross (Fig. 5B). Brains from Grb10Δ2-4:Insr DKO (0.0930±0.0044g) pups were similarly some 11% larger than wild type, whereas Insr KO brains (0.0856±0.0039g) were almost indistinguishable at only 2% larger. This meant that Grb10Δ2-4 KO and Grb10Δ2-4:Insr DKO brains were disproportionately small within larger bodies (Fig. 5G), compared with wild type (p<0.0001 and p<0.05, respectively) and Insr KO (p<0.0001 and p<0.05) brains. Thus, Grb10Δ2-4:Insr DKO brain size followed the pattern of the Grb10Δ2-4 KO and not the Insr KO single mutant phenotype. This interpretation is supported by two-way ANOVA which showed Grb10 (p<0.0074), but not Insr to be a significant influence on brain weight.

Liver displayed a particularly interesting pattern of weight differences (Fig. 5C). Wild type (0.0533±0.0017g) and Insr KO (0.0475±0.0036g) liver sizes were very similar, while Grb10Δ2-4 KO (0.1112±0.0049g) liver was more than twice normal size, at 109% larger than wild type (p<0.0001), as seen in the previous crosses. However, in this case Grb10Δ2-4:Insr DKO liver (0.0622±0.0023g) was only 17% larger than wild type and was significantly different to Grb10Δ2-4 KO liver size (p<0.05) but not to wild type or Insr KO liver, indicating that the disproportionate liver overgrowth associated with loss of Grb10 expression was largely Insr-dependent. This conclusion was reinforced by the finding that only Grb10Δ2-4 KO liver was disproportionately enlarged, in comparison with wild type (p<0.0001), Insr KO (p<0.001) and Grb10Δ2-4:Insr DKO (p<0.0001) (Fig. 5H). Further, a two-way ANOVA test found an interaction between the genotypes for liver weight (p=0.0013) but not for any other organ (Fig. 5A-F). To investigate the liver phenotype further we carried out histological analysis and found that the accumulation of excess lipid previously observed in neonatal Grb10Δ2-4 KO pups [26] was abrogated in Grb10Δ2-4:Insr DKO pups. Viewed at lower magnification (100x), the enlargement of hepatocytes through excess lipid storage was seen throughout Grb10Δ2-4 KO (Fig. 5N), but not wild type (Fig. 5L), Insr KO (Fig. 5M) or Grb10Δ2-4:Insr DKO (Fig. 5O) liver sections. A degenerate fatty histopathological phenotype, that has previously been described in neonatal liver of Insr KO homozygotes [45, 50], is seen more clearly at higher magnification (300x) (Fig. 5M’). This was also evident in Grb10Δ2-4:Insr DKO (Fig. 5O’) sections, is distinct from the lipid engorged cellular phenotype of Grb10Δ2-4 KO liver (Fig. 5N’) and absent in wild type sections (Fig. 5L’). Thus, the disproportionate hepatic overgrowth in Grb10 KO neonates was due to Insr signalling-dependent lipid deposition.

Lungs and heart followed a pattern of size differences like that of body mass. Grb10Δ2-4 KO (0.0442±0.0015g) and Grb10Δ2-4:Insr DKO (0.0431±0.0034g) lungs were similar in size, being 31% (p<0.0001) and 28% larger, respectively than wild type (0.0337±0.0013g), whereas Insr KO (0.0350±0.0011g) lungs were only 4% larger (Fig. 5D). Lungs from animals of all four genotypes remained proportionate with body weight (F igure 5I). Similarly, Grb10Δ2-4 KO (0.01378±0.0004g) and Grb10Δ2-4:Insr DKO (0.0112±0.0006g) hearts were both larger than wild type (0.0091±0.0002g) hearts by 51% (p<0.0001) and 23%, respectively, while Insr KO (0.0087±0.0006g) hearts were 4% smaller and indistinguishable from wild type (Fig. 5E). Hearts from animals of all four genotypes were proportionate with body weight (Fig. 5J). In this cross, Grb10Δ2-4 KO (0.0163±0.0005g) kidneys were 14% larger than wild type (0.0143±0.0004g) (Fig. 5F) but remained disproportionately small (p<0.0001) (Fig. 5K). Conversely, Insr KO (0.0163±0.0013g) kidneys were 14% larger than wild type and disproportionately large. Grb10Δ2-4:Insr DKO (0.0173±0.0013g) kidneys were 21% larger than wild type controls and roughly proportionate such that relative to body mass they were intermediate between the two single KOs. This once again reinforced the sparing of kidneys from the general overgrowth associated with loss of the maternal Grb10 allele.

Grb10Δ2-4 KO x Insr KO offspring e17.5 embryo and placenta

We next investigated the potential for interaction between Insr and Grb10 within the placenta by analysing weights of the whole embryo and placenta at e17.5 (Fig. 6). Similar to pups at PN1, compared to wild types (0.9245±0.0240g), Insr KO (0.8034±0.0569g) embryos were 13% smaller, though not significantly so, whereas Grb10Δ2-4 KO embryos (1.3010±0.0445g) and Grb10Δ2-4:Insr DKO embryos (1.2130±0.0741g) were larger, by 41% (p<0.0001) and 31%, respectively (Fig. 6A). This meant Grb10Δ2-4 KO (p<0.0001) and Grb10Δ2-4:Insr DKO (p<0.05) embryos were both significantly larger than Insr KO embryos but not different from each other.

Fig. 6

Weight analysis of e17.5 conceptuses from crosses between Grb10Δ2-4 KO and Insr KO mice. Data were pooled into four groups for analysis as described in the Methods, wild type, Insr KO, Grb10 KO and Grb10:Insr double knockouts (DKO). Weights are shown for the four offspring genotype groups for embryo (A) and placenta (B) and these have been used to calculate the embryo:placenta weight ratio as a measure of placental efficiency (C). Values represent means and SEM, tested by one-way ANOVA using Kruskal-Wallis and Dunn’s post hoc statistical tests. Summaries of Two-way ANOVA outcomes beneath each graph show the percentage of total variation (%var) and a p value (values significant at p<0.05 in bold) for each source, namely the two single KO genotypes and any interaction (Inter.) between the two. Sample sizes were, for wild type (WT) n=51, Insr KO n=13, Grb10 KO n=52, Grb10:Insr DKO n=8. Asterisks indicate p-values, *p <0.05, ***p <0.001, ****p<0.0001

In the case of placental weights, wild type (0.0899±0.0024g) and Insr KO (0.0915±0.0040g) differed by only 2% while Grb10Δ2-4 KO (0.1162±0.0026g) and DKO (0.1028±0.0051g) were 29% and 14% larger than wild type, respectively (Fig. 6B). The only statistically significant size difference was between Grb10Δ2-4 KO and either wild type (p<0.0001) or Insr KO placentae (p<0.001). When the ratio of embryo to placental mass was calculated as an estimate of placental efficiency, there were no significant differences between genotypes (Fig. 6C), though Grb10 KO (11.35) and DKO (12.0) were slightly higher than wild type (10.55), and Insr KO (9.09) slightly lower. Two-way ANOVA found no evidence of an interaction between the genotypes for either embryo (Fig. 6A) or placenta (Fig. 6B) weight.

Survival of Grb10Δ2-4 KO x Insr KO progeny at PN1 and e17.5

Data from the Grb10Δ2-4 KO x Insr KO cross was subject to Chi-squared statistical testing. This indicated that offspring genotype ratios were not significantly different from expected Mendelian ratios at either PN1 (n=101) or e17.5 (n=124) (Additional file 2: Table S2), even though pups lacking Insr expression are destined to die within a few days post-parturition of diabetic ketoacidosis [43, 45, 50]. To establish if this was also likely to be true for Grb10Δ2-4:Insr DKO animals we measured blood glucose levels during dissection of pups on PN1 (Fig. 7). Mean glucose concentrations were relatively low and indistinguishable between wild type (2.9mM±0.1) and Grb10Δ2-4KO (2.8mM±0.2) animals. Mean glucose levels were also indistinguishable between Insr KO (9.1mM±2.5) and Grb10Δ2-4:Insr DKO (6.4mM±1.7) animals and were significantly higher than wild type (p<0.05 for both comparisons) and Grb10Δ2-4KO (p<0.01 for both comparisons) pups, indicating incipient ketoacidosis in both types of animals lacking Insr expression.

Fig. 7

Blood glucose levels of PN1 progeny from crosses between Grb10Δ2-4 KO and Insr KO mice. Glucose concentration (mM) is shown for progeny of the four genotype groups wild type, Insr KO, Grb10 KO and Grb10:Insr double knockouts (DKO). Values represent means and SEM, tested by one-way ANOVA using Kruskal-Wallis and Dunn’s post hoc statistical tests. Sample sizes were, for wild type (WT) n=40, Insr KO n=6, Grb10 KO n=37, Grb10:Insr DKO n=6. Asterisks indicate p-values, *p <0.05, **p <0.01

Litter size had relatively little impact on pup weight

An inverse correlation between birth weight and litter size has long been established [51]. To assess whether litter size might affect our results we plotted mean PN1 pup weights against litter size for each genotype of offspring from this, the largest dataset for which such data was available (total n = 159 pups from 33 litters (Additional file 1: Fig. S4). The tendency for pups from larger litters to be smaller than those from smaller litters is clearly evident, at least for wild type, Grb10Δ2-4 KO and Grb10Δ2-4:Igf1r DKO pups. A two-way ANOVA test showed litter size to be responsible for 8% (p<0.0001) of the variation while genotype was responsible for 59% (p<0.0001) of the variation, making it unlikely that litter size variation has contributed substantially to comparisons of pup birth weight by genotype.

Discussion

Using a mouse genetic approach we found no evidence that the Grb10 signalling adaptor protein negatively regulates fetal growth through interaction with either Igf1r or Insr. Growth regulation by Grb10 inhibiting the Igf1r, in particular, has become the prevailing view because of evidence that Grb10 can physically interact with both receptors [14, 15, 21] and can modulate their activity and downstream signalling including in vivo, at least in adult mouse tissues [19, 32, 38]. Should an interaction between Grb10 and Insr or Igf1r be responsible for regulation of fetal growth the clear prediction is that mice lacking both Grb10 and either receptor gene will be small at birth to the same extent as the homozygous receptor KO alone, reportedly 60% for Igf1r [41] or 90% for Insr [43] relative to wild type (Fig. 1B). This is because Grb10 will have no influence in the absence of the cognate receptor. However, in crosses between Grb10 KO and Igf1r KO or Insr KO mice this was not what we observed and instead the influence of Grb10 on growth was clearly present in the double knockout offspring both at the level of the whole body, individual organs, and even the gross morphology of Grb10Δ2-4:Igf1r DKO neonates. This conclusion was strongly supported by two-way ANOVA analysis of body and organ weight data. Across the five datasets presented, evidence of an interaction between the genotypes for body or organ weight was found in only two cases. First, there was evidence of a weak interaction between Grb10Δ2-4 and Igf1r for PN1 body weight (p=0.0106), that was out of line with the four other datasets including that involving the same cross analysed at e17.5. The other exception was more interesting, for PN1 liver weight among offspring of the Grb10Δ2-4 KO x Insr KO, supporting an interaction that explains the disproportionate weight of Grb10Δ2-4 KO liver through Insr-dependent lipid storage.

Litter size has long been known to inversely correlate with mean pup birth weight, both for different mouse strains of wild type animals [51, 52] and for at least one growth deficient strain (anemic dwarf) [53]. To assess the potential impact of litter size we plotted it against mean PN1 pup weight for Grb10Δ2-4 KO x Igf1r KO PN1 data, where the largest amount of litter size information was available. The expected decline in birth weight as litter size increased was clearly seen for wild type, Grb10Δ2-4 KO and Grb10Δ2-4:Igf1r DKO pups. In the case of Igf1r KO pups only three PN1 animals were captured in the analysis and these were each from litters of different sizes. These three did not obviously follow the inverse correlation, but whether this was because of the small number or extreme growth deficiency of Igf1r KO pups, is unclear.

In crosses involving the Igf1r KO strain and either the Grb10ins7 or Grb10Δ2-4 KO strains, the birth weight of DKO pups was closer to that of wild type than either the small Igf1r KO or large Grb10 KO pups. Also, the rate of perinatal lethality and cannibalisation of these DKO mice was much reduced in comparison with that for Igf1r KO pups, perhaps reflecting the attainment of an overall size sufficient for a critical function, such as temperature regulation, or the functional rescue of one or more vital organs. Previously, evidence was presented indicating that failure of the Igf1r KO lungs to inflate caused death by asphyxia [42] and in support of this we found the Igf1r KO lungs to be disproportionately small at 79-80% lighter than wild type, whereas Grb10:Igf1r DKO lungs were only some 16-26% smaller than wild type and only marginally disproportionate with body size. Due to ethical permissions in place when the work was conducted all offspring from crosses generating DKO pups were culled on PN1 at latest, consequently we do not know if Grb10:Igf1r DKO pups would have survived beyond the perinatal period.

In contrast to lungs, the heart from Grb10ins7 KO (p<0.05) and Igf1 KO (p<0.05) single mutants were disproportionately large, whereas the Grb10ins7:Igf1r DKO heart was not (Fig. 2J). The dopa decarboxylase gene (Ddc), neighbouring Grb10, also has a role in promoting growth of the developing heart and is expressed in the developing myocardium, specifically, using a paternally expressed transcript, Ddc_exon1a [54]. Despite sharing the imprinting control regions within Grb10 [55, 56], Ddc_exon1 and Grb10 may be expressed in distinct cell populations through the use of separate tissue-specific enhancers [54]. Thus, while the dosage of the two genes is coordinated through genomic imprinting it is not clear whether they regulate fetal heart growth through a shared molecular mechanism.

Relative sparing of brain and kidney, seen here in all three crosses involving the Igf1r KO or Insr KO mice, has been seen in previous crosses involving the Grb10Δ2-4 KO [25, 26] and Grb10ins7 [24] strains. Brian sparing is in keeping with very limited expression of the maternal Grb10 allele in the developing CNS [24, 25]. This lack of Grb10 expression means the result could be considered uninformative. However, the paternal Grb10 allele is strongly expressed in the developing central nervous system and its knockout also has no significant effect on PN1 brain size [24,25,26], indirectly supporting the idea that Grb10 does not interact with Igf1r to limit fetal brain growth. In developing kidney, maternal Grb10 is widely expressed, being lower in the mesenchyme andstrongest in the epithelial component as judged at the level of both mRNA and protein [25]. Since kidney growth is driven primarily by expansion of the metanephric mesoderm to fuel nephrogenesis [57], this expression pattern may explain the relatively limited effect of Grb10 KO on fetal kidney growth. In support of a predominantly epithelial role, human GRB10 has been shown to be a tumour suppressor in clear cell renal cell carcinoma, a prevalent epithelial kidney cancer [58].

Liver followed an interesting pattern of growth changes across the three crosses. In progeny of crosses between either of the two Grb10 KO strains and the Igf1r KO strain, Igf1r KO liver was reduced in size, albeit to a slightly lesser extent than the body. In contrast, Grb10 KO livers were disproportionately enlarged, as previously observed [24,25,26], and Grb10:Igf1r DKO livers were disproportionately enlarged, to a similar extent. Thus, loss of Grb10 expression dominated the DKO phenotype, confirming that Grb10 regulates fetal liver size independently of Igfr1. The Grb10Δ2-4 KO x Insr KO cross provided further information. While Insr KO offspring had livers of normal size and Grb10Δ2-4 KO livers were again disproportionately enlarged, those of Grb10Δ2-4:Insr DKO offspring were indistinguishable in size from wild type and Insr KO livers. Liver histology revealed that excess lipid accumulation, associated with grossly distended hepatocytes, seen in Grb10Δ2-4 KO liver was not seen in Grb10Δ2-4:Insr DKO liver. Instead, Grb10Δ2-4:Insr DKO hepatocyte histology was indistinguishable from that of Insr KO liver, which had a distinct degenerate fatty appearance, as previously reported [45, 50]. This indicates that during gestation Grb10 normally acts on the Insr to suppresses hepatic lipid storage, perhaps to maximise availability of energy for growth. The result demonstrates for the first time a physiological interaction between Grb10 and the Insr other than in adult tissues [32, 33, 38, 39] . An increase in cell number, mediated by a different tyrosine kinase receptor, as in other tissues, cannot be excluded in the Grb10 KO liver but is potentially masked by the Insr-mediated hypertrophic expansion of hepatocytes. Interestingly, transgenic restoration of Insr expression in liver is sufficient to partially rescue the Insr KO phenotype [59,60,61] supporting that liver failure is a major contributor to Insr KO perinatal lethality. This relates to the vital role of liver-derived ketones as an energy source as pups transition from a placental nutrient supply (high in carbohydrates and low in free fatty acids) to post-natal life, to milk (high-fat and low-carbohydrate) and the need for Insr signalling to suppress gluconeogenesis and promote glycogen storage [62]. The lack of a catastrophic metabolic phenotype pre-term may be due to a combination of redundancy between Insr and Igf1r, supported by experiments showing that insulin can stimulate glucose uptake via Igf1r in Insr KO cells [63] , and the reliance of the fetus on placental exchange of nutrients and waste products.

Hepatic Grb10 expression is gradually lost over the first 2-3 weeks after birth and with it the excess weight and lipid accumulation in Grb10 KO liver [26]. Differentiated adipocytes capable of lipid storage emerge relatively late in development, either in late fetal development (subcutaneous white adipose tissue (WAT)) or in the early post-natal period (gonadal WAT) [64]. Interscapular brown adipose tissue is in place at birth and is important for non-shivering thermogenesis. The transition to energy storage in WAT and utilisation in brown adipose tissue (BAT) during the early post-natal period perhaps obviates the need for Grb10 to suppress hepatic lipid storage and fits with the idea that imprinted genes are important for the transition from maternal dependence to independence [65]. Curiously, in different models of hepatic steatosis Grb10 expression is induced, including through exposure to cadmium during gestational development [66] or post-natal exposure to tunicamycin or a high fat diet [67]. A liver-specific Grb10 KO model was used to prove this expression was necessary for steatosis to occur [67]. This indicates a switch in the role of Grb10 from inhibiting to facilitating hepatic lipid accumulation between fetal and adult life. Using tunicamycin to induce ER stress-mediated steatosis, Luo et al., (2018) [67] showed that loss of Grb10 had little effect on insulin-stimulated AKT phosphorylation but significantly down regulated levels of proteins involved in fatty acid synthesis. This suggests involvement of a non-canonical insulin signalling mechanism, in contrast to what we report here in neonatal liver. Steatosis can begin in the fetal or neonatal liver [68] and is recognised as an early indicator of non-alcoholic fatty liver disease, the most prevalent liver disease worldwide [69]. Given the evidence from mouse studies, involvement of GRB10 in steatosis and NAFLD merits further investigation.

In crosses involving either receptor KO and Grb10Δ2-4 KO we evaluated embryo and placental weights at a single late gestational time-point, e17.5, when placental size is maximal. Compared to wild type, we have previously shown that Grb10Δ2-4 KO conceptuses had a significant difference in mass, evident in the fetus from e12.5 and in the placenta from e14.5 [25]. Also, in a study of wild type litters, Grb10 expression was found to be higher in the smallest placentae, relative to the largest [70]. Overgrowth of the Grb10 KO placenta was found to be disproportionate, with greater expansion of the labyrinthine exchange tissue relative to the marginal and junctional zones [49]. This was associated with increased placental efficiency, such that more fetal mass was supported per gram of placental tissue by the Grb10 KO, likely due to the expanded labyrinthine zone allowing increased nutrient transfer from mother to offspring. Previous studies have concluded that there is no significant difference from wild type in the mass of placentae from Igf1r KO, Insr KO or even Igf1r:Insr DKO conceptuses [41, 43]. Our data are consistent with this, and favour the interpretation that Grb10 controls growth independently of Igf1r in the placenta as well as the embryo.

Insr KO progeny from our crosses did not display a significant growth deficit, in terms of whole-body mass at PN1 (-6%) or e17.5 (-13%), or in the mass of any individual PN1 organs. This at first appears to contrast with a reported 10% growth deficiency in e18.5 Insr KO progeny of an Insr KO x Igf1r KO cross [43], where the numbers of embryos weighed (n = 121, including 9 Insr^-/-) were very similar to our PN1 sample size (n = 101, including 6 Insr^-/-). However, it should be noted that the previous report [43] used a student's t test without any correction for multiple testing to find a significant difference in body weight between the genotypes at p<0.05. That said, the fact that Insr KO pups were consistently smaller by 6-13% across 3 different crosses and two separate studies, suggests the impact of Insr KO on fetal growth could be biologically relevant. Indeed, it seems feasible to assume that disruption in energy regulation should impact fetal growth and perhaps surprising that such an effect is not more obvious. In part, this can be explained by mouse genetic experiments showing there is redundancy in Ins/IGF signalling and, particularly, whereas Igf1 promotes growth exclusively through Igf1r, Igf2 uses both Igf1r and Insr [6, 42, 43]. Interestingly, mice with 80-98% mosaic Insr inactivation are normal in size at birth and survive for a few months but display severe post-natal growth restriction, a complete absence of mature adipocytes in BAT and WAT, and are hypoglycaemic [71]. This phenotype resembles Donohue syndrome (formerly leprechaunism), caused by homozygous INSR disruptions (reviewed in [72]).

The lack of a clear growth deficit associated with Insr KO did not affect the interpretation of our data since the well characterised overgrowth of Grb10 KO pups was still evident in Grb10Δ2-4:Insr DKO pups, ruling out Insr as a major receptor through which Grb10 mediates fetal growth regulation. This was evident through examination of individual organ weights as well as whole body weights. Most straightforwardly, lungs and heart were enlarged to a similar extent in Grb10Δ2-4 KO and Grb10Δ2-4:Insr DKO pups and differed from both wild type and Insr KO organs, though not always significantly. In this cross, Grb10Δ2-4 KO brain and kidneys again exhibited sparing from the general overgrowth of the body, which meant there were only small weight differences across the genotypes for these organs, though both Grb10Δ2-4 KO and Grb10Δ2-4:Insr DKO brain and kidneys were disproportionately small relative to the whole-body overgrowth exhibited by pups of these genotypes. At PN1 there was no obvious deficit in the number of Insr KO or Grb10Δ2-4:Insr DKO pups but both had significantly elevated blood glucose levels, indicative of incipient ketoacidosis, as previously observed for Insr KO neonates [45, 50]. In summary, the Grb10Δ2-4 KO x Insr KO cross data has established that increased fetal growth associated with loss of maternal Grb10 expression is not mediated through interaction with the Insr. Any impact of the Insr alone on fetal growth regulation is modest and instead it is primarily or solely a regulator of glucose homeostasis, including lipid storage in the fetal liver, which we have shown is normally inhibited by Grb10. The effects of Grb10 KO on liver at the cellular and molecular level merit further investigation.

As well as optimising body size during fetal growth, the growth of individual tissues and organs must be coordinated to achieve a size compatible with efficient function. Tissue proportions can be influenced by the environment. For instance, when nutrient supply is limited during development proportions can be altered in order to preserve brain growth over other organs in animals ranging from Drosophila to human (see [73]) which has been termed brain sparing. By limiting growth in only peripheral tissues Grb10 could, therefore, be an important determinant of brain sparing. More generally, our work shows how body proportions, as well as size, is altered through the actions of two independent growth regulatory pathways. Although we have not identified the ‘growth’ receptor, or receptors, on which Grb10 acts, the findings allow us to make some important inferences. In at least two pathways growth and energy homeostasis are intimately linked through Insr and mTOR signalling. While it was initially anticipated that Grb10 would prove to be the third imprinted gene influencing the Ins/IGF signalling pathway, we have shown instead that imprinting has evolved to influence more than one growth regulatory pathway. Theories for the evolution of imprinting, including the conflict hypothesis, tend to focus on individual genes rather than pathways. It is generally agreed that the benefits of voluntarily shutting down one of the two parental alleles must outweigh the cost, most obviously the risk of losing the one active copy but also, once adapted to the single gene dose, the risk of the silent copy becoming active. These risks may explain why more genes are not subject to imprinting within a single pathway, with the consequences of the resulting imbalances amply illustrated by imprinting disorders such as BWS and SRS [12].

Our data highlight that the coordination of organ size regulation during fetal development can be disrupted through maternal Grb10 KO in a manner that is not apparent through disruption of Igf1r expression. In Igf1r KO PN1 pups, organs derived primarily from each of the three germs layers, ectoderm (brain), mesoderm (heart, kidneys) and endoderm (liver, lungs) were all reduced in size. This is consistent with Igf1r, which mediates signalling of Igf1 and Igf2 [41, 42], impacting growth during early embryogenesis. A study of Igf2 KO embryos supports this, finding that disruption of cell proliferation and survival in a narrow window between e9-e10, resulted in significant changes in cell number, detectable from e11 [74], which is a few days earlier than a difference in mass can be properly discerned [25, 41, 74]. This window coincides with the early post gastrulation period when there is rapid expansion of the three germ layers and the initial events in organogenesis are taking place. We predict that by acting within a similar developmental window and engaging with one or more different receptors, Grb10 influences growth of a more limited set of tissue lineages. One possibility is that Grb10 acts on lineage-specific progenitors as they emerge during early organogenesis, since their expansion is known to regulate organ size as demonstrated, for instance, by genetic ablation experiments (e.g. [75]). Further work will be needed to identify the receptor(s) with which Grb10 interacts to influence fetal growth. Interactions between Grb10 and RTKs have been established using various techniques, most often involving co-immunoprecipitation of native or over-expressed protein in cultured cells (reviewed [14,15,16]). Since biological outcomes from these interactions may be cell type- and context-dependent the identification of the physiological growth receptor(s) may require the use of fetal tissue for the testing of candidates or application of an unbiased proteomics screen.

Conclusions

Our epistatic tests involving Igf1r KO mice show that the fetal overgrowth phenotype of Grb10 KO mice is not mediated primarily through Grb10 interaction with the Igf1r, contrary to expectation within the field. While we cannot rule out minor involvement of Igf1r, the major effect on fetal growth must involve one or more separate receptors. Similarly, we were unable to detect any growth effect of Grb10 mediated by the Insr, except for the disproportionate overgrowth of the liver. This liver expansion was associated with Insr-mediated accumulation of excess lipid in hepatocytes, indicating a metabolic basis consistent with the known role for Grb10 as an inhibitor of Insr signalling in adult tissues. Fundamental understanding of fetal growth regulation has potential benefits for the development of novel interventions that improve neonatal outcomes and life-long health for the wider population, including those with rare growth disorders.

Methods

Mice

Generation of the mouse strains Grb10Δ2-4 (full designation Grb10^{Gt(β-geo)1Ward}) and Grb10ins7 (previously referred to as Grb10 KO; full designation Grb10^{Gt(β-geo)2Ward}) from gene-trap embryonic stem cell lines has previously been described [24, 25]. Both lines are predicted null alleles and contain a functional LacZ reporter gene insertion expressed under the control of endogenous Grb10 regulatory elements. Detailed characterisation has shown that in Grb10Δ2-4 the LacZ reporter gene has replaced some 36kb of endogenous sequence, including the first 3 protein coding exons (exons 2-4), while the Grb10ins7 insertion site is associated with a 12bp deletion at the 3’ end of exon 7 [44]. Null alleles have also been described for the Insr KO [45] and Igf1r KO [46] strains. To generate experimental animals, first Grb10Δ2-4^+/p and Grb10ins7^+/p males were each crossed with Igf1r^+/- females to generate double heterozygous animals, Grb10Δ2-4^+/p: Igf1r^+/- and Grb10ins7^+/p: Igf1r^+/-. Double heterozygous females were then crossed with Grb10^+/+:Igf1r^+/- males to produce offspring of six genotypes (Table 2A). Mice were genotyped by PCR using primers and conditions previously described for Grb10 [44] and Igf1r [46].

Grb10Δ2-4^+/p males were also crossed with Insr^+/- females to generate double heterozygous animals, Grb10Δ2-4^+/p: Insr^+/-. These double heterozygous females were intercrossed to produce offspring of 12 genotypes (Table 2B). In addition to using PCR to genotype offspring for wild type and mutant Grb10Δ2-4 [44] and Insr [43] alleles, carcasses were LacZ stained [24] to determine the parental origin of mutant Grb10 alleles. Embryos and placentae were collected on embryonic day e17.5, where e0.5 was the day on which a copulation plug was observed. Otherwise, experimental animals were collected on the day of birth, designated post-natal day 1 (PN1). Wild type littermates are considered the control group and single animals the biological replicate, noting that multiple litters were generated in each cross, with the aim of having enough of the least common genotypes for robust statistical analysis. Experimental offspring were derived solely from previously nulliparous dams since we have shown previously that first and second litters from the same dam are non-equivalent [47]. All animals were maintained on a mixed inbred (C57BL/6J:CBA/Ca) strain background and housed under conditions of 13 hours light:11 hours darkness, including 30-minute periods of dim lighting to provide false dawn and dusk, a temperature of 21±2°C and relative humidity of 55±10%. Standard chow (CRM formula; Special Diets Services, Witham, Essex, UK) and water was freely available.

Tissue collection, histology and blood glucose measurements

Whole bodies and organs were collected, any surface fluid removed from embryos or dissected organs by gently touching them onto absorbent paper, and weights obtained using a fine balance accurate to 4 decimal places (Sartorius BP61S). Paired organs (lungs and kidneys) were weighed together. Organs for histology were fixed by immersion in 4% (w/v) paraformaldehyde in PBS at 4°C for 16-24 hours, then processed by machine (Leica TP1020) for wax embedding. Sections were cut at approximately 8-10 μm using a microtome (Leica Histocore Biocut), prior to staining with haematoxylin and eosin as previously described [48]. Images were collected using a digital colour camera (Olympus SC50) and software (Olympus cellSens Entry), attached to a compound microscope (Nikon Eclipse E800), then scored with the operator blind to genotype. Glucose measurements were obtained using a One-Touch ULTRA (Lifescan, CA) glucometer immediately following collection of whole blood by decapitation of PN1 pups.

Statistical analysis

Chi-square tests were applied to determine whether the genotypes of experimental groups were present in the expected Mendelian ratios. Otherwise, numerical data were subject to one-way analysis of variance (ANOVA), using a Kruskall-Wallis test with post-hoc Dunn’s test to determine p-values between groups. This test allowed us to detect significant differences associated with either of the single knockout groups in each set of progeny as well as any significant interaction between them. This relatively conservative non-parametric test was chosen because in some experiments one or more genotype group was represented by a small samples size (n=<5). In order to test for an interaction between mutant genotypes we also applied a two-way ANOVA test where indicated. All statistical tests were applied using GraphPad Prism (v10 GraphPad, La Jolla, CA, USA) software. Graphs show arithmetic means ±standard error of the mean (SEM). Differences with p-values of <0.05 were considered statistically significant.